=Paper=
{{Paper
|id=Vol-2104/paper_180
|storemode=property
|title=Modeling the Social and Human Capital Factors Effect on the Cross-Country Income
Differences
|pdfUrl=https://ceur-ws.org/Vol-2104/paper_180.pdf
|volume=Vol-2104
|authors=Olena Bazhenova,Ihor Chornodid
|dblpUrl=https://dblp.org/rec/conf/icteri/BazhenovaC18
}}
==Modeling the Social and Human Capital Factors Effect on the Cross-Country Income
Differences==
https://ceur-ws.org/Vol-2104/paper_180.pdf
Modeling the Social and Human Capital Factors Effect
On the Cross-Country Income Differences
Olena Bazhenova1, Ihor Chornodid2
1
Taras Shevchenko National University of Kyiv,
olenabazhenova@univ.net.ua
2
Academy of Labor, Social Relations and Tourism
chornodid@ukr.net
Abstract. The paper is devoted to the investigation the effect of social and hu-
man capital factors on the cross-country income differences among countries
that have similar with Ukraine income levels based on the economic and math-
ematical models construction. It was constructed four panel data models to es-
timate the effect of human capital accumulation on cross-country income dif-
ferences. To study the influence of social factors on the dynamics of gross na-
tional income per capita of the countries and the causes of its volatility it was
constructed the panel vector autoregression model.
Keywords. Social infrastructure, human capital, cross-country income differ-
ences, VAR model, panel data model.
1 Introduction
Dynamic development of the national economy is impossible without ensuring its
social competitiveness as an important element of the overall competitiveness of the
country and raising the population standard of living.
National social competitiveness is determined both by economic and socio-
political factors, as well as by the infrastructure of the country, its scientific potential,
the level of education of the population [1].
Quantitative and qualitative assessment of the potential of social competitiveness
helps to determine the existing and potential competitive advantages and competitive
status of the country in international comparisons.
In the context of the countries’ social competitiveness research, the attention
should be paid to the investigation of cross-country income differences determinants.
Besides physical capital and labor, it should be considered indicators that characterize
investment in human capital and social aspects in particular, health care and R&D
expenditures, etc. As known, human capital represents an additional production factor
along with labor and physical capital, which includes education, work experience and
other aspects. According to some macroeconomists, only by investing in human capi-
tal poor countries may converge to wealthy ones. Some economists focus on cross-
country income differences that explained by social infrastructure [2]. The aspects of so-
cial infrastructure include features of fiscal policy conducted by the governments, envi-
ronment that surrounds the private agents in economy, etc.
The one of the most prevalent method of investigation of social aspects and human cap-
ital effects on cross-country income differences is the regression framework.
The purpose of the paper is to determine the influence of social and human capital fac-
tors on the cross-country income differences among countries that have similar with
Ukraine income level based on the economic and mathematical models construction.
2 Analysis of Recent Research and Publications
The importance of social infrastructure and human capital in explaining the cross-
country output differences is empirically tested by using regression techniques.
Research on importance of human capital and social infrastructure using the re-
gression toolkit is devoted to the works of many scientists. Among them, we can em-
phasize works of Hall and Jones [2], Klenow and Rodrigues-Clare [3], Acemoglu,
Johnson and Robinson [4, 5], Sachs and Warner [6] etc.
These papers explores measuring differences in human capital accumulation and
social infrastructure aspects and estimation its influence on income differences with
the regression framework based on the use of production functions, for example
Cobb-Douglas production function [2, 3]. The authors also estimated the proportion
of cross-country volatility in income due to volatility of these factors.
For example, Hall and Jones [2] propose to estimate social infrastructure influence
on cross-country income differences using OLS regression. According to results of
their research, the influence of social infrastructure on income is significant. Above
all, volatility of social infrastructure cause a great volatility of cross-country income
differences. To measure the social infrastructure Hall and Jones use two indexes: an
index of government anti-diversion policies and index of openness or market-
orientation (proposed by Sachs and Warner [6]).
Klenow and Rodrigues-Clare [3] suggest estimating the effect of human capital ac-
cumulation on income given the Cobb-Douglas production function with two factors:
physical capital and effective labor services. The results of estimation show that the
gap between poorest and wealthiest countries due to differences in human capital
accumulation is less than 25%.
3 Research Methods
For the investigation of human capital and social aspects influence on the cross-
country income differences, we propose to use panel data models and vector auto-
regression models toolkits.
To measure the social aspects we suggest using the group of social aspects indica-
tors mentioned in p.4.
