Census 2000 PUMS Report Results Appendix C

Census 2000 PUMS Report
Appendix C: Reduction of PUMS Detail
and Your Research

Responses to Question 10: Please comment on ways in which reduction of PUMS detail might affect your research. Be specific as possible. For example, if you specialize in aging research, comment on the sorts of analyses that would be precluded by grouped age data.

Confidential Response return to top
I am particularly interested in data available by American Indian reservation areas and trust lands, as Census is really the only place where data are available at that level of analysis.

Alair MacLean, University of Wisconsin - Madison, Sociology (Graduate Student) return to top
I do research on occupational change over time. Providing only broad occupational groupings would make it impossible for me to update the research I've already done, which depends on comparability based on the IPUMS recode to the three digit occupational categories of 1950.
I also do research on education and immigration, with specific reference to metropolitan areas. At larger geographic levels, differences are often smoothed over or not apparent. If this level of geographic detail were no longer available, I would not be able to do historical comparisons.

J. Wesley Null, The University of Texas at Austin, Other Academic (Graduate Student) return to top
I conduct research on the ages and levels of educational attainment for teachers. Reductions in the level of specificity for each of these categories would severely hamper the levels of analysis that I could complete.

Confidential Response return to top
The PUMS is the best source of reliable income data by age, race, and geography since it is the most complete with the largest sample size. To lose this would cut the most meaningful variable of interest in the field of economics.

Dowell Myers, University of Southern California, Other Academic (Faculty) return to top
Comparability across time is critical for analysis of trends in meaningful ways. Any added limitations for 2000 data jeopardize all our past understandings.
Our knowledge of occupational mobility could disappear if we cannot identify specific occupational niches. Without exact ages and races we cannot define groups in ways that are consistent with prior studies.

Confidential Response return to top
My area of research is in urban planning, hence it is important for me to access data at a high level of geographic detail. Given that neighborhood oriented programs are now becoming popular, understanding the structure of neighborhoods and localities is essential to our field. Also, the detailed structure of neighborhood populations and comparability across space and time is critically important.

Christina Paxson, Princeton University, Economics (Faculty) return to top
I am currently doing research on socioeconomic status, health, and mortality. The PUMS provides extremely useful information on socioeconomic status that can be matched (by birth cohort, race, and geographical location) to health and mortality information. Removing or reducing geographical identifiers renders the data nearly useless for this purpose. It is also crucial to know the exact year of birth (i.e. age) of each respondent.
Reducing the amount of detail in income would also make it more difficult to construct accurate measures of the distribution of incomes within groups of individuals, to assess the relationships between income distribution and health.

Jeff Hayes, University of Wisconsin, Sociology (Graduate Student) return to top
Aging researchers need to be able to distinguish different age groups over 65.
Modeling incomes would be much more difficult -- both to estimate and to explain findings to various audiences -- if incomes were grouped beyond the use of topcodes.
Detailed occupations are very important to much of the work in my field. I suppose we would all live if the 1950 classification scheme (or a similar one) were used, but I like the IPUMS strategy of having it all! That is, I like having both the 1950 based for time series, but I also think that sometimes one needs an occupational structure that reflects the period in which the data are collected.

Jesse Rothstein, University of California, Berkeley, Economics (Graduate Student) return to top
I have recently been working on a project using 1960-1990 PUMS data to estimate differences in wages across industries. The central challenge in the empirical implementation of this research was the construction of comparable industry and occupation measures from the four census data years. In the end, the only thing that allowed the research to go forward was the fact that the 1990 PUMS had the most detailed industry and occupation codes of the four data sets; if it had been less detailed, I wouldn't have been able to do the research. As it was, the limited detail in earlier years severely limited the usefulness of estimates for those years.

Steve Doig, Arizona State University, Media (Faculty) return to top
Journalists plan to use PUMS to do stories on the widest possible array of topics. Any limiting of detail will invariably degrade the accuracy and quality of the data and the resulting stories.

Charles Haynes, The University of Texas at Austin, Demography (Graduate Student) return to top
My research delves into the various factors which influence migration decisions and the probability of becoming an interstate migrant. It is therefore essential to my work that I have detailed data on race/ethnicity and ancestry, state, age, and income. Reducing data availability in those areas would be highly detrimental to my work. Occupation and industry categories, for me, are much less important. I typically only use major categories.

Eric Monkkonen, UCLA, History (Faculty) return to top
I worked on age standardized homicide and use the IPUMS censuses to establish age distributions for relatively small geographic units. I regroup the age data into different categories, depending on the categories of the various comparison populations. Although age categories are relatively stable recently, it is clear that they will change as different standards come into use. As a minimum, age needs to be recorded by year.

Rachel Deyette, Harvard University, Other Academic (Graduate Student) return to top
I have used the PUMS to examine participation in voucher programs. These programs generally have quite specific criteria -- age, income, geography, public school enrollment, etc. For research purposes, it is CRITICAL that the appropriate characteristics be available to IDENTIFY the eligible groups, and it is extremely important that as many detailed characteristics about these groups -- occupation, working status, citizenship/residency status, age, receipt of public assistance/social benefits, race, etc. -- be available so that we can see who is taking advantage of these programs. Especially with a political hot potato like vouchers, we really need to know who is benefiting and who is being left behind in education.

Kristen Peterson, Brown University, Sociology (Graduate Student) return to top
My current research involves Japanese American occupational concentration. If the race or ancestry categories were limited, I would not be able to accurately identify my study population. Having an aggregate category, such as "Asians", would be unacceptable due to the diversity among Asian American groups. Also, Japanese Americans have very high rates of intermarriage, and having both the race and ancestry variables have been essential to capturing this population.
Because I'm researching occupational networks, it is important to have the most specific occupational and industrial classifications. For example, knowing that Japanese Americans are concentrated in a particular industry without knowing what type of occupation (or vice-versa), I would be unable to identify anything meaningful about occupational networks. The specificity of BOTH of these variables is necessary for my research. In addition, providing only broad geographical categories would also inhibit the study of networks or communities.

Karen Smith, The Urban Institute, Other Nonacademic return to top
I do retirement research. Loss of age above 65 would render the PUMs useless for retirement analysis.

Dan Devroye, Harvard University, Economics (Graduate Student) return to top
With regard to question 7: There is simply no way to choose among these different aspects of occupational status. Each of the categories -- SES, industry, and type of work -- is vital for research in labor economics and related fields. My own research relies on data that are both historically comparable and comparable across countries, and I shudder to think that either time series or cross section research should be compromised. What we need is more detail, not less!

Confidential Response return to top
My research deals with educational markets and their variation across metropolitan areas. I sometimes need detail at the PUMA level, and I would be greatly affected if I couldn't get that level of detail. At least, please keep the metro area identifiers.
Elimination of race and income would be a great source of trouble, too. Educational quality differs greatly across income levels and ethnic groups. If these variables were eliminated, I would be unable to capture this very important source of variation.

Matthew Sobek, University of California, Riverside, History (Researcher) return to top
Occupations are the single most important social identifiers in the census, speaking to a whole range of behaviors and life chances. Moreover, occupations are available in detailed form in every PUMS, allowing long-term measures of economic and social status before such questions as income are available. There is no substitute for historical analyses of social structure, economic attainment, etc. Even in the modern period, occupation has much information not available from other questions such as income. Detailed occupations allow the application of socioeconomic scores, which have proved to be powerful analytical tools in sociology, history, and economics.
Grouping age data would cripple cohort analyses tracking particular groups as they age across the censuses. In defining generations, one cannot be constrained by pre-determined age groupings. Studies of retirement and aging more generally would be compromised by less detail -- and this when an aging population poses one of the greatest public policy concerns of the new century.

Confidential Response return to top
I study the segregation of men and women across occupations and industries. Without information on detailed occupational categories, it would literally be impossible to study this important topic. In fact, it is well known in my field that the "detailed" occupational categories are too gross to obtain more than a guestimate of the actual degree of occupational or industrial segregation.
Note, too, that for much of the research in this area, one analyses will only use three, perhaps four variables out of the following list: detailed occupation, sex, race, education, maybe age and industry. Assuming that one only has national-level or even state-level data, it is virtually impossible to identify a particular person based on these variables alone. Yet files like the PUMS are indispensable, because cross-tabulations of these variables are not available in published tables.

Richard Campbell, University of Illinois at Chicago, Sociology (Faculty) return to top
I do specialize in aging and the loss of age detail would be extremely unfortunate at a time when the number of the "old old" is expanding dramatically. Regarding the occupation issue, again if you want to study occupational data on the elderly, you need detail. It is hard for me to answer question 7 because it depends in part on what is provided for income. Assuming decent income data were available, I would want the occ. data to maximize historical comparability.

April Eaton, University of Washington, Sociology (Graduate Student) return to top
If the data on occupations were to become less detailed, I would not be able to continue a project on the labor market impact of immigration. Further work depends on the ability to compare the labor market structure of MSAs in 2000 with that on 1980 and 1990.

Peter Meyer, Northwestern University, Economics (Graduate Student) return to top
I have been just today downloading data on particular occupations in the 1880 census to see if technological change at different time periods has some similar effects on income inequality. I will be downloading similar data for 1970-1990 soon. I need to be able to distinguish occupational categories in the recent time period to do this -- the difference between electrical engineers and computer scientists or mechanical engineers is probably minor for most research but it may really matter to studies of technological change because we *know* that electrical engineers are really ahead of the others in receiving effects from the invention of the microprocessor.
I study technological change partly because I think it is so important, and will continue to have major effects on development worldwide indefinitely. So I hope there will not be a reduction in occupational precision; we may need that sort of evidence to understand the spread of technologies.

Kim Rueben, Public Policy Institute of California, Economics (Nonacademic) return to top
I do work on the effects of policy on state and local government. I am working on a project examining teacher shortages across different states and districts. I was looking forward to using the pums 2000 data to examine changes in the number and qualifications of public school teachers as compared to other people employed in similar areas. If information on occupation was reduced it would be impossible to examine these questions.
I am also beginning work on a project using 1990 census information to investigate how linking public school attendance to housing decisions affects peoples decisions. Studying similar issues across states using the 2000 census will be more informative given in the last decade more states and districts have adopted both open enrollment systems and licensed charter schools. Investigating these questions and how different ethnic communities are served by having more school choice will be incredibly important and impossible to study if information on individuals' ethnic and racial status is not released.

Confidential Response return to top
Our research on aging requires full age identification. The differences between younger seniors (65 to 70) and older seniors (85 or older) are profound.

Confidential Response return to top
I used the PUMS and PUMA data to look travel patterns and disability for elderly. A category of over 65 is not detailed enough since the disability rates increase as one ages over 65. As the elderly pop. increases, this data will be more important.

Katherine Condon, Florida International University, Other Academic (Faculty) return to top
Issues of aging need more detail than just a "65+" or "60+" age category. I have given presentation where I have been told "65 isn't old" Yet, then one turns around and senior citizen discounts start at age 60. It is increasingly clear to me that being able to have 85+ age category, and eventually will need 100+ age category will be absolutely imperative given the increasing longevity of people. In Miami-Dade County the percentage of 85+ elders of the 60+ population has increased from 4.6% in 1970 to 9.0% in 1990. It is the 85+ that have the highest probabilities of frailty. Being able to plan for services in the community for such "needy" populations is important.
While the majority of elders in the past have been white and non-Hispanic, Miami-Dade County was already showing its unique demographic flavor in 1990 with 40% of elders being Cuban, and another 8% being non-Cuban Hispanic. Approximately 9% of elders in Dade County were black non-Hispanic. Unlike it's neighboring county of Broward in which over 90% of the 60+ population was white non-Hispanic. Not being able to get this kind of racial/ethnic break-out would be absolutely devastating to doing policy research for the county. Further, it has been projected nationally that the number of minority elders will increase astronomically in the next 40 years. While many urban areas may not have large heterogeneous elder population already (unlike Miami) being able to have the racial/ethnic along with age breakouts as detailed as possible has been absolutely important in doing policy research and examining need in the Dade County.
With respect to income, I realize that this is a very touchy issue for many individuals to answer, particularly among elders. This is also key to being able to help make decisions on policy issues as well as understand health care service needs. The difference in percentage of elders who have physical limitations for those who live above poverty and those who live below poverty was astounding even for me as a health researcher.
Historical consistency is very important - I just finished a paper using IPUMS data from 1970, 1980, and 1990 to examine historical demographic trends of elders in Miami-Dade County. I was not able to use 1960 data because there was no county identifier. It would have been very useful to have 1960 data because this census occurred before the Cuban influx into Miami-Dade County. Being able to have at the very least a county identifier is important for policy planning since in Florida if any planning is done it is done at the county level ... not state-wide.

Lisel Blash, Public Research Institute/San Francisco State University, Other Academic (Researcher) return to top
PUMS is absolutely vital to my work. It is so difficult to find such detailed information at this geographic level (PUMA's). I have used these data to look at the economic impact of a local living wage ordinance; a historical survey of the changing race, ethnicity and pay scale of hotel and restaurant workers in San Francisco and Los Angeles; at the changing demographics in the SF Bay Area; at the need for special services for people with limited English speaking skills (by county), at the need to establish adult care facilities with specific cultural and language needs in one PUMA in San Francisco, etc. I have looked at the demographics of people who commute into San Francisco compared to the demographics of those who both live and work here; I have been able to ask questions like "Which racial and ethnic groups in the city are most likely to live in overcrowded households?" I have been able to look at the specific demographics of people who work in below-"living wage" jobs by industry and by occupation. If the geographic detail were eliminated, I would not be able to do any of this work. If the occupational information were removed or made more broad, the value of these data for policy analysis on jobs and wages would be greatly diminished. Less detailed information on race and ethnicity would certainly impact my ability to do relevant research on the San Francisco Bay Area, since we have such a diverse population.

Frank Braconi, Citizens Housing & Planning Council & New York University, Economics (Faculty) return to top
I am primarily involved in urban labor research and community development policy. It is critical for me to be able to relate changes in incomes, work status, and occupations to specific sub-metropolitan geographic areas. If the detail on those issues is diminished, it would impair our ability to track changes in urban economies and their effects on specific communities, and also to evaluate government policies aimed at promoting community development.

Ileana Ciobanu, Harvard University, Economics (Undergraduate) return to top
In general, more information is better than less information in economic research. Economists need to control for as many factors as possible in order to better argue that the independent variable of interest causes, and is not merely correlated with, the dependent variable of interest. Economists can better control for as many factors as possible by having more detailed rather than less detailed data. Thus, in my opinion, reducing the detail of available data would greatly undermine economic research.

Celina Schocken, UC Berkeley- Graduate School of Public Policy, Other Academic (Graduate Student) return to top
As a graduate student in public policy, I research in several areas, so I cannot say exactly how it would affect my research. However, having less detail means being less able to measure some of the changes in public policy.

Mark Smith, The MEDSTAT Group, Economics (Nonacademic) return to top
My research using PUMS concerns the effects of welfare programs on household composition. Age and race/ethnicity distinctions are key elements because the effect of welfare differs substantially among women of different ages and racial/ethnic categories. At a time when the United States is aging and becoming more diverse ethnically, reducing the level of detail in any of these areas would be quite harmful to future social-economic research.

Darius Lakdawalla, RAND, Economics (Researcher) return to top
My research involves relative wage comparisons between different occupations. Eliminating the detailed occupation coding makes it impossible to extend my work to apply to 2000 Census data.

Joseph Doherty, UCLA School of Law, Other Academic (Researcher) return to top
We recently examined the cost of living for garment workers in Central Los Angeles. It would have been impossible without PUMS, which allowed us to estimate housing and utility costs and household composition among garment workers. We also used CPS to fill in the gaps, but its geographic limitations make it only marginally useful. No other data source is comparable to PUMS.

John Johnson, University of Illinois, Economics (Faculty) return to top
The use of detailed data is really one of the strengths of the PUMS sample. The number of important and vital questions that academic researchers can answer would be greatly reduced by cutting back on regional or localized data. Any number of evaluation projects- including issues on the impact of localized labor markets, welfare reform, minimum wage- would be severely limited by cutting back on the detailed data available. I am strongly against it.

FlorenceMae Waldron, University of Minnesota, History (Graduate Student) return to top
The staff of the Minnesota Population Center must already expend a great deal of time and energy to make historical data compatible across years. The proposed reductions in detail would, I believe, affect compatibility to earlier years to such a great extent in my work that the 2000 data would become totally useless. In order to study a population as precisely focused as persons of French Canadian ancestry currently residing in New England, many of whom are quite elderly, it is essential that I and other researchers have access to highly detailed data on age, geographical region, and racial and ethnic background. Without this level of detail, the 2000 PUMS is unusable for our research; the same would be true of researchers examining contemporary ethnicity, race, and immigration issues within numerous populations throughout the United States.

Confidential Response return to top
I conduct research on the geriatric population in Manhattan for the purpose of tracking trends, identifying needs, and developing programs. As most people know, people over 64 are not one group. Not only are they very different based on their living circumstances, past occupational history and income, but they are also very different within different subset age groups. Combining age groups or eliminating/streamlining some of the other variables that have been mentioned would make my type of research impossible.

Dan Black, Syracuse University, Economics (Faculty) return to top
I try to ascertain the earnings of various racial and ethnic groups. Restrictions on income questions might render the PUMS useless

Confidential Response return to top
In education and immigration research, research is increasingly observing trends at the level of PUMAS as a way of controlling for unobserved factors that might affect outcomes such as earnings, education, mobility etc. Aggregating beyond the level of the PUMA would set back social science research, and policy analysis, by decades. Further, aggregating age groups and racial/ethnic groups will make it extremely difficult to examine social and economic trends. What are they thinking?!!

Scott Drewianka, University of Chicago, Economics (Post-Doc) return to top
Reduction in detail could render the PUMS essentially useless for me. Detailed information on age, education, income, labor supply, and occupation is essential for virtually any good empirical labor economics study. Much of my own work also involves factors that vary across locations, so it is extremely important to be able to identify the geographic location of observations as finely as possible.

Rebecca Sandefur, University of Chicago, Sociology (Graduate Student) return to top
Most of the research in stratification, inequality, and work and occupations is moving toward recognizing the need for GREATER detail about occupations, rather than less detail. At the very least, a less fine-grained coding scheme would need to allow the researcher to identify not only type of work, but also detailed industry.
As the population of this country ages, more and more people end up in the 65+ category, and these people become less "selected" on other traits. Research -- and policy!!! -- cannot afford to lose the information we would lose by treating this group as homogenous with respect to age.

Confidential Response return to top
The reduction of detail would make it far more difficult to look at how important factors such as income vary across ages and occupations as well as geographical areas. Census data is a valuable indicator in educational research because it allows broad comparisons across a population with highly specified characteristics. Reducing the level of detail is always harmful in this sense. For example, reducing age data may make it impossible to split groups correctly as to whether they were of school age when certain legislative acts were passed.

Confidential Response return to top
The PUMS is the ONLY source with enough cases to analyze some of the fastest growing immigrant groups in the United States! If we are to be able to study the long-term impact of immigration in the United States and the adaptation of immigrants from various groups to the United States, we must have detailed information on ancestry, race and country of origin! Some point to the Current Population Survey as an alternate and cheaper source for such detail but while the CPS samples may be sufficient for some larger groups or panethnic groups (Asians as a whole, for example) they are NOT large enough for analyzing individual immigrant groups.
Research on immigration and race/ethnicity that comes from PUMS data has applications for academic research but also important policy implications as well! (Age is important too since the impact of migration varies across the life course but I will let those in aging research speak to that specifically).

David Clark, Marquette University, Economics (Researcher) return to top
I use the PUMS to model individual migration propensities, and well as factors which determine wage levels. Occupation and locational attributes are definitely needed for that research. Reducing the geographic resolution would severely diminish the usefulness of this data set. There are alternatives, but none have the scope of geographic coverage combined with the detail of occupational and industry status that the PUMS provides. PUMS is seen as a very rich data set among urban and regional economists, and I strongly urge you to maintain consistency between the Census years.

Patricia Kelly Hall, University of Minnesota, History (Graduate Student) return to top
I analyze long-term trends in immigration and internal migration. The variables you describe as under threat of detail loss are some of the most important personal characteristics associated with migration differentials. The loss of the mother's/father's birthplace question was already a severe blow to generational analysis of migration trends. Further loss of personal characteristics would be -- as the survey so aptly states -- "catastrophic." I understood the purpose of the PUMS was to provide social scientists with access to individual-level data. It sounds like the entire concept is at risk. At least for my work, that may well be the case.

Carole Shammas, University of Southern California, History (Faculty) return to top
Income and wealth are so poorly sampled already by govt agencies that further grouping would make them all but useless. As the growth of the old, old has been the big news over the past 25 years, it makes little sense from a policy standpont to start obscuring that change by grouping older ages.

Joy Williams, Environmental Health Coalition, Other Nonacademic (Nonacademic) return to top
We are an environmental justice organization. We are interested in the toxic chemical exposures of people both at work and in their communities. We can make broad generalizations about chemical exposure from occupation, but, the more process-specific this information is, the more accurately we can gauge what the chemical exposure is. We are also interested in tracking the racial and economic status of the neighborhoods with high levels of pollution, and in identification of neighborhoods with high concentrations of children under 6, the group at highest risk for lead poisoning.

Confidential Response return to top
Now I am studying the effects of immigration on the U.S. economy. To do this, I need ancestry, birth place, age, year of immigration. Moreover, to investigate the economic performance of immigrants in the U.S., I need the income level and the type of occupation of them. It is obvious that the detailed data help ma research.

David Burress, University of Kansas, Economics (Researcher) return to top
We build impact models for regional effects of development within Kansas. We need the PUMAs to be smaller, not larger, so we can localize effects better. We also look at effects of development on human capital by occupation and industry within Kansas. If the occupation categories are too coarse, then we can't distinguish between types of industry, so we lose the ability to recommend industrial targeting.

Susan Carter, University of California, Riverside, Economics (Faculty) return to top
Some of my research is focused on gender, race, and ethnicity differences in occupational attainment. All the standard measures of difference are quite sensitive to the level of detail by which occupations are reported. A reduction of PUMS occupational detail would severely restrict scholarly work in this important area.

Kate Antonovics, University of Wisconsin, Economics (Graduate Student) return to top
My research focuses on racial discrimination. Detailed information on race, education, age and income are all absolutely essential for the work that I do. Broad groupings make rigorous empirical work extremely difficult. For example, it is essential for my work to have data on hourly wages. Because the "number of hours usually worked per week last year" variable is only available in later Census years, I am forced to use weekly wages. Given the tremendous variation in the number of hours per week that people tend to work, this measure of wages is extremely noisy.

Lawrence Nitz, University of Hawaii at Manoa, Political Science, Other Academic (Faculty) return to top
Hawaii has only a handful of PUMAs. As it is, the 100,000 rule combines three islands which are nearly two hundred miles apart (Maui, Lanai, Molokai v. Kauai). We have a largely Asian-Pacific population in which ancestry is as important as race--since there are different Chinese cultures, varying sources of immigration from Asia, and radically different characteristics among folks of the same ethnicity and/or heritage as a function of their respective waves of immigration to Hawaii.
One of he most pressing problems in a small state with a tradition of friendliness is demonstrating systemic failure. When our friends in the business world complain about unqualified job applicants streaming from the high schools, we have no external way of validating either the skill level of the job or that of the candidates. Occupational data should be reflected in a way that maps to some small extent onto the average skill and disposition categories developed in the new Dictionary of Occupational Titles.
Until we can document the relationship between the general skill levels associated with jobs and the wages and benefits paid, we cannot respond adequately to questions of equitable distribution of educational facilities or expectations.

Joelle Saad-Lessler, Columbia University, Economics (Graduate Student) return to top
The more you reduce PUMS detail, the fewer observations we have to work with, and that makes all the results of the research less robust. When you group age data, we are precluded from being able to do significant amounts of research by age category. This will become more and more significant of a restriction in today's labor market, where people are not retiring early, and are working longer and living longer as well. For example, if the age category 65 and older replaces more age detail, we would be precluded from conducting any research about labor participation trends among those over 65, by their age. Any such research will have to content itself with grouping a 65 year old with an 80 year old, thereby reducing the usefulness of any findings. This will affect policy on the reform of the social security system, for example. It will also affect research on the connection between working and longevity, or working and health outcomes. Finally, with the changing skill requirements of the labor market, we would be precluded from learning more detail about the rate of skill depreciation, and the value of experience.

Leora Friedberg, University of California, San Diego, Economics (Faculty) return to top
I used the IPUMS to analyze the impact of Old Age Assistance on retirement. This was a very informative study because OAA benefits varied across states. Social Security benefits do not very across states, so it is difficult to understand how Social Security benefits, which depend on income and other personal characteristics, affect retirement ages. The study of OAA showed definitively that benefits affect when people retire. This helps us understand the extent to which Social Security has lowered retirement ages, which has in turn increased the drain on the Social Security system.

Patricia Oslund, University of Kansas, Economics (Researcher) return to top
1. Research on migration. We have on-going models that include parameters for sensitivity of migration to overall household income and to wage income. Reduction in geographic scope would make this analysis less meaningful, and reduction in income detail might make it impossible.

2. Social accounting model. We have an on-going social accounting matrix model that uses PUMS data (we aggregate to state level) to find: average full time wages by industry; breakdown of income by household income category by industry - for each industry, what share of labor income goes to households with total income < 40th percentile of total income, etc. We need income and industry detail for this. Geographic detail is less important here.

Reeve Vanneman, University of Maryland, Sociology (Faculty) return to top
For the last several years I and my colleagues have used the PUMS, the detailed occupational (EEO) tables, and the STF3c, building up from county records in the STF3C and EEO and from PUMAs in the PUMS to analyze how individual and metropolitan area characteristics JOINTLY affect earnings, labor force participation, and occupation. We have completed a set of papers analyzing the impact of local labor market occupational segregation and the percent of female occupations in the local labor market. We have published this in the top sociology journals and continue to generate new research. All of this will become impossible with the kinds of restrictions you suggest. Just as we get the statistical ability to do multilevel analyses, the Census is thinking about not including sufficient geographic detail to make this possible. What a shame.

Stephanie Aaronson, Columbia University, Economics (Graduate Student) return to top
There are several issues. First of all, I work primarily on young people. Since the living and working circumstances of young people change very quickly, it is important to be able to identify them by specific age. Even just grouping 16 to 18 year olds together would obscure a great deal of variation.
With respect to occupational data it is very difficult to detect changes in the distribution of occupations if the occupations are highly aggregated. A similar issue arises with geographic data. Furthermore, much analysis relies on cross-geographic variation in the data for identification. The fewer geographic areas that can be identified the worse the statistical analysis.

