Consistent with mental accounting, we document that investors sometimes choose the asset allocation for one account without considering the asset allocation of their other accounts. The setting is a firm that changed its 401(k) matching rules. Initially, 401(k) enrollees chose the allocation of their own contributions, but the firm chose the match allocation. These enrollees ignored the match allocation when choosing their own-contribution allocation. In the second regime, enrollees simultaneously selected both accounts' allocations, leading them to mentally integrate the two. Own-contribution allocations before the rule change equal the combined own- and match-contribution allocations afterwards, whereas combined allocations differ sharply across regimes.

Social identities prescribe behaviors for people. We identify the marginal behavioral effect of these norms on discount rates and risk aversion by measuring how laboratory subjects' choices change when an aspect of social identity is made salient. When we make ethnic identity salient to Asian-American subjects, they make more patient choices. When we make racial identity salient to black subjects, non-immigrant blacks (but not immigrant blacks) make more patient choices. Making gender identity salient has no effect on intertemporal or risk choices.

We compare the liquidity that six developed countries have built into their employer-based defined contribution (DC) retirement schemes. In Germany, Singapore, and the UK, withdrawals are essentially banned no matter what kind of transitory income shock the household realizes. By contrast, in Canada and Australia, liquidity is state-contingent. For a middle-income household, DC accounts are completely illiquid unless annual income falls substantially, in which case DC assets become highly liquid. The US stands alone in the universally high liquidity of its DC system: whether or not income falls, the penalties for early withdrawal are low or non-existent.

Do laboratory subjects correctly perceive the dynamics of a mean-reverting time series? In our experiment, subjects receive historical data and make forecasts at different horizons. The time series process that we use features short-run momentum and long-run partial mean reversion. Half of the subjects see a version of this process in which the momentum and partial mean reversion unfold over 10 periods ('fast'), while the other subjects see a version with dynamics that unfold over 50 periods ('slow'). Typical subjects recognize most of the mean reversion of the fast process and none of the mean reversion of the slow process.