مخاطرات اخلاقی، بیمه بهینه بیکاری و نرخ گذاری تجربه

This paper is concerned with evaluating alternative unemployment insurance (UI) schemes in a dynamic economy with moral hazard. We consider changes in the size and duration of UI benefits, and the effects of experience rating, and use a dynamic contracting approach to determine a benchmark optimal allocation. Radical changes in the current UI system increase welfare, but not by much. A move to full experience rating has distributional effects, but the aggregate effects are negligible.

The incentive problems associated with moral hazard have long been recognized as being of critical importance for the design of unemployment insurance (UI) systems. A dynamic model with moral hazard is constructed here to evaluate UI along several dimensions. In particular, we are interested in the effects of the replacement rate, the duration of UI benefits, and experience rating on the search behavior of the unemployed, shirking by the employed, unemployment, and welfare. Taking the solution to a dynamic contracting problem under moral hazard as a benchmark, we evaluate changes in the UI system in the U.S. In most countries, UI benefits are tied in some way to wages when employed, and there is often a maximum length of time for which benefits can be collected. In these two respects, the United States is not unusual. In the U.S., UI systems differ by state, but typically can be characterized as providing benefits at a replacement rate of about 50% for 26 weeks at the maximum. A key difference in the U.S. UI system, relative to other countries, is in its financing. In the U.S., UI benefits are funded through payroll taxes on firms, and an individual firm's payroll tax rate tends to increase as the firm generates more insured unemployment. This feature of the UI system is referred to as experience rating. A replacement rate below 100% and a limit on the UI benefit period are apparently intended as incentive devices in the context of moral hazard. That is, since the search effort of the unemployed is costly or impossible to monitor, too much insurance implies that the unemployment rate would be too high. As in any contracting relationship with moral hazard, there is a tradeoff between incentives and insurance, and reducing benefits and/or cutting off benefits after some period of time are two simple ways to limit insurance. It is an open question whether UI is in fact limited in the U.S. in a way that comes close to achieving an optimal incentive structure. In the literature, one approach has been to ask what replacement rate would be optimal in the context of a model where UI benefits continue indefinitely. For example, Hansen and Imrohoroglu (1992) study an incomplete markets model where there is a moral hazard problem concerning the observability of job refusals by unemployed agents. They find that UI benefits in the U.S. are in general too generous. In a related model with heterogenous agents and voting, Pallage and Zimmerman (1998) argue that current UI benefit levels can be supported in a voting equilibrium. Davidson and Woodbury (1998) construct a two-sided search model, and look for an optimal UI program by varying the replacement rate. Their conclusions, while predicated on the degree of risk aversion, indicate again that the replacement rate is too high in the U.S. In a principal-agent environment with moral hazard, where UI benefits are interpreted as the consumption of an unemployed agent, Shavell and Weiss (1979) and Hopenhayn and Nicolini (1997) find that a welfare-maximizing UI program features benefits which decline monotonically during a spell of unemployment. Topel 1983, Topel 1985 and Topel 1990 has argued that experience rating plays an important role in the UI system in the U.S. In the absence of experience rating, it is argued, firms have the incentive to lay off employees temporarily, and these laid-off workers then collect UI. With full experience rating, the firm ultimately bears the entire cost of the UI benefits paid out, in which case the firm might prefer to keep employees on the payroll who might otherwise be laid off. In the U.S., experience rating is not perfect, since there are upper and lower bounds on payroll tax rates. Thus, firms in industries with high average unemployment tend to be subsidized at the expense of firms in industries with low average unemployment. Topel (1985) has argued that the efficiency losses from imperfect experience rating are large, and that the unemployment rate in the U.S. would drop by about a percentage point if full experience rating were introduced. These conclusions are arrived at without the aid of a formal general equilibrium model. Here, we construct a dynamic model with moral hazard where agents make unobserved effort decisions when employed and when unemployed. In this sense, our model differs from the related literature, in that it accounts for the effects of UI on job-retention effort, and not just on the search effort of the unemployed. We allow for heterogeneity among workers, with the interpretation being that workers of different types are employed in different industries. We thus have a structure that can capture the effects of experience rating. Our approach consists of studying the allocations that arise in this environment under two allocation mechanisms. First, we consider competitive equilibrium allocations in a setup with incomplete markets where agents have access to UI and can also self-insure by accumulating assets. Under this allocation mechanism, the model is calibrated to labor market data, and we evaluate the effects of changes in the UI system. Second, we compute the optimal steady state allocation in this economy through an application of dynamic contracting methods (Green, 1987; Spear and Srivastava, 1987; Phelan and Townsend, 1991; Atkeson and Lucas 1992 and Atkeson and Lucas 1995; Wang, 1995; Aiyagari and Williamson 1999 and Aiyagari and Williamson 2000). This optimal allocation is then a useful benchmark for determining the potential for welfare improvements from changes in UI, and in suggesting features of the allocation which would improve incentives. The difference between our work here and that in Wang and Williamson (1996), is that here we allow for savings in the incomplete markets economy, we provide for agents in the model who are ineligible for UI, we examine more detailed changes in the UI system, and we allow sufficient heterogeneity to study the effects of experience rating. Also, our model here has the feature that, as an unemployment spell progresses, search effort by an unemployed worker is less effective in becoming re-employed. This is important for fitting the labor market data, and can be interpreted as capturing the depreciation of human capital during unemployment. This phenomenon has been studied by Addison and Portugal (1989) and Neal (1995). Ljundqvist and Sargent (1998) study a search model where productivity on the job is affected by the length of an agent's unemployment spell. Sleet (1997) considers a model closely related to Wang and Williamson (1996), but focuses more on cyclical issues. Gomes et al. (2001) study the general equilibrium effects of unemployment insurance in an incomplete markets environment with job search. There are two important conclusions from our work. First, the potential welfare benefits from fine-tuning the UI system in the United States are very small. This is because, in our model, agents can accomplish a high degree of consumption smoothing through precautionary saving and by varying effort when employed and unemployed. Second, experience rating has important distributional implications, but little impact on aggregate welfare or the aggregate unemployment rate. In Section 2, the environment is constructed. Section 3 contains a discussion of alternative allocation mechanisms, Section 4 involves a consideration of the one-agent economy, and in Section 5 the two-agent economy and experience rating are analyzed. Section 6 is a summary and conclusion.

