Results Directory UMM :Data Elmu:jurnal:E:Economics of Education Review:Vol20.Issue1.2001:

66 M.L. Walden, Z. Sogutlu Economics of Education Review 20 2001 63–70 SUPP5f[COL or FMR, EDU, EXP, FEM, SECOND, PUPTCH, ASST, SCHSIZE, READING, REV, PERCAPY]. The variable definitions are given in Table 1. The dependent variable, SUPP, the local teacher salary sup- plement, is explained more fully in the next section.

3. Data

The intrastate teacher salary model is implemented with data from North Carolina for the 1993–94 school year. The salary and other educational information are taken for public schools only. Private and home schools educate less than 6 of North Carolina pupils North Carolina Department of Public Instruction, Division of Non-Public Education, 1998. Local markets are defined as counties. There are 100 counties in North Carolina. There are 129 public school districts in the state. Two-thirds of these districts are sin- gle-county districts. No district crosses county lines. For multi-district counties, data for one district were formed by taking a pupil-weighted average of the multiple dis- tricts. Public school teachers in North Carolina are paid from two sources, the state and the local county government. The state pays a ‘base’ salary dependent on the teacher’s education and experience levels. The local county government pays a supplement to this base salary. The dependent variable in our study, SUPP, is the local county salary supplement. This means the impacts of teacher education and experience through the state base salary are removed. Teacher education and experience will influence SUPP only if local governments place additional value on these two characteristics. Teacher education EDU is measured as the percent- age of teachers in the district with a post-baccalaureate degree. Teacher experience is measured by the average years of teaching experience EXP. The variables SECOND, PUPTCH, ASST, SCHSIZE, and GRADRT are measured for the school district. The vari- ables FEM, REV, and PERCAPY are measured for the county. 8 Sources and descriptive statistics for the variables are given in Table 1. Note that the ACCRA-based cost-of- living index for North Carolina counties is a relative measure that has no intrinsic meaning. 8 Note that PERCAPY is measured in nominal terms, rather than deflated by the cost-of-living measure to be expressed in real terms. This is done for comparability to previous studies, where dollar amounts are in nominal terms.

