Employment effects of two northwest minimum wage initiatives.

Singell, Larry D., Jr. ; Terborg, James R.

I. INTRODUCTION

The impact of a minimum wage on employment has long been of interest to economists such as Stigler (1946), because it provides a relatively direct test of a conventional theoretical prediction that a binding increase in the minimum wage should reduce employment among low-skilled workers. Although this theoretical prediction has not gone unchallenged as in Lester (1946), early empirical work summarized in Brown, Gilroy, and Kohen (1983) confirmed theoretical predictions using aggregate time series data for young workers, which indicated that a minimum wage has a small negative employment effect. By 1990, there was general consensus in the profession that minimum wage laws reduced employment among teenage and low-wage workers, albeit by a small amount.

Ehrenberg (1995) and others have argued that subsequent empirical evidence provides less confidence in both the conventional view that the minimum wage lowers teenage or low-skilled employment and that neoclassical theory can account for the relationship between them. For example, Card and Krueger (1995) show in a 30-study meta-analysis that past time series results are sensitive to the specification and suggest that, when appropriately specified, the minimum wage has had no significant effects on employment over the last 25 yr. Alternatively, Neumark and Wascher (1998) find negative employment effects of the minimum wage consistent with structural change using time series data and specifications that do not suffer from the biases present in earlier work. Cross-sectional studies, as in Currie and Fallick (1996), Burkhauser, Couch, and Wittenburg (2000), and Neumark and Wascher (2001), also find mixed evidence of employment effects on low-wage workers due to the minimum wage. Thus, traditional empirical assessments of the minimum wage yield mixed evidence on the employment impact of the minimum wage.

The "natural experiment" methodology was introduced as an alternative to the cross-sectional and time series approaches and exploits both the time series wage variation before and after the treatment of a minimum wage increase and the cross-sectional wage variation between the treatment state and a control state that did not experience a change in the minimum wage. For example, Card and Krueger (1994) analyzed employment growth in 410 fast-food restaurants in New Jersey and Pennsylvania before and after the 1992 minimum wage increase in New Jersey from $4.25 to $5.05. They find no evidence linking the increase in the minimum wage to reduced employment in fast-food restaurants both when comparing New Jersey with Pennsylvania and when comparing impacted versus nonimpacted restaurants in New Jersey. Neumark and Wascher (2000) reexamine the impact of the 1992 New Jersey minimum wage increase using payroll data for hours worked by fast-food employees and Bureau of Labor Statistics (BLS) eating and drinking employment data and find that the minimum wage significantly reduced employment in New Jersey versus Pennsylvania. Thus, even studies that apply similar natural experiment techniques and data for a given minimum wage treatment, as in Katz and Krueger (1992), Deere, Murphy, and Welch (1995), Kim and Taylor (1995), and Card and Krueger (2000), can yield different conclusions regarding its employment effect.

Our empirical analysis exploits a unique natural experiment initiated by Oregon and Washington voter initiatives that raised the minimum wage over three successive years (1997-1999 in Oregon and 1999-2001 in Washington) by approximately 37% in both states. Following prior work, we focus on the eating and drinking industry (SIC 58), which is a low-wage industry that has been extensively studied in prior work due to an expectation of a binding minimum wage. BLS wage data available between 1997 and 2001 indicate that the successive minimum wage increases become increasingly binding for most, but not all, eating and drinking jobs at the lower end of the wage distribution. Thus, this article addresses some of the criticisms of the natural experiment methodology by Hamermesh (1995) that stem from the short duration in the measured effects and those levied by Freeman (1995) due to a focus on relatively small and nonbinding minimum wage increases.

The empirical analysis uses monthly BLS employment data to estimate several regressions that show that the northwest minimum wage initiatives lowered employment growth in Oregon and Washington. In addition, the negative employment effects are shown to be robust to the inclusion of the lag of the minimum wage and insensitive to specification tests regarding the possible exclusion of unobserved time-varying state-specific effects. Thus, contrary to prior work using eating and drinking data including those by Card and Krueger (1995) and Neumark and Wascher (2000), the empirical results consistently indicate negative employment effects and employment growth effects arising from the minimum wage.

To examine whether the negative employment effects arise in other low-wage industries, the employment growth specifications are replicated using data for the hotel and lodging industry (SIC 70). BLS wage data indicate that jobs in this industry pay slightly higher wages that are less impacted by the minimum wage than the eating and drinking industry. Consistent with a less binding minimum wage, several employment growth models yield mixed evidence for an employment effect of the minimum wage in the hotel and lodging industry. Thus, the results support findings by Yuen (2003) and Neumark, Schweitzer, and Wascher (2004) that suggest that the employment effects of the minimum wage are sensitive to the wage distribution in the industry prior to the introduction of the minimum wage that determines the extent to which the minimum wage binds.

A final set of analyses uses unique data for want ads collected from the Portland Oregonian and the Seattle Times for specific eating and drinking and hotel and lodging jobs over the same period as our employment data. Want-ad regressions indicate that the minimum wage initiatives reduced the amount of job vacancies (and related hiring efforts), particularly for those jobs for which the minimum wage is relatively binding. Jointly, the empirical results provide some of the first formal evidence that the minimum wage is a blunt instrument that yields different employment effects within and between low-wage industries.

Section II describes the minimum wage initiatives in Oregon and Washington, descriptively examines the wage distribution in the eating and drinking industry, and develops and presents the employment results for the eating and drinking industry. Section II also descriptively examines the wage distribution in the hotel and lodging industry and tests the sensitivity of the employment effects of the minimum wage using the eating and drinking specifications and data for the hotel and lodging industry. Section III describes the want-ad data and presents the findings for the want-ad regressions. The final section concludes.

II. EMPLOYMENT IN OREGON AND WASHINGTON

A. Wages and Employment in the Eating and Drinking Industry

The empirical analysis uses monthly employment data for the eating and drinking industry between 1994 and 2001 for Oregon and Washington. In Oregon, the minimum wage remained constant at $4.75 an hour between 1994 and 1996, but a November 1996 referendum approved an increase in the minimum for three successive years to $5.50, $6.00, and $6.50 starting in January 1997. Alternatively, in Washington, the minimum wage was held constant at $4.90 from 1994 to 1998, while a November 1998 referendum raised the minimum wage for three successive years to $5.70, $6.50, and $6.72 starting in January 1999. Thus, the employment data span a time period when the minimum is constant in both states (1994-1996), rising in Oregon but not in Washington (1997-1998), rising in both states (1999), rising in Washington but not in Oregon (2000-2001). (1)

Oregon and Washington are natural comparators to evaluate the employment impact of a minimum wage change because these northwestern states share similar economies and institutions. In particular, the employment demand for the eating and drinking industry in the northwest has been found to depend primarily on gross state population and income growth, which do not differ statistically over the period of the data (Washington Department of Labor and Industries Report, 1999). (2) The close proximity, shared regional economies, and similar minimum wage history make Oregon and Washington the most suitable control states for each other among the alternatives and provide a unique natural experiment regarding the employment effects of the minimum wage on the eating and drinking industry.

Prior studies testing for employment effects of the minimum wage focus on the eating and drinking industry because it is a low-wage industry where the minimum might be expected to bind. Evidence for the presence of a binding minimum wage in Oregon and Washington over the 1997-2001 period is presented in Table 1 using two-digit job-level wage survey data evaluated at the median, 10th, and 90th percentiles, which are collected by the BLS starting in 1997. (3) For the period 1997-1999 when the Oregon minimum wage is increasing, Table 1 shows a higher 10th-percentile and median wage in Oregon versus Washington for four of the five eating and drinking jobs, with a growing wage gap for each successive increase in the Oregon minimum wage. Cooks are the exception, where the minimum wage is well below the 10th-percentile wage and does not appear to bind. However, Washington's 90th-percentile wages are higher for three of the five jobs over this period, indicating that its lower pay in the left tail of the wage distribution is not the result of lower pay for all points in its wage distribution relative to Oregon.

