THE ECONOMIC IMPACT OF A HIGH NATIONAL MINIMUM WAGE: EVIDENCE FROM THE 1966 FAIR LABOR STANDARDS ACT

Abstract

This paper examines the short and longer-term economic effects of the 1966 Fair Labor Standards Act (FLSA) which increased the national minimum wage to its highest level of the 20th Century and extended coverage to an additional 9.1 million workers. Exploiting differences in the “bite” of the minimum wage due to regional variation in the standard of living and industry composition, this paper finds that the 1966 FLSA increased wages dramatically but reduced aggregate employment only modestly. However, the disemployment effects were significantly larger among African-American men, forty percent of whom earned below the new minimum wage in 1966.

The 1966 Amendments to the Fair Labor Standards Act (1966 FLSA) capped almost 15 years of real minimum wage increases in the U.S., leading to the highest national minimum wage of the 20^th Century. In addition to raising the nominal hourly minimum by 28 percent to $11.83 (2019 dollars) for covered workers, the 1966 Amendments expanded coverage to 9.1 million workers in the economy’s lowest earning industries (Martin 1967).¹ Changes in coverage increased the share of private sector workers under the FLSA by 14 percentage points to 77 percent and the share of government employees under the FLSA from zero to 40 percent (Brown 1999).

This moment in history presents a unique opportunity to study the short and lagged economic effects of a very high national minimum wage with effects that persisted for newly covered sectors. Under both competitive and monopsonistic labor-market models, the sustained increase in wages could generate larger employment responses than more recent minimum wage changes that were rapidly eroded by inflation (Boal and Ransom 1997, Brown 1999).² Understanding the employment responses to the 1966 FLSA is important for evaluating the economic theory of labor markets and as a point of reference for contemporary policy proposals to raise federal, state, and local minimum wages to levels similar to the 1966 FSLA (Cooper, Schmitt, and Mishel 2015).

This paper quantifies the wage and employment responses to the 1966 FLSA by contrasting states that were more affected versus those that were less affected. Adapting Card (1992), our research design relies on the idea that the 1966 FLSA had a larger “bite” in states where wages and coverage were lower in 1966, thus allowing a “dose-response” analysis. To capture the impact of the 1966 FLSA on previously covered workers as well as for newly covered workers, we exploit differences across states in the share of workers below the new minimum wage of $1.60.³ Although nationally representative surveys of workers in our period do not ask about hourly wages, the 1960 Census (Ruggles et al. 2015a) and 1962-1974 March Current Population Surveys (CPS) (Ruggles et al. 2015b) show that the share of workers with implied hourly wages below the new minimum wage is highly and robustly correlated with state-level wage increases after 1966. This relationship allows us to examine locations where more workers were affected by the 1966 FLSA and, therefore, where we expect the legislation’s effects on wages and employment to be largest. Similar to Cengiz et al. (2019), a key benefit of our approach is that we can examine the effect of a minimum wage increase on all workers.

Our analysis begins with a quantification of the 1966 FLSA on wages. A dynamic, event-study framework estimates the wage and employment effects in the years before the Amendments took effect (leads provide a placebo test) as well as for the first seven years after implementation (lags characterize the post-legislation responses). The internal validity of the research design is consistent with wages in states with higher shares of workers earning wages below $1.60 following similar trends as less affected states from 1959 to 1966. However, the March CPS shows that men’s hourly wages increased significantly more in more affected states after 1966. Our estimates imply that states like Texas, where 26 percent of workers earned less than $1.60 per hour in 1966, experienced a 6 percent larger increase in average wages relative to New York, where 11 percent of workers earned less than $1.60 per hour. These results are robust to the inclusion of individual covariates for age, race, marital status, and metropolitan residence to account for changing composition of birth cohorts. This relationship holds within states as well. Hourly wages in lower-earning industries (that would have been disproportionately affected by the 1966 Amendments) increased by significantly more after 1966, even after including state-by-year fixed effects to account for differential, exogenous changes at the state level in the demand for or supply of workers. Across the U.S., our estimates suggest that average wages increased by 6.5 percent due to the 1966 FLSA, with around one fifth of the increase due to a higher minimum wage for previously covered workers and the remainder due to coverage increases and spillovers to higher earning workers.

In terms of hiring and hours, the March CPS shows that employment during the year fell by a modest 0.7 percent more in lower earning states and annual hours worked by 0.4 percent more, even though the 1966 FLSA increased wages significantly more in these areas. The implied demand elasticities are −0.14 for employment (a one-sided test rejects zero at the 5-percent level; 95-percent confidence interval (CI) is −0.29 to 0.02) and −0.07 for annual hours worked, which cannot be distinguished from zero. Interestingly, employment in the reference week fell little in response to the 1966 FLSA, suggesting that the legislation’s impact on employment was concentrated among workers with less attachment to the labor force (i.e., workers less likely to be employed for the full year).

An important alternative explanation for these findings is that areas more affected by the 1966 FLSA experienced exogenously slower growth in the demand for labor after 1966, which would lead our research strategy to overstate the negative employment response. To account for this possibility, we include time-varying, state-level controls for gross state product. Contrary to this hypothesis, areas with lower wages (which were more affected by the 1966 FLSA) were growing more quickly. Accounting for this faster growth results in slightly larger demand elasticities: −0.18 for employment during the year (95-percent confidence interval is −0.36 to −0.05), −0.28 for annual hours worked (a one-sided test rejects zero at the 5-percent level, 95-percent CI is −0.59 to 0.03), and a larger but statistically insignificant −0.10 for employment in the reference week.

A final analysis disaggregates these estimates by subgroups to examine the incidence of the legislation. For teens, we estimate larger but imprecise elasticities of employment with respect to wages. Among the 46 percent of men with less than a high school education, the long-run employment elasticity is −0.14 (95-percent CI is −0.34 to 0.06). The evidence is more decisive for African-American men. Their employment during the year decreased by 3.4 percent and annual hours worked fell by 5 percent after the 1966 FLSA was implemented when moving across the interquartile range. Employment in the reference week decreased by 1 percent for the same comparison. In contrast, changes in employment for white men were considerably smaller and statistically insignificant. The resulting demand elasticities for black men are statistically significant and range from −0.142 for employment at any point during the year (CI −0.35 to −0.14) and −0.421 for annual hours worked (CI −0.72 to −0.12). In summary, even if aggregate employment responded little to the 1966 FLSA, the legislation engendered compositional changes in employment and impacted some of the more disadvantaged workers in the economy.

I. The History and Expected Effects of the 1966 Amendments to the FLSA

At the time of their enactment, the 1966 Amendments were regarded as the most wide-ranging changes to the FLSA since 1938 (Levin-Waldman 2001, p. 112). The purpose of the legislation related closely to President Lyndon Johnson’s War on Poverty agenda. Proponents of this legislation stressed how increases in the coverage and level of the minimum wage would alleviate poverty and help struggling low-wage workers. The president of the AFL-CIO noted in June 1965 that, “The minimum wage law amendments now pending before Congress are ‘anti-poverty’ legislation, designed to improve the lot of the ‘working poor’.” Opponents of the legislation like the National Association of Manufacturing countered that the proposed “minimum [would]…be increased to a point where it would cause difficulty to those employing unskilled and inexperienced” (Levin-Waldman 2001, p. 113). Ultimately the proponents won the day. The 1966 Amendments were passed on September 23, 1966, with their first provisions effective in February 1967 (P.L. 89-601).⁴ This national minimum wage was binding, with its level exceeding the state minimum in all but a handful of cases (Quester 1981, Sutch 2010).⁵

The impact of the 1966 Amendments was expected to be large enough that they were challenged as unconstitutional. In Maryland v. Wirtz (1968), the state of Maryland (later joined by 27 other states and a school district) argued that the Supreme Court should enjoin the Act on the basis that its provisions exceeded Congress’s authority to regulate under the Commerce Clause—in particular, the states objected to requirements that they meet federal pay and overtime standards in their schools and hospitals. The 1966 Amendments survived this challenge. On June 10, 1968, the Warren Court affirmed the 1966 Amendments and instructed states to enforce them.

