« Back to Results

Earnings Dynamics in the United States: Trends and Reconciliation

Paper Session

Saturday, Jan. 5, 2019 10:15 AM - 12:15 PM

Atlanta Marriott Marquis, L507
Hosted By: Econometric Society
  • Chair: David Johnson, University of Michigan

Estimating Models of Trends in Income Volatility with the PSID: New Results and Comparisons to the Literature

Robert Moffitt
Johns Hopkins University
Sisi Zhang
Jinan University


This Abstract serves two functions: (1) An Introduction to the entire Session and a description of the motivation of the Session, and (2) An Abstract for this specific paper.

(1) Introduction to the Session. There is by now a very large literature on estimating earnings dynamics models in the U.S., most of which uses error components models of various types which decompose the variance of earnings over the life cycle into permanent and transitory components (Lillard and Willis, ECMA, 1978 is usually cited as the original). A particular focus of this literature has been whether earnings instability and volatility has increased in the U.S. over time. Gottschalk and Moffitt (BPEA, 1994) was the first paper in this literature but there has been a stream of subsequent estimates in the literature (reviewed by Moffitt and Zhang, May AER P&P, 2018). However, while most estimates have been performed with the Panel Study of Income Dynamics (PSID), estimates are also available for other data sets, including matched Current Population Survey (CPS), the Survey of Income and Program Participation (SIPP), and administrative data sets such as the Longitudinal Employer-Household Dynamics (LEHD) or Social Security earnings data.

As the review by Moffitt and Zhang showed (this paper is attached to this submission), the estimates of trends are often different across data sets. Whether estimated by what Moffitt and Zhang call “gross volatility,” which is just the variance of some measure of dispersion of the change in earnings from t to t+1, or by the variance of the transitory component in a model which has a permanent-transitory decomposition, instability in annual male earnings rose in the U.S. in the late 1970s and early 1980s, experienced a flat or declining period through the mid-2000s, and then resumed its increase. Matched CPS data show somewhat similar trends but the turning points are not the same. SIPP data show a very different trend, and administrative data on earnings show downward trends.

This is an important issue, for whether earnings instability has been going up, down, or neither is clearly something that needs to be established. The problem is that the different studies have used different models of earnings dynamics and have used different measures of instability, and have used different samples, variable definitions, and other differences. The goal of this session is to make headway on this problem by bringing together 4 papers, each using one of the major data sets, to estimate common earnings dynamics models, using similar sampling frames and exclusions. Further, assuming there are still differences in instability trends, the papers in the session will attempt to explore reasons for those differences, including possible biases of various kinds.

The four data sets are the PSID, matched CPS’s, the SIPP, and an administrative data set, the LEHD, drawn from wage records in the UI system. There will be a separate paper on each. The four authors will work jointly in a two-stage process. First, the authors will estimate common earnings dynamics models on their data sets, using as similar sample compositions and variable definitions as possible, and then exchange their results in Fall, 2018. Second, they will then work jointly, as a team, to explore different hypotheses for any differences in findings. A partial list of differences to explore will include biases due to attrition, imputation for missing data, measurement error, and biases due to selection into the particular samples. If successful, the papers in the session will jointly establish strong evidence on what the trend in earnings instability in the U.S. actually has been.

(2) This specific paper. This specific paper will contribute to this literature by providing new estimates of trends in earnings volatility with the PSID and by (1) attempting to reconcile estimates from the PSID with those of other data sets and (2) using common models used by other papers in the session with other data sets. Differences in trends in different data sets could be the result of either some underlying difference in the data or a difference in the models used. The paper will investigate both using PSID data from 1970 to 2015.

With regard to new estimates, we will provide new estimates of trends in instability and volatility of female earnings and family income, which have rarely been examined with the PSID. This will also help us make new comparisons to trends in other data sets, which often have computed trends in those variables.

