A new evaluation of impacts of prevailing wage law repeal

Journal of Labor Research

Volumn 17, Issue 2, 1996, Pages 297-322

  Thieblot, A J ^a  

^a Consulting Economist ^*  (United States)

Author keywords

[No Author keywords available]

Indexed keywords

EID: 0030556106     PISSN: 01953613     EISSN: None     Source Type: Journal    
DOI: 10.1007/BF02685847     Document Type: Review

References (56)

1. There is no room here for a full explanation of the Davis-Bacon act and other prevailing wage laws, for which see, inter alia, A. Thieblot, Prevailing Wage Legislation: The Davis-Bacon Act, State "Little Davis-Bacon" Acts, The Walsh-Healy Act, and Contract Act, Labor Relations and Public Policy Series No. 27 (Philadelphia: Wharton Industrial Research Unit, University of Pennsylvania, 1985), pp. 1-43 and A. Thieblot, State Prevailing Wage Laws, An Assessment at the Start of 1995 (Rosslyn, Va.: State Relations Department, Associated Builders and Contractors, 1993), pp. 1-39. In application, all prevailing wage laws set minimums based on the modal rate (the rate paid to the greatest number), usually provided it is the majority or some other set percentage of the similar craft and class, or it is based on the average, or it is set at the union rate. Even when not specifically mandated, the union rate is frequently found to be prevailing because that is the only rate other than the minimum wage likely to be paid to any significant number of different persons in the same job classification.
2. There is no room here for a full explanation of the Davis-Bacon act and other prevailing wage laws, for which see, inter alia, A. Thieblot, Prevailing Wage Legislation: The Davis-Bacon Act, State "Little Davis-Bacon" Acts, The Walsh-Healy Act, and Contract Act, Labor Relations and Public Policy Series No. 27 (Philadelphia: Wharton Industrial Research Unit, University of Pennsylvania, 1985), pp. 1-43 and A. Thieblot, State Prevailing Wage Laws, An Assessment at the Start of 1995 (Rosslyn, Va.: State Relations Department, Associated Builders and Contractors, 1993), pp. 1-39. In application, all prevailing wage laws set minimums based on the modal rate (the rate paid to the greatest number), usually provided it is the majority or some other set percentage of the similar craft and class, or it is based on the average, or it is set at the union rate. Even when not specifically mandated, the union rate is frequently found to be prevailing because that is the only rate other than the minimum wage likely to be paid to any significant number of different persons in the same job classification.
3. The breakdown of state prevailing wage laws at the beginning of 1995 was that 9 states had never had one, 9 states had repealed their laws between 1979 and 1988, and one state, Michigan, had a law inactivated by constitutional problems. Thus, 31 states plus the District of Columbia have active laws. About half of these specifically or effectively use union rates as prevailing, and the rest take Davis-Bacon or some central-tendency rate as prevailing. Additional prevailing wage laws exist in U.S. Territories and at the local level in a number of states, but I do not consider those here.
4. P. Philips, G. Mangum, N. Waitzman, and A. Yeagle, Losing Ground: Lessons from the Repeal of Nine "Little Davis-Bacon" Act (Salt Lake City: University of Utah, February 1995). I shall generally call this the Utah study.
5. Within weeks of its first appearance, the study's findings were quoted in deliberations of the U.S. Senate and House of Representatives and of state legislative bodies in Pennsylvania, Michigan, Indiana, California, and Ohio, among others.
6. This number is actually more than a little elusive in the Utah study. It is presented in the executive summary, as a "real" earnings loss of $1,477 as denominated in 1994 dollars (i.e., conformed to the purchasing power of 1994 dollars by inflating earlier dollars by the amount of change in the Consumer Price Index). An alternate number, also said to be denominated in "real" 1994 dollars, is found on page 16, where the cost to the same construction workers in the same states is said to be $2,016, or 36 percent more. This deconstructs to $1,835 of "real" earnings loss as denominated in 1991 dollars. No explanation is provided for the difference. On page 23, in the regression analysis, the loss is said to be $1,350 in "real" 1991 dollars. Finally, in another summary on page 68, it is alternatively $2,169 in 1991 dollars (fall from $24,317 before repeal to $22,148 after) - roughly $2,386 dollars of 1994 size - or $1,477 in 1994 dollars, depending on the analysis used to determine it.
