Working With Missing Values

Less than optimum strategies for missing values can produce biased estimates, distorted statistical power, and invalid conclusions. After reviewing traditional approaches (listwise, pairwise, and mean substitution), selected alternatives are covered including single imputation, multiple imputation, and full information maximum likelihood estimation. The effects of missing values are illustrated for a linear model, and a series of recommendations is provided. When missing values cannot be avoided, multiple imputation and full information methods offer substantial improvements over traditional approaches. Selected results using SPSS, NORM, Stata (mvis/micombine), and Mplus are included as is a table of available software and an appendix with examples of programs for Stata and Mplus.     

Toward Best Practices in Analyzing Datasets with Missing Data: Comparisons and Recommendations

Although several methods have been developed to allow for the analysis of data in the presence of missing values, no clear guide exists to help family researchers in choosing among the many options and procedures available. We delineate these options and examine the sensitivity of the findings in a regression model estimated in three random samples from the National Survey of Families and Households (n = 250–2,000). These results, combined with findings from simulation studies, are used to guide answers to a set of 10 common questions asked by researchers when selecting a missing data approach. Modern missing data techniques were found to perform better than traditional ones, but differences between the types of modern approaches had minor effects on the estimates and substantive conclusions. Our findings suggest that the researcher has considerable flexibility in selecting among modern options for handling missing data.     

DOES DELEGATION INCREASE WORKER TRAINING?

We model a principal‐firm offering training to its agent‐worker under two alternative organizational structures: integration, where the principal retains authority to overrule the investment project recommended by the worker; and delegation, where the principal cannot overrule the worker's preferred investment project. We assume that training reduces the worker's effort cost of assembling information about alternative projects' payoffs and identify the conditions under which delegation increases the profit‐maximizing intensity of training. Empirical estimates from matched employer–employee data show that workplaces delegating authority do provide more worker training. This result persists in two cross sections, in panel fixed‐effect estimates and, critically, in an instrumental variable exercise that also controls for establishment fixed effects. (JEL D21, D22, D23, M53, M54)     

The economic significance of laws relating to employment protection and different forms of employment: Analysis of a panel of 117 countries, 1990–2013

The authors use time series econometric analysis applying non‐stationary panel data methods to estimate the relationships between employment protection legislation and legal protection of different forms of employment (part‐time, fixed‐term and agency work), and economic outcomes, with a data set based on the Centre for Business Research Labour Regulation Index (CBR–LRI), covering 117 countries from 1970 to 2013. Findings suggest that these laws have become significantly more protective over time and that strengthening worker protection is associated with an increase in labour's share of national income, rising labour force participation, rising employment, and falling unemployment, although the observed magnitudes are small when set against wider economic trends.     

MALE MARITAL WAGE DIFFERENTIALS: TRAINING,PERSONAL CHARACTERISTICS,AND FIXED EFFECTS

Using the National Longitudinal Survey of Youth 1979, we replicate previous estimates of the marital wage differential for white men, extend the analysis to African American men, then explain the within and between race differentials. We first control for formal job training, then for cognitive skills, parental background, and self‐esteem with little effect. By contrast, the white differential but not the black differential disappears in fixed‐effects estimation. We reconcile the cross‐section/panel differentials by focusing on the distinct identification conditions employed by each technique. Men who never change marital status play a significant role in white cross‐sectional estimates. (JEL J31, J12)     

Welfare Reform and Teenage Pregnancy,Childbirth, and School Dropout

This study estimates the effect of welfare reform on adolescent behaviors using a difference‐in‐differences approach. After defining the prereform and reform cohorts and considering the life course development of adolescent behavior by following each cohort from age 14 to age 16, we compare the welfare‐target and nontarget populations in the two cohorts. The difference‐in‐differences estimates are obtained using an event history model. Our analysis suggests that welfare reform has not reduced teenage fertility and school dropout. We find modest evidence that welfare reform is associated with higher risk of teenage births for girls in welfare families and higher risk of school dropout for girls in poor families. A combination of a difference‐in‐differences approach and a life course perspective can be a useful way to delineate the effect of societal‐level change on family phenomena.     

Methods for Handling Missing Secondary Respondent Data

Secondary respondent data are underutilized because researchers avoid using these data in the presence of substantial missing data. The authors reviewed, evaluated, and tested solutions to this problem. Five strategies of dealing with missing partner data were reviewed: (a) complete case analysis, (b) inverse probability weighting, (c) correction with a Heckman selection model, (d) maximum likelihood estimation, and (e) multiple imputation. Two approaches were used to evaluate the performance of these methods. First, the authors used data from the National Survey of Fertility Barriers (n = 1,666) to estimate a model predicting marital quality based on characteristics of women and their husbands. Second, they conducted a simulation testing the 5 methods and compared the results to estimates where the true value was known. They found that the maximum likelihood and multiple imputation methods were advantageous because they allow researchers to utilize all of the available information as well as produce less biased and more efficient estimates.     