To put in another way, excluding cooks in sit-down restaurants, 10th-percentile wages are no more than 11 cents from the minimum wage for the other restaurant jobs in Oregon between 1997 and 1999, whereas they are no less than 28 cents higher than the minimum in Washington (even after its 1999 increase in the minimum). On the other hand, over the period between 2000 and 2001 when the minimum wage is rising in Washington but not in Oregon, 10th-percentile wages are within 10 cents of the minimum wage for Washington. However, in Oregon, wages are no less than 25 cents higher than the Oregon minimum by 200l. Thus, for eating and drinking jobs, the minimum wage appears to be binding at the lower tail of the wage distribution in the northwestern state that has an increasing minimum wage. (4)

Figure 1 plots the average annual employment in the eating and drinking industry for Oregon and Washington and presents the average growth rates for the period 1994-1997 (i.e., the period prior to the minimum wage increases), 1997-1999 (i.e., the period of the Oregon minimum wage increases), and 1999-2001 (i.e., the period of the Washington minimum wage increases). The figure demonstrates that employment in the eating and drinking industry is growing in both Oregon and Washington prior to 1997. However, between 1997 and 1999, employment growth in Oregon stagnates at less than 1%, whereas employment growth in Washington increases to nearly 3% over the same period. However, starting in 1999, employment growth in Washington slows to just over 1%, whereas concurrently Oregon experiences an increase in employment growth (particularly in 2000 and 2001). Of course, it is possible that the post- 1997 employment growth differences between the two states correspond to changes in broad economic trends or other economic factors. It follows that the subsequent empirical analysis includes controls for time-varying economic conditions and state-specific fixed effects and tests for presence of unobserved (and serially correlated) state-specific economic conditions.

[FIGURE 1 OMITTED]

B. Empirical Models of Employment and Employment Growth

Prior work, such as Neumark and Wascher (1992) and Katz and Krueger (1992), has examined the employment impact of the minimum wage by estimating models of employment growth. Thus, as a point of departure, we estimate a monthly employment growth ([DELTA]ln [E.sub.t]) specification of the form:

(1) [DELTA]ln [E.sub.t], = [[beta].sub.0] + [[beta].sub.1] [X.sub.t] + [[beta].sub.2] [DELTA]ln([m.sub.[tau]]) + [[epsilon].sub.t].

The vector [X.sub.t] includes an Oregon fixed effect, the monthly growth rate in population and per capita income for each state, a cubic time trend, and monthly dummies and their interaction with the Oregon binary variable. It follows that the empirical model controls for time-varying economic factors (observed and unobserved) as well as the seasonality observed in employment data. In addition, [DELTA]ln([m.sub.[tau]] is the annual first difference in the log of the real minimum wage measured in 1994 (Consumer Price Index-adjusted) dollars. It follows that [tau] denotes an annual difference for the minimum wage measure, where t denotes a monthly difference for both the dependent variable and the other explanatory variables. (5) As specified, a negative sign on [[beta].sub.2] would support the hypothesis that the minimum wage yields a negative employment impact. (6)

The specification in Equation (1) models an immediate employment response to the minimum wage. However, Brown (1982) argues that hiring and training costs or limited adjustability of nonlabor inputs could yield a lagged employment effect from an increase in minimum wage, whereas the higher turnover rate among low-wage workers may work against finding a lagged response. Neumark and Wascher (1992) show that the omission of a lagged minimum wage variable can yield specification error in first-difference employment growth models, particularly when the estimation focuses on the short-term effect of the minimum wage. Thus, we also modify the specifications in Equation (1) to include the 1-yr lag of the minimum wage variable, [DELTA]ln([m.sub.[tau]] - 1):

(2) [DELTA]ln [E.sub.t], = [[beta].sub.0] + [[beta].sub.1] [X.sub.t] + [[beta].sub.2] [DELTA]ln ([m.sub.[tau]]) + [[beta].sub.3] [DELTA]ln ([m.sub.[tau]) + [[epsilon].sub.t].

in order to examine the possible biases in the estimates from a failure to model the dynamics of the minimum wage employment response.

Specification bias is an important concern with the natural experiment methodology that typically uses a relatively limited set of time-varying state-specific controls (e.g., the population and personal income growth rate) and a state-specific dummy to control for employment differences between states. Specification issues are particularly important in evaluating the minimum wage because Neumark and Wascher (1992) demonstrate that if the time-varying omitted state effect is positively serially correlated with the minimum wage, the employment effect of the minimum wage may be negatively biased.

To explore this possibility, we conduct a specification test proposed in Heckman and Hotz (1989) that examines whether Equations (1) and (2) adequately control for state-specific (time-varying) employment heterogeneity. (7) The test entails reestimating the models including the dependent variable from the period prior to that over which the empirical analysis is estimated. In particular, we include the employment growth measured between 1992 and 1993. If the earlier level of the dependent variable enters significantly into the specification, it is likely that there is an omitted time-varying state effect such that the empirical model of the eating and drinking industry is misspecified.

C. Empirical Results for Eating and Drinking

The employment data for the empirical analysis are drawn from the BLS Employment Series for the eating and drinking industry (SIC 58). These data are used to estimate the employment growth specifications in Equations (l) and (2), presented in Table 2. The results in Table 2 show that the empirical model explains between 65% and 69% of the variation in employment growth. In addition, each of the explanatory variables is significant at traditional levels for at least one of the specifications and, when significant, has the expected sign. For example, the coefficients on the population and personal income growth measures are both positive and significant, indicating that employment growth in the eating and drinking industry tends to rise with personal income and/or population growth. Thus, for brevity, the discussion focuses on the minimum wage variables that are of primary interest.

Models 1 and 2 in Table 2 present the growth regressions including the change in the log of the minimum wage. Following Heckman and Hotz (1989), Model 2 differs from Model 1 by including the change in the log of employment between 1993 and 1992, which has an insignificant coefficient, suggesting that Model 1 does not suffer from an omitted time-varying state effect. The coefficients on the minimum wage variables in both specifications are positive and significant, indicating that increases in the minimum wage in Oregon and Washington lowered employment growth. In particular, the coefficient on the minimum wage variable predicts that a 1% growth in the minimum wage reduces employment growth by 0.07%. Oregon and Washington's minimum wage initiatives raised the minimum wage by approximately 37%. Thus, although the coefficient on the minimum wage variable is small, it implies a 2-percent-point reduction in growth due to policy change, which roughly corresponds to the observed reduction in the average growth rates presented in Figure 1.

Models 3 and 4 in Table 2 present the growth regressions including the change in the log of the minimum wage along with its lagged value. Again, following Heckman and Hotz (1982), Model 3 differs from Model 4 by including the change in the log of employment between 1993 and 1992, which has an insignificant coefficient--again suggesting that Model 3 does not suffer from an omitted time-varying state effect. The lag specifications indicate a negative and significant coefficient for both the difference in the log of the minimum wage and its lag, suggesting that the negative employment growth effect rises when the eating and drinking industry has more time to respond. In fact, the joint contemporaneous and lag effects imply approximately a 0.2% reduction in employment growth for each 1% increase in the minimum wage, which is more than double the impact predicted in the contemporaneous specifications in Models 1 and 2. Thus, collectively, the results suggest that the northwest minimum wage initiatives yielded an economically meaningful reduction in employment growth in the eating and drinking industry.

