# The effect of minimum salaries on employment of teachers: a test of the monopsony model.

1. IntroductionThe market for schoolteachers has long been cited as an example of labor monopsony, where employers--that is, school districts--enjoy market power (Ehrenberg and Smith 2000; Hyclak, Johnes, and Thornton 2005). The reasons offered are plausible. By definition, school districts enjoy exclusive territories. Teachers have specialized training that might not be interesting to nearby nonschool employers. Many teachers are second earners in their families and might be unwilling to relocate in response to distant job offers. There is some evidence that teachers have strong personal geographic preferences (Boyd et al. 2005). It would appear that the set of potential employers available to a certified teacher is limited to the set of school districts within commuting distance of current residence. For these reasons, it is often argued that the greatest monopsony power is probably held by rural school districts because of their large geographic size.

An empirical literature dating back to Landon and Baird (1971) has tried to document monopsony in this market by regressing teacher salaries on measures of geographic concentration of school districts. These studies often appear to find a negative effect of geographic concentration on teacher salaries (see Luizer and Thornton [1986] for a survey). A major difficulty with the salary-concentration research strategy is that concentration is inversely related to urbanization--because urban areas tend to have many school districts in close proximity--and salaries for all workers tend to be higher in urban areas. The apparent negative estimated effect of concentration may disappear when sufficient measures of urbanization are included (Ransom, Boal, and Beck 1998). Moreover, the urban-rural difference in teacher salaries appears similar to the urban-rural difference for other professionals (Hirsch and Nottage 1998). (1)

A second difficulty with the salary-concentration research strategy is the increasing presence of teachers' unions. (2) A monopsonist faces, or at least perceives, an upward-sloping supply of labor. This upward slope raises marginal labor cost above the current wage, causing the monopsonist to reduce employment. Equilibrium occurs on the labor supply curve but below the monopsonist's labor demand curve. By contrast, a union tries to raise the wage by moving the equilibrium above the supply curve. Estimates based on 1970s data show that unionized teachers enjoyed an average wage differential of roughly 10%. (3) In some cases, a union might even move the equilibrium above the demand curve--this is the prediction of the so-called contract curve model, for which Eberts and Stone (1984, 1986) found some evidence among New York State school districts. Any test for monopsony ought to exclude unionized districts from the sample or control for unionism in some way. This has not generally been done. (4)

A recent, seemingly unrelated literature has investigated the effects of legal minimum wages on employment in fast-food restaurants (Katz and Krueger 1992; Card and Krueger 1994). This literature has apparently documented slight increases in employment following increases in minimum wages. The authors note that this result is incompatible with competitive labor markets and may be evidence for employer monopsony. A modest increase in a minimum wage may increase employment in a monopsonized market because it flattens the employer's perceived labor supply curve and thereby lowers the employer's marginal labor cost curve. (5) These results hold whether the monopsony follows the old textbook model of an isolated employer or the "new monopsony" model of equilibrium search, proposed by Burdett and Mortensen (1998) and analyzed at length by Manning (2003). The new monopsony model assumes that market frictions rather than employer concentration are the source of employer monopsony. Thus, the new monopsony suggests that monopsony power could be held even by urban employers in unconcentrated markets. While the econometric results for fast-food restaurants have been challenged by other researchers (Neumark and Wascher 2000), the minimum-wage method offers a promising alternative research strategy for evaluating employer monopsony power.

Apparently unnoticed in the literature on teacher labor markets is that teachers in many states are subject to their own "minimum wages." (6) Many states set minimum salaries for public schoolteachers; although, local districts are free to set salaries above the legal minima. In some states these minimum salaries take the form of an elaborate schedule, varying by educational attainment and experience, so that they are effectively binding on senior teachers as well as new hires. Thus, a change in the minimum salary schedule can directly affect every teacher in a district. By contrast, a change in the U.S. federal minimum-wage law directly affects only the least skilled and least experienced workers.

This study measures the employment effects of increases in state minimum salaries in two nonunion states with different minimum-salary schemes: South Carolina and Texas. (7) The specification used here is similar to those used in the studies of fast-food restaurants previously cited, so the results may be directly compared. In fact, the results prove to be quite different. Increases in state minimum salaries decrease the number of teachers employed, even when (as in Texas) school districts are given additional state funding to pay for the minimum salary increases. This result is clearly incompatible with movement along a supply curve, so the monopsony model is rejected. Assuming that this result represents movement along a demand curve, the short-run elasticity of teacher demand is estimated to be between -0.2 and -0.4, holding constant districts' enrollment and tax base.

The paper is organized as follows: Section 2 presents alternative models of employment determination by school districts, contrasting the responses to minimum salaries by competitive and monopsonistic school districts. Section 3 analyzes a panel data set of South Carolina school districts. Section 4 analyzes a panel data set of Texas school districts. Section 5 concludes.

2. Models of Competitive and Monopsonistic Employment Determination

This section presents a comparative-static analysis of the effects of state minimum salaries on the employment of teachers. The standard textbook analysis of supply and marginal labor cost is extended to develop budget constraints with and without a corresponding change in state aid to assist the transition to the minimum salary. Throughout the discussion, the effects on competitor and monopsonist districts are contrasted.

Assume that the representative school district maximizes output of school services subject to a budget constraint. Denote the production function for school services by U(L, K), where L denotes teachers, and K denotes all other inputs. Assume this function is characterized by diminishing marginal rates of substitution, and that all inputs are normal.

Simple Minimum Salary

Suppose the district is a competitor in the market for teachers. Its perceived "labor supply curve" is therefore a horizontal line at the going salary level (Figure la). Denote this going salary level [W.sub.0]. Let B denote the district's budget, assumed given, and let the price of other inputs equal one dollar. Then the district's budget constraint is simply

B [greater than or equal to] K + [W.sub.0] L. (1)

When a state minimum salary level, [W.sub.min] > [W.sub.0], is imposed, the district's "labor supply curve" rises but remains horizontal, and the competitor's new budget constraint becomes

B [greater than or equal to] K+ [W.sub.min] L. (2)

[FIGURE 1a OMITTED]

[FIGURE 1b OMITTED]

Budget constraints without and with a state minimum salary are depicted in Figure 2a. Both budget constraints have the same intercept on the vertical (K) axis. However, the budget constraint with a state minimum salary has a steeper slope, reflecting the higher but still constant salary level. Clearly, the minimum salary forces the district to a lower isoquant, and it will produce fewer educational services than before. Let [L.sub.0] denote the initial level of teacher employment before the minimum salary. Substitution and income effects of the new minimum salary will both be negative, so that under the state minimum salary, teacher employment will decline to something less than [L.sub.0].

Alternatively, suppose the district is a monopsonist in the market for teachers. Its labor supply curve, denoted f(L), is upward sloping, so that f'(L) > 0 (Figure lb). For an unregulated monopsonist, marginal labor cost is given by

d / dL (f(L) x L) = f(L) + f' (L) x L, (3)

which is necessarily greater than W = f(L). Let [L.sub.0] denote the initial level of teacher employment, and [W.sub.0] denote the initial salary. When a state minimum salary level, [W.sub.min] > [W.sub.0], is imposed, the district's marginal labor cost acquires a discontinuity. For employment levels below [L.sub.1], defined by f([L.sub.1] = [W.sub.min], it becomes a horizontal line with height [W.sub.min]. For employment levels above [L.sub.1], the district's marginal labor cost is the same as before the minimum salary.

[FIGURE 2a OMITTED]

[FIGURE 2b OMITTED]

Budget constraints without and with a state minimum salary are depicted in Figure 2b. Without a state minimum salary, the monopsonist's budget constraint is given by

B [greater than or equal to] K + f(L) L, (4)

a curve concave to the origin because f'(L) > 0. With a state minimum salary, the monopsonist's budget constraint acquires a kink at [L.sub.1] as follows:

B [greater than or equal to] K + [W.sub.min] L, for L [less than or equal to] [L.sub.1] (5a)

B [greater than or equal to] K + f(L) L, for L [greater than or equal to] [L.sub.l]. (5b)

Clearly, the minimum salary forces the district to a lower isoquant, and it will produce fewer educational services than before, but it will likely employ more teachers. If the minimum salary, [W.sub.min], is only slightly higher than the original salary, [W.sub.0], both the minimum salary constraint, Equation 5a, and the labor-supply constraint, Equation 5b, will bind, and the district will now employ [L.sub.1] > [L.sub.0] teachers at the kink in its budget constraint. If the minimum salary, [W.sub.min], is slightly larger than this but still less than its original marginal labor cost, [W.sup.*.sub.0] = [W.sub.0] + f'([L.sub.0]), only the minimum salary constraint, Equation 5b, will bind, but the monopsonist will still employ more teachers than [L.sub.0]. Only if the minimum salary [W.sub.min] is greater than [W.sup.*.sub.0] will the district hire fewer teachers than [L.sub.0].