A. Increases in the Statutory Minimum Wage for Previously Covered Workers

The 1966 Amendments raised the real minimum wage for covered workers to its highest level in the twentieth century, as shown in Figure 1A. To minimize the burden on firms, they were phased over two years (Martin 1967). On February 1, 1967, the statutory minimum wage for covered workers increased from $1.25 to $1.40 ($9.60 and $10.76 in 2019 dollars). In its report to Congress, the Department of Labor estimated that 3.72 million covered workers would benefit from this increase (Ibid, p. i). The second minimum wage hike occurred the following year on February 1, 1968 and increased the statutory minimum wage for covered workers to $1.60 ($11.83 in 2019 dollars). This amounted to a 28-percent nominal increase over two years, or a 23-percent increase in real terms. The effective wage increase for many of the lowest earning, previously uncovered workers was significantly larger (Kocin 1967), as we discuss next.⁶

Figure 1. Real Minimum Wages in the U.S., 1938 to 2018.

A: Real Minimum Wage and Minimum Wage Relative to Median Full-Time Wage

B: The 1966 FLSA and the Increase in the Effective Minimum Wage

Notes: Nominal minimum wages are inflated to 2019 dollars using the CPI-U (U.S. city average for all items, CUUR0000SA0, https://data.bls.gov/timeseries/CUUR0000SA0). In Panel A, the solid black line displays the statutory federal minimum wage in effect for the majority of the year, and the dashed black line shows the minimum wage relative to the median full-time wage from the OECD (2019). The bold red line in Panel B is constructed as the weighted average of the real minimum wage levels and the share of workers employed in industries first covered by each FLSA amendment (see Derenoncourt and Montialoux 2019, Appendix Table A.1).

B. Increases in Coverage and Statutory Minimum Wages for Previously Uncovered Workers

A major feature of the 1966 FLSA was its dramatic expansion of coverage. Figure 1B shows the federal statutory minimum wage in 2019 dollars for workers covered under the 1938 FLSA and workers added in the 1961 FLSA. In April of 1967, the Monthly Labor Review estimated that the 1966 Amendments had expanded the FLSA’s coverage to an additional 9.1 million workers, up from 32.3 million workers covered under previous legislation (Martin 1967). This happened through the 1966 FLSA’s narrowing of exemptions as well as its expansion of industries covered under the “enterprise volume test.” Figure 1B shows the changes in the statutory minimum wage for workers newly covered under the 1966 FLSA. (Note that the pre-1967 wages for newly covered workers were not zero—we use zero to represent the absence of the federal statutory minimum wage.)

The increase in coverage occurred through a direct expansion of the legislation to include employees on large farms, federal service contracts, federal wage board employees, and certain Armed Forces employees (e.g., post exchanges). It also narrowed or repealed exemptions for employees of hotels, restaurants, laundries and dry cleaners, hospitals, nursing homes, schools, auto and farm implement dealers, small loggers, local transit and taxi companies, agricultural processing, and food services. The 1966 FLSA also included an indirect expansion of coverage through its reduction in the enterprise volume test from $1 million (in the 1961 Amendments) to $250,000 within three years.⁷ This meant that employees of smaller firms engaged in “interstate commerce” gained coverage by February 1, 1969.⁸ As a consequence of both these changes, 95 percent of newly covered workers were employed in five industries (Martin 1967). Just over three million (3.1) of the newly covered workers were in services,⁹ 2.4 million in government,¹⁰ 2.2 million in retail trade, 0.6 million in construction, and 0.5 million in agriculture.

The 1966 FLSA specified different wage increases for newly covered workers. Newly covered non-farm workers began at a minimum wage of $1.00 per hour in 1967 with increases of $0.15 per year to reach $1.60 by 1971.¹¹ Newly covered farm workers began at a minimum wage of $1.00 in 1967 and increased by $0.15 per year to reach $1.30 in 1969, which is why the series in Figure 1B diverges for farm and non-farm workers after 1969.

The 1966 FLSA also applied overtime provisions to newly covered workers. As of February 1, 1967, newly covered workers working more than 44 hours per week were paid time and a half. In 1968, this maximum fell to 42 hours per week and, in 1969, to 40 hours per week.¹²

Documenting the impact of the 1966 Amendments on the wages of previously uncovered workers is difficult because (as we show in the Online Appendix) measurement error in the March CPS hourly wage is particularly acute near the minimum wage. To place our subsequent estimates in context, we entered Bureau of Labor Statistics (BLS) tabulations of industry surveys both before and after the 1966 Amendments took effect. Because these studies did not rely on nationally representative samples (they are localized to certain cities, regions and industries) and because non-compliance may be underreported to the federal government, extrapolating from these findings to the state and national impact of the 1966 FLSA is difficult. These reports, nevertheless, cover changes in the wages of approximately two-thirds of the newly covered workers, including about half of the service industry (employees of laundries, schools, nursing homes, and hospitals); about two thirds of the newly covered government workers (employees in public schools and government hospitals); all workers in retail trade; and all workers in agriculture.

In laundries in 1966, 72.5 percent of all U.S. employees and 89.3 percent in the South earned under $1.60 per hour. By 1968, those figures had fallen to 48.7 and 73.6, respectively. Between 1966 and 1968, the average industry wage increased by 16 percent in the U.S. and 23 percent in the South. Similarly, average weekly hours fell from 38.7 to 36 as compliance with new overtime provisions increased.

Because nursing homes, hospitals and public schools received public funding like Medicaid, Medicare, and Title I funds from the Elementary and Secondary Education Act, we expect even greater compliance in these industries (Almond, Chay, and Greenstone 2003, Cascio et al. 2010). Data on hospitals accord closely with this hypothesis. For instance, 43.4 percent of nonsupervisory employees in non-governmental hospitals earned less than $1.60 per hour in July 1966 and average hourly earnings were $1.83. By March 1969, the share of workers earning below $1.60 per hour had fallen to 11.2 percent and average hourly earnings increased by 35 percent to $2.47. Average weekly hours fell from 36 to 34.7, as the share of employees working over 40 hours fell from 15.7 to 10.9 percent.

C. Expected Effects of the 1966 FLSA on Wages and Employment

The literature on the minimum wage is so vast that “we are almost at the point where there are meta-studies of meta-studies” (Manning 2016).¹³ One area of consensus is that increases in the wage floor should raise wages. However, quantifying the magnitude of the effects of the 1966 Amendments on wages in the U.S. economy is difficult owing to a lack of information on the number of directly affected previously covered individuals as well as the number of newly covered individuals. Our analysis uses the 1960 Census and the March CPS, nationally representative datasets of workers for our period of interest, to estimate the national impact of the 1966 Amendments on wages as well as the lag structure of these adjustments.