Confidential Response return to top
I am a labour economist with a particular interest in geographic labour markets. Comparability of geographic areas is vital, as is the ability to control for differences in workforce and employment composition.

Confidential Response return to top
If the data surrounding age, especially for children and race/ethnicity is reduced, our agency would have a difficult process in providing critical demographic information for policy makers and legislative bodies. The loss of the data would greatly impact the ability to design and implement programs focused on overrepresentation or underserved population especially for children's issues.

Gabriela Galescu, Cornell University, Sociology (Graduate Student) return to top
I am working on a structural model of occupational segregation by race and city. David Grusky, my adviser, did a similar research on cross-national occupational segregation by sex and the results showed that there is no such thing as a general pattern of segregation, that the underlying structure of that phenomenon is more complex than previously thought.
I have started on working on a similar research on occupational segregation by cross-cities. Now I'm only working on 13 big cities, but my research will be really significant only when covering most of the cities of the U.S. Also I need detailed race and occupation data in order to give meaningful results. I am claiming that segregation is more subtle than "black people don't get as good jobs as the white ones", but how more subtle and interesting the pattern is depends on the data I can use.
Besides, I grouped the 505 occupations in 42 categories to make the analysis easier for me for a final paper due one of these days and I was just amazed how much of the actual difference I lost even at the low-level of cross-tabulation. Really, for what I am doing, detailed data are absolutely fundamental.

Huey-Chi Chang, University of Wisconsin-Madison, Sociology (Graduate Student) return to top
Age: In the study of the living arrangement of elderly immigrants, I won't be able to identify precisely at what age these immigrants came to US even with the information on the year of entry. Without detailed age information on the elderly, I won't be able to use age as an indicator to explore how the aging process or changes in the health condition might influence living arrangement. Age is an important variable especially if we would like to disentangle the cohort, period and age effects. Unless we can develop a grouping scheme that is consistently used throughout all such time variables (or across all censuses for synthesis cohorts), any arbitrary grouping may render the information useless.
Occupation: 505 titles are preferred. Detailed classification is more likely to insure the comparability with older censuses. It's easier to collapse categories and to reorganize the data with detailed classification.

Confidential Response return to top
One project I am currently working on involves examining the differences in income and education of individuals residing in large, mid-sized, and small cities. This analysis requires a significant level of detail of the data for geographic variables as well as income, race, and age variables.

Charles Wetherell, University of California, Riverside, History (Faculty) return to top
Single year of age is becoming increasing important in the demography of the aging population as well as in social and economic analysis of the elderly behavior. Fine grained geographic analyses are also becoming more important as sociologists and social psychologists try to assess environmental and neighborhood effects. As a Co-PI of a five-year, NIH funded study of functioning in Mexican-American and Euro-American families in Southern California, my work would be seriously impaired by any reduction in information in the 2000 PUMS

Gunnar Almgren, University of Washington, Demography (Faculty) return to top
In fact, reductions in the number age categories would greatly reduce the validity of mortality and morbidity rate estimates by age. This is a particular concern for researchers attempting to link social structure with health outcomes.

Robert Adelman, University at Albany-SUNY, Sociology (Graduate Student) return to top
PUMS data are, without question, an integral part of my research. As a sociologist/demographer interested in racial and ethnic inequality in the United States, any reduction in PUMS detail will hurt my (and many others') ability to perform adequate analyses of important, policy-related issues.

Lindsay Rossano, The Joseph L. Mailman School of Public Health at Columbia University, Sociology (Researcher) return to top
For public health research (specifically social epidemiology) such as I do, detailed information with regard to SES is vital. This, of course, includes information with regard to race, education, occupation, and where such people live. As it stands we know that census data likely underestimates low income and minority populations. Efforts to study these populations will only be further hindered by generalizing the data available.

Confidential Response return to top
I need to be able to match PUMS data to specific local public policies in order to assess the policy impacts. More detail about geographic/political jurisdictions greatly eases this task.

Confidential Response return to top
Loss of detail is loss of information. Loss of information means statistics are less likely to capture micro behavior.

Carolyn Liebler, University of Wisconsin at Madison (sociology & demography), Sociology (Graduate Student) return to top
I do research on American Indians, including people who say their ancestry is partly American Indian. I am interested in the meaning behind reporting a tribe. Also, I use tribal affiliation to add context to my work.
Also, I do research on the meaning and measurement of race. If you were to combine categories in the race question, I would not be able to understand the differences, for example, between people who are American Indian/White/Black and those who are American Indian/White. The 2000 Census race question is very exciting to people who study the meaning of race because it gives us significantly more information than we have ever had about people's racial identification.
Whether or not you combine categories in the race question, the ancestry question is also key to understanding the complex racial nature of our country. In my research, I have found that people who say they are American Indian by both race and ancestry are different in interesting ways from those who say they are American Indian by race only. I hope to use the 2000 PUMS data to see if people who are American Indian/White by race and only American Indian by ancestry are different from those who are American Indian/White and only White by ancestry.

Mark Ellis, University of Washington, Geography (Faculty) return to top
My research involves making comparisons of the social and economic situation of immigrants in major cities of immigrants over time. I spent considerable time (with NSF money) maximizing the compatibility of the 1980 and 1990 microdata geographies to build metropolitan areas that were similar in both years (see Ellis, Mark, Michael Reibel and Richard Wright (1999) "Comparative Metropolitan Area Analysis: Matching the 1980 and 1990 Census Public Use Microdata Samples for Metropolitan Areas". Urban Geography 20: 75-92). This is essential to remove any effect of geographic boundary changes on measurement of migration, employment and so on. I have used these synthesized metro areas (and others have too) to publish a number of papers on immigration/migration linkages, immigrant wages, and immigrant employment. I will need to build synthetic metro areas in the 2000 PUMS that match those I built for 1980 and 1990. This will make results from the 2000 PUMS compatible with my earlier work and allow me to infer trends about immigrants in specific metro areas with greater confidence. Reduced information on race and ancestry and income may also have devastating effect on immigrant analysis if one cannot identify the specific economic circumstances of groups or if these data are not compatible with earlier years.

Confidential Response return to top
My thesis research focuses on the effect of family migration on wage growth, and specifically how this effects married women differently from other workers. The proposed changes would harm my ability to use the census in the following ways: 1. Geographic areas differ in terms of number of jobs available in various fields. Not having labor market information (PUMAs) would limit the ability to make use of this information. 2. Labor markets are thinner in certain occupations. Having specific occupational information helps identify fields most helped/hurt by family migration.

John Sandberg, The University of Michigan, Demography (Graduate Student) return to top
I work primarily with age, race and education data to make population comparisons across time, and geographic region. Aggregated age-especially at older ages would be disastrous for analyses of the older and oldest-old (one of the fastest growing segments of the population). The dynamic relation of income and wealth accumulation among these elderly as they pass through these years is of critical importance, and would be hamstringed by grouping the data as proposed.

Michael Katz, University of Pennsylvania, History (Faculty) return to top
My colleague Mark Stern and I have a contract from the Russell Sage Foundation to write a book putting the 2000 census in the context of social and economic trends in the 20th century. The PUMS is the basic source. Detail is crucial both to perform the level of analysis we need on the 2000 census and to compare trends across time. A reduction in the level of detail would cripple our ability to write meaningfully about major changes across time.

Confidential Response return to top
Reduction in the detail of IPUMs data would be very detrimental to my research. My studies involve the effects of inequality on socioeconomic indicators by race and ethnic backgrounds. Since there is an increasing amount of interest in how different groups are affected by inequality, you can see that detail by race/ethnic background/ancestry are important to my work.
Furthermore, I use income to calculate measures of inequality. Yet income is not the only measure of inequality. Detail on education and occupation are also important to provide robust measures of inequality. In addition, age detail is indispensable because in comparing different groups based on race, ethnicity, etc., one must age standardize. Clumping 65+ excludes analysis of the elderly population. During a time in which there will be more 65+ people, limiting analysis to the younger age cohorts will disallow the many studies that could provide useful policy for Social Security. Also, geographic detail is important because this allows me to combine IPUMS data with other data sets at the aggregate level when I need to utilize indicators such as the infant mortality, death rates, etc. Aggregate measures at geographically differing levels are essential to providing sound testing and robust results.
Concerning the timing release of the data, if less detailed data is available earlier, followed by more detailed data later, that would be very good. This allows for preliminary analysis followed by more substantial investigation.
Finally, I am sure legitimate researchers will not divulge results such as to reveal the personal information of individuals. If the fear is that researchers do this unknowingly, perhaps the Department of Human Subjects at each institution should expend more efforts in disseminating rules concerning privacy.

David Belkin, City of New York Independent Budget Office, Government (Nonacademic) return to top
Given the paucity of alternative sources of earnings and income data at the city level, the construction of such data out of PUMS detail is absolutely vital for urban economic research. Not only should there be NO reduction in the level of geographic detail, there would on the contrary be great value in having GREATER geographic detail.

William Donlin, University Of Minnesota, History (Undergraduate) return to top
Grouping ages for the elderly would definitely work against my research. I am currently working on a project that includes World War II Veterans in the second half of the 20th century. This group would be adversely affected with these new modifications and my project would loose continuity for the latest census year.

Confidential Response return to top
PUMS data with detailed geographic codes are essential to policy researchers who are trying to understand the baseline characteristics of the populations subject to new policies. Limiting the availability of these data would make it necessary to collect more data for individual policy evaluations-- an expensive and inefficient approach given that much of the necessary is potentially available from PUMS.

Steve Holden, Johns Hopkins University, Economics (Graduate Student) return to top
Most important is the regional variables, race, age and gender.

Confidential Response return to top
Much of our current research requires identification of lawyers or judges. Our research also requires clear identification by race and income

Ernest Schreiber, Lancaster New Era, Media (Nonacademic) return to top
Any decrease in detail makes the data less useful. When the bureau goes to the expense and effort to compile good data, any program that degrades its usefulness makes utterly no sense.

Robert Hauser, University of Wisconsin-Madison, Demography (Faculty) return to top
As far as I am concerned, elimination of the detail of age, race, ancestry, income, occupation, and geography would essentially eliminate the value of data from the long form. This is a shameful, cowardly, and ludicrous proposal. I hope it will disappear promptly and not be raised again.

Cristian deRitis, Johns Hopkins University, Economics (Graduate Student) return to top
I consider historical evolutions of wage patterns in my research. Research in this area is significantly compromised by vague definitions of skill, education and occupation with the data currently available. Results are challenged regularly on the basis of historical consistency, relevance and significance. To broaden categories further would only serve to decrease the power labor market analyses may have.

Seth Sanders, University of Maryland, Department of Economics, Economics (Faculty) return to top
It is imperative that all of the data be entered even if restrictions are necessary. First, there are many questions of data quality that can only be answered by cross referencing data items and checks for consistency and error are greatly diminished when the data is preclassified. The exact coding of age seems quite important as so many studies use the age structure of families in creative ways. For example, without exact ages identifying twins would be impossible, identifying when kids meet compulsory school age is impossible and tracing benefits on social programs that are age dependent is also impossible. The geographic detail would be a total disaster if it is obscured. It would just about eliminate the field of rural sociology and greatly limit policy studies in which geography has a role.

Eugene Hammel, University of California Berkeley, Demography (Faculty) return to top
Reduction in specificity at higher ages will make it very difficult to track ongoing changes in survivorship at these ages, just as that population is increasing and just as the results of analysis become more important.

Confidential Response return to top
As a geographer it is particularly spatial distribution on the neighborhood level I am interested in. PUMS detail is thus EXTREMELY important

Confidential Response return to top
For my own research, the 2000 census is not all that important as I mainly look at the 1940 census, but I could see that any reduction would harm the general trend of scholarly research.

Amanda Brown, UC Berkeley, Other Academic (Graduate Student) return to top
I am studying occupational effects of ethnicity and immigrant status, so it is very important for me to be able to see the ethnic break down by as many occupational categories as possible.

Confidential Response return to top
Ability to track racial trends in lieu of new multi-race question. Ability to track social/demographic/ local research for business, industry and government.

Anupam Tyagi, University of Utah, Economics (Graduate Student) return to top
An area of my research interest is social and economic inequalities. This requires detailed data on all aspects listed above that characterize a social and economics group. Loss of details means loss of criteria on which such groups can be differentiated to begin such research. Loss of geographic details will be a big loss since my future research will heavily depend on geographic variations in the sample. It has been delightful to use the well organized samples by IPUMS. With the long form in 2000 Census I was hoping to have more detailed information in the samples available for research, NOT LESS!!!

Confidential Response return to top
Since I do geographic research, an increase in PUMA size would have a great impact on geographic detail in less densely settled areas. The effect would also be a headache in regional studies.

Confidential Response return to top
Socioeconomic studies of behavior in group contexts requires information about relatively small differences between individuals: just as "non-white" doesn't capture much information about culture, "over $50,000" doesn't capture much about standard of living--and neither helps one understand the socio-economic environment faced by the individual or group in American society. Ask your Lexus-driving friends what it's like to own a new Ferrari and you'll see what I mean...

Gunnar Thorvaldsen, University of Tromsoe, History (Faculty) return to top
Since the most reliable detail on ancestry comes from old people, grouping them agewise reduces the value of the data set for immigration and ethnicity research substantially.

Confidential Response return to top
I would only use PUMS data in comparative research with Britain, therefore the more detail available, the easier comparisons would be made.

Barbara Reskin, Harvard, sociology, Sociology (Faculty) return to top
Among my research areas is the study of trends in occupational sex and race/ethic segregation, using 3-digit census codes. There would be no way to monitor these trends or explain cross-sectional variation in trends with more aggregated data. It is well known that occupational segregation (like God) is in the details, and when one sacrifices detailed data, one obtains the illusion of considerably less segregation. It would be impossible for me and others who pursue related research to continue to monitor segregation trends. This is critical because segregation is both the cause of other forms of gender/race/ethnic inequality and it is an important indicators of the extent of inequality.

Robert Kaufman, Ohio State University, Sociology (Faculty) return to top
My own current and long-time research using PUMS data analyzes employment segregation and earnings gaps among race-sex groups. In my current research and analyses (funded by a multi-year NSF grant under the Human Capital Initiative and now by a grant from the W.E. Upjohn Institute for Employment Research), I constructed 1961 labor market positions (comparably defined in 1990 and 1980) starting from 6 digit occupation-industry combinations and aggregating combinations only as needed to have sufficient cases in a position to analyze positions as units. Reduction in the occupation or industry detail would be disastrous for this and other similar research in the field because employment segregation is very sensitive to the level of data available. More broadly the extensive and growing research literature on sex segregation has relied heavily on measurement of sex composition at the detailed occupation level. Reducing occupational detail would seriously hinder such research and analyses of trends would be rendered impossible. Similarly, segregation researchers have just recently begun to move beyond black-white-hispanic comparisons to study segregation using more detailed definitions of race and ethnicity. Again such research would be rendered impossible if race-ethnic detail were eliminated.

Confidential Response return to top
I have analyzed occupational group trends over time, and the reduction of data on occupations would be difficult to continue doing this historical analysis.
I also plan to look at racial and ethnic differences in occupational mobility over time, for my next paper and possibly in future research. By not maintaining historical consistency in occupation, race and ethnicity, or income, my research would be jeopardized.
Finally, changes in age groups, particularly for the elderly 65+, would inhibit me from doing future research with the IPUMS on changing work trends among the elderly.

John Kenny, University of Chicago, Other Academic (Graduate Student) return to top
My research looks at immigrant associations. Without accurate information about ancestry and occupation there is no way I can accurately estimate the size of an immigrant community or the level of occupational segregation among immigrants from particular countries. Detailed occupational classifications are especially important. Segregation can't be found in broad occupational categories.

Louis Chauvel, Institut d'Etudes Politiques de Paris, Sociology (Faculty) return to top
For age period cohort analysis of standard international classification of occupation, I need a complete Occupation code. I need also exact age for the calculation of exact cohorts. Discrepancies in these information will disturb the measure and break long term series.

Carl Schmertmann, Florida State University, Demography (Faculty) return to top
(1) Aggregating the elderly into larger age groups would completely eliminate our ability to study many important policy questions.
Some of my recent research with PUMS data concerns living arrangements of men and women 60+ years old, by single year of age. There are important changes in living arrangements at ages 75+ that would be undetectable in age-aggregated data. For an example of analysis that would be compromised, see C.P. Schmertmann, M. Boyd, W. Serow, and D. White. "Elder-Child Coresidence in the United States: Evidence from the 1990 Census", Research on Aging, Jan 2000.
As the US population ages in the early 21st century, it would be extremely irresponsible of the Census Bureau to eliminate age detail from one of the primary sources of policy information on elders.
(2) Migration research demands geographical detail and large samples. There is no source other than PUMS data for many migration-related issues. Aggregation to higher geographical levels would destroy our ability to study many important policy questions, particularly those related to intra-state population distribution.

Gordon Dahl, University of Rochester, Economics (Faculty) return to top
Income measures are vital for the research economists carry out. Reducing detail on the key income variables would greatly reduce the value of this dataset for economists.

Robert McCaa, U. of Minnesota, History (Faculty) return to top
The proposed changes seem ill-considered. The brouhaha this past Spring was not about microdata, it was about questions on the long form.
Before the census bureau embarks on a radical new policy that could seriously harm one of the largest, most useful, and most used social science databases in the world, there should be a study to assess confidentiality risks in the microdata.
The confidentiality threshold must be extremely high (say the odds of linking microdata to another database at 1 in a million), but there should be some study of the risks before hand.

William Collins, Vanderbilt University, Economics (Faculty) return to top
Much of my research concerns changes over the course of the 20th century in racial income, education, and occupational gaps. This kind of research is crucial for understanding how policy has (and has not) affected differences in well-being across groups. The IPUMS data has been indispensable to this research because the empirical analysis demands detailed information on age, race (and ethnicity), nativity, education, occupation, industry, migration history, and location. Moreover, the analysis requires a strong degree of comparability over time. More generally, compromising the richness of the IPUMS data would substantially hamper economists efforts to do the research that guides (or should guide) policy evaluation and design. Having worked with the data extensively, I'm fairly certain that any gains associated with more stringent confidentiality restrictions will be far outweighed by losses in research quality and quantity.

Steve Coble, Vermont Agency of Commerce and Community Development, Government (Nonacademic) return to top
In Vermont the population is very small, 600,000 for the state. We currently have only 5 pums areas for the entire state. Fewer geographical areas would reduce our ability to identify areas of need for the programs that the Agency of Commerce offers.

Monte Aaker, Minnesota Housing Finance Agency, Government return to top
The reduction in detail would cripple our efforts to target housing resources. We are trying to provide housing resources to households traditionally underserved by the private market. We need this detailed data at the smallest geographic area possible. Any reduction would harm our efforts to better target our resources. We need the age detail to provide better location decisions for our housing investments. Housing lasts a long time and having detailed data is needed for long term planning. This is especially important with the aging of the baby boom.

Confidential Response return to top
I have used the PUMS data for two purposes within the last month. First, to detail the number of school-aged children by race in the Philadelphia Metropolitan Area (city and suburbs) for the last 100 years. Losing detailed age or geographic information would have made this analysis impossible. The second way in which the data were used was to predict private school attendance across PUMA areas of different race and socioeconomic compositions for people of different races, occupations and poverty/income levels. Losing any of this information--but particularly increasing the size of the PUMA areas--would have made the analyses far less useful. As it is, it may be that PUMA areas are too large in size (and don't match with other geography in useful ways) for this particular analysis to be useful enough to be considered for publication or policy-related issues.

Susan De Vos, University of Wisconsin-Madison, Sociology (Researcher) return to top
A major activity at present is to examine the living arrangements of elderly Hispanics. Age is very important as the "old old" (80+) is often of special concern. Much hinges on being able to examine people by detailed age, even among this oldest group. Also, ancestry is very important as there are marked differences among Hispanic groups. Finally, source of income is very important as SSI and Social Security are differentially received. Whether income comes from investments or wages also is crucial, but if this detail were kept, categorizing income in increments of $5,000 (current $) would probably be okay. On the other hand, occupation is dealt with simply and my research probably would not suffer from a much less detailed coding. The same is true for geographic location.

Confidential Response return to top
USDOT sponsors long-term research on travel forecasting models that rely heavily on Census data, including PUMS. Commercialization of these research products, under TRANSIMS, is commencing this year and data availability will strongly influence the success of this effort.
These models are used by States and local agencies to identify transportation needs and develop transport programs. In addition, new requirements on social equity analyses will have to be supported by demographic data that can be compared in a time series using a consistent geography.

Irene Bloemraad, Harvard University, Sociology (Graduate Student) return to top
My own research is on immigration. I need to have the most detailed place of birth and ancestry codes as is possible to have. If census data are only released with the major immigrant/ancestry countries (as is done in Canada), it is almost impossible to track changes in immigrants' demographic and socio-economic characteristics over time since the key immigrant groups change. For example, Portuguese immigrants were numerous in New England in the 1960s and 1970s. If, however, no identifying information is provided on them in the 2000 census, it is impossible to see how they are doing, nor how their descendants are doing. Since the Portuguese on average have low levels of education, it is important from a public policy and academic standpoint to know whether they were nonetheless able to achieve economic success (what the data imply) or whether they have had problems. This would give us an indication of what might happen for other immigrant groups with low levels of schooling. Given these concerns, it is also very important to have the most detailed information possible on income, quite detailed information on occupations, and extremely detailed information on year of migration to the US, citizenship status and age. I am currently comparing immigrant populations in the US to those in Canada, an important exercise for understanding the institutional dynamics of immigrant integration. Canadian data are often aggregated in large groups (e.g. just 28 place of birth/ancestry codes are supplied) which makes my research very, very difficult. It has gotten to the point that I have formally requested access to the confidential Canadian statistics (those NOT released as PUMS) because otherwise my research is impossible. If the US Census Bureau were to reduce the detail of the PUMS data, many academic researchers would desire assurances that they could have access to the detailed data through other channels, such as going to the Census Bureau in the summer to run statistical models. Does the Bureau (and the federal government) have the money to set up a mini-statistical center in order that academics and other interested officials can use the more detailed data? If so, how fair is it to restrict access to such data to a few privileged individuals? The beauty of IPUMS is that many people, in the US or elsewhere, professors or graduate students, can use the data. Greater use increases the chances for significant academic findings that could be important for public policy.

Caroline Hoxby, Harvard University, Economics (Faculty) return to top
For analyzing the rate of return to education, it is absolutely necessary to know a person's exact age, so that we can associate him with the regime (laws, funding) under which he attended school. We also need the exact age for estimating consistent wage equations. The geographic detail that was provided in 1980 and 1990 was too aggregated for many of the projects related to urban poverty, crime, and education. If it becomes more aggregated, it will be a disaster.

Allen Glicksman, Polisher Research Institute of the Philadelphia Geriatric Center, Sociology (Researcher) return to top
I try to understand the way immigrants assimilate into the American mainstream, especially in terms of employment. This has important policy as well as academic implications. Without ancestry, we cannot see how members of different ethnic origins are doing. Without specific employment data comparable with previous census data, we cannot look at assimilation over time.

Myron Gutmann, University of Texas at Austin, History (Faculty) return to top
I work on relationships between population and environmental conditions, which requires as much geographic detail as possible (because areas are always so different). This is especially so in rural areas, where the population is sparse and PUMAs are spatially large. If PUMAs were much larger than they were in 1990, it will be impossible to do research on this important area of the relationship between agricultural land use and population, and therefor it will be impossible to make a policy contribution in this area that is vital to Americans who live in cities and on farms.

Confidential Response return to top
(The first sentence was removed for the purpose of maintaining the author's confidentiality.) Most of our patrons ask for detailed age groupings of the variables of interest. The variables of income and occupation are used most by HR students and income and rent and other expense variables by students of Marketing or finance. Sociology researchers seem to use a wide cross section of variables depending on their area however age and income seem to appear frequently there as well. Education levels are also used as grouping variables in many analyses.

Confidential Response return to top
I use the income variables to make sure that other survey data are reporting reasonable amounts. Thus a reduction in precision for the PUMS data with respect to income could greatly reduce the ability to determine precision in other data sets. Additionally, I use geographic data in order to analyze specific proposals affecting only certain areas of the country. Again, a reduction in geographic precision would reduce the ability to effectively analyze policies.

Confidential Response return to top
I have been studying differences in the gender wage gap across metropolitan areas over time. Clearly, grouped income information would completely preclude this kind of analysis, and grouped geographic information would reduce its usefulness.

Ruth Turley, Harvard University, Sociology (Graduate Student) return to top
I'm specifically interested in city-level, race-specific variables (i.e., income by race, education by race, etc.). I cannot proceed without race-specific data.

David Cutler, Harvard, Economics (Faculty) return to top
Perhaps there could be a process where researchers could apply for more detailed data and use it with special permission, if they have a secure computer. I know that HRS/AHEAD use this method. The Census Bureau also has Census locations around the country that people can use. That might be another solution.

Confidential Response return to top
My research includes earnings differences by sex and race. I also examine the effect of working conditions and job injury risk on earnings. I could not estimate reliable earnings equations without adequate data on income, occupation, race, and age. Some specific examples: I used PUMS detailed occupational and industry data to calculate gender-specific injury incidence rates by 3-digit occ and ind. The BLS injury data is provided at the 3-digit level and I needed the corresponding number of workers in the 3-digit category. I anticipate doing research on discrimination in hiring and the PUMS racial and occupational data provided information on the availability of potential workers. I have a Ph.D. student who is looking at compensating differentials for commuting and is using the detailed geographical information.

Confidential Response return to top
We are interested in behavior of pre-retirement and post-retirement individuals that directly affect their economic outcome. Grouping data for age or income will destroy what microdata are good for in this instance. Race and ancestry information are also crucial in determining how policy can affect different groups, and it is also crucial for us to retain this level of detail

Confidential Response return to top
I frequently need to know information on target audiences based on age, income, occupation, and occasionally race. If any of these statistics are affected, it could have impact on my decisions.

Mark Hayward, Population Research Institute, Penn State University, Demography (Faculty) return to top
Social scientists are making great strides in incorporating geo-referenced concepts into their research. Spatial-temporal perspectives, though relatively new, hold considerable promise for new insights into endemic social problems. Increasingly sophisticated applications of GIS technology and spatial statistics are the new analytic tools of social scientists working in this area. Without high quality, geographically-refined data, the application of these tools will be seriously hampered. Potential scientific advances will be eliminated.

Confidential Response return to top
I specialize in analysis of racial and ethnic fragmentation. A reduction of the available categories would greatly harm the meaning and informativeness of the heterogeneity indexes I construct.