7. The study says (p. 17) that "bringing [the revenue losses] to 1995 values using the consumer price index yields an estimated loss of $8.2 million in state taxes in Utah in 1991 evaluated in 1995 dollars." Two comments: first, this was published in February 1995, and it is extraordinary that the Utah study authors were able to anticipate the value of the Consumer Price Index to establish the worth of the 1995 dollar before the year even began. Second, "real" 1995 dollars are inappropriate to measure tax collections in earlier periods. See for example. Neter and Wasserman, Fundamental Statistics for Business and Economics, 2nd. ed. (Boston: Allyn and Bacon, Inc., 1961), "Some Problems in using index numbers to obtain constant-dollar data," pp. 629 et. seq.
8. The study goes on to claim that besides the lower rates, the repeal also exposed government purchasers of construction services to cost overruns, and included several charts to the effect that cost overruns on highway and heavy jobs in Utah increased after repeal (although, of course, from a lower base). Since the time frame of this examination spans more than two decades, and since allowing cost overruns and "costly change orders" is a matter entirely under the control of the contracting authority and is unrelated to prevailing wages, this part of the Utah study was not further analyzed. Also neglected here are claims in the Utah study of intensely negative impacts of repeal on apprentice training and some interference with minority development in the industry. These matters, though important, are secondary, and proper analysis of them would unduly lengthen my analysis.
9. In the typical depiction of a demand curve for labor, it is downward sloping to the right, indicating that if the price of labor decreases the quantity demanded will rise. The slope of this curve is what one of the Utah study's regression analyses set out to discover when estimating the increased employment that might come about as a result of reduced wage rates. What we are discussing here, however, is exogenous demand, which results is the demand curve shifting outward to the right. If an upward sloping supply curve can shift in proportion, the market-clearing price remains constant; if not, the price rises. See, for example, G. Bloom and H. Northrup, Economics of Labor Relations, 9th ed. (Homewood, III.: Richard D. Irwin, 1981), p. 333.
10. It is commonly assumed that when a union-rate prevailing wage law is repealed and rates on government work are no longer artificially sustained, all workers will thereafter receive the market rate, but that is not true. A union contractor will still have to pay the rates required by his union contract. Thus, union employees who continue to find work continue to be paid their union contract rate. However, it may be more difficult for union employees to find work as the (state) government sector becomes like the private one and union employers leave the market, and over time the union rate may trend towards the open market level.
11. The Utah study's authors rely on anecdotal information to assert that displaced union employees will take jobs away from nonunion employees, but fail to consider that the displaced union employees' jobs are now being filled by those same nonunion workers. Thus repeal of prevailing wages does not produce a cascade of job displacements, but an exchange of jobs between union and nonunion employees, or a three-way exchange of union employees becoming unemployed, nonunion workers taking over the jobs of the union workers, and the unemployed filling in where the nonunion workers left.
12. A market condition does, in fact, exist in which wages can be bid down. It is known as monopsony in economic theory, and is the rarely found situation of the single buyer in a competitive marketplace. Some have argued that monopsony characterized construction labor markets during the Depression. Between 1926 and 1933, for example, the volume of new construction in the United States declined from over $12 billion to under $3 billion, and employment in the industry fell from over 800,000 to under 400,000. Earnings plummeted, from $1,700 a year down to $820. Simultaneously, the proportion of public construction shot up from the traditional 20 percent to over 55 percent. Contractors on public works in that time period could offer below-market wages which workers would have to accept because there were no alternatives. It could be argued that because government construction was about the only game in town, government contractors were agents of a monopsony. Today, public construction is again only about 20 percent of the industry's activity, so there are about four times as many jobs in the private sector at any given time as in the public sector. Furthermore, there are active safety nets in place for the unemployed. No monopsony power lodges in the federal government's 20 percent market share, and therefore there is no need for the likes of a prevailing wage law to protect wage earners from its monopsonistic excesses.
13. These figures were provided by the State of Utah, Department of Community and Economic Development, citing in turn the Utah Department of Employment Security, Labor Market Information and Research. They cover average monthly employment, average monthly wages, and average number of establishments in the first quarter for all employees on nonagricultural payrolls and for various combinations of 2-digit SIC codes in the state, for 1953 to 1993. (Data in the author's possession.)
