Unemployment Insurance Taxes and the Cyclical and Seasonal Properties of Unemployment

We combine Current Population Survey microdata for 1979-1987 with a newly assembled database of tax rates for the Unemployment Insurance system to measure the effects of imperfect experience-rating on temporary layoffs and other types of unemployment. We find a strong negative association between the degree of experience-rating and the rate of temporary layoff unemployment, with the largest effect in recessionary years and the smallest effect in expansionary years. Increases in the degree of experience-rating are also associated with dampened seasonal fluctuations in temporary layoffs, particularly in construction and durable manufacturing. The correlation between the degree of experience-rating and the unemployment rate of permanent job losers is smaller but also negative, whereas the correlation with the unemployment Me of job quitters and re-entrants is negligible. Attempts to concrol for the endogeneity of unemployment insurance taxes are consistent with a causal interpretation of our findings. David Card Department of Economics Princeton University Princeton, NJ 08544 and NBER Phillip B. Levine Department of Economics Wellesley College Wellesley, MA 02181 Unemployment Insurance Taxes and the Cyclical and Seasonal Properties of Unemployment A unique feature of the U.S. unemployment insurance system is its experiencerated tax structure. Firms whose previous employees have collected more unemployment insurance (UI) benefits are charged higher payroll taxes. Nevertheless, employers in most states are only partially experience-rated: the tax increases generated by an additional UI claim cover only a fraction of the benefits paid. Because of the implicit subsidy on UI benefits, it is argued that firms have an incentive to hold an excess inventory of workers and cycle them through spells of temporary unemployment.' Indeed, estimates presented by Feldstein (1978) and Topel (1983) suggest that up to onehalf of all temporary layoff unemployment in the US can be attributed to imperfect experience-rating. While the effects of imperfect experience-rating are easily identified in a stationary environment, in a cyclical or seasonal environment the effects vary with the state of demand. Experience-rated UI taxe~ introduce a layoff or firing cost into the firm's inter temporal optimization decision. Increases in the degree of experience-rating raise this adjustment cost, creating an incentive for firms to layoff fewer workers in a recession and hire fewer workers in a boom. Imperfect experience-rating therefore generates more temporary layoffs and greater unemployment in a trough, but higher employment (and possibly lower unemployment) at the peak of the cycle. 'This point is developed formally in models by Feldstein (1976), Baily (1977) and Brechling (1977). 2 In this paper we use a newly assembled database of experience-rating factors for individual states and industries to measure the effects of imperfect experience-rating at different points in the demand cycle. Vsing Current Population Survey microdata from 1979 to 1987, we estimate the effect of experience-rating on temporary layoff unemployment rates during the cyclical downturn in the early 1980s, and in the expansionary periods before and after. We also examine the effects in different months of the year. As an informal check on the results, we estimate similar models for the unemployment rates of permanent job-losers and non-jab-losers (job-quitters and labor force re-entrants). We find a strong negative correlation between the degree of experience-rating and the rate of temporary layoff unemployment in recessionary years, but smaller and unsystematic correlations in expansionary years. Likewise, temporary layoff rates in high-unemployment months are strongly negatively correlated with the degree of experience rating. By comparison, there is no relation between experiencerating and the unemployment rate of quitters and labor force re-entrants. A major difficulty in the empirical analysis is the potential endogeneity of VI tax rates. Since each state's VI system has a maximum tax rate, industries with the highest unemployment rates may not be experience-rated at all. We present instrumental variables estimates that use only the features of the state tax system (and not the actual unemployment experience of firms in the state) to identify the experience-rating effect. We also examine the effects of experience rating within industries, and find negative effects in services and trade, where few firms are at the maximum of the tax schedule. 3 On balance, we believe the evidence is broadly consistent with a causal link between experience rating and the use of temporary layoffs. J. EMPLOYMENT AND TEMPORARY LAYOFFS IN A CYCLICAL ENVIRONMENT This section presents a simple model of employment determination in the presence of cyclical demand shifts. To capture the effects of experience-rating in the VI tax system we model the pool of recently laid-off workers "attached" to each firm. Temporarily laid-off workers have a value to the firm, since they can be re-employed in an upturn without incurring any training or recruiting costs. They also have a cost, since the VI benefits drawn by these workers are charged back to the firm at a rate that depends on the degree of experience-rating. In a model with a simple alternating demand cycle we show how changes in the degree of experience rating affect the firm's employment and unemployment decisions at the peak and trough of the demand cycle. In any period t a firm is assumed to have a pool P, of attached workers, consisting of N, employed workers and V, temporarily unemployed workers. The firm pays a wage rate w to its employed workers, which we take as exogenous? Its laid-off workers receive an average unemployment benefit of $B per workeLl It is convenient to express B as a fraction of the average wage: B = R ow, where R is the net replacement ratio. 2In contrast, Feldstein (1976) and Baily (1977) take the utility level of workers as exogenous and treat the wage as endogenous. lOnly a fraction of laid-off workers receive any unemployment benefits. Thus B represents the product of the fraction receiving benefits and the ave~;age benefit per worker, conditional on receiving benefits. See below. 4 The cost to the firm of the unemployment benefits received by its laid-off workers is (Is) of the total, where s is the implicit subsidy generated by incomplete experience-rating of the VI tax system. For simplicity, we treat s as fixed and exogenous, although longrun differences in the firm's layoff policy can obviously affect the value of s. We defer discussion of this point to the next section. Assume that a fraction 0 of workers who are employed in period t quit during the period and that a fraction 0+L\ of workers on temporary lay-off move to other jobs. If laid-off workers have a higher attrition rate than those with jobs, then L\ > O. The available pool of workers in period t depends on the size of the pool in the previous period and on the number of new-hires (AJ; P, = (I-O)P'l L\U,.IP•.I + A." (I) here U t == VJPt is the fraction of workers on lay-off. We assume that A,~ 0, which implies that the firm cannot freely dispose of workers. Rather, excess workers are released into the pool of attached unemployed workers. We also assume that new-hires are costly and denote the per-worker recruiting and training cost by pw. Finally, we assume that the revenues of the firm in period t have the simple form e,f(N,), where f( ) is an increasing and strictly concave production function and e. denotes a relative demand shock in period t. With this set-up, the net profits of the firm in period tare