4804
Views
875
Downloads |
Composition of public investment and economic growth: evidence from Turkish provinces, 1975-2001
Gökçen Yilmaz* 
Article | Year: 2018 | Pages: 187 - 214 | Volume: 42 | Issue: 2 Received: November 11, 2017 | Accepted: February 12, 2018 | Published online: June 5, 2018
|
FULL ARTICLE
FIGURES & DATA
REFERENCES
CROSSMARK POLICY
METRICS
LICENCING
PDF
|
Variable
|
|
Mean
|
Std. Dev.
|
Min.
|
Max.
|
Obs. ɫ
|
|
γ
|
overall
|
0.018
|
0.032
|
-0.091
|
0.178
|
N=1474
|
|
between
|
|
0.012
|
-0.019
|
0.049
|
n=67
|
|
within
|
|
0.030
|
-0.080
|
0.161
|
T=22
|
|
θt&c
|
overall
|
0.148
|
0.161
|
0.000
|
0.929
|
N=1474
|
|
between
|
|
0.086
|
0.022
|
0.427
|
n=67
|
|
within
|
|
0.137
|
-0.183
|
0.874
|
T=22
|
|
θen
|
overall
|
0.267
|
0.278
|
0.000
|
0.987
|
N=1474
|
|
between
|
|
0.200
|
0.044
|
0.830
|
n=67
|
|
within
|
|
0.195
|
-0.563
|
1.005
|
T=22
|
|
θed
|
overall
|
0.216
|
0.144
|
0.000
|
0.887
|
N=1474
|
|
between
|
|
0.078
|
0.023
|
0.357
|
n=67
|
|
within
|
|
0.122
|
-0.077
|
0.790
|
T=22
|
|
θc&s
|
overall
|
0.300
|
0.184
|
0.004
|
0.915
|
N=1474
|
|
between
|
|
0.109
|
0.068
|
0.511
|
n=67
|
|
within
|
|
0.148
|
-0.139
|
1.072
|
T=22
|
|
θhe
|
overall
|
0.069
|
0.085
|
0.000
|
0.891
|
N=1474
|
|
between
|
|
0.044
|
0.011
|
0.259
|
n=67
|
|
within
|
|
0.073
|
-0.147
|
0.867
|
T=22
|
|
g/y
|
overall
|
0.037
|
0.063
|
0.002
|
0.873
|
N=1474
|
|
between
|
|
0.040
|
0.009
|
0.204
|
n=67
|
|
within
|
|
0.048
|
-0.159
|
0.735
|
T=22
|
|
η
|
overall
|
0.016
|
0.015
|
-0.035
|
0.101
|
N=1474
|
|
between
|
|
0.012
|
-0.019
|
0.046
|
n=67
|
|
within
|
|
0.008
|
-0.026
|
0.071
|
T=22
|
|
k/y
|
overall
|
0.013
|
0.027
|
-0.005
|
0.377
|
N=1474
|
|
between
|
|
0.019
|
0.000
|
0.092
|
n=67
|
|
within
|
|
0.019
|
-0.078
|
0.371
|
T=22
|
† The summary statistics are expressed in decimal numbers. Thus, “0.018” should be read as “1.8%”. ɫ Obs.: The number of observations; N: the number of observations in the sample; n: the number of panels (provinces) in the sample; T: the number of time periods (years) in the sample.