Olivia Mitchell, University of Pennsylvania, Economics (Faculty) return to top
To do any sort of retirement research or analysis on the wellbeing of the older population it is indispensable to have detail on age. To do policy research on the most important of all national programs -- Social Security, Medicare and Medicaid, long-term care, tax policy -- one must have access to geographic detail, income and wealth, race, and age. I cannot emphasize strongly enough how critical it is to NOT group age data.

Confidential Response return to top
My opinion is that confidentiality can be preserved by large groups at the local level and it is useless to reduce detail in other variables like age, occupation ... which will affect the precision or make impossible thousands of studies. If small geographic units are of interest to many, it is necessary to use large samples to have significant results, so that PUMS are not relevant. Trying to accommodate everyone might destroy everything.

Confidential Response return to top
I do not actually do research, I am a statistical consultant and help the faculty and students do their research. My answers here are based on the research I have helped my clients with. We have many faculty and students (graduate and undergraduate) who use the PUMS data for many different purposes. What would be catastrophic to one may be of no consequence to another. Without a doubt, though, income and geographic location would seriously affect everyone.

Stewart Tolnay, University at Albany-SUNY, Sociology (Faculty) return to top
Reduction in race or socioeconomic detail would seriously affect my research, especially if comparability with previous PUMS files was compromised. Also, I often merge individual- and household-level data with information for geographic areas (for example MSAs) to do multilevel analyses. Therefore, changes in the level of geographic detail could also interfere with my research objectives.

Richard Wright, Dartmouth College, Geography (Faculty) return to top
I do research occupational segregation by race/ethnicity/nativity. I also use the income data to study immigrant impacts on the native born. I use PUMAs to build local labor markets. If these data were compromised from the versions we have for 1990, that would compromise my ability (and that of related research) to investigate employment, income, and migration effects of the foreign born. These are crucial policy issues facing the USA

Richard Kleban, Ohio State University, Sociology (Graduate Student) return to top
I focus on understanding the relationship between social disadvantage, race, and community crime. Without detailed race, occupation, income, and geography information I am unable to measure the key concepts that are thought to link social disadvantage, race, and crime. To lose this detail would seriously handicap our efforts to understand how the social disadvantage of minority communities leads to high crime.

Tomas Jimenez, Harvard University, Sociology (Graduate Student) return to top
Questions about race and ethnicity are vital to my research. The demographics of this country are changing quite rapidly and complete data on race and ethnicity are crucial for understanding the impact that these changes are having and will have on the United States.

Ellen Meara, Harvard Medical School, Economics (Faculty) return to top
A significant portion of my research deals with income inequality. The PUMS is really the best source of information for deriving reliable, comparable, measures of inequality over time. By reducing the level of detail on income, geographic detail, age, or race, this would limit one's ability to assess at what level inequality might matter (states, MSAs, PUMAs, within race, within age groups, etc.).

Susan Brudvig Other Nonacademic return to top
We use the data for school districts' applications. Geographic detail is INDISPENSIBLE for local areas and policy decisions. Also, we have found socio-economic characteristics to be CRITICAL in accurately projecting school-age populations for small geographies. We need this data for local governments!

Giovanni Peri, Bocconi University, Milano, Economics (Faculty) return to top
I study local human capital externalities in US cities, therefore reducing information on income and schooling would be very harmful. I inquire the effect of the neighbors on the productivity of a worker, so that the very detailed geographical level is very useful.

Confidential Response return to top
Much of my research concerns issues of racial segregation and urban poverty. Without detailed information on race, income, and geographic identification it becomes substantially more difficult to test hypotheses. For example, I am interested in determining the impact of central city gentrification on the poor. To adequately conduct this important research, it is essential to know (1) what city, if any, a respondent lives in, (2) what part of that city, if it is large, (3) detailed information on that person's occupation and income. Item (3) is what allows me to differentiate persons of low "current" income from persons of low "permanent" income. My work on segregation and its impacts on individual outcomes requires detailed geographic identification and accurate measures of income.

Confidential Response return to top
We use PUMS for a variety of purposes. The whole point is to get the kind of detail and cross tabulations that are not obtainable in standard tables. My past research on the elderly in Minnesota consistently shows powerful differences among the elderly population when broken down by age, i.e. people 90+ versus those 70-74. These in income, living arrangements, etc., show service providers that just talking about "the elderly" is not very useful because of these great differences within the elderly group.

Confidential Response return to top
Most of my research involves analysis of income and occupation so any reduction of that information would make it necessary to use more detailed data that are not as extensive as the PUMS. Also, as the population ages and becomes more diverse it will be necessary to look at these analyses in terms of age and race which would again necessitate detailed PUMS data.

Darcy Hango, Ohio State University, Sociology (Graduate Student) return to top
I do population/demography research in sociology and find it very distressing that ages might be grouped into categories. The researcher needs the flexibility to be able to choose their own cutpoints for ages. For example, grouping together ages 15 to 19 is not very useful if you are trying to study high school completion.

Roland Daeumer, The Pennsylvania State University; Population Research Institute, Demography (Post-doc) return to top
The Social Security administration expects us to evaluate pension policies. This implies the necessity to analyze retirement decisions at every age of individuals 65 years of age and older. Grouped age data would severely reduce the predictive quality and accuracy of our forecasts regarding retirement decisions of the elderly population and their outcomes. This is especially the case if we are to evaluate retirement policies intended to alter pension and social security eligibility. Thus, the successful implementation of policies aimed at securing the financial integrity of government subsidized health care and retirement programs will be at stake. will be

Confidential Response return to top
I perform analysis on economic opportunity which requires detailed information on the income and occupational categories, and then I look into certain racial/ethnic/age categories by MSA and analyze changes over time. So, all of the variables that are in question would affect my research, and compatibility of the data over time is extremely important to me.

Confidential Response return to top
This would drastically reduce the ability of researchers attempting to conduct cross-level research in many areas of sociology.

Confidential Response return to top
I have used census data in many areas of my research. Calculation of single year of age specific labor force and disability rates has been important in my work on the effect of health on early labor force exit, the geographic detail has been essential for my work on local labor markets, the industry detail has been important for understanding changes in local labor markets, the occupational detail has been essential for my own work on racial earnings differentials, and has been essential for some of my students work on gender differentials. The geographic and age detail has also been central to my work on health differentials across rural and urban populations.
The main advantage of the census over the SIPP or CPS is its large sample size. This allows researchers to use the data to study phenomena that one can not study with the CPS. The geographic, occupation, age etc detail available on the Census is crucial for all of this.

Seth Ovadia, University of Maryland College Park, Sociology (Graduate Student) return to top
My current research is on occupational and residential segregation across metropolitan areas. My plan for the next stages of this research are highly dependent on the availability of 2000 Census data which are comparable with the 1980 and 1990 data, as change over time is an essential part of both my and other segregation researchers' plans for doing policy-relevant analysis. Any changes in the occupational and geographic data formats would make such work simply impossible and cut off a vibrant research tradition before it could take root.

Confidential Response return to top
As an economist who does research on the changing economic position of women, it is absolutely necessary to know not only the age of each individual, but particularly their specific occupation (to understand whether the differences in economic outcomes are due to occupational segregation versus lower wages, etc. within occupation). In addition, it important to have specific income information; grouped income data would make it very difficult to back out women's relative positions.

Suzanne Duryea, Inter-American Development Bank, Economics (Researcher) return to top
My research utilizes the PUMS to establish a baseline of outcomes for comparison with household survey data and census data for Latin America. We are currently examining issues of social exclusion in Latin America and it is critical that reliable information race and ethnicity be available in the PUMS. Perhaps more importantly for my research is the potential indirect effect of the elimination or reduction of data coverage in the PUMS. In high-level dialogues with governments and statistical offices in LAC, the PUMS is used as a model survey for detailed socio-economic questions. Other governments will be much less willing to include critical questions in their surveys if the US government restricts either the collection or accessibility of certain socio-economic and demographic variables.

Jamie Strickland, Western Kentucky University, Geography (Faculty) return to top
I am a geographer who is primarily involved with aging issues. Therefore, I am extremely concerned over a possible decision to reduce BOTH the geographic detail as well as the age detail in the 2000 PUMS. PUMS represents one of the very few datasets which allows me to examine older populations at the state and substate level. I would like to stress that I consider anyone over the age of 40 as being a part of the "older" population. Only considering people over the age of 65 as being part of the "older" or elderly population is incredibly short-sighted. Specifically, my research considers the changing nature of retirement and its potential impacts on the spatial mobility of older persons. It is becoming increasingly common for retirees to be under the age of 65. In addition, not all individuals over the age of 65 have left the paid labor force. These processes do not occur evenly over space nor do they have the same economic and social impacts in all locations. As such, there are important local, state and national policy implications for this research that absolutely require a detailed age and geographic coding. I certainly respect the Census Bureau's desire to maintain a high level of confidentiality for respondents, but I do not feel that presenting a "watered down" version of PUMS for 2000 will accomplish this goal.

Eric Millis, Federal Reserve Bank of Dallas, Economics (Support) return to top
There is a particular interest now in the effects of policies encouraging the elderly population to enter the labor force. Clumping all 65+ categories together, which I have heard being mentioned as a possibility, would be a significant setback to such studies.

Gregory Reeves, The Kansas City Star, Media (Nonacademic) return to top
We use PUMS for detailed information on race and income. Those variables are crucial.

Scott Adams, Michigan State University, Economics (Graduate Student) return to top
The research that I do tends to focus on the comparisons of outcomes at the city level. For example, one project that I am working on concerns access to health care for the elderly. Because certain characteristics of cities (like public transportation) influence this and vary by city, I exploit this cross-city variation. Census data is quite helpful in terms of providing the basic characteristics of city residents, such as race, age, and income. I was hoping the 2000 census would allow me to do this. I could use the Current Population Survey (and have done so), but some of my estimation techniques require very large samples.

Joshua Rosenbloom, University of Kansas, Economics (Faculty) return to top
I have used PUMS data for two purposes. One is labor economics research on the impact of childbearing on female labor force participation and work effort. This research used the occurrence of twins in the first birth as a "natural experiment" but we had to stop our analysis after 1980 because the lack of data on quarter of birth for children made it impossible to identify mothers of twins in census data after that date. The other use is for economic history research on a variety of aspects of the geography of American labor markets. I am interested both in measuring economic integration--for which geographic, occupational, ethnic and income detail is needed--and in studying long-run trends in migration for which all of the same is needed along with detailed information about age and family structure.

Confidential Response return to top
Since I am interested in neighborhoods and neighborhood effects, linking IPUMS' individual records to neighborhood level data is very important to me. The Census is the only source of reliable neighborhood characteristics.

Liana Sayer, University of Maryland, College Park, Demography (Graduate Student) return to top
I have used PUMS data to examine differences in rural and urban poverty between the North and South regions of the U.S. I have also used the PUMS-L, which allows examination of rural non-metro labor market areas. If the PUMS contained broader geographic areas, I would be unable to examine how the link between labor market areas and poverty rates for rural areas has changed since 1990.

Thomas Wilson, Florida Atlantic University, Sociology (Faculty) return to top
I am interested in the interplay of race, class, and gender, with particular emphasis on class. The usefulness of IPUMS data for me resides largely in its detailed class indicators, including occupation, income, SEI, etc. A reduction in this detail would be catastrophic. Further, it is hard to imagine how confidentiality could be compromised by maintaining the current level of detail in these data.

David Kantor, Institute for Policy Studies, Johns Hopkins University, Other Academic (Support) return to top
1: I'm not prepared to answer question 5. 2: The geographic identification is poor enough already. We don't need to make it any less detailed. I would prefer in IMPROVEMENT in the geographic identification.

Confidential Response return to top
We use The Census data for Program administration, Community level Industrial Recruitment and other Economic development activities. We are the source link with the local & regional Chambers of Commerce, Industrial & Workforce boards. Prior to the 1990 census we have request special reports for the local areas and wait for them to be compiled. Since the development of the IPUMS Web server, we have been able to access the information in a more direct timely manor. And I am represented by the Oklahoma Second District Congressman who started this mess. But he has never listened to anyone anyway.

Joshua Angrist, MIT, Economics (Faculty) return to top
A key feature of my work is the ability to identify cohort (exact age), states of residence, and states of birth. Also important is the comparability of data on income, employment status, weeks worked, and schooling.

Confidential Response return to top
I do research on the structure of wages, and changes in wage inequality: a reduction in ANY of these characteristics essentially makes the Census 2000 data useless for research in this area. Perhaps this is exactly what proponents of reducing these details would want, because I find it absolutely crazy to think that characteristics like age, ancestry, race, and occupation could ever be used to identify specific individuals. These variables are vital to expanding our knowledge of how the labor market works. The confidentiality issue is a red herring.

Confidential Response return to top
I specialize in labor market research -- anything related to earnings or occupation must be as finely detailed as possible. Race, age, and gender are also important as control variables.
Geographic information and ancestry are less important for my research.

John Hipp, University of North Carolina, Chapel Hill, Sociology (Graduate Student) return to top
My research focuses on internal migration and its effect on labor markets. Thus, geographic detail is extremely important for me in looking at the location of movers. In addition, information on both occupation and income are necessary for this. As well, much of this research looks at migration by racial/ethnic groups, so losing this level of detail (i.e. particular Latino groups) would dramatically reduce the usefulness of the data.

David Cotter, Union College, Other Nonacademic (Faculty) return to top
I am particularly concerned about the reduction in detail in the geographic codes. These have been absolutely critical to both of my major research projects which examine variations in gender inequality and poverty across labor markets. Without this information there would be no way to construct reliable estimates of several key indicators, nor would it be possible to extend this research into the future without the PUMS. Much of the work we have done already with the 1980 and 1990 PUMS needs to be extended with the 2000 data, and it plays a central part in several grants we have proposed to NICHD and NIH.

Confidential Response return to top
I am a university statistical consultant but have helped people work through compatibility issues between 1980 and 1990 data for studying ethnicity and education, and changes in employment demographics between these years. Addition of the 2000 data will require as much consistency as possible; we already had several problems with compatibility between 1980 and 1990!
As the US population ages it seems nonsensical to group ages over 65. How else are we going to understand changes in aging structure from 2000 forward?

Confidential Response return to top
I use single year of age data and data on race and ethnic group to construct life tables in specific geographic areas. Grouped age data would make this difficult.

Confidential Response return to top
I currently work with a combination of ancestry, race, and age variables. Grouping age would preclude me from determining whether the age structure of various elderly ethnic populations differ. For example, are elderly Japanese Americans much older, on average, than elderly Chinese Americans, due to dramatic differences in foreign-birth, duration in the U.S., and life expectancy? Are Japanese Americans concentrated in the "oldest old" group (those 85+), compared to elderly Chinese Americans who may be more likely to be clustered in the 65-74 group? What of elderly Italian Americans, compared to Mexican Americans?
Reducing geographic detail would also complicate research on, for example, Marriage Markets, which generally attempts to look at the most detailed information to ascertain area data most closely linked to risk populations. For example, having an area of 100,000 is a much more reasonable marriage market to contemplate than an SMSA.

Mark Stern, University of Pennsylvania, History (Faculty) return to top
In our study of historical change in work, family, and community during the 20th century, Michael Katz and I would find geography "most scacrificeable" because we will be using the pums to examine metro regions as our basic geographical unit. However, losing detail on individual characteristics that make it difficult to compare 2000 with the earlier files would be "catastrophic' (to use the conventional term from the survey) to the project as we conceive it.

Susan Losh, Florida State University, Sociology (Faculty) return to top
I have already run into excised variables on NSF datasets (race, region). It has been *catastrophic*. I am now trying to get the complete data because my results will not be academically acceptable with these variables omitted.
Please keep the data intact. Don't group if at all possible because different researchers will group data different ways depending on their research agenda.
I think the identifiability issue is somewhat overemphasized. Just be sure the geographic units are large enough, perhaps truncate a few outliers (the local millionaire) and keep these research resources generally intact!
Thank you

Mary Washington, Lehigh University, Sociology (Faculty) return to top
I specialize in the historical changes in racial classification, mixed-race/ethnic identity and the formation of pan-racial and ethnic groups as they relate to social status. Reduction in the detail of racial categories and occupational classifications reported, as well as single year age would greatly harm my ability to make historical projections or.

James Cassell, University of North Carolina at Chapel Hill, Sociology (Graduate Student) return to top
I am particularly interested in changes in the race/ethnic distributions of income, occupations and industries over time for low-levels of geography. The PUMS is the only source of such data that I'm aware of.

Jeff Elsworth, Purdue University, Economics (Graduate Student) return to top
I do wage and occupation research in the hospitality industry. Loss of this detailed occupation data would severely limit my investigation of what drives wages and other occupation specific factors in the food, lodging and travel industry.

Confidential Response return to top
I do a lot of small area economic and health related research. Therefore the any reduction in degree of detail with regards to geographies a/or detail on individuals would be very harmful.

Kristina de los Santos, Ohio State University, Sociology (Graduate Student) return to top
My work focuses on Asian Americans and examines specific ethnic groups under this rubric. PUMS is one of the only sources with large #'s of Asian Americans, and therefore is indispensable to conducting quantitative analyses.

Confidential Response return to top
I use geographic data to overlay macroeconomic variables on my microdata. For instance, unemployment rate in the area that a person lives in often proves useful in explaining labor market behavior.

Confidential Response return to top
My research focuses on Asian Pacific Americans and specifically, on APA neighborhood formations and characteristics. Reduction in PUMS will seriously affect my ability to look at various APA ethnic groups and their community patterns with respect to socioeconomic and other demographic qualities. More generally, my research interest is racial economic inequality and to be able to examine this issue on the micro-level requires that detailed data be available on the smallest geo level - PUMAs - possible.

Stephen Tordella, Decision Demographics, Demography (Nonacademic) return to top
I do labor force research that depends upon having access to exact occupation.

Walter Carroll, Bridgewater State College, Sociology (Faculty) return to top
Grouped age, income, and occupational data would seriously compromise my research on historical family change and the processes of immigrant adjustment to the United States.

Confidential Response return to top
Occupation, age, income, and race are key variables in controlling for net diffusion of population across geographical space. Less detail on these variables would mean that one could control less well for these demographic differences, which potentially increases the bias of the analysis. In particular, standard regression approaches (in spatial analysis) are susceptible to this problem. Semiparametric approaches, such as matching on propensity scores, which is a key approach to get better results than in regression, crucially rely on these background variables to reduce bias.

Arthur Wilke, Auburn University, Sociology (Faculty) return to top
Grouped data precludes developing what may turn out to be more meaningful classifications of people based on a number of indicators.

John Hall, Mathematica Policy Research, Other Nonacademic return to top
I use PUMS in designing samples. It is essential in those cases where the design must consider variation at the sub-state level.

Susan Mayer, University of Chicago, Sociology (Faculty) return to top
Reducing PUMS income detail would prevent the comparison of census data with income data from other data sets thus eliminating one of the most important checks on the quality of data used in estimating poverty rates and income inequality. Reducing geographical detail would eliminate the ability to do some research on inequality and poverty and research on the effects of economic and racial segregation because for some purposes the Census is the only data set large enough for reliable estimates.

Amy Goodin, UNM Institute for Public Policy, Geography (Researcher) return to top
I conduct policy assessments of hazardous facilities as they relate to CERCLA, RCRA, and property value impacts, thus a variety of indicators at sufficient detail is required to assure informed decisions are made with the data I collect (i.e. representative samples, etc.).

Edward Ratledge, University of Delaware, Demography (Faculty) return to top
In doing policy research for the Delaware General Assembly and the Executive in Delaware I frequently use the PUMS to answer questions that cannot be answered by tables. It is also our primary cross-check with the CPS. Most recently, I used the info on households with school age kids and age of housing to develop an estimate for impact fees. Those estimates materially changed what was going to be used in the absence of the information.

Jeff Grogger, UCLA, Economics (Faculty) return to top
Several years ago I proposed that twin births could be used as an instrument to estimate the effects of unplanned fertility. The results, published in the American Economic Review, were quite successful. Without detailed age data (at least age in years and quarter of birth), this strategy will be unavailable to chart changes in the socioeconomic consequences of unplanned childbearing. We would lose a tremendous resource for analyzing the consequences of this important social problem.

Jeff Claassen, Fort Worth Star-Telegram, Media (Nonacademic) return to top
PUMS data allows is essential for preparing crosstabs on categories that are not in the Census 2000 standard release. For journalistic purposes, there are many trends to examine that can only be done through a PUMS analysis. That includes the demographics of long-haul commuters and other crosstabs.

Susanna Loeb, Stanford University, Economics (Faculty) return to top
Geographic detail is important for my research for matching people to labor markets and government services (child care or schools for examples). Occupational detail is important so that I can identify teachers (elementary, secondary etc) and child care worker as well as identify alternative occupations for workers who choose these fields.

Confidential Response return to top
We use the data to develop quota samples and make estimates of people impacted by state and local level programs. Our work would become almost impossible to defend w/o the PUMS detail.

Russell Davis, Louisiana State University, Sociology (Graduate Student) return to top
I study rural labor markets. Without detailed occupation codes, income, and geographic details, it becomes impossible to identify key demographic groups. With out the detail, all you have is a bunch of people and areas that all look the same. Large population aggregation often times eliminates the place specific detail of rural areas more so than metro.

Mark Yeisley, Florida State University, Sociology (Faculty) return to top
I specialize in inequality research as it relates to crime and crime policy. I require geographically-specific data regarding income, specific occupation, education, race, and other socioeconomic variables relevant to inequality.

Confidential Response return to top
I live in, and do some of my research in, a sparsely populated state in which age and other demographic characteristics are crucial to analyses of the potential or lack thereof of various areas of the state. At a PUMA size of 100,000 it is possible to use Census data as part of such analyses: at anything more we cannot do analysis on anything but a state-wide level.

Confidential Response return to top
Detailed occupations are used to identify individuals in science occupations. Ancestry is very important in categorizing immigrants and immigrant assimilation.

Andrew Beveridge, Queens College and Grad Center CUNY, Sociology (Faculty) return to top
Since 1993 I have had a contract with the New York Times. We have used the PUMS data, often from the IPUMS project, to do a variety of analyses. To lose the sort of detail contemplated in terms of PUMA's (NYC is divided in 54 PUMA's that basically correspond to Community Planning Districts), ancestry, occupation, foreign born place of birth, income, age, etc. would be a disaster. We have had at least six analyses make it to the front page of the Times and one to the front page of the Week In Review. In each case, very detailed PUMS categories were used. For example, for a story on the earnings of male and female college graduates we used single year ages to look for those who had just graduated from college. For a project on income of blacks and whites in Queens we carefully set-up the households in such a way that included PUMA's, as well as age and detailed household characteristics. For a project on earning by occupation for blacks and whites we used the 1950 detailed occupation codes from 1950, 1970 and 1990. Such an analysis would be impossible to replicate for 2000, if these restrictions go into effect.
Currently, I am working on a paper comparing immigrants to New York from 1910 to those here in 1990. The sort of restrictions planned would preclude an analysis using 2000 data.

Lu Chou, Data and Program Library Service, University of Wisconsin Madison, Other Academic (Support) return to top
I am a data librarian and help graduate students, researchers and faculty members on my campus by providing PUMS data to them. I know PUMS are used on our campus for policy research and academic research. Aging and poverty researches are big on our campus. As a service provider I would not want to see the detail level of PUMS be reduced and my users lose some very crucial data for their researches.

Michael Sullivan, Freeman, Sullivan & Co., Sociology (Nonacademic) return to top
I specialize in the analysis of the statistical representativeness of jury venires sampled from the general population. The PUMS data allow estimation of the fraction of the population falling within given race/ethnic groupings, over 18 who are citizens with different levels of language capability within given geographical locations.
I can't imagine trying to do this sort of analysis without this level of detail.

Diane Cowper, Dept of Veterans Affairs, Demography (Researcher) return to top
My research focuses on the migration patterns of retired veterans (and, subsequently, the impact of that migration on the VA health care resources at their destination communities). Life-cycle is important in the types of moves older people make -- early retirees are usually amenity migrants, while the old-old tend to be assistance migrants. It is important to know the types of migrants an area is attracting (or, losing) to plan for adequate services for those individuals. Therefore, age data are extremely important, as are other sociodemographic variables, including (but not limited to): race, gender, income, education, marital status, living arrangements, etc.

William Bridges, University of Illinois at Chicago, Sociology (Faculty) return to top
The best of way of tracking information on specific JOBS in the economy is by a crosstab of detailed occupation by detailed industry (without regard to geography). This is essential for understanding segregation by race and gender. Any tinkering with the current categories would be a disaster.

John Blandford, University of Chicago, Economics (Post-doc) return to top
I research the labor effects of sexual orientation, specifically how sexual orientation impacts earnings and occupational outcomes. Because of the ability (beginning in 1990) to identify same-sex, co-habitating, unmarried partners, the PUMS provides one of the few reliable sources of information on lesbian and gay workers. The other sources for data are the General Social Survey and the National Health and Social Life Survey, both of which I also use extensively. These too are invaluable but suffer from two limitations for the purposes of my research. Firstly, they provide only categorical income data and, secondly, the sample of the target population is rather limited. In contrast, the PUMS provides a full complement of earnings and supporting data on nearly 10,000 full-time workers who identify as same-sex partners. The 1990 PUMS has proven an invaluable complement to other sources of employment data on this population. In particular, occupational clustering patterns significantly determine differential earnings outcomes for gay and lesbian workers, but these clusters are highly nuanced and an accurate estimate of their impact requires occupational data down to the three-digit level (505 titles). Studies on this topic that rely on GSS/NHSLS data measure large wage differentials attributable to sexual orientation. Because of the limited sample size, however, they are unable to generate a meaningful measure of the impact of occupational clustering patterns as a determinant of the earnings differentials. The outcome has been biased estimates of the impact of sexual orientation in labor-market outcomes. It has been precisely the availability of the PUMS, with its full complement of nuanced employment-related data, that has allowed me to conduct research in this area. Indeed, I eagerly anticipate the release of the 2000 PUMS which, if it provides complete data compatible with those of 1990, will provide the first opportunity to see how the economic status of this population has changed over the course of a decade. Needless to say, it would be a horrible loss to many researchers were the Census Bureau to release 2000 PUMS data in a form that would make cross-year comparisons impossible.

Michael Haines, Colgate University, Economics (Faculty) return to top
I will have to revise Historical Statistics of the United States using the 2000 PUMS. Since I believe that the published material will be very limited, I think that a very detailed PUMS will be indispensable to anything with the census that requires any geographic, age, race, occupational, income, marital status, nativity, or other detail.

