Printer Friendly

The dissipation of minimum wage gains for workers through labor-labor substitution: evidence from the Los Angeles living wage ordinance.

1. Introduction

Economic theory suggests that firms will replace low-skill workers with high-skill workers if the two are substitutes in production and the wages of low-skill workers increase as the result of a minimum wage. Labor-labor substitution may also take place along dimensions other than skill if, for example, employers have a "taste for discrimination" and the minimum wage reduces the wage premium they must pay for favored personal characteristics of workers. If affected firms replace low-skill workers with high-skill workers following enactment of a wage minimum, the initial benefits of the regulation for workers at these firms will be dissipated over time as the wage gains for new hires are less than for members of the original workforce. A similar dissipation of wage gains will occur if employers replace workers that are discriminated against in the labor market with those that are not.

Most theoretical depictions of labor-labor substitution focus on employer-initiated causes: It may result from changes in firms' demand for worker skills, from employer efforts to dissipate worker minimum wage rents by requiring greater work effort or less absenteeism from workers (and by seeking workers that better match these new employer goals), or from employers invoking their "taste for discrimination." However, successful labor substitution rests on workers' supply decisions as well. Indeed, such substitution could result solely from labor supply effects if, for example, firms were to hire nonselectively from an altered more highly skilled or otherwise favored applicant pool, which is brought on by the higher wage.

The empirical literature on the employment effects of wage minimums has found little employment loss among affected jobs or more highly impacted worker populations, such as teenagers (Brown 1999). This suggests that there may be little labor-labor substitution across broad skill/occupational categories--for example, adults for teenagers, workers with some college experience for high-school graduates, or backhoe operators for ditch diggers. (1) However, labor-labor substitution may still take place along more refined dimensions of workers' skills (Abowd and Kilingsworth 1981). For example, Neumark and Wascher (1996) find evidence that minimum wages lead firms to substitute away from younger teenagers and toward older, more experienced teenagers.

Labor substitution following a minimum wage may dissipate the initial wage gains for workers. Empirical evidence suggests, however, that, just as in the case of unions (Card 1996), the rent dissipation is by no means complete. It is now well established empirically that a "spike" exists in the wage distribution around the relevant minimum wage (Card and Krueger 1995; DiNardo, Fortin, and Lemieux 1996), suggesting that the wages of low-skill workers are indeed affected by wage mandates. (2) Moreover, job queues exist for minimum wage jobs, suggesting that economic rents are earned by low-wage workers in these positions (Holzer, Katz, and Krueger 1991). Although the worker rents created by minimum wage regulations may not be completely dissipated by labor substitution, the extent of such dissipation remains an open question empirically.

Although the theory of labor-labor substitution is well-developed terrain, we are aware of no empirical tests of the proposition at the establishment level. (3) The empirical literature on this topic has mainly focused on estimates of the elasticity of substitution, typically relying on time series data on broad industry aggregates (e.g., manufacturing) to test for substitution either across broad skill categories (e.g., nonproduction for production workers) or across worker characteristics (e.g., older age for younger age, white for black, or male for female) using labor demand equations or production or cost function specifications. All approaches pose significant econometric challenges to the researcher, none are directly concerned with substitution resulting from a minimum wage increase, and none are able to discern whether the observed substitution is the result of within-firm as opposed to across-firm or across-industry adjustment.

An exception to this basic approach is the literature on wage spillovers, which looks at the impact of minimum wage increases on the wages of workers earning more than the minimum. However, as noted by Grossman (1983), the wages of workers higher up in the wage distribution may increase as the result of internal wage norms rather than demand shifts because of substitution. Thus, the evidence on wage spillovers is difficult to directly link to labor substitution. Neumark, Schweitzer, and Wascher (2000) extend the wage spillovers approach by exploring the impact of an increase in the minimum wage on the hours of work and employment of workers earning more than the minimum. They find some evidence for a positive and statistically significant contemporaneous impact on the hours of work of these workers, but the effects do not persist over time, and there appears to be little statistically significant employment effects. None of these results is directly attributable to labor substitution at the establishment level.

In this paper, we utilize a unique employer-employee matched dataset to explore the extent of labor-labor substitution and dissipation of wage gains for workers resulting from a significant wage mandate--a living wage ordinance within city contract establishments in Los Angeles. We capture labor substitution by comparing the skill and nonskill attributes of a sample of stayer workers with those of a sample of joiner workers following a period of adjustment to the initial minimum wage increase. The demographic and human capital characteristics of these workers allow us to test for substitution along observable demographic and skill dimensions. In addition, we also know the wages of these workers before the wage mandate, which allows us to measure the extent of substitution among unobservable skill and demographic characteristics. Exploring unobservable differences across the stayer and joiner populations is important because the skill substitution that interests us is perhaps less likely to occur across observable dimensions, such as schooling.

The worker-firm match aspect of the data allows us to estimate within-firm measures of labor substitution and thereby gives some insight into whether firms behave as the microeconomics theory of substitution suggests. The establishment survey data offer information on the city contract firms that were affected by the ordinance. In addition to numerous establishment characteristics, the survey asked employers whether they changed hiring standards in response to the living wage, thereby allowing us some insight into the relative importance of demand-side (i.e., employer-initiated) versus supply-side (i.e., changing applicant pool) effects in the overall substitution response.

One of the major challenges posed by these data in adequately testing the labor-labor substitution hypothesis is that we know nothing about the leavers--that is, those workers who left city contract work between the time their firm became subject to the Living Wage Ordinance and the time of the survey. A proper test of the hypothesis would compare joiners with leavers rather than joiners with stayers. However, to the extent firms actively seek to alter the skill or demographic features of the workforce, it seems reasonable to assume that stayers would be drawn disproportionately from among the more highly valued, and thus highly paid workers in the original workforce, which means that a comparison with stayers leads to an underestimate of true labor-labor substitution. The information on the hiring standards change allows us to produce some suggestive evidence on whether or not this is the case.

Finally, estimates of the conditional average wage increase for stayers and joiners resulting from the Living Wage Ordinance are used to derive a measure of the dissipation of minimum wage gains for workers at affected firms. Specifically, we begin with the average wage gain for stayers and assume that it represents (a lower-bound estimate of) the average wage gain for the initial workforce at the time the ordinance took effect. This measure is then compared to a measure of the overall wage gain to workers following a period of adjustment and labor substitution, using a weighted average of the wage gains of stayers and joiners. The difference between the two measures provides an estimate of the dissipation of wage benefits for the workforce at affected firms due to labor-labor substitution. Note that this is not a measure of dissipation in the labor market as a whole--which, at a minimum, would require knowing something about the wages of leavers in their new positions--but rather an establishment-level measure of dissipation, grounded in the labor substitution behavior of affected firms and focused on the workers employed in those firms.

The outline of the paper is as follows: We begin with a brief discussion of the Los Angeles Living Wage Ordinance. This is followed by a discussion of the data and empirical methodology. The results of the analysis of labor-labor substitution on observable and unobservable characteristics are then discussed. We conclude with an estimate of the wage gain for workers resulting from the Living Wage Ordinance, accounting for the impact that labor-labor substitution has had on the dissipation of worker rents. The results suggest substitution toward male, Latino, and black workers, and workers possessing prior formal training. All are characteristics that generate a wage premium in this particular segment of the low-wage labor market in Los Angeles. The results also suggest rather significant substitution on unobservable worker characteristics, as evidenced by the finding that the "before" wages of joiners are significantly higher than those of stayers. Finally, we estimate that the wage gain to the initial group of city contract workers as a result of the Living Wage Ordinance is dissipated by almost one-third following a period of adjustment and labor substitution.

2. The Los Angeles Living Wage Ordinance

The Los Angeles Living Wage Ordinance (LWO) was passed in 1997 by the Los Angeles City Council and went into effect in May of that year. It was the 10th such ordinance to be passed in the country in what has become a movement that currently encompasses over 120 localities. Living wage ordinances typically strive to increase the hourly wages of workers in city contract firms so that, on a full-time basis, they are equal to or greater than the amount required to bring a family of four above the federal poverty line. In addition, some ordinances give encouragement to employers to provide health benefits to workers, and some stipulate a minimum number of paid days off for workers per year. The Los Angeles ordinance is the third largest in the country (behind New York and San Francisco). Employers are covered in Los Angeles if they lease land from the city, including, most importantly, businesses at the Los Angeles and Ontario International Airports.

