What Does the Minimum Wage Do?

3 - Hours of Employment

119 3 Hours of Employment Perhaps evidence for the employment response is weak because of the way that employment is being measured—as the number of jobs or percent of people with jobs.1 Employees in minimum wage jobs are paid by the hour. Unless the number of hours worked is the same in all minimum wage jobs, jobs and hours worked are not equivalent measures of employment. Measuring employment by the number of jobs is like measuring the total amount of water in different-sized pitchers by counting the number of pitchers. Instead, the volume of water should be measured directly by adding the amount in each pitcher to get the total. If the number of hours that individual employees work varies, then measure the number of hours. Perhaps rather than reduce staffing in response to the minimum wage, employers should reduce hours of some or all employees. Underlying this hypothesis is another: that up to a point, hiring and firing workers is more difficult or costly than raising and lowering the hours that current employees work. Hiring involves some or all of the following actions: get the word out about openings, review applications to decide whom to interview, interview, check references, evaluate the information acquired from the interview, and make a decision. Further, even if no training is necessary, new hires in many jobs will initially be less productive until they become familiar with the particular workplace and its routines. If employers lay off people when the minimum wage rises, they will lose the value in these implicit training costs as well as any skills developed on the job. Given this, it seems likely that raising and lowering the number of hours individuals work is an easier, less expensive way to adjust the amount of paid labor that is employed (again, up to a point). In the NMWR, Zavodny (2000) was the first to examine whether these two measures of employment (jobs and hours) give different answers to the question, “Does the minimum wage (necessarily) reduce employment?”2 Coming several years prior to Bertrand, Duflo, and Mullainathan (2004), this study does not report a clustered standard, Belman and Wolfson.indb 119 Belman and Wolfson.indb 119 6/16/2014 12:09:09 PM 6/16/2014 12:09:09 PM 120 Belman and Wolfson so we do not include it in the discussion below, but a number of studies followed her lead, including several already discussed in Chapter 2.3 U.S. STUDIES Demographic Groups Youth Table 3.1 shows the five studies that examine the effect of the minimum wage on teenagers’ hours of employment: Orrenius and Zavodny (2008) and Sabia (2009a,b), which use state-level panels; and Allegretto, Dube, and Reich (2009, 2011), which use repeated cross sections of individual -level data. We will consider first the studies of aggregate data. Both Orrenius and Zavodny and Sabia aggregate CPS ORG data to the level of the state; the first use annual data and the second monthly, and both report results for unconditional average usual hours.4 Sabia also reports estimates for conditional usual hours.5 Because a main theme of Sabia’s analysis is the resolution of disagreement about the best way to control for business cycle effects, he reports several results. Orrenius and Zavodny (2008) estimate a conventional equation, with two-way fixed effects (for state and year), the adult male unemployment rate and the teenage share of the population as the control variables for demand and supply conditions, and one or the other of two different minimum wage variables. The first is constructed using a deflator that is common to all states in the same year, and the second uses the average adult wage in each state and year, which therefore has the same endogeneity problems as the Kaitz index (see Chapter 2). On the basis of the common deflator, they report a negative response of the hours of teenage girls to the minimum wage.6 The elasticity of total hours with respect to the minimum wage is −0.31, with a standard error of −0.12.7 Sabia (2009b) presents results for different log-log specifications where the list of regressors always includes state and month dummies, the adult male unemployment rate, the fraction of 16–64-year-olds who are teenagers, and nominal values of the minimum wage and the mean Belman and Wolfson.indb 120 Belman and Wolfson.indb 120 6/16/2014 12:09:11 PM 6/16/2014...