Search

THE AMERICAN DREAM IS THAT each generation should live twice as well as its predecessor. During the hundred years before 1973, real average hourly earnings rose by 1.9 percent a year. ' At that rate earnings doubled every thirty-six years, and the dream was realized. The dream no longer holds. Since 1973 the United States has failed to match its historic track record. In 1973 average real hourly earnings, measured in 1982 dollars by the consumer price index (CPI), were $8.55. By 1992 they had actually declined to $7.43-a level that had been achieved in the late 1960s. Had earnings increased at their pre-1973 pace, they would have risen by 40 percent to more than $12.00. Or consider average real hourly compensation. This is a more comprehensive measure of the payments to labor because it includes fringe benefits as well as earnings. Between 1973 and 1991, real hourly compensation rose by only 5 percent. However the growth of labor income is measured, it clearly has slumped since 1973. A second ominous development in the American economy has accompanied this slump: a dramatic increase in the inequality of earnings. In

In 1970, when Robert Hall asked, "Why Is the Unemployment Rate So High at Full Employment?" the unemployment rate for adult men stood at 3.5 percent. That rate, which had been substantially below that level throughout the late 1960s, would climb to 4.4 percent in the recession of 1971. More recently, after the longest economic expansion of the post-war period, the unemployment rate of prime-aged men in the late 1980s settled at just below 5 percent of the labor force. What changes in the American labor market led to this apparent secular increase in the natural rate of unemployment. Twenty years later, we revisit Hall's question and turn up some new answers.

employment, and they are generally viewed as programmes of social insurance wherein the costs of financing UI benefit payments are shared by all workers and firms in the covered (insured) sector. As is common in such situations, however, problems of "moral hazard" may arise: the frequency and severity of periods of unemployment are to an important degree products of the employment decisions of workers and firms, and these may be affected by the provision of benefits. There may therefore be important causal links between the provision of UI benefits and the magnitude of unemployment. In this paper, we survey some of the major theoretical and empirical issues relating to causal links between UI and unemployment, with particular emphasis on the role played by systems of UI financing. It has long been recognized that the flow of UI benefit payments to unemployed individuals will affect job search decisions and hence their probability of leaving unemployment. More recently, following the work of Joseph Becker (1972), methods of financing UI benefit payments have found a prominent place in the literature. The issues in this area focus primarily on the behaviour of firms in influencing an individual's probability of entering unemployment, and are related to methods of experience rating (or the lack thereof) in determining Ul tax rates. This practice, which is meant to alter UI tax liabilities on the basis of a firm's history of generating unemployment, is unique to UI systems in the United States, although its policy implications are quite applicable elsewhere.2 We believe that the effects of experience rating represent the most important set of unresolved empirical issues in the analysis of UI, and we therefore devote a large portion of our discussion to the theoretical implications of this practice and to potential methods of measuring its effects. By way of introduction, it is worthwhile to consider some simple empirical background, including the magnitude of UI programmes relative to other social insurance and welfare programmes, and the types and sources of potentially compensated unemployment. While these data are for the United States, we expect that the thrust of the argument applies to other UI systems as well. Table 1 gives the percentage distribution of benefit payments by social insurance beneficiary, and each category's benefits as a percentage of total personal income, for the United States over the period 1973-1976. All categories other than UI are dominated by programmes under the Social Security Act, and hence it is clear that UI is small relative to social security. As one might expect, UI's share of social insurance benefits and of personal income moves counter-cyclically, being

A commonly used procedure in a wide class of empirical applications is to impute unobserved regressors, such as expectations, from an auxiliary econometric model. This two-step (T-S) procedure fails to account for the fact that imputed regressors are measured with sampling error, so hypothesis tests based on the estimated covariance matrix of the second-step estimator are biased, even in large samples. We present a simple yet general method of calculating asymptotically correct standard errors in T-S models. The procedure may be applied even when joint estimation methods, such as full information maximum likelihood, are inappropriate or computationally infeasible. We present two examples from recent empirical literature in which these corrections have a major impact on hypothesis testing.