Income risk over the life cycle and the business cycle: New insights from large datasets.

• Author: Guvenen, Fatih
• Position: Research Summaries

Millions of young men and women enter the labor market annually. Over the next 40 years, each of them goes through a unique journey that involves surprises as well as disappointments: searching for the dream career, finding and losing jobs, getting promotions, salary raises, or demotions, and experiencing the recessions and booms of the macro economy.

In recent research, I try to understand the nature of the uncertainty that major labor market events generate for workers. There are three main dimensions of this research, which studies how individuals' income uncertainty and risk varies over the business cycle and over the life cycle, and how it has changed over the last four decades. The answers to these questions are of immediate relevance for both deepening our knowledge of labor market dynamics and for informing social insurance debates, such as those surrounding Social Security reform, unemployment insurance policy, the degree of job protection, and the progressivity of the tax system. Each of these policies seeks to moderate various types of individual risk.

In this summary, I discuss in detail my colleagues' and my findings on the variation of income risk over the business cycle. I also briefly describe our findings about life cycle risk and changes in risk over time.

Because of its central role for policy questions, the nature of individual income uncertainty has received significant attention from academics since the 1970s, when panel datasets on individual incomes started to become available. However, those datasets--as well as the majority of newer ones--were overwhelmingly based on surveys and therefore suffered from the usual problems of small sample sizes, sample attribution, and survey response error. The data problems forced researchers to focus on simple, parameterized statistical models to examine these questions. Perhaps not surprisingly, the combination of data issues and restrictive methods and assumptions often yielded a wide range of answers to these questions, resulting in wide disagreements. My earlier research on these topics also relied on these survey-based datasets and methods; I became increasingly uncomfortable about their use and this motivated the current work.

My research on income uncertainty builds on two main elements. First, it makes extensive use of large administrative panel datasets on individuals from various countries, some of which have become more widely available in the last decade. Second, because these datasets do not suffer from the shortcomings of survey data such as small sample, attrition, and measurement error, my research relaxes many of the econometric assumptions made in prior literature. For example, my collaborators and I relax the strong focus in earlier work on just the variance--he second moment--as a measure of risk and uncertainty. We find that most of the interesting and substantively important variation happens in "higher-order moments," in particular in the third- and fourth-order moments. The risk from these components, "higher-order risk," matters a great deal for a range of substantive economic questions.

The Datasets

One dataset my coauthors and I have used in this research comes from the Master Earnings File (MEF) of the U.S. Social Security Administration (SSA). The MEF currently covers the entire U.S. population with a Social Security number from 1978 to 2013. It contains data on each individual's labor earnings (wage/salary income from W-2 forms and self-employment income from Schedule SE), as well as some key demographic variables and employer identifiers. The substantial sample size, 600 million individual-year observations in a 10 percent subsample, allows us to employ fully nonparametric methods and take what amounts to high-resolution pictures of individual earnings histories. The relaxation of parametric assumptions is a key part of this research agenda.

In addition, we use data from Swedish, German, and French administrative records (Linda, IAB, and DADS, respectively) and complement them with various survey-based datasets (PSID for the U.S. and GSOEP for Germany) as well as firm-level datasets (Compustat Global, OSIRIS, and ORBIS).

Income Risk over the Business Cycle

Conventional wisdom among economists was that income shocks become much larger in recessions, and that this property was captured by a rise in the variance of such shocks. The most widely-cited papers on...