Charlotte Steeh, Georgia State University, Sociology (Faculty) return to top
Loss of detail deprives researchers of the flexibility to create the groups that are most relevant to their hypotheses. In addition, predefined groups rob analysts of the powerful set of statistical methods that rely on interval and ratio level measurement.

Christopher Lyttle, Rush Primary Care Institute, Sociology (Researcher) return to top
Census data, including PUMS, are important to us in our research into the sociological covariates of asthma. Sometimes we use these data to provide the sociological context of medical events extracted from other data sets. In such analyses we do not match individuals, so it is important to have area as specific as possible. For some of our work, counties are the minimum level of analysis, for other projects we work with Chicago community areas.

Charles Brown, University of Michigan, Economics (Faculty) return to top
For research in labor economics, much of the point of PUMS is the detail. If I want to estimate relationships using only variables that have many respondents, I can use CPS. PUMS is useful for those who want to study relatively small groups (e.g., individual national origin groups) or groups that become small because of cross-tabulation (e.g., older black workers). The direction of the proposals would endanger both types of studies.

Ada Haynes, Tennesse Tech University, Sociology (Faculty) return to top
Since my analysis is a longitudinal study of non-metropolitan areas any changes would be devastating.

Confidential Response return to top
Q#6: Occupational categories can always be collapsed by the researcher, later, but if it's collapsed in the data set, then any refined analysis would be impossible. Q#7: I use both SEI and manufacturing codes, as well as historical comparability. Q#: PUMAs that also allow for classification by region, state, etc. are important as well.

Confidential Response return to top
Comparison of census demographic data to sample survey demographic data.

Nadine Misiaszek, San Diego State University, Other Academic (Support) return to top
Pre-grouped data is a limitation. I need to be able to group the data in the ways that the survey requires. For example, a study on child care needs to be able to get at children to age 12 (not age 18); a study on binge drinking needs to pull out data for unique years within the age span of 18 to 30.

Joyce Jacobsen, Wesleyan University, Economics (Faculty) return to top
One of the main things I do is to analyze workers by occupations within industries, including calculating segregation indices and comparing them across time and space. Having a lot of categories, and historically comparable categories for occupation and industry, is very important for me.

Carl Mason, UC Berkeley, Demography return to top
The most important use to which I put IPUMS right now is in teaching an undergraduate course on the History of Immigration. In that course *undergraduates* use IPUMS data to investigate labor market mobility, intermarriage patterns and residential segregation. Since the US is presently in the midst of a wave of immigration that in so many ways is comparable to the one that took place in the late 19th and early 20th centuries, the comparability of data between that earlier time and the present is extremely valuable. To lose detail on ancestry, occupation, race or income in the 2000 census would mean that the clock would stop in 1990. For undergraduates, this will soon be ancient history.

Robert Morgan, Houston VAMC/Baylor College of Medicine, Other Academic (Faculty) return to top
I do a lot of work on health care utilization and outcomes of care among the elderly and poor (e.g., Medicare, Medicaid, and uninsured). In particular, my current work involves differences associated with geographic and race/ethnic variability among the elderly. Thus, I need to be able to compare data across years (especially recent years, e.g., since 1980), across geographic regions (at least down to PUMAS of 100,000), and across race/ethnic categories; and I need to be able to match data for specific geographic areas with aggregated data obtained from other sources. The finer-grained the data available to me through PUMS are, the more flexibility I have in developing and answering my research questions.

Jon Lorence, University of Houston, Sociology (Faculty) return to top
My research has focused on occupations and income inequality. Any reduction in the level of specificity of such information would make it impossible to conduct any research that would be useful in this area. Failing to provide detailed PUMS information would gut social science research in the United States.

Confidential Response return to top
A less detailed PUMS with harm greatly my current and future research, because I am focusing on comparing the trends in self-employment among Hispanic groups, especially Puerto Ricans, Cubans, Mexicans..

Confidential Response return to top
I study the characteristics of immigrants, and mainly their earnings pattern. The reduction in detail on age, income and country of origin would affect badly the analysis. It is very difficult to work with variables that are defined using groups, that is especially true when we talk about numbers (age, income) and may not be such a very big problem when you talk about categories (occupation etc.) - though depending on the study a range not broad enough for the categories may render the variable itself useless.

Karen Woodrow-Lafield, Mississippi State University, Demography (Faculty) return to top
I would like to use this area to give more detail about my response to item 9. I am reluctant to see any delay to release of public-use microdata files from the 2000 Census, even if a less detailed dataset were initially released. A fuller microdata file release could be made available to users who had purchased the earlier files. Another possibility is that more detailed microdata files could be available through the various Census Research Data Centers, or through the NCHS Research Data Center.
My research specialization in immigration means I am very interested in highly detailed data as to country of origin, race and ethnicity, timing of immigration, and various measures of socioeconomic assimilation within the United States and specific geographic areas. The PUMS files are already limited for several purposes for which current data sources are needed to study recently immigrated origin groups, some of which are highly concentrated within certain metropolitan areas. Asian immigrants are more heterogeneous than our studies may examine.

Elizabeth Mustaine, University of Central Florida, Sociology return to top
I specialize in crime-related topics. If I did not have detailed information about the size of geographic locations, I would not be able to do my research. Also, categories like age, race, occupation, and income are always important indicators of crime rates. With out specificity in these areas, criminological research would, essentially come to a stand still. At best, we would simply be unable to make comparisons between research done on this data and research done previously. Consistency is also important. The only way to provide a better understanding of a social issue, or to make more efficient and effective policy is to continue to study an issue over time, using similar/the same measures. My view is that changing the way variables are categorized, and losing the specificity of variables will be a huge loss to the social sciences and, thus, to policy makers.

Leif Jensen, Penn State University, Sociology return to top
I have done work on poverty among elders, and grouping age cats, esp at oldest ages would be extremely problematic. I also do work on poverty and income, and grouping income categories would greatly reduce the flexibility to experiment with alternative poverty definitions, and would greatly reduce ones ability to examine inequality. I also am a rural sociologist and even as is the PUMS is woeful in providing the geographic detail needed to understand the circumstances of rural populations in the US. It would be devastating to lose the geographic detail on rural/urban, metro/nonmetro, size of metro, state, etc., etc.

Confidential Response return to top
PUMA size increase would preclude the ability to run custom crosstabulations for New York City neighborhoods (community board districts are for the most part equivalent to current PUMAs) and some smaller counties in the New York region.

Catherine Weinberger, University of California Santa Barbara, Economics (Researcher) return to top
Careful research based on detailed information is vital to the design and implementation of reasonable government policies. When I study differences in earnings between technically trained immigrants from different countries and U.S. natives of different ethnicities, detailed information about country of origin, ethnicity and earnings is essential. When I compare the labor market conditions faced by men and women in different fields of engineering, very detailed information on occupation and age is essential. When I studied the relationship between mathematics education and later labor market outcomes, it was valuable to have detailed information on occupation and earnings that could be matched across Census cohorts.

Rebecca Dennison, University of Minnesota, Other Academic (Researcher) return to top
unknown

Masao Suzuki, Public Policy Institute of California, Economics (Researcher) return to top
My research is on economic achievement of racial and immigrant groups, so I would be most concerned with any loss of detail about race, ancestry, and immigration. Since my work is mainly historical, comparability with past IPUMS/census records is critical.

James Walker, University of Wisconsin, Economics (Faculty) return to top
The PUMS plays a major role in analyzing individual determinants of migration. Geographical locations need to be measured with as much detail as possible to correctly characterize the source and destination locations. That is, it is important to know whether an individual moved to New York City versus upstate New York, or for that matter that the destination was Houston and not Dallas. Aggregation to census region, or state is too broad to be of anything other than the most basic descriptive use. Summary by rural/urban is also too broad. Reduction in the geographical level reported in the 2000 PUMS will have a serious deleterious effect on migration research.

Confidential Response return to top
Some of my research is on the marriage patterns of older women (especially related to the Social Security re-marriage penalty). I need to calculate marriage rates by age for women ages 55-75, and this would be impossible with age-grouped data.
Detailed geography data are also important for two reasons. First, the Census is the only data set researchers have that is large enough to be representative at the state level. Second, identifiers for PUMAs make it possible to adjust for differences in local cost-of-living, which is important in any study concerning the economic status of the elderly.

Mark Fossett, Texas A&M University, Sociology (Faculty) return to top
My work often focuses on comparisons across metropolitan areas and/or tracking patterns in metropolitan areas over time. Loss of geographic detail would make this work much harder.
As it presently stands, my wish is to get easier access to even greater levels of geographic detail (e.g., census tracts within metropolitan areas).

Aimee Chin, MIT, Economics return to top
I have been doing some research on immigration, and it would be devastating if I would no longer have details on country of ancestry and year of immigration.
Also, in all my research, I prefer using as fine a geographic category as possible. The state level is too large, especially because large states tend to be very heterogeneous. When examining the effect of policies, sometimes one needs to be able to distinguish NYC from Buffalo, for example.

Frank Lenk, Mid-America Regional Council, Economics (Nonacademic) return to top
Our subject of study is the Kansas City metropolitan area. Geographic detail must at least maintain large metropolitan areas.

Halliman Winsborough, ICPSR and University of Wisconsin, Madison, Demography (Faculty) return to top
Truncation of the age coding would obviate research on the oldest old, the most important policy age category in this period. Reducing geographic coding will nip in the bud the fruits of GIS.

Confidential Response return to top
We are particularly interested in PUMS because it allows us to better understand age/race/income dynamics, and age/migration dynamics.

Francine Blau, Cornell University, Economics (Faculty) return to top
The detailed occupational categories in the PUMS have been used extensively by scholars (including me) to study changes over time in the extent of segregation of women by occupation. A recent study I completed examining determinants of marriage rates used the variation in labor market conditions across MSAs to study their effects on marriage. Data at the PUMA level was necessary to ensure comparability across census years due to changes in MSA boundaries. Important differences were found by race making this data also very important. The study focused on younger individual and controlled in a detailed way by age.

Diane Lauderdale, University of Chicago, Demography (Faculty) return to top
My interests are in cohort effects on aging; eliminate age detail were preclude most of my use of census data. I also need state-level socioeconomic data for the elderly.

Michael Conaway, Institute for Social Science Research, University of Alabama, Demography (Researcher) return to top
My interest in the PUMS data lies in using it to characterize small geographic areas of the state and as a ground truth for correcting for the inevitable selection biases within those areas. The geographic breakdown is already too large to be useful to me in many cases.

Harold H. Kassarjian, UCLA and Cal State University, Northridge, Other Academic (Faculty) return to top
Reliability and validity of research results would be lessened, along with credibility of the findings.

James T. Ault, III, Creighton University, Sociology (Faculty) return to top
We are exploring spatial analysis of demographic attributes (both contemporary and historic), so detail is essential to enable us to match data from disparate sources. Ethnicity, age, and occupational details increase the likelihood of reasonable and logical comparisons.

Elsie Harper-Anderson, UC Berkeley, Other Academic (Graduate) return to top
Data which has details on race is limited. Especially when studying a specific region or city. This is one of the few datasets which provided enough detailed to both break down into the region in question and the race variables needed while maintaining enough cases to be statistically valid. Without it, my last three research projects would not have been possible.

Leann Tigges, University of Wisconsin-Madison, Sociology (Faculty) return to top
I think of my research area as involving the sociology of work and labor markets. Thus, it is crucial to me to have all of the social and economic characteristics of individuals and as much detail as possible on the labor market area in which they work. We need to understand better the relationships between human capital, occupations and industry, family and personal characteristics, and the community or local economy that matches workers with their jobs. PUMS has provided rich data to understand these relationships across time and space. Restricting data on these dimensions would lead to great difficulty understanding the dimensions of inequality (especially those associated with race, gender, and age).

Confidential Response return to top
Reduction or simplification of race, age, and ancestry data would greatly reduce my ability to collect data on intermarriage and other structural variables concerning immigrant assimilation after 1965.

Robert Fairlie, University of California, Santa Cruz, Economics (Faculty) return to top
1. I would not have been able to conduct my previous research on the impact of immigrants on the U.S. education system and labor markets without the geographic detail. 2. I would not have been able to conduct my previous research on the returns to education without the income, age, occupation, and industry detail. 3. I would not have been able to conduct my substantial research on racial, ethnic, and immigrant inequalities in the U.S. without the income, race, Hispanic origin, and ancestry detail.

Confidential Response return to top
I study various aspects of the "Digital Divide" as they relate to all age groups and socio-economic status for small areas. I rely on PUMS data to teach my students in research methods and government information courses.

Charles B. Nam, Florida State University, Demography (Researcher) return to top
I am a developer of the Nam-Powers socioeconomic scale of occupations and provide a decennial update of the scores, which are used by a broad research public. The scale depends on a detailed occupation classification (although comparability with the occupational classification in earlier censuses is not important), as well as a fair amount of detail on education and income (determinants of the score for each detailed occupation). Without the detail for those three variables, it would not be possible to create the scale (nor would it be possible for competing scales, e.g., Duncan SEI, to be constructed).

Helen Marrow, Princeton University, Sociology (Undergraduate) return to top
I do work on immigration, and so reductions in race/ethnicity, ancestry, national origin group, and/or birthplace categories seriously precludes my ability to evaluate foreigners and minorities according to census data.

Judith Rowe, Princeton University (retired), Economics (Support) return to top
Significant policy research has been done using the PUMS in addition to academic research. It would be a great loss to lose the comparability that we have had for so many years.

Confidential Response return to top
I do research on migration and work. I need the most detailed data I can find on where people live now and where they came from. I'm interested in the effects of work on migration, so I need detailed information here as well. I'm especially interested in the self-employed, too. My interest in migration and self-employment are not necessarily always linked but I might want to look into it. Another important variable would be age. With the changes in age at retirement, I would really like more detailed information on those over 65, more detailed than the categories mentioned earlier.

Scott Czepiel, SUNY - Stony Brook, Sociology (Graduate) return to top
Occupation codes are absolutely requisite for work in social stratification. Broad groupings are sometimes helpful but more often than not, researchers must recode the 505 list themselves for very specific reasons. Eg., % of workers in a given occupation who are women, or who earn higher than the average median income of all workers. Such stats are indispensable and would be irrevocably lost if detail is lost in the occ codes.

JoAnn Dionne, University of Michigan Library, Other Academic (Faculty) return to top
I am a data librarian and, while I don't do research myself, have over 20 years experience in helping academic researchers. I've answered as a combination of most of the users I've helped over the years. You may want to disregard my answers to 4. It is critical that historical comparability be maintained. Age, income, race and occupation have been among the most important variables in _any_ of the analyses that I've provided support for.

David Brownstone, University of California, Irvine, Economics (Faculty) return to top
I do research on transportation demand, and PUMS is the only source of joint distributions of socioeconomic variables. It is critical to have this available at the highest level of geographic desegregation.

Patricia Becker, APB Associates/Southeast Michigan Census Council, Demography (Nonacademic) return to top
My #1 focus in PUMS is on using the PUMAs to identify special demographic groups within a geographic area of interest. I can live with somewhat less detail and will be considering all those options in my role in advising the bureau. However, I will fight against cutting back on ancestry or occupation detail because to do so really spoils the utility of PUMS in identifying demographic subgroups. Race will be analyzed from here to Sunday in the main publications and I doubt that we need all 63 categories at the PUMA level. Income can be slightly grouped (rounded to 100 or 1000) and possibly topcoded. I see no point to making any changes on age.

Richard O'Reilly, Los Angeles Times, Media (Nonacademic) return to top
I have used PUMS extensively during the 1990s as background research for a number of demographic, socio-economic, and political issue news articles my newspaper has published. For instance, it was critical in a series of articles we published on proposed welfare-to-work reforms, showing how the persons most likely impacted would be the working poor.
The data also has provided extensive help in profiling various sub-groups in the diverse culture of Southern California, fleshing out opinions and observations with real numbers.
The Los Angeles Times Poll also has made extensive use of PUMS data in planning public opinion polling of non-traditional population subgroups. The PUMS data provided geographic information and became the basis for weighting the responses.

Cheryl Elman, The University of Akron, Sociology (Faculty) return to top
My research pertains to aging and living arrangements and aging and retirement. It is important to my research that census data provide information about who people are (whether young, old, urban, rural, married or not), about what they do in the domains of family and work (census questions pertaining to family formation, economic activity), as well as about the institutional settings of their behaviors (indicators pertaining to occupation and industry at the least). Also important are census questions that ascertain level of education. This type of research identifies the types of older persons who remain economically active, not only by gender, ethnicity, and prior educational background, but also by the kinds of skills non-retirees have (whether they are managers or laborers), the types of families that they live in, and the types of industries they are employed in. With regard to aging and the family, census data allow us examine patterns and trends in the living arrangements of the elderly (or younger persons). I am saddened that the amount of information available to researchers is being threatened. The identification of patterns and trends in work and living arrangements (in the short and long-term) can be very helpful to state and national-level policymakers, for the purposes of planning and the targeting of subgroups for services or incentives. Also, the identification of which types of mature workers remain economically active --or which types of families incorporate older kin- should have great significance for potential employers and/or state-level policy-makers because such information may help us predict the shape of the future labor force or need for long term care. Macro-economic forces over the past two decades, such as firm-level downsizing and job skill changes associated with a globalizing economy, have altered the opportunity structure of work, particularly with respect to the quantity and the quality of jobs available to older workers. These forces may somewhat limit individuals' abilities to improve their employment position. These forces, however, are being counterbalanced by change in the supply of workers. These competing trends INCREASE the need to have variables pertaining to occupation, industry, economic activity, education, and age if we are to learn how to best cope with these economic changes.

Confidential Response return to top
Grouping all individuals over age 65 would be disastrous for aging-related research using PUMS data. For aging-related research, most of the interesting questions involve people over age 65 and it is important to be able to differentiate between 90 year-olds and 65 year-olds. With life expectancies in the high 70s-80s now, the proportion of people over age 65 is sufficiently large that identification based on age is extremely unlikely. It is also critical to retain ages for analysis across years -- a lot of research involves considering how 70-year olds have changed over time and grouped age data will eliminate our ability to track such changes.

Tracy Dietz, University of Central Florida, Sociology (Faculty) return to top
I am a social gerontologist who focuses primarily upon socioeconomic and racial/ethnic minority status in my research. Although I am interested in and conduct research on a number of areas within gerontology, recently, for example, I have written manuscripts that deal with mortality among older adults by ethnic group status. It is imperative that I be able to differentiate between people of different ages (ie 65 to 66, even). This is particularly important when we begin to look at age effects by ethnic group membership because such things as the mortality crossover effect is believe to occur at different ages for different ethnic groups (even, for example, between different Hispanic populations). So, given that the census represents our nation's most comprehensive survey of the population, it is very important that as much detail as possible be retained so that reliable, valid estimates and reports can be made. Any limits of detail only provide for less generalizability of any findings that result from research using the data and will certainly limit our ability to continue to make important strides in understanding our nation's population and the issues impacting specific communities and populations. Please, recognize that researchers should be given the opportunity to make important decisions about collapsing categories as it fits with their theoretical and empirical underpinnings because once the data become less detailed, we cannot go back.

Susan Thistle, Northwestern University, Sociology (Faculty) return to top
I look at women's changing relationship to household and paid work, and differences in such changes by race/ethnicity and cohort. The proposed reductions in data would end my ability to carry out such research, and would do so for many other scholars. The recent changes in women's lives are of unprecedented and profound importance. To lose data, especially for the census year 2000, which marks the end of one century and beginning of another, as well as the opening of a new millennium, would be a sad comment upon our society. Understanding of the recent and radical changes taking place in the lives of women and their families is essential if we are to address the new social problems as well as potentials such changes are creating.

Donald J. Bogue, University of Chicago, Demography (Faculty) return to top
The need to introduce age cohorts into demographic analysis requires that age groups be as refined as possible. The occupational categories used in 1990 were about as minimal as acceptable. Income distributions need detail in order to meet the needs of a variety of studies, but also to adjust income distributions between censuses for comparability. Race and ethnicity are powerful variables in differentiating the population (especially for change over the 1990-2000 decade), and the detail should be as great as for 1990, and as comparable as possible.

John Blodgett, University of Missouri (OSEDA), Sociology (Support) return to top
I am a programmer who primarily assists other researchers in accessing the PUMS files. So it is difficult to comment on any one specific area of research - I wind up dabbling in many areas. PUMS is a very important tool in letting us provide users with answers to the questions that are not available on the summary files. Frequently, we are asked to generate data for smaller areas such as counties with populations between 100,000 and 250,000 which we can normally do with the current PUMA geography. Raising the size limit to 250,000 would significantly reduce the utility of the files for many of our users.

Confidential Response return to top
I specialize both in aging research and in family occupational research. Both types of analyses require detailed information that is only available in the PUMS. Grouping data may minimize issues of confidentiality, but it will preclude most analyses that researchers want to do. For example, with aging research, it is necessary to know ages at least in five year intervals. Anything grouped more broadly than this (especially all people over age 65!) would result in the inability to look at changes in living arrangements over time for the elderly.

Confidential Response return to top
Detailed information on ancestry is very important for the research for immigration.

Confidential Response return to top
I do a lot of research on racial composition and racial change over time for cities across the United States. I am interested in looking beyond crude categories of white-non-white and focusing on previously ignored groups such as Asians and Hispanics (within the literatures I am familiar with). I also look at intersections of race and class so SES measures are crucial.

Tim Wadsworth, University of Washington, Sociology (Graduate) return to top
Much of my research looks at how the economic context of areas affects the employment and criminal behavior of individuals. Without having access to information concerning the economic context I could not do this research.

Marie Mora, New Mexico State University, Economics (Faculty) return to top
A lot of my research focuses on the economics of language. I HEAVILY utilize the PUMS, particularly the questions regarding language, English proficiency, education, ethnicity & race, income, age, occupation, geography, immigration status & time since arrival, and gender. My empirical work usually sheds light on policy recommendations.

Confidential Response return to top
For our research we need detailed information American ethnic subgroups at the county level.

Confidential Response return to top
My work requires extensive occupational data, and losing the ability to construct my own class categories would negatively impact my work.

Confidential Response return to top
Losing comparability would make the PUMS data useless for my research area. If the occupational data are not made comparable to the best data now available, any differences in findings concerning the determinants and effects of occupation would be easily attributable to differences in method. The same would be true of race, if not more so.

Rodolfo Gutierrez, University of Minnesota, History (Graduate) return to top
My principal investigation is oriented in the analysis of hispanic population, in a comparative perspective. It is indispensable to count with enough information to compare, problem already present in the oldest censuses. We should expect at least that the newest allows a major comparability among other aspects, particularly when the ethnic composition is now one of the most important issues in the political and economic arenas. The knowledge of the process that derives to the actual situation, is indispensable in a deep sense.

Danielle Longerbone, Indiana University, Geography (Graduate) return to top
I would not be able to conduct studies containing temporal trends on the depression era population because the level of detail necessary in the age category would be eliminated. This, however, is not the most devastating effect the reduction would have on my research. A greater problem for my research would occur from the reduction of the level of geographic detail. I would not be able to conduct any research regarding local (intrastate) migratory patterns.

Stephen Snyder, University of Maryland, College Park, Economics (Graduate) return to top
We know that income and longevity are correlated, but does income really increase longevity or is it just that high income people are more careful with their health? This is an especially important question when Congress is considering changes to social security, and corportations are substantially changing their pension plans. Some very significant changes have been made in benefits that affect very specific population groups. The 1977 Social Security Amendments, for example, give different incomes to individuals with identical work histories, based on their calendar year of birth. I am comparing the mortality rates of those born in late 1977 with those born in early 1978. This would be completely impossible if I could not identify quarter of birth!

Robert Duran, Charlton Research Company, Industry return to top
Grouped demographic data would limit analytical capabilities.

Glenn Steigbigel, University of Chicago, Economics (Graduate) return to top
Specialize in regulation on specific work. Need current employment categories

Hayward Derrick Horton, Dept. of Sociology, SUNY-Albany, Sociology (Faculty) return to top
A reduction in the age, race, and occupation data would be catastrophic to my research. I am currently comparing race and class over time. The availability of a detailed occupation variable is absolutely necessary to have comparability with prior years.

Jeremy Arkes, Rand Corporation, Economics (Nonacademic) return to top
My research focuses on the effects of education on labor market and health outcomes. I need to know when they were born, so I know what years they were in high school. The 1990 Census caused problems over the 1980 Census because it didn't include the quarter of birth. To eliminate the actual age would make this type of analysis impossible.

Susan Wierzbicki, University of Washington, Sociology (Graduate) return to top
I look at immigrants' settlement patterns within metro areas and already must make tradeoffs in the precision of geography vs. ethnicity. Reduction in data on specific ethnic groups, their employment and their residences would compromise understanding on how, for instance, refugee populations are handling the transition to American life as opposed to groups immigrating for primarily economic reasons.

Devah Pager, University of Wisconsin, Sociology (Graduate) return to top
Some of my work has focused on occupational segregation (by race) and within occupational wage inequality. Because there is such a great amount of heterogeneity within broad occupational categories, they are virtually useless for determining the sources of inequality and for disentangling the multiple factors underlying racial inequality in the labor market.

Abbott Ferriss, Emory University, Sociology (Faculty) return to top
In constructing indices of the quality of life, results are needed by state. The indicators show trends in the QOL. Trends by demographic characteristics, such as race, age, sex, etc., are the objective. In addition, some studies focus upon counties within a state. County data, consequently, are required. Data on occupation, income, education, etc., are combined to create an index of the QOL.

Ross Stolzenberg, University of Chicago, Demography (Faculty) return to top
Any analyses of the effects of schooling on occupation and earnings, or the effects of occupation and earnings on health and well-being, would be severely compromised by reductions in the quality of data on age, occupation, and geographic location.

Michael Dennis, UC BERKELEY, Economics (Graduate) return to top
My research details the interactions between socio-economic status and mortality as well as socioeconomic status and fertility. I find the IPUMS to be a good resource for these policy-relevant projects.
I also assign a term project using IPUMS data for undergraduate students. Many of these students are interested in how people of their ethnicity or socio-economic status are doing. It provides a wonderful and personalized way for students to be introduced into social science, statistics and data analysis. Erosion of the ability to identify people like themselves would probably decrease their interest in the project.
Regarding occupation, something which would enable one to distinguish industry AND occupation (so we can tell the autoworkers at Ford from the secretaries for Ford from the secretaries for Microsoft) seems to be the important thing--more important than a large level of detail in occupation per se.