With regard to the differences in the underlying data, we will conduct a wide variety of tests of estimates from the PSID which will address some of its potential deficiencies. One leading possibility is that there is attrition bias in the PSID, which by now has reached over 50 percent. Tests for selection on observables and unobservables will be conducted to determine how estimates of earnings instability are affected. Another is that methods of imputation for missing earnings--a common issue in all survey data sets, including the others in this session--may affect estimates of earnings volatility. We will examine the effect of the PSID’s methods of imputation, consider alternative methods, and will attempt to reach conclusions about whether is likely that estimates of earnings instability are affected by this issue. We will also explore other ways to test for measurement error. A third issue is that the PSID data typically provide only information on heads of household and their spouses, which makes comparability difficult with administrative data sets like the LEHD and SSA which cannot identify headship. However, The PSID has recently released a public file of earnings on other members of the family in 2005-2015, and we will use that to explore this issue in our work. Our examination of family incomes instability trends and how they compare to those in other data sets will also help in this regard, because family income includes all family members’ earnings. We will also endeaver to mimic the sampling frames of the CPS, SIPP, and LEHD papers in this session.

With regard to the use of common models, we will estimate econometric models of earnings dynamics that have been used in other data sets to determine if this could be a source of difference. The decomposition of earnings levels or changes into permanent and transitory components requires assumptions which can differ across studies. We will also extend a new model with more flexible forms and less restrictive assumptions which we introduced in Moffitt and Zhang (2018). Finally, we will work with the authors of the other papers in this session to estimate common models of earnings dynamics.

Trends in Earnings Volatility using Linked Administrative and Survey Data

Christopher Bollinger
University of Kentucky
Charles Hokayem
U.S. Census Bureau
James Ziliak
University of Kentucky


The aim of this project is to reconcile the diverging trends in earnings volatility obtained from survey and administrative records using a restricted-access dataset that links the Current Population Survey Annual Social and Economic Supplement (ASEC) to the Social Security Detailed Earnings Records (DER). Understanding the trends in earnings volatility is important because of the possibility that changes in human capital, labor supply, and public policies may have shifted more idiosyncratic and business cycle risk onto families, which could have negative welfare consequences if it falls predominantly on those who face liquidity constraints and are less able smooth income shocks. Starting with the seminal work of Gottschalk and Moffitt (1994), the focus on volatility trends centered on identifying whether rising cross-sectional income inequality stemmed from transitory instability or from permanent shocks. The preponderance of evidence on volatility was obtained using data from the Panel Study of Income Dynamics (PSID), and the general consensus from the PSID was that transitory instability increased from the early 1970s until the mid 1980s, and stabilized until 2000, and permanent (“lifetime”) instability rose primarily in the 1980s (Gottschalk and Moffitt 1994, 2009; Haider 2001; Gundersen and Ziliak 2003; Hacker and Jacobs 2008; Keys 2008; Dynan, Elmendorf, and Sichel 2012; Shin and Solon 2013; Moffitt and Zhang 2018). While there is corroborating evidence on the basic PSID trends up until 2000 from matched panels of the ASEC (Gittleman and Joyce 1996; Cameron and Tracy 1998; Ziliak, Hardy, and Bollinger 2011), and from the Survey of Income and Program Participation (Dahl, DeLeire, and Schwabish 2011; Celik, Juhn, McCue, and Thompson 2012; Carr and Wiemers 2017), there is a sharp departure from the PSID starting in the mid 2000s from the other survey datasets. More worrying is recent evidence from the DER that calls into question the basic conclusion of whether volatility increased at any point since 1980 (Sabelhaus and Song 2010; Bloom, Guvenen, Pistaferri, Salgado, and Song 2017).

