Minimum wage employment effects

In this study, the employment to population ratios and unemployment rates of 16- 19- and 20-24-year-olds are modelled as a function of the minimum wage. The federal minimum remains constant while the state minimum increases in some states from 2010- 2019. Such a research question is useful because some labor economists claim minimum wages reduce welfare by reducing employment. Thus, the purpose of this study is to replicate previous research designs, and compare results with different studies using different methodologies. It finds a slightly positive bias in the estimated effect of the minimum wage which likely accrues from omitted variable bias, such as state-level employment trend variables. An interpretation of the results is discussed below, and reference is made to the Danube, Lester, and Reich study which claims to have found a solution to the problem of unobserved heterogeneity in minimum wage-employment effect studies.

ABSTRACT In this study, the employment to population ratios and unemployment rates of 1619- and 20–24-year-olds are modelled as a function of the minimum wage. The federal minimum remains constant while the state minimum increases in some states from 20102019. Such a research question is useful because some labor economists claim minimum wages reduce welfare by reducing employment. Thus, the purpose of this study is to replicate previous research designs, and compare results with different studies using different methodologies. It finds a slightly positive bias in the estimated effect of the minimum wage which likely accrues from omitted variable bias, such as state-level employment trend variables. An interpretation of the results is discussed below, and reference is made to the Danube, Lester, and Reich study which claims to have found a solution to the problem of unobserved heterogeneity in minimum wage-employment effect studies. ii TABLE OF CONTENTS ABSTRACT ii INTRODUCTION 1 LITERATURE REVIEW 2 DATA/METHODS 4 FINDINGS 5 DISCUSSION 11 CONCLUSION 13 REFERENCES 14 iii INTRODUCTION The minimum wage is an important component of labor market regulation. If too aggressive of a minium wage law reduces employment, it may imply the minimum wage is less suitable of a policy tool than if the minimum wage ensures higher wages without reducing employment. For this reason, this study seeks to model employment as a function of minimum wages. It contrasts the methodology and findings with more credible studies with larger datasets. The economy experienced an expansion characterized by falling unemployment rates, low inflation, and stagnant or declining real wages for the bottom decile from 2010-2019. Many states raised their minimum wages, and many did not, as the federal minimum wage stayed unchanged. Moreover, the difference in minimum wages between states increased to greater levels than previously. Most minimum wage earners are young and female; 43% of minimum wage and below earners are between 16-24, as of a report from the BLS in 2021 i . A further examination shows there are 159,000 male as opposed to 325,000 female minimum wage and below earners. And, of the 484,000 minimum wage and below earners, 373,000 earned below the federal minimum wage. This suggests that an increase in the federal minimum wage would not directly raise wages for the 373,000 wage earners paid below the federal minimum, because these workers are regulated by state law rather than federal law. State minimum wages would need to increase in a number of states with no minimum wage laws or minimum wages below the federal amount. Lastly, although this analysis relies on aggregate tabulations, instead of gender differentiated ones, this remains an important area for extension. LITERATURE REVIEW In 1994, Card and Krueger surveyed employment levels in 402 fast food restaurrants in New Jersey and Eastern Pennslyvania before and after New Jersey’s minimum wage took effect on April 1st, 1992, increasing the minimum wage from $4.25 to $5.05 ii. They found no differences in employment between the fast food establishments located in New Jersey and those in Pennslyvania after the increase went into effect. In response to the Card and Krueger study, David Nuemark and William Wascher conducted a study which found adverse effects for at-risk groups iii, and the economists most closely associated with findings of negative elasticity of labor demand since have been David Neumark and William Wascher. Neumark and Wascher conducted a review of 102 minimum wage studies encapsulating the U.S and beyond at state and federal levels and argue the evidence overwhelmingly supported the conclusion that minimum wage 2 increases have negative employment effects, consistent with labor elasticities of -.1 to -.3 for low-wage iv, low-skill or young workers. According to Lester et. al, one key dividing point between studies is along methodological lines v. They are argue fixed-effects models with time and state effects dummy variables fail to control for unobserved heterogeneity influencing the results. They highlight that most of the studies that Neumark and Wascher review consist of fixed-effects models which fail to adequately control for unobserved heterogeneiety, and employment trends that are exogenous to minimum wage policies contaminate the minimum wage estimates. To illustrate this point, they