Social and Economic Convergence Across Districts in Indonesia: A Spatial Econometric Approach

Figures & data

Table 1. Descriptive Statistics of the Growth and Initial Levels of HDI and GDP Per Capita, 2010 and 2018

Source: Authors’ calculations based on public data and official publications BPS.

Notes: The data set covers 514 districts and 4,626 observations over 2010–18. Data from the newly split-up districts have been recalculated and are provided in official BPS publications. Nevertheless, over the 2010–18 period, we found some missing observations. To overcome this issue, we imputed those observations by using a moving average method. The HDI is calculated as the geometric mean (equally weighted) of life expectancy, education and GNI per capita as follows. The education dimension is the arithmetic mean of the two education indices (mean years of schooling and expected years of schooling). HDI is measured as an index, scaled between 0 and 100. The highest possible level of human development is 100. SD = standard deviation.

FIGURE 1 Centroids of 514 Indonesian Districts and Associated Administrative Boundaries

Source: Authors.

FIGURE 2 Centroids of 514 Indonesian Districts and Associated Thiessen Polygon Boundaries

Source: Authors.

FIGURE 3 Global Spatial Dependence and Sigma Convergence in Social and Economic Performance
Source: Authors.

FIGURE 4 Local Indicators of Spatial Association of HDI in 2010

FIGURE 5 Local Indicators of Spatial Association of HDI in 2018

FIGURE 6 Local Indicators of Spatial Association of GDP Per Capita in 2010

FIGURE 7 Local Indicators of Spatial Association of GDP Per Capita in 2018

Table 2. Pesaran’s Test of Cross-Sectional Dependence

Notes: The null hypothesis is cross-sectional independence. *** p < 0.01, ** p < 0.05, * p < 0.1

Table 3. Spatial Dependence Results from the Absolute Convergence Models of HDI

Notes: Robust standard errors are in parentheses. The spatial connectivity structure (W, weight matrix) is based on the Thiessen polygon approach described earlier. Wg is the spatial lag of dependent variables; Wx is the spatial lag of independent variables.

*** p < 0.01, ** p < 0.05, * p < 0.1.

Table 4. Spatial Dependence Results from the Absolute Convergence Models of GDP Per Capita

Notes: Robust standard errors in parentheses. The spatial connectivity structure (W, weight matrix) is based on the Thiessen polygon approach described earlier. Wg is the spatial lag of the dependent variables; Wx is the spatial lag of the independent variables.

*** p < 0.01, ** p < 0.05, * p < 0.1.

Table 5. Spatial Dependence in the Conditional Convergence Models of HDI and Determinants

Notes: Robust standard errors in parentheses. The spatial connectivity structure (weight matrix, W) is based on the Thiessen polygon approach described earlier. Wg is the spatial lag of dependent variables; Wx is the spatial lag of independent variables.

*** p < 0.01, ** p < 0.05, * p < 0.1

Table 6. Spatial Dependence in the Conditional Convergence Models of GDP per capita and Determinants

Notes: Robust standard errors are in parentheses. The spatial connectivity structure (weight matrix, W) is based on the Thiessen polygon approach described earlier. Wg and Wx represent the spatial lags of the relevant variables. W error (spatial error ) represents the error terms across different spatial units.

*** p < 0.01, ** p < 0.05, * p < 0.1.

Table 7. Comparison of Spatial Durbin Model Using Two Different Weight Matrices

Notes: Robust standard errors are in parentheses. W1 is a contiguity-based Thiessen polygon matrix; W2 is an inverse distance matrix.

*** p < 0.01, ** p < 0.05, * p < 0.1.

Table 8. Direct and Indirect Effects of Key Variables on HDI and GDP per capita under the Spatial Durbin Model

Notes: Robust standard errors are in parentheses. W1 is a contiguity-based Thiessen polygon matrix; W2 is an inverse distance matrix.

*** p < 0.01, ** p < 0.05, * p < 0.1.