Printer Friendly

PREVAILING WAGE LAWS AND CONSTRUCTION LABOR MARKETS.

LAWRENCE F. KATZ [*]

Prevailing wage laws, which require that construction workers employed by private contractors on public projects be paid wages and benefits at least equal to those "prevailing" for similar work in or near the locality in which the project is located, have been the focus of an extensive policy debate. The authors, analyzing Current Population Survey data and Census data, find that the relative wages of construction workers decline slightly after the repeal of a state prevailing wage law. However, the small overall impact of law repeal masks substantial differences in outcomes for different groups of construction employees. Repeal is associated with a sizable reduction in the union wage premium and an appreciable narrowing of the black/non-black wage differential for construction workers.

Prevailing wage laws, which require that construction workers employed by private contractors on public projects be paid at least the wages and benefits that are "prevailing" for similar work in or near the locality in which the project is located, have been the focus of an extensive policy debate. Early empirical research on this topic sought to estimate the direct financial costs of prevailing wage laws to governments. To the extent that the prevailing wage is above the market wage, the laws may impose financial costs both through increased wage bills for construction projects (for example, Gujarati 1967; U.s. Congress, GAO 1979; Bourdon and Levitt 1980; Goldfarb and Morrall 1981; Fraundorf et al. 1982, Thieblot 1996) and through an inefficient mix of capital and labor and of different types of workers (for example, U.S. Congress, CBO 1982).

However, because public construction accounts for between one-fifth and one-quarter of all construction, and because prevailing wage laws cover a substantial number of private projects undertaken with public financing or assistance (for example, U.S. Congress, CBO 1982; U.S. Department of Commerce 1996), prevailing wage laws may also affect construction labor markets more broadly. These broad effects of the laws may have public policy implications beyond those implied by the direct financial costs of the laws to government. If governments are the marginal purchasers of construction services, then laws that require governments to pay a supra-market wage on all projects may aid unions by reducing the potential cost advantage of nonunion labor. Furthermore, if worker training or safety is under-supplied from a societal perspective by competitive construction markets, then prevailing wage laws may provide incentives for optimal training and safety (Philips et al. 1995). On the other hand, prevailing wage laws may aggravate discrimination against blacks in the construction industry by reducing nonunion-worker competition with members of historically discriminatory trade unions. Bernstein (1993, 1994) has argued that the federal prevailing wage law, the Davis-Bacon Act, was passed with discriminatory intent (but see Belman and Philips [1996] for an opposing view), and Keyes (1982) found that Davis-Bacon partly accounts for the low percentage of skilled black construction workers.

Despite this policy importance, there is little evidence on the effects of prevailing wage laws on construction labor markets. Identifying the impact of the Davis-Bacon Act on construction costs is difficult because its national scope and long history mean that there are no suitable construction labor markets to serve as a "control" group not subject to the Act. Thus, this paper investigates how the repeal of state prevailing wage laws--which specify that state governments must pay prevailing wages on state- and locally financed construction--affected labor markets for construction workers in the years 1970-93. We employ a modified difference-in-difference-in-difference approach to estimate the impact of the laws. Specifically, we estimate the effect of law repeal as the change over time in labor market outcomes for blue-collar construction versus non-construction workers from states that repealed their laws relative to the change over time in outcomes for workers from states that did not. We also assess the extent to which law repeal has differential effects across groups of construction workers: union versus nonunion, black versus non-black.

Background on Prevailing Wage Laws

Prevailing wage laws exist at the federal, state, and local levels. The federal prevailing wage law, the Davis-Bacon Act, was passed in 1931. [1] The current Act, modified by amendments in 1935 and 1964, requires private contractors to pay workers the prevailing wage/benefit package on all contracts of more than $2,000 for construction, alteration, or repair of federal public buildings or public works. The "prevailing wage" referenced in the Act is defined by the Secretary of Labor as the package of wages and benefits paid to the majority of workers in a given occupation grouping in the geographic area of the project; if the majority of workers do not earn the same wage/benefit package, then the prevailing wage is equal to the average wage/benefit package paid to these workers. [2] The "public projects" covered by the Act include all construction purchased directly by the federal government, plus most private but federally financed or assisted construction: the terms of the Act cover construction undertaken under the authority of more than 58 other laws (U.S. Congress, CBO 1982).

State prevailing wage laws set a minimum wage for construction workers on state (and usually municipal) works projects. Their terms differ across states in multiple dimensions, and are frequently defined customarily rather than by reference to written statutes or regulations (Thieblot 1986, 1995). Some state prevailing wage laws are almost non-binding; others set wages for virtually all contracts at the collectively bargained wage level. In addition, different states' laws treat jointly financed projects (for example, state/federal, local/federal, private/public) differently. Some states defer to the federal Act; others preempt the federal Act; others set the state prevailing wage at the higher of the state or federal prevailing wage. The scope of projects and workers covered under state laws also varies. States explicitly include or exclude specific types of projects (such as road construction), or workers, or projects above or below a given value.

Even if state prevailing wage laws covered purely state and local public construction only, they would have a potentially significant influence on construction labor markets. State and local public construction accounted for 16-22% of all construction over the 1970-1993 period, as compared to 2-3% for federal public construction (U.S. Department of Commerce 1996, Current Construction Reports C30, Tables 1-2). However, the federal prevailing wage law is likely to affect labor markets disproportionately, because state and local construction projects are often partially federally funded. (Indeed, even though federal public construction is only 2-3% of total construction, Congressional Budget Office estimates suggest that 20-25% of all construction is covered by the Davis-Bacon Act [U.S. Congress, CBO 1982].)

The early empirical literature was concerned with assessing the impact of the Davis-Bacon Act on the federal government's construction costs (for example, Gujarati 1967; U.S. Congress, GAO 1979; Bourdon and Levitt 1980; Goldfarb and Morrall 1981; Fraundorf et al. 1982). These studies estimated the impact of Davis-Bacon as the difference between the Department of Labor's posted Davis-Bacon prevailing wage and the average wage for construction workers of a given occupation in a given geographic area. Although the studies agreed that Davis-Bacon increased the government's labor costs for construction, they reported a wide range of point estimates (from 4% to 38%).

However, identifying the impact of the Davis-Bacon Act on construction costs is difficult, because its national scope and long history mean that there are no suitable construction labor markets to serve as a "control" group not subject to the Act. In particular, because the average wage in an area is a function of the Davis-Bacon prevailing wage, the difference between the Davis-Bacon wage and the average wage is not an unbiased estimate of the additional labor costs borne by the government, over what they would have paid in the absence of Davis-Bacon (but see Allen [1983] for a novel way to correct for this bias). Most seriously, however, this identification problem limits the degree to which the findings from the early literature can be generalized to construction labor markets more broadly, and thereby limits the extent to which the early literature can illuminate the policy questions of interest.

In recognition of these gaps, and mindful of the large magnitude of state and local public construction as a share of all construction, more recent work has investigated the impact of state prevailing wage laws, by comparing labor market outcomes across states with differing laws. Building on prior research (Thieblot 1986, 1996; Philips et al. 1995), in Table 1 we present the effective dates for the adoption and repeal of state prevailing wage laws. As of 1969, forty states had prevailing wage laws that covered construction financed by state and local governments. Between 1969 and 1993, nine states repealed their prevailing wage laws, and Minnesota enacted a prevailing wage law. As discussed above, because of the multidimensional variation across states in the prevailing wage statutes and regulatory policies, there is no way a priori to categorize the laws more finely.

