Private School Competition and Public School Teacher Salaries RICHARD VEDDER and JOSHUA HALL Ohio University, Athens OH 45701 Teacher unions have fiercely fought public policy measures (e.g., vouchers, tuition tax credits) that might increase the proportion of students attending private schools. Yet increased competition in the educational service market should also lead to greater labor market competition, reducing any quasi-monopsony tendencies depressing teacher salaries. Using detailed data on over 600 Ohio school districts, we find that increased private school competition leads to higher salaries for public school teachers. It may be that union leaders disregard the interests of their members in trying to maximizing union size and power. An alternative interpretation is that unions sacrifice short-run income gains for their members in order to maintain long-term economic rents associated with substantial political power.
I. What Does Private School Competition Do to Public School Teacher Salaries? Teacher unions, especially affiliates of the National Education Association (NEA), have fought strongly and very often successfully to oppose public financial support for students attending private schools (Lieberman, 1997). They argue that vouchers, tuition tax credits, and other forms of governmental support for private instruction will destroy public education. Presumably, they are acting out of self-interest. Do they believe that by draining funds from public schools, vouchers or tax credits would lower the salaries of the public school teachers that the unions represent? Yet a body of rather elementary economic theory suggests that a vibrant competitive market in providing educational services should raise the salaries of public school teachers. Specifically, competition in the market for educational services also introduces labor market competition, reducing the monopsony power of public school districts. In many communities, there is one public school district within a low-cost commuting range of a large pool of teachers. By introducing viable private school alternatives, there should be some salary competition between educational providers in order to lure better teachers. Hence we hypothesize that the larger the presence of private school competition, the greater the typical public school teacher salary will be. There has been some previous empirical evidence to support this proposition (Hoxby, 1994a). Indeed, it can be argued that teacher unions historically have used their collective bargaining power to overcome the quasi-monopsonistic market structure in the indusJOURNAL OF LABOR RESEARCH Volume XXI, Number 1
Winter 2000
162
JOURNAL OF LABOR RESEARCH
try. A number of studies show that teacher unions have raised teacher wages and per pupil expenditures (Eberts and Stone, 1984, 1986, 1987; Zwerling and Thomason, 1995). If the theoretical proposition above holds, why would public school teacher unions fight attempts to introduce greater private school competition? It is true that on average private school teachers earn about 40 percent less than their public school counterparts (National Center for Educational Statistics, 1997). A compositional shift in the teaching labor force towards the private sector could lower average salaries for all teachers even if reduced monopsony had modest positive impact on salaries in both the public and private sector. For example, suppose that public school teachers average $40,000 a year and constitute 88 percent of all teachers and that private school teachers make typically $24,000 annually and are 12 percent of teachers. The average salary for all teachers would be $38,080. If increased public school competition raised public and private school average salaries by $280 (to $41,000 and $25,000 respectively), but the public school share of all teachers fell to 80 percent, the average salary for all teachers would fall by $1,000 to $37,800. The magnitude and direction of the impact of increased competition on the average salary for all teachers depends on the size of composition shift and the degree to which competition raises salaries. But unions like the NEA and the American Federation of Teachers (AFT) represent almost exclusively public school teachers. If their interest is in the welfare of their membership, why should union leaders fight privatization — if in fact such added competition raises salaries of the predominantly union members in public schools? Should not union leaders embrace privatization, regardless of what it does for average salaries in the entire teacher profession? It is true, of course, that increased competition would likely reduce aggregate union membership and revenues if private schools remained nonunion. Therefore, if it can be demonstrated that enhanced private school competition raises public school salaries, this could be viewed as support for the position that union leaders are seeking to maximize the union organization’s economic power (and possibly their own income and authority), not the interests of individual rank-and-file members nor even their collective incomes. An alternative possibility exists as to why unions might rationally oppose privatization even though its immediate impact is to raise public school teacher salaries. Since public school teachers make dramatically more than those in private schools with similar attributes (Chambers and Bobbitt, 1996), one might argue that teacher unions have used the political process to extract economic rents for their members. An erosion in union membership would lower the long-run ability of the unions to use political power to maintain such rents. Thus short-run economic gains from increased private competition might be eroded over time.