[FIGURE 3a OMITTED]

[FIGURE 3b OMITTED]

Minimum Salary with Salary Transition Aid

In Texas, beginning in the 1995-1996 school year, increases in the state minimum salary were accompanied by "salary transition aid" (STA), calculated to cover the difference between old salaries and the new minimum at the old level of employment. (8) Thus, STA increased the school district's budget by

[DELTA]n = ([W.sub..min] - [W.sub.0]) x [L.sub.0]. (6)

Consider the effect of a minimum salary with compensating STA.

Suppose again that a district is a competitor in the market for teachers. When a state minimum salary level, [W.sub.min] > [W.sub.0], is imposed, the district's "labor supply curve" rises but remains horizontal, as shown in Figure la, and the district's budget constraint becomes steeper. The competitor's new budget constraint, depicted in Figure 3a, becomes

[DELTA] B + B [greater than or equal to] K + [W.min] L. (7)

Note that STA ensures by definition that the old input combination is still feasible for the district. Clearly, the minimum salary with STA will permit the district to reach a higher isoquant, and it will produce more educational services than before, but it will employ fewer teachers. The substitution effect of the minimum salary with transition aid is negative, and there is no income effect.

Alternatively, suppose that the district is a monopsonist in the market for teachers. With a state minimum salary, the monopsonist's budget constraint acquires a kink at [L.sub.1] as follows:

[DELTA] B + B [greater than or equal to] K + [W.sub.min] L, for L [less than or equal to] [L.sub.1] , (8a)

[DELTA] B + B [greater than or equal to] K + f(L) L, for L [greater than or equal to] [L.sub.1]. (8b)

Again, STA ensures that the old input combination is still feasible for the district, as depicted in Figure 3b. If the new minimum salary, [W.sub.min], is less than [W.sup.*.sub.0] = [W.sub.0] + f'([L.sub.0]), the new budget constraint will be linear through the old input combination until it reaches a kink to the right of it, at [L.sub.1]. In that case, the minimum salary with STA will permit the district to reach a higher isoquant, and it will produce more educational services than before and employ more teachers. The district may choose the kink point [L.sub.1], where both the minimum salary constraint, Equation 8a, and the labor-supply constraint, Equation 8b, bind, or it may choose an interior solution between [L.sub.0] and [L.sub.1], where only the minimum salary constraint, Equation 8a, binds. In any case, a revealed preference argument ensures that in no case will the district reduce employment of teachers in response to the minimum salary with STA.

Summary

The effect of minimum salaries on employment of teachers depends critically on whether the school district is a competitor or a monopsonist. If the district is a competitor, a rise in the state minimum salary will cause it to employ fewer teachers. STA will attenuate but not reverse this effect. By contrast, if the school district is a monopsonist, a rise in the state minimum salary will cause it to employ more teachers, provided that the state minimum salary is not too large. STA will augment this effect.

3. Results for South Carolina

This section presents estimates based on a fairly long panel data set of South Carolina school districts. South Carolina provides a good setting for testing monopsony because public education is nonunion. The state has no statute regulating collective bargaining by teachers (Lund and Maranto 1996), and unionism is extremely low. Only 0.1% of school district employees were represented by bargaining units in 1987, while the U.S. average was 44.1%. (9) This reflects the nonunion character of the state as a whole. In 2000, only 13% of South Carolina public-sector workers were covered by union contracts, while the U.S. average was 42%, and only 3.4% of its private-sector workers were covered by union contracts, while the U.S. average was 9.8%. (10) South Carolina is a right-to-work state.

School District Data

School district data were taken from Rankings of the Counties and School Districts of South Carolina, published annually by the South Carolina State Department of Education. South Carolina currently has 86 school districts. Data were obtained for all districts in the school years 1985-1986 through 1998-1999, the last year for which complete data could be obtained at the time the analysis was conducted. Several mergers of school districts took place during the sample period, and a few observations were therefore excluded to ensure that mergers did not distort the results. (11) The available data include enrollment, number of teachers employed, total assessed value per pupil (a measure of school district wealth), and teacher salaries. Summary statistics are shown in Table 1.

Salaries of public schoolteachers in the United States are usually defined by local school districts according to a grid, whose dimensions are teaching experience and educational attainment. The Rankings publication reports salaries for four cells from each district's grid: a bachelor's degree with zero years of experience, a master's degree with zero years of experience, a bachelor's degree with 18 hours of graduate course work and 10 years of experience, and a master's degree with 17 years of experience. "Experience" means total teaching experience in any school district in South Carolina or elsewhere. (12)

South Carolina's state minimum salaries are defined using a similar grid. These minimum salaries are not reported in earlier issues of the Rankings publication and were obtained directly from the State Department of Education. Minimum salaries were increased in all but two years of the sample. (No increases were legislated for the years 1991-1992 and 1993-1994.) Whenever minimum salaries increased, they were always increased by the same percentage for all cells in the grid.

A reasonable way to measure the force of minimum salaries is by the "gap," or percent difference, between state minimum salaries in a particular year and actual salaries for the previous year. This is the definition used by Katz and Krueger (1992) and Card and Krueger (1994). Accordingly, define

[GAP.sub.i,t] = max {[WMIN.sub.t] - [W.sub.i,t-1] / [W.sub.i,t-1], 0}, (9)

where [W.sub.i,t-1] denotes the salary paid by district i to teachers with a bachelor's degree with zero years of experience in year t-1, and [WMIN.sub.t] denotes the state minimum wage for such teachers in year t. A positive sign for GAP indicates that a district must raise salaries to remain in compliance with state law. The size of GAP indicates the size of the required salary increase. By contrast, the value of GAP is zero for districts whose old salary is greater than the new state minimum. Figure 4a shows that, even though increases in the state minimum salaries were small, roughly half of South Carolina districts faced positive salary gaps whenever minimum salaries were increased (Figure 4b). South Carolina school districts were not given any STA to cope with these gaps. (13)

Rural school districts are more likely to hold monopsony power, according to the traditional monopsony model. Slightly more than half of South Carolina's school districts lie outside metropolitan statistical areas (MSAs). Table 2 shows summary statistics separately for districts inside and outside MSAs. On average, rural school districts in South Carolina are less than half as large as urban districts and pay slightly lower salaries.

Econometric Estimates

The econometric analysis seeks to determine whether state minimum salaries had a negative or positive effect on teacher employment. The basic equation to be estimated is

[DELTA] ln([L.sub.i,t]) = [[beta].sub.1] + [[beta].sub.2] [GAP.sub.it]. (10)

Here, [L.sub.i,t] denotes the number of (full-time equivalent) teachers employed in district i in school year t. [GAP.sub.i,t] is the percent difference between state minimum salaries for year t and actual salaries for year t-1 in district i, as defined in Equation 9.

The key parameter is [[beta].sub.2], the coefficient of GAP. Under the monopsony model, [[beta].sub.2] is positive, reflecting a movement along the district's labor supply curve. Formally, [[beta].sub.2] equals the short-run elasticity of labor supply to the individual school district, provided that the increase in the minimum salary is small.

[FIGURE 4a OMITTED]

[FIGURE 4b OMITTED]

By contrast, under the competitive model, [[beta].sub.2] is negative, reflecting a movement along the district's labor demand curve. Formally, without STA, [[beta].sub.2] equals the short-run gross or uncompensated elasticity of teacher demand by the individual school district, assuming that other determinants of teacher demand are held constant. Since these other determinants of teacher demand, such as enrollment or district wealth, might in fact be correlated with GAP, an expanded equation was also estimated as follows:

[DELTA] ln([L.sub.i,t]) = [[beta].sub.1] + [[beta].sub.2] [GAP.sub.i,t] + [[beta].sub.3] [DELTA] 1n([E.sub.i,t]) + [[beta].sub.4] [DELTA] ln ([V.sub.i,t]. (11)

Here, [E.sub.i,t] denotes district enrollment and [V.sub.i,t] total assessed valuation per pupil, a measure of district wealth.