While our analysis indicates that optimal UI systems look radically different from the one currently in place in the United States, the welfare gains from changing the UI system are very small. The upper bound we obtain for the potential welfare gains from altering the UI system in the U.S. is about 1.5% of aggregate consumption. Since this estimate is based on the assumption that private insurance and financial arrangements are extremely primitive, i.e. agents are restricted to self-insuring by holding a zero-interest asset, the actual potential gains are likely much smaller than 1.5% of consumption. We have shown that optimal UI benefit schedules in the context of our incomplete markets environment should be anything but monotonically decreasing with the duration of an unemployment spell. The best arrangement was one where an unemployed worker has a low replacement rate for one quarter, benefits drop essentially to zero for three quarters, and then the replacement rate rises to a rate higher than the initial replacement rate, remaining at that level indefinitely. This back-loading of benefits is quite different from the front-loading inherent in the current UI system in the U.S. In contrast to assertions in the literature, we obtain almost no effect on aggregate welfare or the unemployment rate due to a move from the current system of partial experience rating to one with full experience rating. Experience rating matters only for the distribution of wealth and income (though the welfare effects are small), with high-unemployment workers losing and low-unemployment workers gaining from a move to full experience rating. Our results show that imperfect experience rating actually gives the UI system a negative effect on welfare for the segment of the population which has a low average unemployment rate. Our analysis of optimal allocations is suggestive of optimal UI systems that go beyond systems where UI benefits depend only on the duration of unemployment and on wages when employed. In the structure of the optimal allocation, consumption depends on an agent's entire employment history. In terms of how the UI system operates, this indicates that the system should be experience-rated at the level of the individual, not at the firm level. That is, UI benefits should depend on the worker's entire employment history, and the taxes financing UI should be individual-specific and tied to the same history. To see how this could be implemented in a simple way, we can again borrow from the structure of the solution to the social planner's problem, in that there can be a state variable which serves as a sufficient statistic for employment history. Thus, each worker could hold a UI account which is credited when the worker is employed, and debited when the worker is unemployed and receiving benefits. Credits and taxes paid to the UI system would depend on the level of the individual's account balance and the worker's current wage, and debits and benefits would depend on the account balance.