The Variability of the Okun Coefficient

Abstract

Law combines a host of variables and parameters into a bivariate model originally designed to estimate potential output. It is currently used in economic policy analysis as a rule of thumb for relating changes in the unemployment rate to changes in output. Based on the assumption that potential output is realized when the unemployment rate is 4 percent, Okun originally estimated that a change in the actual unemployment rate should approximate one-third the difference between 4 percent and the annual growth rate in output [4; 15]. Given Okun's procedure, changes in his estimate of the sensitivity of output to unemployment (referred to hereafter as the Okun coefficient) have no behavioral significance. All that is required is that the forecaster regularly update his estimate. Attempts have been made to improve Okun's estimate by analyzing the components of potential output and to revise the full employment unemployment rate [2; 9; 11; 17; 26; 29]. Recent efforts have been prompted by the rising female labor force participation rate, a slowdown in the growth of labor productivity since the 1974-75 recession and the sharp increases in the prices of raw materials, particularly oil imports [3; 5; 6; 7; 13; 16; 18; 21; 22; 23]. The components approach drops Okun's restrictive assumption that all linkages between output and unemployment can be captured in a single linear equation. Consequently, this approach implies that the Okun coefficient may vary over time depending on the behavioral characteristics of the particular model used to estimate it. However, studies using the components approach ignore the question of variability. When they report an estimate of the coefficient, it is an average for the sample period. In general, recent estimates reaffirm Okun's original three-to-one estimate through the 1960s. However, estimates for the 1970s show that the value of the coefficient has declined. For example, Tatom [27] used Okun's methodology to determine if the 1973-75 energy crisis altered the value of the coefficient. He found that a 2.22 percent output loss is associated with each 1 percent of unemployment in excess of the full employment level.