This paper investigates the link between social capital and economic growth across countries. The goal is to develop recommendations for developing countries using a social capital measurement model.
Based on a sample of high income and high social capital 15 countries between 2012–2022, this work build to evaluate social capital contributions to economic growth among the other forms of capital. First, the comprehensive natural capital, intellectual capital and social capital score of each country is gathered, and the lineer regression analysis is used to analyze three type of capital. Second, a regression equation model between the natural capital, intellectual capital and the economic status indicator GDP per capita is established, and a descriptive statistical test, a correlation test, and a robustness test are carried out on the model.
First, in order to choose high income countries, from World Bank high income countries list, 15 countries with top highest social capital scores were chosen from SolAbility “Global Sustainable Competitiveness Index 2023”. The list of countries are: Iceland, Finland, Japan, Norway, South Korea, Portugal, Denmark, Sweden, Netherlands, Austria, Italy, Belgium, Spain, France and Singapore.
In this analysis, the data set was created from SolAbility, Global Sustainable Competitiveness Index 2012–2022 period is used because this index provides the yearly set of social capital scores. The scores of Natural Capital Index, Intellectual Capital Index and Social Capital Index for the 2012–2022 period analyzed. The data for GDP per capita (at 2015 fixed prices, $) was gathered from World Bank WDI for the period of 2012–2022.
The purpose of the model to be used in the study is to determine which forms of the capital; social capital, natural capital, and intellectual capital is more effective on GDP per capita. In the linear regression model created, the model shows the level of development of the dependent variable
The term "panel data" refers to the collection of observations on a sample from households, countries, firms, and so on over time. This can be accomplished by surveying and following a certain amount of individuals or households over time. In comparison to a cross-sectional or regular time-series regression, the panel data regression has a double subscript upon its variables, i.e.
yit = α + X’itβ + uit i = 1,..., N; t = 1, ..., T (1)
If we are concentrating on a particular set of N firms, such as IBM, GE, Westinghouse, etc., and our inference is limited to the behavior of these sets of firms, then the fixed effects (FE) model is a suitable specification. As an alternative, it might be a group of N OECD nations or N US states. In this instance, inference is dependent on the specific N businesses, nations, or states that are under observation. In the event that N people are chosen at random from a sizable population, the random effects model specification is suitable. For household panel studies, this is typically the case. The panel's design goes to great lengths to ensure that it is "representative" of the population from which we are drawing conclusions. A fixed effects model could end up in a significant loss of degrees of freedom in this situation because N is typically large. The inference relates to the population for which the sample was randomly selected, and the individual effect is described as random. According to Hausman (1978), comparing βGLS and βWithin, which, if H0 is false, will result in different probability limits but which are both consistent according to the null hypothesis H0: E(uit/Xit) = 0. Based on the difference between the fixed effects (FE) and the RE (feasible GLS) estimator, the Hausman test's statistic can be appropriately calculated in Stata. The Hausman test has a good power to detect the endogeneity of the regressors, according to Monte Carlo experiments (Baltagi, 2005).
GDPPCit = α + β1 NCit + β2 ICit + β3 SCit + eit (2)
Table 3
lnpci | Coef. | Std. Err. | t | P>|t| |
lnnc | .0195443 | .1067248 | 0.18 | 0.855 |
lnin | .7876169 | .1760635 | 4.47 | 0.000 |
lnsc | 1.780988 | .2712448 | 6.57 | 0.000 |
_cons | .1092661 | 1.126092 | 0.10 | 0.923 |
sigma_u | .11494684 | | | |
sigma_e | .27991095 | | | |
rho | .14430271 | | | |
F test that all u_i = 0: F(10, 150) = 1.65 Prob > F = 0.0987 corr(u_i, Xb) = − 0.4119 |
Table 4
lnpci | Coef. | Std. Err. | z | P>|z| |
lnnc | .1659533 | .1008946 | 1.64 | 0.100 |
lnin | .8702225 | .1734977 | 5.02 | 0.000 |
lnsc | 1.151515 | .2226769 | 5.17 | 0.000 |
_cons | 1.769997 | 1.057507 | 1.67 | 0.094 |
sigma_u | 0 | | | |
sigma_e | .27991095 | | | |
rho | 0 | | | |
corr(u_i, X) = 0 (assumed) |
Table 5
Hausman endogenous test fe vce (r)
lnpci | Coef. | Robust Std. Err. | t | P>|t| |
lnnc | .0195443 | .078479 | 0.25 | 0.808 |
lnin | .7876169 | .0921518 | 8.55 | 0.000 |
lnsc | 1.780988 | .246325 | 7.23 | 0.000 |
_cons | .1092661 | .873986 | 0.13 | 0.903 |
sigma_u | .11494684 | | | |
sigma_e | .27991095 | | | |
rho | .14430271 | | | |
Table 6
Hausman endogenous test re vce (r)
lnpci | Coef. | Robust Std. Err. | z | P>|z| |
lnnc | .1659533 | .0500439 | 3.32 | 0.001 |
lnin | .8702225 | .0931658 | 9.34 | 0.000 |
lnsc | 1.151515 | .1896619 | 6.07 | 0.000 |
_cons | 1.769997 | .6769803 | 2.61 | 0.009 |
sigma_u | 0 | | | |
sigma_e | .27991095 | | | |
rho | 0 | | | |
Social capital has a more robustly positive effect on GDPPC. All of the three forms of capital are found significant for the model. A 1 percentage point increase in social capital boosts GDPPC by 1.15 percentage points, natural capital 0.16 percentage points, and intellectual capital 0.87 percentage points, respectively.
Figure 1 shows the relationship between per capita income and social capital. As seen from the figure when social capital increases, per capital income increases.