Conclusions from the state prevailing wage law studies are not as uniform as results from analyses of the effects of Davis-Bacon. Allen and Reich (1980) reported that state prevailing wage laws have no significant effect on school construction costs, holding constant price levels, urbanization, and climate variables, although Allen (1987) suggested that prevailing wage laws enable union contractors to receive a higher price than nonunion contractors for similar school and hospital projects. Philips et al. (1995) found that the average construction wage declined more in states that repealed their prevailing wage laws than in states that did not, based on 4-digit SIC average wages. They also found that repeal of state prevailing wage laws was associated with an increase in injury rates for plumbers and pipe fitters, although Thieblot (1996) showed that average construction injury rates in repeal states declined by more than average rates in non-repeal states.

However, concerns about unobserved differences across states and over time complicate interpretation of the estimated effect of interest from all of the state law studies. The studies fail to control for fixed differences across states in laws, regulatory policies, and other characteristics; for state/time-varying macroeconomic factors affecting all blue-collar labor markets; and for state/time-varying microeconomic factors such as the occupational and skill mix of construction workers. For example, if states that repeal their prevailing wage laws also have unobserved fixed differences in policies that lead to lower construction wages, then the estimated effect of the laws might represent, in part, unobserved fixed differences across states, leading to an overstatement of the impact of prevailing wage law repeal. Along these lines, if states repeal prevailing wage laws in response to weak blue-collar labor markets (to improve competitiveness), then the estimated absolute effect of the laws on construction wo rkers might represent, in part, the effect of unmeasured labor market conditions for all blue-collar workers. Either of these biases would be magnified if contractors respond to repeals by employing less-skilled workers, because the estimated effects might partly represent unobserved changes in work force composition. Further, the existing work does not address other policy questions of interest, such as the differential impact of prevailing wage laws on black and union workers.

Models and Data

We investigate the impact of state prevailing wage law repeal with methods that directly address these concerns. Our basic specification compares time trends in blue-collar construction and non-construction labor market outcomes across repeal and non-repeal states during a 24-year period. We model wages and unionization rates as nonparametric functions of worker characteristics, state-fixed-effects, time-fixed-effects, and prevailing wage laws; thus, we control for fixed differences across states and over time. We use individual data on workers from the census and Current Population Survey (CPS), which enable us to control for changes in work force composition. Finally, we estimate the impact of prevailing wage law repeal as the difference between the change over time in the relative blue-collar construction/non-construction wage in repeal states and non-repeal states, to control for other, unobserved time-varying factors that affect all blue-collar labor markets and may be correlated with the status of labo r market regulation.

While this is fundamentally a difference-in-difference-in-difference (DDD) approach to estimating the impact of the laws, we modify conventional DDD strategies in several ways. First, our models include few restrictive parametric or distributional assumptions about the relationship between laws, individual characteristics, and labor market outcomes. Second, we do not only model the impact of law repeal as having a one-time effect on the levels of outcomes. For example, to the extent that multi-year contracts influence construction wages, changes in prevailing wage laws may not have instantaneous effects on labor markets. In addition, to the extent that prevailing wage laws affect the bargaining positions of workers and employers, the laws may affect future wage growth as well as current wage levels. Third, in order to investigate the impact of repeal on black workers and union workers, we use a difference-in-difference-in-difference-in-difference (DDDD) approach to estimate the relative growth in the blue-co llar black/non-black (union/nonunion) construction/non-construction wage differential in repeal versus non-repeal states.

To illustrate DDD estimation of the impact of law repeal, in Table 2 we present descriptive statistics from both individual and establishment-level data sets on wage trends during the 1980s for construction and non-construction workers, for states repealing and not repealing prevailing wage laws. [3] In addition to census and CPS analyses, Table 2 reports, for comparison purposes, descriptive statistics from the Department of Labor's ES-202 data gathered by establishment on employees covered by various unemployment insurance programs. The "before law repeal" columns (1) and (4) present average wages for 1979 (1979 CPS and ES-202 1980 census); the "after law repeal" columns (2) and (5) present average wages for 1993 (CPS and ES-202) and 1989 (1990 census). The "time difference for location" columns (3) and (6) present the proportional change in wages before versus after law repeal in a given type of state.

The top panel of Table 2 compares the change in average wages for construction workers in the states that repealed their laws ("experimental" states) to the change in average wages for construction workers in the states that did not ("non-experimental" states). In the census, for example, there was a 17.5% fall in construction workers' wages over the 1980s in states that repealed their laws, compared to a 12.9% fall in wages in states that did not. Thus, there was a 4.7% relative fall in construction workers' wages in states that repealed their prevailing wage laws; this is the differences-in-differences (DD) estimate of the impact of law repeal.

However, if there were unobserved time-varying factors that affected all blue-collar labor markets and were correlated with the status of labor market regulation, this DD estimate would not identify the impact of the law: it would reflect a combination of the law's impact and the impact of the unobserved factors affecting all blue-collar labor markets. The bottom panel of Table 2 investigates this possibility by examining the change in average wages for non-construction workers in experimental relative to non-experimental states. In fact, in the census, there is a slight fall in relative wages in experimental states of 2.2%, which suggests that controlling for factors correlated with law repeal that influence all blue-collar workers may be important.

Taking the difference between the first and seventh rows of column (7) of Table 2 shows that there is a 2.5% fall in the relative wages of construction workers in states that Repealed their prevailing wage laws, compared to the change in relative wages in the non-experimental states. This DDD estimate suggests that prevailing wage law repeal has a causal impact on construction labor markets. This conclusion is supported by replication of this result with other data sets, collected for different types of workers. Simple DDD estimates from the CPS and ES-202 are similar, at-1.3% for blue-collar workers in the CPS (-0.020 + 0.007), -1.9% for all workers in the CPS (-0.016 - 0.003), and -1.8% for all workers in the ES-202 (-0.041 + 0.023).

Our formal models analyze repeated cross-sections of blue-collar workers from the 1970, 1980, and 1990 census and the 1977-93 CPS. In state s = 1, ..., S during year t = 1, ..., T, one observation in our model is an individual worker i = 1, ..., [N.sub.st]. Each worker has occupational and demographic characteristics [X.sub.ist], which we describe as a set of binary variables. We describe worker i's industry of employment with [C.sub.ist], where [C.sub.ist] = 1 if worker i was employed in construction in state s and year t and 0 otherwise. Depending on the year of the survey and the worker's state of residence, the worker's labor market outcome may be affected by a prevailing wage law. We define [L.sub.s] = 1 if state s repealed its prevailing wage law over the 1970-90 period ([L.sub.s] = 0 otherwise), and [A.sub.st] = 1 if the repeal occurred before year t ([A.sub.st] = 0 otherwise). [4] We study two labor market outcomes: real hourly wage [W.sub.ist] and union status [U.sub.ist], where [U.sub.ist] = 1 if w orker i was a union member and 0 otherwise.