It is estimated that as many as 8000 workers in 150 firms were directly affected by the wage provision of the ordinance (Fairris et al. 2005). Although the Los Angeles ordinance is among the largest in the country, it directly affects only 2% of the low-wage labor force in the area. We might expect greater labor substitution in this case, compared to one in which the wage mandate covers a significant portion of the low-wage workforce because any substitution that does take place is unlikely to put significant upward pressure on the wages of more highly valued workers. Landscape laborers, janitors, security guards, food service workers, and parking attendants are among the more prominent occupational groups affected by the ordinance.

The ordinance covers (i) companies with a city service contract of $25,000 or more and their subcontractors; (ii) companies that receive economic development subsidies (i.e., "business assistance") of $1 million or more in one year, or $100,000 or more annually on an ongoing basis, and their subcontractors; and (iii) companies that have a lease with the city, or are granted a license or permit, and their subcontractors. (4) Firms that had a contract or lease at the time the ordinance was passed did not become covered until that contract or lease was renewed. Finally, unionized firms are exempt from the provisions of the ordinance.

Beginning in 1997, covered firms were required to pay their employees who worked on city contracts or on city property a minimum of either $8.50 per hour (roughly 70% above the then state minimum wage) or $7.25 per hour plus a $1.25 hourly contribution to employee health benefits. The two-tier wage structure was intended to give employers an incentive to provide health insurance to their workers. The incentive comes largely in the form of tax savings because taxes are paid on the wage component of labor costs but not on the value of the fringe benefits package. The living wage levels rise every year and are indexed to the annual increase in the city employee pension fund. Firms were also required to provide their covered employees with 12 paid days off and 10 unpaid days off per year.

3. Data

This paper utilizes original employer-employee matched data. The Survey of Los Angeles Living Wage Employers (SLWE) was the first stage in an intentional two-part survey design that included a survey of affected workers at these very same establishments as a second stage. The survey response rate for the employer survey was 84%. Structured, in-person interviews of roughly 1-2 hours were initially conducted with 82 living wage employers at their offices, and then approval and assistance was solicited in contacting, sampling, and surveying their workers. The SLWE began in the fall of 2001 and was largely completed by the fall of 2002.

The sampling frame for the SLWE was developed from a database maintained by the City of Los Angeles, and in particular from its list of "priority one" firms--those deemed by the city to employ significant numbers of low-wage workers. A two-stage stratified cluster approach was used to generate the sample of survey establishments. The "priority one" firms were first stratified into the following industrial groups:


Airline services



Retail and food service

Security and parking

Social services



Each industrial group was then further stratified by firm size: large firms, with 50 or more workers, and small firms with less than 50 workers. Large firms were oversampled for cost and clustering reasons. Weights were developed to render the results of statistical analyses representative of the general population of living wage firms.

In the second stage of the project, 320 workers from 65 living wage establishments responded to surveys from the winter of 2002 to the summer of 2003. The survey response rate was 81%. Worker surveys were typically administered away from the place of employment and lasted from 45 minutes to 1.5 hours. The target population of the Survey of Living Wage Workers (SLWW) was workers directly affected by the living wage ordinance--specifically, workers in jobs wherein the wages for their jobs were mandated to increase as a result of the ordinance. The sampling design in this case was intended to have two stages of stratification, with the establishments functioning as a cluster of workers. For this purpose, information on the population of affected workers was requested from the sampled establishments via lists that included name, occupation, and hiring date. The first level of intended stratification was based on occupational groups. However, because not every establishment supplied complete information as requested, the stratification at this stage was possible in only 11 establishments.

The second level of intended stratification was by date of hire, to capture both workers who were at the firm before the Living Wage Ordinance took effect (stayers) and workers who were hired afterwards (joiners). Because information on hiring date was not provided by all establishments, the stratification at this level was applied to only eight of the establishments that were first stratified by occupations. In five other establishments, the only stratification applied was the one on date of hire. No stratification was applied in the 49 establishments in which neither the occupation nor the hiring date of the workers was supplied. In these cases workers in the sample were selected randomly. Sampling weights were developed for the worker survey to render statistical results representative of the population of workers affected by the living wage ordinance. (5) The establishment and worker survey data were merged to form the matched employer-employee dataset.

Table 1 lists the variables from the employer and worker surveys to be used in the analysis. We possess a host of worker skills and personal characteristics, such as experience at hire, years of schooling, female, and Latino, with which to test for observable differences across the joiner and stayer populations. In addition, for most of the workers in our sample we possess wage histories, some of which date as far back as the previous five jobs. From this information, we can retrieve the hourly wage paid in the job held immediately before the living wage--before wage--the date on which the worker was first paid this wage, and a host of variables associated with the job, including whether it was unionized--union (before)--and whether it offered employer-provided health benefits--health benefits (before). Thus, it is possible to identify for each worker in our sample the wage earned immediately before coverage by the LWO. For joiners, this "before" information applies to previous employers; whereas, for stayers, it represents wage information at their current living wage employer before the onset of the living wage. We use this information to test for differences between stayer and joiner populations in unobserved wage-determining characteristics, controlling for those observed characteristics contained in the data.

In combination with the after wage information, it is possible to calculate the wage gain experienced by each worker. In the case of stayers, the after wage observation occurs when a worker's firm became subject to the provisions of the LWO; whereas, for joiners it occurs when the worker joins the city contract sector. Differences in wage gains across the stayer and joiner populations form the basis for our estimate of the dissipation of intended wage benefits for workers due to labor substitution.

Table 2 gives the basic descriptive statistics of the establishments from the SLWE dataset. The total number of useable establishments in the dataset is 44, following deletion of firms with missing responses to pertinent information and firms that were new to the city contract sector following the living wage. New firms entered city contract work following the ordinance, and they are part of our data but are not included in the sample analyzed in this paper. The labor-labor substitution hypothesis is most appropriately applied to firms that held a contract with the city before passage of the living wage ordinance because it is in these firms where true "labor substitution" may take place.

The two largest industry categories are miscellaneous, which includes a hodge-podge of establishments, from production firms having benefited from "business assistance" to operators of amusement park equipment, and social services, which includes establishments operating youth programs, half-way houses, and child-care facilities through the city. Thirty-two percent of firms changed hiring standards as a result of the living wage ordinance. There are, on average, five sampled workers per establishment.

Table 3 gives the descriptive statistics for variables from the SLWW dataset. The total number of useable observations in the data is 231, after deletion of workers with missing responses to pertinent information and workers in establishments new to city contract work following the living wage ordinance. The descriptive statistics are also presented by worker type: stayer and joiner. The unconditional means do not appear to be greatly different across most of the worker skill and demographic characteristics of stayers and joiners. However, several differences do stand out. There are slightly fewer females among the joiner population but also slightly more Latino and black workers, especially relative to white workers. The first is consistent with labor substitution based on discrimination; whereas, the latter two are generally not thought to be so.

There is a sizeable quantitative difference in the unconditional before wages of the two groups, with joiners possessing statistically significantly higher before wages, consistent with the labor substitution hypothesis. However, joiners are also significantly more likely to have been in part-time jobs as compared to stayers, suggesting that controlling for part-time status will be important in the analysis of before wage comparisons. Looking at the union and health benefits attached to before wages and comparing to current status, while stayers experienced a slight increase in both, joiners witnessed a very dramatic rise in the incidence of both unionization and possession of health benefits. Thus, it will be important to control for these differences for both groups in the analysis of wage gains owing specifically to the Living Wage Ordinance. (6)

Finally, in the analysis to follow we control for occupation using very detailed occupational categories. Among the more prominent occupations are childcare provider, landscape laborer, janitor, security guard, parking attendant, cashier, and counselor for youth activities. The detailed occupations come from 11 major occupational groups: child-care workers, cleaning service workers, landscape workers, parking service workers, restaurant workers, retail workers, social service workers, security workers, airline service workers, supervisors, and miscellaneous occupations. These categories follow closely the industry where the job is performed.

4. Empirical Methodology

Labor Substitution and Observable Characteristics

The first phase consists in the estimation of differences in the conditional mean of observable characteristics between joiners and stayers. The focus in this analysis is on the following skill-related characteristics: experience at hire, hire age, years of schooling, currently enrolled, native English speaker, and prior formal training. We focus here on a comparison of skills at the time of hire. In addition, we explore substitution toward a more male- and white-dominated workforce by analyzing differences between the joiner and stayer populations with regard to the variables female, Latino, and black, where the latter two are measured relative to white workers only.

