Skip to main content

Table 2 Estimates for public education spending per (FTE) student

From: An empirical inquiry into the determinants of public education spending in Europe

SUR estimates for the model specifications:

Extended & GDP per capita

Extended & GDP per capita

-adding general gov’t expenditure-

Specification

(4)

(7)

Δlog (spending per student, FTE, at ISCED 0–1) [eq1]

  

 Δlog (ULC)

1.03***

1.04***

(5.22)

(5.29)

 Δlog (real GDP per capita)

0.64***

0.65***

(3.49)

(3.57)

 Δlog (pupil/teacher ratio) (−1)

−0.12

−0.03

(−0.92)

(−0.22)

 Δlog (share of capital expenditure) (−1)

0.05

0.05

(1. 35)

(1.57)

 Δlog (share of teachers’ wages expenditure) (−1)

0.46**

0.37**

(2.36)

(2.00)

 Constant

3.26***

3.29***

(3.89)

(3.73)

Δlog (spending per student, FTE, at ISCED 2–4) [eq2]

  

 Δlog (ULC)

0.64***

0.65***

(2.97)

(3.01)

 Δlog (real GDP per capita)

0.60***

0.61***

(3.02)

(3.06)

 Δlog (pupil/teacher ratio) (−1)

−0.29**

−0.24*

(−2.07)

(−1.73)

 Δlog (share of capital expenditure) (−1)

0.03

0.03

(0.72)

(0.72)

 Δlog (share of teachers’ wages expenditure) (−1)

0.48**

0.44**

(2.29)

(2.09)

 Constant

3.03***

3.05***

(3.32)

(3.34)

Δlog (spending per student, FTE, at ISCED 5–6) [eq3]

  

 Δlog (ULC)

0.82***

0.82***

(3.88)

(3.88)

 Δlog (real GDP per capita)

0.49**

0.49**

(2.48)

(2.48)

 Δlog (share of capital expenditure) (−1)

0.06*

0.06*

(1.65)

(1.65)

 Δlog (share of teachers’ wages expenditure) (−1)

0.46**

0.47**

(2.22)

(2.23)

 Constant

0.90

0.90

(1.00)

(1.00)

Δlog (general govt. spending per capita) [eq4]

  

 Δlog (ULC)

 

0.70***

 

(6.96)

 Δlog (real GDP per capita)

 

0.47***

 

(5.05)

 Constant

 

2.31***

 

(5.36)

Observations

206

206

R 2 [eq1]

0.22

0.22

R 2 [eq2]

0.14

0.14

R 2 [eq3]

0.15

0.15

R 2 [eq4]

 

0.27

AIC

4556.55

5774.34

BIC

4613.12

5840.90

Breusch-Pagan test of independence stat. (p-val.)

21.27 (0.00)

38.23 (0.00)

  1. Note: The R 2 is just a summary measure of the overall in-sample predictive power of the model; it is simply computed as 1-residual sum of squares/total sum of squared residuals. The null of the Breusch-Pagan tests is that the residuals across the estimated equations in the model are independent
  2. Standard errors are adjusted for small sample sizes. t statistics are given in parentheses, * p < 0.10, ** p < 0.05, *** p < 0.01