These equations were estimated by ordinary least squares, with the results reported in Table 3. Estimates of Equation 10 are given in columns 1 and 3, and estimates of Equation 11 are given in columns 2 and 4. Fixed effects for years are included in all columns. Column 1 shows that for Equation 10, the coefficient of GAP is about -0.37 and easily less than zero at conventional levels of significance. Column 2 shows that when enrollment and assessed valuation are introduced as covariates, the coefficient of GAP drops to -0.26 but remains significant. Columns 3 and 4 show that when fixed effects for districts are introduced, the coefficient of GAP drops even further and is no longer statistically different from zero at conventional levels of significance. District fixed effects are intended to control for unobserved characteristics of districts, constant through time, that might be correlated with GAP.(14) However, those district fixed effects are themselves insignificant for the specifications with covariates, according to simple F-tests. (15) There appears to be a negative effect of minimum salaries on employment of teachers for South Carolina as a whole.

To investigate the traditional argument that monopsony power is stronger in rural areas, the same equations were estimated separately on urban and rural subsamples, with the results reported in Table 4. Columns 1 through 4 report estimates using only urban districts--that is, districts in MSAs. Estimates of Equation 10 are given in columns 1 and 3, and estimates of Equation 11 are given in columns 2 and 4. Estimates of the coefficient of GAP without district fixed effects for this subsample are roughly similar to estimates for the entire sample in Table 3, but the standard errors are nearly twice as large. Introduction of district fixed effects causes the estimates of the coefficient of GAP to become small and positive and raises the standard errors further. According to simple F-tests, the fixed effects are jointly significant for this subsample, so one must conclude that there is no evidence of a negative effect of minimum salaries on teacher employment in urban South Carolina. However, there is little evidence of a positive effect, either, because the estimated coefficients of GAP are only about half as large as their standard errors.

Columns 5 through 8 of the same table report estimates using only rural districts--that is, districts outside MSAs. Estimates of Equation 10 are given in columns 5 and 7, and estimates of Equation 11 are given in columns 6 and 8. Estimates of the coefficient of GAP without district fixed effects are roughly -0.2 and about twice their standard errors for this subsample, with p-values of 0.035 and 0.061, respectively. Introduction of district fixed effects does not change the coefficient estimates much but raises their standard errors substantially. However, the fixed effects are themselves not significantly different from zero. There is weak evidence for a negative effect of minimum salaries on teacher employment in rural districts, with a short-run elasticity of teacher demand of about -0.2.

Similar results were found when quadratic terms in enrollment (E) and assessed valuation (V) were introduced, or when Equations 10 and 11 were estimated separately on subsamples of large and small districts (see Appendix). In particular, estimates of the coefficient of GAP for "large" districts (observations with more than 4000 students) were close to zero. Estimates of the coefficient of GAP for "small" districts (fewer than 4000 students) were slightly larger than those for rural districts.

In summary, there is weak evidence in South Carolina of a negative effect of minimum salaries on teacher employment but--in contrast to the traditional argument--only for rural school districts.

4. Results for Texas

This section presents estimates based on a short panel data set of Texas school districts. As in South Carolina, public education in Texas is nonunion. The state has no statute regulating collective bargaining by teachers (Lund and Maranto 1996), and unionism is very low. Only 0.9% of school district employees were represented by bargaining units in 1987. (16) This reflects the nonunion character of the state as a whole. In 2000, only 21.6% of Texas public-sector workers were covered by union contracts, while the U.S. average was 42%, and only 4.7% of its private-sector workers were covered by union contracts, while the U.S. average was 9.8%. (17) Texas is also a right-to-work state.

Despite the absence of state legal protection, teachers in a few Texas school districts have achieved informal agreements for "exclusive consultation." These districts were excluded from the sample. (15)

School District Data

School district data for Texas were taken from Snapshot: School District Profiles, published annually by the Texas Education Agency. Texas has many more school districts than South Carolina--over 1000 total--but their average size is smaller and their size range greater. Data were obtained for the school years 1994-1995 though 1998-1999. These years were chosen because minimum salaries were not changed for many years before 1995 1996 and because a change in the state aid formula makes it impossible to calculate the GAP variable after 19981999. The available data include enrollment, number of teachers employed, total assessed value per pupil, and average teacher salaries. (Salaries for particular cells in the experience-education grid are never reported.) Texas has 1042 public school districts, but the actual number of districts available varies from year to year because of missing data. Summary statistics are shown in Table 5. (19)

Texas's state minimum salaries are defined using a simpler grid than South Carolina's grid. Since the 1995-1996 school year, Texas minimum salaries have varied only with experience, not educational attainment. ("Experience" means total teaching experience in any school district in Texas or elsewhere.) Unfortunately, the state minimum salary grid is difficult to compare with individual district average salaries reported in Snapshot.

Fortunately, an alternative measure of state minimum salaries is available in STA. From 1995-1996 through 1998-1999, Texas offered additional aid to local school districts to cover the difference between old salaries and the new minimum salaries at the old level of employment. Thus, the amount of STA received by a district in any given year, divided by the salary bill in the base year, equals the average gap between the year's state minimum salaries and the base year's actual salaries. The salary bill may be computed by multiplying average salary by employment, giving the following tentative definition for GAP:

[GAP.sub.i,t] = [STA.sub.i,t] / [L.sub.i,t]-1] x [WAVG.sub.i,t-1]. (12)

STA is not reported in the Snapshot publication--these data were obtained directly from the Texas Education Agency. (20)

In computing the GAP for Texas school districts, two complications require attention. The first is spurious correlation between the GAP variable and the dependent variable (i.e., the change in log employment). Note that the denominator of GAP as defined previously will be negatively correlated with the dependent variable because the base level of employment ([L.sub.i.t-1]) appears in the formulas for both variables. Using current employment ([L.sub.i,t]) in Equation 12 is no better--it will cause GAP to be positively correlated with the dependent variable for the same reason. To minimize spurious negative or positive correlation, the average of employment in the base and current years will be used to compute the GAP.

The second complication is the way Texas calculated STA. Texas operates on a fiscal biennium. Although minimum salaries were increased every year of the sample, the base year for calculating STA remained fixed for two years at a time. In particular, 1994-1995 served as the base year for computing STA in both the 1995-1996 and the 1996-1997 school years. Similarly, the 1996-1997 school year served as the base year for computing STA in both the 1997-1998 and the 1998-1999 school years. Thus, for the years 1995-1996 and 1997-1998, STA reflects a one-year increase in minimum salaries; whereas, for the years 1996-1997 and 19981999, STA reflects a two-year increase in minimum salaries.

[FIGURE 5a OMITTED]

[FIGURE 5b OMITTED]

To address these complications, the final definition of GAP becomes

[GAP.sub.i,t] = [STA.sub.i,t[ / (L.sub.i,base] + [L.sub.i,t] / 2) x [WAVG.sub.i,base], (13)

where [STA.sub.i,t] denotes STA received by district i in year t and [WAVG.sub.i,base] denotes the average teacher salary paid by district i in the base year. To accommodate the fiscal biennium, "base" is defined as the immediately prior school year when t = 1995-1996 or 1997-1998 and defined as the year before that when t = 1996-1997 or 1998-1999.

Figure 5a shows that, even though increases in the state minimum salaries were fairly small, a large and increasing number of districts had positive salary gaps (Figure 5b). By the end of the sample period, nearly all districts had positive gaps.

About 60% of Texas's school districts lie outside MSAs. Summary statistics are shown separately for districts inside and outside MSAs in Table 6. On average, rural school districts in Texas are only about a fifth as large as urban districts, and they pay slightly lower salaries.

Econometric Estimates

Equations 10 and 11 were estimated by ordinary least squares on the panel data set of all Texas school districts, with the results reported in Table 7. Again, estimates of Equation l0 are given in columns 1 and 3, and estimates of Equation 11 are given in columns 2 and 4. Fixed effects for years are included in all four columns. Columns 1 and 2 show that, without district fixed effects, estimates of the coefficient of GAP for Texas are negative and slightly larger than corresponding estimates for South Carolina. They are moreover statistically different from zero at conventional levels of significance. Introduction of district fixed effects raises the standard errors but also raises the coefficient estimates slightly, and the estimates remain statistically significant with p-values well below 0.01. The district fixed effects are themselves significant for every specification, according to simple F-tests. There is a clear negative effect of minimum salaries on employment of teachers for Texas as a whole.

Table 8 reports separate estimates for urban and rural subsamples. As in Table 4, columns 1 through 4 report estimates using only urban districts, while columns 5 through 8 report estimates using only rural districts. Estimates of Equation 10 are given in colunms 1, 3, 5, and 7, and estimates of Equation 11 are given in columns 2, 4, 6, and 8. In contrast to the results for South Carolina, the Texas results show virtually no difference between urban and rural school districts in the estimates of the coefficient of GAP. Introduction of district fixed effects, which are always significant in Texas, raises the estimates of the coefficient of GAP to between -0.35 and -0.39. P-values for the estimated coefficient of GAP are below 0.05 in every column but one. Similar results were obtained when quadratic terms in enrollment (E) and assessed valuation (V) were introduced (see Appendix). There is strong evidence for a negative effect of minimum salaries on teacher employment in both urban and rural districts, with a short-run elasticity of teacher demand of nearly -0.40.