The empirical approach is to perform a series of ordinary least squares regressions, the general specification of which is

[y.sub.i] = [[beta].sub.0] + [[gamma].sub.J][joiner.sub.i] + [[beta]'.sub.2][E.sub.j] + [[epsilon].sub.ij] (1)

where [y.sub.i] is the observable characteristic of worker i, [joiner.sub.i] is a dummy variable indicating whether worker i is a joiner, [X.sub.i] is a vector of individual attributes, and [E.sub.j] is vector of establishment dummy variables. The error term is indexed by both i and j, suggesting possible interdependence between individuals working for the same establishment. All regressions are estimated utilizing a correction for this possible clustering effect (as well as White-corrected standard errors to address possible heteroskedasticity). The indicator of labor-labor substitution is given by the coefficient [[gamma].sub.J], which reflects the mean conditional within-establishment difference in observable characteristics between joiners and stayers.

A set of detailed categorical variables indicating the occupation of workers is included in all regressions to ensure that the comparison of observable characteristics is restricted to variation within occupations. A host of worker demographic characteristics also appear as independent variables, including part-time job and the other observable, skill, gender, race, and ethnicity categories.

Labor Substitution and Unobservable Characteristics

"Before" wage comparisons are used to uncover differences in characteristics between joiners and stayers that are unobservable to the researcher but perhaps observable to participants in the hiring decision. Human capital theory suggests that, all else constant, a higher wage is indicative of greater skills. However, a higher wage might also be an indication of non-skill-related characteristics that are favored by employers. There is also ample empirical evidence of the existence of unexplained wage differences across similar workers, presumably owing to differences in establishment human resource policies (Groshen 1991; Gibbons and Katz 1992; Abowd, Kramarz, and Margolis 1999). In the end, we are unable to clearly distinguish wage differences that reflect unobservable worker characteristics from wage differences due to unobserved establishment pay policies. This is an important matter for future research. Note, however, that this inability in no way compromises our estimate of the dissipation of minimum wage gains for workers through worker replacement.

To detect differences in unobservable characteristics between stayers and joiners, we utilize a wage regression of the following form:

[] = [[beta].sub.0] + [[gamma].sub.J][joiner.sub.i] + [[beta]'.sub.1][] + [[beta]'.sub.2][E.sub.j] + [[beta]'.sub.3][Z.sub.k[i]] + [[delta].sub.t][] + [[epsilon].sub.ik] (2)

where [] is the log "before" wage of individual i at time t, [] is a vector of worker characteristics, [E.sub.j] is a vector of living wage establishment dummy variables, and [Z.sub.k[i]] is a vector of characteristics of the establishment where the before wage was earned. The disturbance term has similar properties as the disturbance term in Equation 1. The index t is used here with more liberty than is customary in the literature. It indicates that the observations are dated--that is, not all wages are observed in the same time period--and not that there is a time series component to the analysis. The difference in the within-establishment conditional mean wages of joiners and stayers is captured by the coefficient [[gamma].sub.J].

To eliminate the effect of observable skills and other individual characteristics on wages, we control for experience at hire, hire age, Jemale, years of schooling, native English speaker, prior formal training, and race and ethnicity variables. Part-time job is added to capture the significant difference observed in work hours across the stayer and joiner populations before the LWO. To remove important differences in before establishment characteristics, controls are added for health benefits and union.

Differences in the timing of wage observations pose a challenge. Not only are immediately before wages observed at different points in time, these time periods are not uniform across stayers and joiners. The before wages of stayers are observed before their city contract employer became subject to the LWO. Firms become subject to the ordinance once their contract with the city is renewed. The LWO took effect in 1997, but a majority of firms were not affected by the ordinance until after 1999. The before wage observations of joiners typically come from the period immediately before the worker joined the living wage sector--that is, after (and sometimes long after) their city contract employer was affected by the ordinance. Thus, the stayers tend to be overrepresented in the early years and joiners in the later years.

Controlling for time in this analysis is especially important because most individuals in the sample are low-wage workers, and the California minimum wage changed five times over the period, increasing from $4.75 at the end of 1996 to $6.75 in January 2002. Any observed difference in wages between stayers and joiners might be an artifact of the different minimum wage regimes affecting these workers. Joiners might possess larger before wages simply because their before wage observations come disproportionately from later in the period. We address this challenge by including a set of categorical variables--[]--indicating minimum wage periods. A dummy variable in this set equals one if the wage observation corresponds to minimum wage period l and equals zero otherwise. This is equivalent to redating the wages using a calendar based on minimum wage regimes instead of years. (7)

Wage Gains Estimation and the Measure of Dissipation

To estimate the wage gains for workers from the LWO, we use an approach that is similar to the methodology used in the unobservables analysis above but with a few changes in specification and data. (8) We wish to isolate separately the wage gains of stayers and joiners due solely to the Living Wage Ordinance and not to other factors such as changing union status or the different time periods during which stayers and joiners experienced their wage increases. Estimating a regression in wage levels, instead of wage gains, allows us to link each wage observation (both before wage and after wage) to the particular temporal features (e.g., union status or minimum wage period) that pertained at the time.

Ignoring for the moment controls for minimum wage periods, our strategy thus consists of estimating


where [] is the before wage or after wage observation, [t.sub.w] is the time in month and year when the wage is reported, [t.sup.*] is the time in month and year when the worker became affected by the LWO, 1{.} denotes an indicator function that equals one if the parenthetical temporal condition holds, [E.sub.j] is a vector of living wage establishment dummy variables, and [] and [Z.sub.k[i]t] are vectors of worker and establishment characteristics, respectively, that are temporally attached to the before and after wage observations. The wage gain for both groups of workers is easily obtained from the coefficients of the indicator functions. The estimate of the wage gain for stayers attributable to the LWO is

E[[]|t [greater than or equal to] [t.sup.*], joiner = 0, [], [Z.sub.kt]] - E[[]|t < [t.sup.*], joiner = 0, [], [Z.sub.kt]] = [[[phi].sub.S].

Similarly, the wage gain for joiners can be estimated as

E[[]|t [greater than or equal to] [t.sup.*], joiner = 1, [], [Z.sub.kt]] - E[[]|t < [t.sup.*], joiner = 1, [], [Z.sub.kt]] = [[[gamma].sub.JA] + [[phi].sub.S] - [[gamma].sub.JB].

We allow observed and unobserved characteristics to vary across the stayer and joiner populations, thereby leading to different wage gains for the two groups, but hold other factors constant in the analysis. We know, for example, that the city contract sector stands out among similar employers in the region by possessing a higher union density and greater provision of employer-paid health benefits (Fairris 2005). Both could be important factors in the estimated wage gains of joiners, but perhaps of stayers as well, if their union status or health benefits changed over the period. Thus, to attribute the wage gain to the LWO specifically and not to unionization or changes in health benefits, we control for union and health benefits associated with each wage observation.

Some of the difference in estimated wage gains for workers might be related to differences in the length of elapsed time between the before and after period of recorded wages, or differences in before and after part-time status. We also want to make wage gain comparisons across worker groups within current occupations. For these reasons, we control for time elapsed, part-time job, and occupation. Establishment fixed effects are employed to arrive at within-firm wage gain differences.

Another source of the estimated difference in wage gains might be related to the fact that joiners tend to come under the provisions of the LWO later than do stayers. One way to free the wage gain estimates of this temporal effect is to add the minimum wage categorical variables to the regression, just as we did for the unobservables analysis. However, that approach would be inappropriate in this case. It would yield within-period estimates of the overall wage gain; whereas, we wish to capture within-period estimates of before and after wages separately. To achieve this, we add to Equation 3 a set of interactives of the minimum wage variables with the indicator functions--1 {.}--that identify before and after wages.

To estimate the dissipation of wage gains, we calculate a population-weighted average of the wage gains for stayers and joiners resulting from the LWO and compare this to a counterfactual wage gain assuming no labor substitution took place. The counterfactual wage gain is derived by attributing to the affected-worker group the average wage gain received by stayers. This is likely to be a lower bound on the actual wage gain of the original, preadjustment workforce if stayers are a select draw from that workforce. We ignore the dissipation of wage gains resulting from employment loss in this calculation. Survey estimates put this loss of employment at less than 2%.

5. Results

Observable Substitution

We begin with an analysis of differences in observable skill and demographic features across the stayer and joiner populations. The results are reported in Table 4. (9) The characteristics that stand out in the column 1 results are formal training, gender, ethnicity, and race. Joiners are more likely to have had prior formal training and to be male, Latino, or black. Thus, with the exception of prior formal training, we find little significant evidence of labor substitution on the observable skills of workers in these results.