14. The claim that construction wage rates fell is predicated on their being measured in "real" dollars of 1991 or 1994 value, inflated to earlier years by the Consumer Price Index. Not only construction but almost all other average wage rates - in Utah, in other prevailing wage repeal states, in states that never had a prevailing wage law, and in states that have functioning prevailing wage laws - also declined in "real" terms during (and for a few years after) the period of double-digit inflation of the Carter years.
15. Georgia, Iowa, Mississippi, North Carolina, North Dakota, South Carolina, South Dakota, Vermont, and Virginia have never had prevailing wage laws.
16. In the order in which their prevailing laws were repealed, the repeal states are Florida (1979), Alabama (1981), Utah (1981), Arizona (1984), Colorado (1985), Idaho (1985), New Hampshire (1985), Kansas (1987), and Louisiana (1988). Michigan's law was also found to be unconstitutional and was rendered inactive in 1994, but it was active during all of the statistical comparisons made here or in the Utah study.
17. Explanations for the listings of increasing severity are found in Thieblot (1995), pp. 12-18. The groupings are as follows: weaker (nearer to market rate) laws, Nebraska, Tennessee, Oklahoma, Kentucky, Maine, Maryland, Montana, and Texas; average laws. Delaware, Connecticut, District of Columbia, Wyoming, New Mexico, Arkansas, Indiana, Pennsylvania, Alaska, Nevada, Oregon, West Virginia, and Wisconsin; more severe (near union rate) laws, Michigan, Illinois, Missouri, Rhode Island, Minnesota, Ohio, Washington, Hawaii, California, New Jersey, New York, and Massachusetts.
18. The number of data points, 27,778, claimed by the Utah study is puzzling, as is the collection of data on construction earnings by four-digit SIC code. Since there is no indication that inappropriate SIC codes (such as Operative Builders who build single-family housing for their own account) have been eliminated, there seems to be no reason not to use industry-wide annual averages, which would require fewer than 900 data points. The number is also puzzling because it requires that there be 32.04 SIC code classifications to reach the total number of observations (51 states × 17 years × 32.04 SIC codes = 27,778) where only 26 exist. (Office of Management and Budget, Standard Industrial Classification Manual, 1972.)
19. The range of control amounts by region is quite large, from positive $15,628 in Alaska to a negative $2,360 in the South area. It ignores such items as urbanization of different states within regions.
20. The secular trend amount is apparently not itself affected by regional differences or regional differences in inflation. Being in fixed dollar amounts, it has a much higher percentage effect on low-wage states than on high-wage ones, although no valid case could be made for why it should act this way. For 1991, the secular control variable would be .59 percent of the Hawaiian construction worker wage, but 1.17 percent of the New Mexican one.
21. The proposed relationship between statewide unemployment rates and construction earnings is unexplained in the text.
22. Note that in 1991 dollars the negative effect ascribed to repeal of prevailing wage law in the regression analysis (p. 23) is the $1,350 reported here. The difference is not overwhelming, but on page 20, the negative effect of repeal, in 1991 dollars, was said to be $1,835. See also note 6, above.
23. Many other important complications are also ignored by the Utah study. For example, construction workers are not paid annual incomes, but hourly wage rates, and the two are not conformable without taking work hour changes into account. This factor is not mentioned in the study, although there is growing evidence that nonunion employees work fuller years than union ones. See H. Northrup, Open Shop Construction Revisited (Philadelphia: Industrial Research Unit, The Wharton School, University of Pennsylvania, 1984), p. 520. Another matter is the wage dynamic associated with the changing number of persons employed. Since most employees enter or leave an industry at the bottom of the wage scale, expanding work forces tend to have lower average wages, and shrinking ones higher average wages. This factor, though mentioned in the text, is not included in the regression equations.
24. Unemployment rates are found in U.S. Department of Labor, Bureau of Labor Statistics, Geographic Profile of Employment and Unemployment, 1991. (Also available for other years.) Earnings information is from U.S. Department of Labor, Bureau of Labor Statistics, Employment and Earnings, 1991. This series is available, sometimes called Employment and Wages, for a number of years, some exclusively on microfisch. As of this writing, a few years from the mid-1980s do not seem to be available in any distributed format. Their lack is not critical to the analysis.
