Santa Fe’s Living Wage Ordinance and the Labor Market

September 22, 2005

Dr. Aaron S. Yelowitz

Department of Economics

University of Kentucky

335 Business and Economics Building

Lexington, KY 40506-0034

Email: aaron@uky.edu

URL: http://gatton.uky.edu/faculty/yelowitz

Introduction

            In February 2003, the Santa Fe City Council approved the most expansive living wage ordinance to date. After sixteen months of legal wrangling, on June 24, 2004, a New Mexico state court judge upheld Santa Fe’s “living wage” law, and the ordinance immediately went into effect. Unlike most living wage ordinances, the Santa Fe living wage ordinance (hereafter, “LWO”) required all businesses within city limits with at least 25 workers to pay workers $8.50 an hour, rather than just businesses with city contracts. Hourly pay rates will increase to $9.50 on January 1, 2006, to $10.50 on January 1, 2008, and will be indexed to inflation starting on January 1, 2009.

            More than a year has passed since the LWO was enacted and it is appropriate to explore the labor market impact. The idea that the LWO could affect the labor market is well grounded in the economic literature. Santa Fe’s ordinance raised the wage floor from $5.15 per hour to $8.50, a 65% percent increase. Even with a fairly modest employment elasticity (such as the -0.22 elasticity estimate found in Neumark and Wascher’s (2000) compelling study), such a large change in the minimum wage is still likely to lead to substantial job loss.

Nonetheless, living wage advocates in Santa Fe paint a rosy picture of the law’s impact. The home page of the Santa Fe living wage advocacy group proclaims:

“Two Reports Show Living Wage Working in Santa Fe. Since the Santa Fe Living Wage has come into effect, public assistance is down sharply and employment is up.”

As evidence, they cite time-series data on welfare caseloads and employment. For example, their proof on the labor market effect is:

“According to the New Mexico Department of Labor, Montly (sic) News Release, Employment & Unemployment, August 25, 2005, job growth for Santa Fe was 2.0 percent, adding 1,200 jobs. The State had the same job growth rate, which makes it the 13th highest in the country. Most important, for the Santa Fe hospitality industry, which has the largest number of low-wage workers, the growth rate was even higher, 3.2%, or 300 new jobs.”

This conclusion should be surprising to those familiar with the minimum wage literature, since virtually all advocates for minimum wage increases claim they have zero effect on employment; virtually no serious economist would argue that at 65% increase in the wage floor would lead to employment growth. A key problem (true of time-series studies in general) is that other, unaccounted time-varying factors could create the illusion that the living wage is have a zero (or even a positive!) effect, when it reality the ordinance is having a negative effect. For example, a growing statewide economy could mask the true negative effects of the ordinance.

This study provides a more careful examination of Santa Fe’s ordinance. I explore the impact of the Santa Fe LWO on the labor market, drawing on publicly available microdata from the Current Population Survey (hereafter, “CPS”). To examine the impact of the LWO, my research examines more than 21,000 individuals aged 16 to 64 who were in the labor force in New Mexico. A key contribution is that individuals outside of Santa Fe metropolitan area (and those in Santa Fe before the LWO) serve as “control groups.” Such control groups provide a much more credible framework for evaluating the LWO than the evidence currently offered by living wage advocates. Having these control groups allow me to separately disentangle the effects of the LWO from other confounding factors (such as statewide economic growth or the fact that Santa Fe’s labor market is different from other areas of the state).

The results from the empirical models confirm the straightforward predictions from recent, credible minimum wage studies: higher wage floors hurt the labor market. I find that the likelihood of being unemployed increases and that usual hours worked decreases. Both results are statistically significant and economically meaningful. Moreover, the CPS reveals that the entire disemployment effect was concentrated amongst individuals with 12 or fewer years of education – precisely the group for whom the 65% increase in the wage floor should be the most binding.

Data Description

The CPS is a monthly survey of about 50,000 households conducted by the Bureau of the Census for the Bureau of Labor Statistics (hereafter “BLS”). The survey has been conducted for more than fifty years; the current analysis draws on the thirty months of CPS data between January 2003 and June 2005 (which is the latest available). The CPS data is free to download from the internet.

According to the BLS, the CPS is the primary source of information on the labor force characteristics of the U.S. population. The sample is scientifically selected to represent the civilian noninstitutional population. Respondents are interviewed to obtain information about the employment status of each member of the household 15 years of age and older. The sample provides estimates for the nation as a whole and serves as part of model-based estimates for individual states and other geographic areas.

The CPS asks about employment, unemployment, earnings, and hours of work. Some of these labor market questions are asked of the full monthly sample, and others are only asked of one-quarter of respondents. In addition, respondents are asked about their age, sex, race, ethnicity, marital status, veteran status, and educational attainment.

Critical for this study, the CPS also provides geographic identifiers for all states, and for many large metropolitan areas. The CPS allows geographic identification of three New Mexico metropolitan areas throughout the entire 2003 to 2005 time span – Santa Fe, Albuquerque, and Las Cruces. New Mexico residents in other parts of the state, by necessity, are grouped together. , During the period analyzed, the CPS surveyed roughly 1,800 individuals in New Mexico each month – and the 41 percent who were adults in the labor force form the sample that is analyzed. The sample ultimately consists of 21,776 observations, including 9,404 individuals with 12 or fewer years of education and 12,372 with 13 or more years of schooling.

Table 1 provides summary statistics for the sample as a whole, and broken out by educational attainment. Over the entire time period, 6 percent of individuals in the labor force were unemployed during a typical month, while those with less education were more than twice as likely to be unemployed as those with more education. Usual hours of work averaged 39.3 hours per week, with small differences by education group. Nearly three percent of the sample is classified as subject to the LWO – meaning that they participated in the labor force in the Santa Fe metropolitan area in June 2004 or later. Individuals in Santa Fe before June 2004 (as well as all individuals in other areas) are classified as unaffected by the ordinance.

