The effect of unions on hours and employment

Abstract

Increases in interest rates or real wages alter labor supply behavior, motivating both employed and unemployed workers to try to work more hours. The outcome is not a unilateral decision, however; firms also play a role. And union power influences these choices. Do firms choose to hire new workers or are they more inclined to increase the average hours of currently employed workers? This paper focuses on how unions affect firms' labor input decisions.Two opposing strands of literature examine how relative union power affects employment decisions. Blanchflower et al. [The Economic Journal, 1991 ] show a negative relationship between the level of union power and the growth rate of the number of employees. Trejo [AER, 1993] shows that firms prefer increasing employment (and not average overtime hours) when changing labor input to avoid paying overtime wages. Firms perceive firing costs lower relative to overtime pay. By generating two ratios of changes in labor input, this paper looks at the relationship between union strength and how the change in total hours is divided between increased new hires versus increased overtime hours. Union density ratios for nine standard industry code groups are calculated by dividing the membership levels by the number of workers employed. The groups' union densities range from 0.41 to 1.30 (some unions report retirees as members). The sample period runs from 1958-78. Unlike some other studies, this approach uses output behavior to gauge relative changes in employment. In addition, the level of disaggregation allows for distinguishing between regular and overtime hours and production workers versus the total number of workers.Specifically, the ratio o calculates how the change in total hours (regular plus overtime) is divided between new workers versus an increase in average total hours. The other ratio, 13, shows the percentage of the change in total hours due to changing overtime hours. Pooling the data for the union groups, a panel data model with both time series and cross sectional observations for each ratio is estimated. For example:6°it : c~it + ~lit qit + ~2itWit + ~3~tdertsitya + ~ta ' where q is output, w is the real wage, and density is union densities for standard industry code groups. For both equations, one would expect a larger portion of the error to be explained by cross sectional effects if unions do affect firms' employment decisions.Results for the equation using co as the endogenous variable produce residuals that are noisy and show that neither unions nor time explain a large portion of the error. For the 11 equation, both cross sections and time explain a larger portion of the error. That is, unions play a greater role in explaining the variance in ~q, lending support to the position that firms prefer to employ more workers and pay less overtime. Still, using F-tests for restricted versus unrestricted versions of the models, one cannot reject the hypothesis that the coefficients are stable across union groups. 225