Naomi Hsu, UCLA, Sociology (Undergraduate) return to top
I do research on ethnic identification, that is, on how people choose their ethnicity. If there is a reduction of PUMS race/ancestry detail, I would no longer be able to distinguish between, say, those who identify as being 'Taiwanese' and those who identify as being 'Chinese' because both would be generalized under the group 'Chinese'. Consequently, it would be impossible to identify those factors that are associated with identification one way or the other.

Marco van Leeuwen, NEHA/IISG, Sociology (Faculty) return to top
To study the degree to which the social structure and the social fluidity of the USA is unique or to what degree it conforms to that of other Western regions, and to trace the causes and historical paths, one absolutely needs detailed occupational data that are as detailed as possible.

Confidential Response return to top
A major advantage of having PUMS data is the opportunity to examine detailed age distributions. Grouping the age data (as, for example, into 65+) would render the data worthless.

Confidential Response return to top
We are planing a research project with the title: "Migration to North-America, Internal Migration, and Demographic Structures in the Late Period of the Austrian Empire." The head of the project is Josef Ehmer. So for us the information on geographic details will be very helpful. I only used the ipums-data once for a short overview. So at the moment I can not say more.

Jan Mutchler, University of Massachusetts Boston, Demography (Faculty) return to top
I am a specialist in aging and minorities. Research in these areas has clearly demonstrated the importance of detailed comparisons by age, and the diversity of detailed racial and ethnic groups. Losing these aspects of PUMS detail would be a major step backward in our understanding of differences throughout the later life course. The assumptions that are often made through grouping decisions (e.g., that 65-year-olds are more like 69-year-olds than they are like 64-year-olds) are often refuted by research using detailed age classifications--loss of the detail means we cannot assess these potential differences. As well, loss of the geographic detail would mean that meaningful contextual analyses would be almost impossible, at a time when the importance and techniques for these analyses are finally being established.

Amy Guptill, Cornell University, Sociology (Graduate) return to top
For community development research, detailed local information is very important for understanding which kinds of development strategies succeed or fail in different places, and who benefits and loses from those changes. Without that individual level data on age, race, ancestry, income, and occupation, it is impossible to know how different social groups are impacted by development policies. The PUMS is the single best source of comprehensive high-quality data for this research.

Mark Melder, Louisiana State University, Sociology (Graduate) return to top
As most of my work involves LMA data and the attending industrial categories as well as racial disparity in income, the reduction of any detail for these categories would be very detrimental to my work and the work of my colleagues.

Martha Hill, ISR University of Michigan, Economics (Researcher) return to top
As for possible cutbacks in detail, there are important international projects that would be seriously harmed. The Population Activities Unit of the United Nations Economic Commission for Europe is launching a project to facilitate international comparisons of quality of life for the different generations (children/prime age/elderly but with concern as well about age differentiation within those groups and at the seams where they overlap) and genders. Many European countries tend to collect even more complete environmental/contextual data in their censuses than does the US, and this currently restrains international comparisons involving the US. If further cutbacks are made in the US census data, the possible research will be limited even more. International comparisons offer vast potential for understanding the role of cultural differences in achieving quality life, and having internationally comparable data gives a handle on the state of the world within which the US must operate. As for timing, if adequate detail is not designed into the data, then no matter when it is publicly available it will fall short of what is needed. Better to design in the needed detail even if it delays the release of the data.

William Hart, Saint Louis University, Other Academic (Faculty) return to top
The fastest growing segment of our population is in the aged. Percentage increases increase as you define older and older categories. Lumping together everyone into a category of 65+ would significantly inhibit important research into the problems faced by the older groups. For example, I am researching the effects of retirement on demand elasticities in broad categories of expenditure. Given the wide range of age of retirement, lumping into a category of 65+ would significantly inhibit my own research if not make some comparisons impossible.

Dirk Early, Southwestern University, Economics (Faculty) return to top
Research on housing issues requires detailed information regarding characteristics of tenants. Any loss of detail will reduce my confidence in the results.

Amy Pienta, Wayne State University, Sociology (Faculty) return to top
PUMS data have been used recently to compare active life expectancy across major racial/ethnicity groups in the U.S -- a key question in aging research. Important questions about population health and aging can be addressed with PUMS data only if detailed information is available for all ages after 65. If age data were grouped after age 65, this work would be measurably less precise.

Confidential Response return to top
In defining target populations for Federal, State and Private Grants.

Confidential Response return to top
Changes in income and occupational detail would affect our ability to analyze many trends, especially regional and racial variations in occupational hiring and income.

William Dow, UNC-Chapel Hill, Economics (Faculty) return to top
Example: My research on economic determinants of mortality at different ages would be severely compromised if I did not have detailed population denominators for age groups above 65. Losing detailed geographic identifiers would also make it impossible to study how mortality rates are related to local economic and environmental conditions.

Mary McCall, St. Mary's College of California, Other Academic (Faculty) return to top
Clearly, with grouped data of "over 65" there would be no way to do detailed analyses of young-old, old, and old-old, not to mention the growing group of "oldest-old" over 100 years of age. The same would apply for racial/ethnic groupings, especially with the influx of various Asian older populations.

Kathryn Hornberger, HealthFirst Management Group, Industry (Nonacademic) return to top
I am a healthcare consultant. Because my clients range from those needing help with policy research (towards legislative lobbying) to those needing help with service area development, my use of population data is crucial. The more specific the data, the more accurate my findings, often down to zipcode and census tract. When working with State agencies, for example, I have needed clear delineation between the elderly and frail elderly in regard to assessing program development, i.e. home health, rural clinic development. With the rapidly changing demographics among age 50+, it seems especially crucial in coming years to have the data segmented with detailed sets. This will become even more important as we study alternatives and innovative programs in longterm care, assisted living as well as subacute care. The delivery of healthcare is becoming more complicated, not less so - and the data needs to become more specific, not more generalized. It is rather incredible with our level of technical sophistication to consider that data categories may be consolidated rather than expanded.

Charles Rambeck, St. John's University, Economics (Faculty) return to top
I do research on migration. Geographical detail is obviously important. Age and, to a lesser extent, occupation are important variables in my analysis. Grouped age data would force me to use statistical techniques (dummy variables on age categories) that give less satisfactory results.

Wilbert Van der Klaauw, UNC-Chapel Hill, Economics (Faculty) return to top
Many government programs (social security, Medicare, SSDI) have eligibility criteria that depend on a person's actual age (ie 62,65 etc). By grouping ages in broader intervals very important information is lost that is essential for evaluating the impact of such programs on behavior (retirement etc).

R.S. Oropesa, Penn State University, Sociology (Faculty) return to top
Although it is hard to comment without information on specific proposals, efforts to collapse existing classifications in several areas would seriously compromise future efforts to document patterns of ethnic/racial incorporation and the processes that underlie the incorporation of specific ethnic/racial groups into American society. The geographic detail is also important because it is a common practice link responses of individuals in surveys to information from the PUMS about the areas in which they live to understand how socioeconomic contexts influence behaviors and attitudes. The loss of geographic detail would seriously hamper the ability to understand how individuals are affected by the contexts within which they live.

Janet Wilmoth, Purdue University, Sociology (Faculty) return to top
I have used the PUMS data to study older immigrants. A strength of the 1990 data is that the race, ethnicity, ancestry, and place of birth items provide sufficient detail to construct a variable that allows within group comparisons. Reducing the detail of the race and ancestry items would severely limit the usefulness of the 2000 PUMS. This would be unfortunate since the PUMS is one of the few data sources that contains enough cases to make extensive within group comparisons.
Another strength of the PUMS data is the detailed information that is provided for age. I have used this item extensively in my research. For example, in my research on older immigrants I have used the age item to construct a measure of age at immigration. It would be impossible to contruct this variable if the age item were grouped. Measuring age in years is critical to researchers like myself who are interested in the older population. Grouping the age categories would result a loss of information (and variance) that is essential to our research. It would limit our understanding of the rapidly growing older population.

Alicia Adsera, University of Illinois at Chicago, Economics (Faculty) return to top
Any reduction in geographic detail would affect my research in housing markets and spatial economics. Any reduction in occupation detail would affect my ability to understand differences in productivity of similar workers across geographic areas.

Carl Backman, Buffalo State College, Sociology (Faculty) return to top
I have been involved in evaluating jury selection procedures. It is essential to have as much geographic detail as possible. It is important to have age detail that allows identifying those old enough to vote (18+), those old enough to drive (16+), and the old old (say 80+) who might reasonably be excluded from jury duty. Migration is a critical issue for evaluating lists maintained by the courts, so migration detail (already weak, given a 5 year delay) needs to be maintained (particularly same house, same county). Poverty status is also valuable.

Shiro Horiuchi, Rockefeller University, Demography (Faculty) return to top
In order to understanding demographic, social, health, and other aspects of aging processes and in order to serve demands of the elderly population better, we need to investigate AGE-RELATED CHANGES of various characteristics in the elderly population. People 65 years old and those 95 years old have completely different profiles, characteristics, problems, and needs. To understand age-related changes of behaviors and characteristics, the AGE IN SINGLE YEAR [65, 66, 67 etc.] is essential.

John McHenry, Demographic Data For Decision-Making, Inc., Demography (Nonacademic) return to top
My firm's local government and business clients require the best possible local area (i.e. sub-county) demographic data in order to make the best possible local area business and policy decisions. My firm is embarking on an SBIR-sponsored line of research to produce local area demographic estimates and, later, local area demographic projections. Not surprisingly, our preliminary research indicates that lower-level geo-demographic inputs produce higher quality demographic outputs. To restrict PUMS and other census outputs to higher levels of geography and/or higher levels of demographic aggregation will cripple my research and like research efforts. Regional health-care providers, banks, local businesses, school districts, utilities, local governments and all their direct and indirect clients will suffer if PUMS databases are no longer readily available.

Shelley Lapkoff, Lapkoff & Gobalet Demographic Research, Inc., Demography (Nonacademic) return to top
I am a consultant to many school districts. My company provides school enrollment forecasts for effective planning. Perhaps surprising, understanding the senior population is very important to enrollment forecasts in mature communities, because housing turnover is an important variable predicting migration and increases in the school-aged population. Also, which housing turns over is important - seniors who move out of their homes create the biggest impact on school enrollments. We use PUMS to study relationships between age and housing turnover generally, which we can then apply to specific school districts. Please do not eliminate ANY age detail from the PUMS data - this is critical (65 year olds act a lot differently than 85 year olds!) Also, we hope no change will occur on providing characteristics of the housing.

Joseph Persky, University of Illinois at Chicago, Economics (Faculty) return to top
I first used PUMS data in 1967 as a graduate student and have been using them ever since. Most critical to my current work are the puma and powpuma income data within metropolitan areas. These data allow us to research the distributional implications of different economic development scenarios. The same data are key to analyses of suburban sprawl.
As to historical comparability, I recently supervised a thesis on interregional income convergence since 1950. The IPUMS allowed the PH.D. candidate to carry out detailed comparisons of incomes adjusted for education and occupation over time. I can't imagine another source of data that could have played this role.

Jeanne L. Anderson, Jeanne Anderson Research, Industry (Nonacademic) return to top
Difficult to specify. See suggestion below for issuing several versions of the PUMS file.

William Yancey, Temple University, Sociology (Faculty) return to top
It appears that you may be planning on combining occupational and industrial categories. My interest in the consequences of industrial transformation could not be conducted if this were done.
Given variation in industrial and occupational profiles between urban areas it is also critical that detailed geography is maintained.

Rachel Boaz, Graduate School, City University of New York, Economics (Faculty) return to top
Aging research: Do NOT compress age-classes; the aged population is a very heterogeneous with substantial differences by health status, income, marital status, living arrangements, and race. Income: Narrow income classes are needed for meaningful analysis. The open-ended class should be at the 99th percentile.

Debra Erdley, Tribune-Review, Media (Nonacademic) return to top
As a newsreporter covering issues in one of the nation's oldest communities, any reduction in specificity with regard to age would hurt me tremendously in trying to make comparisons over the years. When you live in an aging community that has faced economic hardships, the boosters always try to paint a picture that is slightly rosier than reality. Good solid data is a tremendous reality check. It also helps community leaders make informed decisions.

Melissa Binder Binder, University of New Mexico, Economics (Faculty) return to top
My recent research involves the effect of local labor market conditions on social and economic outcomes such as fertility, school enrollment and poverty. The identification of localities over time, as well as detailed individual information regarding race, age, income and occupation is vital to understanding the effect of labor markets on particular population groups. I believe that this type of research is crucial in making informed policy decisions.

Jeff Burr, UMASS Boston, Demography (Faculty) return to top
The kinds of grouping being contemplated would make historical (recent) research nearly impossible. My aging research focusing on race groups would be finished as far as national data are concerned. AS you know, the census PUMS offer the only source of individual data for research on certain kinds of race/ethnic groups. It is yet another great irony that the federal government and politicians are trying to fix something that is not broken. Where are the examples of any person being specifically identified with these data?

Steve Pischke, MIT, Economics (Faculty) return to top
Since one of the main advantages of the PUMS is it's large sample size, it allows analysis of very detailed subgroups, which would be compromised by aggregation. A specific example in economics is a very influential paper on the returns to schooling by Angrist and Krueger using the QUARTER OF BIRTH variable available until 1980. Even if the proposed aggregation is not particularly harmful to a specific project now, it might always compromise the ability to implement good research ideas in the future.

Mary Grace Kovar, NORC, Demography (Researcher) return to top
People age 65 and older exhibit more diversity than any other age, and are the subject of more national policy (except in times of war) than any other age. While the population pyramid narrows considerably, it is a 40-year age span and is much like having a 25-64 year age group, which no one considers.
Death rates are much higher at older ages. Higher for men than women - same as earlier ages Higher for white men than black men - reversal of earlier ages Hospitalization rates are much higher. Medical visit rates are not. Nursing home residence rates are much higher. There is a radical change in living arrangements and in marital status - women are widowed and live alone.

Confidential Response return to top
Age, income, and race/ethnicity are crucial factors in the work that I do on low-income communities and populations. For example, how many 3 and 4 year-olds living in poverty (or near) poverty are enrolled in preschool programs. An aggregated less than 5 category would not provide this type of detail. Similarly, the prevalence of disabilities and the need for supportive services vary greatly by age for older persons. An aggregated 65+ category would not allow this crucial analysis. Income levels and race/ethnic factors are similarly needed to target programs and services to persons who are most in need.

Confidential Response return to top
PUMS data have been used by my doctoral students in their dissertation research, and have been a major source of insights into demographic characteristics of people in the PUMS regions that have informed the development of primary survey instruments. Any reduction in detail would have major negative impacts on basic research of this type.

Myron Floyd, Texas A&M University, Other Academic (Faculty) return to top
My specialty area is race and ethnic studies. My research examines how race/ethnic impacts natural resource policy and management. Reduction of PUMS limits development of my GIS databases.

Peter Philips, University of Utah, Economics return to top
I research the construction industry. Occupational and industry detail are essential to understand the dynamics of this industry. The income, employment and mobility patterns vary considerably comparing electricians to carpenters to roofers to laborers. These are very distinct labor and commercial markets. They cannot be condensed meaningfully.
Race and age are key variables in predicting outcomes. Construction workers in outside work, operating engineers, laborers wear out much faster than inside workers, plumbers electricians (as opposed to linemen). Restricting age information would hide these important differences.
There are key policy questions that will not be well informed if these information are surpressed. examples:
Should retirement age for social security vary by occupation? Should prevailing wage regulations for construction continue? (The Davis Bacon Act and 31 state regulation plus many municipal regulations of construction. The census plays a key role in the debate regarding these laws.
I also use the census pums as the key data set for forensic research on the effects in cases involving wrongful termination, wrongful death and personal injury cases. No other data set provides the occupational detail at a state level that allows for predicting the effects of these events on income. The justice system relies on these data for fair determination of damages in thousands of cases each year.

Jeff Tayman, San Diego Association of Governments, Government (Nonacademic) return to top
Our agency uses PUMs for a number of activities, including the preparation of detailed demographic and economic estimates that support multi-billion dollar capital facility projects, the development of local housing elements, plans and grant applications, and for infrastructure and health care planning and analysis.

Ronald Lee, University of California at Berkeley, Demography (Faculty) return to top
I work on aging, and the loss of age detail above age 65 would be devastating; it would mean that PUMS would no longer be of use for research in this area. Loss of information on source of income, eg public assistance or Social Security, labor earnings, etc would also be very damaging.

Charles Rynerson, San Diego Association of Governments, Government (Nonacademic) return to top
We do estimates and forecasts of population (by age, sex, ethnicity, and military/civilian status), labor force participation (by age, sex, ethnicity, and military/civilian status), employment by occupation and industry, and household income. We use PUMS to develop basic assumptions and inputs for all of our models. The race and single year age detail is particularly important, because we need to crosstabulate race and age by military status, labor force, and migration history (both the 5 year mover and the born abroad questions). Migration history is important to us at all age levels. Maintaining the occupational categories is important, because it would allow flexibility to group them to be compatible with occupation groups from BLS' Occupational Employment Survey. The geographic detail is less important, because most of our use for PUMS is at the county (almost 3 million people) level, and we are crosstabulating too many variables to get significant results from a sample of an area with 100,000 people.

Virginia Appl, Houston-Galveston Area Council, Government (Nonacademic) return to top
I use PUMS data to determine "synthesized households" for use in Urban Land Use Modeling ... less detailed means broader assumptions regarding population characteristics which is not good when you are trying to produce estimates at the block level!

Mark Hill, Pennsylvania State University, Demography (Faculty) return to top
Grouped age data would severely limit the usefulness of the PUMS data. I have developed a survivorship method ideally suited for the IPUMS data that requires the same cohort members be identified in various PUMS files. With out age in single years, this method could not be used appropriately. Furthermore, with the aging of the population the key demographic issue in the US, grouping data at the oldest ages poses a key threat to research on aging.

Haydar Kurban, University of Illinois at Chicago, Economics (Researcher) return to top
It will be almost impossible to have a representative sample in different income, age, race and geographic classification.

Jenifer Bratter, University of Texas at Austin, Sociology (Graduate) return to top
I research family dynamics, particularly marriage and cohabitation. Currently I am examining geographic variability of interracial marriage for 1990. Detailed race and ancestry data is crucial to identify the ethnicity of whose marrying whom, when the broad categories are too general. I also greatly appreciate the geographic detail the census provides. Currently, I can match the detail applied in other studies which increases the comparability of my results with other work.

Bill O'Hare, Casey Foundation, Demography (Nonacademic) return to top
Loss of geographic detail would be very difficult for me. The only reason the PUMS are more useful than CPS is because PUMS provides data for smaller geographic areas. If that strength of the PUMS is lost, it becomes relatively useless.

Jennifer Garner, BBC Research & Consulting, Industry return to top
PUMS data is a critical input to consolidated housing plans and other assessments of regional demographics, particularly on the PUMA scale. We often analyze PUMAs to segment demographic markets within a small geographic region in order to assist public sector clients in prioritizing spending.

Jeanne Gobalet, Lapkoff & Gobalet Demographic Research, Inc., Demography (Nonacademic) return to top
We examine small geographical areas and reduction of geographic detail would greatly reduce our ability to provide the answers our clients need. The more detailed the information, the better, as far as meeting needs of our clients, who include public schools, public health agencies, and those assessing child care availability. We have also used PUMS data to assess whether housing discrimination exists.

Yoshinori Kamo, Louisiana State University, Sociology (Faculty) return to top
Chinese/Japanese elderly's living arrangement (JMF 1996): Age cut off at 65 made it virtually impossible to do research like that. Married women's labor force participation affected by the husband's income, own education, etc. 1950 data had to be dropped since education was not available. 1960 data categorized income (rather than dollar value), which makes the husband's income measure quite primitive. I don't want to see it again for 2000 data, since income is critical in sociology.

Mark Mather, Population Reference Bureau, Demography (Nonacademic) return to top
I have often used the 1990 PUMS to create customized tabulations for states, that can then be combined with national-level data from the Current Population Survey to arrive at post-census state estimates for various sub-groups of the population. Combining age groups or race/ethnicity data would be harmful if these variables could no longer be used in combination with other federal data sources, like the CPS.
For me, the main advantage of PUMS is the availability of data for customized sub-groups of the population -- for example, data on persons in a specific age group that aren't published in summary tables. If these data are grouped, then the PUMS data become less useful.

Tim Smeeding, Maxwell School, Syracuse University, Economics (Faculty) return to top
PUMS is a once a decade occurrence; there is considerable interest in detailed age and location material on income, family status, age etc. During the 1990's the interest was at the level of the school district; during the next decade the interest will be in aging populations. Please do not skimp on the PUMS

Julie Hudson, The Urban Institute, Economics return to top
I study behavioral responses to welfare and other policies at the national, state and local levels. It is very important for me to have detailed information on a person's geographical location of residence. In some cases, state level is enough, but in others it is not. I also must break down my groups of interest by age, race, sex and income levels. Income is also an important outcome variable in some of my studies.

James Holland Jones, Harvard University, Demography (Post-doc) return to top
What can I say? Making projections about mortality and fertility changes with respect to income and geographic mobility simply can not be done without data on age structure, income (and, to a lesser extent, occupation) and geographical location.

Sharon Ju, Houston-Galveston Area Council, Economics (Post-doc) return to top
Employment details

Harry Holzer, Urban Institute/Georgetown University, Economics (Faculty) return to top
I and my coauthors use race/ethnicity data and geographic information, especially at the census tract level from the STF3 files.

Roy Treadway, Illinois State University, Sociology (Faculty) return to top
For my research detailed geographical units are more important than detailed economic and social characteristics since we more often are involved with general profiles for smaller (100,000 population or so) areas.

Charles Purvis, Metropolitan Transportation Commission, Government (Nonacademic) return to top
1. PUMS is needed in preparing county-level data on socio-economic characteristics by market segment, for use in travel demand forecasting models in use by regional and county-level government agencies. For example, PUMS is used to calculate mean household income, mean household size, and mean workers per households for households stratified by household income quartile by auto ownership level (0,1,2+ vehicles per household). Only PUMS provides this critical information.
2. PUMS is used in analyzing characteristics of commuters by means of transportation, including age, sex, occupation, earnings, race/ethnicity, ancestry, year of immigration to US, and industry.

John Straub, University of Wisconsin-Madison, Economics (Graduate) return to top
The lack of geographic detail in the 1990 PUMS data poses many challenges to researchers wishing to use the data. Mapping between PUMAs and other geographic entities can be quite complicated. This is a reasonable price to pay to ensure confidentiality, however. I'm not sure how hard it is to ensure confidentiality with PUMA-level geocoding, but I do know that more coarsely geocoded data would be useless for many purposes.

Julie Lee, National Bureau of Economic Research, Economics (Researcher) return to top
I do research in health economics. Very often health-related data sources lack social economic variables, such as income, education, occupation, etc. which influence choices people make regarding health. PUMS is a good source of such variables. Age, race and geographical details are especially important because they allow me to merge social economic information with individual level health-related data; finer the age-race-sex-geographical cells are, more accurate the merging would be.

Jim Hinterlong, Washington University, Other Academic (Graduate) return to top
I use 5 year age-cohorts to examine patterns of engagement in paid employment, while limiting my sample to those aged 60 and over. Aggregation beyond the current level (which is "lumpy" enough) would preclude me from using this data for future research.

Confidential Response return to top
We use seed matrices derived from PUMS to establish joint distributions of householder age, household size and income, workers, and vehicles. We would like to maintain as much specificity as possible within these categories.

Simon Choi, Southern California Association of Governments, Government (Nonacademic) return to top
My research generally might not need a detailed categorization of some items see below). However, the more information we have, the better we can analyze. This will help us to devise a better planning or public policy.

1. Cohort component model run: I use age and race/ethnic specific birth rate and migration rate from PUMS to be applied to cohort component model. Generally 18 age groups (0-4, 5-9,...) and 4 different race/ethnic groups (Non-hispanic white, non-hispanic black, non-hispanic asian & others, hispanic)
2. Household projection model run: I use age and race/ethnic specific headship rate from PUMS to do household projection.
3. Travel behavior of commuters by race/ethnic group and income level.
4. Household overcrowding by age group, race/ethnic group, income, and occupation.

Lisa Catanzarite, U.C. San Diego, Sociology (Faculty) return to top
My research field is stratification, with a primary focus on labor market inequality by race/ethnicity, immigration status and gender. Occupational segregation is central to my research. Any reduction of detailed occupational categories would be disastrous to my work, as well as that of others in my field. Information on ancestry, race, and place of birth is critical to research pertaining to immigration, race, and ethnicity. Both my work and that of others would be adversely affected by loss of detail on these variables. Much of the research on immigration has been conducted at the level of metropolitan areas or PUMAs. Such detail is crucial for work in this field. I also conduct research on other aspects of stratification, including poverty. Both earnings and income data are essential to this work.

Francine Atwell, Hawaii SBDC Network Business Research Library (under the Research Corporation of the University of Hawawii), Economics (Researcher) return to top
I work with small business people. They often want to know the number of likely customers for their products/services in their neighborhood, and want to get what often is a cross-tab of 2 or more characteristics. I can't get that from the usual STFs.

Deb Greene, City of Berkeley Health Dept., Other Academic (Researcher) return to top
My major focus is racial disparities in health outcomes. I am currently working on broader historical models that take into account income, occupation and am analyzing trends over time. Geographically, I am making city level (Pop >100,000) models. Any reduction in geographic detail or racial/ethnic detail would prevent much of my research

Susana Adamo, University of Texas at Austin - Population Research Center, Demography (Graduate) return to top
My current works deals mostly with non-metropolitan populations. The reduction of geographical detail (I mean, less detail that in 1990 census) may, eventually, reduce the possibility of meaningful analysis other than at national level. In addition, grouped age-data (for example, in 5-years groups) limits the analysis of life transitions where a difference of one or two years may be potentially important, like age at marriage, or entering and exiting from the labor force.

Karen Lamphere, San Diego Association of Governments, Government (Nonacademic) return to top
One of the most valuable features of PUMS, obviously, is that users can determine their own groupings of categories. Although the files are large and cumbersome, I would encourage the Bureau to limit the grouping of variables as much as possible.

Shawn McBrien, University of Michigan, School of Public Health, Other Academic (Graduate) return to top
Reduction in detail of data would significantly hamper accurate research in the area of social epidemiology. Accurate research requires the greatest accuracy at the smallest scale possible.

Confidential Response return to top
The new format of the census race question, which will allow multiple responses, will involve the potential reclassification of millions of Americans. In order to separate the artifactual changes introduced by question format from real changes in the socio-economic status of various minority groups, it will be extremely important to have as much detail as possible on which people are marking more than one race.

