Abstract
In this paper, I explore a novel method of estimating the elasticity of labor supply. Economists have traditionally based their estimates on employment-related outcomes such as hours worked, earnings, or employment status. But the drawback to these outcomes is that they are conditional on decisions of an employer and therefore do not purely reflect the labor supply response of workers. Instead, I examine how labor force participation changes in response to revisions to state minimum wage laws. Since labor force participation includes active job seekers, not just current workers, it more purely reflects willingness to work.
This paper builds off the work of economists like Mincer (1976) and Cave (1983) who theorized that higher minimum wages may draw some individuals into the labor force while causing others to drop out, discouraged by a reduced probability of finding employment. I use Current Population Survey data from 2006 to 2017 and a multi-year difference-in-differences model and find that, overall, there is no statistically significant effect on labor force participation. However, this coarse measure masks heterogeneous responses between men and women and across different levels of human capital: younger individuals with less education are significantly more likely to withdraw from the labor market, while workers with higher levels of human capital are more likely to enter. Further, I examine American Time Use Survey data and find that the minimum wage is not related to the amount of time people spend looking for work.
Introduction
Studies estimating the elasticity of labor supply typically examine the decisions of current workers in response to an exogenous change in their effective wage. Approaches include, among others: case study-like analyses of how a narrow group of workers responds to natural variation in daily earnings (Carrington, 1996; Camerer et al, 1997; Chou, 2002; Farber, 2005); randomized control trials that impose wage shocks in a single industry and geographic market (Fehr &
Goette, 2007); and natural experiments using government survey data to estimate the effect of changes in marginal income tax rates on employment decisions (Eissa, 1995; Eissa and Liebman, 1996; Blau and Wolfson, 2007). But these methods, which examine labor supply elasticity exclusively among current workers, are inherently limited in a way that is not immediately obvious: all the outcomes measure employment or are conditional on employment, and therefore
such studies fail to capture the willingness to work among individuals who are not actively
employed. These existing attempts at measuring elasticity of labor supply either at the extensive margin (becoming employed) or the intensive margin (adjusting one’s work effort or hours) are imperfect because they, naturally, require a favorable decision by another party: an employer who extends a job offer or accommodates a worker’s desire for a change in hours.
This paper considers a different approach of estimating labor supply elasticity on the extensive margin by focusing on labor force participation as the outcome variable. Because labor force participation includes individuals who are not only currently working but also those
actively looking for work, it better captures the underlying desire to be employed. I hypothesize that changes in state minimum wage rates, my independent variable of interest, affects the decision of whether to join or drop out of the labor market since the minimum wage represents not just the wage floor of those currently employed but also the earnings potential for those not
employed and those sitting on the sidelines of the job market. My empirical test uses a multi-year difference-in-differences model and examines 14 million individual observations from the 2006- 2017 Current Population Survey (CPS) published by the Bureau of Labor Statistics (BLS). I link the survey data to a state-month panel of minimum wage levels and other covariates.
My analysis includes several additional sub-analyses. First, recognizing that changes in minimum wage rates will most likely affect individuals near the bottom of the wage distribution, I subdivide my observations according to common measures of human capital: age (as a proxy for experience) and education. Similarly, I examine separate responses for men and women since the latter are often second earners in a household and therefore may exhibit more flexibility in their work decisions. Second, recognizing that not all workers immediately learn of and are able to respond to minimum wage changes, I examine elasticities over before and after the laws’ effective dates to determine the extent to which there are either anticipatory or lagging effects in decision-making. Finally, as a measure of labor supply elasticity at the intensive level
(conditional on being in the labor force), I consider a second outcome: the amount of time individuals spent in job search activities following minimum wage increases. For this last part, I use 2006-2017 individual-level survey data published by the American Time Use Survey (ATUS).
The questions explored here are important for policy makers. Their policy relevance derives in part from the large pool of affected workers, which can be categorized into three ripples. First are the 1.8 million individuals earning at or below the federal minimum wage (BLS 2017) and the presumably higher number who earn at or below their state wage floors. The number who earn at or below the state wage floors would presumably be considerably higher. But the effects ripple out more widely to workers earning above the old minimum wage but less
than the new, higher wage rate and who are, therefore, eligible for an increase. Researchers estimating the size of this second group suggest that more than one in ten workers would be swept up by a 10 percent increase in the minimum wage (Belman & Wolfson 2014). The final ripple of affected workers consists of those being paid more than the new minimum wage level. Employers who use a hierarchy of pay levels to differentiate workers of varying skills and contributions may wish to maintain the hierarchy by increasing the wages of, say, immediate supervisors already earning above the new wage floor. Research suggests that this last ripple effect may extend to workers earning up to 123 percent of the new minimum wage (Wicks-Lim 2006).