The likelihood of having received prior formal training is 11 percentage points higher for joiners, which is a considerable magnitude given that only 12% of stayers report having received prior formal training. Although 57% of the stayers are female, only 43% of joiners with similar characteristics are female. Relative to white workers, Latinos and blacks make up 82% and 75%, respectively, of the stayer population. Those percentages increase to 97% and 88% in the joiner population, ceteris paribus.

Theory suggests that a minimum wage may reduce job training if, due to the wage floor, low-wage workers are unable to accept a temporarily lower training wage to pay for the acquisition of skills on the job. Thus, the prior formal training result could merely indicate that firms are now seeking workers that already possess the skills that can no longer be imparted through on-the-job training programs. Joiners pay for and receive elsewhere the general human capital skills that stayers used to pay for and receive in-house. If true, it is not that more skilled workers are being utilized as a result of the minimum wage, in accordance with the labor substitution hypothesis, but rather that the training of workers no longer takes place within firms.

Information is available for shedding some light on this alternative explanation. The worker survey contains a question regarding the existence of job training after being hired (received training). Regression results reveal no statistical difference in the likelihood of initial training for stayers and joiners, controlling for an array of worker characteristics and firm fixed effects. Moreover, adding this training variable to the column 2 specification of the prior formal training regression does not substantively alter the finding that joiners are significantly more likely than stayers to have received formal training before taking their jobs. Although we know nothing about differences in the amount of job training workers receive after being hired, these results offer at least suggestive evidence that joiners are indeed more highly trained than stayers, consistent with the labor substitution hypothesis.

Although the hiring of more male workers following the LWO is consistent with the discrimination variant of the labor substitution hypothesis, the increased preponderance of Latinos and blacks in the joiner population is puzzling if the latter are discriminated against in the labor market and employers prefer white workers. One possibility is that these differences merely represent changing demographic trends in the southern California labor force over this period an especially viable explanation for the Latino result. (10) However, further analysis of wage differences for these groups yields a different, more surprising explanation.

Curiously, a simple before wage regression reveals that while females earn significantly lower wages than men (by about 5%), ceteris paribus, Latinos and blacks earn higher wages than whites (by 11% and 8%, respectively). (11) These results suggest that women and whites are either discriminated against or possess lower unobserved abilities than men, Latinos, and blacks in this portion of the low-wage labor market in southern California. Regardless of which of the two explanations drives the results, this analysis suggests that the substitution we observe favors the higher-wage group, consistent with the labor-labor substitution hypothesis.

In the theoretical formulation of the labor substitution hypothesis, both demand- and supply-side factors play a role. The minimum wage mandate gives firms an incentive to demand more high-skill labor or an opportunity to act on discriminatory tastes at the same time that it fosters an increased supply of high-skill applicants and those possessing nonproductive characteristics favored by employers. Indeed, given supply-side effects, it is possible for labor substitution of either sort to occur without employers altering their hiring standards or acting on their discriminatory tastes. A mere random draw from a higher-skill or more favored applicant pool would result in an increased use of high-skill or more favored labor. (12)

We are able to explore this issue in greater depth with the worker-firm matched data because the establishment survey asked firms whether they changed hiring standards as a result of the living wage. The hiring standards change variable captures whether firms sought to hire workers with greater skills, experience, schooling, or English language skills or who are more responsible. Firms that did not alter hiring standards and yet attained more skilled workers following the LWO arguably did so at the prompting of supply-side forces rather than as a direct, demand-driven substitution response to the wage increase. Of course, as more skilled workers swelled the ranks of the applicant pool in response to the increased wage, even employers who did not alter hiring standards may have discovered and acted on this increased availability. However, in the extreme, supply-side forces may be entirely responsible for the observed labor substitution if firms merely drew nonselectively from the skill-enhanced applicant pool.

In column 2 of Table 4, we present the results of specifications in which the firm-level hiring standards change variable is interacted with the joiner variable in the observable skills regressions. The estimated coefficient on the stand-alone joiner variable captures supply-side effects in that it reflects labor substitution in firms that did not alter hiring standards. The results suggest that supply-side forces are an important part of the explanation for our earlier findings. The difference between stayers and joiners in prior formal training is statistically significant in firms that did not alter hiring standards.

Interestingly, the signs on the joiner and joiner interaction terms suggest that, compared to stayers, joiners are less likely to have had prior formal training in firms where hiring standards did change than in those where hiring standards did not change. Indeed, the quantitative magnitudes of these coefficients imply that in establishments that changed hiring standards, joiners are less likely than stayers to possess prior formal training. These results may seem counterintuitive, but the first result has a plausible explanation. In firms that changed hiring standards, the stayer group is perhaps far more likely to reflect the best workers from among the pre-LWO workforce than in firms that did not change hiring standards, precisely because the former firms will have acted on their hiring standards intentions in dismissal decisions before our survey. (13) (Note that, if true, the stayer/joiner difference for this group of firms yields a downward-biased estimate of the leaver/ joiner difference--the true measure of labor-labor substitution.)

The second result--that joiners in firms that changed hiring standards are less likely than stayers to possess prior formal training--is more difficult to explain and perhaps leads to a concern about the overall integrity of these findings. However, models of asymmetric information regarding worker quality suggest that firms may know more about the quality of stayers than they do about the quality of new hires. If firms are unaware of the prior formal training of workers, and yet this training translates into increased worker productivity, it may not be wholly implausible to find that a select group of stayers possesses more of such training than do recent hires, and especially so in firms where standards have been recently ratcheted up.

Although we are unable to shed similar light on the hiring dynamics for favored demographic characteristics, supply-side forces may play a central role in explaining these results as well. Systematic differences across stayers and joiners with regard to demographic features such as .female, Latino, and black may reflect the increased opportunity for labor market discrimination that minimum wage mandates confer on employers or merely more passive hiring from an altered applicant pool made possible by the higher living wage (14)

Unobservable Substitution

Worker characteristics that are unobservable to the researcher and yet observable to participants in the hiring decisions of firms may also form the basis for labor-labor substitution. Strength, mental agility, diligent work habits, intensity of labor effort, physical attractiveness, and charm are among the many worker characteristics that are unobserved in the data. Is there any evidence of labor substitution based on these and other unobservable worker characteristics as a result of the living wage ordinance?

We tackle this question by comparing the "before" wages of stayers and joiners within establishments, holding constant observable worker characteristics provided in the data. Table 5 gives the results of this analysis. In column 1 we see that the before wages of joiners are statistically significantly different from, and roughly 20% above, those of stayers, all else constant. (15) To the extent this substitution is based largely on unobservable features related to worker productivity, comparing these results to those for observables suggests that labor substitution on unobservables such as mental agility or work ethic is much more important in minimum wage contexts than is substitution on observable skills such as education or experience. This is precisely what we might expect given the minimal employment loss observed within broad age and acquired skill categories in the empirical literature on minimum wage effects.

Is there evidence that labor-labor substitution on unobservables takes place in firms that did not alter hiring standards as a result of the LWO? The results of column 2 of Table 5 provide the answer. Once again, we find strong evidence for labor substitution among firms that did not change hiring standards, suggesting that supply-side, applicant pool effects are important. We also find evidence of greater labor substitution in firms that did not change hiring standards than in firms that did, and statistically significantly lower before wages of joiners compared to stayers in firms that changed hiring standards--both plausibly reflective of selection on highly valued stayers in firms that changed hiring standards. If true, the comparison of joiners with stayers yields an underestimate of true labor substitution which is found in a comparison of joiners with leavers. (16)

Wage Gain

We have found evidence to suggest that labor-labor substitution takes place on both observable and unobservable worker skill and personal characteristics in city contract firms as a result of the living wage ordinance. What does this imply about the dissipation of initial wage gains for workers in affected firms? Table 6 reports the conditional mean within-establishment raise experienced by joiners and stayers as estimated from the wage gain analysis. (17)

The findings reveal that although stayers experienced a 25% increase in pay as a result of the ordinance, joiners received only an 11% increase in pay in joining the city contract sector following the ordinance. These results may be used to estimate the wage impact of the ordinance both with and without labor substitution. The initial benefits of the ordinance to workers in affected firms can be estimated, assuming stayers resemble leavers, by applying the average wage increase for stayers to the entire post-LWO affected workforce in city contract establishments. This number may then be compared to the actual average wage increase with labor-labor substitution the population-weighted sum of the average wage increases for the two groups. Using this method, our estimate is that living wage gains for workers were dissipated by roughly 27% through the impact of labor-labor substitution. To the extent stayers are more highly valued than leavers, as seems plausible given our earlier findings, this represents an underestimate of the extent of wage dissipation.