It is important to emphasize that the results are robust to a variety of alternative specifications. For example, Oregon voters also passed a ballot initiative in 1997 that raised the earned income tax credit and could have mitigated the impact of the Oregon minimum wage law. However, a specification that includes an interaction of the Oregon binary variable with a dummy variable that equals 1 for the post-1997 period (not presented) does not change the sign, significance, or the magnitude of the minimum wage coefficients presented in Table 2. Likewise, specifications that include the nominal rather than real minimum wage yield qualitatively similar conclusions. Finally, specifications estimated in levels rather than in differences yield negative and significant coefficients on both the level of the minimum wage and its lag, contrary to prior work by Card and Krueger (1994) and Neumark and Wascher (1998). Thus, the findings are robust to a variety of alternative specifications.

In addition, the findings do not appear to be driven by the presence of state-specific serial correlation in the errors. For example, Durbin's test from Oregon- and Washington specific regressions using specifications in Table 2 yields a chi-square statistic that rejects the presence of serial correlation at the 1% level. The model is also estimated permitting the errors to be correlated each trimester, quarter, or semiannually to examine for the possibility that general forms of time series trends are present in the data that are not necessarily associated with omitted variables. However, both the qualitative and the statistical conclusions do not differ from those presented in any of these cases and indicate no remaining autocorrelation in the error beyond that controlled for in the cubic time trend and the state-specific monthly dummies. It follows that the apparent negative employment effect of the minimum wage in the eating and drinking industry in the northwest does not appear to be a statistical artifact of the particular specification adopted or assumptions regarding the error structure.

On the other hand, the above results also do not necessarily suggest that the observed negative employment effect of the minimum wage is wide spread for other low-paying industries in these states. In fact, the descriptive evidence suggests that, even in the eating and drinking industry, there is a nonbinding minimum wage for a significant portion of the wage distribution. Thus, the subsequent analysis examines whether the estimated employment impact depends on the choice of industry and the extent to which the minimum wage binds. In particular, the eating and drinking analysis is replicated using BLS data for the hotel and lodging industry (SIC 70) over the same time period, which we turn to next.

D. Empirical Results for the Hotel and Lodging Industry

The hotel and lodging industry is a low-wage industry that pays slightly higher wages than those paid in the eating and drinking industry. For example, in 1994, the average annual wage in the eating and drinking industry was $9,252 in Oregon and $9,401 in Washington that, respectively, compares to $11,409 and $13,150 in the hotel and lodging industry. However, just like the eating and drinking industry, the hotel and lodging industry hires workers in a variety of jobs that pay distinctly different wages. Thus, to understand whether the minimum wage might bind for the industry as a whole (as well as in particular jobs within the industry), it is useful to examine pay across several jobs in the hotel and lodging industry.

Table 3 summarizes the median, 10th-, and 90th-percentile wages for three hotel and lodging jobs between 1997 and 2001 when the minimum wage was rising in Oregon or Washington. For 1997, the wage data show that 10th-percentile wages, although within 15 cents of the minimum for maids and housekeepers in Oregon, are at least 32 cents higher than the minimum for the other jobs. By 1999, the minimum wage appears to be relatively binding in Oregon as its 10th-percentile wage is within 5 cents of the minimum for maids and housekeepers and within 16 cents for hotel desk clerks, whereas this difference is 51 and 75 cents, respectively, for these same jobs in Washington. On the other hand, unlike in Oregon's eating and drinking industry, the gap between the minimum and 10th-percentile wages shows an increase in the hotel and lodging industry after 1999. For example, for 2001, the 10th-percentile wage in Oregon is 38 cents higher than the minimum for maids and housekeepers and 20 cents higher in Washington, with the difference becoming even larger for other jobs. Thus, overall, there is less compelling evidence of a binding minimum wage in the hotel and lodging industry in comparison to the eating and drinking industry.

Descriptive evidence with regard to employment also suggests that the minimum wage was not binding in the hotel and lodging industry. For example, Figure 2 plots the average annual employment in the hotel and lodging industry and reports the employment growth rate for the period prior to the minimum wage increases (1994-1997), the period of the Oregon minimum wage increases (1997-1999), and the period of the Washington minimum wage increases (1999-2001). For Oregon, employment growth continued at a rate of 2.7% from 1994 through 1999 and slowed to less than 1% after 1999. In Washington, employment growth in the hotel and lodging industry was relatively modest in comparison to that in Oregon and slowed well prior to the increase in the state's minimum wage in 1999. Thus, the descriptive evidence regarding the hotel and lodging industry, unlike like that found for eating and drinking jobs, does not suggest a deleterious effect from the minimum wage.

[FIGURE 2 OMITTED]

Table 4 replicates the specifications presented in Table 2 for the eating and drinking industry using the hotel and lodging data. Most of the coefficients on the nonminimum wage variables are similar in sign to those presented in Table 2. For example, the Oregon binary variable is negative and significant indicating a differential employment trend between Oregon and Washington, whereas the coefficient on population growth in the state is positive and significant suggesting that hotel and lodging demand is positively related to population growth. Alternatively, the coefficient on personal income growth is negative (and significant in Models 1 and 2); thus, collectively the results suggest that increases in wealth lead people to eat out more at local establishments but travel more frequently out of state. Again, similar to the eating and drinking results, the coefficient on the change in the log of employment between 1993 and 1992 is insignificant in Models 2 and 4, indicating that specifications pass the Heckman-Hotz test for an omitted time-varying state effect. Thus, the hotel and lodging and the eating and drinking industries do not appear to respond distinctly to the time-varying controls.

On the other hand, the coefficients on the minimum wage variables in the hotel and lodging industry presented in Table 4 differ distinctly from those regarding the same specifications in the eating and drinking industry presented in Table 2. First, contrary to the minimum wage findings, the coefficient on the difference in the log of the real minimum wage is positive and significant, suggesting that a rise in the minimum wage is positively associated with employment growth in the hotel and lodging industry. The positive coefficient is consistent with efficiency wage and monopsony power models advanced in the literature, as in Wessels (1997). However, the fact that the descriptive evidence suggests that the minimum wage was not binding in the hotel and lodging industry may well suggest that voters simply pass legislation when "times are good" such that the cost of the minimum wage (at least for industries where the minimum wage is not strongly binding) is relatively low. Second, the coefficient on the lag of the minimum wage is insignificant in Models 3 and 4 (although the joint effect of the minimum wage and its lag is significantly positive). This finding is consistent with the fact that a nonbinding minimum wage would not be expected

to yield a lagged employer response. Collectively, the findings provide inconclusive or inconsistent evidence regarding the employment effects of the minimum wage and suggest that the employment effect of the minimum wage is sensitive to the choice of industry and the extent to which the minimum wage is observed to bind. (8)

III. WANT ADS IN OREGON AND WASHINGTON

A. The Want-Ad Data

The employment results suggest that the impact of the minimum wage depends on the extent to which the minimum wage binds in an industry. The wage data for the eating and drinking and the hotel and lodging industries also suggest that wages differ distinctly across jobs within the same industry. Thus, one might also expect that the employment impact of the minimum wage would differ across jobs within the same industry reflecting that the minimum wage binds at different points on the job-specific wage distribution. Nonetheless, the BLS did not begin collecting. job-specific survey data for the eating and drinking or the hotel and lodging industries until 1997. It follows that BLS data do not span the pre-- versus post-minimum wage periods and, therefore, cannot be used to examine job-specific responses to the minimum wage.

As an alternative, we study the potential differential employment effect of the minimum wage on eating and drinking and hotel and lodging jobs using data from the "Help Wanted" sections of the Portland Oregonian and the Seattle Times. Specifically, want ads are collected for the second Sunday of each month starting in January 1994 and ending in December 2001. (9) Want-ad listings are counted for the following eating and drinking jobs: waitpersons (excluding counter persons), bus persons (including server assistants), dishwashers, cooks (includes all cook positions other than chef or chef/manager), and host persons (including greeter). (10) In addition, to provide a point of comparison for low-wage jobs in the eating and drinking industry, want-ad data are also collected for hotel housekeepers. Tables 1 and 3 suggest that hotel housekeepers earn comparable wages to many workers in eating and drinking jobs. Thus, in total, there are listings for five specific jobs in the restaurant industry and one job in the hotel and lodging industry over 96 mo from 1994 through 2001 in Portland, OR, and Seattle, WA.