The magnitude and speed of the wage responses also inform expectations about the 1966 Amendments’ effects on employment, which are theoretically ambiguous. In the classic (perfectly competitive) labor-market case, the aggregate labor demand and labor supply curves pin down wages and employment at the competitive equilibrium. In the monopsonistic case, the marginal cost of hiring additional workers lies above the aggregate labor supply curve. The intersection of the marginal cost curve and demand curve pin down the labor market equilibrium, where both employment and wages lie below the perfectly competitive equilibrium. A key result in standard monopsonistic models is that the imposition of a wage floor up to the perfectly competitive level could raise employment to the perfectly competitive level. In both models, however, raising wages above the wage set in a perfectly competitive labor market would lower employment. In standard two-sector models of the labor market, increasing the coverage rate (or the probability of finding a covered sector job) should exacerbate the effects of raising the minimum wage (Brown 1999). Finally, monopsonistic firms may also engage in wage discrimination. Assuming they have some information about the labor supply elasticities of different groups, firms could pay workers with lower labor-supply elasticities (potentially due to fewer outside options or lower incomes) lower wages (Boal and Ransom 1997).¹⁴

It is doubtful that the labor market is a pure form of perfect competition or monopsony, so these predictions benchmark extremes with the actual labor market lying somewhere between. The important theoretical prediction is that both competitive and monopsonistic labor-market models suggest that a high enough minimum wage should reduce employment. There is, however, less agreement on the point at which this high level of wages would be reached. The 1966 FLSA presents a unique opportunity to study the short and lagged economic effects of the highest national minimum wage of the 20^th Century—a level similar to recent policy proposals. In addition, the 1966 FLSA represents a permanent increase in the minimum wage for a large number of newly covered workers. Our analysis considers both the magnitude of disemployment effects and whether these effects varied by group of worker.

II. Evaluating the Economic Effects of a National Minimum Wage

Our research design follows the spirit of Card (1992), who makes use of the long-standing criticism of the national minimum wage—namely, that geographic variation in the cost of living makes the impact of a national minimum wage larger in some areas (Stigler 1946). For instance, the same nominal minimum wage in New York would be effectively much higher in Texas after accounting for the cost of living. This geographic variation in cost of living means that imposing a high, uniform, and national minimum wage should have differential real impacts on local economies, allowing a dose-response style analysis.

Card (1992) exploits this fact in a simple two-period model to study the 1990 national minimum wage increase. Focusing on teens, a group largely earning the minimum wage, Card uses variation in the fraction of workers affected by the change in the national minimum wage, $F_s^{\ast }$, as an instrumental variable in the following two-equation model:

$$\Delta \log W_s={\gamma }_1+{\gamma }_2{F}_s^{\ast }+{\mathit{X}}_s^{\prime }{\gamma }_3+{\epsilon }_s\text{and}$$ (1)

$$\Delta E_s={\beta }_1+{\beta }_2\Delta W_s+{\mathit{X}}_s^{\prime }{\beta }_3+{\omega }_s.$$ (2)

The dependent variables, Δ log W_s and ΔE_S, capture the change in mean log wages or employment rates (employment-to-population ratio) among teens in state s during a period before and after the minimum wage increase. In some specifications, X_s is the employment-to-population ratio among all workers or the overall unemployment rate. $F_s^{\ast }$ is the number of workers earning above the old minimum wage and below the new minimum wage, divided by the number of workers in the state. Thus, $F_s^{\ast }$ captures the “bite” of the minimum wage as the fraction of workers in a state who would be affected by the 1990 national minimum wage increase. Card finds evidence that an increase in the federal minimum wage generates greater wage gains in states with a greater fraction of workers affected, showing that γ₂=0.15. Then, Card tests whether employment falls more in places where the fraction of workers affected by the minimum wage was higher, or β₂ < 0. As he notes, β₂ is proportional to the labor demand elasticity in this simple model.

Our analysis uses a nationally representative sample of prime-age (16-64), male workers from the 1960 Census and annual 1962-1974 March CPS.¹⁵ This broad age range is important for capturing the national effects of the legislation, as employers may have substituted hiring across age or skill groups in response to the 1966 FLSA. We also exclude self-employed workers, who are not covered under the FLSA.¹⁶ To increase consistency between the CPS and Census, we also restrict the Census sample to individuals not living in institutional group quarters. Finally, we convert income and wages into 2019 dollars using the CPI-U and index wages and employment to the relevant year (annual earnings and weeks worked refer to the year before the survey, while employment in the reference week does not). See our Online Appendix for more details.

The CPS allow us to extend Card’s (1992) methodology in several ways. First, we estimate a dynamic version of his two-period model to examine how wages and employment changed from 1959 to 1973 in response to the 1966 FLSA.¹⁷ Second, we use the share of workers earning below $1.60 in 1966 (rather than the share of workers earning between the old and new minimum) as a measure of the state’s labor market that is “potentially affected.” This measure ameliorates concerns regarding measurement error in implied hourly wages in the March CPS. In periods where we observe both reports of hourly wage earnings (Monthly Outgoing Rotation Group, ORG) and annual wage earnings (March CPS) for the same year, the distribution of implied hourly wages is very similar for workers earning just above the minimum wage (see Online Appendix), whereas the March CPS measure of implied hourly wages severely misstates the share of workers earning between the old and the new minimum wage. In addition, our cumulative measure captures the 1966 FLSA’s increase in coverage that impacted wage earners below the old minimum wage of $1.25 per hour.¹⁸ While the available data do not allow us to precisely measure workers exposed to the coverage expansion, the share of wages below $1.60 captures this better than using the share of wages between $1.25 and $1.60.¹⁹

Figure 2 illustrates the spirit of this approach, plotting kernel density estimates of the implied hourly wage in different states in 1966. We construct hourly wages in the March CPS by dividing annual wage earnings in the previous year by the mean of weeks worked within each reported category in the previous year and hours worked in the week before the survey.²⁰ For a given change in the nominal minimum wage, the share of workers affected (approximated as the share with wages between $1.25 and $1.60) is larger in lower earning states (such as Texas) than in higher earning states (such as New York). Notably, however, Card’s (1992) fraction affected does not capture changes in the FLSA’s coverage that also extended to workers earning below $1.25 per hour—a crucial feature of the 1966 legislation that motivates our use of the cumulative share of workers below $1.60 per hour. This measure is largely correlated with the share of workers between the old and the new minimum but also accounts for concentration of low wages outside the covered range. Because the impact of the 1966 FLSA should be larger in lower earning states, economic theory predicts that the law’s effects on wages and employment should also be larger.

Figure 2. The 1966 Average Hourly Wage Distribution, March CPS.

Notes: Figure displays log real March wage densities (in 2019 dollars) for men ages 16 to 64. The densities are estimated only among wages between the 0.5th and 99.5th percentiles of the aggregate wage distribution. Densities are weighted by the product of the CPS weight and the annual number of hours. Texas and New York are at the 25^th and 75^th percentiles of the share of wages below $1.60 in 1966. The vertical lines correspond to the federal minimum wage before and after the 1966 FLSA. Source: 1967 March CPS.

Table 1 displays the variation in Fraction Affected, the share of workers earning below the 1966 FLSA new minimum wage in the year before it took effect, and Figure 3A presents this information as map form, where darker shades capture a higher share of wages below $1.60 in 1966. As noted in Department of Labor wage studies, the share of wages below $1.60 in 1966 was much higher in the South and interior states. However, there is substantial variation within the South and interior states in the bite of the statute, which our study leverages.

Table 1.

Share of Workers with Hourly Wages below the 1966 Minimum Wage of $1.60, by State Group

State group	Fraction affected: Share of workers with hourly wages below $1.60 in 1966
New Jersey	0.083
Alaska-Hawaii-Oregon-Washington	0.090
California	0.091
Illinois	0.094
Ohio	0.098
New York	0.107
Pennsylvania	0.109
Michigan-Wisconsin	0.111
Connecticut	0.117
Indiana	0.130
Maine-Massachusetts-New Hampshire-Rhode Island-Vermont	0.152
Delaware-Maryland-Virginia-West Virginia	0.166
Arizona-Colorado-Idaho-Montana-Nevada-New Mexico-Utah-Wyoming	0.176
Iowa-Kansas-Minnesota-Missouri-Nebraska-North Dakota-South Dakota	0.193
Washington, DC	0.223
Texas	0.257
Georgia-North Carolina-South Carolina	0.259
Kentucky-Tennessee	0.279
Florida	0.291
Arkansas-Louisiana-Oklahoma	0.319
Alabama-Mississippi	0.392
United States	0.161

Notes: Table reports the share of men ages 16-64 with average hourly earnings below $1.60 in 1966. Sample includes men not residing in group quarters or in the military and for whom self-employment income accounts for no more than 10 percent of total income. The rows indicate the 21 state groups consistently identified in the CPS for our sample period. Source: Authors’ calculations using the 1967 March CPS.