In regression terms, the basic DDD models of wages and unionization rates are of the form

(1) ln ([W.sub.ist]) = [[theta].sub.t] + [[[theta].sup.c].sub.t] * [C.sub.ist] + [[alpha].sub.s] + [[[alpha].sup.c].sub.s] * [C.sub.ist] + [L.sub.s] * [A.sub.st] [[gamma].sub.1] + [L.sub.s] * [A.sub.st] * [C.sub.ist][[gamma].sub.2] + [X.sub.ist][beta] + [[epsilon].sub.ist],

where [[theta].sub.t] is a time-fixed-effect, [[alpha].sub.s] is a state-fixed-effect, [[epsilon].sub.ist] is an error term, [5] and [[gamma].sub.2] is the effect of interest--the change over time in the construction wage premium in states that repealed their prevailing wage laws relative to the change over time in states that did not.

We use the CPS to investigate the dynamic effects of prevailing wage law repeal by estimating separately the change over time soon after repeal and long after repeal in the construction wage premium in states that repealed their prevailing wage laws relative to the change over time in states that did not, where

(1a) ln ([W.sub.ist]) = [[theta].sub.t] + [[[theta].sup.c].sub.t] * [C.sub.ist] + [[alpha].sub.s] + [[[alpha].sup.c].sub.s] * [C.sub.ist] + [L.sub.s] * [SA.sub.st][[gamma].sub.1s] + [L.sub.s] * [SA.sub.st] * [C.sub.ist][[gamma].sub.2s] + [L.sub.s] * [LA.sub.st][[gamma].sub.1l] + [L.sub.s] * [LA.sub.st] * [C.sub.ist][[gamma].sub.2l] + [X.sub.ist][beta] + [[epsilon].sub.ist].

In one time-since-adoption specification, we defined [SA.sub.st] = 1 if the repeal occurred 1-2 year before year t ([SA.sub.st] = 0 otherwise), and [La.sub.st] = 1 if the repeal occurred 3 or more years before year t ([LA.sub.st] = 0 otherwise). In an alternative time-since-adoption specification, we defined [SA.sub.st] = 1 if the repeal occurred 1-4 years before year t ([SA.sub.st] = 0 otherwise), and [LA.sub.st] = 1 if the repeal occurred 5 or more years before year t ([LA.sub.st] = 0 otherwise).

Finally, we estimate DDDD models to assess the impact of prevailing wage law repeal on the change time in differences by race and union status in the construction wage premium ([[gamma].sub.4]):

(2) ln ([W.sub.ist]) = [[theta].sub.t] + [[[theta].sup.b].sub.t] * [B.sub.ist] + [[[theta].sup.c].sub.t] * [C.sub.ist] + [[[theta].sup.b*c].sub.t] * [B.sub.ist] * [C.sub.ist] + [[alpha].sub.s] + [[[alpha].sup.b].sub.s] * [B.sub.ist] + [[[alpha].sup.c].sub.s] * [C.sub.ist] + [[[alpha].sup.b*c].sub.s] * [B.sub.ist] * [C.sub.ist] + [L.sub.s] * [A.sub.st][[gamma].sub.1] + [L.sub.s] * [A.sub.st] * [B.sub.ist][[gamma].sub.2] + [L.sub.s] * [A.sub.st] * [C.sub.ist][[gamma].sub.3] + [L.sub.s] * [A.sub.st] * [B.sub.ist] * [C.sub.ist][[gamma].sub.4] + [X.sub.ist][beta] + [[epsilon].sub.ist],

where [B.sub.ist] = 1 if worker i was black and 0 otherwise, [B.sub.ist] [epsilon] [X.sub.ist]. We estimate the effect of law repeal on trends in differences in the construction wage premium soon after versus long after repeal with versions of equation (2) analogous to equation (1a). Analogues to equation (2) that substitute [U.sub.ist] for [B.sub.ist] assess the impact of law repeal on the change over time in differences by union status in the construction wage premium.

We estimate the parameters of models (1) and (2) with nonfarm private-sector blue-collar wage and salary workers aged 16-64 using the Current Population Survey (CPS) and the census. Our principal CPS results are based on the currently employed workers in the Merged Outgoing Rotation Groups (MORGs) for 1979-93 plus such workers in the May CPS survey groups from 1977-78; we appended the earlier May CPS samples to the CPS MORGs to expand the time period covered by our analysis in the years before law changes occurred. [6] For the same reason, our CPS results estimating the impact of repeal on the relative union wage premium for construction workers (equation 2) are based on the 1983-93 MORGs matched with May CPS survey groups from 1977-81 (because the CPS did not begin asking all outgoing rotation groups about their union status until 1983, and the CPS did not ask any workers about their union status in 1982). [7] The vector of control variables X in the CPS includes indicators for age (16-20, 21-25, 26-30, 31- 35, 36-40, 41-45, 46-50, 51-55, 56-60, 61-64), educational attainment (less than high school, high school, some college, college graduate or more), marital status (married or unmarried), gender, black or non-black race, and one-digit occupational classification (craftsmen and kindred workers, operatives except transport, transport equipment operatives, nonfarm laborers, service workers), plus an interaction effect between each element of X and a variable indicating whether the observation was from the 1986 CPS or later.

Our census results are based on workers employed at any time during the previous year from the 1970, 1980, and 1990 public use census files. Thus, the wage information in the three censuses pertains to the years 1969, 1979, and 1989. To calculate the hourly wage for workers in 1970 with missing hours, we imputed average hours per week based on reported hours per week for workers in 1980 with the same age, sex, race, ethnicity, marital status, educational attainment, industry, and occupation. For all years, we analyze a 1% sample of blue-collar non-construction workers. For 1980 and 1990, we analyze a 5% sample of bluecollar construction workers; for 1970, a 5% sample with state information is not available, so we analyze a 2% sample of blue-collar construction workers (U.S. Department of Commerce 1972). Because Minnesota enacted a prevailing wage law in 1973, we omit workers from Minnesota from our analysis of the census data. The vector of control variables X in the census includes all of the variables from the CPS analysis plus indicators for hispanic/nonhispanic ethnicity and foreign/U.S. country of birth, plus interactions between year indicators and each element of X. Analyses of the census data weight each observation by the inverse of its sampling probability.

Results

Basic Difference-in-Difference-in-Difference Results

Table 3 presents estimates of [[gamma].sub.2], the effect of prevailing wage law repeal on construction wage differentials. In contrast to the simple DDD estimates from Table 2, our regression estimates of the impact of law repeal control for changes in workforce demographic and occupational composition across states and over time (except column 1), for time- and state-fixed-effects, and for construction*time- and construction*state-fixed-effects (except columns 2 and 4, which control for construction*time and construction*[L.sub.s]). In addition, our regression estimates allow returns-to-characteristics to differ in 1970, 1980, and 1990 (census) and to differ before and on/after 1986 (GPS). Columns (1)-(3) provide estimates based on the 1970, 1980, and 1990 census (column 1 reports DDD estimates on the effect of repeal on the change over time in wage levels from equation 1 without controls for demographic and occupational characteristics [X.sub.ist]). Columns (4)-(7) provide estimates based on the 1977-93 C PS: columns (4) and (5) report estimates from equation (1), and columns (6) and (7) report estimates of the time-since-adoption specification given by equation (1a).