The question is also relevant because labor force participation rates are hovering near a 40-year low, having never recovered from the Great Recession. Demographic factors such as an aging population are commonly believed to explain most of the decline (Krueger 2017, Fujita 2013, Kudlyak 2013). Even after accounting for unfavorable demography, however, labor force participation is between 0.2 and 1.2 percentage points lower than what would have been
expected before the Great Recession (Aaronson et al 2014). And the rate for prime-age working
males (89.3 percent) is about five percentage points lower than 40 years ago (BLS).7
Understanding drivers to low labor force participation rates is important to research because of its effect on economic growth (Aaronson et al 2014). Further, the results from low labor force participation rates have been found to be similar to the outcomes of long term unemployment: more people applying for disability, an erosion of skills, and a drying up of job networks (van Zandweigh 2012).
I conclude that the overall labor supply elasticity with respect to minimum wage changes is negative but not statistically different from 0. Similarly, the effects of the minimum wage on the amount of time people spend in job searches is also statistically insignificant. These results differ from much of the existing literature that estimate elasticity by measuring the effect of actual wages (as opposed to potential wages effected by minimum wage laws) on employment. However, this broad, null result conceals a great deal of heterogeneity among responses.
Importantly, higher minimum wages are more likely to push out individuals with lower levels of human capital (younger, less educated workers) while drawing in those with higher levels. Also, as women age, the path of their labor force participation rates in response to minimum wage laws differs slightly from the path of men over their life cycle. Finally, higher minimum wages have a positive but statistically insignificant effect on the amount of time people spend looking for work.
Theoretical Underpinning
The simplest depiction of economic theory would suggest that minimum wage legislation would unequivocally increase labor market participation: supply increases as the reservation wages of more individuals are swept under the new legislated wage floor. The higher opportunity cost of leisure (or any other alternatives to paid employment, like unpaid homemaking) would incentivize more individuals to seek paid work. In an oversimplified analysis of theory, as illustrated by Marshallian supply and demand curves in Figure 2.1, the imposition of a wage
floor, wm, higher than the equilibrium level, wo, would result in upward movement along the
original supply curve, S, which, combined with reduced quantity of labor demanded from
employers would create a wide excess supply of job-seekers represented by the distance between
This naïve description, though, oversimplifies worker responses and the direction of their mobility. Mincer (1976, 1981), in more nuanced arguments, notes that workers not only respond to the obvious increase in wages but also to a reduced probability of employment (represented by
an inward pivot of the labor supply curve from S to S1 or even S2). Where the labor supply curve
ultimately settles is ambiguous since it is the sum of two opposing responses: a positive effect (entrants to the labor market, attracted to higher earnings potential) and a negative effect (withdrawals from the labor market, discouraged by a reduced likelihood of finding
employment). “The more interesting question,” Mincer concludes, “is not whether the supply of
labor is less than (quantity L on supply curve S), but whether it will be increased or decreased
from the (original) level,” that is, whether labor entrants dominate and the supply curve settles at
S1 or the withdrawals dominate with supply settling at S2 (1981, page 2).
Cave (1983) also addresses the opposing responses by labor. Workers with less human capital have a greater incentive to enter the labor force when wages at the bottom of the distribution are increased, particularly when firms have imperfect information about workers’ skill levels. At the same time, when minimum wage laws introduce rationing behavior by firms, awareness of this rationing discourages entry or continued job searching by those least likely to obtain jobs. The hypothesized direction of the minimum wage’s effect on labor force
participation is therefore ambiguous.
Theory would also suggest heterogeneous responses according to a number of parameters. I discuss two here. First, on the theoretical premise that individual labor force decisions are made at the household level and take into consideration the earnings of other family members, greater flexibility in entering or leaving the labor force may be predicted for non-primary earners and for those members whose non-employment contributions at home are
more valued, such as homemakers or caregivers (Blau & Khan, 2017). This would suggest that women, generally, would be more responsive to their own wage rates and to the wage rates of their partners. Second, minimum wages may have heterogeneous effects on individuals depending on where they are in the earnings distribution. A minimum wage increase should trigger more robust responses for those whose reservation wage is in between the original, lower wage floor and the new, higher rate. Greater flexibility is predicted therefore for individuals with lower levels of human capital who generally lie towards the bottom of the wage distribution. To illustrate heterogeneous responses, I examine differences in effects between the sexes and among individuals with differing levels of human capital as measured by education and age (as a proxy for experience).
Literature and My Contribution
Effects of the minimum wage on numerous outcomes, most prominently end
employment, have already been comprehensively studied (see meta-analyses by Neumark & Wascher, 2007; Doucouliagos & Stanley, 2009; Wolfson & Belman, 2016). And a substantive number of existing papers have used a variety of methods to estimate the labor force decisions of workers in response to changes in actual wages. Frequently those papers employed methods that were limited to individuals conditional on employment or focused their analyses on very
narrowly defined, homogenous groups of workers. My contribution to the literature is a less restricted estimate of labor supply elasticity. I leverage an under-explored method that includes all working age individuals, not just those with paid jobs. By measuring responses to variation in
state minimum wage rates, this analysis captures the willingness to work for both the employed
and unemployed, as opposed to outcomes that are conditional on employment and, therefore, encumbered by the decisions of an employer. Additionally, a majority of the labor supply
literature examines a specific group of individuals, such as married women or workers in narrowly defined occupations. In contrast, my analysis includes 14,167,090 individual observations from surveys of the general population.