Figure 3. The Share of Workers in 1966 Earning below the 1966 FLSA Minimum Wage of $1.60.

A. The Share of Hourly Wages below $1.60 in 1966, by State Group

B. The Share of Hourly Wages below $1.60 in 1966, by State Group and Industry

Notes: Panel A shows the share of hourly wages below $1.60 in each state group (see also Table 1). Panel B shows the share of hourly wages below $1.60 in each state-group/industry cell (y-axis) by state-group fraction below $1.60 (x-axis). Variation in the vertical dimension shows variation in the bite of the 1966 FLSA within state groups. We use 10 one-digit industries. Source: 1967 March CPS.

Our analysis presents the reduced-form estimates using the following event-study (equation 3) and differences-in-differences (equation 4) specifications:

$${\mathrm{Y}}_{s,b,t}=\sum \begin{array}{c}\hfill \\ k\hfill \end{array}{\theta }_{k}1(t=k)F_{s,1966}+{\mathit{X}}_{s,t}^{\prime }\beta +{\gamma }_{s,b}+{\delta }_t+{\epsilon }_{s,b,t}$$ (3)

$${\mathrm{Y}}_{s,b,t}=\tilde{\theta }1(t>1966)F_{s,1966}+{\mathit{X}}_{s,t}^{\prime }\beta +{\gamma }_{s,b}+{\delta }_t+{\epsilon }_{s,b,t}.$$ (4)

Outcomes in the March CPS are average log hourly wages and employment during the year, reference week, or average annual hours worked in state group s, birth cohort b, where b ranges from 1895 (age 64 in 1959) to 1958 (age 16 in 1974), and year t, where t ranges from 1959 to 1973 for employment during the year or hourly wages and 1960 to 1973 for employment in the reference week.²¹ In equation 3, we normalize θ₁₉₆₆ = 0 for 1966, the year before the FLSA took effect and the year we measure the share of wages below $1.60, F_s,1966. State fixed effects, γ_s, account for time-invariant differences across states, such as unchanging differences in legislation, geography, resource endowments, and cost of living. Year fixed effects, δ_t, account for national changes across years that may also affect wages: large tax cuts (1964), the Civil Rights Act (1964) and Voting Rights Act (1965), Medicare (1966), as well as other Great Society legislation (Bailey and Danziger 2013, Bailey and Duquette 2014).

In some specifications, we also include state-by-birth cohort fixed effects, γ_s,b, to account for time-varying characteristics of each state’s labor force. For instance, these fixed effects would account for the differential evolution of school quality (Card and Krueger 1992b) and racial discrimination (Donohue and Heckman 1991, Wright 2013) across birth cohorts within states. Finally, we include “gross state product” to account for potentially different exogenous rates of economic growth across states unaccounted for by changes across birth cohorts.²² This final covariate intends to reduce omitted variables bias due to differential changes in the demand for workers in states differentially affected by the 1966 FLSA. However, due to concerns about endogeneity to the effects of the 1966 FLSA, which could affect economic growth directly, we omit this variable from our preferred specification. For computational reasons, we partial out covariates to adjust for potentially confounding changes in individual characteristics in some specifications using the Frisch-Waugh-Lovell theorem (Frisch and Waugh 1933, Lovell 1963).²³

The point estimates of interest, θ, capture the regression-adjusted, reduced-form co-movements of the outcome variable with the “bite” of the 1966 FLSA. Because the 1966 FLSA should only affect outcomes after the Amendments took effect, one test of the validity of the research design is whether θ_t = 0 jointly for all t < 1966 in equation (3). Of course, the 1961 FLSA may have differentially impacted wages in states with a higher share of wages below $1.60 in 1966, which may lead to a slight pre-trend. Similarly, because the 1966 FLSA should increase wages after 1966, we should only observe θ_t > 0 for t > 1966. Standard errors are corrected for an arbitrary within-state covariance structure (Arellano 1987).

In addition to presenting the estimates for the reduced form, we estimate the labor demand elasticity by estimating equation (4) using two-stage least squares as a two-equation system, with log wages as the outcome in the first stage and the employment rate (in levels) as the outcome in the second stage. We calculate the elasticity by dividing the resulting second-stage point estimate of β by the mean employment rate in 1966.

III. Results: The Effects of the 1966 FLSA on Wages and Employment

Documenting the aggregate impact of the 1966 FLSA is key to understanding its impact on the economy. Although the BLS conducted a number of surveys to address this question, these studies were specific to certain industries and are not representative of all U.S. workers. Our analysis, therefore, uses the March CPS to quantify, first, the wage effects of the 1966 Amendments to the FLSA for a nationally representative sample and, second, any resulting changes in employment.

A. Wages

Figure 4A plots the event-study reduced-form results for all men ages 16 to 64 using the baseline specification of equation (3) (all covariates except for gross state product). Dashed lines represent the 95-percent, point-wise confidence intervals. In addition, we report the comparable reduced-form, differences-in-differences estimate from equation (4), summarizing the effect averaged over all years after 1966 (also shown in column 3, Table 2). Consistent with these estimates reflecting the 1966 FLSA itself (rather than potentially confounding policy changes), hourly wages in lower and higher wage states followed similar trends before the 1966 FLSA and these increases appear exactly when the 1966 FLSA was implemented. Our baseline estimates imply that lower earning states like Texas (the lower quartile of the interquartile range of share of workers with wages below $1.60 in 1966) experienced a 6.0 percent larger increase (0.397 x 0.15) in average wages relative to states like New York (upper quartile), where wages were higher and the 1966 FLSA was less binding. The increase in wages persists through the end of our sample in 1973.

Figure 4. The Effects of 1966 FLSA Amendments on Log Hourly Wages.

A. Results Averaged over All Industries

B. Results Disaggregated by High and Low Coverage Expansion Industries

Notes: Panel A plots point estimates and 95-percent confidence intervals for equations 3 and 5 using the log hourly wage as the dependent variable. All regressions include indicators for state-by-birth cohort, year, age, nonwhite, marital status, and metropolitan residence status. Model 1 plots estimates of equation 3 and includes state and year fixed effects. Models 2 and 3 plots estimates of equation 5. Both models include state-by-industry and year fixed effects, and Model 3 additionally includes state-by-year fixed effects. Panel B plots estimates of equation 3, separately for industries with a large coverage expansion in the 1966 FLSA and other industries (see text for definition). Sample includes men ages 16-64 not residing in group quarters or in the military for whom self-employment income accounts for no more than 10 percent of total income. Standard errors are clustered at the state group level. Sources: 1960 Census and 1962-1974 March CPS.

Table 2.