Table 5 summarizes the want-ad data by year and job category, which support the employment findings. For example, in Oregon, the initial increase in the minimum wage in 1997 does not appear to have a large direct effect on the average number of want ads, consistent with the relatively strong lag effect found for employment. Specifically, in Oregon, the average number of want ads declined for bus persons and cooks, but actually grew in the other job categories. However, the growth in the number of want ads for Washington is substantially larger for each of the job categories between 1996 and 1997. In addition, the employment impact of the Oregon minimum wage legislation appears larger in 1998 and 1999. For example, the average annual number of want ads is at or below their 1996 level in all eating and drinking job categories for Oregon in 1998 and 1999, whereas the average annual number of want ads increased for those same jobs in Washington between 1996 and 1999.

In Washington, the increase in the average number of want ads for all eating and drinking jobs in 1999 (except dishwashers) and the subsequent decline in the average number of want ads in 2000 and 2001 suggest that the Washington minimum wage law lowers employment with a lag. Interestingly, however, the average number of want ads in Oregon also declines for all jobs in 2001, which reflects the marked post--September 11th reduction in want ads and related recession that reduced employment in all industries. (11)

Table 5 also summarizes the want-ad data for hotel housekeepers. Similar to the descriptive evidence for the eating and drinking industry, the number of want ads declines for hotel housekeepers in 1998 and 1999 for Oregon and in 2000 and 2001 for Washington, but by a smaller percentage than that for each of the five eating and drinking jobs in both states. Thus, the effect of the minimum wage on the number of want ads appears to be relatively larger in the eating and drinking industry that pays slightly lower wages than that for hotel housekeepers.

B. Want-Ad Regressions

Following the employment specifications, the empirical analysis examines whether the apparent reduction in the number of want ads for jobs in the eating and drinking industry and for hotel housekeepers remains after controlling for state-specific differences and time-varying cyclical effects. However, because want ads indicate job vacancies that reflect a desire to change the stock of employment (i.e., want ads are a measure of a desired employment flow), the empirical analysis focuses exclusively on the level of want ads as opposed to the employment regressions that are estimated in log differences to measure employment growth. However, the dependent variable is scaled by the employment level in the industry to reflect the fact that it will require a larger level of want ads just to maintain a fixed level of employment. Thus, the dependent variable is measured as the number of monthly want ads divided by the monthly employment level in either the eating and drinking or the hotel and lodging industry.

The want-ad regressions are presented in Tables 6 and 7. Following the employment growth models, the specifications presented in Table 6 focus exclusively on the contemporaneous effect of the (real) minimum wage, and the specifications in Table 7 include both the contemporaneous and 1-yr lag of the minimum wage. The results in Tables 6 and 7 indicate that the empirical models explain between 28% and 78% of the variation in the log number of want ads with the coefficients on each of the explanatory variables significant in at least one specification. For brevity, the discussion focuses on the results for the minimum wage coefficients, which are of primary interest. In addition, there is no further discussion of tests for misspecification other than to note that each specification passes the Heckman-Hotz test as indicated in Tables 6 and 7.

The coefficients on the log of the minimum wage are significant at traditional levels in each of the specifications presented in Table 6 with the exception of host staff, indicating that a 1% increase in the minimum wage reduces the number of want ads by between 1.0% and 4.7%. With the exception of host staff, the relative magnitudes on the log of the minimum wage coefficients are consistent with the largest percentage want-ad reductions occurring for those jobs where the minimum wage is relatively binding. Thus, among the eating and drinking jobs significantly impacted by the minimum wage, cooks experience the smallest percentage decrease in the number of want ads, whereas dishwashers experience the largest percentage decrease in the number of want ads.

Further, consistent with the employment findings, housekeepers experience a smaller percentage reduction in the number of want ads than that for all of the eating and drinking jobs except host persons. The smaller minimum wage impact on housekeepers (who are comparably paid to many eating and drinking workers) may reflect that some of the reduction in employment for higher paid jobs in the eating and drinking industry (e.g., cooks) arises from the fact that the minimum wage is binding for a greater percentage of jobs in the industry, which depresses overall employment even for jobs that may not directly be impacted by the minimum wage. For example, Terborg (1999) finds that the minimum wage induced restaurant managers to raise wages across the broad spectrum of jobs due to internal equity concerns. The cost pressures arising from such internal equity concerns are likely to be larger for industries where a greater percentage of workers are affected by the minimum wage. This may be particularly important in the eating and drinking industry where inputs may be relatively more substitutable across jobs. Overall, the want-ad results support the employment findings that the minimum wage markedly lowered employment for the eating and drinking industry in Oregon and Washington, whereas it had a decidedly smaller effect in the hotel and lodging industry.

However, contrary to the findings for employment growth, the results in Table 7 yield small and insignificant coefficients on the log of the lag of the minimum wage in both the eating and drinking industry and the hotel and lodging industry. Thus, while the employment regressions suggest that stock of employment responds with a lag, the flow of employment as reflected in want ads appears to respond immediately. On the other hand, joint significance tests conducted on the log minimum of the minimum wage and its lag indicate a negative employment effect at the 10% level or higher for all jobs except host staff and housekeepers with an implied elasticity that ranges between 2.5 and 5. Thus, the findings suggest an immediate elastic reduction in hiring efforts as a result of an increase in the minimum wage as opposed to the inelastic response found for employment growth. This finding may reflect that reductions in employment are achieved in these low-skilled industries through attrition, whereas hiring through want ads is more directly related to maintaining or growing employment. In any case, employment responses to a binding increase in the minimum wage may take time because reduced hiring efforts, although adopted immediately, do not immediately impact the actual level of employment.

IV. CONCLUDING REMARKS

Recent minimum wage studies have used the natural experiment approach that compares employment in low-wage industries before and after a one-period increase in the minimum wage and conditioned on employment in a similar state that does not enact a minimum wage change. These studies have sparked a heated debate among economists because the empirical evidence regarding the expected negative employment effect of a minimum wage is decidedly mixed despite the use of similar techniques and data. This article extends prior work by examining whether the observed mix of evidence could arise from using a relatively small one-time increase in the minimum wage or from pooling across low-skill industries or jobs that are differentially affected by the minimum wage.

The empirical analysis exploits a unique natural experiment from two voter initiatives in Oregon and Washington that raise the minimum wage in each state by approximately 37% over three successive years. The timing of the minimum wage increases permits us to examine the employment effect of the minimum on the eating and drinking and the hotel and lodging industries over a period between 1994 and 2001 when the minimum is constant in both states (1994-1996), rising in Oregon but not in Washington (1997-1998), rising in both states (1999), and rising in Washington but not in Oregon (2000-2001). BLS wage data available between 1997 and 2001 indicate that the successive minimum wage increases become increasingly binding for most eating and drinking jobs, but not for hotel and lodging jobs, at the lower end of the wage distribution.

Consistent with a binding minimum wage, employment growth specifications using BLS data indicate that the minimum wage uniformly generates a negative employment effect for the eating and drinking industry. However, comparable specifications for the hotel and lodging industry indicate either an insignificant or even a positive employment effect, suggesting that the minimum wage is not binding in a low-skill industry that pays marginally higher wages than those in the eating and drinking industry. Nonetheless, when binding, the magnitude of the employment effect of the minimum wage appears to be nontrivial. For example, our employment growth specifications imply that the Oregon and Washington minimum wage legislation, which raised the minimum wage by an average of 37%, reduced employment growth by between 2 and 7 percentage points in the eating and drinking industry. Such reductions in employment growth imply that the minimum wage could potentially have large cumulative effects on employment over time.