6. Empirical Concerns

In this section, we address some of our concerns with the results of the empirical analysis. One concern regarding the unobservables analysis is that, on average, the before wages of stayers are observed at an earlier point in time than the before wages of joiners, thereby giving rise to a possible positive bias in the before wage comparisons. Differences in the timing of wage observations have been accounted for through the introduction of minimum wage categorical variables, but these may not be entirely satisfactory. With the wage histories provided in the data, we are able to build a longer panel of hourly wage observations for 254 workers in the sample; more than half of the sampled workers offered three or more historical wage observations, and 10% offered five or more distinct wage observations. Utilizing the longer panel of wage histories allows us to create greater uniformity across these worker categories in the timing of before wage observations. (18) In comparing not just immediately before wages but rather entire wage profiles that contain before wage observations distributed more evenly over time for both stayers and joiners, we are comforted to report that the statistical and even quantitative significance of the results are largely unchanged by this extension.

A second concern with the results, on both unobservables and observables, regards the absence of information on leavers. If stayers are more skilled or possess more favored personal characteristics than leavers, as we suspect might be true from the analysis of hiring standards changes, comparing joiners with stayers leads to an underestimate of the extent of labor substitution due to the LWO. However, if the opposite is true, then our results cannot be taken as a convincing test of the labor substitution hypothesis. (19) Perhaps joiners possess higher before wages than stayers because the former replace leavers who also possessed higher before wages than their stayer counterparts. Even in the absence of information on leavers, it may be possible to shed some light on this matter by asking where in the intrafirm, intra-occupational wage distribution joiners enter. If joiners fill the positions that leavers vacate, then their after wages, relative to others in the same firm and same occupation, will reflect the position of leavers in the before wage distribution.

For this exercise to be valid, stayers must not have been promoted into the jobs of leavers, and, despite the enactment of a wage minimum, there must exist significant intrafirm, intraoccupational wage dispersion. We address the first condition by interacting the joiner variable with the promotion from within variable in an analysis of after wages, which allows us to identify the after wage differences between joiners and stayers in establishments with no internal promotion policy. Regarding the second condition, an analysis of variance exercise reveals that 25% of the variation in after log wages remains even after controlling for occupation and firm fixed effects.

The results of this analysis reveal that no statistically significant difference exists in the after wages of joiners and stayers regardless of the propensity toward internal promotion. (20) This result is robust to a number of changes in specifications and samples, including an after wage regression in which each establishment is constrained to have at least two stayers and two joiners in the same occupation (wherein the sample size drops to 113 workers). These results leave us somewhat more confident that leavers are not a selective, high-skill draw from the pre-LWO workforce.

A final concern with these results is whether the differences we observe across stayers and joiners are truly due to labor substitution in reaction to the Living Wage Ordinance as opposed, for example, to some shock to the local labor market such as a slowdown in economic activity. This concern arises because we do not possess a control group of unaffected firms with which to establish a baseline counterfactual and to generate difference-in-differences estimates of the impact of the ordinance. We offer several responses to this concern.

First, recent research suggests that our finding that joiners possess significantly smaller wage gains than stayers is exactly the reverse of what we would expect to find in a control group of firms. Andersson, Holzer, and Lane (2005) have conducted a careful analysis of comparative wage gains for job stayers and job changers in the low-wage labor market, and their results strongly support the view that earning growth is greater for job changers than for stayers. (21) If we accept the Andersson, Holzer, and Lane (2005) result as a definitive control-group finding for this analysis, the labor substitution we observe is likely to be entirely attributable to the enactment of a living wage.

Second, we can exploit the fact that different firms come under the influence of the Living Wage Ordinance at different times and are thus hiring joiners under different macroconditions, to further explore this concern empirically. The late 1990s brought lower unemployment to Los Angeles, with the local unemployment rate averaging 5.5% for the years 1999 and 2000. However, unemployment jumped in 2001 and averaged over 6% for the years 2001 and 2002. Let us suppose that we designate the late 1990s as the "good macro years" and the period 2001 and 2002 as the "bad macro years." Any firm coming under the influence of the Living Wage Ordinance in the "good years" may hire joiners during both the good years and the bad years that follow. Firms coming under the influence of the ordinance in the "bad years" will hire joiners only during the bad years. Is there any significant difference in estimated labor substitution, on either observable or unobservable skills or favored demographic characteristics, across the two time periods?

The results reveal some clear differences in the extent of labor substitution over the two periods, but nothing that would cause us to fundamentally change the conclusions reached in the paper. There are differences in the joiner/stayer effect in only three cases, and only one of which pertains to our earlier findings. The results suggest that joiners are much more likely than stayers to be enrolled in school, to be native English speakers, and to be male in bad times as compared to good. This seems consistent with what we might expect: more highly valued workers are seeking positions in the low-wage labor market during bad times than during good times. The only finding that gives us some pause with regard to our earlier results is that the preference for hiring men takes place solely during the "bad years" and so indeed may be a macrophenomenon.

7. Conclusion

This paper utilizes an original employer-employee matched dataset on city contract establishments following the Los Angeles Living Wage Ordinance to explore the extent of labor-labor substitution resulting from a minimum wage. We test for substitution on observable and unobservable skills and worker demographic characteristics and measure the extent to which such substitution dissipates the benefits of a wage minimum for workers. The results suggest labor substitution toward male, Latino, and black workers and workers with a greater propensity to possess prior formal training. We also find evidence of significant substitution toward workers with more valuable unobservable skills and worker characteristics. This is evidenced by the finding that the "before" wages of workers new to city contract work following the ordinance are significantly higher than the "before" wages of city contract workers preceding the ordinance, controlling for differences in observables.

We find evidence of significant labor substitution even in firms that did not increase hiring standards following the ordinance, suggesting that applicant pool effects are important. More highly valued workers are drawn to employment in the city contract sector following the ordinance by virtue of the increased wage, and this appears to influence the extent of labor substitution even in firms with unaltered hiring standards. Curiously, contrary to expectations, we find that labor substitution is less extensive in firms that increased hiring standards. We take this to be suggestive evidence that stayers are a select draw from the pre-living wage workforce in these firms. If true, our overall estimates of labor substitution may be biased downward as the differences between joiners and stayers that form the basis of our analysis will underestimate the differences between joiners and leavers--the true measure of labor-labor substitution.

Finally, evidence suggests that the labor substitution we observe in these data is toward workers who are more highly valued in the labor market. This is trivially true in the analysis of unobservables, where wages proxy skills and preferred worker characteristics. But it is also true in our analysis of observables: men, Latinos, blacks, and workers with prior formal training all receive a wage premium in this segment of the low-wage labor market in Los Angeles. Because of this, the initial wage gains resulting from the Living Wage Ordinance will be dissipated over time with worker replacement. Firms may appropriate some of the initial worker rents by hiring more productive workers or workers with more preferred personal characteristics. In some cases the initial rent may be dissipated without any significant benefit to firms, simply because of nonselective draws from the more highly valued applicant pool following the ordinance. We estimate that initial living wage gains for workers were dissipated by roughly 27% through labor-labor substitution.