Thus, in the research we suggest construction following panel data models:
yit = α i + β1 x1it + ... + β k x kit + ε it (1)
2
where yit - the resulting variable, xit - k- dimensional vector of explanatory varia-
bles that does not include a constant [7]. The social aspects indicators are observed for
N (i=1…N) observation units (countries) during T periods (t=1…T). In turn, the ef-
fects of change in x are the same for all units of observation. At the same time, the
average levels for each unit of observation are different. Elements α i characterize the
influence of individual factors for a i-th observation unit that is constant throughout
the time period; perturbations ε it are independent, equally distributed random varia-
bles with mean 0 and variance σ ε . If α i are fixed, the model is called a fixed-effect
2
panel model. And if α i are random variables with mean µ and variance σ α , we
2
have the panel model with random effects. Consequently, the error in this model has
two components: independent of time α i and residual components - ε it .
So, the model with random effects can be written as follows:
yit = µ + α i + β1 x1it + ... + β k x kit + ε it (2)
where µ - free term or intercept [7].
To analyze the influence of human capital and social aspects on the countries’ in-
come dynamics and investigating its volatility we propose to use the vector auto-
regression models toolkit.
Thus, the p-th order vector autoregression model or VAR(p) has the following
form:
Yt = C 0 + C1Yt −1 + ... + C p Yt − p + vt , (3)
where Yt is the k-dimensional vector of the endogenous variables of the model, C 0 -
k-dimensional vector of constants, Cj - the matrix of coefficients of kxk (j=1…p)
dimension, vt - is the k-dimensional perturbation vector with the covariance matrix Σ
[7].
The stability or stationary of the vector autoregression model is the decay of exter-
nal shocks over time. So, the VAR(p) model to be stationary the characteristic roots
that are found by solving the equation
λ p I − λ p −1C1 − ... − λC p −1 − C p = 0 (4)
in absolute value must be less than one or lie within a single circle.
If there is a shock to the system (one of the vector vt element changes), model
variables should deviate from their equilibrium state and eventually return to it. The
trajectory of returning variables to its equilibrium state is an impulse response.
Impulse response functions are calculated by finding partial derivatives
Θ i = ∂Yt / ∂vt − i . (5)
The (m,s)-th element of this matrix shows how the error in the m-th equation of the
system affects the S-dependent variable in the presence of a lag in i periods.
The analysis of the decomposition of the predictions errors variances of the model
variables allows determining the sources of their volatility.
4 Results
The indicator that characterizes cross-country income is the gross national income per
capita. As regressors that influence the gross national income per capita, we selected
such indicators as gross capital formation, labor force share in the total population
over 15 years, health care and education expenditures, education level of the popula-
tion that measured as the proportion of the population entering the higher education
institutions.
Thus, as variables measuring investment in human capital, along with physical
capital (gross capital formation) and labor (the share of labor force in the total popula-
tion over 15 years) the model includes the expenditures on education and health care.
Moreover, the proportion of the population enrolled in higher education reflects the
level of countries’ human capital. For example, according to S. Kuznets [8], advanced
technology is only a necessary but not sufficient condition for economic growth. The
production of own innovations is based on the institutional transformations that are
stimulated only by the accumulated amount of human capital. Therefore, the main
source of economic growth is "breakthroughs" in raising the level of human capital
("epochal innovations").
The data source of the research is the World Bank data during 2000-2015 for 14
countries: Armenia, Belarus, Bulgaria, Estonia, Indonesia, Kosovo, Mongolia, Para-
guay, Poland, Russia, Samoa, Serbia, Ukraine and Chile [9]. We selected these coun-
tries according to their gross national income per capita similar to Ukraine’s one and
due to interest in the context of comparing the results of the research. The selection of
indicators and time period also was limited by the availability of the data. As a further
way of research it will be of interest to model the influence of human capital and so-
cial factors on economic growth of low-income and high-income countries and due to
measure the level of human capital to include such indicators as labor force with
basic, intermediate and advanced education.
All variables are modeled in logarithms. Moreover, the variables were tested for a
unit root with tests for models with panel data such as Lewin, Lina, and Chu and Brei-
tung criteria for the existence of common process of a unit root and Ima, Pesaran, and
Tina criteria, criteria based on the use of ADF and PP statistics that include individual
processes of unit root.
The results of the tests showed that all variables are first order integrated (Table 1).
Therefore, we include the variables in the models in the first differences.