The conclusions regarding the employment effects are shown to be robust to the inclusion of lag values of the minimum wage and insensitive to specification bias arising from the possible exclusion of unobserved time-varying state-specific effects. Moreover, analyses that use monthly want-ad data collected from the Portland Oregonian and the Seattle Times for specific eating and drinking and hotel and lodging jobs between 1994 and 2001 indicate that increases in the minimum lead to reduced job vacancies (and related hiring efforts) particularly for those jobs for which the minimum wage is relatively binding. Thus, overall, these findings suggest that some of the mixed evidence regarding the employment impacts of the minimum wage may result from the fact that it is a blunt instrument that differentially affects low-wage workers within and across industries. It follows that public policy must be sensitive to the possible trade-offs between a rise in the minimum wage and its employment effects that likely vary across types of workers, jobs, and industries.

ABBREVIATIONS

BLS: Bureau of Labor Statistics

OES: Occupation and Employment Statistics

doi: 10.1111/j.1465-7295.2006.00018.x

REFERENCES

Brown, C. "The Effect of the Minimum Wage on Employment and Unemployment." Journal of Economic Literature, 20, 1982, 487-528.

Brown, C., C. Gilroy, and A. Kohen. "Time-Series Evidence of the Effect of the Minimum Wage on Youth Employment and Unemployment." Journal of Human Resources, 18, 1983, 3-31.

Burkhauser, R. V., K. A. Couch, and D. C. Wittenburg. "A Reassessment of the New Economics of the Minimum Wage Literature with Monthly Data from the Current Population Survey." Journal of Labor Economics, 18, 2000, 653-80.

Card, D., and A. B. Krueger. "Minimum Wages and Employment: A Case Study of the Fast-Food Industry in New Jersey and Pennsylvania." American Economic Review, 84, 1994, 772-93.

--. Myth and Measurement: The New Economics of the Minimum Wage. Princeton: Princeton University Press, 1995.

--. "Minimum Wages and Employment: A Case Study of the Fast-Food Industry in New Jersey and Pennsylvania: Reply." American Economic Review, 90, 2000, 1397-420.

Currie, J., and B. D. Fallick. "The Minimum Wage and the Employment of Youth." Journal of Human Resources, 31, 1996, 404-48.

Deere, D., K. M. Murphy, and F. Welch. "Reexamining Methods of Estimating Minimum Wage Effects: Employment and the 1990-1991 Minimum Wage Hike." American Economic Association Papers and Proceedings, 85, 1995, 232-7.

Ehrenberg. R. G. "Review Symposium: Myth and Measurement: The New Economies of the Minimum Wage, by David Card and Alan B. Krueger." Indus-trial and Labor Relations Review, 48, 1995, 827-8.

Fraker, T., and R. Maynard. "Evaluating Comparison Group Designs with Employment-Related Programs." Journal of Human Resources, 22, 1987, 199-227.

Freeman, R. B. "Review Symposium: Myth and Measurement: The New Economics of the Minimum Wage. by David Card and Alan B. Krueger." Industrial and Labor Relations Review, 48, 1995, 830-34.

Hamermesh, D. S. "Review Symposium: Myth and Measurement: The New Economies of the Minimum Wage, by David Card and Alan B. Krueger." Industrial and Labor Relations Review, 48, 1995, 835-8.

Heckman, J. J., and V. J. Hotz. "Choosing among Alternative Nonexperimental Methods for Estimating the Impact of Social Programs: The Case of Manpower Training." Journal of the American Statistical Association, 84, 1989, 862-74.

Katz, L. F., and A. B. Krueger. "The Effect of the Minimum Wage on the Fast-Food Industry." Industrial and Labor Relations Review, 46, 1992, 6-21.

Kim, T., and L. J. Taylor. "'The Employment Effect in Retail Trade of California's 1988 Minimum Wage Increase." Journal of Business Economies and Statistics, 13, 1995, 175-82.

LaLonde, R. "Evaluating the Econometric Evaluations of Training Programs with Experimental Data." American Economic Review, 76, 1986, 604-20.

Lester, R. S. "Shortcomings of Marginal Analysis for Wage-Employment Problems." American Economic Review, 36, 1946, 62-82.

Neumark, D., M. Schweitzer, and W. Wascher. "Minimum Wage Effects Throughout the Wage Distribution." Journal of Human Resources, 39, 2004, 425-50.

Neumark, D., and W. Wascher. "Employment Effects of Minimum and Subminimum Wages: Panel Data on State Minimum Wage Laws." Industrial and Labor Relations Review, 46, 1992, 55-81.

--. "Is the Time-Series Evidence on Minimum Wage Effects Contaminated by Publication Bias?" Economic Inquiry, 36, 1998, 458-70.

--. "Minimum Wages and Employment: A Case Study of the Fast-Food Industry in New Jersey and Pennsylvania: Comment." American Economic Review, 90, 2000, 1362-96.

--. "The Employment Effects of Minimum Wages: Evidence from a Prespecified Research Design." Industrial Relations, 40, 2001, 121-44.

O'Connor, P., and A. Ayre. "Oregon's Minimum Wage Rises." Oregon Labor Market Information System, 17 January 2003, 1-8.

Stigler, G. J. "The Economics of Minimum Wage Legislation." American Economic Review, 36, 1946, 358-65.

Terborg, J. R. "The Impact of Increases in the State Minimum Wage on the Oregon Restaurant Industry: 1997 to 1999," Eugene, Oregon: Lundquist College of Business, University of Oregon, 1999, 1-16.

Wessels, W. J. "Minimum Wages and Tipped Servers." Economic Inquiry, 35, 1997, 334-40.

Yuen, T. "The Effect of Minimum Wages on Youth Employment in Canada: A Panel Study." Journal of Human Resources, 38, 2003, 647-72.

(1.) Over the period of the data, neither state has a tip-wage credit, which permits employers to pay subminimum wages to those employees who earn tips. Given the focus on restaurant workers, the absence of a tip-wage credit is important because it implies that all workers are subject to the minimum regardless of the prevalence of tips for certain jobs in the industry.

(2.) Population grew at an average annual rate of 1.6% in both Oregon and Washington over the period between 1994 and 2001, whereas per capita income grew at a rate of 2.7 and 2.8% in Oregon and Washington, respectively.

(3.) The Occupation and Employment Statistics (OES) survey is a semiannual mail survey measuring occupational employment and wage rates for wage and salary workers in nonfarm establishments in the United States. OES estimates are constructed from a sample of 1.2 million establishments. Forms are mailed to about 200,000 establishments in May and November of each year for a 3-yr period. Response rates are generally high. For example, the nationwide response rate for the November 2004 survey was 78.7% for establishments, covering 73.0% of employment. Additional information is provided on the BLS Web site at http://www.bls.govloeslcurrentloes_tec. htm.

(4.) Prior studies have found that the minimum wage applies to a relatively large fraction of the workforce in the eating and drinking industry. For example, O'Connor and Ayre (2003) estimate that 51% of eating and drinking workers earn an hourly wage at the minimum. Terborg (1999), in a survey of 189 Oregon restaurant owners and managers, finds that the average restaurant experienced a 32% increase in total wages and benefits over a 26-mo period, in part, because internal equity concerns lead to increases in wages and benefits paid to workers who earned more than the minimum wage.

(5.) The annual difference in the log of the minimum wage is used rather than the monthly difference because the three successive annual minimum wage increases that were passed in both Oregon and Washington apply for each month over an entire year. This permits each month in a year without a minimum wage change to be compared to each month in year with a minimum wage change.

(6.) We use logs to focus unit-free employment changes that are directly comparable across specifications rather than the difference in levels used in Card and Krueger (1994) or the employment to population ratio used in Neumark and Wascher (1992).