The rest of this section examines the current literature and is divided into two sections: first, a summary of previous research using other approaches to estimate labor supply elasticity; and second, a summary of labor supply elasticity papers that are more similar to my approach, relying on minimum wage rates as an independent variable, albeit with limited analyses and much older data.
Papers using other approaches to estimate elasticity. A number of natural experiments have used observational data to estimate labor force responses of broad subgroups, primarily women, in response to changes in wages or marginal income tax rates. A discussion of the more prominent studies follows. Blau and Khan (2007) focused their analysis on married women, noting big swings in their labor supply behavior in the 1980s and 1990s. Using wage data from the CPS, they estimated, consistent with economic theory, that cross-wage elasticity with a husband’s earnings is negative and their own-wage elasticity is positive. Eissa and Leibman (1996) studied unmarried women with children following 1986 expansions of the Earned Income Tax Credit that substantively increased their effective wage rates. They found a statistically significant increase in labor supply on the extensive margin. However, their research showed no strong effect on hours worked conditional on being employed, which they posited may be in part due to the non-linear incentive structure of the EITC program. And research by Eissa (1995) estimated labor supply effects following the Tax Reform Act of 1986, focusing on married women at the very high end of the household income distribution (where marginal tax rates fell
the most) relative to women lower in the distribution (where marginal tax rates fell far less). She found positive, significant participation elasticities on both the external and internal margins.
A second group of natural experiments take a more case-study approach, focusing very narrowly on a single group of workers in specific jobs and specific geographical areas. In particular, the jobs are those in which workers exercise an unusually high degree of control over their decision to work on a given day and over their level of effort. These studies, similar to those above, estimate labor supply elasticity by relying on unexpected, transitory variation in wages caused by market shocks and other exogenous factors such as weather. Carrington (1996) performed an analysis on the Alaskan labor market – particularly in the construction industry – during the mid-1970s building of the Trans-Atlantic Pipeline System. The author noted the scope of the market shock: it was the most expensive privately financed construction project in world history and the largest localized demand shock in postwar United States. Moreover, it happened in a relatively isolated location, enhancing the project’s experimental nature. Carrington found positive labor supply elasticities on both the extensive and intensive (hours worked) margins with respect to wages. Oettinger (1999) exploited extensive data on the shifts worked and wages earned by mobile concessions vendors at a major league baseball stadium over an entire baseball season (81 home games). The decisions whether to sign up for a shift and how hard to work during that shift belonged entirely to the workers and were made based on demand shifters such as game time, day of the week, the standing of teams playing on that particular day, and the weather. The author found that workers’ daily labor supply decisions were highly responsive to expected earnings.
Two additional studies focused narrowly on cab drivers in New York City. Camerer et al (1997) found a positive elasticity of labor supply in the daily decision to work a shift but an
intra-day negative elasticity of effort with respect to earnings. That is, drivers were more likely to quit early on days where wages are high (implying that drivers did not have to wait very long between fares, such as during bad weather or during busy conventions) and more likely to work longer hours when wages were low. The authors rely in part on behavioral economics to propose why workers might respond to higher wages by actually decreasing their work schedule. Drivers demonstrate loss-aversion with the respect to their expected earnings on any given day, entering the market each day with a target for total earnings in mind and quit once that target is reached. Similar research by Chou (2002) into Singapore taxi industry also found, counter to neoclassical predictions but consistent with Camerer et al, that drivers worked fewer hours when wages were high. Farber (2005) responded to Camerer et al in a paper of his own on New York City cab drivers, arguing that conclusions in the original paper were limited by (negative) division bias and other sources of endogeneity. Using different methods and new data (but then corroborating his methods and results with data from Camerer et al), Farber found a positive elasticity. Next, using a hazard model, he tested the suggestion of Camerer et al that individuals quit once they reach a target income, but found that hours worked, not earnings, was what mattered to the drivers’ decisions to quit.
Beyond the number of natural experiments observed in the previous research, Fehr and Goette (2007) conducted a randomized control trial to estimate the response of bike messengers in Zurich, Switzerland to a temporary, anticipated wage shock. The authors increased the commission rate (the only form of pay) by 25 percent for four weeks for half the messengers chosen randomly, then repeated the experiment on the other half. Their results showed a larger increase in labor supply (measured by the number of shifts and hours worked) compared to many previous studies. But in their study, higher wages also led workers to exert slightly less effort on
average during those work hours, a negative effort elasticity, especially among workers with greater loss aversion who would be more likely to quit after hitting a daily earnings target (consistent with the Camerer et al finding regarding cab driver intensity of effort).
In summary, the case-study papers find positive measures of elasticity of participation and mixed results on elasticity of effort. These studies though may lack generalizability since the workers held an atypically high ability to make unconstrained choices about their daily
participation and effort. Additionally, each study was limited in time and focused on a single geographic area. Similarly, the randomized control trial of bike messengers generated similar results: strong, positive elasticity on the extensive margin and a slightly negative elasticity on the