General equilibrium effects of social insurance programs: long and short-run sectoral re-allocation

在社会保险文献中，Baily（1978）和Chetty（2006）的开创性工作为部分均衡设置下的福利分析提供了足够的统计数据。他们的工作依赖于包络定理的有力应用，以证明储蓄和其他行为反应的变化与福利无关，但人们自然想知道，当我们摆脱部分均衡假设时，这种见解能在多大程度上成立。McKay和Reis（2016）等人的作品提出了结果可能在一般均衡中被打破或削弱的方式。他们将失业保险嵌入到通用电气的商业周期模型中，发现更高的替代率会导致更低的稳态资本存量和更差的总体抵御总体冲击的能力。Landais等人（2018）的后期工作扩展了Baily-Chetty公式，以在匹配模型中通过劳动力市场紧缩来解释一般均衡效应，但也可以考虑其他扩展。例如，如果各行业产出和就业的波动性存在系统性差异，失业保险可能影响各行业工人的分配，或者失业保险可能与衰退期间的行业调整相互作用，这种想法并非不合理。这些更微妙的道德风险形式可能很难从经验上识别出来，但某些政策变化可能会涉及到它们。作为初步设想，西班牙在失业保险方面存在区域和部门差异，其形式是对安达卢西亚和埃斯特雷马杜拉的农业工人实行一种特殊的、更慷慨的制度，称为PER / PFEA。差异中的差异研究设计可能潜在地确定该政策对农业就业份额的影响。诚然，这是一个比我前面提到的更狭窄的背景，但对于一般社会保险在宏观经济规模上产生类似配置效应的范围，这种做法可能具有指导意义。

In the social insurance literature, the seminal work of Baily (1978) and Chetty (2006) developed sufficient statistics for welfare analysis in a partial equilibrium setting. Their work relies on a powerful application of the envelope theorem to argue that the change in savings and other behavioural responses are irrelevant for welfare, but naturally one wonders to what extent this insight survives when we get rid of the partial equilibrium assumption. Works such as McKay and Reis (2016) suggest ways in which the result may break down or be weakened in general equilibrium. They embed unemployment insurance into a GE business cycle model and find that higher replacement rates lead to lower steady state capital stock and worse overall capacity to weather aggregate shocks. Later work by Landais et al. (2018) extended the Baily-Chetty formula to account for general equilibrium effects via labour market tightness in a matching model, but there are other extensions that could be considered. For instance, it is not implausible to think that unemployment insurance could affect the allocation of workers across industries if there are systematic differences in the volatility of output and employment across sectors, or that it could interact with sectoral adjustments during recessions. These subtler forms of moral hazard are probably harder to identify empirically, but some kinds of policy variation could get at them. As a preliminary idea, in Spain there are regional and sectoral differences in unemployment insurance in the form of a special, more generous regime for agricultural workers in Andalusia and Extremadura, called PER / PFEA. A differences-in-differences research design could potentially identify the effects of this policy on the share of agricultural employment. This is admittedly a narrower context than the one to which I referred earlier, but the exercise could be instructive about the scope for general social insurance to induce similar allocative effects at a macroeconomic scale.