To verify the robustness of obtained results the proportion of research spending
due to GDP, fertility rate and the proportion of population aged 15-64 were added to
the model.
The results of estimation of panel data models are presented in Table 2. All coeffi-
cients presented in Table 2 are the coefficients of elasticity of gross national income
per capita with respect to the regressors of the models.
4
Table 1. The results of unit root testing of panel data model variables
Variable Name of variable in the Order of
model integration
Gross national income per person GNI_PER_CAPITA_LN І(1)
Gross fixed capital formation GFCF_LN І(1)
Share of labor force in the total popu- LF_PART_RATE_LN І(1)
lation over 15 years
Health care expenditures HEALTH_EXP_LN І(1)
Education expenditures EDU_SPEN_LN І(1)
Population enrolled in higher educa- ENROL_SCHOOL_LN І(1)
tion
Ratio of R&D spending to GDP RD_GDP_LN І(1)
Fertility rate FERTILITY_RATE_LN І(1)
Proportion of population aged 15-64 POP_15_64_LN І(1)
The choice of models with fixed effects is based on the verification of Redundant
Fixed Effect-Likelihood Ratio test. It should be noted that all evaluated models are
significant with sufficiently high values of R-squared, the residuals of which have a
normal distribution and are characterized by the absence of auto-correlation.
Thus, as seen from the table 2, the gross fixed capital formation, the share of labor
force, health care and education expenditures and the level of population education
(measured by the indicator of the proportion of the population admitted to higher
educational institutions) do explain the increase of gross national income per capita.
In three of four models, all indicators are significant. In the last model, the gross fixed
capital formation is not significant.
Due to results of the estimation, the increase in expenditures on research contrib-
utes to the growth of national income with a lag of 3 years. However, the addition of
population aged 15-64 years to the model leads to the insignificance of gross fixed
capital formation and fertility rate.
Analyzing the values of the elasticity coefficients, the largest impact on the gross
national income per capita of the countries has the proportion of the population enrol-
ling the higher education and the share of labor force in the total population over 15
years.
Table 2. The results of estimation of fixed effects panel data models
Dependent variable
Regressors ∆GNI_PER_CAPITA_LN
∆EDU_SPEN_LN(-3) 0.105576* 0.121558* 0.130035* 0.127586*
∆ENROL_SCHOOL_ 0.783125* 0.994216* 0.776312* 0.711054*
LN
∆HEALTH_EXP_LN 0.555596* 0.574627* 0.533300* 0.520187*
∆LF_PART_RATE_ - - -1.584307** -1.482774*
LN 1.320736** 0.804834**
∆GFCF_LN 0.115139** 0.093158** 0.073278** 0.062377
*
Intercept 0.038660** 0.026281* 0.036353** 0.031539
∆RD_GDP_LN(-3) 0.083733*
∆FERTILITY_RATE 0.593319*** 0.484617
_LN
∆POP_15_64_LN 3.646704
Cross-section fixed Yes Yes Yes Yes
effects
R-squared 0.787213 0.829872 0.807529 0.810558
*, **, and *** denote the significance of the coefficients at 1%, 5% and 10% error
Table 3 represents the fixed effects calculated due to constructed models.
Thus, the obtained results show that an increase in investment both in human capi-
tal and in physical capital leads to accelerated economic growth of countries and the
convergence of poorer countries to a richer ones.
To study the social factors influence on the dynamics of national income we use
the VAR approach that concentrates on the research of its volatility causes and reac-
tion on impulses. For this purpose, we construct the panel vector autoregression mod-
el.
Table 3. Fixed effects calculated due to constructed models
Country Model 1 Model 2 Model 3 Model 4
Armenia 0.024058 0.019678 0.027125 0.012702
Bulgaria -0.024964 -0.031207 -0.030838 -0.011481
Belarus 0.023514 0.015108 0.017077 0.015675
Chili -0.014704 -0.061004 -0.000728 -0.009334
Estonia 0.034561 0.036644 0.018565 0.027056
Indonesia -0.046747 - -0.038308 -0.030230
Poland -0.008750 -0.009697 -0.007464 -0.008341
Paragway -0.001606 -0.001958 0.019765 -0.001346
Russian Federation 0.055708 0.052050 0.046846 0.043888
Serbia -0.062459 -0.062453 -0.042929 -0.051374
Ukraine -0.000588 -0.007072 -0.010587 -0.005839
Thus, as indicators that characterize the cross-country income differences in the
models the gross domestic product per capita (in current US dollars) and gross nation-
al income (in current US dollars) are used.