II. Testing the Hypothesis: Data and Methodology About a decade ago, the State of Ohio began what it termed a Educational Management Information System for its public schools, designed to provide information to permit greater public accountability. The 612 public school districts were required to report
RICHARD VEDDER and JOSHUA HALL
163
data on literally hundreds of variables. The data are readily accessible by the general public, in recent years on the Internet. Among the data collected are educational, experience, and salary data for teaching staff, student academic performance indicators, and demographic and economic data on school districts and their residents. In addition, the state has used Census tract data to gather by public school district some less publicly accessible information from the 1990 U.S. Census of Population on attendance in private schools. The interdistrict variation in average public school district salaries is great within Ohio, ranging from $24,070 to $51,352 in 1996, with a mean of $35,458. A two to one ratio of average salaries in the highest vs. lowest paying district has been observed in other states (Callas and McCormick, 1993). Ohio tends to be right in the middle of states in rankings on most educational variables including teacher salaries, suggesting that it is rather typical of all states. For example, the NEA estimates that in 1997 average annual public school teacher salaries in the U.S. were $38,600, whereas the Ohio estimate was $38,700, less than 0.3 percent above the national average (U.S. Bureau of the Census, 1998, p. 175). We used ordinary least squares regression analysis. Our dependent variable was average teacher salaries in 1996 as reported by Ohio school districts. That statistic excludes fringe benefits and supplemental employment contracts (e.g., extra pay for serving as a club adviser or coach). Our measure of private school competition, PRIVATE, was the percent of all students in the district who attended private school. That proportion varied dramatically over the state, ranging from zero to 42.52 percent. The private school data are necessarily somewhat imperfect, being from the 1990 Census, while our public school enrollment data are more recent. Students living in a school district attending a private school in another state or part of Ohio are excluded, a number believed to be small. We introduce eight other variables into the analysis for control purposes. Two variables look at teacher characteristics: EXPERIENCE, the average number of years of teaching experience of teachers in the school district, and MASTER’S, the percentage of teachers possessing at least a master’s degree. Given the explicit link of salaries to these considerations, it is expected that both variables would be strongly positively correlated with SALARY, or average salary per teacher. The percentage of teachers with master’s degrees varied between four and 80 percent across the state, so this factor potentially is quite important. Three school district characteristic variables are included, namely SPENDING, current expenditure per pupil, CLASS SIZE, the ratio of students to teachers, and ENROLLMENT, the average daily membership of the school district. Ceteris paribus, it is expected that more affluent districts pay higher salaries. Historical evidence suggests that the rise in teacher salaries is, indeed, significantly correlated with rising expenditures per pupil (Hanushek and Rivkin, 1997). Where spending per pupil has been curtailed by tax limitations, teacher salaries suffer (Figlio, 1997a). Teacher unions have long complained about large classes, and accordingly some compensating salary
164
JOURNAL OF LABOR RESEARCH
differential may be required to overcome the less desirable teaching setting associated with large classes. An analysis of a large data base of the National Center for Educational Statistics confirms the existence of class size compensating differential payments (Chambers and Bobbitt, 1996). Similarly, the large bureaucracy, regulatory regime, and impersonal treatment associated with large school districts might require a compensation differential. A positive relation is thus hypothesized between both ENROLLMENT and CLASS SIZE and SALARY. Two parental characteristics are incorporated: INCOME and EDUCATION. INCOME is average family income, as measured by Ohio personal income tax returns (which are conveniently filed by school district residence). EDUCATION is the percent of the population over 25 years of age with a college degree. Other things equal, we expect these variables to be positively associated with SALARY. Six school districts are excluded from the analysis because of extremely small size or inadequate data. Finally, just as the existence of private school competition is expected, other things equal, to raise teacher salaries by reducing quasi-monopsony elements in teacher labor markets, the same applies to public school competition. There are counties in Ohio with a single public school district, while at the other extreme, one county had over 30 districts (the state average is about seven). Other things equal, we expect that the greater public school competition, PUBLIC SCHOOL, the greater the average salary in any given district, although that conclusion based on theory is contradicted by some results showing a negative relationship (Hoxby, 1994b).