We use our unique access to internal ASEC files linked to the DER to examine where, how, and for whom estimates of earnings volatility from survey reports differ from administrative reports using an exact survey-administrative link. The basic ASEC-DER link is available for the 1998-2016 survey years, permitting us to use the rotation group structure of the ASEC to construct two-year matched panels to compute basic measures of earnings volatility (log difference, arc percent change) using both survey reports and administrative reports. Importantly, this offers the first volatility estimates using matched CPS data after the Great Recession, and also the first estimates using the DER. We construct these measures separately for men and women, and by education attainment (less than high school, high school, some college, college, and graduate school). Importantly, we also have access to the DER back to 1978, which permits us to link each calendar-year ASEC-DER match to the full administrative time series to compute long-term measures of volatility used in the permanent-transitory literature. The advantage offered by the link to administrative data is that today nearly 45 percent of survey earnings responses in the ASEC are missing due to nonresponse, but the link to administrative data permits us to “fill in” the missing survey responses using actual information from the same worker as opposed to using an imputed value from a randomly selected “donor” as currently practiced by Census. Our previous research using linked ASEC-DER data has demonstrated that earnings nonresponse in the ASEC produces biased estimates of poverty and inequality (Hokayem et al. 2015; Bollinger et al. 2018), and in this study we expand our analysis to the longitudinal outcome of volatility.

A possible concern with matched ASEC is with sample attrition affecting our earnings series. The CPS sample domain is household addresses and not individuals, so that if a person moves between ASEC surveys then the Census Bureau interviews the new occupant at the address and does not follow the original respondent. Moves are more likely among low-income families whose earnings are more volatile, which means we could understate the level and trends in volatility with our sample. Under the assumption that the probability of attrition is unobserved and time invariant (i.e., a fixed effect), or trending very slowly over time, then first differencing earnings as used in the volatility measures based on log-differences will remove the latent probability of attrition and our estimates will be purged of possible attrition bias (Ziliak and Kniesner 1998; Wooldridge 2001). However, if there is time-variation in the factor loading on the unobserved individual-level heterogeneity then differencing will not eliminate potential attrition bias unless the factor loading is randomly distributed across the population. A conservative interpretation is that data from matched ASEC provides estimates of earnings volatility among the population of non-movers. Hardy, Smeeding, and Ziliak (2018) demonstrate that sample demographics between the cross-section of ASEC individuals align fairly closely with those from matched panels (panel individuals tend to be slightly older, more educated, and more likely to be white than the cross-section sample), and re-weighting the sample using inverse probability weights (IPW) had little effect on their model estimates. As their analysis was only for first moments of the distribution, we will examine the effects of possible attrition bias on second moments, again using IPW as well as exploiting the access to administrative records to examine how volatility of attriters in administrative data compares to stayers in both survey and administrative data.

For our analyses of volatility, we extend the summary measure of volatility used in Dynan, et al. (2008) and Dahl, et al. (2008) so it is robust not only to those workers transitioning in and out of the labor market but also to negative earnings commonly found among the self employed. Most of the literature measures earnings volatility in terms of the growth in log earnings, which precludes those with zero or negative earnings. However, there has been trend growth in the fraction of the labor force that is self employed, as well as growth in the fraction of men (and since 2000) women out of the labor force, and our measure captures this shifting composition. In this context, we relate our measure to others used in the literature such as the permanent and transitory decompositions in Gottschalk and Moffitt (1994) and Shin and Solon (2010), including the roles of lifecycle age adjustment, self employment, and non-employment. A simple decomposition of earnings into permanent, μ(i), and transitory, u(it) ,components
y(it)=μ(i)+u(it), serves as a departure point. The literature on income volatility seeks to identify the variation in the transitory component, u(it). While a variety of important identification issues arise, we note that survey, y(it)^C, and administrative, y(it)^A data differ in the data generating process of observed income. At its simplest one can argue that measurement error, ε(it), enters survey data but not administrative data: y(it)^C=μ(i)+u(it)+ε(it). This then would explain higher variation in survey data, as is typically observed. Similarly, survey non-response is known to be concentrated among individuals with extremely low or extremely high earnings (Bollinger et al, 2018) and it appears, among those who have high growth in earnings. Concern arises in treating administrative data as error free. If indeed, substantial under the table earnings are missing from administrative data, then a model of administrative earnings must include a measurement error component, v(it), as well: y(it)^A=μ(i)+u(it)+v(it). At its simplest, the covariance of (y(it)^C-y(it-1)^C) and (y(it)^A– y(it-1)^A) would then identify the variance of the transitory component, u(it). More complex measurement systems are often found to be supported in data (Kapteyn and Ypma, 2007). Our data improve on theirs, as our data provide two observations of the survey process for the same individual, as well as the entire earnings history for the administrative data. Under normality assumptions, a maximum likelihood estimator can be implemented. However, as noted above, under certain assumptions, simple moment conditions will identify the parameters of interest. We propose to investigate those conditions and further refine estimation, including admitting calendar-time effects of the permanent and transitory components. Further, administrative data are not available for all survey observations and over 20% of the survey refused to report earnings (but do have administrative matches). Moreover, not all survey respondents are matchable to administrative data, though we do have DER earnings for those individuals and we can construct two-year measures of volatility for these observations to compare to the other groups. Our approach will specifically model these processes using Bollinger et al (2018) as a departure point.