Incarceration and the Formation and Stability of Marital Unions

Rising imprisonment rates and declining marriage rates among low‐education African Americans motivate an analysis of the effects of incarceration on marriage. An event history analysis of 2,041 unmarried men from the National Longitudinal Survey of Youth suggests that men are unlikely to marry in the years they serve in prison. A separate analysis of 2,762 married men shows that incarceration during marriage significantly increases the risk of divorce or separation. We simulate aggregate marriage rates using estimates from the National Longitudinal Survey of Youth and find that the prevalence of marriage would change little if incarceration rates were reduced.     

REGRESSION WITH MISSING YS: AN IMPROVED STRATEGY FOR ANALYZING MULTIPLY IMPUTED DATA

When fitting a generalized linear model—such as linear regression, logistic regression, or hierarchical linear modeling—analysts often wonder how to handle missing values of the dependent variable Y . If missing values have been filled in using multiple imputation, the usual advice is to use the imputed Y values in analysis. We show, however, that using imputed Y s can add needless noise to the estimates. Better estimates can usually be obtained using a modified strategy that we call multiple imputation, then deletion (MID). Under MID, all cases are used for imputation but, following imputation, cases with imputed Y values are excluded from the analysis. When there is something wrong with the imputed Y values, MID protects the estimates from the problematic imputations. And when the imputed Y values are acceptable, MID usually offers somewhat more efficient estimates than an ordinary MI strategy.     

Women's Attitudes Toward Family Formation and Life Stage Transitions: A Longitudinal Study in Korea

Although it is well documented that family attitudes become less traditional over cohorts, little is known about how individuals' attitudes change over time. More research also is needed on how the within‐individual changes relate to important life stage events such as marriage, childbirth, and transitions in education and work. Evidence is particularly lacking in Asian countries, which have socioeconomic and cultural contexts very different from those in the West. To fill these gaps in the literature, the authors analyzed the attitudes toward family formation of Korean women in their 20s and 30s (N = 6,042). Individual fixed effects regression using the panel data from the Korean Longitudinal Survey of Women and Families revealed that women became more traditional over time and that transitions to marriage and motherhood partly accounted for the change. These findings are explained within the context of very low fertility in Korea and have implications for other rapidly changing societies.     

Marginal Structural Models: An Application to Incarceration and Marriage During Young Adulthood

Advanced methods for panel data analysis are commonly used in research on family life and relationships, but the fundamental issue of simultaneous time‐dependent confounding and mediation has received little attention. In this article the authors introduce inverse‐probability‐weighted estimation of marginal structural models, an approach to causal analysis that (unlike conventional regression modeling) appropriately adjusts for confounding variables on the causal pathway linking the treatment with the outcome. They discuss the need for marginal structural models in social science research and describe their estimation in detail. Substantively, the authors contribute to the ongoing debate on the effects of incarceration on marriage by applying a marginal structural model approach to panel data from the National Longitudinal Survey of Youth 1997 (N = 4,781). In line with the increasing evidence on the collateral consequences of contact with the criminal justice system, the authors find that incarceration is associated with reduced chances of entering marriage.     

Moving Out: Transition to Nonresidence Among Resident Fathers in the United States, 1968–1997

This article provides the first individual‐level estimates of the change over time in the probability of nonresidence for initially resident fathers in the United States. Drawing on the 1968–1997 waves of the Panel Study of Income Dynamics, we used discrete‐time event history models to compute the probabilities of nonresidence for six 5‐year periods. Our sample consists of men (N = 1,388) who are coresident with their biological children at the time of birth. We found that the observed probability of nonresidence doubled over the three decades of the study period, but not linearly. The risk increased substantially in the 1980s and then stabilized in the 1990s. Our multivariate models show that the stabilization was due to changes over time in characteristics such as income; had these remained constant, the likelihood of nonresidence would have increased throughout the study period. Both fathers’ and mothers’ incomes reduce the likelihood of paternal nonresidence, as do mothers’ employment hours.     

Solidarity and Conflict Between Adult Children and Parents: A Latent Class Analysis

Using multiple dimensions of solidarity and conflict in a latent class analysis, we develop a typology of adult child–parent relationships. The data (N= 4,990) are from the first wave of the Netherlands Kinship Panel Study. In descending order of relationship quality, the 5 types are harmonious (akin to relationships with friends), ambivalent (intensive exchange of material support accompanied by strain), obligatory (just keeping in touch), affective (emotionally supportive with few other meaningful exchanges), and discordant (predominantly negative engagement). The types are differentiated by gender, age, family size, geographic distance, and parental marital history, indicating that they are not fixed but are shaped by social‐structural conditions.     

