主观收入与就业动态

Subjective earnings and employmentdynamicsManuel Arellano, Orazio Attanasio,Margherita Borella, Mariacristina De Nardi,Gonzalo Paz-Pardo Abstract We develop a new approach to estimating earnings, job, and employment dy-namics using subjective expectations data from the NY Fed Survey of ConsumerExpectations.These data provide beliefs about future earnings offers and accep-tance probabilities, offering direct information on counterfactual outcomes and en-abling identification under weaker assumptions. Our framework avoids biases fromselection and unobserved heterogeneity that affect models using realized outcomes.First-step fixed-effects regressions identify risk, persistence, and transition effects;second-step GMM recovers the covariance structure of unobserved heterogeneitiessuch as ability, mobility, and match quality.We find lower risk and persistenceof the individual productivity component than in prior work, but greater hetero-geneity in ability and match quality. Simulations show that reduced-form estimatesoverstate persistence and volatility on individual-level productivity due to job tran-sitions and sorting. After accounting for heterogeneity, volatility declines and be-comes flat across the earnings distribution. These results underscore the value ofexpectations data. JEL Codes: C23, C81, D15.Keywords: Subjective expectations, earnings dynamics models. Non-technical summary How uncertain are workers’ incomes, and why do some workers earn persistently morethan others? Workers who face a lot of income risk engage more in precautionary savingsand respond differently to economic shocks and to policy.The answers also matterdirectly for policy design:the appropriate generosity of unemployment insurance, theredistributive role of taxation, and the way monetary policy transmits to householdspending all depend on whether income differences across people mainly reflect risk orheterogeneity. There are many challenges associated with measuring earnings risk. First, the earningsof two workers diverge, it is hard to tell from the data alone whether it reflects genuineuncertainty or predictable differences between the two workers. Second, earnings are onlyobserved for people who work: if those who receive bad shocks are also those who dropout of employment, the observed data give a distorted picture of the underlying risks(“selection bias”).Disentangling these forces from realised earnings histories typicallyrequires following the same individuals for many years, and such data does not exist oris hard to access in many countries. In this paper we develop a new way around these problems: instead of relying onlyon what happened to workers, it uses workers’ expectations about what might happen,including scenarios that never materialise. The data come from the Federal Reserve Bankof New York’s Survey of Consumer Expectations (SCE), a monthly survey of roughly1,300 U.S. household heads. Its labor market module, which we use from March 2014 toNovember 2019, asks both employed and unemployed respondents three key questions:the annual salary of the best job offer they expect to receive over the next four months,the probability that this best offer falls within various salary ranges, and the probabilitythat they would accept an offer at each of those salary levels. By asking how people would respond to a menu of hypothetical wage offers, the surveygenerates variation similar to a controlled experiment. We can directly observe both thewages potentially available to a worker (whether or not the worker takes them) and theworker’s decision rule. This eliminates selection bias, because beliefs about future offersare reported before any outcome is realised and cover counterfactual situations. By using this data and estimating a model of workers’ earnings and employmenttransitions, we find that earnings shocks are smaller and fade faster than previouslythought.Match quality, moreover, largely does not carry over when a worker changesemployer. On the other hand, differences between people and jobs are much larger thanpreviously estimated.Stable heterogeneity in ability and in employer–match qualityexplains far more of the dispersion in earnings than in earlier studies. Intuitively, much ofwhat looked like everyone facing large, long-lasting shocks is better described as differentpeople being on persistently different earnings tracks, and most of the changes in tracksseem to be associated with job transitions. Wage offers strongly drive labor market transitions. Higher offered salaries substan-tially raise the probability that an unemployed worker takes a job and that an employedworker switches employers, while the wage in the current job has a smaller effect onstaying. These results have different policy implications.First, quantitative heterogeneous-agent models require an income process as a key input. Our results are useful in gettingthe decomposition between risk and heterogeneity right and therefore improving ther