The division of labour between central and local governments is not as clear cut as outlined in our theoretical model. Central government tries to influence class size by rules for maximum class size and maximum travel distance to schools. On the other hand, local governments can have some influence on teacher wages by recruiting experienced teachers, promoting teachers into advisory positions and giving them administrative responsibilities. Teachers per class is realistically the result of both national bargaining over teacher working conditions and local priorities of teacher input, in particular related to handicapped students and low achievers. To check if our analysis captures the main local and central determinants of school spending growth, for each variable, we have tested whether a model with local or central explanatory variables performs best. We focus on the influence of income and politics. The encompassing tests are reported in Table 4.
Regarding teacher wages, we cannot reject the hypothesis that a model with national income level as determinant encompasses a model using the local income level, but the opposite hypothesis is rejected. The same result appears for teacher input per class, in accordance with our assumptions. Sorensen (1995) finds that some of the variation in teacher input per class between local governments can be explained with local explanatory variables. We agree that some local variation takes place, but our interpretation is that central decisions dominate the development over time. The ideological impact of socialist share of representatives is negligible, both when measured for national and local parliaments.
Turning to the two variables assumed to be determined at the local level, it seems like both local and national income influence non-teacher spending per student. Statistically GDP cannot be excluded from the equation determining non-teacher spending at 10% level. Local political orientation, and not national politics, is important, in particular for class size. All in all, the interactions between central and local governments in practice implies that both levels have some influence over teachers per class and non-teacher spending.
We have shown how central government, teacher unions and local governments have influenced the remarkable primary school spending growth during the post-World-War II period. The three agents have different roles in the decision-making process, and they all contribute to the expansion of school resources. At the national level, teacher unions struggle for higher teacher wages and reduced teaching load. They have been successful, and these two elements are the main cost factors identified. Future analysis must judge whether teachers have been more successful than other government employees. The central government tries to hold back on the claims of the teacher union. The bargaining strength of central government, both in terms of macro-economic situation and political structure, is shown to matter.
Interestingly, education come out as an income-elastic good in the present study. The income elasticity of primary school spending is above 1.5, implying strong support for Wagner's law. Education is an area cited by Wagner where he saw collective producers as more efficient than private ones. Our analysis suggests a solution to the puzzle that time-series analyses show income-elastic school spending, while cross-sectional analyses offer overwhelming evidence of income-inelastic educational services (see the overview of Inman, 1979). There is limited variation of teacher wages and teacher input per class across municipalities. Spending growth is driven by income-elastic decisions about teacher wages and working conditions at the national level.
Clearly, central government does not work primarily as an institution resisting school spending. Rather the other way around. All major primary school reforms during the period studied, in particular the 9-year compulsory school, were instigated by central government. It is seen too that local governments implement the reforms by channelling more resources to the local public sector. Local governments have been a source of school spending growth by non-adjustment to centrally determined cost factors--teacher wages and teaching load. The inelastic response to rising costs means that national cost expansion is translated directly to higher spending. The non-adjustment has been most pronounced regarding the demographic shifts. The declining share of students in the population during the 1980s has increased spending, since class size has been allowed to go down, thereby contributing to the dramatic rise in teacher-student ratio and per-student spending.
The good news is that political factors can control school spending growth. Strong central governments can hold back teacher wages and maintain teachers' teaching loads. Efficient local governments can adjust the school structure to demographic changes and try to substitute away from ever-more expensive teachers.
The project is funded by the Norwegian Research Council. We appreciate comments from the participants at the conference and in particular Eric Hanushek and Robert Inman. Our research also has benefited from comments made at the European Public Choice meeting in Valencia and the European Economic Association meeting in Maastricht.
1. The enrolment ratio for the age group 7-16 increased from 90% in 1946 to 97.5% in 1960 and 100% since 1971. It follows that the number of students partly increased due to a higher enrolment ratio before 1971.