(7.) Concern regarding misspecification and the validity of nonexperimental control groups can arise from a number of sources. For example, LaLonde (1986) and Fraker and Maynard (1987) find that nonexperimental control groups can be problematic in evaluating manpower training programs because participants sell-select into the program. This type of selection is unlikely to be an issue for referendum-initiated minimum wage law, because neither low-wage workers nor the business owners that employ them are sufficiently numerous to determine the initiative process or the election outcome (particularly since the minimum wage was one of a number of issues that was on the ballot in the two states). Thus, the marginal voters in Oregon and Washington are unlikely to be in either treatment groups or to be markedly different from their demographically similar sister state.

(8.) The minimum wage coefficients are sensitive to the nominal versus real distinction in the hotel and lodging industry. Specifically, the nominal minimum wage specification yields a positive and significant coefficient on the change in the log of the minimum wage and an offsetting negative coefficient of equal magnitude on the lag of the minimum wage. An equal and offsetting effect for the nominal minimum wage is also found in Wessels (1997) that uses restaurant data from the Census of Retail Trade to test a monopsony power model, which predicts that a higher minimum wage can yield a "reverse-C" pattern such that employment can increase over some ranges of the minimum wage.

(9.) Portland and Seattle are relatively comparable cities that comprise approximately two thirds of their state's populations. For example, in 1999, the Portland metro area contained roughly 2 of the 3 million residents of Oregon. whereas the Seattle metro area contained 5 of the roughly 8 million residents in Washington.

(10.) To insure reliability of the data, two coders independently examined all 12 mo of 1994 for both newspapers. The coders had the same number for 118 observations out of a possible 120, or 98% agreement. The two cases of disagreement were minimal with one coder reporting 22 instead of 21 listings for cook and reporting eight instead of nine listings for dishwashers.

(11.) Irrespective of the minimum wage increase in Washington, September 11th had a pronounced effect on the number of want ads in both states. For example, even though Oregon did not experience a minimum wage increase in 2000 or 2001, the average number of want ads for waitstaff was 20 for October, November, and December of 2000 in comparison to four for the same period in 2001. A specification that includes a post September 11th dummy yields a negative and significant coefficient (not presented) but does not alter the qualitative conclusions of the analysis, likely reflecting that the negative employment shock occurred for both states. This dramatic post September 11th reduction in want ads is likely not found for employment due to the observed lag in the employment response.

LARRY D. SINGELL Jr and JAMES R. TERBORG *

* We thank Daniel Hamermesh, David Neumark, Philip Romero, and Joe Stone and the two anonymous referees for helpful comments. We also thank Tami Daley, Raina Lawton, Niran Harrison. and Michael Terborg for assistance with data collection. The authors are responsible for all remaining errors.

Singell: Department of Economics, University of Oregon, Eugene, OR 97403-1285. Phone (541)-346-4672, Fax (541)-346-1243, E-mail [email protected]

Terborg: Department of Management, Charles H. Lundquist College of Business, University of Oregon, Eugene, OR 97403-1285. Phone (541)-346-3354, Fax (541)-346-3341, E-mail [email protected]

TABLE 1
Median, 10th-, and 90th-Percentile Wages for Eating and
Drinking Jobs in Oregon and Washington (a)

 Hourly Wages in Oregon

 Minimum 10th 90th
 Wage Percentile Median Percentile

Occupational title--year 1997
 Hosts and hostesses 5.50 5.60 6.27 8.32
 Waiters and waitresses 5.56 5.84 8.04
 Cooks, fast food 5.57 5.95 8.13
 Cooks, restaurant 5.91 7.66 10.19
Occupational title--year 1998
 Hosts and hostesses 6.00 6.02 7.03 8.48
 Waiters and waitresses 6.01 6.66 8.15
 Cooks, fast food 6.01 6.84 8.34
 Cooks, restaurant 6.25 8.27 10.65
Occupational title--year 1999
 Hosts and hostesses 6.50 6.54 7.26 8.67
 Waiters and waitresses 6.50 6.84 8.53
 Cooks, fast food 6.50 6.86 8.56
 Cooks, restaurant 6.74 8.24 10.70
Occupational title--year 2000
 Hosts and hostesses 6.50 6.56 6.88 8.45
 Waiters and waitresses 6.55 6.82 8.75
 Cooks, fast food 6.62 7.41 8.40
 Cooks, restaurant 6.94 8.49 11.59
Occupational title--year 2001
 Hosts and hostesses 6.50 6.75 7.20 8.76
 Waiters and waitresses 6.76 7.31 9.95
 Cooks, fast food 6.78 7.63 9.16
 Cooks, restaurant 7.23 9.06 12.44

 Hourly Wages in Washington

 Minimum 10th 90th
 Wage Percentile Median Percentile

Occupational title--year 1997
 Hosts and hostesses 4.90 5.31 6.06 8.53
 Waiters and waitresses 5.24 5.58 7.79
 Cooks, fast food 5.28 5.82 8.10
 Cooks, restaurant 5.93 8.24 11.10
Occupational title--year 1998
 Hosts and hostesses 4.90 5.45 6.66 8.76
 Waiters and waitresses 5.39 6.16 7.81
 Cooks, fast food 5.44 6.39 8.40
 Cooks, restaurant 6.19 8.78 11.39
Occupational title--year 1999
 Hosts and hostesses 5.50 5.85 6.64 9.10
 Waiters and waitresses 5.78 6.09 8.01
 Cooks, fast food 5.80 6.33 8.35
 Cooks, restaurant 6.53 8.83 11.59
Occupational title--year 2000
 Hosts and hostesses 6.50 6.58 7.24 9.64
 Waiters and waitresses 6.55 6.74 11.58
 Cooks, fast food 6.57 6.91 8.31
 Cooks, restaurant 7.24 9.43 12.58
Occupational title--year 2001
 Hosts and hostesses 6.72 6.74 7.58 10.10
 Waiters and waitresses 6.74 7.26 14.91
 Cooks, fast food 6.82 7.79 10.56
 Cooks, restaurant 7.72 9.92 13.08

(a) These wage estimates are calculated by the BLS using data from a
sample of employers in all industry sectors in metropolitan and
nonmetropolitan areas in Oregon and Washington. These and other data
elements, including the 10th-, 25th-, 50th-, 75th-, and
90th-percentile wages, are available in the downloadable EXCEL files
from the Occupational Employment and Wage Estimates page on the BLS
Web site (http://www.bls.gov/oes/).

TABLE 2
Regressions for Employment Growth in the Eating and Drinking
Industry (a)

Variables (1) (2)

[DELTA]log(minimum
 wage) -0.0710 *** (0.0237) -0.0713 *** (0.0238)
[DELTA](lag minimum
 wage)
Oregon -0.0028 *** (0.0009) -0.0029 *** (0.0009)
Time -0.0285 *** (0.0094) -0.0285 *** (0.0094)
Time squared 0.0028 * (0.0015) 0.0028 * (0.0015)
Time cubed -0.0001 (0.0001) -0.0001 (0.0001)
Population growth 0.0171 ** (0.0077) 0.0175 ** (0.0077)
Personal income growth 0.0062 *** (0.0007) 0.0062 *** (0.0007)
[DELTA]log(E&D -0.2691 (0.3053)
 employment 1993-1992)
Constant 0.0596 *** (0.0012) 0.0671 *** (0.0010)
Observations 192 192
R squared 0.65 0.65

Variables (3) (4)

[DELTA]log(minimum
 wage) -0.0959 *** (0.0231) -0.0963 *** (0.0231)
[DELTA](lag minimum
 wage) -0.1078 *** (0.0239) -0.1082 *** (0.0239)
Oregon -0.0029 *** (0.0009) -0.0029 *** (0.0009)
Time -0.0286 *** (0.0089) -0.0286 *** (0.0089)
Time squared 0.0030 ** (0.0014) 0.0030 ** (0.0014)
Time cubed -0.0001 (0.0001) -0.0001 (0.0001)
Population growth 0.0120 (0.0073) 0.0123 * (0.0074)
Personal income growth 0.0039 *** (0.0009) 0.0039 *** (0.0009)
[DELTA]log(E&D -0.2875
 employment 1993-1992)
Constant 0.0764 *** (0.0008) 0.0845 *** (0.0008)
Observations 192 192
R squared 0.69 0.69

(a) Standard errors in parentheses. The data include monthly
observations for Oregon and Washington between 1994 and 2001. The
dependent variable measures the monthly difference in log employment.
The differenced log of the minimum wage is defined as the annual
difference in the log of the real minimum wage (in 1994 dollars),
reflecting the fact that the minimum wage changes annually. The
specifications also include monthly dummies and their interaction
with the Oregon dummy variable.