The Reduced-Form Effects of 1966 FLSA Amendments on Wages and Employment

	(1)	(2)	(3)	(4)
A. Log hourly wage (Mean real wage 1966: $22.92)
Post-1966 x fraction affected	0.589 (0.064)	0.491 (0.062)	0.397 (0.058)	0.366 (0.066)
F statistic	85.654	61.821	47.427	30.661
Effect of moving across IQR (0.15)	0.088	0.074	0.060	0.055
Effect of one SD increase (0.09)	0.053	0.044	0.036	0.033
B. Employed during year (Mean in 1966: 0.917)
Post-1966 x fraction affected	0.061 (0.032)	0.041 (0.034)	−0.049 (0.027)	−0.060 (0.027)
Effect of moving across IQR	0.009	0.006	−0.007	−0.009
Effect of one SD increase	0.005	0.004	−0.004	−0.005
C. Employed in reference week (Mean in 1966: 0.819)
Post-1966 x fraction affected	0.108 (0.030)	0.074 (0.030)	−0.002 (0.035)	−0.025 (0.037)
Effect of moving across IQR	0.016	0.011	0.000	−0.004
Effect of one SD increase	0.010	0.007	0.000	−0.002
D. Annual hours worked (Mean in 1966: 1,631)
Post-1966 x fraction affected	294.459 (72.196)	148.391 (67.249)	−42.854 (63.525)	−166.064 (84.225)
Effect of moving across IQR	44.169	22.259	−6.428	−24.910
Effect of one SD increase	26.501	13.355	−3.857	−14.946
State and year fixed effects	x	x	x	x
Demographic covariates		x	x	x
State-by-cohort fixed effects			x	x
Log gross state product				x
State-year observations	294	294	294	294

Notes: Panel titles refer to the dependent variable used for equation 4. Estimates are the coefficient on the interaction between the share of workers with wages in each state below $1.60 in 1966 and an indicator variable for the year being 1967 to 1973 (inclusive). Sample includes men ages 16-64 not residing in group quarters or in the military. In panels A, B, and D, we exclude individuals for whom self-employment income accounts for no more than 10 percent of total income. In panel C, we exclude individuals who report being self-employed in the reference week. Standard errors are clustered at the state-group level. All dollars are adjusted to 2019 using the CPI-U. For the share of wages below $1.60 in 1966, the cross-state standard deviation is 0.090 and the interquartile range is 0.150. Panel C has 21 fewer observations because we focus only on outcomes through 1973. The number of observations is 1,878,830 in panel A; 2,407,230 in panels B and D; and 2,447,550 in panel C. Sources: 1960 Census, 1962-1974 March CPS, BEA Regional Economic Accounts.

One potential threat to the internal validity of our research design is that other state or federal changes after 1966—not accounted for in gross state product—could confound our estimates. Because there is a great deal of within-state, across-industry variation in the share of wages below $1.60 (Figure 4B), we test this hypothesis by refining our estimating equation to examine changes within a state-industry cell using the following event-study specification:

$${\mathrm{Y}}_{j,s,t}=\sum _{k=1959}^{1973}{\pi }_{k}1(t=k)F_{j,s,1966}+{\delta }_{j,s}+{\delta }_{s,t}+{\epsilon }_{j,s,t}$$ (5)

One-digit industries are indexed by j and other notation remains as previously described. The advantage of this specification is that it permits fixed effects by state-year, δ_s,t, as well as by industry-state, δ_j,s. State-year fixed effects flexibly control for any exogenous state-level changes in the demand for or supply of workers (which are not captured in gross state product in equations 3 and 4). The point estimates of interest, π, capture changes after 1966 in lower wage state-industry combinations (which would have been more affected by the 1966 FLSA) relative to higher wage state-industries.

Figure 4A plots the results as model 2 (M2), which changes the key independent variable to a state-by-industry variable and adds state-by-industry fixed effects), and model 3, which adds state-by-year fixed effects to M2. The similarity of these estimates to those from our baseline specification (M1) and to one another (M2 vs. M3) implies that state-year changes in worker demand or supply are not driving (or offsetting) the legislation’s effects—a finding that should ameliorate concerns about the interpretation of employment analyses where we cannot use industry variation and include state-year fixed effects.²⁴

Table 2 presents additional robustness checks as reduced-form differences-in-differences estimates. Similar to the robustness in Figure 4A, Table 2A shows how the combined post-1966 effects are impacted by the inclusion of individual covariates (age, race, marital status, and metro area, column 2), state-by-birth cohort fixed effects to account for unobserved changes within states across birth cohorts (such as improvements in school quality or the cohort-evolving anti-discrimination efforts in the South, column 3), and time-varying, state-level controls for the natural log of gross state product (column 4). The inclusion of state-by-birth cohort effects only modestly reduces the estimated wage effects across the interquartile range by 0.014 when moving from column 2 to 3 and 0.005 when moving from column 3 to 4. Noteworthy is that the estimates change little across specifications and our baseline specification (column 3) is not statistically different when controlling for gross state product in column 4 (p-value = 0.38). Our Online Appendix presents the event-study estimates for these specifications for the interested reader.

These wage increases likely reflect both the 1966 FLSA’s increase in the real minimum wage for previously covered workers and its coverage expansion for previously uncovered workers. To separate these effects, we estimate equation (3) separately for “high coverage expansion industries”—which Martin (1967) indicates to be agriculture; forestry and fisheries; construction; retail trade (eating and drinking establishments and other retail establishments); services (personal, entertainment and recreation, medical, hospitals, educational); and government (postal service, federal, state, and local)—and other industries. The resulting estimates quantify the wage effects of the 1966 FLSA in industries where coverage expanded the most and those where the effects are predominantly driven by increases in the minimum wage (not coverage).

As expected, Figure 4B shows that the wage increase in high expansion industries, which employed 40.6 percent of all workers in 1966, is substantially larger than in industries where most workers were previously covered under the 1966 FLSA: the difference-in-difference estimates are 0.48 (0.09) and 0.25 (0.06), respectively. These estimates are also statistically distinguishable at the 5-percent level. This makes sense, because wages in industries uncovered before the 1966 FLSA increased by much more than wages in industries that were covered. Interestingly, our differences-in-differences estimate of 25 log points for industries not experiencing a large coverage expansion (labeled “Other”) is a bit larger than the 23 percent increase in the real minimum wage (section I.A), which likely reflects cross-industry spill-overs as well as the difficulty in mapping aggregated industries to finer FLSA regulations about coverage. This sizable increase in industries where most workers were previously covered suggests an important role for the statutory minimum wage increase and, potentially, general equilibrium wage adjustments across industries. While the real minimum wage declined over time due to inflation (Figure 1A), the estimated effect on real log wages is persistent, which reflects the large increase in coverage under the 1966 FLSA (Figure 1B).

A final, partial-equilibrium exercise seeks to gauge the plausibility of these effect sizes. As a benchmark, our difference-in-difference estimate of 0.40, scaled by an estimated 16.2 percent of workers having wages below $1.60 in 1966 in the March CPS, suggests that, nationally, average wages rose by 6.5 percent due to the 1966 FLSA. This estimate is also consistent with the following decomposition of the wage effects among employees covered prior to the 1966 FLSA, p; employees newly covered under the 1966 FLSA, n; and employees uncovered by the 1966 FLSA, u,

$$\Delta logW={\theta }_p^{66}\Delta log{W}_p+{\theta }_n^{66}\Delta log{W}_n+{\theta }_u^{66}\Delta log{W}_u.$$ (6)

The weights, θ, represent the share of U.S. employees in each of these groups in 1966 prior to the legislation (which implicitly assumes no disemployment effects due to the 1966 Amendments). According to the Department of Labor, 44 percent of workers were covered by the FLSA before 1966, and 12 percent of these workers would have been directly affected by the minimum wage increase because their wages were between $1.25 and $1.60. Assuming these workers received an average raise of 25 percent (in real terms, see Figure 4B “Other industries”) implies a 1.3-percent average wage increase in the economy. More difficult to quantify is the effect among newly covered workers (roughly 13 percent of all U.S. workers in 1966), whose nominal wages grew in some cases from less than $1 in 1967 to $1.60 in 1971. If half of this group experienced a 48-percent real wage gain (see Figure 4B, “Large coverage expansion industries”), then average wages would have increased by another 3.1 percent.²⁵ Finally, 43 percent of workers remained uncovered after the 1966 Amendments, and we expect their wages to rise in equilibrium assuming no disemployment effects and no spill-overs above the minimum wage. This group would need to have experienced another 2 percent increase in wages (e.g., 20 percent of these workers experienced a 24-percent real wage gain, half of that experienced by covered workers). In short, while there is considerable uncertainty about some of the inputs into this back-of-the-envelope calculation, the magnitudes of our estimates are plausible.