The regression results reported in Table 3 echo the simple DDD estimates reported in Table 2: repeal of prevailing wage laws leads to slight decreases in construction wage differentials. In the census, the DDD impact of repeal on the change over time in wage levels ranges from 2% to 4%, depending on the controls included in the regression. [8] Magnitudes of average effect estimates of the impact of law repeal from the CPS are the same as or smaller than the magnitudes of similar estimates from the census. The estimated impact of repeal from equation (1) calculated controlling for construction*[L.sub.s] (column 4) is -3.9%. Controlling for a full set of construction*state-fixed-effects reduces this estimate substantially in both the basic and the time-since-adoption specification (columns 5- 7), suggesting that state heterogeneity in construction labor markets is an important determinant of wage schedules.

Differential Effects of Prevailing Wage Law Repeal by Race

Estimates of the average effect of law repeal fail to identify the differential effects of repeal across groups of construction workers, which also may be important from a policy perspective. We begin by analyzing the effects of law repeal on racial differences in construction employment. In the 1990 census, blacks comprised 11.7% of blue-collar non-construction workers and 7.4% of blue-collar construction workers, making for a black construction employment differential of-4.3 percentage points (0.074 - 0.117). In the census, we found that the black construction employment differential shrank by 1 percentage point in repeal versus non-repeal states over our sample period, controlling for demographic and occupational characteristics (allowing returns-to-characteristics to vary over time) and for state-fixed-effects, time-fixed-effects, time-fixed-effects*black, and [L.sub.s]*black. However, this result was no longer statistically significant when we included a full set of state*black interactions, and we did not find any statistically significant trends in blacks' construction employment differentials in the CPS.

For this reason, we focus on the relative wage effects of repeal. Table 4 reports DDDD estimates of [[gamma].sub.4], the impact of repeal on racial differences in the construction wage premium from equations (2) and (2a) using the census and the CPS. Except for column (1), the models in the table control for demographic and occupational characteristics (allowing returns-to-characteristics to vary over time), and for state-fixed-effects, time-fixed-effects, state-fixed-effects*construction, time-fixed-effects*construction, time-fixed-effects*black, state-fixed-effects*black, time-fixed-effects*black*construction, and state-fixed-effects*black*construction.

The results in Table 4 show that repeals affect black and non-black construction workers differently. Columns (1) and (2), reporting results of the analysis using census data, show that the construction wage premium earned by black workers rose by approximately 4 percentage points relative to that earned by non-blacks as a result of prevailing wage law repeal; this difference is statistically significant at the 5% level. The analysis using GPS data indicates that law repeal led to statistically significant increases (at the 10% level) in the long-run relative construction wage premium for blacks of 5.6 to 6.8 percentage points, depending on specification (columns 4-5). In both analyses, repeal has a negative effect on non-blacks' construction wage premium, although this effect is statistically significant only in the census analysis.

The differential impact of law repeal on blacks' construction wage premium has two important implications. First, because the raw construction wage premium is 6.5 percentage points larger for non-blacks than for blacks, [9] this finding implies that law repeal eliminates roughly 60% of this difference ([approximate] 4 / 6.5). Second, in no specification does law repeal significantly reduce the absolute level of black workers' construction wage differential, and in some specifications it increases that differential, even as it reduces the construction differential for non-black workers. In terms of equation (2), the impact of law repeal on the absolute level of black workers' construction wage differential is equal to [[gamma].sub.3] + [[gamma].sub.4]. In the census, [[gamma].sub.3] + [[gamma].sub.4] is small in magnitude, but in the CPS, the long-run estimate of [[gamma].sub.3] + [[gamma].sub.4] from column (5) is approximately equal to 4.8%, although not statistically significant at conventional levels.

Differential Impact of Prevailing Wage Law Repeal by Union Status

Based on the CPS, Table 5 presents estimates of the effect of law repeal on the construction unionization differential (equations 1 and la, columns 1-3) and the relative union wage premium for construction workers (equations 2 and 2a, columns 4-6). Regression results suggest that repeal in the long run reduces the relative construction unionization differential by 1.5 percentage points, although this result is not statistically significant (and is actually positive in sign in the short run). Furthermore, compared to the average level of construction industry unionization (approximately 25%, based on the CPS), the effect of repeal on the unionization rate is extremely small in magnitude. However, prevailing wage law repeal does significantly reduce the relative union wage premium for construction workers, controlling for worker demographic and occupational characteristics (allowing re turns-to-characteristics to vary over time), and for state-fixed-effects, time-fixed-effects, state-fixed-effects*construction , time-fixed-effects* construction, time-fixed-effects*union, state-fixed-effects*union, time-fixed-effects*union*construction, and state-fixed-effects*union*construction. The first row of column (4) suggests that prevailing wage law repeal reduces the relative union wage premium paid to construction workers by approximately 5.9 percentage points. Columns (5) and (6) show that this effect grows in magnitude over time: three years after law repeal, the relative union wage premium paid to construction workers has fallen by approximately 9.3 percentage points, and after five or more years it has fallen by 11.2 percentage points. [10] These effects are both statistically and economically significant. Estimates from the 1983-93 CPS MORGs indicate that the additional average union wage differential earned by construction workers over that earned by non-construction workers is approximately 20%. Thus, law repeal roughly halves the relative union wage differential for construction workers. [11] However, law repeal in the long run has no statistically significant negative correlation with non-construction union wage differentials. Indeed, states repealing their prevailing wage laws show statistically significant increases in non-construction unionization rates. Thus, there is no evidence that prevailing wage law repeal is correlated with other state-level policies that have adverse effects on unions.

Understanding Differential Effects by Race and Union Status

We conducted additional analyses to investigate the mechanism by which law repeal affects racial differences in construction wage differentials, and the construction wage premium generally. First, law repeal does not have its most important impact on racial differentials through its impact on the unionization rate. Following Ashenfelter (1972), we calculated the total contribution of unionization to the black/non-black wage differential as [u.sub.bj][[gamma].sub.j]-[u.sub.wj][[gamma].sub.j'], where [u.sub.bj] ([u.sub.wj]) is the share of black (non-black) workers who are union members in industry j and [[gamma].sub.j] ([[gamma].sub.j']) is the union/nonunion wage differential for blacks (non-blacks) in industry j. In other words, we estimate what the change in the racial wage differential would be if the unionization rate were zero, relative to the current racial differential. Since the estimated total contribution of unionization to racial differences in the construction wage premium is approximately 3 perce ntage points, [12] even a 6% decrease in the unionization rate (1.5 percentage points/25 points total unionization rate, where 1.5 percentage points is approximately the maximum impact of the law repeals) would reduce the black/non-black construction wage differential by at most 0.18 percentage points (0.06 * 3). Thus, law repeal, which has an estimated total effect an order of magnitude larger, must affect wage differentials primarily through means other than reduced unionization.

Second, repeal achieves its approximately 2-percent effect on the aggregate construction wage differential primarily by reducing the union wage premium paid to construction workers. If the average construction unionization rate is equal to u, then the average effect for all construction workers of law repeal in terms of the parameters from equation (2) (substituting [U.sub.ist] for [B.sub.ist]) is equal to (1 - u)*[[gamma].sub.3] + u*([[gamma].sub.3] + [[gamma].sub.4]). Since u [approximate] 0.25, [[gamma].sub.3] = 0.007 and [[gamma].sub.4] = -0.112 (Table 5, column 6), most of the effect on the aggregate wage differential comes through [[gamma].sub.4] rather than through [[gamma].sub.3].