Bollinger, Christopher R., Barry Hirsch, Charles Hokayem, and James P. Ziliak. 2018. “Trouble
in the Tails? What We Know About Earnings Nonresponse Thirty Years After Lillard, Smith, and Welch.” Under final revision at Journal of Political Economy, 2017

Bloom, Nicholas, Fatih Guvenen, Luigi Pistaferri, Sergio Salgado, and Jae Song. 2017. “Decline
in Micro Volatility,” Mimeo.

Carr, Michael, and Emily Wiemers. 2017. “Recent Trends in the Variability of Men’s Earnings:
Evidence from Administrative and Survey Data,” Mimeo.

Celik, Sule, Chinhui Juhn, Kristin McCue, and Jesse Thompson. 2012. “Recent Trends in
Earnings Volatility: Evidence from Survey and Administrative Data,” The B.E. Journal of Economic Analysis and Policy 12(2): 1-26.

Dahl, Molly, Thomas DeLeire, and Jonathan Schwabish. 2011. “Estimates of Year-to-Year
Variability in Worker Earnings and in Household Incomes from Administrative, Survey, and Matched Data.” Journal of Human Resources 46(4): 750-774.

Dynan, Karen E., Douglas W. Elmendorf, and Daniel E. Sichel. 2012. “The Evolution of
Household Income Volatility.” The B.E. Journal of Economic Analysis & Policy: Advances, Volume 12, Issue 2: Article 3.

Gottschalk, Peter and Robert Moffitt. 1994. “The Growth of Earnings Instability in the U.S.
Labor Market.” Brookings Papers on Economic Activity 1, 217–254.

Gottschalk, Peter, and Robert Moffitt. 2009. “The Rising Instability of U.S. Earnings.” Journal
of Economic Perspectives 23(4): 3-24.

Gundersen, Craig, and James P. Ziliak. 2003. “The Role of Food Stamps in Consumption
Stabilization,” Journal of Human Resources 38(Supplement): 1051–1079.

Hacker, Jacob S., and Elisabeth Jacobs. 2008. “The Rising Instability of American Family
Incomes, 1969-2004: Evidence from the Panel Study of Income Dynamics.” EPI Briefing Paper 213, Economic Policy Institute

Haider, Steven. 2001. “Earnings Instability and Earnings Inequality of Males in the United
States: 1967–1991.” Journal of Labor Economics 19(4): 799–836.

Hardy, Bradley, Timothy Smeeding, and James P. Ziliak. 2018. “The Changing Safety Net for
Low Income Parents and Their Children: Structural or Cyclical Changes in Income Support Policy?” Demography, https://doi.org/10.1007/s13524-017-0642-7

Hokayem, Charles, Christopher R. Bollinger, and James P. Ziliak. 2015. “The Role of CPS
Nonresponse in the Measurement of Poverty.” Journal of the American Statistical Association 110(511): 935-45.