Table A1. Full Regression Results: Observable and Unobservable

                                                           Table 4

Independent variables         Prior Formal Training         Female

Joiner                          0.11 ** (0.06)         -0.14 *** (0.06)
Experience at hire             -2.74e-3 (0.01)         -0.01 (0.01)
Hire age                        7.84e-4 (4.43e-3)      -1.97e-4 (0.01)
Female                         -0.08 * (0.05)                 --
Years of schooling              0.01 (0.01)            -0.03 *** (0.01)
Currently enrolled              0.08 ** (0.05)          0.28 *** (0.11)
Native English speaker         -0.05 (0.07)             0.09 (0.12)
Prior formal training                  --              -0.21 ** (0.12)
Black                           0.03 (0.05)             0.04 (0.17)
White                           0.11 (0.10)             0.12 (0.17)
Asian                           0.12 ** (0.06)         -0.17 (0.18)
Union status                   -0.23 (0.37)            -0.47 (0.60)
Union                                  --                     --
Health insurance                0.02 (0.06)            -0.07 (0.14)
Health benefits                        --                     --
Current part-time              -0.08 * (0.05)           0.10 * (0.06)
Part-time job                          --                     --
Constant                        0.90 **** (0.22)        1.14 **** (0.41)
Number of occupations          42                      43
Number of LW establishments    40                      39
Number of MW periods                   --                     --
[R.sup.2]                       0.675                   0.585
Number of observations        251                     251

Independent variables                Latino                 Black

Joiner                          0.15 ** (0.08)          0.13 *** (0.06)
Experience at hire              2.27e-3 (3.91e-3)      -0.01 * (0.01)
Hire age                       -0.01 *** (3.40e-3)      0.01 (0.01)
Female                         -0.09 ** (0.04)          3.92e-3 (0.20)
Years of schooling              0.01 (0.01)            -2.42e-3 (0.02)
Currently enrolled              0.18 *** (0.09)         0.26 ** (0.15)
Native English speaker         -0.29 *** (0.11)         0.28 ** (0.15)
Prior formal training          -0.15 (0.18)            -0.19 (0.13)
Black                                  --                     --
White                                  --                     --
Asian                                  --                     --
Union status                   -0.29 **** (0.27)       -0.33 * (0.22)
Union                                  --                     --
Health insurance                0.05 (0.01)             0.12 (0.14)
Health benefits                        --                     --
Current part-time               0.07 (0.08)            -0.18 (0.19)
Part-time job                          --                     --
Constant                        1.25 **** (0.14)        0.22 (0.50)
Number of occupations          41                      23
Number of LW establishments    29                      15
Number of MW periods                   --                     --
[R.sup.2]                       0.624                   0.710
Number of observations        165                      86

                                   Table 5 (1)

Independent variables               Log-Wage

Joiner                          0.18 *** (0.07)
Experience at hire             -2.20e-3 (3.24e-3)
Hire age                        0.01 *** (3.82e-3)
Female                         -0.03 (0.03)
Years of schooling             -8.69e-4 (0.01)
Currently enrolled                     --
Native English speaker         -0.02 (0.07)
Prior formal training          -0.17 (0.14)
Black                          -0.05 (0.13)
White                          -0.11 ** (0.06)
Asian                          -0.15 *** (0.06)
Union status                           --
Union                           0.11 (0.10)
Health insurance                       --
Health benefits                 0.06 (0.04)
Current part-time                      --
Part-time job                  -0.03 (0.07)
Constant                        1.91 **** (0.15)
Number of occupations          32
Number of LW establishments    36
Number of MW periods            7
[R.sup.2]                       0.663
Number of observations        190

* = statistically significant at 15% level.

** = statistically significant at 10% level.

*** = statistically significant at 5% level.

**** = statistically significant at 1% level.

Table A2. Wage Regressions

Independent Variables         Before Wages            Recent Wages

Female                      -0.05 * (0.03)           0.01 (0.02)
Latino                       0.11 ** (0.06)          0.05 (0.05)
Black                        0.08 (0.08)             0.01 (0.04)
Asian                       -0.06 (0.07)             0.03 (0.07)
Experience                  -0.01 (0.01)             3.41e-3 (3.20e-3)
Experience2                  2.34e-4 (2.09e-4)      -4.52e-5 (6.49e-5)
Years of schooling           1.93e-3 (0.01)          2.67e-4 (1.95e-3)
Prior formal training              --                0.02 (0.03)
Native English speaker      -0.02 (0.05)             0.04 *** (0.02)
Union                        0.05 (0.04)             0.02 (0.03)
Health benefits              0.04 (0.07)             0.02 (0.02)
Part-time job                0.01 (0.07)            -0.01 (0.01)
Constant                     2.05 **** (0.16)        2.04 **** (0.09)
Number of MW periods         7                       2
Number of occupations       44                      40
[R.sup.2]                    0.5806                  0.5880
Number of observations     190                     166

* = statistically significant at 15% level.

** = statistically significant at 10% level.

*** = statistically significant at 5% level.

**** = statistically significant at 1% level.

Table A3. Wage Gain Regression

Independent Variables                           Log-Wage

Wage increase for stayers                0.22 **** (0.08)
Before wage difference joiner-stayer     0.10 ** (0.05)
After wage difference joine-stayer       0.02 (0.03)
Time elapsed                             7.48 x [10.sup.-4]
                                        (3.53 x [10.sup.-3])
Part-time job                            4.78 x [10.sup.-3] (0.01)
Union                                    0.06 (0.05)
Health benefits                          0.05 (0.03)
Constant                                 2.14 **** (0.12)
Number of MW periods                     7
Number of LW establishments             26
Number of occupations                   32
[R.sup.2]                                0.71
Number of observations                 306

** = statistically significant at 10% level.

**** = statistically significant at 1% level.

We have received helpful comments from Philip Babcock, Charlie Brown, David Card, Eve Caroli, Arin Dube, Shelly Lundberg, Mindy Marks, David Neumark, and participants in presentations at the Society of Labor Economists, UC-Berkeley, UCLA, and PSE in Paris. We acknowledge financial support from the Ford Foundation and the University of California Labor and Employment Research Fund, and research support of various kinds from researchers at the Los Angeles Alliance for a New Economy.

Received March 2007; accepted December 2007.


Abowd, John M., and Mark R. Killingsworth. 1981. Structural models of the effects of minimum wages on employment by age groups. Report of the Minimum Wage Study Commission. Washington, DC: U.S. Government Printing Office, pp. 143-69.

Abowd, John M., Francis Kramarz, and David Margolis. 1999. High wage workers and high wage firms. Econometrica 67:251-333.

Alpert, William T. 1986. The minimum wage in the restaurant industry. New York: Praeger.

Andersson, Fredrik, Harry J. Holzer, and Julie I. Lane. 2005. Moving up or moving on." Who advances in the low-wage labor market. New York: Russell Sage Foundation.

Brown, Charles. 1999. Minimum wages, employment, and the distribution of income. In Handbook of labor economies, vol. 3, edited by Orley Ashenfelter and David Card. New York: Elsevier, pp. 2101-63.

Card, David. 1996. The effect of unions on the structure of wages: A longitudinal analysis. Econometrica 64(4):957-79.

Card, David, and Alan B. Krueger. 1995. Myth and measurement. The new economics of the minimum wage. Princeton, N J: Princeton University Press.

DiNardo, John, Nicole Fortin, and Thomas Lemieux. 1996. Labor market institutions and the distribution of wages, 1973-1992: A semiparametric approach. Econometrica 64:1001-44.

Fairris, David. 2005. The impact of living wages on employers: A control group analysis of the Los Angeles ordinance. Industrial Relations 44:84-105.

Fairris, David, David Runsten, Carolina Briones, and Jessica Goodheart. 2005. Examining the evidence: The impact of the Los Angeles living wage ordinance on workers and businesses. Accessed 16 April 2006. Available at www.

Gibbons, Robert, and Lawrence F. Katz. 1992. Does unmeasured ability explain inter-industry wage differences? Review of Economic Studies 59:515-35.

Groshen, Erica. 1991. Sources of intra-industry wage dispersion: How much do employers matter? Quarterly Journal of Economies 106:869-84.

Grossman, Jean Baldwin. 1983. The impact of the minimum wage on other wages. Journal of Human Resources 18:359-78.

Hamermesh, Daniel S. 1996. Labor demand. Princeton, N J: Princeton University Press.

Holzer, Harry J., Lawrence F. Katz, and Alan B. Krueger. 1991. Job queues and wages. Quarterly Journal of Economics 106:739-68.

Neumark, David, Mark Schweitzer, and William Wascher. 2000. The effects of minimum wages throughout the wage distribution. NBER Working Paper No. 7519.

Neumark, David, and William Wascher. 1996. The effects of minimum wages on teenage employment and enrollment: Evidence from matched cps surveys. Research in Labor Economics 15:25-63.

Sicilian, Paul, and Adam J. Grossberg. 1993. Do legal minimum wages create rents? A re-examination of the evidence. Southern Economic Journal 60:201-9.

Wessels, Walter J. 1980. Minimum wages, fringe benefits, and working conditions. Washington, DC: American Enterprise Institute for Public Policy Research.

(1) Employment loss in our data was less than 2% following a roughly 20% increase in wage levels. Moreover, only 2% of firms reported changing "the equipment, machinery, or general way of doing the work"--indicative of substitution across broad skill/occupational categories--following the wage mandate.

