Drawing statistical inferences from historical census data, 1850–1950

Virtually all quantitative microdata used by social scientists derive from samples that incorporate clustering, strati cation, and weighting adjustments (Kish 1965, 1992). Such data can yield standard error estimates that differ dramatically from those derived from a simple random sample of the same size. Researchers using historical U.S. census microdata, however, usually apply methods designed for simple random samples. The resulting p values and confidence intervals could be inaccurate and could lead to erroneous research conclusions. Because U.S. census microdata samples are among the most widely used sources for social science and policy research, the need for reliable standard error estimation is critical. We evaluate the historical microdata samples of the Integrated Public Use Microdata Series (IPUMS) project from 1850 to 1950 in order to determine (1) the impact of sample design on standard error estimates, and (2) how to apply modern standard error estimation software to historical census samples. We exploit a unique new data source from the 1880 census to validate our methods for standard error estimation, and then we apply this approach to the 1850–1870 and 1900–1950 decennial censuses. We conclude that Taylor series estimation can be used effectively with the historical decennial census microdata samples and should be applied in research analyses that have the potential for substantial clustering effects.     

America’s Churning Races: Race and Ethnicity Response Changes Between Census 2000 and the 2010 Census

A person’s racial or ethnic self-identification can change over time and across contexts, which is a component of population change not usually considered in studies that use race and ethnicity as variables. To facilitate incorporation of this aspect of population change, we show patterns and directions of individual-level race and Hispanic response change throughout the United States and among all federally recognized race/ethnic groups. We use internal U.S. Census Bureau data from the 2000 and 2010 censuses in which responses have been linked at the individual level (N = 162 million). Approximately 9.8 million people (6.1 %) in our data have a different race and/or Hispanic-origin response in 2010 than they did in 2000. Race response change was especially common among those reported as American Indian, Alaska Native, Native Hawaiian, Other Pacific Islander, in a multiple-race response group, or Hispanic. People reported as non-Hispanic white, black, or Asian in 2000 usually had the same response in 2010 (3 %, 6 %, and 9 % of responses changed, respectively). Hispanic/non-Hispanic ethnicity responses were also usually consistent (13 % and 1 %, respectively, changed). We found a variety of response change patterns, which we detail. In many race/Hispanic response groups, we see population churn in the form of large countervailing flows of response changes that are hidden in cross-sectional data. We find that response changes happen across ages, sexes, regions, and response modes, with interesting variation across racial/ethnic categories. Researchers should address the implications of race and Hispanic-origin response change when designing analyses and interpreting results.     

An evaluation of bridging methods using race data from Census 2000

The question on race from Census 2000 was different from previous censuses because it allowed respondents to select one or more races to indicate their racial identities. Because of this change, the race data from Census 2000 are not directly comparable with data from earlier censuses. Researchers can use `bridging' methods to assign more than one race respondents to single race categories to maximize the comparability of Census 2000 race data with earlier censuses. This paper uses several bridging methods to generate race population estimates and analyzes the variability in those estimates across six single race groups.     

Identifying race and ethnicity in the 1979 National Longitudinal Survey of Youth

The 1979 National Longitudinal Survey of Youth is among the few surveys to provide multiple reports on respondents’ race and ethnicity. Respondents were initially classified as Hispanic, black, or “other” on the basis of data collected during 1978 screener interviews. Respondents subsequently self-reported their “origin or descent” in 1979, and their race and Hispanic origin in 2002; the latter questions conform to the federal standards adopted in 1997 and used in the 2000 census. We use these data to (a) assess the size and nature of the multiracial population, (b) measure the degree of consistency among these alternative race-related variables, and (c) devise a number of alternative race/ethnicity taxonomies and determine which does the best job of explaining variation in log-wages. A key finding is that the explanatory power of race and ethnicity variables improves considerably when we cross-classify respondents by race and Hispanic origin. Little information is lost when multiracial respondents are assigned to one of their reported race categories because they make up only 1.3% of the sample.

