Wage Changes, Establishment Growth, and the Effect of Composition Bias

The correlation between real wages and aggregate employment growth has beenthe object of several empirical studies conducted with both aggregate and micro data.Despite the new availability of linked employer-employee data, however, we still havelimited empirical evidence [Belzil, 2000] to describe how real wage cyclicality can beexplained by what happens between workers and employers at the firm level. Thispaper makes a contribution by making use of Italian data to explore whether a posi-tive relationship between wage growth and employment growth is induced mainly byan establishment effect or by an industry effect, at least as long as these effects aremeasured in terms of employment changes.As in the case of wage cyclicality studies, this research takes into account theconsequences that a composition effect can have on the factors that affect the employeeearnings in each establishment. When a firm is growing, as well as in the growthphase of a cycle, new workers enter the job market. They are traditionally low-skillemployees or young people or previously “discouraged” workers. They earn low wages,and so lower the average wage in the firm. This can explain the negative or insignifi-cant correlation between real wages and employment level that has been found inseveral studies conducted at the aggregate level. For the first time, this study teststhe existence of a composition bias with firm-level data where both employment andwage growth can be measured for each establishment. Checking whether employ-ment growth, within firms and within sectors, differently affects the change in thefirm’s average wage or the mean of the individual wage changes does this.This research makes use of 1981-83 records for a sample of Italian firms. Informa-tion about each establishment is combined with information about its employees. Thestudy explores cases in which firms are experiencing an employment decline, an employ-ment increase, or no more than the national rate of labor turnover. The same analy-sis is conducted for categories of workers that, within the same firm, differ because ofjob qualifications or gender.The paper is organized as follows: part 2 presents some of the theories concerningthe relationship between wage changes and employment growth. It also illustratesthe problem of a possible composition bias. Part 3 presents the empirical framework.Part 4 describes the data set used for the estimations. Part 5 discusses the results.The conclusions summarize the major findings.