III. Empirical Results Table 1 summarizes the empirical findings. The model as a whole explains over threefourths of the considerable interdistrict salary differential. The hypothesized positive relationship between PRIVATE and SALARY is observed and is statistically significant at the one percent level. If a district goes from no competition to a situation where 20 percent of students go to private schools, the model predicts that average public school salaries would rise by $1,084, or about three percent if the district’s average salary was at the state arithmetic mean. Similarly, PUBLIC SCHOOL works as hypothesized and is of a meaningful magnitude. Other things equal, a teacher is estimated to earn $808 more if there are a dozen public school districts in a county instead of one. Comparing a single public school district in a county with no private schools to a county with 12 public school districts and 20 percent of the students in private schools, the average paid teacher is estimated to earn over five percent ($1,892) more in the district with public and private labor market competition. The other control variables work as hypothesized. Interestingly, all of them are statistically significant in the expected direction at the one percent level. Each dollar increase in spending per pupil is estimated, ceteris paribus, to increase teacher salaries by $1.82. Since the typical student teacher ratio observed was about 23 to 1, it implies that about eight percent of increased per pupil school spending is transmitted to exist-
RICHARD VEDDER and JOSHUA HALL
165
Table 1 The Determinants of Public School Teacher Salaries in Ohio N=606 Variable or Statistic
Regression Coefficient Or Statistic
t-Value
Constant
1,734.03
1.09
PRIVATE
54.20
3.47
546.45
10.14
MASTER’S
59.63
6.83
SPENDING
1.82
13.87
CLASS SIZE
228.62
5.06
ENROLLMENT
0.12
5.89
INCOME
0.19
5.33
EDUCATION
58.40
4.50
PUBLIC SCHOOL
73.45
4.20
EXPERIENCE
R2 F-Statistic
0.76 209.43
Source: See text.
ing teachers in the form of salary increases (that does not include other monies going for new additions to staff). Even controlling for per pupil expenditures, there is a rather powerful relationship between INCOME and SALARY. Compare two districts, otherwise identical except that one has average family income of $50,000 a year and the second has an average income of only $25,000. Teachers in the higher income district should average nearly $4,800 more in annual salary. Whether this reflects the fact that richer districts with more affluent teachers get the better teachers, that teachers in rich districts collect greater economic rents, or some combination of both factors, is not possible to ascertain from existing data. Some studies argue that higher salaries are associated with qualitative improvements in teaching (Figlio, 1997b), while others see little or no qualitative improvement (Ballou and Podgursky, 1997). The results above might be somewhat spurious, and that with changes in model specification the observed positive relationship between PRIVATE and SALARY would not be obtained. Accordingly, we did some sensitivity analysis, estimating models with different control variables or samples, consistently obtaining the hypothesized positive relationship between SALARY and PRIVATE. Table 2 reports five additional regression models. In the first two, additional control variables are introduced. In regression one, variables measuring student academic performance (9TH GRADE TEST), the proportion
166
JOURNAL OF LABOR RESEARCH
of students that were nonwhite (NONWHITE), and assessed property valuation per pupil (PROPERTY) were introduced. The hypothesized PRIVATE-SALARY relationship is maintained, albeit with a slightly lower coefficient. Only PROPERTY of the additional variables is statistically significant. In the second equation, three different control variables are introduced, one BIG8, a dummy variable for the eight largest big city school districts (e.g., Cleveland), ADC, the percent of the population receiving Aid to Dependent Children public assistance, and LOCAL, the percent of district revenues derived from local (property tax) sources. Again, PRIVATE is statistically significant at the five percent level, although again the result is somewhat weaker than denoted in Table 1. Interestingly, after controlling for other factors salaries were considerably lower in the big city districts, and a high proportion of local funding was associated with higher salaries; welfare participation was not statistically significant. The last three regressions reduce the sample used to explore the possibility that the results might arise in large part from the diverse nature of school districts examined. The third regression excludes the 17 school districts in the state with more than 10,000 students. The fourth regression has 177 districts with a high proportion of relatively economically disadvantaged students (districts below the state median with respect to income or college education of adults, but above it with respect to welfare participation). The last regression looks at 179 high socioeconomic status districts (all with incomes and college educational participation above the median, but welfare participation below it). In all three cases, PRIVATE is positive and, in two cases, is statistically significant at the five percent level. The results with respect to the low socioeconomic status districts are more dubious, although significant at the 10 percent level if a one-tailed test is used. Still other combinations and permutations of variables and sample definitions were examined, although not reported here. In all of the regressions, a positive relationship was observed between PRIVATE and SALARY, and in most cases that relationship was statistically significant at the five percent level or beyond. In short, the sensitivity analysis increases our confidence in the validity of the results reached above.
IV. Conclusions Competition in the market for teacher services means higher salaries for public school teachers, if our results using Ohio data are representative. If this is correct, why, then, do unions like the NEA and AFT oppose public support of private education? Do unions fight such competition because they are concerned about a possible erosion of average salaries in the teaching profession as a whole associated with increased allocation of human resources into the lower-paid private school market? Possibly, but we believe it is at least as likely that union leaders fear an erosion in their membership, and thus income and power. With that power comes the long-run ability to use the political process to attract economic rents from the taxpaying public. Nonetheless, our results suggest that greater private and public school competition offers an alternative to union-
RICHARD VEDDER and JOSHUA HALL
167
Table 2 Five Alternative Regressions Variable, Statistic Or Sample Size N= Sample
Regression 1
Coefficient or Value (t-Statistics in Parentheses) Regression 2 Regression 3 Regression 4
