How does the insurance firms provide to their workers vary over the business cycle? Using Swedish administrative data, I document that the elasticity of earnings to positive idiosyncratic firm productivity shocks is acyclical, whereas that to negative ones doubles in recessions. A directed search model in which contracts balance insurance against retention reproduces both patterns: low job-finding rates in recessions lead firms to cut pay more after adverse shocks, while free entry caps raises after favorable ones in both states. The same forces generate procyclical skewness in earnings changes and imply that hiring subsidies, unlike unemployment benefits, erode workers' protection.

Administrative data reveal that risky-asset participation, expected returns, and idiosyncratic return risk all increase with wealth. These patterns arise endogenously in an incomplete-markets model with portfolio choice, cyclical skewness in labor income shocks, and preference heterogeneity. Estimating the model on Swedish administrative data, we find that matching portfolio allocations over wealth requires two types: an impatient, highly risk-averse majority with preferences typical of the household finance literature, and a patient, less risk-averse 10% with preferences typical of macroeconomics. The latter sort into the right tail, hold over half of aggregate wealth, and drive rising risk-taking even at the very top. Although untargeted, the model matches wealth inequality — including the top 1% share — wealth mobility, and the decline of MPCs in wealth. We then show that endogenous portfolio choice matters for the dynamics of inequality: a permanent one-percentage-point fall in the risk-free rate raises the top 10% wealth share by 3 pp through the portfolio rebalancing of the rich. By contrast, the same shock with portfolio choices hard-wired as a function of wealth implies a decline instead.

There is a long-standing literature in economics whose goal is to infer properties of individuals’ income and human capital and their impact on consumption-saving decisions by using revealed choices, especially on consumption. While this approach is superior to the utilization of income data alone, it nevertheless relies on very strong assumptions on the form of the stochastic process for income, in particular it hard-wires the relationship between shocks to current income and expected future income, that is, human capital. In this paper we develop a new method that enables to perform this task without imposing any restriction on the latter. Specifically, we log-linearize the recursive relationship defining human capital, insert it into a linearized savings policy function and derive moment conditions which, in turn, we use for GMM estimation of the parameters governing moments of the joint and marginal distributions of savings and income. Using high quality Swedish administrative data on wealth – which enables us to overcome the well-known issues deriving from using imputed or survey data – we find that about 60 percent of human capital corresponds to expected income in the following year. This result suggests that individuals are very short-sighted regarding their future income when they make consumption-saving decisions.