Changes in the earnings of Arab men in the US between 2000 and 2002

Using public-use microdata samples from the American Community Survey, we find that Middle Eastern Arab men and Afghan, Iranian, and Pakistani men experienced a significant earnings decline relative to non-Hispanic whites between 2000 and 2002. Further analyses based on the Juhn–Murphy–Pierce wage decomposition technique as well as quantile regression indicate that this earnings decline is not explained by changes in the structure of wages or in observable characteristics beyond ethnicity. Our interpretation is that the unanticipated events of September 11th, 2001 negatively affected the labor-market income of the groups most closely associated with the ethnicity of the terrorists.

Childbearing within Marriage and Consensual Union in Latin America, 1980–2010

This article compares the fertility patterns of women in consensual union and marriage in 13 Latin American countries, using census microdata from the four most recent census rounds and a methodological approach that combines the own‐children method and Poisson regression. Results show that in all these countries, fertility is slightly higher within consensual union than marriage and that the age pattern of fertility is very similar in marital and non‐marital unions. Further analyses show that over the period considered, childbearing within a consensual union has changed from rare to increasingly common, although not yet mainstream, for highly educated women in most countries examined. Results show that in Latin America, at least since the 1980s, women's childbearing patterns depend on their age and on their being in a conjugal relationship, but not on the legal nature of this relationship. The similarities in reproductive behavior between marital and non‐marital unions are not confined to the socially disadvantaged groups, but apply as well to the better off.     

Pictorial Approaches for Measuring Time Use in Rural Ethiopia

Time use researchers working in least developed countries (LDCs) face difficulties collecting data from illiterate populations who may conceptualize time differently than those in industrialized countries. We identify existing gaps in time use data collection methods and discuss two novel, pictorial methods to collect time use data from these populations. The first method is a modified recall interview modeled on participatory rural appraisal (PRA) methods that asks respondents to place macaroni on pictures of activity categories in proportion to the amount of time spent on that activity during the previous day. The second is a simplified pictorial time diary that uses a timer and sequentially-numbered stickers to re-create the temporal order of activities in 30-min increments. The latter method also avoids recall bias problems. We present time use data collected in 2009 using these methods in a study examining the impacts of water infrastructure on women and children’s time use in rural Ethiopia. In total, we collected information using the first method from 263 household members over age 10, including 167 water collectors, and pilot-tested the pictorial diary approach with 10 adult respondents.     

Big Microdata for Population Research

This article describes an explosion in the availability of individual-level population data. By 2018, demographic researchers will have access to over 2 billion records of accessible microdata from over 100 countries, dating from 1703 to the present. Another 2 to 4 billion records will be available through restricted-access data enclaves. These new resources represent a new kind of data that will enable transformative research on demographic and economic change and the spatial organization of society.     

Measuring social indicators in opinion surveys: A method to improve accuracy on sensitive questions

Asking sensitive questions, without risking a terminated interview or response bias, is a major problem in deriving accurate social indicators based on public opinion surveys. This problem has become particularly acute as the topics that interest researchers have become more personal in nature. Mail and telephone surveys, and methods such as the randomized response technique, have all been used to try and overcome this problem, with varying degrees of success. In this paper, we describe an alternative approach using a sealed booklet. We report results from a question-format experiment that asked respondents the same sensitive questions in a sealed booklet, completed in the presence of the interviewer, and in a standard face-to-face interview. The survey used for the experiment was a personal interview survey of drug use based on a national population sample. The sealed booklet format was found to produce more accurate estimates of drug use compared to direct questions. In addition to assuring the respondent greater anonymity, the sealed booklet permits a wide range of questions to be asked and does not limit the analyses that can be conducted on the data.     

The Effects of Allostatic Load on Racial/Ethnic Mortality Differences in the United States

This study expands on previous findings of racial/ethnic and allostatic load (AL) associations with mortality by addressing whether differential AL levels by race/ethnicity may explain all-cause mortality differences. This study used data from the third National Health and Nutrition Survey public-use file, gathered between 1988 and 1994, with up to 18 years of mortality follow-up (n = 11,733). AL scores were calculated using a 10-biomarker algorithm based on clinically determined thresholds. Results of discrete-time hazard models suggest that AL is associated with increased mortality risks, independent of other factors, including race/ethnicity and SES. The results also suggest that the AL–mortality association is stronger for non-Hispanic blacks than for non-Hispanic whites, and that at low levels of AL observed mortality differences between non-Hispanic blacks and non-Hispanic whites are non-significant. These findings suggest that mortality differences between non-Hispanic blacks and non-Hispanic whites may be the result of how early life exposure causes premature aging and increased mortality risks. More attention to resource allocation and local environments is needed to understand why non-Hispanic blacks experience premature aging that leads to differential mortality risks compared to non-Hispanic whites.     