Nearly 7 percent of this CPS sample is located in Santa Fe, nearly 11 percent in Las Cruces, 43 percent in Albuquerque, and the remainder is dispersed throughout the rest of the state. Fifty-four percent are married, 51 percent are male, and 43 percent have, at most, a high school diploma. Nearly 40 percent are of Hispanic origin, and the average age in the sample is 39 years.

Difference-in-difference estimation

            The basic model estimates an equation of the form:

             ,

where y_ict is the labor market outcome (either unemployment or usual hours of work), LWO_ict is an indicator variable equal to one if individual is subject to the ordinance, X_ict is a vector of other individual characteristics that affect work behavior, and D_it and D_ic

are indicator variables for time (month and year) and location (Santa Fe, Las Cruces, and the rest of the state). In some of the specifications, the month and year dummies are replaced with a time trend, but the results are nearly identical. Individual covariates include household size, a full set of dummy variables for age (from 17 to 64, with 16 as the omitted category), and indicators for married, head of household, male, high school dropout, high school graduate, some college, white, Hispanic, and veteran status.

            When LWO_ict, D_it and D_ic are included, the estimate on β₁ provides the “difference-in-differences” estimate of the impact of the living wage ordinance. The dummy variables for metropolitan area account for fixed, time-invariant differences between Santa Fe and other parts of the state. For example, to the extent that Santa Fe’s economy is more prosperous or dependent on tourism (and this remains fixed), then the metropolitan area controls will account for this heterogeneity on the labor market. The dummy variables for year and month account, respectively, for statewide growth in the economy over time and for seasonality. By including both sets of dummy variables, the true effect of the living ordinance, β₁ is obtained. The equation above essentially estimates how Santa Fe’s labor market changed after the ordinance, relative to other parts of the state.

            Although such a difference-in-difference estimator certainly provides more compelling evidence than time-series data alone, it does have its limitations. In particular, if there were other factors that were changing differently across cities over time, then it will be difficult to separately identify the effect of the living wage from those other factors. I have not been able to pinpoint any obvious explanations that vary in such a way (and affect employment), but the possibility does exist.

Findings

            Tables 2 and 3 present the basic results. Table 2 examines the likelihood of being unemployed in a given month with a probit model, while Table 3 examines hours of work for those who are employed.

            We first look at unemployment in Table 2. As the outcome of interest, unemployment, is a binary dependent variable, a probit model is estimated. In addition, the standard errors are corrected for clustering at the city/month/year level. The first and second columns of Table 2 examine the full sample, and include the all of the demographic controls mentioned above, as well as a time trend or month and year dummies. Before exploring the living wage results, it is important to note that the other independent variables have the expected impact on unemployment. In particular being married or white is associated with large reductions in unemployment, while being Hispanic or less educated is associated with large increase in the likelihood of unemployment. The probability derivatives, in italics, show the economic magnitude of the explanatory variables. The fixed effect for Santa Fe (measured relative to Albuquerque) shows that it has a persistently lower unemployment rate (of 1.7 percentage points). Nonetheless, the LWO reversed Santa Fe’s advantage – the measured impact of the LWO was to increase the unemployment rate by 3.2 percentage points.

            The next two columns focus on those on the less educated, while the final two columns focus on the more educated. The results reveal an extremely large effect for the less educated for whom the LWO is likely to be binding, while there is no effect for the more educated (either statistically or in terms of economic magnitudes). The entire negative effect of the living wage ordinance on unemployment is concentrated amongst the less educated. This conclusion does not change with the inclusion of either a time trend or month and year dummies.

            We next examine those who kept their jobs in Table 3. This table examines usual hours of work per week, which is very close to 40 hours in the full sample. The sample size is somewhat smaller than in the previous tables, because the unemployed and workers who report “variable hours” are excluded. Although some employers in Santa Fe might consolidate several part-time jobs into one full-time job in order to get under the 25-employee limit of the ordinance, the analysis shows that such an effect is clearly dominated by a reduction in hours for employees who are presumably at larger firms that are not close to the 25-employee limit. The ordinary least squares estimates show that for the sample as a whole, the LWO reduced usual hours of work by 1.6 hours per week. When broken out by educational attainment, hours fell by 3.5 hours for less educated workers, while they fell by a trivial amount for more educated workers (and that measured effect is statistically insignificant).

Discussion

            Although the living wage ordinance certainly raised wages for less-skilled workers in Santa Fe who kept their jobs, it had some severe consequences for many less skilled workers who were previously employed. It dramatically increased the unemployment rate for those with 12 or fewer years of education, and it reduced hours of work among this group as well. This hours reduction means that even for those who kept their jobs, total income rose less quickly than their wage rate. Table 1 shows that a typical less-educated worker had usual hours of work per week of 38.2 before the ordinance went into effect; afterwards it would have been 34.6 if he remained employed. For a worker previously earning $5.50 an hour, weekly earnings would go from $210.10 to $294.10, an increase of nearly 40%. But for a worker who initially earned $6.50 per hour, total earnings would have increased by only 18%, and would have fallen by 4% for some initially earning $8.00 per hour.

            The findings in this study should provide a cautionary tale for other localities that are considering such an ordinance. In Albuquerque, for example, voters will soon decide whether to implement a citywide wage floor of $7.50 per hour. In deciding how to cast their vote, these citizens should understand that there is no free lunch with living wages – they cause unemployment.