B. Employment and Annual Hours Worked

These robust increases in wages to a high (real) level would lead to disemployment, both in perfectly competitive and monopsonistic models of the labor market. To investigate this, Figures 5A and 5B present the reduced-form estimates for our baseline specification of equation 3 for several different measures of employment: a measure of employment at any point during the year (positive weeks worked), employment in the reference week, and annual hours worked (including individuals working no hours).²⁶ The first two measures capture different employment adjustments to the 1966 FLSA. The former captures longer-term, persistent employment responses by measuring disemployment only if the individual is not employed at any point during the year. The latter captures employment responses only during the March reference week and so is more sensitive to short-term, transitory fluctuations in employment. Annual hours worked describe changes in the combination of the extensive and intensive margins of work.

Figure 5. The Effects of 1966 FLSA Amendments on Employment.

A. Employed during Year

B. Employed in Reference Week

C. Annual Hours Worked (Including Zeros)

Notes: Figure plots point estimates and 95-percent confidence intervals for equation 3 using the indicated dependent variable indicated in the panel title. Regressions include state and year fixed effects, and indicators for state-by-birth cohort, age, nonwhite, marital status, and metropolitan residence status (corresponding to column 3 of Table 2). Sample includes men ages 16-64 not residing in group quarters or in the military who are not self-employed (see notes to Table 2 for details). Standard errors are clustered at the state-group level. Sources: 1960 Census and 1962-1974 March CPS.

Figure 5A presents our baseline event-study specification as well as specifications that control for log gross state product. The rationale for including this additional time-varying covariate is that we cannot use the state-industry variation in the share of potentially affected workers or, by extension, include state-by-year fixed effects to account for differential, exogenous changes in the demand or supply of workers. Tables 2B to 2D additionally summarize these results using the differences-in-differences specification (equation 4) for the specifications previously discussed.

This analysis shows that the 1966 FLSA’s wage and coverage increases had modest disemployment effects which, interestingly, appear mainly for longer-term employment. The March CPS shows that the share of men employed during the year fell by 0.7 percent in areas like Texas relative to New York, when these areas experienced larger wage changes after the 1966 FLSA (Table 2B and 2C, column 3). In contrast, men’s employment during the reference week fell by only 0.03 percent—or not at all given the event-study estimates in Figure 5B. A natural explanation that reconciles these two findings is that many of the men who no longer work during the year after the 1966 FLSA were less attached to the labor market and, therefore, less likely to be working in the March reference week even in the absence of the 1966 FLSA. Consequently, employment in the reference week shows less of a decline after the 1966 FLSA’s implementation. These findings may also be related to the fact that agricultural employment—where workers are more seasonal—was particularly impacted by the 1966 FLSA’s coverage expansion. Annual hours worked were 6 hours lower per year over a pre-1966 FLSA average of 1,631, implying a 0.4 percent decrease. The weaker effects on annual hours worked (which includes zeros) suggests that some of the extensive margin disemployment is offset by increased hours among men remaining employed.

Table 2 shows that the employment effects are sensitive to the inclusion of state-by-cohort effects. The motivation for these covariates is that school quality—measured by teacher-pupil ratios, term length, and teacher wages—was improving differentially in Southern states during our period of interest (Card and Krueger 1992b). Our Online Appendix maps Card and Krueger’s data into high and low Fraction Affected states and shows that school quality improved faster for cohorts in the former category. In addition, for both white and black men the average years of schooling in more affected states converged to that of less affected states. These trends suggest that—even in the absence of the 1966 FLSA—productivity and employment would have risen differentially in more affected states, at the same time that the black-white earnings gap narrowed (Card and Krueger (1992a). Higher productivity growth in more affected states tends to reduce disemployment effects when omitting these controls.

Consistent with changes in unobserved cohort attributes improving employment differentially in states where the minimum wage’s impact was larger, including the state-by-cohort effects reverses the sign of the relationship of the 1966 FLSA to employment (compare columns 2 and 3). Similarly, controlling for gross state product—which controls for differential productivity growth in the South in the absence of the legislation—leads to slightly more negative employment effects in column 4 of Table 2. Just as Figure 4A suggests that including state-year effects tends to dampen the estimated wage effects, controlling for gross state product tends to make the employment estimates more negative. The estimates for employment during the year are not different from one another (p=0.57). However, estimates for employment in the reference week are marginally statistically different from one another (p=0.107) and annual hours worked are statistically different from one another (p=0.071).

The sensitivity of our results to controlling for gross state product is consistent with two explanations. First, states where the 1966 FLSA had a larger impact tended to experience higher economic growth for exogenous reasons, so accounting for differential employment trends tends to make the observed effects of the 1966 FLSA more negative. In addition, states where the 1966 FLSA had a larger impact tended to experience faster employment growth due to the 1966 FLSA, so controlling for one result of the legislation’s effect tends to make the observed effects smaller. The data do not distinguish between these explanations, so both should be considered in interpreting the results.

C. Aggregate Demand Elasticities

Table 3 presents the demand elasticities implied by these estimates. For employment during the year, the demand elasticity with respect to the wage is −0.135 (column 3, one-sided test rejects zero) to −0.177 (column 4, also statistically significant). The increase in the estimates between columns 3 and 4 implies that accounting for faster employment growth in areas more affected by the 1966 FLSA generates larger demand elasticities. For employment in the reference week, the demand elasticities are −0.009 to −0.099 in the specification without and with controls for gross state product, respectively, and statistically indistinguishable from zero in both cases. For annual hours worked, the respective elasticities are −0.066 to −0.277 (one-sided test for the latter rejects zero). Noteworthy is that the extensive margin elasticities are smaller in magnitude than the employment elasticities for teens presented in Brown (1999, elasticities ranging from −0.5 to −2.7) and at the lower end of the range found by Card (1992, elasticities ranging from −0.12 to 0.39).²⁷ Figure 6 places our demand elasticity estimates within the context of recent papers. Our estimated demand elasticities fall roughly in the middle of elasticities in this literature and the results for employment during the year are identical to Harasztosi and Lindner (2019)’s calculation of −0.18 elasticity (with respect to the workers’ wages in Hungary in 2001), where the real minimum wage increased by around 60 percent.

Table 3.

Elasticities of Employment and Annual Hours Worked with Respect to Wages

	(1)	(2)	(3)	(4)
A. Employed during year
Wage elasticity	0.112 (0.056)	0.091 (0.069)	−0.135 (0.077)	−0.177 (0.086)
B. Employed in reference week
Wage elasticity	0.263 (0.069)	0.217 (0.080)	−0.009 (0.128)	−0.099 (0.154)
C. Annual hours worked
Wage elasticity	0.307 (0.070)	0.185 (0.076)	−0.066 (0.098)	−0.277 (0.158)
State and year fixed effects	x	x	x	x
Demographic covariates		x	x	x
State-by-cohort fixed effects			x	x
Log gross state product				x
State-year observations	294	294	294	294

Notes: This table reports own-wage elasticities from 2SLS estimation of equation 4. Panel titles refer to the dependent variable used in the employment regressions. First-stage estimates and dependent variable means are shown in Table 2. See Table 2 notes for information on sample and sources.