Union Formation Implications of Race and Gender Gaps in Educational Attainment: The Case of Latin America

We use census microdata to assess the levels of educational homogamy in six Latin American countries: Argentina, Brazil, Chile, Costa Rica, Ecuador and Mexico. This paper contributes to the literature on homogamy in three ways. First, by conducting a comparative analysis between countries belonging to the still little-studied region of Latin America, which is still undergoing intense and varied processes of demographic, economic, social, and political modernization. Second, by simultaneously including variables of structural and individual nature. Finally, by making progress with respect to the interactions between educational homogamy and other important variables associated with high levels of social inequality in the region: race, ethnicity and birthplace.     

Labour Underutilisation and Gender: Unemployment Versus Hidden-Unemployment

As labour markets have become more complex there has been increasing interest among researchers in understanding the ways that social and labour market processes and contexts impact on various labour market states. One important area has been in understanding the differences between unemployment and hidden unemployment. This paper considers the ways in which these two labour market states differ for a sample of male and female respondents to the Household Income and Labour Dynamics Australia (HILDA) survey. It presents data related to the reasons why respondents in these two labour force states consider they are jobless and analyses the characteristics of male and female respondents in the two labour market states to consider differences in outcomes. The findings suggest that there are differences in the two states of labour market outcomes and that these are further complicated when one considers processes for males and females.     

A model with long-term survivors for the analysis of current-status nuptiality data

The model proposed in this paper combines a logistic regression model and a Weibull regression model for the analysis of current-status data. This joint model allows a simultaneous estimation of two sets of effects on the covariates: one on the probability that the event occurs (also known as quantum) and the other on the timing of the event. Thus, the model can be seen either as an extension of a survival-analysis model for use with current-status data where a survival fraction is added in order explicitly to take into account the possibility that the event may never occur, or as an extension of survival analysis with long-term survivors to the analysis of current-status data. As an illustrative application we apply our model to a study of nuptiality in seventeenth-century Italy.     

A model with long-term survivors for the analysis of current-status nuptiality data

The model proposed in this paper combines a logistic regression model and a Weibull regression model for the analysis of current-status data. This joint model allows a simultaneous estimation of two sets of effects on the covariates: one on the probability that the event occurs (also known as quantum) and the other on the timing of the event. Thus, the model can be seen either as an extension of a survival-analysis model for use with current-status data where a survival fraction is added in order explicitly to take into account the possibility that the event may never occur, or as an extension of survival analysis with long-term survivors to the analysis of current-status data. As an illustrative application we apply our model to a study of nuptiality in seventeenth-century Italy.     

Gender bias in reproductive behaviour in Georgia,Indonesia, and Vietnam: An application of the own-children method

I present a method for estimating indicators of gender bias in reproductive behaviour, using microdata based on the own-children method. The method is first tested on a large sample from India with both birth history and household records. I then apply the method to Georgia, Indonesia, and Vietnam. My estimates demonstrate that the proportion of Georgian couples expressing a preference for sons in their fertility behaviour closely corresponds to the proportion resorting to sex selection at high parities. I show how individual Indonesian provinces provide examples of both son and daughter preference. The method also allows me to date the onset of sex imbalances at birth in Vietnam. The approach based on the own-children method therefore provides a unique tool for estimating the diversity and intensity of gender bias in reproductive behaviour.     