Sensitivity Analyses

One important concern with any analysis of the effects of law changes is policy endogeneity--the correlation of law repeal with unobserved time-varying determinants of wages. We found no substantial evidence that prevailing wage law repeal was endogenous. First, according to Thieblot (1986, 1996), the ways in which repeal occurred are consistent with an assumption of exogenous timing. Some repeals (for example, Arizona) were imposed judicially (rather than legislatively), and therefore likely to be less dependent on contemporaneous economic conditions. Similarly, most legislative repeals (for example, in Alabama, Colorado, Idaho, Kansas, New Hampshire, and Utah) only occurred after previous failed attempts; repeal efforts often started years before a change in law took place. Second, as discussed above, we found no systematic correlations between repeal and wages or union wage differentials of nonconstruction blue-collar workers, suggesting that repeal was not correlated with other state-level policies that have adverse effects on unions or unionized workers.

Third, to investigate directly the possibility that time-varying state macroeconomic factors correlated with but not caused by repeal were responsible for differential trends in wage differentials, we parameterized the differential impact of state macroeconomic factors by re-estimating equations (1) and (la) with controls for the state unemployment rate and the state unemployment rate*construction and by reestimating equations (2) and (2a) with controls for unemployment, unemployment* black, unemployment*construction, and unemployment*black*construction (for those versions of equations 2 and 2a that model the impact of law repeal on the union/nonunion wage differential, we included controls for unemployment, unemployment*union, unemployment* construction, and unemployment*union*construction). In no case did inclusion of these macroeconomic control variables substantially alter the estimated effects of law repeal.

We further investigated the hypothesis that the estimated effects of law repeal were caused by underlying macroeconomic labor market conditions that differentially affected the construction industry and were correlated with state labor market rules by re-estimating the basic model with altered control groups. We estimated equations (1) and (la) based on the census dropping from the analysis alternately all individuals from states that always had a prevailing wage law in effect during the sample period, and all individuals from states that never had a prevailing wage law in effect during the sample period. Although the point estimates of the effect of law repeal were somewhat larger in those models that used only individuals from states that never had a prevailing wage law as the control group, the substantive conclusions did not change.

Conclusion

We find that state prevailing wage laws have small but significant average effects on construction labor markets. Repeal of prevailing wage laws leads to slight decreases in the relative wage levels of construction workers. However, the effects of repeal differ substantially across groups of construction workers. The negative wage effects are borne primarily by union workers and white workers. Although relative construction unionization rates do not decline significantly in response to repeal, the longrun union wage premium earned by construction workers decreases by approximately 10 percentage points, or almost half of the total union wage premium in construction. Since union members account for approximately 25% of all construction workers, the 10-percentage-point decrease in the union wage premium accounts for essentially all of the (approximately 2-4%) decline in construction workers' wages.

Second, despite the negative overall effects of repeal on construction workers, repeal of prevailing wage laws does not harm--may, in fact, actually benefit--black construction workers. Prevailing wage law repeal raises black workers' construction wage differential relative to non-black workers' differential. [13] Furthermore, in no specification does law repeal significantly reduce the absolute level of black workers' construction wage differential, and in some specifications it increases that level, even as it reduces the construction differential for non-black workers.

The policy implications of these findings, and their applicability to other states considering repeal of their prevailing wage laws, depend crucially on the mechanism causing the differential impact of repeal across groups. On one hand, the differential impact of repeal may reflect a decrease in discrimination due to a weakening of construction unions, or due to a change in unions' behavior arising out of a declining union wage premium. On the other hand, if repeal affects workers in heavy construction (for example, SIC 16, road and sewer construction) more than workers in light construction (for example, SIC 15 and 17, general and special trade contractors), [14] and heavy construction workers are more likely to be white and unionized, then the differential impact of repeal by race and union status may simply reflect the differential composition of workers in the two segments of the construction industry. In this case, law repeal might simply be transferring resources from workers to purchasers of heavy con struction projects. Furthermore, to the extent that the composition of the construction industry differs between repeal states and other states considering repeal, our findings may not even accurately forecast the average effects of repeal in states considering repeal.

Unfortunately, since neither the CPS nor the census includes detailed industry or union information on construction workers, we cannot distinguish definitively between these two hypotheses. Indeed, in our supplementary analyses, we found no clear support for either causal mechanism. We found no evidence that repeal affected racial wage differentials through the unionization rate. The counterfactual of reducing the unionization rate to zero in both construction and non-construction industries would only reduce the magnitude of the raw difference between blacks' and non-blacks' construction wage differential by half. Compared to the contribution of unionization to racial wage differentials in the 1960s, the minimal contribution of unionization in the 1950s and 1990s is striking. Because law repeal has a relatively small impact on the unionization rate, law repeal must affect race-based wage differentials primarily through other means.

On the other hand, we found no evidence that repeal had different effects on different segments of the construction industry. Differential trends in payroll per worker from the ES202 establishment data presented in Table 2 suggest that repeal affects workers in heavy construction and light construction equally: the simple DOD estimates of the effect of repeal in the heavy construction industry are only 0.1% greater than the estimates of the effect of repeal in the light construction industry (-1.9% compared to -1.8%). Of course, because the establishment data do not allow us to control for worker heterogeneity, this finding is not a definitive rejection of differential effects across 2-digit SIC construction industries. We leave further investigation of the mechanism by which law repeal affects wage schedules to future research. One possible avenue might be investigation of the extent to which construction union behavior or decision-making differs in repeal and non-repeal states.

(*.) Daniel P. Kessler is Associate Professor at the Stanford University Graduate School of Business, Research Fellow at the Hoover Institution, and Research Associate at the National Bureau of Economic Research (NBER). Lawrence F. Katz is Professor at the Harvard University Department of Economics and Research Associate at NBER. Kessler gratefully acknowledges funding from the Stanford University GSB and from the John M. Olin Foundation. The authors thank Daniel Kao, Julie Lee, and Ari Zweiman for excellent research assistance, and Steven Allen for helpful advice on union wage contracts.

A data appendix with additional results, and copies of the computer programs used to generate the results presented in the paper, are available from Daniel Kessler at the Graduate School of Business, Stanford University, Stanford, CA 94305.

(1.) 40 U.S.C. Sec. 276(a) (1982).

(2.) Before 1985, the prevailing wage was equal to that paid to more than 30% of the workers in a given industry/occupation/area group, or the average if less than 30% of the workers earned the same amount.

(3.) In this and all of our subsequent analysis, we deleted individuals reporting wages of less than $1.65 or greater than $50 per hour in 1982 constant dollars.

(4.) We assume in all of the models that law repeals take effect beginning in the calendar year after the repeal is adopted.

(5.) We estimate all models allowing for heteroskedasticity and for within state/time-group correlation in [[epsilon].sub.ist], which may be important in models of state policy effects (Moulton 1990).

(6.) Prior to 1989, workers in the CPS earning more than $999 per week were topcoded at $999. We recoded these workers to earning $1,400 per week.