4. Results

Following the procedure used by Walden and New- mark 1995, four specifications of the model are presented in order to ascertain the effects of different sets of variables. First, the teacher salary supplement is regressed only on the cost-of-living measure. Then, the supplement is regressed on the cost-of-living measure and the personal characteristics. In the third specifi- cation, job characteristics are added to the right-hand side variables and, in the fourth and full specification, the demand factors are added to the regressors. The results of the four regressions using the ACCRA- based cost-of-living measure are presented in Table 2. 9 In the first specification, the cost-of-living measure COL is positive and statistically significant. COL explains 14 of the variability in the salary supplement. This is a lower amount of explained variability than found in the Walden–Newmark study of interstate teacher salaries, where the cost-of-living variable alone accounted for almost two-thirds of the salary variation. Calculated at the mean, the elasticity of the supplement with respect to COL is 10.9, and the elasticity of the total teacher salary with respect to COL is 0.2. 10 COL is also positive and statistically significant in the second and third specifications. It is positive, but not stat- istically significant, in the fourth specification, but this is because the nominal per capita income variable likely captures variability in the local cost-of-living. Indeed, when we deflate the nominal per capita income variable by the cost-of-living measure to form real per capita income and include it in the fourth specification, both COL and the real per capita income variable are positive and statistically significant and the results for the other variables are not affected. The second specification adds the personal character- istics to the analysis. The education and experience vari- ables EDU, EXP are not statistically significant in this equation nor in the third and fourth specifications. These results could suggest teacher education and experience are accounted for in the state salary base, and local governments in North Carolina place no additional value on these factors. Alternatively, the results could mean that rightward shifts in the demand curve for better educated and experienced teachers occur simultaneously with compa- 9 Since COL is an estimate and thus measured with error, maximum likelihood estimates are presented. The variables are measured linearly. The equations were also estimated in log– log form, but the results were unchanged. The Breusch–Pagan– Godfrey test found no problems of heteroskedasticity, and the Kelsley–Kuh–Welsch test found no influential observations. 10 The average total teacher salary in North Carolina in 1993– 94 was 29,728. 67 M.L. Walden, Z. Sogutlu Economics of Education Review 20 2001 63–70 Table 1 Variables and descriptive statistics Variable Description Source Mean Standard Minimum Maximum deviation SUPP Local teacher salary supplement, North Carolina Department of Public Instruction, 733.9 754.3 3485.0 1993–94 Statistical Research and Data Center, 1996 COL County cost-of-living index, Walden, 1998 106.8 4.1 94.6 118.2 1993 FMR Fair market rent, 1993 U.S. Department of Housing and Urban 402.1 42.2 320.0 520.0 Development EDU Percentage of teachers with a North Carolina Department of Public Instruction, 34.5 7.3 20.2 54.1 post-baccalaureate Statistical Research and Data Center, 1996 EXP Average years of experience of North Carolina Department of Public Instruction, 12.0 1.2 8.6 15.0 teachers Statistical Research and Data Center, 1996 FEM Females as a percentage of the North Carolina Office of State Planning, 1996 51.7 1.6 41.0 54.6 population, 1993 SECOND Secondary school teachers as a North Carolina Department of Public Instruction, 29.3 8.2 14.9 50.2 percentage of all teachers, 1993 Statistical Research and Data Center, 1996 PUPTCH Pupil–teacher ratio North Carolina Department of Public Instruction, 15.8 1.6 7.7 18.3 Statistical Research and Data Center, 1996 ASST Instructional aides as a North Carolina Department of Public Instruction, 29.7 4.6 7.3 42.4 percentage of teachers, 1993–94 Statistical Research and Data Center, 1996 SCHSIZE Average school size in pupils, North Carolina Department of Public Instruction, 522.5 122.2 206.8 778.6 1993–94 Statistical Research and Data Center, 1996 READING Elementary school reading test North Carolina State Board of Education, 1994 62.8 7.7 42.9 77.7 score, 1993–94 REV Total country revenue per 1000 North Carolina Association of County 43.7 14.8 24.7 114.5 of country personal income, Commissioners, 1997 1993–94 PERCAPY Nominal county per capita U.S. Department of Commerce, Bureau of 16,564 2617.7 11,921 24,612 income, 1993 Economic Analysis, 1995 68 M.L. Walden, Z. Sogutlu Economics of Education Review 20 2001 63–70 Table 2 Determinants of the local teacher salary supplement in North Carolina, using COL as the cost-of-living measure a Equation 1 2 3 4 County cost-of-living index COL 71.6 75.2 51.3 15.7 4.1 4.2 2.5 0.9 Teacher education EDU 253.4 694.3 2121.2 0.3 0.8 20.2 Teacher experience EXP 2104.0 270.9 269.0 21.7 21.3 21.5 Females in population FEM 89.0 73.7 0.9 2.0 1.8 0.1 Secondary school teachers SECOND 29.1 28.0 17.9 3.6 3.9 2.9 Pupil–teacher ratio PUPTCH 212.1 26.3 20.3 20.2 Aides as percentage of teachers ASST 26.7 21.6 20.5 20.1 Elementary school reading test score READING 21.4 23.4 20.1 20.4 Average school size in pupils SCHSIZE 2.8 1.9 5.0 4.0 County revenue per 1000 personal income REV 2.6 0.7 Nominal per capita income PERCAPY 0.2 6.6 Adjusted R 2 0.14 0.30 0.46 0.63 a Notes: 1 dependent variable is local salary supplement, SUPP; 2 t-ratios are in parentheses; 3 the constant term in each specification is not reported to save space; 4 the number of observations is 100. rable shifts in the supply curve due to some factor corre- lated with education and experience. The result would leave no local premium for education and experience. 11 The percentage of females in the local market FEM, a proxy for the potential supply of teachers, is positive in equation 2 as well as in the third and fourth equations. This is contrary to the hypothesis that a greater supply of females will shift the teacher supply curve to the right and lower the equilibrium salary. How- ever, the t-ratio in equation 3 is not quite 2, and the t- ratio in equation 4 is far below 2. Thus, it can be con- cluded that in the modern economy, the supply of females in a local market has no statistically significant relationship to local teacher salaries. The percentage of secondary teachers SECOND is positive and statistically significant in all equations in which it is entered, with t-ratios well above 2. This is evidence that secondary teachers in North Carolina school districts receive salary premiums unrelated to their education and experience. Job characteristics are added in the third specification. 11 We are indebted to a reviewer for this interpretation. Among them, only average school size SCHSIZE has statistically significant coefficients in both equations 3 and 4. Teachers working in North Carolina schools with more pupils receive a pay premium, perhaps to compensate for some disamenities related to school size. Substituting the SAT score or graduation rate for the READING score yielded the same statistically insignifi- cant results for this factor. The fourth and full specification adds the two demand variables. Both the local public revenue REV and local nominal income per capita PERCAPY variables are positive, but only PERCAPY is statistically significant. The full specification accounts for almost two-thirds of the variation in the local teacher salary supplement. The results of the four regressions using the fair mar- ket rent FMR measure of the local cost-of-living are given in Table 3. 12 The results are virtually identical to those in Table 2 using the COL measure of the cost-of- living. FMR is positive and statistically significant in all 12 FMR can be considered to be an instrument for the imper- fectly measured COL. Maximum likelihood estimates are presented. 69 M.L. Walden, Z. Sogutlu Economics of Education Review 20 2001 63–70 Table 3 Determinants of the local teacher salary supplement in North Carolina, using FMR as the cost-of-living measure a Equation 1 2 3 4 Fair market rent FMR 7.3 6.4 3.0 20.3 4.5 4.0 1.9 20.2 Teacher education EDU 998.8 692.0 262.1 1.1 0.7 20.1 Teacher experience EXP 282.4 266.1 271.3 21.4 21.2 21.6 Females in population FEM 33.3 57.2 24.3 0.8 1.4 20.1 Secondary school teachers SECOND 29.7 27.2 16.8 3.6 3.7 2.7 Pupi–teacher ratio PUPTCH 23.6 23.6 20.1 20.1 Aides as percentage of teachers ASST 26.8 22.9 20.5 20.3 Elementary school reading test score READING 12.1 1.3 1.3 0.2 Average school size in pupils SCHSIZE 2.9 2.0 5.1 4.3 County revenue per 1000 personal income REV 2.6 0.7 Nominal per capita income PERCAPY 0.2 6.8 Adjusted R 2 0.16 0.29 0.44 0.63 a Notes: 1 dependent variable is local salary supplement, SUPP; 2 t-ratios are in parentheses; 3 the constant term in each specification is not reported to save space; 4 the number of observations is 100. but equation 4. In equation 1 and calculated at the mean, the elasticity of the supplement with respect to FMR is 5, and the elasticity of the total teacher salary with respect to FMR is 0.1. The teacher education and experience variables are not statistically significant. Average school size and nominal income per capita are positive and statistically significant in all equations in which they appear.

5. Summary and conclusion