Edward Miguel, Harvard University, Economics (Graduate) return to top
I am interested in intergenerational mobility, especially among ethnic and national minority groups. Reduction of PUMS in future censuses would make this research extremely difficult, if there were less detailed data on geographic mobility, income and occupational outcomes, and national origin.

Tara Gibbs, University of Minnesota, Other Academic (Graduate) return to top
My field is applied linguistics
I would like to see MORE detail, not less. 2000 job titles would be better.

I need summaries of data available by neighborhood and small town and rural area, but this isn't currently available. I'd like to be able to search by primary language and get maps of any areas with 5 speakers or 1000 speakers, etc.
I would like a data set that is broad released soon and a more detailed set released later.
Historical compatibility would be essential, if historically they had actually asked the questions that would have provided the information I need to know. It would be nice if certain questions were asked and then kept the same. Specifically, I need to know first and second language information and whether someone is a first or second generation immigrant or later. I need to know if the person speaks no English, very little English, some English, conversationally fluent or is perfectly fluent in English. And I want to know this about the other languages that this person speaks. This is only surmisable from the national origin data using statistical procedures over time, but it is grossly inaccurate to use this, however, there is nothing else available, so this is what we do in this field.
Thus our estimates on what level of interpreter, L.E.P. and other services needed, in what languages, and where, are grossly inaccurate, as is our ability to measure levels of language maintenance, language shift, and bilingualism which would highlight the effects and conditions of various policies, as well as whether a given language group continues to need services.

Alexander Bucur, USC/UCLA, Sociology (Researcher) return to top
Changes in the American society cannot be measured or extrapolated from other surveys unless historical compatibility is achieved. Since we have such good data for 1980 and 1990, and we have experienced such change in our society it seems that we must maintain compatibility to be able to analyze changes. Even more, all other surveys that compare or balance themselves against the US Census will be undermined since they will not be able to properly analyze changes at the societal level.

Cecilia Conrad, Pomona College, Economics (Faculty) return to top
I am investigating historical trends in the economic status of teen mothers by race. It is critical to distinguish between young teens and older teens. Each year of age can matter a great deal.

Kenneth Sylvester, University of Victoria, History (Post-doc) return to top
My research takes me into the exploration of degrees of self-employment in the farm economy, and the shape of regional non-farm economies.
I need to have fairly detailed breakdowns to explore inter-regional and international comparisons.

Confidential Response return to top
I'm most concerned with reduction in geographic level. If it was only for areas 250,000 or greater, then we would be hard pressed to look at cross tabulations for our region.

Marie Cornwall, Brigham Young University, Sociology (Faculty) return to top
The pums data have been most useful in my research on changing gender roles using the 1940 to 1990 data sets. Most data available used 1960 and beyond, but the pums data allowed us to examine the 40s and 50s and see a more consistent trend line.

Karen Norberg, National Bureau of Economic Research, Other Academic (Faculty) return to top
I specialize in epidemiological & economic research concerning children and youth, especially pertaining to high risk behavior, household resource allocation, maternal employment, and factors affecting fertility and teen births. I make use of the finest level of detail about age groups & geographic variation possible, & these are the highest priority areas, for me, in preserving geographic detail. I compare local area estimates from the PUMS with US vital statistics data, & much of the value of the PUMS would be eroded, for my work, if children's ages or geography were further collapsed.

Confidential Response return to top
I study the changing nature of childhood in the United States. A reduction in detail would not allow me to bring my studies into the present, and would reduce historical comparability.

Dennis Yee, Metro (Portland, OR), Government (Nonacademic) return to top
I and others use the PUMS data as the initial seed matrix for creating current and future year estimates of populations for transportation and land use modeling and forecasting. Household, income and age data are extremely important for us as these variables represent key components into our simulation models. In addition, geographic detail is very important to us. Reduction in population size of PUMA's could damage our modeling and forecasting precision.

Confidential Response return to top
I am a labor economist, so I am very interested in income and occupation data. However, given differences in income, etc. between metropolitan areas, even small ones, I think having PUMA information is important. Personally, I'd like to have county data (or county groups for very rural counties).

Richard West, Social Policy Research Associates, Economics (Nonacademic) return to top
We do a lot of analysis for the Department of Labor in which we calculate demographic and economic characteristics of local workforce investment areas. Geographical detail at the county level is highly desirable and indeed necessary for this analysis.

William Sundstrom, Santa Clara University, Economics (Faculty) return to top
I have recently been examining historical changes in occupations by race. One dimension of this is to examine how these changes varied according the characteristics of local markets. Thus I combine detailed information on occupation and location of residence. A reduction in detail would reduce the available variation for analysis considerably. For example, knowing the city is much more informative than merely knowing the state.

Jon Stiles, UC Berkeley, Sociology (Researcher) return to top
One of the many strengths offered by the time scale and sample sizes available in the IPUMS is the ability to construct birth cohorts which can be sliced several ways at the analyst's discretion over time. This virtually requires single years of age to accommodate different analyst's intentions.

Laura Leete, Willamette University, Economics (Researcher) return to top
I work on the issue of how capable local labor markets are of absorbing welfare recipients, among other things. Thus, having maximum detail on occupational classifications by sub-metro area is crucial in being able to characterize the nature of local labor markets. More generally, reducing the precision of any of the variables mentioned here reduces the precision with which we are able to characterize and understand poor, welfare-dependent and at-risk populations.

Rose Nguyen, University of Minnesota - Census Projects, Economics (Support) return to top
If I specialize in occupation research, then the market analysis for the unemployment rate and employment rate is not statistically clear for any occupations.

Confidential Response return to top
PUMS has been one of our sole means of gaining greater understanding of the racial, ethnic, and linguistic diversity in our area. Some of the collapsing of categories already has presented problems in using these data for health assessment. For example, tribal affiliation for American Indians/Alaska Natives was collapsed to the largest tribes nationally. This made information on local tribe affiliation non-existent for our health assessment purposes. We would like greater detail in areas such as this.

Confidential Response return to top
I explore theoretical issues of representing data uncertainty and data fusion. Reducing detail and/or increasing the size of PUMS regions will significantly reduce the ability for researchers to model urban areas.

Jong Won Min, UCLA, Other Academic (Graduate) return to top
With a rapid increase in old-old population (85+), updated information on detailed age group is crucial to understanding their health care and long-term care needs.

Robert O'Connor, Penn State University, Government (Faculty) return to top
My research has used PUMS data to examine the ability of government programs to involve different racial, ethnic, and income groupings. Grouping data beyond that done for the 1990s PUMS would weaken this research.

Maggi Rose, Open University, UK, Economics (Undergraduate) return to top
I examined the effect of home computer ownership on occupation and income. I required detailed economic, geographic and occupational data to complete this. To remove other influences I also required sex, ancestry and race so that I was sure I was comparing like individuals whose only difference was their computer ownership. To have been unable to identify the similarities and differences between individuals would have rendered the study useless.

Confidential Response return to top
My area of research is primarily a comparison of aging and working-aged populations over a number of topics. It is crucial that lumping of persons aged 65 and over does not occur.

Confidential Response return to top
The population estimates and projections produced by the PA State Data Center rely on the single year age data. The model ages the population factors in birth and death data from our Department of Health and applies migration rates. A different model would need to be used if only grouped age data was produced. In addition, the PA State Data Center has studied the income of elderly citizens (detailed income categories for those age 60 and older, both unmarried persons and married couples with age detail for both partners) for the PA Department of Aging. The reduction in the geographic, income or age detail would effect this study as well.

Dawn Hubbs, University of North Carolina at Charlotte, Other Academic (Researcher) return to top
I work in a library and therefore help other researchers. Many faculty at our institution do research on family relationships, and need specific age and race data. We also have several faculty doing research on the aging population, and need to compare 65 yr olds to 70 yr olds to 75 yr olds, etc.

Francine Stephenson, SDC, NC Office of State Planning, Government (Nonacademic) return to top
Government policy makers need information on VERY SPECIFIC age groups in local areas cross-tabulated by a variety of characteristics. Not having the age detail means that policy decisions would be based on guess.

David Wong, George Mason University, Geography (Faculty) return to top
PUMAs are already quite coarse in geographical scale. Many phenomena have significantly variation with a county and PUMAs are county or groups of counties, which fail to reflect the spatial differentiation within county. Grouping PUMAs together will be much worse. The results will be of no significant geographical implication and useful for policy formulation related to differences within regions.

Michael Stover, Hillsborough County Planning Commission, Government (Nonacademic) return to top
No comment.

Confidential Response return to top
The PUMS allows our office to virtually do any analysis that is not available from the standard products. This file has been essential for researching the impact of policy and proposed legislative bills. It's use has escalated in the last couple years (NOT reduced) -- even as the data has aged.

Michael Spar, Weldon Cooper Center, University of Virginia, Demography (Researcher) return to top
I doubt that changes in the PUM will affect me directly. However, as a member of the state data center network for Virginia, we receive a high volume of information requests. Some of these can only be answered by reference to PUM data. If those data are compromised in any way - either by changing the minimum population size for designating a PUM area, or by reducing the level of data detail, then fewer data users will find the answers to their questions in the PUMs data. For some data users, the ability to "create" their own tables is the one and only saving grace of PUMS. Reductions in the number or quality of those home-grown tables will simply reduce the usefulness of the Census to some data users.

Leonard Gaines, Empire State Development (NYS SDC LEAD), Government (Nonacademic) return to top
As a data provider/intermediary we frequently get requests for specialized tabulations (specific age ranges, etc.) for program evaluation purposes. Using different age groupings would bring the entire analysis into question. For example, a user might be looking at a potential program aimed at 16 and 17 year olds not in school. Using any other age group would not be valid. The next user might NEED 17 to 20 year olds. It would be impossible to meet the needs of both users with a single grouped age variable.

William Even, Miami University, Economics (Faculty) return to top
It would be impossible to examine how aging affects the level of income among the aged, as well as how widowhood at different ages affects income.

Duane Marble, The Ohio State University, Geography (Faculty) return to top
The current level of geographic aggregation is already far too high to permit valid inferences to be made about many aspects of behavior. E.g., changes in the journey to work.

Lisa Nelson, Case Western Reserve University, Sociology (Support) return to top
In general, I use the PUMS data to assist community organizations and some governmental agencies with their work -- whether it be proposal writing, program planning or needs assessment. The detail of PUMS is necessary to meet some of their specific questions and needs. Communities need this type of information in order to understand and work toward improving the social and economic conditions in their areas. While the STF3 and STF1 may meet many of the information needs -- it does not meet all of them.

Peter Granda, University of Michigan - ICPSR, History (Support) return to top
While I am not an active researcher and user of PUMS data, as an archivist I interact with numerous researchers who have used the 1980 and 1990 PUMS files. My strong impression is that these individuals want as much if not more detail in the 2000 PUMS files as appeared in either the 1980 or 1990 files. ICPSR expects a very heavy demand for the 200o PUMS files.

Richard Rathge, North Dakota State University, Demography (Faculty) return to top
Often we are asked by congressional staff to assist in understanding the consequences of changes in legislation. The last several times dealt with legislation regarding income. We had to use detailed income breakdowns to offer insight into the consequences of different thresholds by detailed age. Thus age and income are necessary in detailed form.

Susan Alibozek, Montgomery County Dept. of Planning & Development, Government return to top
Our county has a high elderly population (24% over the age of 60 in the 1990 census), and as a result many many programs affecting the elderly rely on age population. Additionally, we have a high hispanic population which also relies on data for that specific category therefore reduction of information on the age and race category would harm the data requested by many organizations who contact our office for information.

Ronald Cossman, University of Central Arkansas, Geography (Graduate) return to top
Both my dissertation and professional research agenda are focused on the relationship between economic structure and wage inequality in the United States over time. This research is critical to both understanding the processes that cause changes in inequality, as well as guiding economic development professionals (from EDA to the city level), in policies and programs that will mitigate, not cause, increased inequality. The PUMS is the only currently available data set to examine the migration of workers between occupational and industrial sectors of our economy.

Lisa Dillon, Institute of Canadian Studies, University of Ottawa, History (Faculty) return to top
My research on the aged has demonstrated strong distinctions among older persons on the basis of age; the living arrangements, income and meaning of old age itself varies as older persons grow even older. Grouped age data would prevent me from tracing fine distinctions among the elderly on the basis of age. This will be particularly true in the case of the 2000 Census, which will record the largest ever elderly population--and largest ever population of old old persons--in the history of the United States. I would prefer that the age variable be available in single years so that I can fashion my own age groups as suits my research needs--including researching the declarations and experiences of centenarians. The other changes being considered for the 2000 Census will also have a grave impact on research. For example, a fine level of geographic detail is necessary for health research. With improvements in Geographic Information Systems technology, scholars can now research correlations between disease and environment by mapping at the level of small communities. By erasing fine-grained geographic distinctions, the 2000 Census will considerably hinder our efforts to understand links between disease and place.

Gastón Repetto, UIC, UNLU, UCC, Economics (Researcher) return to top
I have used IPUMS in my research regarding immigrant assimilation, ethnicity, how different social groups invest in different forms of human capital, and the rewards they get in the labor market and in home production across time. Therefore, all the information concerning: 1) labor market individual's characteristics and outcomes (income, occupational choice, tenure, OJT, etc.), 2) sociodemographic data (age, ancestry, household composition, race, language spoken at home, etc.) and 3) data specific to immigrants are a must for my line of research.

Steven Ehrlich, US Dept of Housing & Urban Development, Economics (Nonacademic) return to top
At the very least, central city status should be identified for large and medium sized metro areas.

Confidential Response return to top
I study age adjusted rates of disease using race, geography and place of birth as markers for increased or reduced likelihood of infection. Since groups from certain countries with certain background are much more likely to contract certain infections it is essential to have this information for health program planning.

Diane Hite, Mississippi State University, Economics (Faculty) return to top
I work in urban, regional and environmental economics, and analyses on race and the environment, for example, would be very adversely affected if the geography level were to be broadened.

Michael McDonald, SUNT--Binghamton, Other Academic (Faculty) return to top
My area is race and politics. Without detailed information on race, by small geographical units, much of what I have been doing for 15-20 years can't be done for 2000. Because my research involves voting--turnout and candidate choice--not having detail information on age (turnout), income and occupation (vote choice controls when looking at race) most inferences would have large clouds of doubt hanging over them.

Craig Olson, University of Wisconsin-Madison, Economics (Faculty) return to top
I'm doing research now which focuses on the supply and demand for technology workers and the impact of higher education on the creation of high skilled jobs. Detailed data over time by geographic area and occupation are essential for estimating the relationship between education and high skilled jobs.

Beverly Allen, Iowa State University, Sociology (Faculty) return to top
I am using the occ score variable over the 1850-1990 period. This is crucial to my current research. Any changes, other than improving comparability over time will hinder planned research for the next few years.

Williams Olatubi, Louisiana State University, Economics (Researcher) return to top
In my work, being able to map earnings by industry to occupational categories is very important. The more the detail, the more plausible the results of my analysis and the better the inferences.

Confidential Response return to top
In the library, we provide very specific census data to researchers, whether they are students, faculty, staff, or members of the public. Topics run the gamut--physical and social sciences, humanities, health, fine arts... No longer providing this service would be a great loss.

William O'Dell, University of Florida, Other Academic (Faculty) return to top
A reduction in detail and a more highly aggregated geography will substantially impair our research which is in housing policy. Our Center works very closely with state and local governments and non-profit organizations. PUMS data has been crucial to these organizations in policy-making, resource allocation decisions and strategic planning. Recently, for example, the state's housing finance agency was contemplating an expansion of elderly housing subsidies (affecting the allocation of millions of dollars annually) and needed information on relative cost burdens across several age sub-categories within the group - PUMS data was absolutely essential to the development of that information. For a number of organizations that we consult with a loss of geographic detail in PUMS would be a significant impediment to informed resource allocation decisions.

Roy Williams, Massachusetts Institute for Social & Economic Research/Umass/Amherst (State Data Center), Government (Researcher) return to top
As a State Data Center we get many requests for tabulations of data for very small geographic areas. PUMAs of 100,000 give us some flexibility to construct geographic areas that approximate users' areas of interest even though a tabulation based on a single such PUMA would rarely have statistical validity.

Ren Farley, University of Michigan, Demography (Researcher) return to top
My work involves describing racial differences nationally and at the PUMA level. It is necessary to have detailed information about race, occupation, earnings and industry. I also work on immigration. Immigrants are now clustering in specific occupations within specific industries in specific geographic locations. Only the census can tell us much about the assimilation process for most immigrant streams.

Kathy Kaufman, Columbia University, Sociology (Graduate) return to top
It is difficult to project how one's future research may be impaired by the lack of data on particular variables because one cannot anticipate in advance what issues will be pertinent in the future. But once a new problem emerges, we can count on the necessity to go back and examine the historical record in order to place it in proper perspective. Even a slight increase in the detail of data exponentially increases the information's utility, because it vastly expands the flexibility with which the data can be applied to the examination of unforeseen questions.

George Sweeting, New York City Independent Budget Office, Economics (Nonacademic) return to top
Our research is largely concerned with changes in tax burdens (both income and property) over time, We do both cross-sectional distributions as well as micro-simulation modeling. Detail on components of income, occupation, age, tenure status, and wealth factors are critical to this work.

Jim D'Amato, Spartanburg County, SC, Other Nonacademic return to top
As a transportation planner for a Metropolitan Planning Organization, I need specific population and population derived variables for 700 separate zones for my county. This data is crucial for the proper working of our travel demand modeling program. Other variables (by zone) include numbers of employees; population, labor force, registered vehicles, and occupied dwelling units by zone of residence, school attendance zone, and retail sales.

Paul Lachance, Department of History, University of Ottawa, History (Faculty) return to top
Since my research is primarily on the 19th century, the 2000 PUMS is less important for me than for colleagues who specialize in the next century. Nevertheless, any problem needs to be addressed in comparative perspective, whether forwards or backwards in time. It is also important to think ahead to the census as a source for historians of the future. What we can observe from work on U.S. censuses is that the more detailed they have become over time, the more valuable they are. It is especially important, in variables such as occupation, to provide as extensive a breakdown as possible; since how historians will regroup that data into categories can be expected to change in the future in unpredictable ways, just as it has in the past. As a rough rule of thumb, I would suggest maintaining as a minimum the amount of detail provided in the IPUMS samples for the latest censuses.

Earl Ruiter, Cambridge Systematics, Inc., Other Nonacademic (Nonacademic) return to top
The trend in transportation research is toward lifestyle classifications. Loss of information on income, age and occupation will limit the usefulness of the PUMS data for these types of studies.

Confidential Response return to top
My work is focused on racial inequality in the US. Reduction of PUMS detail may preclude me from on-going analysis of ways racial inequality changes, intensifies and diminishes over time.

Louis Kyriakoudes, University of Southern Mississippi, History (Faculty) return to top
I study internal migration and economic change. To limit the geographic, occupational, income, racial or education data would greatly impair my research

Janice Madden, University of Pennsylvania, Economics (Faculty) return to top
PUMAs of 100,000 have allowed us to examine the effects of race, household structure and education on job and residential locations. The individual data in PUMS are unique in that they allow us to examine the simultaneity of these location decisions. Given the criticality of spatial mismatch to overall economic performance and to equity, it is important that we understand its dimensions and its causes and effects, both for science and for policy.
Detailed occupations and small PUMAs are also critical to measuring the effects of, and assuring the enforcement of, affirmative action and anti-discrimination policies. While many firms can use aggregated occupations, the new economy is relying on much finer matches between skills and jobs. To measure compliance with non-discrimination norms, it is important to have detailed occupational data on workers in the vicinity of the firm.

David Hollingsworth, Nunez Community College, Other Academic (Researcher) return to top
using the regions sets for prediction of both academic and occupational training needs is crucial in planning the ed services, marketing strats, and overplanogram for resource allocation here...

Confidential Response return to top
I use PUMS data in my traffic research. A reduction in geographic detail and detail on income and other factors which influence travel patterns would seriously impact the accuracy of my forecasts.

Willis Goudy, Iowa State University, Sociology (Faculty) return to top
Gerontological studies need detail for more than a 65-plus category, primarily because of the gross differences between those in various subgroups above that age. At a minimum, 65-74, 75-84, and 85-plus age categories are needed. Housing, health, and many other factors vary substantially across these age groups. Both basic and applied research would be hindered without information by more than one age category for those 65 or older.

Debbie Pitts, Missouri State Library, Demography (Nonacademic) return to top
I am the coordinator for the State Census Data Center Program in Missouri. My network of data users will be greatly affected by a reduction in PUMS data. One of the members of our coordinating group uses PUMS data extensively to support the work of local planners and regional planning commissions in Missouri.

Patricia Beeson, University of Pittsburgh, Economics (Faculty) return to top
I am currently using the pums data in a research project examining differences in wage structure and wage inequality across metropolitan areas. For this we need to have at least metropolitan areas identified. Since we are looking at changes over time we need historical comparability. This is very difficult as is, given changes the census has made over the years in reported substate geography. 1960 data cannot be used for substate analysis, and it would be catastrophic for researchers examining local areas to also lose 2000. I hope the Census will maintain some substate geography. While I prefer MSAs, the PUMAs would be okay. Also, since we are looking at wage structures and wage inequality, we need the detail on individual characteristics that might affect wages (age, occupation, race).

Don Burrell, OH-KY-IN Regional Council of Governments, Demography (Nonacademic) return to top
OKI is a transportation planning agency (Metropolitan Planning Organization). For our use, the occupation, housing and household characteristics, and vehicle ownership data are important. The purpose of PUMS data is to enable development of special cross tabulations for specific purposes that would not have wide use. Therefore the more subject detail, the better. The compromise is in the geographic detail for confidentiality. Hopefully the 100k limits will continue to be sufficient to insure this.

Confidential Response return to top
We have been conducting housing needs analyses for state, county, and local governments for almost two decades. Grouped age data will not give us the level of detail we need to conduct the household formation projections on which these analyses are based.

Confidential Response return to top
Our primary use of data is for economic development and planning. However, all the data is important when you are trying to entice more companies, or job training programs, or social issues. Then bring planning into your work. How can you effectively work towards economic development and planning when you don't know what you have? I believe that you can't have enough various types of data when making decisions that will affect peoples lives.
Response time has always been less than what a person wants but on a wish list, I'd like an overview of data, then I'd be willling to wait for detail.
Please keep answers confidential if you can.

Sara McLafferty, Hunter College, Geography (Faculty) return to top
I am very concerned about the loss of detail in all areas, but for my research, geographic detail is crucial. To understanding commuting patterns, geographical barriers to employment and questions of spatial mismatch, we MUST have data for relatively "small" areas (i.e. populations of 100,000 as in the past) Otherwise it is impossible to describe and analyze the movements of people from residential areas to worksites.

Edward Schafer, San Diego Council of Governments, Government (Nonacademic) return to top
A significant aspect of my work deals with industrial clusters. These are defined from detailed industry classification. The loss of this detail would severely restrict my ability to produce comparable information over time.

Jack Solock, Center for Demography and Ecology and Center for Demography of Health and Aging, University of Wisconsin-Madison, Demography (Support) return to top
I am a data librarian who serves over 100 faculty, students, and researchers in two demography centers. Many of these researchers use PUMS data extensively and would be severely affected in their research by the proposed reduction in PUMS detail. In particular researchers in aging in the Center for Demography of Health and Aging would be hampered by grouped age data.

Don Larrick, Office of Strategic Research (the State Data Center in Ohio), Government (Nonacademic) return to top
Leave the age variable as it was in 1990. Our research and the work we do for clients in-and-outside of state government requires that we be able to create age groups as desired.

Erzo Luttmer, University of Chicago, Economics (Faculty) return to top
In our research we identify the role of network effects in welfare participation decisions among non-native English speakers using the fraction of speakers of the same language in a PUMA as a proxy for the size of the network. Loss of geographical detail (larger PUMAs), loss of detailed ethnicity/language information or loss of detailed income information would be detrimental to this line of research. In other research, I look at the effects of behaviors and racial composition of welfare recipients in one's area on support for welfare spending. This research too relies heavily on PUMS data where detailed information at the PUMA level is critical.

Richard Sherwood, Maine State Planning Office, Sociology (Nonacademic) return to top
I am a policy advisor to the Governor of Maine and his cabinet. I have probably used the PUMS files two or three hundred times over the past twenty years to address issues raised by the cabinet or the legislature. It is essential in this sort of setting to be able to undertake customized tabulations and cross-tabulations. Sometimes, I am asked about a policy pertaining to a narrow age cohort, e.g., ages 2 to 3 or ages 13 to 15. Sometimes, I am asked to focus on a specific industry or occupation or income level. For example, I have used the PUMS file several times to analyze residential property tax burdens for households in which the number of occupants, the householder's age and the household's income all met criteria specified by a legislative committee or by the State Tax Assessor.

Robert Burchell, Center for Urban Policy Research, Rutgers U., Other Academic (Faculty) return to top
We use the PUMS data (CD's) for various projects, although much of my work is related to determining need for affordable housing. In order to use the PUMS we need age data, but grouped by 5 year groups would be work. Reducing geographic detail would be disastrous. Reductions in detail regarding income, beyond grouping to ranges of $100 or maybe $250, would also be disastrous. We already have contracts with two agencies, the New Jersey Council on Affordable Housing and a large county in another state, which rely heavily on the availability of PUMS as it was done in 1990. In the past we have used PUMS with a number of other governmental groups to look at housing affordability. We have also used the PUMS for determining the characteristics of households in new housing and for various journey to work studies.

Stephen Wilson, SRF Consulting Group, Industry (Nonacademic) return to top
Working and commuting patterns of the 65-69 age group very significantly for the 70+ . As the workforce ages this becomes a critical issue for the travel behavior business.

David Kruse, Alamo Area Council of Governments, Government return to top
We need PUMS data for small area forecasting of population and employment in support of transportation planning, especially travel demand forecasting. I have also used PUMS data for social services planning.

Joyce Benson, Maine State Planning Office, Government (Nonacademic) return to top
State level policy research/analysis has relied heavily on PUMS to obtain the level of detail to gauge impacts of various programs and policies. I am often asked questions such as what is the social and economic profile of persons who work as farm laborers by a legislature or state dept. considering the development of programs or services to the farming industry. This kind of profile can only be developed by having detailed information on age, household structure, housing conditions, income, etc. as found in PUMS. I have relied on the data from the 1990 pums file hundreds of times in the past decade. Historical consistency is also critical to my work.