Keys, Ben. 2008. “Trends in Income and Consumption Volatility, 1970–2000.” In Income
Volatility and Food Assistance in the United States, D. Jolliffe and J. P. Ziliak, eds., Kalamazoo, MI: W.E. Upjohn Institute.

Kapteyn, Arie and Jelmer Y. Ypma. 2007. “Measurement Error and Misclassification: A
Comparison of Survey and Administrative Data.” Journal of Labor Economics 25(3): 513-551.

Moffitt, Robert, and Sisi Zhang. 2018. “Income Volatility and the PSID: Past Research and New
Results,” American Economic Review Papers & Proceedings, Forthcoming.

Sabelhaus, John, and Jae Song. 2010. “The Great Moderation in Micro Labor Earnings.” Journal
of Monetary Economics 57(4): 391-403.

Shin, Donggyun, and Gary Solon. 2011. “Trends in Men’s Earnings Volatility: What Does
the Panel Study of Income Dynamics Show?” Journal of Public Economics 95(7): 973-982.

Wooldridge, Jeffrey. 2001. Econometric Analysis of Cross Section and Panel Data. Cambridge,
MA: The MIT Press.

Ziliak, James P., Bradley Hardy, and Christopher Bollinger. 2011. “Earnings Volatility in
America: Evidence from Matched CPS.” Labour Economics 18(6): 742-754.

Ziliak, James P., and Thomas J. Kniesner. 1998. “The Importance of Sample Attrition in Life-
Cycle Labor Supply Estimation,” Journal of Human Resources 33(2): 507–530.

Reconciling Trends in Volatility: Evidence from the SIPP Survey and Administrative Data

Michael Carr
University of Massachusetts-Boston
Robert Moffitt
Johns Hopkins University
Emily Wiemers
University of Massachusetts-Boston


A more complete and easier to read version of this abstract is attached to the PDF of the uploaded paper.

Because of the rapid rise in inequality in the United States beginning in the late 1970s and the explicit link between inequality and earnings into its permanent and transitory components. The early literature, which primarily relies on data from the Panel Study of Income Dynamics (PSID), is in general agreement that earnings volatility increased from the 1970s through the mid 1980s and declined into the early 1990s, with approximately half of the increase in earnings inequality attributable to increasing transitory earnings variances and half to a widening of the distribution of permanent earnings. In more recent work that uses both a wider variety of administrative and survey data sources and a wider variety of methods, a lack of consensus has emerged about trends in earnings volatility both in more recent years and during earlier years where there was previously a consensus.

The lack of consensus in trends in volatility may be the result of true discrepancies in trends across different sources of survey data or across survey and administrative data sources. However, it is difficult to draw such conclusions out of the current literature because of a lack of consistency in sample definitions and methods across studies. The result is a set of estimates that are simply not comparable to each other, despite the fact that comparisons are frequently made across papers that simultaneously change data, method, and sample. For example, Shin and Solon (2011) measure earnings volatility for working-age men using the PSID and find that earnings volatility increased between the 1970s and early 1980s, stabilized and decreased slightly through the 1990s, and increased again in the late 1990s and 2000s. In contrast, Sabelhaus and Song (2009, 2010) show smoothly declining earnings volatility from the 1980s through 2005 in administrative earnings data, but pool working-age men and women together. Guvenen, Ozkan, and Song (2014), using admininstrative data on working age men, find slight declines in volatility overall between 1980 and 2011, but that volatility is cyclical and increases somewhat from 1996 through 2009, a pattern not seen in Sabelhaus and Song (2009, 2010). Ziliak, Hardy, and Bollinger (2011) use the panel component of the Current Population Survey and find that volatility increased from the 1970s through the mid 1980s and stabilized thereafter, but use a sample of 16 to 60 year olds instead of the more customary sample of 25 to 59 year olds. In other words, as new data are added to the literature on earnings volatility so are new sample definitions and methods, resulting in noncomparable estimates.

