64 M.L. Walden, Z. Sogutlu Economics of Education Review 20 2001 63–70
Third, there are important policy implications for local public decision-makers from understanding the determi-
nants of teacher salaries at a local district level. Under- standing the characteristics that affect local teacher salar-
ies will help school administrators and school boards deploy resources more efficiently. For example, if it is
found that larger schools are considered a disamenity by teachers requiring a salary premium, then school districts
may want to account for this finding when constructing new schools.
This paper analyzes intrastate variation in teacher sal- aries in one state, North Carolina. We are interested in
the individual and collective influence of local market factors, such as the local cost-of-living, personal and job
characteristics, and demand factors, on teacher salary. A particular challenge in the study is the selection and test-
ing of alternative cost-of-living indicators.
2. Model
We use a model similar to the one by Walden and Newmark 1995 to analyze interstate variation in
teacher salaries. Salary determinants are categorized into personal characteristics, job characteristics, factors meas-
uring the demand for education, and the cost-of-living.
2.1. Personal characteristics Among personal characteristics, previous research has
clearly established that average teacher salaries are related to teacher education and teacher experience
DeTray Greenberg, 1977; Eberts Stone 1984, 1985; Walden Newmark, 1995; Wentzler, 1981. In
fact, in several states, a teacher’s salary is automatically scaled to hisher educational level and years of teaching.
Indeed, this is the case in North Carolina. A state- wide schedule sets a teacher’s base salary related to that
teacher’s education academic degree and years of teaching experience. However, a local school district can
augment the base salary if it puts a higher value on teacher education and experience than does the state.
Traditionally, teaching has been a female-dominated profession. Teaching was often the secondary job in the
household. The household’s residential location would be determined by the primary jobholder, and if the wife
wanted to work, teaching would be a logical choice. In this type of economy, a larger proportion of females in
a local market would imply a larger relative supply of teachers and, ceteris paribus, lower salaries. In contrast,
as non-teaching employment opportunities have opened for women, any link between the female labor supply
and teacher salaries could have dissipated in local mar- kets. To test these effects, we include the female percent-
age of the local population in the analysis.
Other studies have found significant average salary differences between elementary and secondary school
teachers. For example, the intrastate studies of Eberts and Stone 1984 and DeTray and Greenberg 1977
found higher salaries for secondary teachers than for elementary teachers after controlling for education and
experience. This result could occur for a number of reasons. Secondary school teachers may systematically
have more positive skills and attributes unrelated to edu- cation and experience. Or, the job requirements of sec-
ondary school teachers may be more demanding, and thus the higher salary is a compensating differential.
Also, secondary school teachers may have greater alter- native job opportunities because their skills are closer to
those demanded in other jobs than for primary school teachers. Hence, schools must pay more to secondary
school teachers to attract and retain them. Last, local school districts may place a higher value on secondary
education, and this is reflected in a higher demand and equilibrium salary for secondary school teachers.
1
2.2. Job characteristics Four variables were available to measure job charac-
teristics of teachers in local districts. The first is the pupil–teacher ratio. A higher number of pupils per
teacher should mean a greater teacher workload. Every- thing else equal, a higher pupil–teacher ratio should shift
the teacher supply curve up and result in higher teacher salaries. Conversely, districts with a greater number of
teacher assistants per teacher should be associated with a lower teacher workload and lower teacher salaries.
Another job characteristic we include is average school size in the local market. As explained in Walden
and Newmark 1995, larger schools could be positively or negatively related to teacher salaries. On the one hand,
larger schools may provide a more pleasant work environment for teachers by having the ability to offer
teacher amenities like guidance counselors, librarians, and other support staff. In this case, larger school size
would be a positive job characteristic and would be asso- ciated with lower teacher salaries. The opposite argu-
ment is that larger schools present diseconomies of scale for teachers, such as more administrative work and rules.
Here, larger sized schools would be a disamenity for tea- chers, thereby shifting the supply curve up and resulting
in higher average salaries.
The last job characteristic considered is the ability and behavior of students. This factor has long presented a
challenge for researchers. Clearly, a teacher’s job is eas- ier if students are motivated and well- behaved. Thus, it
would be expected that local school markets with better-
1
Although classified as a personal characteristic, the dis- cussion indicates that the proportion of secondary teachers
could also be termed a job characteristic.