Figure 6. Comparison of Own-Wage Employment Elasticities to Prior Research.

Notes: Figure displays point estimates and 95-percent confidence intervals of own-wage employment elasticities from other papers studying minimum wage increases. Expanding Figure A6 of Harasztosi and Lindner (2019), the figure contains estimates from Table 3, column 3, as well as estimates from Cengiz et al. (2019) and Derenoncourt and Montialoux (2019). The vertical line is at zero.

D. Heterogeneity in Effects by Race, Age, and Education

A final analysis examines heterogeneity in the effects of the 1966 FLSA by demographic and skill groups. Several reasons suggest the effects could vary across demographic groups. First, different workers have different skills, and less-skilled workers (often proxied by less education, younger age, or minority racial group) were more likely to be directly affected by the 1966 FLSA. Second, different workers (even with the same skills) may be more concentrated in different industries due to regional differences in industry concentration or historical or institutional reasons (e.g., both agriculture and the share of African American workers is higher in the South). Because the FLSA affected coverage according to industry of employment, younger and less skilled workers may be more affected. Third, the 1966 FLSA also could lead to different effects for workers with different supply elasticities, which is related to race or age-based discrimination. Although the CPS is not rich enough to separate the mechanisms for differential effects, this final section documents the differential incidence of the FLSA by different race, age, and education subgroups by estimating equations 3 and 4 separately for each subgroup.²⁸

Table 4 presents reduced-form, difference-in-difference estimates of wages for subgroups of workers defined by race, education, and age. Figures 7 and 8 present the complementary event-study specifications which—due to smaller sample sizes—group years to increase precision as follows: 1959-61, 1962-63, 1964-65, 1967-68, 1969-70, and 1971-73 (estimates plotted against the year-group mid-point on the x-axis). Table 4A and Figure 7 show that all groups experienced wage increases and, as expected, groups with lower average skill experienced larger increases in wages. In 1966, for instance, 38 percent of black men and 15 percent of white men earned below the 1966 FLSA minimum wage (see Online Appendix). And, while both groups experienced large wage increases after 1966, the estimate for African-American men was almost three times as large as that for white men, owing to the fact that black men lived in lower-earning regions (e.g., the South) and worked in lower earning industries previously uncovered by the FLSA. Similarly, the wages of men with less than a 12^th grade education (approximately the median in 1966) increased by 33 percent more than men with at least a 12^th grade education. Teenagers experienced a larger wage increase than men ages 20 to 35, who in turn experienced a larger wage increase than those age 36 to 64. However, large wage growth among teenagers comes with the caveat that the event-study estimates in Figure 7C show that their wages were trending upward in more affected states before the 1966 FLSA took effect, which limits the strength of conclusions about causal effects of the legislation. The broad conclusion, however, is that the 1966 Amendments to the FLSA substantially increased wages for a large group of workers across the country.

Table 4.

The Reduced-Form Effects of 1966 FLSA Amendments on Log Wages, Employment, and Annual Hours Worked, by Subgroup

		(1) White men	(2) African- American men	(3) Less than 12 years education	(4) At least 12+ years education	(5) Ages 16-19	(6) Ages 20-35	(7) Ages 36-64
A. DV: Log hourly wage	Means: a	23.60	16.60	19.80	25.20	13.40	21.90	24.70
Post-1966 x fraction affected		0.290 (0.062)	0.792 (0.105)	0.426 (0.074)	0.318 (0.075)	0.625 (0.217)	0.406 (0.112)	0.328 (0.048)
F-statistic		21.7	57.3	33.0	17.9	8.29	13.26	45.9
Effect of moving across IQR (0.15)		0.044	0.119	0.064	0.048	0.094	0.061	0.049
Effect of one SD increase (0.09)		0.026	0.071	0.038	0.029	0.056	0.037	0.030
B. DV: Employed during year	Means:	0.923	0.873	0.866	0.960	0.731	0.964	0.936
Post-1966 x fraction affected		−0.026 (0.027)	−0.200 (0.067)	−0.053 (0.037)	0.011 (0.016)	−0.075 (0.083)	−0.024 (0.035)	−0.057 (0.031)
Effect of moving across IQR		−0.004	−0.030	−0.008	0.002	−0.011	−0.004	−0.009
Effect of one SD increase		−0.002	−0.018	−0.005	0.001	−0.007	−0.002	−0.005
C. DV: Employed in reference week	Means:	0.827	0.760	0.750	0.881	0.405	0.882	0.890
Post-1966 x fraction affected		−0.004 (0.036)	−0.055 (0.063)	0.002 (0.041)	0.027 (0.040)	−0.061 (0.126)	0.058 (0.036)	−0.034 (0.036)
Effect of moving across IQR		−0.001	−0.008	0.000	0.004	−0.009	0.009	−0.005
Effect of one SD increase		0.000	−0.005	0.000	0.002	−0.005	0.005	−0.003
D. DV: Annual hours worked	Means:	1,666	1,356	1,392	1,837	335	1,796	1,871
Post-1966 x fraction affected		19.30 (73.5)	−452 (140.3)	39.9 (80.1)	−61.2 (62.5)	−174.6 (101.0)	−11.56 (87.9)	−102.8 (104.1)
Effect of moving across IQR		2.90	−67.90	5.99	−9.18	−26.19	−1.73	−15.42
Effect of one SD increase		1.73	−40.74	3.59	−5.51	−15.71	−1.04	−9.25
State-year-(age group)b cells		294	266b	273b	273b	882b	882b	882b

Notes: Panel titles refer to the dependent variable used in equation 4 (specification for column 3, Table 2). Column headings identify sub-samples.

^aThe mean is the 1966 average of hourly wages for wages greater than zero and less than the 95 percentile.

^bTo ensure state groups are balanced across all years for a subgroup, we drop CT and AK/HI/OR/WA in column 2. Columns 3-4 exclude 1963, because the 1963 CPS does not contain education. Columns 5-7 pool three age-groups for 294 state-year cells for a total of 882 cells, allowing for interactions between age group dummies and all covariates except for state-by-cohort fixed effects. Standard errors are clustered at the state-group level. The number of observations in panel A by column is 1,693,170; 168,354; 959,314; 910,701; 101,764; 721,594; and 1,055,467. In panels B and D, the number of observations is 2,140,780; 241,222; 1,330,950; 1,064,540; 286,634; 855,638; and 1,264,958. In panel C, the number of observations is 2,179,050; 242,890; 1,351,140; 1,083,870; 286,514; 867,745; and 1,293,288. See Table 2 notes for additional information on the sample and sources.

Figure 7. Effects of 1966 FLSA Amendments on Log Wages, by Subgroup.

A. White and Black Men, 16-64

B. Men with at Least (“High”) and Less than 12^th Grade (“Low”), Ages 16-64

C. Men by Age Group, Ages 16-64

Notes: Panel titles refer to subgroups examined. Figure plots estimates of the interaction between the share of wages in each state below $1.60 in 1966 and indicators for years per equation 3. Because subgroup samples are smaller, we group years 1959-61, 1962-63, 1964-65, 1967-68, 1969-70, and 1971-73 and plot estimates at the midpoint of each interval. Differences-in-differences regressions using the same covariates are in Table 4 and summarized in the legend. See also notes to Figure 5.

Figure 8. Effects of 1966 FLSA Amendments on Employment, by Subgroup.