Equations 10 and 11 were also estimated separately on subsamples of large and small districts (see Appendix). Estimates of the coefficient of GAP for "large" districts (observations with more than 1000 students) were smaller than those for urban districts and not statistically significant at conventional levels when district fixed effects were included. Estimates of GAP for "small" districts (fewer than 1000 students) were much larger--about -0.56, with p-values less than 0.02--when district fixed effects were included.

It is curious that the estimates of the coefficient of GAP for Texas are larger than the corresponding estimates for South Carolina. Recall that the South Carolina coefficients estimate the short-run gross, or uncompensated demand elasticity, while the Texas coefficients estimate the short-run net, or compensated demand elasticity, because districts received STA as minimum salaries were raised. One might have expected the Texas estimates to be closer to zero because they do not include an income effect. Note, however, that the South Carolina estimates are based on one-year differences, while the Texas estimates are based on a mixture of one- and two-year differences. If districts take time to adjust to salary changes, one would expect a larger change in employment in two years than in one year. Moreover, demand for teachers can differ across states for a variety of factors, including local preferences and state regulations.

One possible source of bias in the Texas data should be addressed. In principle, Texas's STA program could have created an incentive for school districts to inflate employment in the base year to maximize STA in subsequent years. The strength of this incentive would have depended on the size of the anticipated salary gap, therefore potentially causing negative bias in the estimated coefficient of GAP and obliterating evidence of monopsony power. (Bias of this type is not an issue in South Carolina, which provides no STA.) In practice, however, minimum salary increases are typically set by the Texas state legislature only at the end of the base year (Walt 1995, 1999), so school administrators would have required excellent foresight to exploit this incentive.

Nevertheless, to investigate the possibility of negative bias, Equations 10 and 11 were reestimated using only the two-year differences from the years 1994-1995 to 1996-1997 and from the years 1996-1997 to 1998-1999. Note that all these years were base years, so that if employment was inflated equally in all base years, the employment data after differencing should be less subject to bias. Results for the two-year differences subsample are shown in Table 9, with district fixed effects included in all specifications. All the coefficients of GAP for this subsample are more negative than the estimates in previous tables. In particular, the coefficients estimated on the entire sample and on the rural subsample are significantly different from zero, with p-values well below 0.01. There is no evidence of bias due to any incentive for districts to inflate initial employment and thereby maximize subsequent STA. (21)

In summary, there is strong evidence in Texas of a negative effect of increases in minimum salaries on employment of teachers. The effect is more negative and statistically stronger in rural and small school districts. The short-run elasticity of demand for teachers for all districts in Texas is probably about -0.4.

5. Conclusion

The effect of state minimum salaries on employment of teachers is definitely negative in Texas and probably negative in South Carolina. The evidence is strongest for school districts in rural areas. The simplest interpretation of these results is a rejection of the monopsony model--whether the traditional isolated employer model or the "new monopsony" model based on market frictions--for the market for schoolteachers. Teacher employment and salaries in these data sets appear to lie on downward-sloping teacher demand curves with short-run elasticities between -0.2 and -0.4. (22)

It might be argued in defense of the monopsony model that state minimum salaries for teachers are so high in these samples that the supply curve is no longer binding--that is, the minimum salary is above [W.sup.*.sub.0] in Figure la. If this interpretation is correct, the monopsony model is not rejected, strictly speaking, but simply irrelevant: State legislatures in South Carolina and Texas have taken away whatever market power school districts may have once possessed by raising minimum salaries into the demand-constrained range. But are minimum salaries in fact above [W.sup.*.sub.0] in Figure 1a? Figures 4 and 5 suggest otherwise because increases in minimum salaries in both states were modest and many districts were above the minimum, at least at the beginning of the samples.

Alternatively, it might be argued that some districts are supply-constrained monopsonists, exhibiting a positive relationship between minimum salaries and employment, while other districts are demand-constrained competitors, exhibiting a negative relationship. Econometric estimates in a study like this one measure only the average relationship. (23) Nevertheless, if this interpretation is correct, the preponderance of school districts in South Carolina and especially Texas have been shown to be demand-constrained.

However interpreted, the negative employment effects found in this paper stand in sharp contrast to the positive employment effects found by previous researchers (Katz and Krueger 1992; Card and Krueger 1994) for the fast-food industry using similar econometric specifications. Of course, the employers and workers in these two industries are quite different. Moreover, the structure of minimum salaries is different. In the fast-food industry, a single minimum salary directly affects only the lowest-paid and least experienced workers; whereas, in public education, a grid of minimum salaries directly affects even the most experienced teachers. From this perspective, one might expect changes in minimum salaries for teachers to have a greater effect on employment of teachers than changes in the minimum wage would have on employment of fast-food workers. On the other hand, the actual increases in minimum salaries analyzed in this study were quite modest--barely greater than contemporaneous increases in the consumer price index, on average--whereas the increases in the minimum wage analyzed by previous researchers were on the order of 15-20%. From this perspective, it is perhaps remarkable that an effect of minimum salaries on employment of teachers could be detected at all.

Appendix

This appendix reports alternative estimates for South Carolina and Texas when the samples were divided into large and small districts and when quadratic terms were added to Equation 11.

Equations 10 and 11 were estimated separately on subsamples of large and small districts. In South Carolina, median district enrollment was 3827, so to create roughly equal-sized subsamples, "large" districts were defined as all observations having enrollment greater than 4000 students and "small" districts as all observations having enrollment less than 4000 students. (Some districts that changed size are represented in both subsamples.) Results are presented in Table A1. They are similar to those for urban and rural districts in South Carolina. Estimates of the coefficient of GAP are of the same magnitude and statistical significance, and the district fixed effects are not significant when enrollment and assessed valuation are included as covariates, according to simple F-tests.

In Texas, median district enrollment was 868, so to create roughly equal-sized subsamples, "large" districts were defined as all observations having enrollment greater than 1000 students and "small" districts as all observations having enrollment less than 1000 students. (Again, some districts that changed size are represented in both subsamples.) Results are presented in Table A2. These results differ from those for urban and rural districts in Texas. Estimates of the coefficient of GAP for "large" districts are smaller than those for urban districts and not statistically significant at conventional levels regardless of whether district fixed effects are included. The fixed effects themselves are jointly significant. Estimates of the coefficient of GAP for "small" districts are larger in magnitude than those for rural districts, with p-values less than 0.02, regardless of whether district fixed effects are included. The fixed effects themselves are jointly significant, according to simple F-tests.

Equation 11 was reestimated with quadratic terms in the covariates:

[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII]. (A1)

Results are reported in Tables A3 and A4. The coefficients of the quadratic terms were sometimes significant, but the estimates of the coefficient of GAP were usually very similar to the estimates for the linear Equation 11.