65 M.L. Walden, Z. Sogutlu Economics of Education Review 20 2001 63–70
motivated and better-behaved students would be pre- ferred by teachers and, in compensation, those districts
could offer lower average salaries. The problem is finding an adequate measure for these
student characteristics. We use a direct measure of stud- ent performance, the standardized elementary school
reading test score, as a proxy for student ability. Other measures were considered, such as the Scholastic Apti-
tude Test SAT score and the graduation rate. Fortu- nately, as will be discussed in the Results section, the
performance of each of these measure was very similar in the analysis.
2
2.3. Demand for teachers The demand for teachers in a local market should also
be related to the locality’s financial ability to pay for education Eberts Stone, 1984; Cohn Geske, 1990.
Two variables are included to measure this ability, the local governmental body’s local tax revenue per 1000
of personal income, and the local market’s nominal income per capita.
3
The tax revenue measure will be related to the locality’s wealth. Both variables are
expected to shift the demand curve for teachers to the right and increase teacher salaries.
2.4. Local cost-of-living Researchers studying interstate variation in teacher
salaries have found a positive relationship to the state’s cost-of-living Fournier Rasmussen, 1986; Nelson,
1991; Walden Newmark, 1995. The same positive relationship would be expected in a study of intrastate
variation in salaries. However, the existing studies of intrastate variation have not explicitly included a cost-of-
living variable Chambers, 1978; DeTray Greenberg, 1977; Eberts Stone, 1984. One reason may be the
difficulty of constructing such a measure at the local level.
One of the goals of our study was to develop local cost-of-living measures and test their relationships to
local teacher salaries. Two alternative measures were used. The first measure is described in detail in Walden
1998. The basis of the measure is quarterly retail price data available from the American Chamber of Com-
merce Researcher’s Association, or ACCRA. Each quar-
2
The Walden and Newmark study 1995 of interstate salary variation included a direct survey measure of work hours
reported by teachers as a job characteristic. Unfortunately, no similar survey information was available for the school districts
in North Carolina.
3
The Walden and Newmark study 1995 used nominal dis- posable income per capita as the income variable, but dispos-
able income was not available at the county level.
ter ACCRA collects price data for 61 retail products and services in 289 locations across the country, including
20 sites in North Carolina. ACCRA forms a cost-of-liv- ing index by taking a weighted average of the product
and service prices.
4
In order to form cost-of-living indices for local mar- kets in North Carolina, such as counties, the ACCRA
cost-of-living indices for the 20 North Carolina locations were regressed on exogenous variables that should deter-
mine the local cost-of-living, including population growth, population density, variables measuring the
spending power of consumers, and public sector tax and output variables.
5
The regression coefficients from this procedure were then applied to values for the exogenous
variables in all North Carolina local markets to construct cost-of-living indices.
However, there are two potential problems with this ACCRA-based local cost-of-living measure. First, it is
an estimate and not a direct measure. Second, ACCRA has been criticized for not always using professional staff
trained in data collection, which could lead to biased price measures.
For these reasons, a second local cost-of-living meas- ure was used, the county’s ‘fair market rent’. This meas-
ure is available annually from the U.S. Department of Housing and Urban Development. The fair market rent
is the average monthly rent, including utilities, for priv- ate rental units of a certain size and with modest and
suitable amenities in the county. Hence, the fair market rent is a quality-adjusted rent. Rents for two-bedroom
units were used.
6
The justification for using the fair mar- ket rent as a measure of the local cost-of-living is two-
fold. First, housing is the largest component of most household’s costs. Second, the geographical variation in
housing costs is much greater than the geographical vari- ation in other goods and services.
7
2.5. Summary In summary, the model of intrastate variation in
teacher salaries is given in the following equation, all variables are measured at the local market level:
4
Technically the weighted index is a price index, since a cost-of-living index requires that consumer utility be held con-
stant. However, for conformity to the terminology used in other studies, the index is called a cost-of-living index.
5
The regression explained 50 of the variation in the cost- of-living index, with all but one of the regressors having a stat-
istically significant parameter estimate Walden, 1998.
6
Fair market rent data were taken from the U.S. Department of Housing and Urban Development web site .
7
For example, the 1993 ACCRA data for North Carolina show a 46 difference between the location with the highest
housing cost index and the location with the lowest housing cost index, compared with an 18 difference for other goods
and services.
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