(2) Studies have also shown that this clustering of wages is not offset by variation in nonwage benefits such as employer-provided health insurance, paid vacations, or bonuses. See, for example, Wessels (1980), Alpert (1986), and Sicilian and Grossberg (1993).

(3) See Hamermesh (1996) for both a recent statement of the theory and an exhaustive review of the empirical literature.

(4) For ease of exposition, we refer to all as "city contract firms."

(5) The weights take into consideration the stratification within establishments as well as the stratification between establishments. In other words, each worker's weight is the inverse of their probability of selection, which is the product of the probability of the establishment being selected among all establishments in the same industry-size stratum and the probability of the worker being selected among all workers in the stratum within the establishment. The whole establishment was a stratum for those cases in which no stratification was applied.

(6) Because unionized establishments are exempt from the provisions of the living wage ordinance, the high union density measures among workers might appear to pose problems for estimating labor substitution effects. However, we offer within-establishment estimates of labor substitution, and union density among establishments is much lower only 17.7% of establishments were unionized before the onset of the ordinance.

(7) Controlling for minimum wage periods presents an advantage over simpler year controls because some legislated minimum wage increases occurred within the year.

(8) The LWO benefited workers primarily through wage gains as opposed to nonmonetary benefits, such as additional health insurance. The only established nonmonetary improvement for workers was two additional paid days off (Fairris 2005).

(9) The full set of results for select regressions is reported in Table A1.

(10) Note, however, that only several years separate the hire date of stayers and joiners. At the time of interview, the average years of tenure for stayers was 5.57 (0.33 standard error): whereas, for joiners, it was 1.76 (0.14).

(11) The estimated coefficient for blacks is not statistically significant. The wage impact of prior formal training is analyzed in a regression using recent wages as the dependent variable to insure that the training is indeed before the current job of joiners. Workers with prior formal training earn wages 1% higher than those without such training. However, the effect is measured with such imprecision that we cannot conclude it is statistically different from zero. See Table A2.

(12) Holzer, Katz, and Krueger (1991) offer evidence that the applicant pool increases with a wage minimum. There is no research of which we are aware that explores the skills or demographic features of new applicants.

(13) Suggestive evidence in support of this explanation is given by the finding that stayers in firms that changed hiring standards are statistically more likely to possess prior formal training than stayers in firms that did not change hiring standards. We thank an anonymous referee for suggesting this test.

(14) These personal characteristics may also proxy for uncaptured human capital and so be entirely unrelated to discrimination based on gender, race, or ethnicity.

(15) Percentage changes are calculated using the formula: ([MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] - 1) x 100.

(16) As further suggestive evidence for the validity of this explanation, we compared the before wages of stayers in firms that did and did not alter hiring standards. The results reveal that stayers are of higher value--that is, possess higher before wages--in firms that altered hiring standards in reaction to the ordinance than stayers in those that did not, ceteris paribus.

(17) The wage gain regression results are given in Table A3.

(18) Note that we are able to accurately attribute only a limited number of worker and firm characteristics to these historical wage observations. The age, sex, and race and ethnicity of the worker at the time of the survey may be imputed without error. However, the timing of acquired schooling is not available in the survey, and so the current level of worker education is attached to each historical wage observation. The union status and health benefits provision variables are assigned to previous jobs based on responses concerning the job immediately preceding coverage by the living wage ordinance.

(19) We know nothing about dismissals versus quits for the leaver population. It is not entirely clear what to expect of the skill set or personal characteristics of workers who voluntarily quit following the ordinance. Those who quit due to bad matches, for example, are perhaps more likely to be lower paid and lower skilled, but voluntary exits might also be high among more highly skilled workers because their outside wage offers are greater.

(20) The control variables include all of those utilized in the before wage analysis above, except that current age and current experience are substituted for age at hire and experience at hire. In addition to the joiner variable and its interaction with the internal promotion variable, there is also a stand-alone variable capturing internal promotion intensity.

(21) Note that because the after wages of the stayer and joiner populations are statistically insignificantly different, the results of Andersson, Holzer, and Lane (2005) suggest we should expect the before wages of joiners to be less than the before wages of stayers in the absence of the LWO--exactly the opposite of what we observe.

David Fairris * and Leon Fernandez Bujanda ([dagger])

* Department of Economics, University of California, Riverside, CA 92521, USA; E-mail; corresponding author.

([dagger]) Oficina de Investigaciones Economicas, Banco Central de Venezuela, Av. Urdaneta, Esq. Carmelitas, Edf. Sede, Piso 2, Caracas, DC, Venezuela; E-mail
Table 1. Variable definitions

Variable                                   Definition

Employer Survey

Hiring standards change   Equals one if hiring standards changed due
                            to LWO; equals zero otherwise
Promotion from within     Percentage of positions beyond entry level
                            that are typically filled through
                            promotion from within the firm
Industry                  Industry categories

Worker Survey

Joiner                    Equals one if worker joined the city
                            contract operation of a former contractor
                            after the establishment was affected by
                            the LWO; equals zero otherwise
Age                       Years of age
Hire age                  Age when worker was hired by LW
Experience                Years of labor market experience after
                            age 16
Experience at hire        Experience when worker was hired by LW
Years of schooling        Years of schooling completed
Currently enrolled        Equals one if currently enrolled in school;
                            equals zero otherwise
Native English speaker    Equals one if English is first language;
                            equals zero otherwise
Received training         Equals one if received any training when
                            started current job; equals zero otherwise
Prior formal training     Equals one if completed any training before
                            starting current job; equals zero
Female                    Equals one if female; equals zero if male
Latino                    Equals one if Latino/a; equals zero
Black                     Equals one if African American or black;
                            equals zero otherwise
White                     Equals one if white; equals zero otherwise
Asian                     Equals one if Asian; equals zero otherwise
Health insurance          Equals one if the LW employer currently
                            provides health insurance for the worker;
                            equals zero otherwise
Current part-time         Equals one if hours per week at current LW
                            job is less than 35; equals zero otherwise
Before wage               Wage earned by stayers immediately before
                            their employer became subject to the LWO,
                            or by joiners before being hired by a LW
After wage                Wage earned by stayers immediately after
                            their employer became subject to the LWO,
                            or by joiners after being hired by a LW
Recent wage               Wage earned by workers at the time of the
Part-time job             Equals one if wage observation is associated
                            with working less than 35 hours per week;
                            equals zero otherwise
Health benefits           Equals one if the wage observation is
                            accompanied by the existence of
                            employer-provided health benefits; equals
                            zero otherwise
Union                     Equals one if the wage observation is from
                            a unionized employer; equals zero
Time elapsed              Years elapsed between before and after wage
Occupation                Occupational categories
LW establishment          Living wage establishments
MW periods                Minimum wage periods

Table 2. Descriptive Statistics from the Employer Survey

                                                Mean (Standard Error)
                                                [Standard Deviation]

Hiring standard change                               0.32 (0.09)
Promotion from within (a)                           48.62 (7.90)
Airline service                                      0.06 (0.03)
Landscape maintenance                                0.08 (0.04)
Janitorial                                           0.09 (0.05)
Miscellaneous                                        0.30 (0.08)
Retail and food                                      0.10 (0.07)
Security and parking                                 0.12 (0.06)
Social service                                       0.25 (0.09)
Number of sampled workers per establishment          5.05 [8.08]
Number of establishments                            44

(a) The sample size for this variable is 36.