David Elesh, Temple University, Sociology (Faculty) return to top
I am an urban sociologist who continually works on issues of occupational and industrial change. The available database--particularly with respect to industrial change--are already very limited. In a period of rapid industrial change and concommitant urban change, the proposal would seriously damage researchers' ability to examine the causes and consequences of change.

John Garcia, University of Arizona, Government (Faculty) return to top
On the race data, reduction of the newer combinations of racial categories would undermine the breadth of the expanding racial make-up of the U.S. also it would be more difficult to conduct "historical" analysis over past censuses.

Sarah Wilhelm, University of Utah, Economics (Post-doc) return to top
Discovering the effect of antidiscrimination laws on occupational segregation by race would be impossible.

Jerome Deichert, University of Nebraska at Omaha, Other Academic (Faculty) return to top
We have had a number of questions from our Congressional delegation concerning the impact of various policies on Social Security recipients in our state. To do this we needed detailed income by detailed age by type of income so we could run different scenarios by different age and income cutoffs.

Joseph Willey, Binghamton University, Other Academic (Support) return to top
Academic support, so focus is others research needs. Many geographers, sociologists and economist relying on PUMS. Most notable needs are in the field of labor economics. Without good quality, COMPARABLE, Pums data cross-temporal comparisons are incredibly difficult. In this, as in all cases, I am pleading with the Census Bureau to actually stick with consistent classifications for a while and not tinker with definitions and indicators every ten years.

Claudia Goldin, Harvard University and the NBER, Economics (Faculty) return to top
I would like historical continuity in the PUMS.

Tara Watson, National Bureau of Economic Research, Economics (Graduate) return to top
I do research that requires detailed race and income information. Grouped income bins make it impossible to accurately determine changes in income across areas or demographic groups. Suppression of racial information would also make it impossible to do my research.

Betty Dodson, university of Nevada, Reno, Other Academic (Faculty) return to top
I live in the State of Nevada. We have had a high in-migration of elders. Geographic information and age data are essential for a state that is experiencing phenomenal percentage growth in its population base. In fact, any and all detailed information is important as we lack a sophisticated infrastructure at the state level for collecting useful information that will support grant applications and help to identify areas of need.

Jeffrey Kamakahi, College of St. Benedict / St. John's University, Sociology (Faculty) return to top
I use the data in my statistics and demography courses as well as for background on other research. Since data can always be truncated into larger categories, the greater the detail in the data the more useful it is for researchers and research questions. Aggregation of income data, for instance, might very well mask curvilinearity and would have an effect upon what statistical procedures the researcher would be able to perform in the analysis.

Walter Treichel, Wilder Research Center, Demography (Nonacademic) return to top
PUMS offers the only way in which sub-populations not included in Census crosstabs can be studied. I find geography particularly important. Minnesota has 30 PUMAs currently and degrading that to about 15-18 would make it almost impossible to draw distinctions between city, first ring, second ring, and rural residents. This level of detail is critical for us in conducting research and program evaluation.

Matthew LaPenta, Federal Reserve Bank of Boston, Economics (Nonacademic) return to top
My research is concentrated on migration patterns by educational attainment, age, race, occupation, and income. Comparability across years is essential for me to look at migration trends over time.

Linda Giannarelli, The Urban Institute, Other Nonacademic (Nonacademic) return to top
In the past, the PUMS data have been used as input to microsimulation models of tax and transfer programs. The PUMS have never had the detail of the CPS, but they do offer the possibility of state-specific analysis. Limiting the amount of detail on income could make it impossible to use the PUMS for microsimulation in the future.

Keith Lawton, Metro, (MPO in Portland, Oregon), Government (Nonacademic) return to top
Note: We do limited "research", But we use the PUMs/PUMAS extensively in the development of synthetic populations by small area (Block Group)for application in the microsimulation of personal travel in models of urban travel demand. (A planning application). In this arena both income and age are related to demand for mobility, and ages above 65 show decreasing demand for travel that we need to understand when forecasting a future with an aging population.

Matthew Ides, UCR, History (Graduate) return to top
My last research with IPUMS was an examination of Asian immigrant groups to the US, measuring their social mobility, occupations and geographical locations. I analyzed this data in order to explore the impact of different immigration restrictions. If this categories were collapsed, IPUMS would not have been useful to me. Grouped age data or any other groupings diminish the specificity of social research. The quality of PUMS is that its multi-variate data allows the researcher to refute theories based on broad based aggregates, as well as allowing the research to explore aspects of American society in greater detail. It is much easier for an individual to pay 25$ for a web service to find out someone's elses personal information, why and how could someone use IPUMS for this purpose?

Jeffrey Reitz, University of Toronto, Sociology (Faculty) return to top
My research examines the economic success of immigrants as this varies across urban areas, looking at specific racial origins groups, and examining the effect of changing occupational characteristics. The analysis requires information needed to construct conventional human capital models (e.g., education, age, gender). Reductions in the detail of racial data would severely reduce my ability to consider and/or eliminate race as a factor in urban area differences. Reductions in the detail occupational data would make it difficult or impossible to examine how labour market change is affecting the success of immigrants.

Pamela Klein, Penn State University, Sociology (Graduate) return to top
I specialize in military sociology so any reduction in occupation and income would have negative effects on how easily military personnel are identified. As it is, I have a very difficult time determining if someone who identifies themselves a military personnel are enlisted or in the Reserves or ROTC, etc. If there was a reduction in detail concerning age, occupational status, and income, I could no longer even attempt to analyze the level of poverty in the current military ranks, nor the differences in poverty levels among veterans of different periods/conflicts.

William Grant, Old Dominion University, Government (Graduate) return to top
GIS analysis of multiyear census data would be seriously impacted resulting in loss of urban analysis and a significant reduction in the effectiveness of this emerging analysis tool.

Patrick Costinett, Parsons Brinckerhoff, Industry (Nonacademic) return to top
PUMS has been and is becoming even more important in the development of behavioral travel demand forecasting models in which households and members are classified by multiple characteristics such as income, occupation, industry, age, household size, and vehicle ownership. Similar information is collected in special surveys but these involve small sample sizes and high cost per sample. PUMS is a critical basis for survey expansion, analyzing sample bias and providing more detailed cross-classifications of households and individuals for model building. Maintenance of socioeconomic detail is more important than greater geographic detail but race and ancestory are generally not of interest in travel demand modeling.

Michael Gillett, Oregon DOT, Government return to top
We use the data for travel estimation and forecasting. The detail that is needed for our work is geographic detail. The household demographic data is often aggregated to three or four income groups, age groups, HH size groups, etc.

John Pitkin, Analysis and Forecasting, Inc., Demography (Nonacademic) return to top
I study housing consumption by birth and immigration cohorts. Loss of age detail for the elderly, or at any age, would limit my flexibility to compare specific cohorts between censuses. This is fundamental to my research. I am especially concerned that with the expected shift to ACS data, I will need more detailed PUMS data to calibrate PUMS-to-ACS changes for non-standard age groups. The ability to look at PUMS data for smaller areas (100,000-plus) has been very important for intercensal comparability, e.g. comparing metro areas with constant boundaries. The larger the "building blocks" the worse the intertemporal comparability.

Irene Browne, Emory University, Sociology (Faculty) return to top
I study race/ethnic differences in labor market outcomes among women. In this area, there are some important trends that have important policy implications. For example, by 1979, the earnings of Black and White women were almost equal. In the 1980s and 1990s, Black-White wage inequality INCREASED. Why are we moving towards more racial inequality? This is particularly important as a policy issue given that many women are supporting their families alone. The detailed occupational and industry categories, income, labor market status and geographic boundaries in the PUMS data are INDISPENSABLE for my research. I create aggregate measures at the MSA level to track changes in economic opportunities. For instance, in my most recent publication, I show that changes in employment opportunities within PARTICULAR TYPES OF SERVICE INDUSTRIES (retail services) benefited White women but not Black women. (Social Forces, March 2000) I COULD NOT HAVE PERFORMED THESE ANALYSES WITHOUT PUMS DATA THAT WAS COMPARABLE ACROSS YEARS. The 2000 PUMS will be even more important, as it is the ONLY dataset that captures the characteristics of relatively small but rapidly growing race/ethnic groups, such as many Asian groups. Only the PUMS, with its 1990 detail, allows for the possibility of creating population and labor market measures that can help determine how changes in labor pools within local labor markets affect women's wages and employment.
I am also intending to look at labor market trends among Asian-American women. This group can ONLY be studied nationally with decennial census data (the CPS is inadequate). The geographic information is important, as different areas of the U.S. have very different economies. The occupation/industry/income detail are important because we cannot determine how Asian women are faring in the labor market without knowing precisely which occupations and industries in which they are employed and what they are earning.

Frank W. Sweet, American Military University, History (Graduate) return to top
My research focuses on the history of U.S. black/white intermarriage; how it varied over time, and across states. Without race, marital status, and geography, the census would be worthless.

Mark Hardt, Montana State University-Billings, Sociology (Faculty) return to top
I do time series research. A major problem I often encounter is a difficulty, if not an inability, to measure variables over time. Reducing the number of categories will have a serious impact on the ability to conduct historical research and to analyze trends.

Jonah Gelbach, University of Maryland at College Park, Economics (Faculty) return to top
Much of my research uses age structure to identify effects of public expenditure programs. Moreover, my research frequently involves using variation across states in public policies. Without these variables I might as well not have the PUMS at all.

Blaine Billings, Boise State University, Sociology (Undergraduate) return to top
Life span longevity (higher mortality age) Higher age requirements on retirement Social Security Bankruptcy scheduled for 2035 (determined by US Department of Social Security) Between the ages of 65 and over covers a minimum of 4 extremely significant socioeconomic categories. 1) Still working 2) Economically stable on retirement 3) Economically assisted by government subsidy 4) Total public subsidy for medical care

Melissa Levitt, San Francisco State University, Other Academic (Faculty) return to top
My research areas are cities, and I primarily concentrate on race-based or ethnic based shifts in population, income brackets and occupational changes. Any modifications that would limit comparability over time would severely damage my ability to conduct thorough research.

Andrew Harvey, Saint Mary's University, Economics (Faculty) return to top
My interests are in regional research, occupational change and the aging population. With the labour market as fluid as it is aggregated data along any of these dimensions would cripple much of the analysis than should be done.

James Gregory, University of Washington, History (Faculty) return to top
I study the effects of migration and need both geographic detail and the full range of demographic detail. To scale back either would be a tragedy with both immediate and long-term implications. I pity the policy analysts, not to mention the historians, who will have to look back on the 20th century as the golden age of social research.

Benjamin Ben-Baruch, StarWorks, Inc., Industry (Nonacademic) return to top
The most catastrophic affect that decisions of this kind could have on market research would be to compromise confidentiality and to cause people to lose faith in the ability of the Census Bureau and/or of researchers to protect privacy, maintain confidentiality, and protect the integrity of research.
Maintaining public confidence that privacy is being protected is -- in the long run -- more important than better data.

Bill Godair, Old Dominion University, Other Academic (Graduate) return to top
With reference to question 6: Need to broaden occupational categories to reflect emerging technical occupation, however at the same time, there is a need to maintain comparability across census years.
If geographic detail is reduced then the ability to track changes and compare changes in educational achievement, industry and occupational changes, between different geographic area would be lost. Since I am primarily interested in changes at the central city and MSA level of detail I would be unable to conduct my research.

Confidential Response return to top
My research tracks men's work involvement relative to their family status from 1940 on. Although I have only been tracking behavior thus far, the next stage of my analysis will be to analyze my data according to various demographic variables. At some stage of my research, I will want to use fairly specific racial and occupational categories. If these variables are made more general, it might hurt my research.
One of the great things about census data is the amount of specific data with such a large sample size. There is nothing else that even comes close. I think it would be a mistake to reduce the level of detail available to researchers.

Anne Jachim, Moraine Valley Community College, Other Academic (Faculty) return to top
Our world and technology are changing at an astronomical rate. Those of us involved in research for either academia or business depend upon the research results of PUMS as a viable resource.

Louise Gunderson, University of Virginia, Other Academic (Graduate) return to top
I am working on the problem of crime prediction. Detailed information is vital to this work.

Confidential Response return to top
The type of research I am currently working on requires detailed information at the level of the census-tract. When this is not available, I have been forced to use metropolitan areas as the unit of analysis. I require social and demographic detail on residents of neighbourhoods over time (1900-2000) and therefore seriously rely on consistency in data compatibility across decennial census years. Race, socio-economic status, and age are integral components of my research, and to limit the categories available would jeopardize analyses over time. Longitudinal research is extremely fruitful, relative to cross-sectional research, in determining time-order and macro-level trends. To alter the type of data released for the 2000 census would, in essence, impose severe limitations to longitudinal research and may even reduce the utility of cross-sectional research by limiting the richness of descriptive data.

Deborah Ward, Columbia University, Government (Faculty) return to top
The most important loss is in comparison over time. We should never go backwards in data collection, only forwards.

Kenneth Sunderland, University of California - Riverside, Economics (Graduate) return to top
I use nonparametric methods to extract and test information on income and earnings distributions across virtually all the categories you named above. In order to do this, I need datasets that are both large and detailed so that I can break them down in a variety of different ways and still have sample sizes large enough for reasonable nonparametric estimations. Without sufficient detail, comparisons across categories and subcategories become difficult or impossible. In particular, I have developed a method of quickly generating distributions for age on a per year basis that reveals considerable detail in the evolution of income/earnings with increasing age that is not available with parametric methods. This would become almost impossible with aggregated data.

Wendy Gordon, Plattsburgh SUNY, History (Faculty) return to top
Because I research the nineteenth and early 20th centuries reduction of detail in the 2000 census would not immediately jeopardize my research. However, loss of detail in any variable endangers the ability of future historians to analyze data by methods and for reasons that we cannot imagine today.

Carrie Yang Costello, University of Wisconsin-Milwaukee, Sociology (Faculty) return to top
I do research on professionalization, for which detailed occupational categorical data is necessary. I also do research on identity, for which ancestral and racial data are vital.

Jill Marlette, Geography (Nonacademic) return to top
I use income, age, occupation and other variables to create maps. The more detail I have, the more informative my maps are.

Charles Starnes, Oregon State University, Sociology (Faculty) return to top
Income distribution data have always been so crudely truncated at higher income levels as to severely damage research on topics where income inequality is a relevant consideration. We need more information on distribution at higher income levels, not less!

Confidential Response return to top
I am a graduate student at Queens College/CUNY and using the IPUMS for my thesis. Any reduction of any sort regarding the data quality would definitely prove to be harmful to me, to other researchers, and to the general citizen who would like to "see" how the society they live looks like.

Paul Voss, University of Wisconsin-Madison, Demography (Faculty) return to top
I mostly do applied demographic work. When the STFs are not detailed enough, one must turn to the PUMS data. The geographic detail is already too large. If it were increased to something larger than areas with minimum 100,000 population, the PUMS would lose much of its usefulness to me.

Jason Fields, U.S. Census Bureau, Demography (Nonacademic) return to top
While I am certainly biased, and have access to internal files for any runs off the 2000 census that I will need to make, I have used IPUMS data to supplement a bureau report and hope to continue using it. The ability to make time series comparisons is invaluable and should be retained in as much detail as possible. Examination of detailed living arrangements will certainly require the detailed age data beyond age 65, for both aging and child related research.

Leslie McCall, Rutgers University, Sociology (Faculty) return to top
On the specific question of how to deal with occupations (question #7), I think a major issue is that certain detailed occupational categories are very "timely" ones and these cannot be determined in advance (e.g., securities analyst, or, in terms of industry, personnel supply industry workers).
On the more general question of the broader impact of less detail, it would be truly catastrophic since my research is on earnings inequality in *PUMAs*, the smallest geography available in the PUMS and the only one that includes non-urban areas. States are too large to use as labor markets and urban areas are not representative of the country as a whole or of the differences between segments within large metro areas (i.e., between central city and suburbs). Since more and more of our theories and policies emphasize the importance of place and locality, and the U.S. has an extremely large and geographically diverse population, it seems that the Census should be moving toward more geographical detail rather than less. Another issue is income. The U.S. has a very poor documentation system when it comes to wealth. Given the large accumulations of wealth over the 1990s, it seems especially important for the 2000 Census to be able to accurately reflect the structure of income (if not wealth), which can't be done if there is top-coding. This has always been a major weakness of the census.

Dunbar Brooks, Baltimore Metropolitan Council, Government (Faculty) return to top
We are an MPO and we use these data for a variety of purposes in transportation planning and to assist local governments in developing policies and/or building a case to apply for federal and state funding based on the detailed crosstabulations that we can make. The smaller the geographic detail the better. PUMS data is only data that allows you complete freedom to develop critical, but non-published data tables that may fit special circumstances. We have used PUMS to assist state and local governments in developing child care programs. We have used PUMs to assess certain conditions among the disabled. I have been making PUMs tabulations that aid both government and private industry for a quarter of a century. It would a travesty to now dump this program or to severely diminish it.

Julie Whittaker, Rutgers University, Economics (Faculty) return to top
Wage distribution by sex and race. As a labor economist I cannot think of what could be worse for my research than these proposed changes. Geography, age, race, sex, occupation...all of these are necessary to understand labor markets and the fundamental changes that occur.

Confidential Response return to top
We used the data to track the educational attainment of teachers, looking for trends and focusing on age, race, and gender. Losing access to this information may hinder the research that influences the teacher training programs across the United States.

David Olive, Southern Illinois University, Other Academic (Faculty) return to top
It could be useful to make year by year comparisons. For example, having all of the variables from the 1800's in the 2000 data set would allow comparisons between historic and current data.

Confidential Response return to top
I have worked on some research on Italian Americans and if there was limited information available on ancestry this research would be severely compromised.

Dr. Gene George, Butler County Community College, Demography (Support) return to top
I do demographic research to support strategic planning at a community college in south-central Kansas. It is critical that we know as much as we can about the six-county region we serve. We especially need to understand the long-term dynamics of population growth and decline and the evolution of an urban society in what was once a rural area. Analyzing the change in income, occupation, and race is vital to our understanding of what's going on in the region. And since we rely on demography to identify more specific qualitative research (focus groups, depth interviews), a loss of demographic detail will diminish the effectiveness of our social research into emerging trends. Thus, the college's capacity to make effective plans for its future will be damaged.

Confidential Response return to top
We use the PUMS data to identify sample bias resulting from non-response. We also rely on PUMS data in the development of our sampling plans, to identify our sampling universe. If reduction is necessary, the sooner we are aware of the new categories, the easier our post-processing will be (i.e., we can design instruments to collect data in line with the new categories). If reduction is necessary, we request that the reduction take place along current categories so that comparisons can be done at some level rather than not at all.

Ken Prandy, University of Cardiff, UK, Sociology (Faculty) return to top
My research, which is part of an international comparative study using census data from a large number of countries, involves an analysis of the occupations of married (or cohabiting) partners in order to develop a graphical representation of social space. The first dimension of this space is one of generalized social advantage/disadvantage, so that the co-ordinates provide a measure of social hierarchy.

Mark Killingsworth, Rutgers University and NBER, Economics (Faculty) return to top
On #9 above, it might be possible to release a less detailed dataset immediately, and LATER release a more detailed dataset -- the two strategies aren't mutually exclusive.
In answer to this question, analyses of earnings need detailed occupation data, and an enormous amount of detail would be lost if the detailed occupation categories were collapsed. Note that this has significant implications for equal pay policies.
Likewise, analyses of earnings and of hiring need detailed data on residence and commuting patterns. The STF files don't allow one to do any meaningful analysis controlling for covariates like education -- you have to take the info. in exactly the form the STF tables give it to you. The alternative, which is far superior, is to use the PUMS data, where you can look at reasonably detailed info. on residence/commuting patterns and you can also control for education, age and the like. Note that here too there are some significant implications for public policy -- for example, "utilization analysis" for EEO purposes often requires info. on the geographic distribution, within rather small areas, of persons with specific characteristics (education, age, occupation, etc.) by race and sex. A lot of this would be lost if the geographic detail were condensed.

Hannah McKinney, Kalamazoo College, Economics (Faculty) return to top
The key issues in US urban policy today are race and class. We need to untangle the interactions between income, education, age, race, and class to try to understand the extent of spatial and labor market discrimination in the market. This data is indispensable for this purpose.

Michael Stoil, Conwal Incorported, Industry (Nonacademic) return to top
Health care delivery research by ethnicity is not available from SSI/Medicare records because SSI Idents lists only White, Black, and Other. As the population diversifies and we become more concerned about the interaction of gender, age, insurance and health care access, and ethnicity on health status, there is no substitute on a nationwide basis for PUMS detail. We simply will not be able to accurately track these variables if detail on race, nativity, age, and ethnicity are not available.

Harriet Duleep, The Urban Institute, Economics (Researcher) return to top
I do research on the socioeconomic determinants of mortality (including occupational experience, income, where one lives), as well as the economic assimilation of immigrants and their effect on the employment opportunities of the native born. Obviously, having very detailed information on all of the variables discussed above is imperative for the quality of this research, and its ability to speak to important policy issues.

Confidential Response return to top
I specialize in research in American economic history that make VERY heavy use of IPUMS data -- for example, data on occupation, educational attainment, home ownership (the latest paper, covering 1900 to 1990), migration, income, and many other variables. Such studies are essential for putting current policy problems in their long-term historical context; without the necessary detail, such studies are difficult or impossible to conduct.

Mary Kritz, Cornell University, Demography (Researcher) return to top
My research focuses on demographic and socio-economic analysis of the foreign-born population in the USA. I have carried out a number of research projects using 1980 and 1990 PUMS files. Most of my analyses involve disaggregation of the foreign-born population by origin and geographic residence in the USA. Several of the research projects that I have done using the 1990 files would be precluded if a higher level of aggregation is utilized and the research that I would like to do has already been constricted because of sample size constraints.
One of my projects has focused on the internal migration patterns of the 20 largest foreign-born groups in the USA, using PUMS files. In 1980, a similar analysis could only be done for 10 foreign born groups. (Some peculiarities in how the Census Bureau classified internal moves also constrained analysis possibilities in 1980.) Internal migration of the foreign-born into non-metropolitan areas is a topic that was already precluded in 1980 and 1990 because of restricted sample size.
Another one of my projects focuses on living arrangements of elderly foreign born from the 10 largest origins. On this topic also, the number of persons aged 60 and over from different countries of origin is already very small. I did some explorations with the 1990 files to determine whether there were sufficient cases to permit comparisons between immigrants and native-born persons of the same ancestry. Using the ancestry classification (i.e. 2nd generation and higher), I discovered that there were only a sufficient number of cases [elderly 60+] from the two largest countries of U.S. immigration (Mexico and the Philippines [N=812 from the combined PUMS 1/1000 and PUMS-O]). If I aggregated Chinese from different origins, not an advisable strategy, I would also be able to obtain a sufficient number of persons of Chinese ancestry for national-level analysis [N=2,507]. However, for the 4th largest U.S. immigrant group, Dominicans, the number of elderly of Dominican ancestry drops to 119 cases. As these figures indicate, the number of cases for detailed subgroup analysis is already too small for certain types of analyses based on the 1990 PUMS.
A third illustration of a research project that I did on which I quickly "ran out of cases" involved comparisons of women's labor force participation for Hispanic foreign-born women from the 5 largest groups (Mexico, Dominican Republic, El Salvador, Cuba, and Colombia). I wanted to compare how their levels of labor force participation differed by metropolitan labor markets. After checking sample sizes, I discovered that I only had a sufficient number of cases in 3 metro areas -- New York, Miami, and Los Angeles.
I have been looking forward to replicating and extending analyses done on the 1980 and 1990 files, based on the expectation that the sample sizes would be larger for the 2000 PUMS given the continued and large volume of U.S. immigration during the 1990s. Apparently that may not be the case. As the U.S. population becomes more diversified, it is important to do subgroup analyses of multi-racial/ethnic populations.

Douglas Gurak, Cornell University, Demography (Faculty) return to top
Other countries, e.g. Canada, already group information for place of birth and those data for national origin are not very useful. Having as detailed information as possible for origin, occupation, location, age, and other characteristics such as income is essential if we are to be able to learn anything about what is really going on in important areas such as: 1) occupational success of minorities, immigrants, or other groups; 2) migratory patterns, causes and consequences of the general population and specific sub groups such as immigrants and the elderly; 3) fertility trends and correlates for the general population and specific ethnic groups. A lack of detail in any of these and similar areas lends itself to findings that are misleading because they are averages of more complex underlying patterns. The greater the detail, the better our ability to identify such complexity.

Matthew DeZee, University of Maryland (at College Park), Economics (Undergraduate) return to top
The reduction of detail would be catastrophic to future economic research.
A general grouping of age or income would strongly deteriorate any research in labor economics.

Warren Barclay, New Jersey State Department of Personnel, Government (Nonacademic) return to top
1. With employees retiring later, past 65, it is necessary to have occupational data by the "above 65" years to determine availability for affirmative action programs.
2. Having less occupational detail will hinder establishing availability percentages for affirmative action programs.

Gary Portenier, Milwaukee County Department on Aging, Government (Nonacademic) return to top
Our use of PUMS data has been limited in the past, but hope to use this information in conjunction with mapping software to improve program planning.

Rowan Wolf, Portland Community College, Sociology (Faculty) return to top
Reduction in PUMS data would impact my research and that of other sociologists dramatically. First, reducing the detail decreases and may invalidate historical comparisons of the population as a whole, and of specific populations. Second, sociological and work force research indicates there are transitions in how inequality works in the labor force which is a trend towards smaller units - rather than industry-level units. Losing the detail masks the ability to see these changes. Third, we have rapidly changing occupations and race/sex stratification within fields (i.e. clerical, decreasing detail makes it impossible to see these changes.

Bobray Bordelon, Princeton University Library, Economics (Faculty) return to top
Princeton University uses the PUMS data for Economics, Sociology, Demography, and Politics. I see a broad spectrum of students and faculty.

Eileen McConnell, University of Notre Dame, Sociology (Graduate) return to top
My research focuses on Mexican immigrants in the Midwest, thus, less detailed information about ancestry would mean that I could not determine whether individuals are foreign born, whether they are children of immigrants, etc. Additionally, states and even MSAs vary widely in proportion of the foreign born etc--thus less detailed geographic areas means that estimates of local/area characteristics are less precise. Similarly, grouped occupational categories means that it would be difficult to compare detailed occupations within the manufacturing/industrial occupational sector across regions of the country--for example, percentages involved in food processing may be similar in CA and NE (they are not--but just as an example)--but employees in CA are involved in canning while in NE they are in meat processing and packing. With grouped occupational data--that might be impossible to differentiate.