Celik et al. (2012) is one of the few attempts in the literature to estimate trends in volatility across several sources of administrative and survey data including the PSID, CPS, PSID, SIPP, and LEHD. Their estimates of earnings volatility in the CPS suggest relatively flat trends since the mid- 1980s though with considerable cyclicality, increasing volatility in the PSID since the mid-1990s, declining volatility in the SIPP since the mid-1980s, and relatively flat trends in volatility in the LEHD since the early 1990s. Their work suggest that trends in volatility may be different across different sources of data but they do not attempt to isolate the causes behind the different trends.

In this paper, we provide estimates of earnings volatility from the Survey of Income and Program Participation Gold Standard File (SIPP GSF) which links SIPP survey data to Social Security Administration earnings records. Estimates from these data will contribute to our understanding of underlying trends in volatility in several ways. First, these data allow us to estimate volatility in earnings using survey-reported earnings from the SIPP and using administrative earnings data, on the same sample of individuals. This comparison is crucial to our understanding of how the use of administrative earnings affects trends in volatility. Second, because the SIPP GSF includes demographic, geographic, and human capital characteristics, we are also able to create samples comparable to the other papers in the session and to estimate earnings volatility of both survey- reported and administrative earnings on these comparable samples. Finally, because the SIPP GSF contains both administrative and survey-reported earnings, these data allow us to understand how the choice of how to trim very low earnings in administrative data affects estimates of earnings volatility in both administrative and survey-reported earnings. Similarly, the data allows us to understand how the ability to match individuals in a survey to their administrative earnings records affects trends in the volatility of administrative earnings. By estimating trends in volatility in both survey-reported earnings and administrative earnings in the SIPP GSF and working with others in the session to identify causes of differences across datasets we are able to extend the work of Celik et al. (2012).

We also note that the same issues that make comparisons in volatility across papers difficult also plague comparisons of studies that decompose earnings inequality into its permanent and transitory components. For example, Moffitt and Gottschalk (2012) show that, despite a surge in transitory earnings variances in the late 1990s and early 2000s in the PSID, total inequality is still roughly 50/50 transitory and permanent. But work decomposing inequality in administrative data shows that the overwhelming majority of inequality is attributable to permanent earnings variances (Debacker et al., 2013, Kopczuk, Saez, and Song, 2010). While there are important methodological differences across papers which surely contribute to differences in results, work using the PSID and administrative data also decompose fundamentally different measures of earnings: Moffitt and Gottschalk (2012) decompose within-group earnings inequality (within age, education, and race) while Kopczuk, Saez, and Song (2010) and Debacker et al. (2013) decompose earnings inequality adjusted only for age. While the short panels of survey-reported earnings in the SIPP make estimation of error components models difficult, the SIPP GSF administrative earnings data allow us to estimate more complex error components models on overall earnings inequality and within group inequality. We will also explore these comparisons as part of this work.

Male Earnings Volatility in LEHD before, during, and after the Great Recession

John Abowd
U.S. Census Bureau
Kevin McKinney
U.S. Census Bureau


Using data from the Census Bureau’s Longitudinal Employer-Household Dynamics infrastructure files, we study the change in log real labor earnings and measures of its volatility for prime-age men over the period 1996 to 2015. We use a consistently defined population frame to facilitate accurate estimation of temporal changes and comparability to designed longitudinal samples of people. The Great Recession reduced earnings primarily through long spells of non-employment. Prime-age males who did not change employers and worked continuously experienced stable real earnings or growth every year. All other prime-age male workers (about 30% of the eligible population) had a cumulative loss over the same period of -0.288 log points during the Great Recession. Those with stable employment experienced very little change in volatility; whereas overall volatility for prime-age males not stably employed was about 15 times as large as for the stably employed, spiked during the Great Recession, and remained elevated thereafter.
Joseph Altonji
Yale University
Michael Keane
University of New South Wales
Dmytro Hryshko
University of Alberta
John Sabelhaus
Federal Reserve Board
JEL Classifications
  • J3 - Wages, Compensation, and Labor Costs
  • C5 - Econometric Modeling