(7.) If the unadjusted CPS sampling weight for individual i in state s during year t is defined as [[omega].sub.ist] (that is, the number of individuals represented by individual i), analyses of the May/MORG sample weight each observation by its adjusted sampling weight [[omega].sub.ist], where

[[omega].sub.ist] = [[omega].sub.ist]/[[sigma].sub.i,s][[omega].sub.[ist.sup.*]]

(8.) The impact of repeal on the change over time in the census in wage growth rates (for example, 1980-90 wage growth relative to 1970-80 wage growth in repealing versus non-repealing states) is small and statistically insignificant, suggesting a need for some caution in assigning a strong causal interpretation to our findings; state repeals may he correlated with preexisting factors that affect construction labor markets.

(9.) Calculations based on the 1990 census.

(10.) Interpreting the long-run rather than the short-run effects as the equilibrium impact of law repeal depends on two key assumptions: that the full effects of repeal on labor markets take more than two years to appear, and that no other changes in the labor market or legal environment correlated with but not caused by repeal occurred three to five years after repeal. On one hand, to the extent that wage levels, and especially wage differences across types of workers, take several years to re-equilibrate in response to changes in state policy, the long-run effects would accurately represent the equilibrium impact of law reform. On the other hand, to the extent that union contracts in the construction industry are shorter than two years in duration, the short-run effects would accurately represent the equilibrium impact of repeal, with fewer potential unobserved confounding factors.

(11.) Correlation between unobserved worker characteristics and the union membership decision may lead to selection bias in the estimated levels of union wage premiums (for example, Freeman and Medoff 1984). As Rose (1987) observed in connection with the deregulation of the trucking industry, the small (negative) estimated effect of law repeal on the unionization rate suggests that this selection is unlikely to bias our results in favor of finding a negative effect of repeal on the relative union wage premium paid to construction workers, as long as shrinkage (or non-expansion) in the unionized construction sector led to increasing (or non-decreasing) concentrations of experienced, high-quality workers. However, if repeal led more able workers to exit the unionized sector more quickly than less able workers, then selection bias may lead to incorrect inferences about the impact of repeal on union wage premiums.

(12.) The small size of the contribution of the unionization rate to racial wage differentials in construction in the 1980s and 1990s is particularly striking by comparison with either its parallel contribution in 1967 or its contribution to the total racial wage differential. According to Ashenfelter (1972), the estimated contribution of unionization to racial differences in construction wage premiums was 8.9 percentage points in 1967. According to calculations based on the 1990 census, the total racial wage differential in the construction industry was 17.6%.

(13.)This finding is analogous to the finding that deregulation in the trucking industry lowered black/non-black wage differentials in that sector (for example, Rose 1987; Peoples and Saunders 1993).

(14.) This would be true if, for example, (prevailing-wage-law-covered) state and local construction projects involve more heavy than light construction, and light construction workers are poor substitutes for heavy construction workers.

REFERENCES

Allen, Steven G. 1983. "Much Ado about Davis-Bacon: A Critical Review and New Evidence." Journal of Law and Economics, Vol. 26 (October), pp. 707-36.

_____. 1987. "Can Union Labor Ever Cost Less?" Quarterly Journal of Economics, Vol. 102 (May), pp. 347-73.

Allen, Steven G., and David Reich. 1980. Prevailing Wage Laws Are Not Inflationary. Washington, D.C.: Center to Protect Workers' Rights.

Ashenfelter, Orley. 1972. "Racial Discrimination and Trade Unionism." Journal of Political Economy, Vol. 80 (May-June), pp. 435-64.

Belman, Dale, and Peter Philips. 1996. "Historical Research on the Davis-Bacon Act." Mimeo.

Bernstein, David. 1993. "Roots of the 'Underclass': The Decline of Laissez-Faire Jurisprudence and the Rise of Racist Labor Legislation." American University Law Review, Vol. 43, pp. 85-138.

_____. 1994. "The Davis-Bacon Act: Vestige of Jim Crow." National Black Law Journal, Vol. 13 (Fall),pp. 276-97.

Bourdon, C. C., and R. E. Levitt. 1980. Union and Open-Shop Construction. Lexington, Mass.: Lexington Books.

Fraundorf, Martha N., et al. 1982. "The Effects of the Davis-Bacon Act on Construction Costs in Non-Metropolitan Areas of the United States." Reprinted in The Impact of the Davis-Bacon Threshold on Small Business Construction Contractors, Hearing before the U.S. Senate, Subcommittee on Government Procurement of the Committee on Small Business (February 2).

Freeman, Richard B., and James L. Medoff. 1984. What Do Unions Do New York: Basic Books.

Goldfarb, Robert S., and John F. Morrall III. 1981. "The Davis-Bacon Act: An Appraisal of Recent Studies." Industrial and Labor Relations Review, Vol. 34, No. 2 (January), pp. 191-206.

Gujarati, D. N. 1967. "The Economics of the Davis-Bacon Act." Journal of Business, Vol. 40, No. 3, pp. 303-16.

Keyes, William A. 1982. "The Minimum Wage and the Davis-Bacon Act: Employment Effects on Minorities and Youth." Journal of Labor Research, Vol. 3, No. 4, pp. 399-407.

Moulton, Brent. 1990. "An Illustration of a Pitfall in Estimating the Effects of Aggregate Variables on Micro Units." Review of Economics and Statistics, Vol. 72, No. 2 (May), pp. 334-38.

Peoples, James, and Lisa Saunders. 1993. "Trucking Deregulation and the Black/White Wage Gap." Industrial and Labor Relations Review, Vol. 47, No. 1 (October), pp. 23-34.

Philips, Peter, et al. 1995. "Losing Ground: Lessons from the Repeal of Nine 'Little Davis-Bacon' Acts." University of Utah Department of Economics Working Paper (February).

Rose, Nancy L. 1987. "Labor Rent Sharing and Regulation: Evidence from the Trucking Industry." Journal of Political Economy, Vol. 95, No. 6 (December), pp. 1146-78.

Thieblot, Armand J. 1986. Prevailing Wage Legislation: The Davis-Bacon Act, State "Little Davis-Bacon" Acts, The Walsh-Healey Act, and the Service Contract Act. Philadelphia: University of Pennsylvania Industrial Research Unit.

_____. 1996. "A New Evaluation of Impacts of Prevailing Wage Law Repeal." Journal of Labor Research, Vol. 17 (Spring), pp. 297-322.

U.S. Congress, Congressional Budget Office. 1983. Modifying the Davis-Bacon Act: Implications for the Labor Market and the Federal Budget. Washington, D.C.: GPO, July.

U.S. Department of Commerce, Department of the Census. 1972. Public Use Samples of Basic Records from the 1970 Census: Description and Technical Documentation. Washington, D.C.: Department of the Census.