show that employment growth in states without minimum wage laws was consistently higher from 1991 to 1996, but virtually identical from 1996 to 2006, a time-variant trend, and employment growth was intially higher in the South for reasons unrelated to the lack of minimum wage laws. To adequately control for unobserved heterogeneity the models must use essentially similar counties according to Dube et al. They solve this problem by using a sample of bordering counties with very similar characteristics, and the geographical proximity ensures that exogenous employment trends are less heterogenous in the sample. A sample with 337 counties with 66 quarters of data is used, and the variation in the minimum wage ranges from 7% to 20%. Testing six different specifications, they find negative employment effects of minimum wages in the national sample of counties with common period effects, but after adding Census-division-specific period fixed-effects and statelinear trends, they fail to find statistically significant findings. They also fail to find 3 statistically significant findings with their contiguous border county pair sample with county-pair specific period effects. DATA/METHODS Annual employment to population ratios and unemployment rate data for each state for the 16-19 and 20-24 year-old groups are obtained using the American Community Survey administered by the U.S Census vi, and data regarding minimum wages for all 50 states, not including Puerto Rico and D.C., is compiled for the years 2010-2019 from the FRED database vii. The minimum wage, the unemployment rate, and the employment to population ratio are all measured contemporaneously without lags. Unlike from 19902006, where the maximum variation in the minimum wage between counties reached 20%, variation in state minimum wages between 2010-2019 reached 50% between states with the lowest minimum wage and states with the highest minimum wage. Next, four specifications are created and run: (1)𝑈𝑈𝑈𝑈𝑈𝑈𝑈𝑈𝑈𝑈(16 − 19)𝑠𝑠,𝑡𝑡 = 𝛼𝛼 + 𝛽𝛽𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑠𝑠,𝑡𝑡 + 𝛾𝛾𝑡𝑡 𝑌𝑌𝑌𝑌𝑌𝑌𝑌𝑌 + 𝛿𝛿𝑠𝑠 𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆 + 𝜖𝜖𝑠𝑠,𝑡𝑡 (2)𝑈𝑈𝑈𝑈𝑈𝑈𝑈𝑈𝑈𝑈(20 − 24)𝑠𝑠,𝑡𝑡 = 𝛼𝛼 + 𝛽𝛽𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑠𝑠,𝑡𝑡 + 𝛾𝛾𝑡𝑡 𝑌𝑌𝑌𝑌𝑌𝑌𝑌𝑌 + 𝛿𝛿𝑠𝑠 𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆 + 𝜖𝜖𝑠𝑠,𝑡𝑡 (3)𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸(16 − 19)𝑠𝑠,𝑡𝑡 = 𝛼𝛼 + 𝛽𝛽𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑠𝑠,𝑡𝑡 + 𝛾𝛾𝑡𝑡 𝑌𝑌𝑌𝑌𝑌𝑌𝑌𝑌 + 𝛿𝛿𝑠𝑠 𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆 + 𝜖𝜖𝑠𝑠,𝑡𝑡 (4)𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸(20 − 24)𝑠𝑠,𝑡𝑡 = 𝛼𝛼 + 𝛽𝛽𝑚𝑚𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑠𝑠,𝑡𝑡 + 𝛾𝛾𝑡𝑡 𝑌𝑌𝑌𝑌𝑌𝑌𝑌𝑌 + 𝛿𝛿𝑠𝑠 𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆 + 𝜖𝜖𝑠𝑠,𝑡𝑡 Where s and t denote states and years, and “Year” and “State” are vectors of binary year and state dummy variables. The base year chosen is 2010, and dummy variables are included for all years other than 2010 to control for year effects. Alabama is chosen as the 4 anchor state, and all other states receive dummy variables to isolate state effects. Thus fixed-year and fixed-state effects are included in the model. A linear regression model is used with the classical OLS assumptions. FINDINGS No controls are added in the initial regression, and the minimum wage is left as the only explanatory variable. The estimated unemployment rate coefficient is -1.24 and -.787 for 16-19 and 20-24 year-olds respectively. This means a $1.00 increase in the minimum wage results in a reduction of the unemployment rate by 1.24 absolute percentage points from the unemployment rate for the 16-19 year-old group, and .787 absolute percentage points for the 20-24 year-old group. Both estimates are statistically significant at the 1% level and have adjusted rsquared values of .042 and .048. The coefficients for the minimum wages are .26 and .82 respectively with adjusted r-squared values of -.0003 and .0022 in the employment to population specification. These coefficients imply a .26 and .82 increase in the employment to population ratio percentage for the two groups. The first estimate is statistically insignificant, and the second estimate is significant at the 1% level. Next, year-effect dummy variables are included to isolate year-effects. The base year is 2010. This changes the unemployment rate coefficient to .47 and .266 for the same groups in respective order, both statistically significant at the 5% level with R-squared values of .47 and .49 respectively. The employment to population coefficients turn into .73 and -.133. The 16-19 coefficient is statistically significant at the 5% level, but the 5 Table 3: Summary of Minimum Wage Effects Explanatory variables included Minimum wage Minimum wage, yeareffects Minimum wage, yeareffects, stateseffects 16-19 20-24 Unemployment Unemployment Minimum wage coefficient: -1.24 P-value: .000 Adjusted Rsquared: .042 Minimum wage coefficient: -.787 P-value: 3.52e07 Adjusted Rsquared: .048 Minimum wage Minimum wage coefficient: .476 coefficient: .266 P-value: .025 P-value: .031 Adjusted RAdjusted Rsquared: .470 squared: .49 Minimum wage coefficient: -.344 P-value: .112 Adjusted Rsquared: .819 Minimum wage coefficient: -.105 P-value: .357 Adjusted Rsquared: .856 16-19 Employment to population ratio Minimum wage