Regression 5
606 All
606 All
589 No Large Dist.
177 Low SES
179 High SES
4,032.77 (2.30)
3,144.23 (1.96)
2,062.21 (1.38)
6,062.75 (1.79)
7,031.96 (2.41)
PRIVATE
46.48 (2.96)
34.19 (2.22)
40.18 (2.70)
56.51 (1.47)
61.62 (2.24)
EXPERIENCE
555.76 (10.41)
526.82 (10.18)
543.07 (10.74)
624.42 (5.62)
336.71 (3.60)
MASTER’S
57.53 (6.60)
58.80 (7.03)
53.79 (6.58)
65.55 (3.63)
64.62 (3.60)
SPENDING
1.37 (7.06)
1.57 (10.46)
1.97 (15.81)
1.52 (3.88)
2.39 (13.21)
CLASS SIZE
228.59 (5.06)
215.80 (4.95)
171.83 (3.97)
247.17 (2.31)
331.28 (3.35)
ENROLLMENT
0.122 (5.54)
0.237 (7.35)
0.615 (10.29)
0.106 (2.37)
0.539 (7.21)
INCOME
0.158 (4.12)
0.134 (3.24)
0.217 (6.40)
EDUCATTION
62.56 (4.77)
44.93 (3.53)
30.52 (2.42)
PUBLIC SCHOOL
74.03 (4.20)
69.74 (4.15)
57.57 (3.48)
144.22 (3.30)
24.46 (0.94)
9TH GRADE TEST
–5.40 (0.64)
NONWHITE
0.28 (0.02)
PROPERTY
0.01 (3.46)
CONSTANT
BIG8
–6,669.83 (4.89)
ADC
15.98 (0.98)
LOCAL
46.40 (5.01)
R2
0.765
0.781
0.788
0.539
0.761
F-Statistic
160.82
176.11
239.75
28.28
77.74
168
JOURNAL OF LABOR RESEARCH
ization for diluting any monopsony effects on salaries arising from public school monopolies. Certainly they suggest that what might be in the best interest of rank-andfile public school teachers, at least in the short run, is in conflict with the political behavior of the leadership of these union organizations.
REFERENCES Ballou, Dale and Michael Podgursky. Teacher Pay and Teacher Quality. Kalamazoo, Mich.: W.E. Upjohn Institute for Employment Research, 1997. Callas, Rosanne and Rod McCormick. A Study of Factors Influencing Teacher Salaries in Vermont. Montpelier, Vt.: Vermont State Department of Education, 1993. Chambers, Jay and Sharon A. Bobbitt. The Patterns of Teacher Compensation: Statistical Analysis Report. Washington, D.C.: U.S. Government Printing Office, 1996. Eberts, Randall W. and Joe A. Stone. Unions and Public Schools. Boston, Mass.: Lexington Books, 1984. ______________.
“Teacher Unions and the Cost of Public Education.” Economic Inquiry 24 (October 1986):
631-43. ______________.
“Teacher Unions and the Productivity of Public Schools.” Industrial and Labor Relations Review 40 (April 1987): 354-63.
Figlio, David N. “Did the ‘Tax Revolt’ Reduce School Performance?” Journal of Public Economics 65 (September 1997a): 245-69. ______________.
“Teacher Salaries and Teacher Quality.” Economic Letters 55 (August 1997b): 267-71.
Hanushek, Eric A. and Steven G. Rivkin. “Understanding the Twentieth-Century Growth in U.S. School Spending.” Journal of Human Resources 32 (Winter 1997): 35-68. Hoxby, Caroline M. “Do Private Schools Provide Competition for Public Schools?” National Bureau of Economic Research Working Paper No. 4978, December 1994a. ______________. “Does Competition among Public Schools Benefit Students and Taxpayers?” National Bureau
of Economic Research Working Paper No. 4979, December 1994b. Lieberman, Myron. The Teacher Unions: How the NEA and AFT Sabotage Reform and Hold Students, Parents, Teachers, and Taxpayers Hostage to Bureaucracy. New York: Free Press, 1997. U.S., Bureau of the Census. Statistical Abstract of the United States: 1998. Washington, D.C.: U.S. Government Printing Office, 1998. U.S., National Center for Educational Statistics. Digest of Educational Statistics. Washington, D.C.: U.S. Government Printing Office, 1997. Zwerling, Harris L. and Terry Thomason. “Collective Bargaining and the Determinants of Teachers’ Salaries.” Journal of Labor Research 16 (Fall 1995): 467-84.
