سطح سود بیمه بیکاری و تغییرات مصرف

We use a Canadian survey of the unemployed to examine how household expenditures after a job loss respond to the level of income replacement provided by UI. We isolate a liquidity constraint or ‘transitory income’ effect from the ‘permanent income’ shock of job loss, and from the costs of working. We find significant effects of varying the replacement ratio among the third of the sample who did not have assets at the job loss. We conclude that the consumption smoothing benefit of UI is concentrated wholly on a sub-group of the unemployed.

Should governments provide unemployment insurance and, if so, what should the provisions be? There have been numerous studies of the costs of unemployment insurance (UI) in recent years but fewer attempts to measure the benefits. Households may benefit from UI in several ways. The ‘consumption smoothing’ benefit of UI is realized by liquidity constrained households that have temporarily low income and are not able to set their current consumption at a level consistent with their expectations of the future. The ‘insurance’ benefit of UI arises because UI, like a progressive tax, 1 reduces the variance of stochastic outcomes. That UI has an ‘insurance’ benefit requires only that households be risk adverse and not otherwise fully insured against idiosyncratic shocks; in particular, borrowing constraints are not a necessary condition. The ‘consumption smoothing’ benefit of UI is often thought to be the most important potential benefit of UI, 2 and it is this benefit which is the focus of this paper. We use a new Canadian panel data set to examine household expenditure changes across a job loss and how those changes vary with the level of replacement income provided by UI. Household expenditure changes with unemployment confound three things: the costs of working (changes in expenditure due to the non-separability of consumption from labor supply), a response to the ‘permanent income’ shock of job loss, and, a response to ‘transitory income’. Responses to transitory income are informative about the ‘consumption smoothing’ benefit of UI. Specifically, if households respond to marginal changes in ‘transitory income’ then they are not on their optimal consumption path, and marginal actuarially fair increases in UI replacement income (that increase current income and lower future income) raise household welfare, moving the household towards that optimal path. The form of our test will be to see if differences in the UI replacement ratio across a sample of households experiencing unemployment correlate with differences in the reported expenditure change from before the beginning of the unemployment spell. There is no variation in labor force status across our sample, so our test is not confounded by non-separability between consumption and labor supply. Consequently, to isolate a ‘transitory response’ the main econometric problem we have is that across our sample the UI replacement ratio is plausibly correlated with the permanent shock from a job loss. To overcome this we use rich controls for the permanent shock. We can also test that we have ‘purged’ the UI benefit of its ‘permanent’ component by performing an exogeneity test using instruments that are based on the ‘quasi-experiment’ afforded by two sets of legislative changes (and one administrative change) to the Canadian UI system over our sample period. These instruments are correlated with the temporary loss of income but not with the permanent shock of job loss. Of the small literature on the benefits of UI, this paper is most similar to Gruber (1997). Our work differs from Gruber’s in several important ways: we use a different measure of expenditure (total, rather than food), a different source of variation in benefits (specific legislative changes rather than state-time variation) and of course a different country (Canada rather than the US) which differs in the parameters of its UI system and in other aspects of its social safety net. In addition, our data are rich enough to allow us to focus attention on particular subsets of households which are likely to be liquidity constrained. Since responses to ‘transitory income’ should only be observed among the liquidity constrained, this focus improves our chance of finding indications of large ‘consumption smoothing’ benefits. An outline of the paper is as follows. In the next section, we derive our estimating equation from a structural model of inter-temporal utility maximization. The presence of a structural model makes explicit the potential sources of gains from UI. Additionally, the structural model allows us to state explicitly and clearly the conditions that allow us to identify the ‘benefit effect’; that is, the impact of marginal changes in the replacement rate on total expenditure. In Section 3 we introduce the data, the Canadian UI system and the legislative and administrative changes that the data capture. Section 4 presents our empirical results. Our principal finding is that differences in the replacement ratio have small mean effects on household total expenditures. For example, a 10 percentage point cut in benefit levels (from 60% to 50% replacement, for example) would lead to an average fall of only 0.8% in total expenditure, and the implied cut in expenditures due to a 1 dollar cut in benefits (at the means of the data) is less than 7 cents. This mean effect is somewhat smaller than those reported by Gruber (1997). Although the mean effect is small we show that it is concentrated among a small number of households and that most households would not react to a change in the UI replacement rate. The strongest effect is for households that did not have liquid assets at the job separation; this is consistent with the presence of liquidity constraints. We also find a significant effect for respondents who were married but whose spouse was not employed; this suggests that the presence of a second income facilitates consumption smoothing (although we do not find any significant effect for singles). In Section 5, we discuss possible reasons for the difference between our results and those of Gruber (1997) and summarise what we think are the important implications of our results for policy analysis. In particular, we emphasise that the fact that benefit effects are heterogenous invalidates most analysis which effectively assumes a representative agent.

To conclude, we have used a survey of unemployed Canadian workers to investigate the effects of changes in the UI benefit on household total expenditure. We emphasise that we consider only medium length unemployment spells (between 4 and 9 months); we have no estimates of the effect of benefit levels on short run or long run expenditures. Moreover, our results are best applied to UI systems that have replacement ratios of between 40% and 60%; we would not want to extrapolate to what would happen if benefits were cut or increased dramatically. Given this, some clear results do emerge. The most important of these is that for most households in our sample (perhaps as much as 90%) small changes in UI benefits would not lead to changes in total expenditure. For a small number of households, however, total expenditure is very sensitive to the level of UI benefits. These households are those that had little in the way of assets or access to credit and which experienced a large fall in total expenditure. Although the implications for policy from this finding have yet to be fully analysed it is clear that the consumption smoothing gain from UI (which many believe to be the most important gain) will depend heavily on the weight given to the small group who have substantial welfare gains from increases in UI benefits.