Ken Blum, Center for Labor & Community Research, Other Nonacademic (Nonacademic) return to top
I have done research using specific occupations. If this detail was eliminated, it would destroy this type of research. We serve poor communities.

Elaine Murakami, US Department of Transportation, Government (Nonacademic) return to top
The transportation community pays for the Census Transportation Planning Package through a cooperative project via AASHTO. However, the CTPP is only part of the Census data used by the transportation profession. PUMS is also a critical Census data source for transportation planners.
We are primarily concerned about the size of the PUMAs. We do not want to see the PUMAs be any larger than 100,000 population, and would, in fact, prefer that the size be decreased as much as possible. When the geography of the PUMA becomes too large, the PUMS becomes less useful for transportation planning. For example, transit use is directly tied to accessibility to transit buses and rail systems. If the PUMA includes areas with both high and low transit accessibility, we are less able to use the data for regression or other analyses on travel mode choice. Similar arguments can be made for other census variables such as departure time to work, vehicle availability, etc., and related models and simulations.
There is a trend in the transportation community toward microsimulation of travel behavior to address congestion and air quality issues. That is, characteristics of individuals and characteristics of their household are used to develop activities and trips which are applied to the transportation network to forecast such things as travel volume and travel speeds, which are then transformed into estimates and forecasts of congestion and air quality.
We would rather give up detail on response categories than give up geographic detail. That is, we would be willing to live with some grouping or collapsing of categories for age, income, and occupation. Because of environmental justice concerns, we DO need race and Hispanic origin data, but we could probably live with a limited number of race categories in the PUMS, since the detailed counts will be available at the block level from PL-94-171.
Submitted by CTPP Working Group

Confidential Response return to top
Often need the PUMS data to assist in comparing other suveys on income by age, and region, race etc. We never know exactly and would have to request special tabluations if not available.

Patricia Fernandez-Kelly, Princeton University/Department of Sociology and Office of Population Research, Sociology (Faculty) return to top
I conduct ethnographic research in urban settings. Because my work emphasizes behaviors found at the local level, I often depend on detailed Census information concerning SES, occupation, ancestry, and other demographic characteristics. Reduction in the level of detail on these and other characteristics would have a very harmful effect on my research.

Lorraine Porcellini, ISR, Temple University, Other Nonacademic (Nonacademic) return to top
To determine sample sizes for a study, I use PUMS data to calculate the proportion of the population I can expect for certain subcategories, most often age, sex, race, Spanish origin, marital status, income, and education. I use these data to determine the number of households I must sample to attain the required subgroup sample sizes. I often need to know the breakdowns by state, region, MSA, or PUMA.

Bruce Western, Princeton University, Sociology (Faculty) return to top
I have used the PUMS to study racial inequality in employment, and earnings and employment among immigrants. The PUMS is an invaluable data source for studying narrowly defined demographic groups, such as young minority males with low levels of education. No other source offers the necessary quantity of data for studying such small groups over a relatively long time span. Such research is only possible because the PUMS provides detailed information about age, region, and schooling. If only aggregated codes were reported, the value of the PUMS would be significantly diminished. Rival data sources, like the Current Population Survey, provide relatively large sample sizes for coarsely defined groups. In my view, eliminating detailed demographic codes in the PUMS would largely undermine the utility of the Census for social research.

Jim Reilly, NJ Office of State Planning, Government (Nonacademic) return to top
If reductions are unavoidable, they must be made in a manner that allows comparability with earlier data.

Confidential Response return to top
Regional modeling requires consistent data at the county level for my research, therefore the loss of detail in the dataset would make future updates difficult if they weren't in some way compatible.

Confidential Response return to top
Need maximum level of crosstabs possible for detailed data on age, income, migration, and housing.

Confidential Response return to top
We are doing transportation planning consulting to all level of governments. Accurate and more up-to-date data are essential to an accurate planning. Age, income, type of job, geographic location of both home and other activity location are essential to the kind of work we do.

Brian Cushing, West Virginia University, Economics (Faculty) return to top
Reducing geographic detail would significantly (maybe severely) limit my ability to study metropolitan migration and the usefulness of my findings. I already lose the smaller metropolitan areas and have limited ability to distinguish central cities from suburban areas. Loss of geographic detail may eliminate most of my research on rural poverty and migration since the PUMAS would be too large to be useful - would have to lump together too many distinct areas. Any loss of detail on income could wreak havoc in research on income distribution, income inequality, and poverty. The decennial census is the only data source that provides the combination of large sample size, detailed characteristics, and geographic detail to allow detailed stratification by personal characteristics and type of place.

Eberts Randall, W.E. Upjohn Institute for Employment Research, Economics (Nonacademic) return to top
One area of research is economic development and the dynamics of local labor market. The PUMS offers the most detailed information about the workforce in smaller metropolitan areas. Without these data, we and others could not provide the analysis that policy makers need and want.

Confidential Response return to top
In Hawaii, we have a great diversity of ethnic groups. There is a great need for the cross-tabulation of data on population and housing characteristics by specific detailed race category. Frequently, these requests involve three-way or more cross-tabulation of data as well and/or using one or more selection criteria. We have had many requests over the years where the PUMS data is the only one available that is able to satisfy specific data needs.

Roger Avery, Brown University, Population Studies & Training Center, Demography (Faculty) return to top
We are working on the new disability questions. Only the census has the huge n that is necessary to study differentials in some prevalence of disability, which now is shown is significantly more detail than before. Detail in age is crucial to our research. Those 65-69 are very different from the 85+, those 5-9 are very different from those 15-19.

Confidential Response return to top
Detailed-level data, by geography, is the foundation in creating the proper sample design in each of the research projects my firm undertakes. By reducing or condensing the level of detail would be detrimental in how we are able to properly sample the market under study (a majority of which are of interest to local transportation planning agencies around the country).

Heidi Wong, Asian & Pacific Islander American Health Forum, Other Nonacademic (Nonacademic) return to top
Our primary level of analysis is at the individual Asian and Pacific Islander ethnic groups (e.g. Chinese, Vietnamese, Hmong, Chamorro, Samoan, etc.). PUMS is the only source from which we can do specialized analyses at this level (e.g. county by Korean by age by income by public assistance). Should PUMS detail be reduced, our ability to impact policy and decision makers around issues concerning the Asian American and Pacific Islander communities. There are many data obstacles our community faces, but in 1990, PUMS data was the one source that we could use to support our advocacy work (e.g. around health care reform, immigration reform, etc.).

Confidential Response return to top
Occupational and SES detail: as the economy changes, the only way we will be able to meaningfully analyze the distribution of work and wealth in our society is to retain as much detail as possible. Currently many different theories and concepts apply to research in this area. Decreasing measures available will preclude us from being able to test existing and new theories. I would say the same about the other variables under discussion as well. At present, we have a dearth of available measures in which to test more and more complex theories. We need more not less. This is especially true for being able to measure the effectiveness of State policies on different groups in society. Please, do not limit yet further our already very limited tools in which to study some of the most pressing social, economic and political questions of our day.

Cynthia Rogers, University of Oklahoma, Economics (Faculty) return to top
PUMS is essential for doing substate level analysis. The geographic detail is essential for making regional research more relevant for practical policy analysis.

Robert Gibbs, Economic Research Service, USDA, Economics (Nonacademic) return to top
Much of my work requires knowledge of the industry and occupational mix of labor market areas (substate metro/nonmetro) This information is indispensable for estimating local skill levels (computed with detailed industry/occupation) and employment opportunity. For instance, estimates of the supply of jobs available to welfare recipients requires attaching average skill and educational information to detailed occupations in a local area. Broad occupational categories are useless for this analysis, since the variation in skill and education are so great within each broad grouping. Similarly, I have computed wage premia for workers in local industry clusters, again requiring detailed industry and geography in order to identify clusters (these must be done at a local labor market level for specific industries).

Judith Stallmann, Economics (Faculty) return to top
I have done research on the elderly. Grouping everyone over 65 does not allow us to look at what happens as people age.
I have done work on rural labor. Geographic detail is essential for rural labor work.
I have done work on the relationship between educational achievement, dropout rates, and the local economic structure. Occupational data is essential for this work. Industry data is not as useful because it includes everyone from the CEO to the janitor.
Occupations are changing rapidly so that the more detail gives an idea of the changes that are taking place and have taken place in occupations. I know that many times researchers use industry data when they really want to use occupational data, but occupational data are only available every ten years, whereas industry data are available annually. We need more emphasis on occupational data.

Confidential Response return to top
The geographic detail is essential!

Jim Ross, CSUB, Sociology (Faculty) return to top
Reduction of PUMS detail would prevent much of my use of census data in my research. The detail is a necessary for my work.

David Pearson, Texas Transportation Institute, Other Academic (Researcher) return to top
Reduction in detail of household information (e.g., size, income, etc.) would virtually render the data for our research as unusable and worthless. When the 1990 census only used household size categories of 1 to 4+, it created some problems in using the data for estimation purposes and in effect, we had to use estimates in some areas.

Confidential Response return to top
My research focuses on urban area (ward/census tract level and up) economic base change and demography across time. Any reduction of industrial classifications or in age, race, income, education, geography, would preclude detailed analysis of across time change in economic base through industrial shift and resulting demographic migration patterns. Given the economic restructuring currently experienced in this country, this research is vital to understanding the services needs of urban areas, the employment pressures on migration etc. My research agenda would be seriously curtailed. While my focus is academic research, much of it has strong policy implications and is developed as such.

Susan Hendricks, KJS Associates, Industry (Nonacademic) return to top
We look at travel behavior and location choice based on geographic location and person and household characteristics, including as age, income, and occupation.

Michael Lahr, Rutgers University, Center for Urban Policy Research, Economics (Faculty) return to top
I do economic research in geographic detail. Mentioning a set number of categories, like 25, is simply not enough. Indeed, 25 may be sufficient but it would have to be something like that displayed in Appendix II of McGranahan, David A. and Linda Gelfi. 1991. "The Education Crisis and Rural Stagnation in the 1980's," Education and Rural Economic Development: Strategies for the 1990's, Agriculture and Rural Economy Division, Economic Research Service, U.S. Department of Agriculture, ERS Staff Report No. AGES 9153, pp. 40-92.

Confidential Response return to top
I analyze the income and poverty status of persons, families, and households in rural or nonmetropolitan areas. If the 2000 PUMS does not contain detailed information on sources of income, poverty status, and the employment and occupational status of persons by place of residence, it will be of no use in my research.

Jon Moen, University of Mississippi, Economics (Faculty) return to top
A loss of age detail will produce truncation and censoring problems for most empirical work, rendering research into retirement useless.

Lionel Beaulieu, Southern Rural Development Center, Sociology (Faculty) return to top
My work focuses primarily on human capital and workforce development issues. The PUMS data allow us to examine these critical issues across various spatial areas and labor market areas. The lost of specificity in the PUMS data would seriously compromise our ability to understand the unique dynamics associated with urban/rural areas in various LMAs.

Tom Trozzolo, University of Notre Dame, Sociology (Researcher) return to top
Any attempt at comparability between the "youngest old" (65-74), "middle old" (75-84), and "oldest old" (85+) becomes impossible without birth year information remaining as is.

Confidential Response return to top
Since I assist an area agency on aging to utilize major population data on trends for providing services needed by the growing Aging cohorts, any major deletions would be critical. For example, we cannot accept a lump sum category like ages 65 & over. We MUST have more definite groupings to meet different trends that we know have been in progress over the last two censuses. Age groupings, economic groupings, health status and other changes within counties are REQUIRED to meet both service needs and planning requirements.

Marie Bousfield, City of Chicago, Government (Nonacademic) return to top
To create a forecast of population and employment as required by ISTEA and the Clean Air Act, we used PUMAS as basic units of geography. The model we use (DRAM-EMPAL created by Professor Putnam) requires input that is not available from the Bureau's prepared tables. The Chicago City Council has approved several programs for middle-income homeowners and has developed guidelines for scattered sites for public housing. These require custom-made tables generated from the PUMS file. Q9 reformulated: Do you want the electric chair or the guillotine.

Rose Martin, New Jersey Department Of Health, Government (Nonacademic) return to top
The population over the age of 65 is growing rapidly and is projected to continue this growth well into this century, in view of the increasing life expectancy. In terms of health indicators, we have found major differences between the "young" elderly (e.g., 65-84 years of age) and the "older" elderly (85+ years). If we do not have estimates of the population beyond age 65 broken down into ten or twenty year age groups, we can not assess health status in these groups or plan health programs to improve the quality of life for this population.

Pamela Perlich, Bureau of Economic and Business Research, Economics (Researcher) return to top
In our economic and demographic modeling, age detail in the Census is our only cross-sectional data that can be cross classified with other characteristics. It is the most basic input to our modeling work.

Confidential Response return to top
I do housing research and all of the suggested variables are important.

Confidential Response return to top
My research on changing occupational sex composition across time would be absolutely impossible without occupation data at the detailed occupation level. I have a dataset of occupational characteristics aggregated from the 1970, 1980, and 1990 PUMS datasets. Updating it further with 2000 data would be impossible with any loss of detail.

Arthur Wittman, Broward County School Board, Government (Nonacademic) return to top
I am the demographer for a large school district and ideally I would like even a smaller level of geography than 100,000. So making it 250,000 or even higher is a move in the wrong direction, in my opinion.
The most need piece is 0-4 years with cross tabs - ideally at the Census Tract level - but no more aggregated than 100,000 population.

Confidential Response return to top
Teaching requires detailed information about historical times so students can understand what types of occupations were available and how they might have been gendered. For example, in Family Problems and Social Change, I lecture on what percentage of jobs women and men occupied and so forth. All demographic information is pertinent.

Donald Hirasuna, House Research Department, Economics (Nonacademic) return to top
We are asked a variety of questions regarding different demographic groups in different regions. For example, I have been asked about the distribution of occupations across geographic regions. Another example is to determine the number of single parents by real income categories for different regions across the state.
The geographic, age and income detail is also indispensable for monitoring policy issues such as access to jobs in central cities. Without the current detail of the PUMS data, much research cannot be properly conducted.

Alexandra Hill, University of Alaska Anchorage, Economics (Researcher) return to top
Geography: Currently Alaska has 2 PUMAs in Anchorage (101 and 102), which I always combine; PUMA 200 that can best be categorized as "other non-bush" and PUMA 300 that is very rural. If we were to lose PUMA 300 (which is the smallest in population) our ability to do any urban/rural comparisons or to look at rural Alaska populations would be eliminated. Combining 200 and 300 would cause Fairbanks and Juneau to overwhelm the rural data.
Race: separating Eskimo, Aleut, Athabaskan, etc is often the only way we have of using PUMS data to look at problems in one or another area of rural Alaska. Western Alaska has a very different economy and culture than northern or southeast Alaska. These racial designations allow us to look at how differences in these regional cultures (differences that often persist in urban dwellers) affect policy questions.

Aileen Duldulao, Other Nonacademic (Graduate) return to top
I'm involved in projects that document different characteristics of the Asian and Pacific Islander community in Los Angeles. In Los Angeles County alone there are over 25 distinct Asian and Pacific Islander identities with over 40 languages spoken. In my research, which focuses on domestic violence, I already find a dearth of quantitative research that uses the category of "Asian and Pacific Islander" and instead lumps this community into "Other." Even when research or data does separate Asians and Pacific Islanders into a distinct category, this does not capture the diversity of this racial group. Wide variances exist within this community between ethnic groups. For example, Japanese have the highest per capita income of all ethnic groups in Los Angeles county, while Cambodians and Laotians have the lowest per capita income in the county. To group these two groups, who have drastically different languages, histories, and socioeconomic statuses, in one monolithic group is a disservice to these communities, particularly communities in need of services. Greater detail in census data would allow social service providers like myself to make more informed decisions regarding what communities are in greatest need for our services. Without this detail, especially breakdowns in race and geographic location, many communities and families will go without services they so desperately need.

James Kinard, Government (Post-doc) return to top
In social policy determination, critique, analysis, etc. details are extremely important for gaining insight into the quantitative and qualitative aspects of populations with different characteristics, needs, and so forth. The "lumpen" achieved during data analysis almost invariably hides important population groups. Less detailed data hide such populations even further. Perhaps more important is the fact that the unavailability of less detailed data will leave researchers, policy makers, advocates, etc. without the data needed to verify general conclusions and/or assumptions. Many populations would be excluded from the benefits of policy-making processes because they are just not accounted for.
For instance; there is a serious problem in service-delivery systems because people with multiple disabilities practically have no where to go. Persons with multiple disabilities are forced into disability-specific programs where they don't get adequate treatments. If they have vision disabilities, they must go into programs designed for the blind; if they have emotional disabilities, they must go mental health programs designed for the mentally ill; if they have mobility disabilities, they must go to programs designed for people with mobility disabilities; and son on. This population is forced to identify such resources on their own and try to determine which ones are more appropriate for their different disabilities. They receive no professional help in this time-consuming and frustrating process.
The patient wastes invaluable time needed to receive the services that could optimize his/her rehabilitation; while programs like Medicaid, Vocational Rehabilitation, and others, make inefficient use of their resources by haphazardly providing services to this population, or by not addressing all their needs properly.
More detailed data is needed to know about those populations that suffer paralysis and blindness, emotional problems, and other disabilities. The same applies to the poor that need shelter AND medical help AND employment AND food AND economic aid, etc.
In conclusion, more detailed information is needed to know what is going out there in the real world and to give analysts what they need to design efficient and humane programs.

Hans D. Laux, Geography Department, University of Bonn, Germany, Geography (Faculty) return to top
My research as a social and population geographer is mainly concentrated on the aspects of spatial and social integration of racial and ethnic minorities in US American cities. Detailed information about ethnic and racial categories as well as about demographic and soio-economic characteristics is indispensable. Moreover, I need information on a very detailed geographic level. And last but not least, intercensal comparability is requested.

Confidential Response return to top
Most of my work focuses on inner-city, economic development. Statistics on age and income are critical to assessing the community dynamics in a given neighborhood. In addition, these data are necessary to identify "qualifying recipients" for various federal programs that fund inner-city economic development, especially HUD's Community Development Block Grant program -- funded agencies need to show that the majority of their clientele is "low or moderate income" (a number which is geographically specific . . . something like 75% of median household income for low income, 50% for moderate income). Geography is also critical as regions (NE, South, Mid-West, West, etc) vary dramatically in their income and age statistics, but even within these regions there is great variance in statistics (e.g., Atlanta, Charlotte, and Memphis could have very different age populations). Thus, the smaller the region for PUMS data, the more specific and appropriate the analysis is from a policy perspective.

Confidential Response return to top
I study the determination of wages and the impact of discrimination on wages and occupational attainment. Reducing the detail in the age, occupation, income, race, and geographic variables (and in the ancestry variables to some extent) would greatly reduce the precision of studies of this kind. It would make it very difficult to compare findings across time, which is essential in analyzing the impact of policy and the need for further policy development.

Confidential Response return to top
I am particularly concerned about any reduction in geographic detail as well as grouping of any other data. Greater geographic detail is vital to understanding the spatial patterns and behavioral motivations of everyday human activity. My own research looks at intermetropolitan migration by race, gender and the influence of labor market dynamics. You lose so much detailed information of market structures at the state level - labor markets operate at the metropolitan scale. Distinguishing by race, gender and age are paramount since each sub-category displays unique patterns and motivations. Loss of data detail would be a step backwards in the sort of research I conduct.

Richard Ogburn, South Florida Regional Planning Council, Government (Nonacademic) return to top
The extraordinary age, race and ethnic diversity in South Florida, and the pace of change, make it critical to be able to carry out socio-economic analysis at the greatest level of subject and geographic detail consistent with the preservation of confidentiality. Much analysis depends on PUMS data because published tabulations do not cover the full range of need.

Marianne Hill, Mississippi Center for Policy Research, Economics (Nonacademic) return to top
Maybe 1/3 of women on welfare here in MS live with someone else, based on PUMS data: approximated as females under 40 with own child present not head of household and receiving public assistance. I also look at occupational segregation by race and sex--black and white female childcare workers are one concern, but generally want to know if the Duncan index has improved or not. Need detail on ages of children under 18, and each year from 1 to 6 for childcare. Also projecting occupational demand, detail on occupations is critical, though I welcome the replacement of some old occupations with new.

Carolyn Lauer, Michigan Information Center/DMB, Government return to top
PUMS, as they have been released in the past, have not provided the level of detail we require to provide service to our clients

Marilyn Cavell, Virginia Tech, Other Academic (Faculty) return to top
We use PUMS data to create more detailed demographic estimates (generally at the county level) than are available through STF tables.

Alicia Sasser, Harvard University, Economics (Graduate) return to top
As a labor economist, my central focus is analyzing data on income. However, in order to analyze issues such as the gender gap one needs to be able to control for age, race, ethnicity, occupation, geographic region in order to determine whether income differences stem from differences in the characteristics of men and women in the labor force or from differences in the returns that men and women earn for having these characteristics - being a certain age, race or ethnicity, working in certain occupations or regions. Without such detail one cannot know whether gender differences in income could be alleviated by encouraging women to enter certain occupations or by continuing a policy of affirmative action.

George Hough, Portland State University, Demography (Faculty) return to top
With the new Disability Questions, age is of utmost importance. I utilize this data to provide information necessary for policy planning for a number of aging and disability service agencies.

Deborah Garvey, Santa Clara University, Economics (Faculty) return to top
The PUMS is the only nationally representative database for demographic and economic policy research on immigrant incorporation and assimilation into the United States, one of my current research areas. The PUMS is the single most important data source for my work. Significant reduction in detail would render it much less useful, and possibly completely useless, for my work.
Loss of geographic detail would make it impossible to provide estimates of short-run and lifecycle fiscal impacts of immigrants on state and local governments. It would be more difficult to track migration patterns and identify the jurisdiction where an individual lives. Loss of income detail beyond the current topcoding would also make fiscal impact estimates less precise and potentially very misleading, due to an inability to calculate tax remittances and government benefit receipt with reasonable precision. Age groupings beyond the current classification will make it difficult to track changes in economic behavior (e.g., labor force participation or government transfer receipt) over the life course.
Of course, all these problems would be compounded with the loss of comparability to previous PUMS data sets.

Glenn Fuguitt, University of Wisconsin-Madison, Sociology (Faculty) return to top
I make a lot of comparisons between metropolitan and nonmetropolitan areas for smaller subregions of the country. Its important to me that these distinctions be maintained. Also, I consider elderly migration, and wish to be able to divide the 65 and over category into at least two groups.

Philip Wuest, Clark County, Washington, Government (Graduate) return to top
I deal with transportation and development models. Location and specific characteristics associated with that location are everything. Without location and detailed information assigned to that location, it is extremely difficult to build useful and meaningful public policy analysis models/tools.

Sue Wong, National Economic Development & Law Center, Other Nonacademic (Nonacademic) return to top
We are a national intermediary that works with local nonprofits to develop workforce development strategies to link low-income residents to high growth, high wage occupations. We need specific workforce profiles for specific targeted occupations and to demographic information of our target population (e.g. low-income youth) to inform the development of an effective workforce strategy. Factors such as age, race, occupations, and geography and their current breakdowns (1990) allow us to compare and contrast the target population's demographics with the existing workforce to determine how far apart or similar these two labor forces are which, in turn, will determine if the employment linkage strategy should target the occupation in question. Furthermore, the information will be used to develop the specific workforce strategies to include components such as basic remediation, technical skills training, diversity issues and/or transportation modes, for example.

Janette Kawachi, University of California, Santa Barbara, Sociology (Graduate) return to top
My research focuses generally upon issues of labor and economy in both a domestic and international context. I focus particularly upon issues dealing with inequality and urban poverty. Much of my work would be impeded if they were to reduce data collection on variables like industry, occupation, income, wage, race, age, sex, housing, education, socioeconomic status, geography, etc. I depend upon IPUMS data as a reliable and comprehensive source of information since census aggregation techniques for city or state level data is extremely rough and untrustworthy.

Diane Muehl, University of Illinois at Urbana-Champaign, Sociology (Graduate) return to top
I am doing research in the area of older immigrants and need maximal details of age, income, race, ancestry, etc. As my teenager just told me when I explained your questionnaire to him: BAD!! BAD!! BAD!! I concur.

Lori Hunter, University of Colorado at Boulder (as of summer, 2000), Sociology (Faculty) return to top
My research entails linkages between geographic context and demographic processes, therefore the loss of geographic detail in the PUMS would render these data unsuitable for this type of research.

Judy Olson, Michigan State University, Geography (Faculty) return to top
I have been working with a student interested in ancestry, the interest having developed as a result of some work in another of my classes in which less detailed data were used for mapping. It is early to say where our efforts will go upon completion of the current project.

Stephanie Kellner, National Economic Development and Law Center, Other Nonacademic (Nonacademic) return to top
My organization works to revitalize low-income communities. We use PUMS data to develop a demographic profile of specific communities. We require a fairly small level of geographic detail (i.e. counties and smaller) so we use PUMA data. If PUMS data is grouped in larger geographic areas, the data becomes much less useful. Also, as part of our work, we develop occupational profiles of the workforce in our target area (for example, if we are looking at nursing assistants, we develop crosstabs to show the median age of nursing assistants in our target area, race breakdown, the number of years of schooling, how they get to work, etc.).

John Foster-Bey, The Urban Institute, Other Nonacademic (Nonacademic) return to top
I do a good deal of research examining the differences across metropolitan areas by race, age, gender, industry and occupation.

Confidential Response return to top
Reducing the number and quality of variables available would destroy the possibility of testing previous researchers' work to determine if their results are still valid for more current data.

Confidential Response return to top
I need to identify children age 2 years in my work. I need counts of children by various characteristics for many small geographic areas.

William Mead, Gloucester County New Jersey Planning Division, Demography (Nonacademic) return to top
Primarily in the allocation of financial resources necessary for support services affecting TANF recipients, i.e., Child Care Planning. It would also affect analyses regarding the distribution of resources to meet labor force maximization objectives within age and income categories.

Kenneth Mitchell, University of Minnesota, History (Graduate) return to top
I would be lost without PUMS data (hopelessly lost...)

Jeremy Atack, Vanderbilt University, Economics (Faculty) return to top
My work is primarily historical and I think that the Census needs also to consider the permanent long-term impact of these decisions on research when issues of confidentiality have become irrelevant. This decision has long-term consequences.

Kate Kelly, University of Central Arkansas, Other Academic return to top
My research will be dealing with children, those of day care age, and those in each grade level age. To be most useful, it is necessary to know how many children are in what grade or if they are in day care or care for at home.
I also intend to work with the aging demographics not only for today, but also for 30 years from now when the baby boomer generation begins to die.