25. I could not simply rerun the analysis on the Utah study's data points to determine the regional weighting coefficients, because its authors declined to make the data set reasonably available.
26. It is possible to have a regression model that uses only unemployment rate, secular trend, legal status, and regional location as variables, and with such a model, the least squares regression line fit to the data would no doubt produce the coefficients for legal status found in the Utah study. But these coefficients are simply the ones that give the best fit to the data for that particular model, and, as already indicated, the model used by the Utah study is of questionable validity. The Utah study also has a second regression model, using what may be a different set of another 27,778 data points (it requires different data than the first, but is said to be the same size). It is intended to show the relationship between level of construction employment per Standard Industrial Classification (SIC) code and changes in average construction wages. For unexplained reasons, this analysis uses four-digit SIC codes, assigning an average employment level of 3,540 workers to each code, per state. This breakdown is strange, because the SIC code system classifies employers, not employees. For example, SIC code 1781 is for contractors who specialize in drilling water wells, whereas SIC code 1522 is for "general contractors primarily engaged in construction of residential buildings other than single-family housing." (Executive Office of the President, Office of Management and Budget, Standard Industrial Classification Manual, 1972.) The groups of contractors represented by these codes are similar neither in the number of persons employed by them, nor in their size from state to state or year to year. There is no reason to suppose that the size of these disparate groups might be affected in the same percentage or absolute way by existence or repeal of prevailing rate laws. Thus, to divide the industry's employees arbitrarily among these 26 groups of employers, "control" the groups' size to a national average, make assumptions about the impact of lower wage rates or the way the numbers might be increased or decreased by prevailing wage law repeal, statewide unemployment rates, and the rest, then expand the result back to the universe of construction workers is senseless. One could equally well divide construction workers by alphabetical listing or shoe size. The components of this model are even less compelling than those used previously for the economic model, and will not be covered further.
27. There is no valid reason for performing such an average over a variable number of years, just as there is no valid reason for deducing Utah construction workers' incomes from the average wage rates of construction workers in nine states when those figures for Utah construction workers are available directly.
28. My numbers use a somewhat different data set from the same source, the Bureau of Labor Statistics Employment and Earnings series.
29. The Utah study, using data from 1975-1991, found the nine repeal states' average incomes to decline from $24,317 to $22,482, a difference of $1,835. (Figure 2.5, p. 19.) I do not consider the variation from the numbers in the text significant.
30. The inflator used to translate 1979 and 1980 dollars into "real" 1991 dollars is the Consumer Price Index (CPI). On the basis of 1991 = 100, the CPI stood at 195 in 1979, and 165 in 1980. Following the study's example (at p. 20), I increased the purported loss 20 percent to restate it in 1995 dollars. (Actual CPI changes were 4 percent for 1991-1992 and 3 percent for both 1992-1993 and 1993-1994, so the Utah study must anticipate 10 percent additional inflation for 1994-1995.) The study further expands the losses by imposing them on the 31,528 persons employed in construction in Utah in 1991, rather than to the smaller numbers employed in the periods either before or after repeal - 27,224 and 29,723 respectively. The entire analysis for the State of Utah is repealed in slightly different form on p. 29 of the study, using the regression coefficient for "repeal states" of $1,477, upped from $1,350 to express it in 1994 dollars as a measure ot the loss. This amount is applied as above, but to the actual number of construction employees in Utah in each year from 1987 to 1993, less a small offset. Total tax losses are balanced against a range of hypothetical savings from repeal, which depend on the value of state-financed construction. The dollar amounts of state financed construction are not adjusted to constant 1994 dollars, but are in current, and therefore smaller, dollars.
31. According to Employment and Earnings, in 1993 the average annual employment in construction, counting 636,518 proprietors and an indeterminate number of management personnel who would not be subject to prevailing rates, wax 4,589,541. Thus no more than 3,953,023 persons could possibly be affected by prevailing rate repeal under the broadest assumptions of the Utah study. This number was readily available to the study's authors, who chose instead to use a hypothetical and unsupported number of 6,000,000.
32. These new employees go from unemployment or work in other industries to construction, where they are paid an average of around $24,000, much of which is new money on which taxes would be paid. This should certainly represent a net gain in tax receipts. Instead, the Utah study credits them with losing $1,477 each (and government is credited with losing several hundred dollars in taxes). Even with a model as crude as this one, the new employees should be stripped from the cost side.