The social aspects of society development in the models are described by following
indicators: food production index, tuberculosis incidents (per 100 thousand popula-
tion), life expectancy at birth for men and women (in years), infants mortality rate
(per 1,000 newborns), AIDS rate (percent of the population aged 15-49), unemploy-
ment rate, population growth rate, fertility rate (births per woman), share of labour
force in the population aged 15 and over, mortality rate (per 1 thousand people),
health care expenditures per capita (current US $), percent of population aged 15-64,
R&D expenditures (in percent to GDP), fraction of high-tech products exports in total
exports of products, Internet users (per 100 people), fraction of population with access
6
to improved sanitary conditions, carbon dioxide emissions (metric tons per capita),
the fraction of the population entering the high education institutions (percent of pop-
ulation), public expenditure on education (as a percentage to GDP), the number of
mobile communication users (per 100 people).
The testing of these variables for the presence of a unit root indicated that almost
all variables are first order integrated, except for the growth rate of population that is
second order integrated (Table 4). Therefore, in the model the first differences of
variables will be used.
At the next stage of the research, we analyze the relationship between the variables
based on the cause and effect relationship analysis by Granger causality test and con-
structing the correlation matrix.
According to the constructed correlation matrix, the indicators having a close rela-
tionship with the variables that measure the income of countries - GDP and GNI per
capita - are the following: the amount of carbon dioxide emissions, health care ex-
penditures, Internet users and the life expectancy at birth for women and men.
Table 4. The results of unit root testing of VAR model variables
Variable Name of variable in the Order of
model integration
Carbon dioxide emissions CO2_EMIS_LN I(1)
Death rate DEATH_RATE_LN I(1)
Public expenditures on education EDU_SPEN_GDP_LN I(1)
The share of the population entering the ENROL_SCHOOL_LN I(1)
high education institutions
Fertility_rate FERTILITY_RATE_LN I(1)
Food production index FPI_LN I(0)
GDP per capita GDP_CAP_LN I(1)
GNI per capita GNI_PER_CAPITA_LN I(1)
Health care expenditures per capita HEALTH_EXP_CAP_LN I(1)
Fraction of high-tech products exports in HIGH_TECH_EXPORT_ I(1)
total exports of products LN
Tuberculosis incidents INC_TUB_LN I(1)
Internet users INTERNET_USERS_LN I(1)
Life expectancy at birth for women LEB_FEM_LN I(1)
Life expectancy at birth for men LEB_MALE_LN I(1)
Fraction of labour force in the popula- LF_PART_RATE_LN I(1)
tion aged 15 and over
Number of mobile communication users MOB_SUBS_LN I(1)
Infants mortality rate MORT_RATE_INF_LN I(1)
Fraction of population aged 15-64 POP_15_64_LN I(1)
Population growth rate POP_GR_LN І(2)
AIDS rate PREV_HIV_LN I(1)
R&D to GDP ratio RD_GDP_LN I(1)
Fraction of population with access to SANIT_FAC_LN I(1)
improved sanitary conditions
Unemployment rate UN_RATE_LN I(1)
At the same time Granger's causality test indicated that the number of Internet us-
ers and cellular networks, the fraction of the population aged 15-64 and the health
care expenditures (with lags 1-5), birth rate (with lags 3 and 4) and R&D expenditures
(lags 2 and 5) contributed to the explanation of the countries income.
The Johansen test indicated no cointegration relations between variables.
All these indicators were included in the model as endogenous variables. In addi-
tion, the rate of foreign direct investments growth is added to the model as indicator
measuring the attractiveness of the country for the foreign investors.
Thus, the vector model of autoregression in the reduced form was estimated with
an intercept and one lag that was determined based on the application of the Schwarz
information criterion. The obtained vector autoregression model satisfies the condi-
tion of stability as evidenced by the non-exaggeration by the inverse roots of the char-
acteristic autoregression polynomial the unit value (in absolute values).
Based on the constructed model, impulse response functions were generated for
differenced gross national income per capita (fig. 1).
Analyzing the impulse response functions, we conclude that the positive shocks in
all variables (except R&D expenditures, the number of Internet users and the fraction
of population aged 15-64) lead to an increase in the gross national income per capita
and its further stabilization. At the same time, the change of these three indicators
provokes a slight deterioration of GNI per capita during the first two years, its further
growth and stabilization after the 6th period.