|
γ
|
θen
|
θt&c
|
θed
|
θhe
|
θc&s
|
g/y
|
η
|
k/y
|
|
γ
|
1.000
|
|
|
|
|
|
|
|
|
|
θen
|
0.118
|
1.000
|
|
|
|
|
|
|
|
|
θt&c
|
-0.206
|
-0.381
|
1.000
|
|
|
|
|
|
|
|
θed
|
-0.099
|
-0.584
|
-0.061
|
1.000
|
|
|
|
|
|
|
θhe
|
0.014
|
-0.291
|
-0.093
|
0.108
|
1.000
|
|
|
|
|
|
θc&s
|
0.073
|
-0.585
|
-0.209
|
0.103
|
-0.028
|
1.000
|
|
|
|
|
g/y
|
0.086
|
0.510
|
-0.094
|
-0.310
|
-0.170
|
-0.367
|
1.000
|
|
|
|
η
|
-0.024
|
0.095
|
-0.009
|
-0.149
|
-0.020
|
-0.010
|
0.061
|
1.000
|
|
|
k/y
|
0.116
|
-0.077
|
-0.004
|
-0.006
|
0.022
|
0.115
|
-0.149
|
0.166
|
1.000
|
|
|
(1)
|
(2)
|
(3)
|
(4)
|
(5)
|
(6)
|
|
θen
|
|
0.004
|
|
|
|
|
|
|
(0.006)
|
|
|
|
|
|
θt&c
|
|
|
-0.016†
|
|
|
|
|
|
|
(0.007)*
|
|
|
|
|
θed
|
|
|
|
-0.004
|
|
|
|
|
|
|
(0.010)
|
|
|
|
θhe
|
|
|
|
|
0.011
|
|
|
|
|
|
|
(0.016)
|
|
|
θc&s
|
|
|
|
|
|
0.006
|
|
|
|
|
|
|
(0.008)
|
|
g/y
|
0.089
|
0.080
|
0.084
|
0.085
|
0.091
|
0.092
|
|
(0.016)**
|
(0.020)**
|
(0.016)**
|
(0.017)**
|
(0.017)**
|
(0.018)**
|
|
η
|
-0.082
|
-0.083
|
-0.088
|
-0.083
|
-0.085
|
-0.084
|
|
(0.129)
|
(0.127)
|
(0.127)
|
(0.125)
|
(0.129)
|
(0.129)
|
|
k/y
|
0.102
|
0.105
|
0.105
|
0.106
|
0.103
|
0.100
|
|
(0.032)**
|
(0.033)**
|
(0.032)**
|
(0.033)**
|
(0.032)**
|
(0.032)**
|
|
Year
Dummies (Tj)††
|
|
|
|
|
|
|
|
Constant
|
0.004
|
0.004
|
0.006
|
0.005
|
0.003
|
0.002
|
|
(0.003)
|
(0.003)
|
(0.003)
|
(0.005)
|
(0.004)
|
(0.005)
|
|
Observations
|
1474
|
1474
|
1474
|
1474
|
1474
|
1474
|
|
Number of
panels
|
67
|
67
|
67
|
67
|
67
|
67
|
|
Wald
|
771.65
|
832.68
|
829.83
|
773.86
|
829.34
|
797.86
|
|
Prob>
Wald χ2
|
0.0000
|
0.0000
|
0.0000
|
0.0000
|
0.0000
|
0.0000
|
|
R2
|
0.17
|
0.17
|
0.18
|
0.17
|
0.17
|
0.17
|
|
Mean VIF†††
|
1.15
|
1.29
|
1.24
|
1.33
|
1.20
|
1.40
|
Standard errors in parentheses, * significant at 5%; ** significant at 1%. † The coefficients show the effect of a one-unit change in the value of an indicator on the dependent variable. The values of the variables are expressed in decimal numbers in table 1. This means that a unit change in table 3 corresponds to a 100% change in the shares of public investment. ††The results for year dummies are not reported in the table for conciseness. ††† VIF: Variance Inflation Factor.