A. White and Black Men, Ages 16-64

B. Men with at Least (“High”) and Less than 12^th Grade (“Low”), Ages 16-64

C. Men by Age Group, Ages 16-64

Notes: Panel titles refer to subgroups examined. Figure plots estimates of the interaction between the share of wages in each state below $1.60 in 1966 and indicators for years per equation 3. Because subgroup samples are smaller, we group years 1959-61, 1962-63, 1964-65, 1967-68, 1969-70, and 1971-73 and plot estimates at the midpoint of each interval. Differences-in-differences regressions using the same covariates are in Table 4. See also notes to Figure 5.

The remaining panels of Table 4 present the reduced-form effects of the 1966 FLSA on employment, and Figure 8 shows the corresponding event-study estimates.²⁹ Owing to smaller sample sizes, most employment outcomes are imprecisely estimated. Similar to the estimates in Table 2, the absolute magnitudes of the disemployment estimates are generally larger for employment during the year relative to employment in the reference week and annual hours worked.

One noticeable exception is for African-American men, who experienced a sharp and statistically significant decline in employment during the year and annual hours worked. Moving across the interquartile range implies that employment during the year was 3.4 percent lower (3 percentage points over a baseline of 87 percent) and their annual hours worked fell by 5 percent (68 hours over a baseline of 1,356) after the 1966 FLSA became effective. Notably, employment in the reference week decreased by less, at 1.1 percent (or 0.8 percentage points over a baseline of 76 percent). By comparison, changes in employment for white men when moving across the interquartile range were much smaller and statistically insignificant. In addition, the black-white differences in Table 4 for log wages, employment during the year, and annual hours worked are statistically significant (p-values of 0.001, 0.052, 0.012), but the difference in employment in the reference week is not (p-value of 0.44). The event-study estimates show the persistence of the employment responses for black men during the year as well as in annual hours worked, although these effects fade modestly after 1969 as inflation reduced the magnitude of the wage effects for previously covered workers. The persistence is likely driven by reductions in employment in newly covered areas of the economy with permanently higher wages.

The results for employment in the reference week provide a direct comparison to Derenoncourt and Montialoux (2019), who use employment in the reference week conditional on labor-force participation as their primary outcome. In addition, their preferred specification does not control for state-by-birth cohort or gross state product, which Table 2 demonstrates are important in accounting for differential skill and economic growth in more affected states. Similar to our findings, Derenoncourt and Montialoux (2019) report that the 1966 FLSA had no detectable effect for employment in the reference week.³⁰

Figure 9 summarizes the relevant demand elasticities across groups, and Table 5 reports them for each group. The demand elasticities for black men are −0.289 (0.089) for employment during the year, −0.107 (0.124) for employment in the reference week, and −0.421 (0.153) for annual hours worked. The elasticity for annual hours worked is especially large and stems from a reduction in weeks worked and the usual number of hours worked per week (see Online Appendix). Notably, the black-white difference in the own-wage elasticity is statistically different for annual hours worked (p-value of 0.074), but the elasticities for employment during the year and employment in the reference week are not (p-values of 0.27 and 0.73). Employment elasticities for teens are also larger but imprecisely estimated owing to smaller sample sizes, at −0.138 (standard error: 0.128) for employment during the year, −0.029 (0.300) for employment in the reference week, and −0.606 (0.465) for annual hours worked. An important caveat to the findings for teenagers is that their wages began to rise and employment began to fall before the 1966 FLSA took effect. The largest own-wage employment elasticities for black men or young men are at the bottom of the range of the early time series literature but are also consistent with more recent work as shown in Figure 6.

Figure 9. Own-Wage Employment Elasticity in the late 1960s, by Subgroup.

A. Employed during Year

B. Employed in Reference Week

C. Annual Hours Worked

Notes: Panel titles refer to measure of employment used in equation 4. Figure displays point estimates and 95-percent confidence intervals from regressions in column 3 of Table 3 and columns 1-7 of Table 4.

Table 5.

Elasticities of Employment and Annual Hours Worked with Respect to Wages, by Subgroup

	(1) White men	(2) African- American men	(3) Less than 12 years education a	(4) At least 12+ years education a	(5) Ages 16-19	(6) Ages 20-35	(7) Ages 36-64
A. DV: Employed during year
Wage elasticity	−0.098 (0.100)	−0.289 (0.089)	−0.142 (0.101)	0.035 (0.052)	−0.138 (0.128)	−0.063 (0.080)	−0.187 (0.119)
B. DV: Employed in reference week
Wage elasticity	−0.023 (0.179)	−0.107 (0.124)	0.008 (0.153)	0.118 (0.161)	−0.029 (0.300)	0.207 (0.126)	−0.166 (0.175)
C. DV: Annual hours worked
Wage elasticity	0.040 (0.147)	−0.421 (0.153)	0.067 (0.129)	−0.104 (0.105)	−0.606 (0.465)	−0.021 (0.112)	−0.169 (0.176)
State-year-(age group)c cells	294	266c	273c	273c	882c	882c	882c

Notes: This table reports own-wage elasticities from 2SLS estimation of equation 4. Panel titles refer to the dependent variable used in the employment regressions. First-stage estimates and dependent variable means are shown in Table 4. See Table 2 and Table 4 notes for additional information on the samples and sources.

In summary, we find that the employment of African-American men and, perhaps, younger men fell with implementation of the 1966 FLSA. Broadly speaking, the magnitude of the demand elasticities and disemployment effects suggest that—although the aggregate effects are not large—the 1966 FLSA may have had adverse consequences for some workers.

IV. Conclusions

This paper examines the effects of the 1966 Amendments to FLSA, which raised the federal minimum wage to its highest level of the 20^th Century while significantly expanding the Act’s coverage. Using variation across states in the “bite” of the 1966 FLSA, we find that the amendments led to large increases in wages. Our estimates imply that wages rose 6 percent faster after 1966 in states like Texas, where many more workers had wages below the new federal minimum, than in states like New York, where fewer workers did. Extrapolating from our results suggests that, nationwide, wages increased by 6.5 percent on average because of the FLSA.

Notably, we estimate relatively small aggregate employment responses to this legislation. The average employment rates and annual hours worked decreased by 0.7 percent and 0.4 percent more in lower earning states, both statistically indistinguishable from zero. Our estimates also imply labor demand elasticities of −0.14 for employment (measured as positive weeks worked during the year) and −0.07 for annual hours worked with respect to wages. These elasticities are smaller than those implied by the early time series literature covering the U.S. in the same period as our analysis (Brown 1999) but are similar to more recent estimates of very large wage increases in other countries (Harasztosi and Lindner 2019). These aggregate effects mask the incidence of the 1966 FLSA on different subgroups. For instance, substantial decreases in employment and annual hours for African-American men suggest that large changes in the minimum wage could shift the composition of employment and harm certain groups of workers. Interestingly, the disemployment effects are concentrated on measures of employment during the year and, likely, workers with less attachment to the labor force as we consistently find little effect on employment in the March CPS reference week.

This evaluation of the 1966 FLSA offers a unique opportunity to evaluate the economic effects of a large increase in the real minimum wage to a high level for a sizable share of the economy that persisted in the short-term in the face of inflation due to the large increase in coverage. Although we find disemployment effects for some groups in the economy, the magnitude of these effects appears fairly modest in magnitude. Also noteworthy is the persistence of wage effects over time, alongside relatively stable impacts on employment. Although putty-clay models in Sorkin (2015) and Aaronson et al. (2018) imply that disemployment effects would increase over time as capital-intensive firms enter, our estimates show little evidence of this. During the 1960s, the increase in the minimum wage and its coverage may have instead led to capital deepening that enhanced the output of less skilled workers as well as a reallocation of employment to more productive or less discriminatory establishments (Dustmann et al. 2020). The available data do not allow us to study the mechanisms underlying these wage and employment dynamics, which is left for future research.