* Significant at 10%, ** significant at 5%; *** significant at 1%.

TABLE 3
Median, 10th-, and 90th-Percentile Wages for Hotel and Lodging Jobs in
Oregon and Washington (a)

 Hourly Wages in Oregon

 Minimum 10th 90th
 Wage Percentile Median Percentile

Occupational title--year 1997
 Maids and housekeeping 5.50 5.65 6.97 6.97
 Hotel desk clerks 5.82 7.19 7.19
 Lodging managers 7.31 14.02 14.02
Occupational title--year 1998
 Maids and housekeeping 6.00 6.05 7.31 7.31
 Hotel desk clerks 6.14 7.43 7.43
 Lodging managers 8.27 15.42 15.42
Occupational title--year 1999
 Maids and housekeeping 6.50 6.55 7.39 7.39
 Hotel desk clerks 6.66 7.68 7.68
 Lodging managers 9.27 15.93 15.93
Occupational title--year 2000
 Maids and housekeeping 6.50 6.68 7.71 7.71
 Hotel desk clerks 7.00 7.93 7.93
 Lodging managers 10.44 16.08 16.08
Occupational title--year 2001
 Maids and housekeeping 6.50 6.88 7.95 7.95
 Hotel desk clerks 7.21 8.26 8.26
 Lodging managers 11.01 16.63 16.63

 Hourly Wages in Washington

 Minimum 10th 90th
 Wage Percentile Median Percentile

Occupational title--year 1997
 Maids and housekeeping 4.90 5.61 7.48 10.13
 Hotel desk clerks 5.86 7.30 9.21
 Lodging managers 6.39 10.90 17.69
Occupational title--year 1998
 Maids and housekeeping 4.90 5.67 7.79 10.32
 Hotel desk clerks 5.91 7.67 9.70
 Lodging managers 8.50 16.18 19.33
Occupational title--year 1999
 Maids and housekeeping 5.50 6.01 7.52 10.39
 Hotel desk clerks 6.25 7.85 10.05
 Lodging managers 9.85 16.37 25.98
Occupational title--year 2000
 Maids and housekeeping 6.50 6.68 7.89 10.64
 Hotel desk clerks 6.75 8.13 11.01
 Lodging managers 11.31 21.79 40.66
Occupational title--year 2001
 Maids and housekeeping 6.72 6.92 8.34 10.91
 Hotel desk clerks 7.10 8.50 11.86
 Lodging managers 12.74 25.38 45.18

(a) These wage estimates are calculated by the BLS using data from
a sample of employers in all industry sectors in metropolitan and
nonmetropolitan areas in Oregon and Washington. These and other data
elements, including the 10th-, 25th-, 50th-, 75th-, and
90th-percentile wages, are available in the downloadable EXCEL files
from the Occupational Employment and Wage Estimates page on the BLS
Web site (http://www.bls.gov/oes/).

TABLE 4
Regressions for Employment Growth in the Hotel and Lodging Industry (a)

Variables (1) (2)

[DELTA]log(minimum
 wage) 0.1543 *** (0.0388) 0.1543 *** (0.0389)
[DELTA](lag minimum
 wage)
Oregon -0.0028 * (0.0015) -0.0029 * (0.0015)
Time -0.0283 * (0.0153) -0.0287 * (0.0154)
Time squared 0.0059 ** (0.0024) 0.0060 ** (0.0024)
Time cubed -0.0003 *** (0.0001) -0.0003 *** (0.0001)
Population growth 0.0416 *** (0.0126) 0.0422 *** (0.0126)
Personal income growth -0.0030 ** (0.0012) -0.0030 ** (0.0012)
[DELTA]log(H&L 0.1717 (0.3194)
 employment 1993-1992)
Constant -0.0146 (0.0357) -0.0128 (0.0359)
Observations 192 192
R squared 0.43 0.43

Variables (3) (4)

[DELTA]log(minimum
 wage) 0.1611 *** (0.0400) 0.1611 *** (0.0401)
[DELTA](lag minimum
 wage) 0.0296 (0.0414) 0.0295 (0.0415)
Oregon -0.0028 * (0.0015) -0.0029 * (0.0015)
Time -0.0283 * (0.0153) -0.0286 * (0.0154)
Time squared 0.0059 ** (0.0024) 0.0059 ** (0.0024)
Time cubed -0.0003 *** (0.0001) -0.0003 *** (0.0001)
Population growth 0.0430 *** (0.0127) 0.0436 *** (0.0128)
Personal income growth -0.0024 (0.0015) -0.0024 (0.0015)
[DELTA]log(H&L 0.1714 (0.3199)
 employment 1993-1992)
Constant -0.0192 (0.0363) -0.0174 (0.0366)
Observations 192 192
R squared 0.43 0.43

(a) Standard errors in parentheses. The data include monthly
observations for Oregon and Washington between 1994 and 2001. The
dependent variable measures the monthly difference in log employment.
The differenced log of the minimum wage is defined as the annual
difference in the log of the real minimum wage (in 1994 dollars),
reflecting the fact that the minimum wage changes annually. The
specifications also include monthly dummies and their interaction
with the Oregon dummy variable.

* Significant at 10%; ** significant at 5%; *** significant at 1%.

TABLE 5
Average Number of Want Ads Each Month in the Portland
Oregonian and the Seattle Times by Job Type (a)

 No Change in Minimum

 1994 1995 1996

Variables--Oregon (observations = 96)
 Waitpersons 22.50 (7.39) 28.58 (8.19) 8.00 (12.07)
 Bus persons 8.75 (4.39) 11.00 (2.98) 8.66 (3.82)
 Dishwasher 8.33 (2.74) 12.16 (5.13) 9.41 (3.20)
 Cooks 38.66 (13.52) 47.50 (16.84) 44.58 (12.85)
 Host persons 8.50 (3.37) 10.66 (3.22) 8.83 (3.90)
 Hotel housekeepers 23.75 (5.67) 28.25 (8.23) 32.00 (5.54)
Variables--Washington (observations = 96)
 Waitpersons 15.41 (4.90) 17.25 (5.72) 24.25 (5.32)
 Bus persons 5.91 (3.14) 4.90 (2.39) 5.00 (2.92)
 Dishwasher 6.00 (3.07) 5.16 (1.89) 7.66 (4.18)
 Cooks 30.58 (7.26) 32.08 (11.29) 34.41 (9.81)
 Host persons 7.25 (2.66) 4.83 (2.88) 7.91 (3.44)
 Hotel housekeepers 24.16 (10.05) 30.66 (4.57) 40.08 (10.41)

