Regression Analysis of STC
Utilization in 17 States
Table 6B.1 displays results for 17 state-level regression equations typically spanning the years 1989–2010. The dependent variable is annual STC equivalent weeks claimed measured as a percentage of annual weeks in the regular UI program of each state. Each regression equation has three explanatory variables: 1) the current year’s state total unemployment rate (TUR), 2) the TUR lagged one year, and 3) a linear time trend that starts in 1989. Adjacent to each estimated slope coefficient is the absolute value of its t-ratio. The summary measures on the right-hand side of Table 6B.1 are the adjusted R2s, the standard error of estimate, and the Durbin-Watson statistic (DW). The final columns display the mean of the dependent variable and the maximum percentage. The table also identifies four states where STC data were not available from 1989 due to later start dates for the programs (1991 for Iowa and Rhode Island; 1992 for Connecticut and Minnesota).
For most states the fits are quite satisfactory, with adjusted R2s of at least 0.50 for 13 states and standard errors below 0.30 for 15 states. The generally small scale of STC is vividly illustrated by the small means in the right-hand column of Table 6B.1. Rhode Island is the only state where the mean over the full period exceeds 1.0 percent of regular UI claims. Only four other states have means that exceed 0.40 (California, Kansas, Missouri, and Vermont). The small scale of STC is also illustrated by the maximum annual percentages during the estimation period. While most exceed 1.0 percent, only two maxima exceed 2.0 percent, Kansas and Rhode Island at 3.24 and 4.17 percent, respectively.
While the summary measures in Table 6B.1 show the small scale of STC, the regression results point to a pattern of strong cyclical sensitivity. All 17 slope coefficients on the current year TUR are positive and their t-ratios all exceed 2.0, the threshold for statistical significance. The t-ratios in eight states even exceed 5.0; the slopes are highly significant. When unemployment increases, utilization of STC increases relative to utilization of regular UI claims.
The patterns for the lagged TUR coefficients are nearly as consistent. All 17 are negative and 13 have t-ratios of 2.0 or larger. In the year after the TUR increases there is a sharp falloff in STC usage. Short-time compensation usage decreases noticeably in the second year of a recessionary period. This falloff