Unemployment and Okun’s Law in the United States, 1948-2005

The torpid recovery of the job market in the wake of the 2001 recession has continued to be the focus of much discussion among economists and policy makers as job growth continues to only creep upwards from its low four years after the start of the recovery. Correspondingly, unemployment numbers remained seemingly high for much longer than forecasters had predicted, even as the rest of the economy continued to gather speed. Anecdotally, this behavior has been attributed to high productivity growth during the ‘90’s, growth which enabled firms to meet recovering demand after the recession without hiring new workers. However, only a statistical analysis of the behavior of unemployment—historical and recent—can reveal to what extent this behavior really does differ from historical norms, and to what extent it can be attributed to productivity gains. This paper will follow the approach taken by Robert Gordon in a 1984 study that poses almost exactly these questions. In his case, however, the provocation was not unusually slow employment growth, but the unusually fast employment growth during the 1983-84 recovery. Gordon’s approach is to link unemployment with GNP via the estimation of an Okun’s Law coefficient, which, more accurately speaking, links the unemployment and output “gaps.” (Since this paper will make use of GDP—as opposed to GNP—data, the following description of the proposed model will also use that term.) The ‘GDP gap’ refers to the “percentage difference between actual and potential real [GDP],” (Gordon, 1984) potential GDP being the level of output the economy could produce under conditions of full employment (Okun, 1970). Likewise the unemployment ‘gap’ can be understood as the difference between the actual unemployment rate and the ‘full-employment’ unemployment rate. Okun’s Law states that the “unemployment gap is a constant fraction, k, of the output gap”, approximately equal to 0.3 (Gordon, 1984). The recent recovery’s seemingly slow decrease in unemployment implies a similarly slow decrease in the unemployment gap, and therefore—via Okun’s law—a correspondingly slow decrease in the output gap. This paper will examine the extent to which the recent behavior of unemployment differs from historical norms—or perhaps the extent to which it has altered historical norms. This analysis, after Gordon’s original, will begin with a simple identity that relates real GDP (Q) with the unemployment rate (E/L), hours per employee (H), labor productivity (Q/EH), the labor force participation rate (L/N), and population (N) (539).