_____. 1996. Value of New Construction Put in Place: May 1996. Washington, D.C.: Department of the Census.

U.S. General Accounting Office. 1979. The Davis-Bacon Act Should Be Repealed. Washington, D.C.
                      Chronology of State Prevailing
                          Wage Laws through 1993.
               Year Effective
State            Enactment    Repeal
Alabama             1941       1980
Alaska              1931
Arizona             1912       1979
Arkansas            1955
California          1931
Colorado            1933       1985
Connecticut         1935
Delaware            1962
D.C.                1931
Florida             1933       1979
Georgia
Hawaii              1955
Idaho               1911       1985
Illinois            1931
Indiana             1935
Iowa
Kansas              1891       1987
Kentucky            1940
Louisiana           1968       1988
Maine               1933
Maryland            1945
Massachusetts       1914
Michigan            1965
Minnesota           1973
Mississippi
Missouri            1957
Montana             1931
Nebraska            1923
Nevada              1937
New Hampshire       1941       1985
New Jersey          1913
New Mexico          1937
New York            1894
North Carolina
North Dakota
Ohio                1931
Oklahoma            1909
Oregon              1959
Pennsylvania        1961
Rhode Island        1935
South Carolina
South Dakota
Tennessee           1953
Texas               1933
Utah                1933       1981
Vermont
Virginia
Washington          1945
West Virginia       1933
Wisconsin           1931
Wyoming             1967
                   Wage Trends in Construction and Non-
               Construction Industries, for States Repealing
                  and Not Repealing Prevailing Wage Laws.
                                        Experimental States: States That
                                         Repealed Prevailing Wage Laws
                                                     Before
                                                      Law
                                                     Repeal
Variable                                              (1)
Treatment Group: Construction Workers
Census:
  BC Hourly Wage                                     9.654
CPS:
  Hourly Wage                                        8.932
  BC Hourly Wage                                     8.717
ES202:
Payroll/Employment
  All Construction                                   19266
  Light Construction
  (SIC 15 and 17)                                    18318
  Heavy Construction
  (SIC 16)                                           22642
Control Group: Non-Construction Workers
Census:
  BC Hourly Wage                                     7.875
CPS:
  Hourly Wage                                        7.439
  BC Hourly Wage                                     6.690
ES202:
  Payroll/Employment                                 15799
                                        After    Time
                                         Law   Diff. for
                                        Repeal Location
Variable                                 (2)      (3)
Treatment Group: Construction Workers
Census:
  BC Hourly Wage                        7.960   -0.175
CPS:
  Hourly Wage                           7.302   -0.182
  BC Hourly Wage                        6.828   -0.217
ES202:
Payroll/Employment
  All Construction                      15981   -0.171
  Light Construction
  (SIC 15 and 17)                       15408   -0.159
  Heavy Construction
  (SIC 16)                              18632   -0.177
Control Group: Non-Construction Workers
Census:
  BC Hourly Wage                        6.838   -0.132
CPS:
  Hourly Wage                           7.099   -0.046
  BC Hourly Wage                        5.588   -0.165
ES202:
  Payroll/Employment                    15432   -0.023
                                        Non-Experimental States: States That
                                        Did Not Repeal Prevailing Wage Laws
                                                      Before
                                                        Law
                                                      Repeal
Variable                                                (4)
Treatment Group: Construction Workers
Census:
  BC Hourly Wage                                      10.685
CPS:
  Hourly Wage                                         10.098
  BC Hourly Wage                                       9.955
ES202:
Payroll/Employment
  All Construction                                     21546
  Light Construction
  (SIC 15 and 17)                                      20657
  Heavy Construction
  (SIC 16)                                             25468
Control Group: Non-Construction Workers
Census:
  BC Hourly Wage                                       8.442
CPS:
  Hourly Wage                                          8.068
  BC Hourly Wage                                       7.229
ES202:
  Payroll/Employment                                   17297
                                        After    Time    Difference
                                         Law   Diff. for   (in %
                                        Repeal Location   Change)
Variable                                 (5)      (6)       (7)
Treatment Group: Construction Workers
Census:
  BC Hourly Wage                        9.311   -0.129     -0.047
CPS:
  Hourly Wage                           8.414   -0.167     -0.016
  BC Hourly Wage                        7.997   -0.197     -0.020
ES202:
Payroll/Employment
  All Construction                      18749   -0.130     -0.041
  Light Construction
  (SIC 15 and 17)                       18229   -0.118     -0.041
  Heavy Construction
  (SIC 16)                              22029   -0.135     -0.042
Control Group: Non-Construction Workers
Census:
  BC Hourly Wage                        7.516   -0.110     -0.022
CPS:
  Hourly Wage                           7.679   -0.048      0.003
  BC Hourly Wage                        6.087   -0.158     -0.007
ES202:
  Payroll/Employment                    17300    0.000     -0.023


Notes: ES-202 includes private nonagricultural firms; CPS and census include private-sector nonfarm wage and salary workers aged 16-64. For the census, 1980 is the "before" period; 1990 is the "after" period. For the CPS and ES202, 1979 is the "before" period; 1993 is the "after" period. "Time difference for location" represents the percentage change after versus before law repeal in a given type of state. Observations from Minnesota are omitted from the census analysis because Minnesota enacted a prevailing wage law in 1973. Observations from Arkansas, Delaware, New York, and Texas are omitted from ES202 analysis because of suppression of 2-digit SIC construction industry data from those states. In the census, non-construction workers are a 1% sample; 1970 construction workers are a 2% sample; 1990 construction workers are a 5% sample. CPS means are calculated using CPS sampling weights. All dollar amounts are in 1982 constant dollars.
                 Effects of Prevailing Wage Law Repeal on
                Blue-Collar Wages and the Construction Wage
                                 Premium.
Variable                          (1)       (2)       (3)       (4)
Repeal State*After Repeal       -0.023    -0.039    -0.034    -0.039
*Construction                   (0.013)   (0.025)   (0.013)   (0.016)
Repeal State*After Repeal        0.008    -0.001    -0.001    -0.012
                                (0.017)   (0.018)   (0.017)   (0.007)
Repeal State*Shortly after
Repeal*Construction
Repeal State*Long after
Repeal*Construction
Repeal State*Shortly after
Repeal
Repeal State*Long after
Repeal
Demographic Controls?             No        Yes       Yes       Yes
Full Set of State*Construction
Interactions?                     Yes       No        Yes       No
Data Set                        Census    Census    Census      CPS
N                              1,468,033 1,468,033 1,468,033 1,017,875
Variable                          (5)       (6)       (7)
Repeal State*After Repeal       -0.002
*Construction                   (0.013)
Repeal State*After Repeal       -0.015
                                (0.007)
Repeal State*Shortly after                -0.005    -0.004
Repeal*Construction                       (0.015)   (0.014)
Repeal State*Long after                   -0.005    -0.008
Repeal*Construction                       (0.014)   (0.015)
Repeal State*Shortly after                -0.020    -0.021
Repeal                                    (0.009)   (0.007)
Repeal State*Long after                   -0.014    -0.010
Repeal                                    (0.007)   (0.008)
Demographic Controls?             Yes       Yes       Yes
Full Set of State*Construction
Interactions?                     Yes       Yes       Yes
Data Set                          CPS       CPS       CPS
N                              1,017,875 1,017,875 1,017,875