Table 3. Descriptive Statistics from the Worker Survey

                                         Mean (Standard Error)

                                             Whole Sample

Age                                       39.48 (0.93)
Hire age                                  35.77 (0.87)
Experience at hire                        16.62 (0.91)
Years of schooling                        11.96 (0.37)
Currently enrolled                         0.17 (0.03)
Native English speaker                     0.37 (0.06)
Received training                          0.84 (0.04)
Prior formal training                      0.15 (0.04)
Female                                     0.54 (0.04)
Latino                                     0.47 (0.07)
Black                                      0.33 (0.07)
White                                      0.07 (0.03)
Asian                                      0.13 (0.03)
Latino (relative to white) (a)             0.87 (0.05)
Black (relative to white) (b)              0.82 (0.08)
Health insurance                           0.47 (0.12)
Current part-time                          0.23 (0.04)
Union status                               0.70 (0.10)
Wage before (a)                            7.37 (0.25)
Wage after (b)                             9.01 (0.16)
Recent wage                                9.22 (0.19)
Part-time job for before wages (c)         0.25 (0.04)
Health benefits for before wages (c)       0.31 (0.07)
Union for before wages                     0.37 (0.08)
Elapsed time (d)                           2.55 (0.40)
Childcare workers                          0.03 (0.01)
Cleaning service workers                   0.14 (0.07)
Landscape workers                          4.82e-3 (2.84e-3)
Parking service workers                    0.07 (0.04)
Restaurant workers                         0.05 (0.03)
Retail workers                             0.01 (0.01)
Social service workers                     0.01 (0.01)
Security service workers                   0.04 (0.03)
Miscellaneous occupations                  0.27 (0.09)
Airport service workers                    0.35 (0.14)
Supervisors                                0.03 (0.01)
Number of observations                   251

                                         Mean (Standard Error)


Age                                       42.27 (1.34)
Hire age                                  36.70 (1.47)
Experience at hire                        17.35 (1.47)
Years of schooling                        11.89 (0.50)
Currently enrolled                         0.15 (0.04)
Native English speaker                     0.35 (0.07)
Received training                          0.85 (0.05)
Prior formal training                      0.12 (0.04)
Female                                     0.57 (0.06)
Latino                                     0.44 (0.09)
Black                                      0.30 (0.06)
White                                      0.10 (0.04)
Asian                                      0.16 (0.05)
Latino (relative to white) (a)             0.81 (0.08)
Black (relative to white) (b)              0.75 (0.10)
Health insurance                           0.46 (0.14)
Current part-time                          0.16 (0.05)
Union status                               0.70 (0.11)
Wage before (a)                            6.24 (0.17)
Wage after (b)                             9.21 (0.23)
Recent wage                                9.47 (0.21)
Part-time job for before wages (c)         0.15 (0.05)
Health benefits for before wages (c)       0.39 (0.11)
Union for before wages                     0.65 (0.11)
Elapsed time (d)                           4.48 (0.32)
Childcare workers                          0.03 (0.02)
Cleaning service workers                   0.18 (0.08)
Landscape workers                          0.01 (4.44e-3)
Parking service workers                    0.04 (0.03)
Restaurant workers                         0.04 (0.03)
Retail workers                             4.87e-3 (0.01)
Social service workers                     0.01 (0.01)
Security service workers                   0.02 (0.02)
Miscellaneous occupations                  0.24 (0.10)
Airport service workers                    0.37 (0.15)
Supervisors                                0.05 (0.02)
Number of observations                   133

                                         Mean (Standard Error)


Age                                       36.57 (1.67)
Hire age                                  34.81 (1.63)
Experience at hire                        15.85 (1.31)
Years of schooling                        12.04 (0.37)
Currently enrolled                         0.19 (0.05)
Native English speaker                     0.39 (0.07)
Received training                          0.85 (0.05)
Prior formal training                      0.18 (0.06)
Female                                     0.50 (0.05)
Latino                                     0.50 (0.09)
Black                                      0.35 (0.08)
White                                      0.04 (0.03)
Asian                                      0.10 (0.05)
Latino (relative to white) (a)             0.92 (0.05)
Black (relative to white) (b)              0.89 (0.08)
Health insurance                           0.48 (0.13)
Current part-time                          0.31 (0.08)
Union status                               0.69 (0.12)
Wage before (a)                            8.43 (0.37)
Wage after (b)                             8.82 (0.21)
Recent wage                                8.99 (0.25)
Part-time job for before wages (c)         0.35 (0.04)
Health benefits for before wages (c)       0.24 (0.07)
Union for before wages                     0.11 (0.05)
Elapsed time (d)                           0.73 (0.18)
Childcare workers                          0.02 (0.02)
Cleaning service workers                   0.10 (0.06)
Landscape workers                          3.85e-3 (3.10e-3)
Parking service workers                    0.11 (0.06)
Restaurant workers                         0.06 (0.04)
Retail workers                             0.01 (0.01)
Social service workers                     0.01 (0.01)
Security service workers                   0.06 (0.05)
Miscellaneous occupations                  0.30 (0.11)
Airport service workers                    0.31 (0.14)
Supervisors                                0.01 (0.01)
Number of observations                   118

Latino (relative to white) is the proportion of Latinos in the
subsample of whites and Latinos. Similarly, black (relative to
white) is the proportion of blacks in the subsample of whites
and blacks.

(a) The sample size is 165. The number of stayers and joiners is
88 and 77, respectively.

(b) The sample size is 86. The number of stayers and joiners is
44 and 42, respectively.

(c) The sample size is 190 due to missing before wages. The number
of stayers and joiners is 96 and 94, respectively.

(d) The sample size is 182 due to missing wages. The number of
stayers and joiners is 94 and 88, respectively.

Table 4. Observable Characteristics Regressions

                                        (1)                 (2)

Experience at hire
  Joiner                           0.10 (0.79)         0.11 (0.75)
  Joiner X HS change                    --            -0.23 (0.85)

Hire age
  Joiner                           0.07 (0.66)         0.18 (0.63)
  Joiner X HS change                    --            -0.49 (0.95)

Years of schooling
  Joiner                          -0.09 (0.52)         2.86e-3 (0.63)
  Joiner X HS change                    --            -0.13 (1.01)

Currently enrolled
  Joiner                          -0.04 (0.06)         0.01 (0.07)
  Joiner X HS change                    --            -0.22 (0.16)

Native English speaker
  Joiner                          -0.07 (0.05)        -0.07 (0.06)
  Joiner X HS change                    --            -0.11 (0.19)

Prior formal training
  Joiner                           0.11 ** (0.06)      0.15 *** (0.07)
  Joiner X HS change                    --            -0.30 *** (0.12)

  Joiner                          -0.14 *** (0.06)          --

Latino (relative to white) (a)
  Joiner                           0.15 ** (0.08)           --

Black (relative to white)b
  Joiner                           0.13 *** (0.06)          --

Number of observations           251                 242

Control variables include union status, heath insurance, current
part time, occupations, and establishment fixed effects, in addition
to the other observable characteristics.

(a) The sample size in regression (1) is 165. The sample size in
regression (2) is 157.

(b) The sample size in regression (1) is 86. The sample size in
regression (2) is 82.

** = statistically significant at 10% level.

*** = statistically significant at 5% level.

Table 5. Unobservable Characteristics Regressions

                                  (1)                   (2)

Joiner                       0.18 *** (0.07)       0.26 **** (0.07)
Joiner x HS change                --              -0.35 **** (0.11)
Number of observations     190                   182

The dependent variable is the before wage. Control variables include
female, experience at hire, age at hire, years of schooling, native
English speaker, health benefits, union, part-time job, race/ethnicity
variables, occupations, minimum wage periods, and establishment fixed

*** = statistically significant at 5% level.

**** = statistically significant at 1% level.

Table 6. Wage Gain Regression Results

           Before LW      LW      LW Raise     Increase (%)     N

Stayer       6.20        7.73       1.53           24.7         77
Joiner       6.85        7.57       0.72           10.5         76

The Before LW of stayers is the average of immediately before wages
for this group. The Before LW of joiners is the Before LW of stayers
x (1 + [g.sub.JB]), where [g.sub.JB] = ([e.sup.[gamma]JB] - 1) and
[[gamma].sub.JB] is estimated from Equation 3. The percentage wage
increase of stayers is the transformed coefficient [[phi].sub.s] in
Equation 3, where the transformation is ([e.sup.[phi]s] - 1). The
percentage wage increase of joiners is ([e.sup.[gamma]JA] +
[[phi].sub.s] - [[gamm].sb.JB] - 1) x 100 from the same equation.
After LW is the sum of Before LW and the wage increase. The dependent
variable is before wage and after wage pooled together. The control
variables include time elapsed, union, health benefits, part-time
job, occupations, minimum wage periods, and establishment fixed
COPYRIGHT 2008 Southern Economic Association
No portion of this article can be reproduced without the express written permission from the copyright holder.
Copyright 2008 Gale, Cengage Learning. All rights reserved.

Article Details
Printer friendly Cite/link Email Feedback
Comment:The dissipation of minimum wage gains for workers through labor-labor substitution: evidence from the Los Angeles living wage ordinance.
Author:Fairris, David; Bujanda, Leon Fernandez
Publication:Southern Economic Journal
Article Type:Statistical table
Geographic Code:1USA
Date:Oct 1, 2008
Previous Article:Sources of economics majors: more biology, less business.
Next Article:The effects of retirement on physical and mental health outcomes.

Terms of use | Privacy policy | Copyright © 2019 Farlex, Inc. | Feedback | For webmasters