Following up on yesterday's post about the discrepancies of micro- and macro-estimates, another recent paper caught my eye. Sungbae An, Yongsung Chang and Sun-Bin Kim argue that for a representative household model to replicate the behavior of aggregates, on needs often unreasonable parameter values. This may be due to the fact that representative agents do not aggregate well, or because heterogeneity matters for aggregates. They demonstrate this for the second case, modeling heterogeneous households subject to credit constraints and indivisible labor, and find that this leads to imperfect aggregation.
Of particular interest here is that the departures from the representative agent economy can be shut down one by one to understand what matters and how. This counterfactual exercise shows that there is considerable bias in the risk aversion and uncertainty in the labor supply elasticity. Thus, heterogeneous agents cannot by properly summarized with a representative agent. And what all this highlights is that aggregation is the main issue, not, as surmised by some like Gregory Mankiw, Julio Rotemberg and Lawrence Summers, that the labor market does not clear or that preferences shift.