33. The Utah study does state (p. 25) that although savings may be higher or lower, "The Congressional Budget Office [CBO] favors an estimate of a 1.5 percent cost savings associated with the wage effect plus a 0.2 percent cost savings because of paperwork associated with Davis-Bacon." Where I need to illustrate a point with a concrete example, I use this level, but point out in passing that it is inappropriate for the Utah repeal. Unlike the Utah's study's retroactive analysis of repeal impact, the CBO estimate is prospective (after 1995) because it supposes implementation of a 1983 court determination requiring rates to be set for construction helpers (a worker category found mostly in open-shop construction). CBO anticipates this to occur in 1995, decreasing the excess cost of Davis-Bacon by 1.6 percentage points, and consequently lowering the savings from repeal if repeal were to take place later. (Letter of June E. O'Neill, Director, Congressional Budget Office, to Honorable Cass Ballenger, Chairman, Subcommittee on Workforce Protection Committee on Economic and Educational Opportunities, April 21, 1995, p. 21.) As applied to historical figures such as those used in the Utah study, the proper CBO percentage estimate would be between 3.3 percent and 3.8 percent for a law that sets rates similarly to Davis-Bacon. Prior to repeal, however, Utah set rates effectively at the union rate, so the impact of repeal, using the CBO's methodology, would be considerably higher, probably on the order of 6 percent.
34. The $11.58 billion value for federal construction used by the Utah study (p. 30) is the figure for the amount of federally owned construction put in place in about 1985. (See U.S. Department of Commerce, International Trade Administration, Construction Review, Spring, 1994, Table A-3.) By fiscal 1993, that figure had grown to $17.2 billion, but still massively understated the amount of construction that might be affected by repeal, because Davis-Bacon applies to state grants-in-aid as well as to direct federal spending. The total figure for 1993 is actually $47.7 billion, budgeted to rise to $55.8 billion by 1995. The Utah study's figures thus understate the value of relevant construction (and therefore the savings that would result from repeal) by about a factor of five.
35. The original of this figure is found on page 29 of the Utah study.
36. See the Utah study executive summary. Curiously, in testimony prepared for delivery to the Ohio Legislature in June 1995, Professor Philips, the lead author of the Utah study, said: "We found that injury rates rose by 15 percent after the nine states repealed their prevailing wage laws. We did not calculate the increased worker compensation costs these injuries created nor the cost to the state for caring for uninsured construction workers." P. Philips, Wages and Benefits as a Percent of the Net Total Costs in the Construction Industry, (Salt Lake City, Utah: Economics Department, University of Utah, undated), p. 6.
37. Utah study, endnote 82.
38. U.S. Department of Labor, Occupational Safety and Health Administration, Construction Accidents: The Workers' Compensation Data Base, 1985-1988, April 1992, p. 13, cited in Utah study at endnote 78.
39. Same source page. The public relations flak who turns the theater reviewer's pan, "It was wonderful if you have trouble sleeping," into praise by quoting only the first three words is doing similar work.
40. U.S. Department of Labor, Bureau of Labor Statistics, Occupational Injuries and Illnesses in the United States by Industry, 1978, Bulletin 2078, August 1980, Table 6. Although the numbers shown in our Figure 6 are for 1978, there should be no material change from year to year.
41. The Utah study (p. 62) says that plumbers and pipefitters (SIC 171) were chosen because this specialty trade has injury rates in the mid-range of rates for construction. This is not supported by the sources cited (U.S. Department of Labor, Bureau of Labor Statistics Occupational Injuries and Illness Survey reports, various years), where this category is typically the second highest of all 3-digit codes within the construction industry. Furthermore, since all construction workers are presumably affected by prevailing wage laws, it is more reasonable to evaluate the injury and illness rates of all construction workers, which numbers are presented in the same reports, than to deliberately measure a nontypical subset's patterns and apply them to all.
42. If one were to assume there actually was an impact on injury rates as a result of inexperienced or untrained workers taking over prevailing rate work after repeal, there would be no reason to suppose a proportionate impact on segments such as roadbuilding, where plumbers and pipefitters are represented hardly at all, as on commercial construction, where they are numerous. This is a different statistical problem from that of extending the injury rate results from specialty contractors (plumbers and pipefitters) to other construction segments or to other industrial segments, like operative builders or residential contractors, whose employees or employees' activities would be very little influenced by prevailing rate repeal. Both analyses are neglected by the Utah authors.