Network sampling coverage II: The effect of non-random missing data on network measurement

Missing data is an important, but often ignored, aspect of a network study. Measurement validity is affected by missing data, but the level of bias can be difficult to gauge. Here, we describe the effect of missing data on network measurement across widely different circumstances. In Part I of this study (Smith and Moody, 2013), we explored the effect of measurement bias due to randomly missing nodes. Here, we drop the assumption that data are missing at random: what happens to estimates of key network statistics when central nodes are more/less likely to be missing? We answer this question using a wide range of empirical networks and network measures. We find that bias is worse when more central nodes are missing. With respect to network measures, Bonacich centrality is highly sensitive to the loss of central nodes, while closeness centrality is not; distance and bicomponent size are more affected than triad summary measures and behavioral homophily is more robust than degree-homophily. With respect to types of networks, larger, directed networks tend to be more robust, but the relation is weak. We end the paper with a practical application, showing how researchers can use our results (translated into a publically available java application) to gauge the bias in their own data.     

Latinas’ Transition to First Marriage: An Examination of Four Theoretical Perspectives

National Longitudinal Survey of Youth and census data are used to examine the effect of both individual‐ and contextual‐level determinants on Latinas’ transition to first marriage (n= 745). Hypotheses derived from 4 leading theories of marriage timing are evaluated. Discrete‐time event‐history models that control for clustering within Labor Market Areas suggest that foreign‐born Latina and Anglo women have virtually identical marriage trajectories. Analyses further demonstrate that Latinas’ individual human capital, and residence in areas characterized by a relatively large supply of single foreign‐born Latino men, are associated with higher probabilities of marriage, whereas women’s aggregate economic opportunities are correlated with the predicted postponement of first marriage.     

YOUTH DEPRESSION AND FUTURE CRIMINAL BEHAVIOR

While the contemporaneous association between mental health problems and criminal behavior has been explored in the literature, the long‐term consequences of such problems, depression in particular, have received much less attention. Using data from the National Longitudinal Study of Adolescent Health (Add Health), we examine the effect of depression during adolescence on the probability of engaging in a number of criminal behaviors later in life. In our analysis, we control for a rich set of individual‐, family‐, and neighborhood‐level factors to account for conditions that may be correlated with both childhood depression and adult criminality. One novelty in our approach is the estimation of school and sibling fixed effects models to account for unobserved heterogeneity at the neighborhood and family levels. Furthermore, we exploit the longitudinal nature of our data set to account for baseline differences in criminal behavior. The empirical estimates show that adolescents who suffer from depression face a substantially increased probability of engaging in property crime. We find little evidence that adolescent depression predicts the likelihood of engaging in violent crime or the selling of illicit drugs. Our estimates imply that the lower‐bound economic cost of property crime associated with adolescent depression is approximately 227 million dollars per year. (JEL I10, K42)     

Handling missing data in numeric analyses

ABSTRACT

Social science datasets usually have missing cases, and missing values. All such missing data has the potential to bias future research findings. However, many research reports ignore the issue of missing data, only consider some aspects of it, or do not report how it is handled. This paper rehearses the damage caused by missing data. The paper then briefly considers eight different approaches to handling missing data so as to minimise that damage, their underlying assumptions and the likely costs and benefits. These approaches include complete case analysis, complete variable analysis, single imputation, multiple imputation, maximum likelihood estimation, default replacement values, weighting, and sensitivity analyses. Using only complete cases should be avoided wherever possible. The paper suggests that the more complex, modelling approaches to replacing missing data are based on questionable methodological and philosophical assumptions. And they may anyway not have clear advantages over simpler approaches like default replacements. It makes sense to report all possible forms of missing data, report everything that is known about the characteristics of cases missing values, conduct simple sensitivity analyses of the potential impact of missing data on the substantive results, and retain the knowledge of missingness when using any form of replacement value.     

Analytic approaches for assessing long-term treatment effects. Examples of empirical applications and findings

Analytic approaches, including the structural equation model (autoregressive panel model), hierarchical linear model, latent growth curve model, survival/event history analysis, latent transition model, and time-series analysis (interrupted time series, multivariate time-series analysis) are discussed for their applicability to data of different structures and their utility in evaluating temporal effects of treatment. Methods are illustrated by presenting applications of the various approaches in previous studies examining temporal patterns of treatment effects. Recent advancements in these longitudinal modeling approaches and the accompanying computer software development offer tremendous flexibility in examining long-term treatment effects through longitudinal data with varying numbers and intervals of assessment and types of measures. A multimethod assessment will contribute to a more complete understanding of the complex phenomena of the long-term courses of substance use and its treatment.     

INFLATION DYNAMICS AND THE OUTPUT‐INFLATION TRADE‐OFF: INTERNATIONAL PANEL DATA EVIDENCE

We explore the impact of inflation and its variability on the output‐inflation trade‐off using a single‐step approach in a panel data context. Previous empirical approaches focus on either cross‐country or country‐by‐country time‐series analyses. We employ a dynamic heterogeneous panel specification using an all‐encompassing estimation framework accounting for parameter heterogeneity, cross‐sectional dependence, dynamics, and nonstationarity. Our sample covers 60 countries from 1970 to 2010. While inflation variability reduces the trade‐off for specific periods and country groups, an unambiguous and more pronounced negative relation emerges between the inflation rate and the trade‐off. The findings are consistent with the New Keynesian view. (JEL E31, E12, C23)