Table A1. Ordinary Least Squares Results for South Carolina School Districts by District Enrollment Size Regressor 1 2 3 4 Sample definition Large Large Large Large Salary gap -0.4041 -0.2147 0.1095 0.0141 (0.1218) (0.1192) (0.1422) (0.1493) Change in log enrollment 0.6995 0.3681 (0.0563) (0.0780) Change in log total 0.0419 0.0345 assessed value per pupil (0.0147) (0.0145) Fixed effects for No No Yes Yes districts? Adjusted [R.sup.2] 0.1913 0.3807 0.3714 0.3962 Number of districts 47 47 47 47 Number of observations 559 559 559 559 F-statistic for district 4.3941 1.3038 fixed effects p-value 3.04E-17 0.0933 Regressor 5 6 7 8 Sample definition Small Small Small Small Salary gap -0.2790 -0.2146 -0.2384 -0.2551 (0.1025) (0.0940) (0.1661) (0.1674) Change in log enrollment 0.3517 0.1866 (0.0776) (0.0843) Change in log total 0.0101 0.0060 assessed value per pupil (0.0095) (0.0098) Fixed effects for No No Yes Yes districts? Adjusted [R.sup.2] 0.0881 0.1334 0.1179 0.1262 Number of districts 51 51 51 51 Number of observations 603 603 603 603 F-statistic for district 1.3985 0.9034 fixed effects p-value 0.0413 0.6629 Dependent variable is change in log full-time equivalent teacher employment. Standard errors (in parentheses) are computed using formulas robust to heteroskedasticity and second- order serial correlation. Year dummies included for all years (coefficient estimates not shown to save space). "Large" subsample contains all observations with more than 4000 students. "Small" subsample contains all observations with fewer than 4000 students. Table A2. Ordinary Least Squares Results for Texas School Districts by District Enrollment Size Regressor 1 2 Sample definition Large Large Salary gap -0.4114 -0.1438 (0.0874) (0.0737) Change in log enrollment 0.7273 (0.0411) Change in log total 0.0678 assessed value per (0.0185) pupil Fixed effects for No No districts? Adjusted [R.sup.2] 0.0774 0.4779 Number of districts 488 488 Number of observations 1894 1894 F-statistic for district fixed effects p-value Regressor 3 4 Sample definition Large Large Salary gap -0.1787 -0.1532 (0.0959) (0.0929) Change in log enrollment 0.4560 (0.0568) Change in log total 0.0549 assessed value per (0.0219) pupil Fixed effects for Yes Yes districts? Adjusted [R.sup.2] 0.5582 0.6123 Number of districts 488 488 Number of observations 1894 1894 F-statistic for district 5.2209 2.3434 fixed effects p-value 8.2E-130 3.63E-34 Regressor 5 6 Sample definition Small Small Salary gap -0.3390 -0.3133 (0.1294) (0.1223) Change in log enrollment 0.3925 (0.0809) Change in log total 0.0674 assessed value per (0.0230) pupil Fixed effects for No No districts? Adjusted [R.sup.2] 0.0133 0.1160 Number of districts 572 572 Number of observations 2233 2233 F-statistic for district fixed effects p-value Regressor 7 8 Sample definition Small Small Salary gap -0.5955 -0.5562 (0.2198) (0.2231) Change in log enrollment 0.2495 (0.1043) Change in log total 0.0682 assessed value per (0.0337) pupil Fixed effects for Yes Yes districts? Adjusted [R.sup.2] 0.2551 0.2814 Number of districts 572 572 Number of observations 2233 2233 F-statistic for district 2.2662 1.8972 fixed effects p-value 8.37E-37 7.05E-23 Dependent variable is change in log full-time equivalent teacher employment. Standard errors (in parentheses) are computed using formulas robust to heteroskedasticity and second order serial correlation. Year dummies included for all years (coefficient estimates not shown to save space). Data for the years 1995-1996 and 1997-1998 are one-year differences, while data for the years 1996-1997 and 1998-1999 are two-year differences. "Large" subsample contains all observations with more than 1000 students. "Small" subsample contains all observations with fewer than 1000 students. Table A3. Ordinary Least Squares Results for South Carolina School Districts with Quadratic Terms in Enrollment and Assessed Valuation Regressor 1 2 3 Sample definition All All Metro Salary gap -0.2764 -0.1760 -0.2376 (0.0718) (0.1137) (0.1255) Change in log enrollment 0.4370 0.2115 0.5691 (0.0598) (0.0739) (0.0746) Change in log total 0.0212 0.0182 0.0706 assessed value per pupil (0.0083) (0.0080) (0.0191) Change in log enrollment 0.9684 0.9058 1.4805 squared (0.6825) (0.6970) (0.9150) Change in log total assessed -0.0097 -0.0134 -0.0359 value per pupil squared (0.0031) (0.0034) (0.0114) Fixed effects for districts? No Yes No Adjusted [R.sup.2] 0.2087 0.2175 0.3307 Number of districts 91 91 40 Number of observations 1163 1163 516 F-statistic for district 1.1419 fixed effects p-value 0.1805 Regressor 4 5 6 Sample definition Metro Nonmetro Nonmetro Salary gap 0.0690 -0.1865 -0.2413 (0.1603) (0.0938) (0.1663) Change in log enrollment 0.2809 0.3111 0.1711 (0.0837) (0.0891) (0.1074) Change in log total 0.0719 0.0019 0.0024 assessed value per pupil (0.0167) (0.0124) (0.0121) Change in log enrollment 0.6693 0.7587 0.9880 squared (0.8926) (0.8568) (0.9086) Change in log total assessed -0.0467 -0.0106 -0.0099 value per pupil squared (0.0104) (0.0045) (0.0047) Fixed effects for districts? Yes No Yes Adjusted [R.sup.2] 0.3863 0.1353 0.1095 Number of districts 40 51 51 Number of observations 516 647 647 F-statistic for district 2.1570 0.6350 fixed effects p-value 0.0001 0.9764 Regressor 7 8 Sample definition Large Large Salary gap -0.2187 0.0135 (0.1183) (0.1495) Change in log enrollment 0.6604 0.3680 (0.0616) (0.0868) Change in log total 0.0380 0.0281 assessed value per pupil (0.0224) (0.0246) Change in log enrollment 1.1626 -0.0330 squared (1.1481) (0.9475) Change in log total assessed 0.0149 0.0315 value per pupil squared (0.0766) (0.0838) Fixed effects for districts? No Yes Adjusted [R.sup.2] 0.3799 0.3939 Number of districts 47 47 Number of observations 559 559 F-statistic for district 1.2722 fixed effects p-value 0.1150 Regressor 9 10 Sample definition Small Small Salary gap -0.2344 -0.2804 (0.0963) (0.1677) Change in log enrollment 0.3275 0.1676 (0.0813) (0.0908) Change in log total 0.0137 0.0135 assessed value per pupil (0.0104) (0.0094) Change in log enrollment 0.9120 0.8397 squared (0.8061) (0.8323) Change in log total assessed -0.0067 -0.0151 value per pupil squared (0.0047) (0.0040) Fixed effects for districts? No Yes Adjusted [R.sup.2] 0.1329 0.1275 Number of districts 51 51 Number of observations 603 603 F-statistic for district 0.9278 fixed effects p-value 0.6164 Dependent variable is change in log full-time equivalent teacher employment. Standard errors (in parentheses) are computed using formulas robust to heteroskedasticity and second- order serial correlation. Year dummies included for all years (coefficient estimates not shown to save space). "Large" subsample includes observations with more than 4000 students. "Small" subsample includes observations with fewer than 4000 students. Table A4. Ordinary Least Squares Results for Texas School Districts with Quadratic Terms in Enrollment and Assessed Valuation Regressor 1 2 3 Sample definition All All Metro Salary gap -0.2792 -0.4093 -0.2414 (0.0804) (0.1367) (0.1492) Change in log enrollment 0.5058 0.2954 0.5734 (0.0669) (0.0936) (0.0826) Change in log total 0.0910 0.0743 0.0941 assessed value per pupil (0.0191) (0.0212) (0.0299) Change in log enrollment -0.6533 -0.4720 0.4575 squared (0.4380) (0.4972) (0.4330) Change in log total assessed 0.0059 0.0011 -0.0171 value per pupil squared (0.0554) (0.0488) (0.0755) Fixed effects for districts? No Yes No Adjusted [R.sup.2] 0.2006 0.3577 0.3387 Number of districts 1033 1033 402 Number of observations 4127 4127 1606 F-statistic for district 1.9699 fixed effects p-value 6.76E-45 Regressor 4 5 6 Sample definition Metro Nonmetro Nonmetro Salary gap -0.3645 -0.2088 -0.3951 (0.1536) (0.1029) (0.1843) Change in log enrollment 0.1151 0.4025 0.2816 (0.0681) (0.0811) (0.1034) Change in log total 0.0622 0.0548 0.0673 assessed value per pupil (0.0326) (0.0248) (0.0263 Change in log enrollment 1.2631 -0.9108 -0.7220 squared (0.3500) (0.4794) (0.5422) Change in log total assessed -0.0528 0.0548 0.0320 value per pupil squared (0.0637) (0.0667) (0.0612) Fixed effects for districts? Yes No Yes Adjusted [R.sup.2] 0.4986 0.1513 0.2958 Number of districts 402 631 631 Number of observations 1606 2521 2521 F-statistic for district 2.2536 1.8092 fixed effects p-value 2.79E-26 8.56E-22 Regressor 7 8 Sample definition Large Large Salary gap -0.1430 -0.1554 (0.0717) (0.0920) Change in log enrollment 0.7250 0.4262 (0.0381) (0.0593) Change in log total 0.0791 0.0631 assessed value per pupil (0.0246) (0.0304) Change in log enrollment 0.0272 0.3016 squared (0.4979) (0.4242) Change in log total assessed -0.0956 -0.0767 value per pupil squared (0.0726 (0.0804) Fixed effects for districts? No Yes Adjusted [R.sup.2] 0.4792 0.6132 Number of districts 488 488 Number of observations 1894 1894 F-statistic for district 2.3254 fixed effects p-value 1.81E-33 Regressor 9 10 Sample definition Small Small Salary gap -0.3038 -0.5697 (0.1210) (0.2159) Change in log enrollment 0.4217 0.2712 (0.0777) (0.1058) Change in log total 0.0634 0.0708 assessed value per pupil (0.0266) (0.0300) Change in log enrollment -0.8164 -0.5573 squared (0.4624) (0.5189) Change in log total assessed 0.0446 0.0164 value per pupil squared (0.0629) (0.0533) Fixed effects for districts? No Yes Adjusted [R.sup.2] 0.1487 0.2916 Number of districts 572 572 Number of observations 2233 2233 F-statistic for district 1.7761 fixed effects p-value 1.33E-18 Dependent variable is change in log full-time equivalent teacher employment. Standard errors (in parentheses) are computed using formulas robust to heteroskedasticity and second order serial correlation. Year dummies included for all years (coefficient estimates not shown to save space). Data for the years 1995-1996 and 1997-1998 are one-year differences, while data for 1996-1997 and 1998-1999 are two-year differences. "Large" subsample includes observations with more than 1000 students. "Small" subsample includes observations with fewer than 1000 students.