Notes: Standard errors in parentheses are robust with respect to heteroskedasticity and with respect to within state/year group correlation. Census data are from 1970, 1980, and 1990. CPS data are from 1977-78 May surveys and 1979-93 MORGs. See text for full list of demographic and other controls. Relevant models allow returns to demographic characteristics to differ in 1970, 1980, and 1990 (census) and to differ before and on/after 1986 (CPS). Columns (2) and (5) control for repeal state*construction. In column (6), "shortly after repeal" is 1-2 years, "long after repeal" is 3 or more years; in column (7), "shortly after repeal" is 1-4 years, "long after repeal" is 5 or more years. Columns (1)-(3) using census data Omit observations from Minnesota because Minnesota enacted a prevailing wage law in 1973. Observations are weighted with sampling weights described in the text.
                  Effect of Prevailing Wage Law Repeal on
                  the Construction Wage Premium by Race.
Variable                                       (1)       (2)       (3)
Repeal State*After
Repeal*Construction*Black                     0.043     0.041     0.055
                                             (0.019)   (0.012)   (0.032)
Repeal State*After Repeal*Construction       -0.040    -0.040    -0.013
                                             (0.013)   (0.013)   (0.014)
Repeal State*After Repeal*Black               0.003    -0.011    -0.036
                                             (0.030)   (0.014)   (0.011)
Repeal State*Shortly after
Repeal*Construction
*Black
Repeal State*Long after Repeal*Construction
*Black
Repeal State*Shortly after
Repeal*Construction
Repeal State*Long after Repeal*Construction
Repeal State*Shortly after Repeal*Black
Repeal State*Long after Repeal*Black
Full Set of State*Black Interactions?          No        Yes       Yes
Data Set                                     census    census      CPS
N                                           1,468,033 1,468,033 1,017,875
Variable                                       (4)       (5)
Repeal State*After
Repeal*Construction*Black
Repeal State*After Repeal*Construction
Repeal State*After Repeal*Black
Repeal State*Shortly after
Repeal*Construction                           0.051     0.037
*Black                                       (0.034)   (0.031)
Repeal State*Long after Repeal*Construction   0.056     0.068
*Black                                       (0.034)   (0.036)
Repeal State*Shortly after
Repeal*Construction                          -0.006    -0.006
                                             (0.016)   (0.015)
Repeal State*Long after Repeal*Construction  -0.015    -0.020
                                             (0.015)   (0.016)
Repeal State*Shortly after Repeal*Black      -0.040    -0.039
                                             (0.016)   (0.013)
Repeal State*Long after Repeal*Black         -0.035    -0.034
                                             (0.012)   (0.012)
Full Set of State*Black Interactions?          Yes       Yes
Data Set                                       CPS       CPS
N                                           1,017,875 1,017,875


Notes: Standard errors in parentheses are robust with respect to heteroskedasticity and with respect to within state/year group correlation. Census data are from 1970, 1980, and 1990. CPS data are from 1977-78 May surveys and 1979-93 MORGs. See text for full list of demographic and other controls. Relevant models allow returns to demographic characteristics to differ in 1970, 1980, and 1990 (census) and to differ before and On! after 1986 (CPS). Column (1) controls for repeal state*black and repeal state*black*construction; columns (2)-(5) include controls for state-fixed-effects*black and state-fixed-effects*black*construction. In column (4), "shortly after repeal" is 1-2 years, "long after repeal" is 3 or more years; in column (5), "shortly after repeal" is 1-4 years, "long after repeal" is 5 or more years. Columns (1) and (2) using census data omit observations from Minnesota because Minnesota enacted a prevailing wage law in 1973. Observations are weighted with sampling weights described in the text.
                  Effect of Prevailing Wage Law Repeal on
                  Unionization Rates and the Construction
                       Wage Premium by Union Status.
                                       Dependent Variable
                                             Union        Union    Union
                                             Member       Member  Member
Independent Variable                          (1)           (2)     (3)
Repeal State*After Repeal*Construction
*Union Member
Repeal State*After Repeal*Construction       0.003
                                            (0.016)
Repeal State*After Repeal*Union
Repeal State*After Repeal                    0.019
                                            (0.005)
Repeal State*Shortly after Repeal
*Construction*Union Member
Repeal State*Long after Repeal
*Construction*Union Member
Repeal State*Shortly after Repeal                          0.029   0.026
*Constructjon                                             (0.028) (0.019)
Repeal State*Long after Repeal                            -0.004  -0.015
*Construction                                             (0.017) (0.018)
Repeal State*Shortly after Repeal
*Union
Repeal State*Long after Repeal*Union
Repeal State*Shortly after Repeal                          0.015   0.013
                                                          (0.008) (0.006)
Repeal State*Long after Repeal                             0.020   0.024
                                                          (0.005) (0.006)
N                                           754,609       754,609 754,609
                                       ln(wage) ln(wage) ln(wage)
Independent Variable                     (4)      (5)      (6)
Repeal State*After Repeal*Construction -0.059
*Union Member                          (0.026)
Repeal State*After Repeal*Construction -0.001
                                       (0.019)
Repeal State*After Repeal*Union        -0.015
                                        0.012)
Repeal State*After Repeal              -0.007
                                       (0.008)
Repeal State*Shortly after Repeal                0.051   -0.006
*Construction*Union Member                      (0.035)  (0.030)
Repeal State*Long after Repeal                  -0.098   -0.112
*Construction*Union Member                      (0.026)  (0.028)
Repeal State*Shortly after Repeal               -0.026   -0.012
*Constructjon                                   (0.026)  (0.019)
Repeal State*Long after Repeal                   0.006    0.007
*Construction                                   (0.020)  (0.021)
Repeal State*Shortly after Repeal               -0.043   -0.020
*Union                                          (0.026)  (0.018)
Repeal State*Long after Repeal*Union            -0.006   -0.011
                                                (0.013)  (0.014)
Repeal State*Shortly after Repeal               -0.004   -0.009
                                                (0.011)  (0.009)
Repeal State*Long after Repeal                  -0.008   -0.006
                                                (0.009)  (0.009)
N                                      754,609  754,609  754,609


Notes: Standard errors in parentheses are robust with respect to heteroskedasticity and with respect to within state/year group correlation. Data are from the 1977-81 May CPS surveys and the 1983-93 GPS MORGs; data on unionization are unavailable for 1982. All columns control for state- and time-fixed-effects, state-fixed-effects*construction, time-fixed-effects*construction. They also control for demographic characteristics (see text) allowing returns-to-characteristics to differ before and on/after 1986. Columns (4)-(6) also control for state-fixed-effects*union member, time-fixed-effects*union, state-fixed-effects*union*construction, and time-fixed effects*union*construction. In columns (2) and (5), "shortly after repeal is 1-2 years, "long after repeal" is 3 or more years; in columns (3) and (6), "shortly after repeal" is 1-4 years, "long after repeal" is 5 or more years. Observations are weighted with sampling weights described in the text.
COPYRIGHT 2001 Sage Publications, Inc.
No portion of this article can be reproduced without the express written permission from the copyright holder.
Copyright 2001 Gale, Cengage Learning. All rights reserved.

Article Details
Printer friendly Cite/link Email Feedback
Title Annotation:bibliography included
Author:KESSLER, DANIEL P.; KATZ, LAWRENCE F.
Publication:ILR Review
Article Type:Statistical Data Included
Geographic Code:1USA
Date:Jan 1, 2001
Words:9879
Previous Article:THE USE OF FLEXIBLE STAFFING ARRANGEMENTS IN CORE PRODUCTION JOBS.
Next Article:IMMIGRATION REFORM AND THE EARNINGS OF LATINO WORKERS: DO EMPLOYER SANCTIONS CAUSE DISCRIMINATION?
Topics:

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