|
|
(1)
|
(2)
|
(3)
|
(4)
|
(5)
|
(6)
|
|
θen
|
|
0.009†
|
|
|
|
|
|
|
(0.004)*
|
|
|
|
|
|
θt&c
|
|
|
-0.019
|
|
|
|
|
|
|
(0.008)*
|
|
|
|
|
θed
|
|
|
|
-0.015
|
|
|
|
|
|
|
(0.011)
|
|
|
|
θhe
|
|
|
|
|
0.013
|
|
|
|
|
|
|
(0.010)
|
|
|
θc&s
|
|
|
|
|
|
0.005
|
|
|
|
|
|
|
(0.006)
|
|
g/y
|
0.074
|
0.054
|
0.068
|
0.064
|
0.077
|
0.079
|
|
(0.008)**
|
(0.008)**
|
(0.007)**
|
(0.007)**
|
(0.009)**
|
(0.013)**
|
|
η
|
-0.091
|
-0.093
|
-0.095
|
-0.104
|
-0.091
|
-0.095
|
|
(0.098)
|
(0.097)
|
(0.101)
|
(0.101)
|
(0.099)
|
(0.101)
|
|
k/y
|
0.167
|
0.165
|
0.166
|
0.163
|
0.168
|
0.166
|
|
(0.060)**
|
(0.057)**
|
(0.055)**
|
(0.057)**
|
(0.059)**
|
(0.060)**
|
|
Year
Dummies (Tj)††
|
|
|
|
|
|
|
|
Constant
|
0.003
|
0.003
|
0.006
|
0.009
|
0.002
|
0.001
|
|
(0.002)
|
(0.002)
|
(0.002)**
|
(0.003)**
|
(0.002)
|
(0.004)
|
|
Observations
|
1474
|
1474
|
1474
|
1474
|
1474
|
1474
|
|
Number of panels
|
67
|
67
|
67
|
67
|
67
|
67
|
|
F
|
35.98
|
26.60
|
41.45
|
30.92
|
32.05
|
55.26
|
|
Prob > F
|
0.0000
|
0.0000
|
0.0000
|
0.0000
|
0.0000
|
0.0000
|
|
R2
|
0.17
|
0.18
|
0.18
|
0.18
|
0.17
|
0.17
|
|
Mean VIF†††
|
1.15
|
1.29
|
1.24
|
1.33
|
1.20
|
1.40
|
† The coefficients show the effect of a one-unit change in the value of an indicator on the dependent variable. The values of the variables are expressed in decimal numbers in table 1. This means that a unit change in table 4 corresponds to a 100% change in the shares of public investment. ††The results for year dummies are not reported in the table for conciseness. ††† VIF: Variance Inflation Factor.
Table 1Descriptive statistics DISPLAY Table
Table 2Pairwise correlation matrix for the variables DISPLAY Table
Table 3Composition of public investment and economic growth: random-effects technique-standard errors corrected for heteroscedasticity and serial correlation DISPLAY Table
Table 4Composition of public investment and economic growth: pooled OLS technique-standard errors corrected for heteroscedasticity, and serial correlation both within and between panels DISPLAY Table
* The author would like to thank to two anonymous referees for useful comments and suggestions.
1 Public investment corresponds to public capital expenditure. This paper adopts the former term whenever it refers to its sample. This is because the State Planning Organisation (now, a section of the Ministry of Development) reports these data under the title “public investment”).
2 Note that, with the real GDP data for the years between 1975 and 2001, the dependent variable can be calculated for the years between 1975 and 1996.
4 The figure is provided as part of the post-estimation diagnostics on request.
5 The post-estimation diagnostics show the presence of serial correlation in residuals both within and between panels.
6 Deverajan, Swaroop and Zou (  1996) and Haque ( 2004) correct the standard errors using the methodology in Hansen and Hodrick ( 1980). This paper uses the built-in commands in the statistical software (Stata) used for the empirical analysis of the data. For panel data, the command “xtreg” offers the “robust” option which corrects the standard errors to heteroscedasticity and serial correlation, while the command “xtscc” allows for correcting the standard errors to heteroscedasticity and serial correlation between panels, and within panels up to a specific number of lags. In this paper, the standard errors obtained from the “xtscc” command are corrected for serial correlation within panels up to five lags, as the five-year forward-moving geometric average of per-worker real GDP growth rate introduces correlation to error terms between years t and t+5 (Devarajan, Swaroop and Zou, 1996). Note that, while the command “xtreg” offers the random-effects and fixed-effects techniques, the command “xtscc” offers pooled OLS and fixed-effects techniques as econometric methods. In accordance with the results of the Hausman test for model specification in the post-estimation diagnostics, Table 3 uses the random-effects technique, while Table 4 uses the pooled OLS technique.
7 Robustness of the results to alternative specification of the dependent variable is discussed in the end of the results section.
8 This, indeed, appears to be the case, as the results remain similar if one uses the between-effects estimator provided by the statistical software Stata.
9 The dataset has been examined for errors in data entry and the calculation of the dependent variable but neither of these appears to be the case.
10 Between tables 3 and 4, the statistical significance of the coefficient of the share of energy infrastructure in total public investment is sensitive to the treatment of residuals for heteroscedasticity, and serial correlation between and within panels. For this reason, the robustness analysis regarding alternative specifications of the dependent variable focuses on the share of transportation and communication in total public investment, which is the only public investment component that has a statistically significant coefficient in both Table 3 and Table 4.
11 All the results are available on request.
|
|
June, 2018
II/2018 |