2. Matching grants to health care are excluded from the analysis and the model because data have not been available.
3. The elasticity of total school spending (SS) is calculated as
ElSS is approximately equal to lambda (ElW + ElT/Cl -ElSt/Cl) + (1- lambda)ElSp/St
where El is a symbol of elasticity and lambda is the teacher spending share of total spending. lambda is calculated at post-1973 mean, where lambda=0.78. At the local level ElW=ElT/Ct = 0, and the income elasticity of total school spending is about 0.2.
4. The local government demand functions related to primary school (equations (3) and (4)) and health care inserted into (1) define the indirect utility function of the local government:
Multiple line equation(s) cannot be represented in ASCII text
where m is excluded from the model.
- 5. Matching grants have been excluded in this section. Our simplification of the decision-making situation has produced a problem: matching grants have no role in a model with one representative local government and where local and central preferences agree.
- 6. The outcome of the bargaining game between the central government and the teacher union depends on the effects of exogenous variables on the utility levels. We therefore use the indirect utility function of the central government. The demand functions for the public good B, the private good C and grants G are derived by maximizing the central government utility function X subject to the budget constraint and the local government indirect utility function. Together with the national government utility function, they define the indirect utility function of the national government.
- 7. A formal analysis of the model, and a discussion of the partial effects, is available from the authors.
- 8. In order to test for other comparison wages than manufacturing wages, we tested the effects of average wage measures from the national accounts, only available from 1962. It seems like teacher wages follow both the wage level of central government employees and GDP with an elasticity of 0.5 in this period, while it is a clear short-run effect of wages in private services. More thorough investigation of wage comparison effects is necessary before drawing definite conclusions.
- 9. This elasticity is calculated in accordance with Note 3. The precise elasticity depends on the effect of GDP upon local government income. Borge and Rattso (1994) calculate this to be 1.4, implying an income elasticity of total school spending of 1.9.
Table 1. School spending growth 1946-1990 (1990-NOK)
Legend for Chart:
A - Spending per student
B - 1946
C - 1955
D - 1965
E - 1975
F - 1985
G - 1990
B C D E F G
Teacher expenditure per student
3 198 4 113 7 766 17 238 23 652 28 311
Non-teacher expenditure per student
1 801 2 592 5 389 5 396 7 066 6 643
Legend for Chart:
A - Decomposition of total teacher expenditure growth
B - 1946-1965
C - 1965-1980
D - 1980-1990
E - 1946-1990
A B C D E
Teacher wage expenditure 48.2 64.7 27.3 53.7
Teacher-class ratio 31.2 29.2 200.0 37.8
Class size -17.2 -0.4 81.8 -6.4
Student enrolment 37.8 6.5 209.1 14.9
Sum 100.0 100.0 100.0 100.0
Table 2. Local government decision
Legend for Chart:
A - Dependent variable
B - Class size
C - Non-teacher expenditure per student
A B C
Local government income, LRE -- 1.00
School price, LRE -- 1.00
Student share, LRE 0.43 --
Student share, SRE 0.45 -2.98
Spending share, SRE -- -0.29
Share of elderly, SRE 0.48 --
Divorce rate, LRE -0.07 --
Centralization, LRE 0.14 --
Socialist share local parliaments, SRE 0.04 0.38
Social democratic hegemony, 1946-1961 0.01 --
Notes: Only significant effects reported. The calculations are based on regressions 3a and 3b in Table B2. LRE is long-run elasticity and SRE is short-run elasticity. Elasticity at mean for political variables. School price is WT*T/Cl and spending share is (1- m).
Table 3. Central government bargaining
Legend for Chart:
A - Dependent variable
B - Teacher wages
C - Teachers per class
A B C
GDP, LRE 1.00 1.00
Student share, LRE 0.44 --
Student share, SRE -1.99 -0.38
Inflation -0.66 --
Vacancy rate, LRE 0.09 --
Female share, SRE -0.35 0.39
Social democratic hegemony, 1946-1961 0.02 --