coefficient: .260 P-value: .360 Adjusted Rsquared: -.000 Minimum wage coefficient: -.735 P-value: .014 Adjusted Rsquared: .096 Minimum wage coefficient: .366 P-value: .032 Adjusted Rsquared: .903 20-24 Employment to population ratio Minimum wage coefficient: .823 P-value: .000 Adjusted Rsquared: .022 Minimum wage coefficient: -.133 P-value: .576 Adjusted Rsquared: .172 Minimum wage coefficient: .392 P-value: .011 Adjusted Rsquared: .885 DISCUSSION The exercise above illustrates the variety of results that might arise in minimum wage analysis as we alter how we measure the variable likely to be impacted (unemployment or the employment-to-population ratio) and as we control more rigorously for common time and state effects. Of course, one implication of the finding may be the presence of unobserved heterogeneity arising from differential employment time trends within each state. A second potential source of unobserved heterogeneity may be the effect 11 of changes in the population between ages 16-19 and 20-24, which may affect employment to population ratios, and if correlated with states with minimum wage increases, could explain some of the variation in employment to population ratios. The findings also suggest the employment to population ratio is a superior measure of employment levels than the unemployment rate, given the statistically insignificant results of the unemployment data, the lower R-squared values, and the condition in which workers exit the labor market after involuntary job loss, but aren’t counted in the official unemployment rate. Part of the reason the unemployment rates have a lower fit may be that the unemployment rate only includes individuals actively seeking a job. Discouraged workers temporarily, and even permanently, drop out of the workforce, no longer being counted in the unemployment rate, which tends to understate joblessness by not counting discouraged workers. The unemployment rate specification has a significantly higher Rsquared value in the model without state-fixed effects. This may be because employment to population ratio variation is substantially greater between states, and less strongly correlated with time. The two specifications with the employment to population ratios and fixed-time and fixed-state effects are better fits with greater R-squared values. The adjusted R-squared values of the employment to population ratio of 16-19 year-olds is .903 and .885 for 20-24 year-olds, while the adjusted R-squared value of unemployment rates of 16-19 year-olds is .819 and .856 for 20-24 year-olds. These specifications with fixed-time and fixed-state effects explain, in the case of the employment to population ratios model, over 90% of the 12 variation for 16-19 year-olds and 88.5% of the variation in employment rates for 20-24 year-olds, a strong fit. The findings also suggest the minimum wage is not a strong predictor of joblessness for the 16-19 and 20-24 year-old age group, or reduced job growth over time, as evidenced in employment to population ratios across time. CONCLUSION These findings add support to the conclusion that increases in the state minimum wages do not reduce the employment levels of 16-19 year-olds and 20-24 year-olds. Future increases in state minimum wages, or increases in the federal minimum, should not reduce youth employment levels. One implication of this finding is that instead of relying on measures like earned-income tax credits to reduce income inequality, by means of shifting the burden to taxpayers, minimum wages can simply shift the burden to employers, and go as far as to reverse the process of declining real wages in low-wage sectors without reducing employment for vulnerable groups. A $.50-1.00 increase in the federal minimum is a far greater percent increase in nominal wages than the average annual inflation rate, and would increase real wages for the bottom decile of wage earners. 13 REFERENCES i U.S. Bureau of Labor Statistics. (2022, April 1). Characteristics of minimum wage workers, 2021. U.S. Bureau of Labor Statistics. Retrieved May 2, 2023, from https://www.bls.gov/opub/reports/minimum-wage/2021/home.htm. Card, D., & Krueger, A. B. (1994). Minimum Wages and Employment: A Case Study of the Fast-Food Industry in New Jersey and Pennsylvania. The American Economic Review, 84(4), 772–793. http://www.jstor.org/stable/2118030 ii Neumark, D., & Wascher, W. (2000). Minimum Wages and Employment: A Case Study of the Fast-Food Industry in New Jersey and Pennsylvania: Comment. The American Economic Review, 90(5), 1362–1396. http://www.jstor.org/stable/2677855 iii Neumark, D., & Wascher, W. (2006). Minimum Wages and Employment: A review of Evidence from the New Minimum Wage Research. NBER Working Paper 12663. https://doi.org/10.3386/w12663 iv Dube, A., Lester, T. W., & Reich, M. (2010). Minimum Wage Effects Across State Borders: Estimates Using Contigous Counties. The Review of Economics and Statistics, 92(4), 945–964. http://www.jstor.org/stable/40985804 v US Bureau of the Census (2021) Table S2301: Employment Status (ACS 1 Year Estimates). https://data.census.gov/table?q=S2301%3A%2BEMPLOYMENT%2BSTATUS&tid=ACSST1Y 2021.S2301 vi Federal Reserve of St. Louis (nd), State Minimum Wage Rates. https://fred.stlouisfed.org/searchresults?st=minimum%2Bwages vii 14 Name of Candidate: Ali Cem Erturk Date of Submission: August 6th, 2023 15