Adolescents’ Sexual Initiation: The Interaction of Race/Ethnicity and Immigrant Status

This paper provides an in-depth examination of the joint effects of race/ethnicity and immigrant status on adolescents’ intercourse risk. We employ a sample of 4,535 females and 3,759 males from the National Education Longitudinal Study (NELS 88/94) who were followed for 6 years beginning in the eighth grade. We use discrete-time logistic regression models to estimate the associations of race/ethnicity and immigrant generational status with first intercourse hazard, and to evaluate the statistical interactions between race/ethnicity and immigrant status. Overall, Asian and Hispanic girls had lower and non-Hispanic Black girls had higher estimated risks relative to non-Hispanic White girls. Hispanic boys and White non-Hispanic boys had similar intercourse risks, but Black boys had higher and Asian boys lower relative risks. However, these patterns are contingent on immigrant status. Among girls, the protective effects of Asian or Hispanic identity are found only among second generation youth. Risk profiles for boys are more complex: being a third-plus generation Hispanic is associated with a higher risk while an Asian identity is associated with a lower risk only among first- and second-generation youth. These findings confirm the importance of accounting for the overlap between race/ethnicity and immigrant status in models of adolescent behavior. As the demographic diversity of the US population grows, researchers must include both race/ethnicity and immigrant status in their models of adolescent behavior.     

Mortality and fertility: How large is the direct child replacement effect in China?

Olsen (1980) proposed a method for quantifying the fertility response to child mortality. He showed how to correct for bias in the OLS estimator. He also proposed the use of mortality rates as an instrumental variable. This method is applied here to a new Chinese microdata set. It appears that the method works well. The bias-corrected direct replacement effect based on the Chinese data is about 0.6, which is three times as large as the effects found by Olsen (1980, 1988) using data for Colombia and Malaysia. Several explanations are provided for this result.I am grateful to Frank Denton, Martin Dooley, Ronald Lee, Lonnie Magee, Byron Spencer and three anonymous referees for many helpful comments and suggestions. I would also like to thank the International Statistical Institute Research Center for supplying the data used in this study.     

Frequency of Sexual Activity With Most Recent Male Partner Among Young,Internet-Using Men Who Have Sex With Men in the United States

Sex frequency, defined here as the number of oral or anal sex acts with the most recent partner in the past year, is a potential driver of risk for sexually transmitted infections. However, few data on sex frequency have been reported for men who have sex with men (MSM). Data from an Internet survey of MSM were used to describe sex frequency with most recent main and casual male partners and to estimate factors associated with higher sex frequency. Among 5,193 MSM, higher sex frequency was associated with younger age, shorter relationship duration, and reporting a main (vs. casual) partner; and lower sex frequency with male partners was associated with heterosexual or bisexual (vs. homosexual) identity or Black race (vs. non-Hispanic White). Secondary analyses of estimates of sex frequency from 2 publicly available nationally representative datasets comprised of primarily heterosexual survey respondents (the 2008 General Social Survey and the 1992 National Health and Social Life Survey) were performed. Sex frequency among MSM respondents was similar to that reported by heterosexuals.     

The Sibsize Revolution and Social Disparities in Children’s Family Contexts in the United States, 1940–2012

This article points to a sharp decline in children’s sibling numbers (sibsize) that occurred in the United States since the 1970s and was large enough among children with lower socioeconomic status (SES) (particularly black children) to amount to a revolution in their family circumstances. It interprets sibsize decline as a source of social convergence in children’s family contexts that ran counter to trends toward social divergence caused by the rise of lone parenthood. The article is based on new estimates of differences in children’s sibsize and lone parenthood by race and maternal education generated from public-use samples from the Census of Population and Current Population Survey (CPS), focusing especially on the period 1940–2012. I discuss some methodological and substantive challenges for existing scholarship arising from the findings and point to questions for future research.     

Welfare Use Among Older Hispanic Immigrants: The Effect of State and Federal Policy

This study examines the effect of state and federal welfare policy changes on the use of Supplemental Security Income (SSI) among older Hispanic immigrants in the US. The results contribute to the ongoing debate about whether the Personal Responsibility and Work Opportunity Reconciliation Act (PRWORA) of 1996 had a “chilling effect” on welfare use. Using data from the 1990 and 2000 US Censuses of Population, we use multi-level binomial logistic regression to evaluate whether state levels of welfare generosity contribute to the probability of SSI use among eligible older Hispanics. Our results show that the decline in SSI receipt between 1990 and 2000 is greater among older Hispanic non-citizens from Mexico, Central and South America as compared to their naturalized and US-born Hispanic counterparts. We also demonstrate that state differences in welfare generosity did not moderate the effect of federal policies during this time period as was expected.