 Both States
 Oregon Raises Minimum Raise Minimum

 1997 1998 1999

Variables--Oregon (observations = 96)
 Waitpersons 28.75 (8.67) 22.75 (7.92) 21.16 (5.70)
 Bus persons 8.33 (3.62) 6.00 (3.27) 3.42 (1.78)
 Dishwasher 9.75 (3.64) 6.91 (2.46) 4.08 (2.35)
 Cooks 45.41 (17.98) 39.58 (9.08) 36.83 (12.44)
 Host persons 11.25 (5.10) 8.16 (4.34) 6.33 (3.68)
 Hotel housekeepers 37.41 (8.91) 34.16 (6.19) 33.83 (6.60)
Variables--Washington (observations = 96)
 Waitpersons 29.33 (8.12) 31.57 (7.54) 35.33 (10.65)
 Bus persons 9.50 (4.62) 10.83 (4.21) 11.33 (4.00)
 Dishwasher 8.33 (4.90) 12.41 (2.93) 12.00 (5.18)
 Cooks 36.66 (11.31) 40.50 (12.95) 50.41 (10.43)
 Host persons 9.33 (4.51) 10.33 (4.05) 14.50 (4.18)
 Hotel housekeepers 39.33 (8.91) 42.27 (6.27) 48.25 (6.60)

 Washington Raises Minimum

 2000 2001

Variables--Oregon (observations = 96)
 Waitpersons 19.58 (6.77) 9.25 (4.71)
 Bus persons 4.50 (2.46) 1.91 (1.56)
 Dishwasher 6.00 (3.04) 2.33 (1.77)
 Cooks 37.83 (7.44) 19.08 (7.17)
 Host persons 6.91 (2.81) 2.17 (1.47)
 Hotel housekeepers 30.25 (8.76) 16.83 (5.65)
Variables--Washington (observations = 96)
 Waitpersons 27.92 (10.02) 12.58 (5.28)
 Bus persons 8.25 (5.66) 2.75 (2.00)
 Dishwasher 9.92 (4.81) 3.33 (13.02)
 Cooks 44.42 (13.37) 23.00 (9.01)
 Host persons 9.58 (4.78) 4.08 (2.94)
 Hotel housekeepers 41.67 (6.60) 20.17 (8.06)

(a) Standard errors are in parentheses. Want-ad data are collected
from the Portland Oregonian and the Seattle Times. The minimum wage
in Oregon was $4.75 between 1994 and 1996, increased to $5.511, $6.00,
and $6.50 in 1997, 1998, and 1999, respectively, and remained constant
at $6.50 in 2000 and 2001. The minimum wage in Washington was constant
between 1994 and 1998, increased to $5.50, $6.50, and $6.72 in 1999,
2000, and 2001, respectively.

TABLE 6
Job-Specific Want-Ad Regressions (a)

Variables Waitstaff Buss Staff

Log(minimum wage) -2.4369 *** (0.6382) -3.4347 ** (1.4498)
Oregon 0.5869 *** (0.1348) 0.6403 ** (0.3061)
Time -0.3177 (0.2333) -0.6233 (0.5300)
Time squared 0.0734 ** (0.0371) 0.0979 (0.0844)
Time cubed -0.0043 ** (0.0017) -0.0045 (0.0040)
Population growth -0.2106 (0.2233) 0.1909 (0.5072)
Personal income growth 0.0755 *** (0.0200) 0.1303 *** (0.0453)
Constant 2.0904 (1.2964) 2.5135 (2.9449)
Passes Heckman-Hotz test yes yes
Observations 192 192
R squared 0.78 0.48

Variables Dish Staff Host Staff

Log(minimum wage) -4.7616 *** (1.5731) 0.7113 (2.8208)
Oregon 0.4001 (0.3322) 0.2731 (0.5956)
Time -0.2397 (0.5751) -0.2953 (1.0312)
Time squared 0.0431 (0.0915) 0.0540 (0.1641)
Time cubed -0.0020 (0.0043) -0.0029 (0.0077)
Population growth -0.6479 (0.5503) 0.3885 (0.9868)
Personal income growth 0.0830 * (0.0492) 0.0247 (0.0882)
Constant 5.9515 * (3.1953) -8.7736 (5.7297)
Passes Heckman-Hotz test yes yes
Observations 192 192
R squared 0.34 0.27

Variables Cooks Housekeepers

Log(minimum wage) -1.7636 *** (0.5944) -1.0526 * (0.5917)
Oregon 0.7722 *** (0.1255) 0.2650 ** (0.1249)
Time -0.5935 *** (0.2173) -0.1772 (0.2163)
Time squared 0.1073 *** (0.0346) 0.0633 * (0.0344)
Time cubed -0.0058 *** (0.0016) -0.0044 *** (0.0016)
Population growth -0.3014 (0.2079) -0.2810 (0.2070)
Personal income growth 0.0202 (0.0186) 0.0201 (0.0185)
Constant 2.5018 ** (1.2074) 2.1800 * (1.2019)
Passes Heckman-Hotz test yes yes
Observations 192 192
R squared 0.78 0.67

(a) Standard errors in parentheses. The data include monthly
observations for Oregon and Washington between 1994 and 2001. The
dependent variable measures want ads in each job category in logs.
The regressions also include monthly dummies and monthly dummies
interacted with the Oregon dummy variables.

* Significant at 10%; ** significant at 5%; *** significant at 1%.

TABLE 7
Job-Specific Want-Ad Regressions Including Lag of the Minimum Wage (a)

Variables Waitstaff Buss Staff

Log(minimum wage) -2.1742 *** (0.7848) -3.5956 ** (1.7844)
Log(lag of the minimum
 wage) -0.3828 (0.6627) 0.2344 (1.5068)
Oregon 0.5929 *** (0.1354) 0.6366 ** (0.3079)
Time -0.2606 (0.2539) -0.6583 (0.5772)
Time squared 0.0652 (0.0398) 0.1029 (0.0906)
Time cubed -0.0039 ** (0.0019) -0.0047 (0.0043)
Population growth -0.1730 (0.2330) 0.1679 (0.5297)
Personal income growth 0.0766 *** (0.0201) 0.1296 *** (0.0457)
Constant 2.0748 (1.2994) 2.5230 (2.9544)
Passes Heckman-Hotz test yes yes
Observations 192 192
R squared 0.78 0.48

Variables Dish Staff Host Staff

Log(minimum wage) -4.4507 ** (1.9358) -0.0008 (3.4707)
Log(lag of the minimum
 wage) -0.4531 (1.6346) 1.0376 (2.9307)
Oregon 0.4070 (0.3341) 0.2571 (0.5990)
Time -0.1721 (0.6262) -0.4501 (1.1227)
Time squared 0.0333 (0.0983) 0.0762 (0.1762)
Time cubed -0.0015 (0.0047) -0.0040 (0.0083)
Population growth -0.6035 (0.5747) 0.2867 (1.0303)
Personal income growth 0.0844 * (0.0496) 0.0216 (0.0889)
Constant 5.9331 * (3.2052) -8.7314 (5.7465)
Passes Heckman-Hotz test yes yes
Observations 192 192
R squared 0.34 0.27

Variables Cooks Housekeepers

Log(minimum wage) -1.1306 (0.7266) -1.2590 * (0.7278)
Log(lag of the minimum
 wage) -0.9223 (0.6135) 0.3007 (0.6145)
Oregon 0.7865 *** (0.1254) 0.2603 ** (0.1256)
Time -0.4559 * (0.2350) -0.2220 (0.2354)
Time squared 0.0875 ** (0.0369) 0.0697 * (0.0369)
Time cubed -0.0048 *** (0.0017) -0.0047 *** (0.0017)
Population growth -0.2110 (0.2157) -0.3105 (0.0017)
Personal income growth 0.0230 (0.0186) 0.0191 (0.0186)
Constant 2.4643 ** (1.2030) 2.1923 * (1.2050)
Passes Heckman-Hotz test yes yes
Observations 192 192
R squared 0.78 0.67

(a) Standard errors in parentheses. The data include monthly
observations for Oregon and Washington between 1994 and 2001. The
dependent variable measures want ads in each job in logs. The
regressions also include monthly dummies and monthly dummies
interacted with the Oregon dummy variables.

* Significant at 10%; ** significant at 5%; *** significant at 1%.