This paper has benefited from helpful comments on earlier drafts by David Harrington and Ronald Oaxaca, by seminar participants at Drake University and Iowa State University, and by anonymous referees.

Received November 2006: accepted February 2008.

References

Boyd, Donald, Hamilton Lankford, Susanna Loeb, and James Wyckoff. 2005. The draw of home: How teachers' preferences for proximity disadvantage urban schools. Journal of Policy Analysis and Management 24:113-32.

Burdett, Kenneth, and Dale T. Mortensen. 1998. Wage differentials, employer size, and unemployment. International Economic Review 39:257-73.

Card, David, and Alan B. Krueger. 1994. Minimum wages and employment: A case study of the fast food industry in New Jersey and Pennsylvania. American Economic Review 84:772-93.

Clark, Catherine. 2001. "Texas." Public School Finance Programs of the U.S. and Canada 1998-1999. Accessed 20 February 2003. Available http://nces.ed.gov/edfin/state_finance/StateFinancing.asp.

Currie, Janet, Mehdi Farsi, and W. Bentley MacLeod. 2005. Cut to the bone? Hospital takeovers and nurse employment contracts. Industrial and Labor Relations Review 58:471-93.

Dickens, Richard, Stephen Machin, and Alan Manning. 1999. The effects of minimum wages on employment: Theory and evidence from Britain. Journal of Labor Economics 17:1-22.

Eberts, Randall W., and Joe A. Stone. 1984. Unions and public schools: The effect of collective bargaining on American education. Lexington, MA: Lexington Books.

Eberts, Randall W., and Joe A. Stone. 1986. On the contract curve: A test of alternative models of collective bargaining. Journal of Labor Economics 4:66-81.

Ehrenberg, Ronald G., and Robert S. Smith. 2000. Modern labor economics: Theory and public policy. 7th edition. New York: HarperCollins.

Freeman, Richard B. 1986. Unionism comes to the public sector. Journal of Economic Literature 24:41-86.

Hirsch, Barry T., and David A. Macpherson. 2003. Union membership and coverage database from the Current Population Survey: Note. Industrial and Labor Relations Review 56:349-54.

Hirsch, Barry T., and Brian A. Nottage. 1998. Interarea differences in teacher salaries: The role of collective bargaining laws and employer concentration. Paper presented at the annual meeting of the Western Economic Association, Lake Tahoe, NV.

Hirsch, Barry T., and Edward J. Schumacher. 2005. Classic or new monopsony? Searching for evidence in nursing labor markets. Journal of Health Economics 25:969-89.

Hyclak, Thomas, Geraint Johnes, and Robert Thornton. 2005. Fundamentals of labor economics. Boston: Houghton Mifflin.

Katz, Lawrence F., and Alan B. Krueger. 1992. The effect of the minimum wage on the fast-food industry. Industrial and Labor Relations Review 46:6-21.

Landon, John H., and Robert N. Baird. 1971. Monopsony in the market for public school teachers. American Economic Review 61:966-71.

Luizer, James, and Robert Thornton. 1986. Concentration in the labor market for public school teachers. Industrial and Labor Relations Review 39:573-84.

Lund, John, and Cheryl Maranto. 1996. Public sector labor law: An update. In Public sector employment in a time of transition, edited by Dale Belman, Morley Gunderson, and Douglas Hyatt. Madison, WI: Industrial Relations Research Association.

Manning, Alan. 2003. Monopsony in motion: Imperfect competition in labor markets. Princeton, NJ: Princeton University Press.

Neumark, David, and William Wascher. 2000. Minimum wages and employment: A case study of the fast-food industry in New Jersey and Pennsylvania: Comment. American Economic Review 90:1362-96.

Ransom, Michael R., William M. Boal, and Paul M. Beck. 1998. Empirical tests of labor monopsony: School teachers in Missouri. Paper presented at the annual meeting of the Western Economic Association, Lake Tahoe, NV.

South Carolina Department of Education. 1985-1999. Rankings of the counties and school districts of South Carolina. Columbia: South Carolina Department of Education.

Stigler, George. 1946. The economics of minimum wage legislation. American Economic Review 36:358-65.

Tetreault, Donald R., and Daniel Chandler. 2001. "South Carolina." Public School Finance Programs of the U.S. and Canada 1998-1999. Accessed 20 February 2003. Available http://nces.ed.gov/edfin/state_finance/StateFinancing.asp.

Texas Education Agency. 1994-1999. Snapshot: School district profiles. Austin, TX: Texas Education Agency.

U.S. Bureau of the Census. 1975. 1972 census of governments, volume 3, public employment, number 3, labor-management relations. Washington, DC: Government Printing Office.

U.S. Bureau of the Census. 1984. 1982 census of governments, volume 3, public employment, number 3, labor-management relations. Washington, DC: Government Printing Office.

U.S. Bureau of the Census. 1991. 1987 census of governments, volume 3, public employment, number 3, labor-management relations. Washington, DC: Government Printing Office.

U.S. Bureau of the Census. 2007. Statistical abstract of the United States: 2007. Washington, DC: Government Printing Office.

Walker, Billy D., and Daniel T. Casey. 1996. The basics of Texas public school finance. 6th edition. Austin, TX: Texas Association of School Boards.

Walt, Kathy. 1995. Lawmakers agree on state-funded teacher pay hike. Houston Chronicle, 13 May, p. A1.

Walt, Kathy. 1999. Texas Senate gives blessing to school funds: $2.6 billion plan would provide raises for teachers, help pay debt. Houston Chronicle, 29 April, p. A1.

William M. Boal, College of Business and Public Administration, Drake University, Des Moines, Iowa 50311, USA; E-mail william.boal@drake.edu.

(1) A related literature examines nursing labor markets, regressing nursing salaries against measures of hospital concentration. Recent contributions include Currie, Farsi, and Macleod (2005) and Hirsch and Schumacher (2005).

(2) About 44% of school district employees in the United States were represented by bargaining units in 1987, and the proportion has likely risen since then (U.S. Bureau of the Census 1991, table 3). More recent editions of the Census of Governments do not tabulate collective bargaining by public employees.

(3) See survey in Freeman (1986, table 6).

(4) Luizer and Thornton (1986) estimate the concentration-salary relationship using data on Pennsylvania school districts in 1981, notwithstanding that more than half of Pennsylvania's school employees were covered by collective bargaining agreements (U.S. Bureau of the Census 1984, table 3). Landon and Baird (1971) estimate the concentration-salary relationship using data on 136 school districts in the United States in 1967, notwithstanding that, in 1972, more than one-quarter of U.S. school districts engaged in collective bargaining (U.S. Bureau of the Census 1975, table 3).

(5) This peculiar effect was apparently first noted by Stigler (1946). Of course, a very high minimum wage causes a decrease in employment in a monopsonized labor market, just as it does in a competitive labor market (see section 2).

(6) Among southeastern states, Alabama, Arkansas, Delaware, Georgia, Kentucky, Louisiana, Mississippi, North Carolina, Oklahoma, South Carolina, Tennessee, Texas, and West Virginia have state minimum salary schedules, but Florida, Maryland, and Virginia do not (Gale F. Gaines, Southern Regional Education Board, personal communication, June 9, 2000).

(7) Despite the absence of unions, teacher salaries in these states are not especially low compared to other states. South Carolina ranked 28th and Texas ranked 31st by average teacher salaries in 2004 (U.S. Bureau of the Census 2007, table 240).