43. The Utah study authors talk about injury rates, serious injury rates, and days lost per year but show no apparent appreciation of the difference between these series and workers compensation statistics. Accident rates reported for workers' compensation purposes (by the only 10 states that provided this data to the Bureau of Labor Statistics each year for the four consecutive years of the study) are quite different from accident rates reported to OSHA, partly because states delay from one to eight days before classifying an injury or illness as serious and reporting it. (U.S. Department of Labor, Occupational Safety and Health Administration. Construction Accidents: The Workers' Compensation Data Base 1985-1988, April, 1992.)
44. The Utah study's source, as cited in its text, p. 62, is the BLS Occupational Injuries and Illness Survey report for various years. But these reports do not give state-by-state data breakdowns and have never done so. A special Occupational Injuries and Illness by State report was issued in 1980, giving data for 1975-1976, but there has been no published report since. Furthermore, because of varying state participation in the survey, those state-by-state data that were available did not contain reports for several key states, including New York, New Jersey, Illinois, Ohio, and Texas, among others. Spokesmen at the OSHA Office of Statistical Studies and Analysis, interviewed July 29, 1995. stated that full state-by-state information is not available and has never been available in any form for either plumbers and pipefitters or for construction workers generally. The greatest concentration of published state-by-state data on construction injuries is found in "Oversight Hearings on OSHA - Occupational Safety and Health for Federal Employees, Part 4: State Plans," 96th Cong., 2d Sess., March 4, 11, 19; April 1, 29; May 28; June 17, 25; July 22; and September 16, 1980, p. 688. This gives incidence rates for construction injuries and illness per 100 fulltime workers in 45 reporting states in the five years, 1974-1978.
45. FrOm the OSHA hearings, 1980. Not all states are included. Georgia, one of the never-had states, declined to participate in the survey, as did Illinois, Ohio, New Jersey, New York, and Texas, all of which did have laws. Additionally, data are missing for Nevada (1975), Michigan (1976), Arkansas, North Dakota, and New Hampshire (1977 and 1978) and Oklahoma, Colorado, North Dakota, Tennessee, and Mississippi (1978). It does not appear that data missing from individual years should affect the results.
46. Utah study, p. 63. For reasons already stated, it is highly unlikely that the Utah study could possibly have reported data for 32 states that retained prevailing wage laws, since data were available from the stated source for a maximum of 27 states. Also note that the chart (Figure 4.1, p. 63) is titled "Injury Rates in Construction," whereas the text speaks of injury rates to plumbers and pipefitters, SIC 171. I can neither confirm nor deny the injury rates presented if they are for plumbers and pipefitters, since I know of no possible source for that information. It seems unlikely, however, that the pattern for the injury rate for this subset would be the inverse of the pattern for construction workers generally.
47. The pre-repeal data, for 1975 through 1978, are from the hearings cited in n. 33, supra. Data for 1990 through 1992 were made available to the author (in Washington, D.C., July 29, 1995) by Joseph DuBois, Director of Data Analysis, U.S. Department of Labor, Occupational Safety and Health Administration. Mr. DuBois, who had also prepared the data used in the 1980 OSHA hearings, verified that both data sets were similarly derived from individual state submissions, using the same definitions and statistical methodologies. He also confirmed that data are not available in any format from the federal government covering construction worker injury rates in Georgia, Illinois, New Jersey, New York, and Ohio for either time period, or from Texas in the earlier period and South Dakota, Colorado, Idaho, New Hampshire, the District of Columbia, Pennsylvania, and Wisconsin in the later one. For comparison purposes here, if no data were available for either time period, no estimates were made. If data were available for one or the other period, I assumed no change in average rate between periods.