(8) Section 1.09 of Texas House Bill 4, part (a). See also Walker and Casey (1996).

(9) U.S. Bureau of the Census (1991, table 3).

(10) Hirsch and Macpherson (2003).

(11) The following mergers took place during the sample period. In the summer of 1986, Dorchester 1 and Dorchester 3 merged to form Dorchester 4. Premerger observations on Dorchester 1 and 3 were dropped from the sample. In the summer of 1997, three mergers took place. Orangeburg 3 and Orangeburg 7 merged to form a new Orangeburg 3. Orangeburg 1, Orangeburg 4, and Orangeburg 8 merged to form a new Orangeburg 4. Orangeburg 2, Orangeburg 5, and Orangeburg 6 merged to form a new Orangeburg 5. Postmerger observations on Orangeburg 3, 4, and 5 were dropped from the sample. Enrollment data for the 1999-2000 school year are now available from the state department of education, but the number of teachers and teacher salary schedules are not available.

(12) South Carolina teachers participate in a statewide retirement plan, South Carolina Retirement Systems (http://www.retirement.sc.gov), so pension vesting does not pose a drag on mobility.

(13) South Carolina does offer across-the-board aid when minimum salaries are increased, but the amount awarded to each district is unrelated to the district's existing salaries. See Tetreault and Chandler (2001).

(14) Unobserved time-varying district characteristics might still be correlated with GAP, of course.

(15) Strictly speaking, these F-tests are invalid in the presence of heteroskedasticity or serial correlation.

(16) U.S. Bureau of the Census (1991, table 3).

(17) Hirsch and Macpherson (2003).

(18) Texas school districts with "exclusive consultation" include Corpus Christi, North Forest, San Antonio, South San Antonio, and Austin (Patricia Olshefski, American Federation of Teachers, personal communication, October 18, 1993).

(19) Texas teachers participate in a statewide retirement plan, the Teacher Retirement System of Texas (http://www.trs.state.tx.us), so pension vesting does not pose a drag on mobility.

(20) STA was not awarded after the 1998-1999 school year. Instead, teacher minimum salaries are now tied to overall state aid. See Clark (2001).

(21) Note that the coefficients in Table 9 estimate two-year elasticities; whereas, the coefficients in Tables 7 and 8 estimate an average of one-year and two-year elasticities. Thus, it is not surprising that the coefficients in Table 9 are larger than those in Tables 7 and 8.

(22) Because South Carolina school districts are substantially larger than Texas school districts, one might argue that the evidence for South Carolina may be more relevant to the traditional monopsony model.

(23) This framework is used by Dickens, Machin, and Manning (1999) in their analysis of British Wages Councils.