48. 2 statistic (.16 to .49). The way the model is constructed, it in the years before prevailing law repeal for states within the same region (so no regional control variable) and the same time frame (eliminating differences in secular trend), states have the same unemployment rate (which, within the regions is an acceptably close estimate) their injury rates (or the log of their injury rates) will differ from one another only based on status of prevailing wage law: States that never had one should show higher rates than states that do have one (even if they will repeal it later). The actual statistics show, however, that such is not the case. BLS Region IV comprises four states that never had laws and four that had them in the period 1975-1978. The Utah regression model says the "never had" states should bave higher injury rates. Here are the figures for the "never had" states: Georgia, no report; South Carolina, 11.4; North Carolina, 12.2; Mississippi, 13.6; the figures for the "had law" states; Kentucky, 14.2; Alaska, 15.1; Tennessee, 15.8, Florida, 17.8. Clearly, the injury rates in "never had" states are lower. Similar results obtain in Region VIII, where North and South Dakota are substantially below their "have law" counterparts in Wyoming, Colorado, Montana, and Utah. In other words, in all cases where statistics are available to check, the model not only does not work, it works backwards.
49. C. Culver, Construction Fatalities: Comparison of Nonunion and Union Contractors (Rosslyn, Va.: National Center for Construction Education and Research, 1995), Table 2 gives comparison of union and nonunion contractor fatalities for the years 1985-1993. In every year, the union fatality rate is three to six percentage points higher than the nonunion rate.
50. With respect to death rates, OSHA's Analysis of Construction Fatalities - The OSHA Data Base 1985-1989, p. 18, notes: ". . . 30 percent of the fatalities occurred at union worksites, 70 percent at nonunion worksites. For that year, union membership accounted for 22 percent of the workforce, 78 percent was nonunion. The distribution of fatalities among union and nonunion worksites is similar to the composition of the construction workforce in tenus of union and nonunion workers for the 5-year period." It also notes (p. 19) that "The percentage of fatalities for the various size construction firms is similar to the representation of the construction workforce in terms of firm size." There is, of course, no statement about prevailing wage states, but a rough correlation between prevailing wage status and unionization can safely be assumed.
51. See Thieblot (1985), pp. 93-113. Although the Utah study cites this work seven times, and its authors can therefore be presumed to be familiar with its savings estimates restated here, one of them leads off a position paper read to the Ohio legislature thus: "It has been claimed that the states can save 20 percent or more on total construction costs by eliminating prevailing wage regulations. This claim is false. . . ." (P. Philips, Wages and Benefits as a Percent of Net Total Cost in the Construction Industry, Economics Department, University of Utah, undated, p. 1.) Philips then attacks this straw man with understandable success.
52. Professor Philips supposes (Wages and Benefits . . . undated, fn. 4): "The historic union-nonunion wage rate differential in the United States has remained fairly constant at around 20 percent." The Congressional Budget Office, to the contrary, states: "The ratio of cash wages for union construction workers to those for nonunion construction workers was 1.62 in 1993. as compared with 1.72 in 1983." (Letter from June O'Neill, Director, Congressional Budget Office, to Honorable Cass Ballenger, U.S. House of Representatives, April 21, 1995, p. 3.) Adding in fringe benefits would make the ratios even higher.
53. The variations in earnings from state to state can be very large and are preserved in this model. In 1993, the lowest average weekly wage was reported by the Bureau of Labor Statistics (in Employment and Earnings, 1993) for employees in South Dakota, at $358 per week. The highest, excluding the special case of the District of Columbia ($754 per week) was Connecticut, 1.8 times higher at $638 per week. For construction workers, the lowest was found in Arkansas ($395) and the highest in Alaska ($846). These differences may result from historical patterns, union strength, population density, urbanization, or other causes.
54. See endnote 16 for taxonomy. Notice that the repeal states, in the pre-repeal time period, showed ratios consistent with the fact that different repeal states had prevailing wage laws of differing severity.
55. Although there is no way to verify this short of actual repeal, there is no reason to believe it would not be true. Indeed, most state prevailing wage laws would probably follow a repealed federal law into oblivion. Furthermore, the observed relative earnings impact in repeal states occurred as though all public works jobs were affected, although federal projects and joint federal-state ones continued to be constructed under Davis-Bacon rates, so the opposite is also likely to occur.
56. According to the Bureau of Economic Analysis, U.S. Department of Commerce, the national income without capital consumption adjustment of the construction industry in 1993 was $228 billion. The International Trade Administration of the U.S. Department of Commerce reports federal and federal-grant outlays for construction in fiscal 1993 to be $47.2 billion. Hence, Davis-Bacon would cover 21 percent of the total. This is not evenly distributed state by state, but should be roughly proportional to the number of construction workers in each state.