Table 1. Descriptive Statistics of Raw Data for South Carolina School Districts Standard Variable Mean Deviation Minimum Teachers (full-time equivalent) 416 514 33 Salary, bachelor's degree, no experience $20,120 $2,143 $16,062 Salary, master's degree, no experience $23,020 $2,453 $18,391 Salary, bachelor's degree, 10 years' experience $26,482 $2,827 $21,202 Salary, master's degree, 17 years' experience $33,183 $3,549 $26,583 Salary gap 0.90% 1.53% 0.00% Enrollment 7056 8702 467 Total assessed value per pupil $11,548 $8,381 $2,947 Number of observations = 1163 Variable Maximum Median Teachers (full-time equivalent) 3728 227 Salary, bachelor's degree, no experience $25,277 $19,985 Salary, master's degree, no experience $28,944 $22,860 Salary, bachelor's degree, 10 years' experience $33,366 $26,276 Salary, master's degree, 17 years' experience $41,835 $32,879 Salary gap 6.41% 0.00% Enrollment 59,135 3827 Total assessed value per pupil $90,421 $9,690 Number of observations = 1163 Includes a maximum of 91 school districts from the years 1986-1987 through 1998-1999. Some observations were lost because of mergers of school districts. "Salary gap" is defined as the maximum of zero and the percent difference between this year's state minimum salary and last year's actual salary for a teacher with a bachelor's degree and zero teaching experience. Source: South Carolina Department of Education (1985-1999), Rankings of the Counties and School Districts of South Carolina, various issues. Table 2. Descriptive Statistics for South Carolina School Districts by Metropolitan/Nonmetropolitan Location Standard Variable Mean Deviation Minimum School districts in metropolitan statistical areas (number of observations = 516) Teachers (full-time equivalent) 630 678 66 Salary, bachelor's degree, no experience $20,484 $2,192 $16,062 Salary, master's degree, no experience $23,463 $2,507 $18,391 Salary, bachelor's degree, 10 years' experience $26,985 $2,893 $21,202 Salary, master's degree, 17 years' experience $33,865 $3,624 $26,583 Salary gap 0.47% 1.13% 0.00% Enrollment 10,801 11,455 1,151 Total assessed value per pupil $13,671 $10,293 $4,312 School districts outside metropolitan statistical areas (number of observations = 647) Teachers (full-time equivalent) 245 205 33 Salary, bachelor's degree, no experience $19,830 $2,059 $16,062 Salary, master's degree, no experience $22,667 $2,351 $18,391 Salary, bachelor's degree, 10 years' experience $26,081 $2,709 $21,202 Salary, master's degree, 17 years' experience $32,638 $3,393 $26,583 Salary gap 1.24% 1.71% 0.00% Enrollment 4069 3385 467 Total assessed value per pupil $9,856 $5,952 $2,947 Variable Maximum Median School districts in metropolitan statistical areas (number of observations = 516) Teachers (full-time equivalent) 3728 402 Salary, bachelor's degree, no experience $24,894 $20,464 Salary, master's degree, no experience $28,505 $23,432 Salary, bachelor's degree, 10 years' experience $32,861 $26,995 Salary, master's degree, 17 years' experience $41,201 $33,867 Salary gap 6.41% 0.00% Enrollment 59,135 7,278 Total assessed value per pupil $90,421 $11,439 School districts outside metropolitan statistical areas (number of observations = 647) Teachers (full-time equivalent) 1019 172 Salary, bachelor's degree, no experience $25,277 $19,720 Salary, master's degree, no experience $28,944 $22,531 Salary, bachelor's degree, 10 years' experience $33,366 $25,905 Salary, master's degree, 17 years' experience $41,835 $32,383 Salary gap 6.41% 0.00% Enrollment 16597 2918 Total assessed value per pupil $46,103 $8,042 Includes school districts from the years 1986-1987 through 1998-1999. "Salary gap" is defined as the maximum of zero and the percent difference between this year's state minimum salary and last year's actual salary for a teacher with a bachelor's degree and zero teaching experience. Source: South Carolina Department of Education (1985-1999), Rankings of the Counties and School Districts of South Carolina, various issues. Table 3. Ordinary Least Squares Results for All South Carolina School Districts Regressor 1 2 Salary gap -0.3674 -0.2566 (0.0778) (0.0701) Change in log enrollment 0.4633 (0.0567) Change in log total assessed 0.0166 value per pupil (0.0079) Fixed effects for districts? No No Adjusted [R.sup.2] 0.1241 0.2074 Number of districts 91 91 Number of observations 1163 1163 F-statistic for district fixed effects p-value Regressor 3 4 Salary gap -0.1224 -0.1640 (0.1116) (0.1138) Change in log enrollment 0.2349 (0.0687) Change in log total assessed 0.0121 value per pupil (0.0082) Fixed effects for districts? Yes Yes Adjusted [R.sup.2] 0.2010 0.2157 Number of districts 91 91 Number of observations 1163 1163 F-statistic for district fixed effects 2.2280 1.1339 p-value 2.95E-09 0.1931 Dependent variable is change in log full-time equivalent teacher employment. Standard errors (in parentheses) are computed using formulas robust to heteroskedasticity and second-order serial correlation. Year dummies included for all years (coefficient estimates are not shown to save space). Table 4. Ordinary Least Squares Results for South Carolina School Districts by Metropolitan/Nonmetropolitan Location Regressor 1 2 3 4 Sample definition Metro Metro Metro Metro Salary gap -0.4186 -0.2278 0.0834 0.0729 (0.1328) (0.1241) (0.1561) (0.1590) Change in log enrollment 0.5940 0.2710 (0.0740) (0.0815) Change in log total 0.0368 0.0296 assessed value per pupil (0.0144) (0.0156) Fixed effects for No No Yes Yes districts? Adjusted [R.sup.2] 0.1606 0.3213 0.3485 0.3732 Number of districts 40 40 40 40 Number of observations 516 516 516 516 F-statistic for district 4.7124 2.0613 fixed effects p-value 1.02E-16 0.0003 Regressor 5 6 7 8 Sample definition Nonmetro Nonmetro Nonmetro Nonmetro Salary gap -0.2050 -0.1708 -0.2044 -0.2316 (0.0970) (0.0913) (0.1671) (0.1665) Change in log enrollment 0.3315 0.1968 (0.0827) (0.1003) Change in log total 0.0006 0.0011 assessed value per pupil (0.0131) (0.0126) Fixed effects for No No Yes Yes districts? Adjusted [R.sup.2] 0.1005 0.1357 0.1018 0.1097 Number of districts 51 51 51 51 Number of observations 647 647 647 647 F-statistic for district 1.0177 0.6311 fixed effects p-value 0.4432 0.9778 Dependent variable is change in log full-time equivalent teacher employment. Standard errors (in parentheses) are computed using formulas robust to heteroskedasticity and second- order serial correlation. Year dummies included for all years (coefficient estimates are not shown to save space). Table 5. Descriptive Statistics for Texas School Districts Standard Variable Mean Deviation Minimum Teachers (full-time equivalent) 231 645 2 Average salary $31,181 $2,617 $18,932 Salary gap 1.65% 2.04% 0.00% Enrollment 3539 10,863 16 Total assessed value per pupil $217,530 $328,768 $9,801 Number of observations = 4127 Variable Maximum Median Teachers (full-time equivalent) 11,935 68 Average salary $49,187 $31,221 Salary gap 14.16% 0.78% Enrollment 210,988 868 Total assessed value per pupil $5,398,362 $136,559 Number of observations = 4127 Includes a maximum of 1033 school districts from the years 1995-1996 through 1998-1999. "Salary gap" is defined as salary transition aid per teacher, as a percent of the average of this year's average salary and last year's average salary. Source: Texas Education Agency (1994-1999), various issues. Table 6. Descriptive Statistics for Texas School Districts by Metropolitan/ Nonmetropolitan Status Standard Variable Mean Deviation Minimum School districts in metropolitan statistical areas (number of observations = 1606) Teachers (full-time equivalent) 456 979 5 Average salary $31,465 $2,440 $22,470 Salary gap 1.14% 1.63% 0.00% Enrollment 7199 16,565 58 Total assessed value per pupil $168,467 $151,857 $9,801 School districts outside metropolitan statistical areas (number of observations = 2521) Teachers (full-time equivalent) 87 134 2 Average salary $31,000 $2,709 $18,932 Salary gap 1.98% 2.20% 0.00% Enrollment 1206 2097 16 Total assessed value per pupil $248,810 $399,768 $19,827 Variable Maximum Median School districts in metropolitan statistical areas (number of observations = 1606) Teachers (full-time equivalent) 11,935 150 Average salary $41,319 $31,494 Salary gap 12.77% 0.36% Enrollment 210,988 2247 Total assessed value per pupil $2,737,836 $131,176 School districts outside metropolitan statistical areas (number of observations = 2521) Teachers (full-time equivalent) 1427 46 Average salary $49,187 $31,020 Salary gap 14.16% 1.25% Enrollment 21,830 549 Total assessed value per pupil $5,398,362 $141,454 Includes school districts from the years 1995-1996 through 1998-1999. "Salary gap" is defined as salary transition aid per teacher, as a percent of the average of this year's average salary and last year's average salary. Source: Texas Education Agency (1994-1999), various issues. Table 7. Ordinary Least Squares Results for All Texas School Districts Regressor 1 2 Salary gap -0.4153 -0.3098 (0.0841) (0.0778) Change in log enrollment 0.4693 (0.0605) Change in log total assessed value per pupil 0.0864 (0.0165) Fixed effects for districts? No No Adjusted [R.sup.2] 0.0287 0.1807 Number of districts 1033 1033 Number of observations 4127 4127 F-statistic for district fixed effects p-value Regressor 3 4 Salary gap -0.4410 -0.4095 (0.1381) (0.1394) Change in log enrollment 0.2713 (0.0849) Change in log total assessed value per pupil 0.0696 (0.0235) Fixed effects for districts? Yes Yes Adjusted [R.sup.2] 0.3206 0.3511 Number of districts 1033 1033 Number of observations 4127 4127 F-statistic for district fixed effects 2.7159 2.0483 p-value 1.71E-98 2.93E-50 Dependent variable is change in log full-time equivalent teacher employment. Standard errors (in parentheses) are computed using formulas robust to heteroskedasticity and second-order serial correlation. Year dummies included for all years (coefficient estimates not shown to save space). Data for the years 1995-1996 and 1997-1998 are one-year differences, while data for the years 1996-1997 and 1998-1999 are two-year differences. Table 8. Ordinary Least Squares Results for Texas School Districts by Metropolitan/Nonmetropolitan Location Regressor 1 2 3 Sample definition Metro Metro Metro Salary gap -0.2985 -0.2293 -0.3939 (0.1475) (0.1478) (0.1405) Change in log enrollment 0.6266 (0.0506) Change in log total assessed 0.0941 value per pupil (0.0263) Fixed effects for districts? No No Yes Adjusted [R.sup.2] 0.0635 0.3357 0.4590 Number of districts 402 402 402 Number of observations 1606 1606 1606 F-statistic for district 3.9186 fixed effects p-value 2.8E-74 Regressor 4 5 6 Sample definition Metro Nonmetro Nonmetro Salary gap -0.3459 -0.2377 -0.2419 (0.1485) (0.1086) (0.1018) Change in log enrollment 0.2662 0.3760 (0.0613) (0.0884) Change in log total assessed 0.0610 0.0614 value per pupil (0.0250) (0.0216) Fixed effects for districts? Yes No No Adjusted [R.sup.2] 0.4812 0.0120 0.1079 Number of districts 402 631 631 Number of observations 1606 2521 2521 F-statistic for district 2.1188 fixed effects p-value 1.19E-22 Regressor 7 8 Sample definition Nonmetro Nonmetro Salary gap -0.3887 -0.3938 (0.1913) (0.1895) Change in log enrollment 0.2634 (0.1098) Change in log total assessed 0.0687 value per pupil (0.0321) Fixed effects for districts? Yes Yes Adjusted [R.sup.2] 0.2467 0.2778 Number of districts 631 631 Number of observations 2521 2521 F-statistic for district 2.2441 1.9389 fixed effects p-value 7.09E-40 5.73E-27 Dependent variable is change in log full-time equivalent teacher employment. Standard errors (in parentheses) are computed using formulas robust to heteroskedasticity and second order serial correlation. Year dummies included for all years (coefficient estimates not shown to save space). Data for the years 1995-1996 and 1997-1998 are one-year differences, while data for the years 1996-1997 and 1998-1999 are two-year differences. Table 9. Ordinary Least Squares Results for All Texas School Districts: Two-Year Differences Only Regressor 1 2 3 Sample definition All districts All districts Metro Salary gap -0.5331 -0.5266 -0.3826 (0.1519) (0.1524) (0.2142) Change in log enrollment 0.2277 (0.0950) Change in log total 0.0666 assessed value per pupil (0.0308) Fixed effects for districts? Yes Yes Yes Adjusted [R.sup.2] 0.1364 0.1616 0.3548 Number of districts 1032 1032 402 Number of observations 2063 2063 803 Regressor 4 5 6 Sample definition Metro Nonmetro Nonmetro Salary gap -0.3759 -0.6058 -0.6030 (0.2291) (0.1927) (0.1909) Change in log enrollment 0.1500 0.2505 (0.0788) (0.1182) Change in log total 0.0456 0.0785 assessed value per pupil (0.0304) (0.0398) Fixed effects for districts? Yes Yes Yes Adjusted [R.sup.2] 0.3600 -0.0069 0.0279 Number of districts 402 630 630 Number of observations 803 1260 1260 Dependent variable is change in log full-time equivalent teacher employment. Standard errors (in parentheses) are computed using formulas robust to heteroskedasticity. Year dummy included (coefficient estimate not shown to save space). Data for the years 1996-1997 and 1998-1999 only: two-year differences.

Printer friendly Cite/link Email Feedback | |

Comment: | The effect of minimum salaries on employment of teachers: a test of the monopsony model. |
---|---|

Author: | Boal, William M. |

Publication: | Southern Economic Journal |

Geographic Code: | 1USA |

Date: | Jan 1, 2009 |

Words: | 12114 |

Previous Article: | Estimating distributions of willingness to pay for heterogeneous populations. |

Next Article: | Two-sector growth models with productive public goods: equilibrium (in)determinacy. |

Topics: |