The volatility of national income per capita is explained by its own fluctuations by
almost 65% since the 8th year. The variation in the number of cellular communication
users and the growth rate of foreign direct investment account for about 18% and 9%
of the fluctuations of GNI per capita growth.
At the same time, the variation in spending on health and R&D explains about 3%
of the variation in gross national income per capita.
In order to verify the robustness of the obtained results, we constructed a similar
model for a gross domestic product per capita that presented similar results. The GDP
per capita behavior is similar to the GNI per capita response to the simulated impuls-
es, except for the response to a positive shock in health care spending.
8
ResponseВідгук
of D(GNI_PER_CAPITA_LN) to
на impulse
імпульс in of D(GNI_PER_CAPITA_LN) на
Відгук
Response to імпульс
impulse in Response of D(GNI_PER_CAPITA_LN) на
Відгук toімпульс
impulse in
D(HEALTH_EXP_CAP_LN) D(INTERNET_USERS_LN) FDI_PC
.06 .06 .06
.04 .04 .04
.02 .02 .02
.00 .00 .00
-.02 -.02 -.02
-.04 -.04 -.04
2 4 6 8 10 12 14 16 18 20 22 24 2 4 6 8 10 12 14 16 18 20 22 24 2 4 6 8 10 12 14 16 18 20 22 24
of D(GNI_PER_CAPITA_LN) to
ResponseВідгук на impulse
імпульс in of D(GNI_PER_CAPITA_LN) на
Відгук
Response toімпульс
impulse in of D(GNI_PER_CAPITA_LN) наtoімпульс
Відгук
Response impulse in
D(MOB_SUBS_LN) D(POP_15_64_LN) D(RD_GDP_LN)
.06 .06 .06
.04 .04 .04
.02 .02 .02
.00 .00 .00
-.02 -.02 -.02
-.04 -.04 -.04
2 4 6 8 10 12 14 16 18 20 22 24 2 4 6 8 10 12 14 16 18 20 22 24 2 4 6 8 10 12 14 16 18 20 22 24
Response of D(GNI_PER_CAPITA_LN) на
Відгук toімпульс
impulse in Response of D(GNI_PER_CAPITA_LN) на
Відгук toімпульс
impulse in of D(GNI_PER_CAPITA_LN) на
Відгук
Response toімпульс
impulse in
D(FERTILITY_RATE_LN) D(CO2_EMIS_LN) D(LEB_MALE_LN)
.06 .06 .06
.04 .04 .04
.02 .02 .02
.00 .00 .00
-.02 -.02 -.02
-.04 -.04 -.04
2 4 6 8 10 12 14 16 18 20 22 24 2 4 6 8 10 12 14 16 18 20 22 24 2 4 6 8 10 12 14 16 18 20 22 24
Fig. 1. Impulse responses functions for differenced gross national income per capita
A positive shock in this variable leads to an initial deterioration of GDP per capita
and its subsequent stabilization after the 6th period.
5 Conclusions
Summing up the results and interpreting them, we note that the inalienable factors of
production included in the aggregate production function are investments in human
capital and social infrastructure.
The largest impact on the countries’ gross national income per capita has the pro-
portion of the population enrolling the higher education and the share of labor force in
the total population over 15 years that prove the role of human capital accumulation
as a driving force of economic growth.
Moreover, positive shocks in all social aspects indicators (except R&D expendi-
tures, the number of Internet users and the fraction of population aged 15-64) lead to
an increase in the gross national income per capita.
In this context, we note that it is possible to accumulate human capital for an arbi-
trarily long time, since its marginal productivity is a constant value. The pace of
growth of a country investing in human capital will increase even on a balanced
growth path.
In turn, insufficient investment in infrastructure also could explain the insignificant
convergence between countries primarily due to lack of its mobility and the impossi-
bility of purchasing in international markets.
As a further research it could be the one based on data for low-income and high-
income countries including such indicators as labor force with basic, intermediate and
advanced education.
References
1. Chornodid, I.: Mechanism of social competitiveness ensuring in national economy for-
mation conditions. Ekonomika ta derzhava 5, 40-44 (2014).
2. Hall, R.E., Jones, C.I.: Why Do Some Countries Produce So Much More Output per
Worker than Others? Quarterly Journal of Economics 114, 83-116 (1999).
3. Klenow, P.J., Rodrigues-Clare, A.: The Neoclassical Revival in Growth Economics: Has it
Gone Too Far? NBER Macroeconomics Annual 12, 73-103 (1997).
4. Acemoglu, D., Johnson, S., Robinson, J.A.: The Colonial Origins of Comparative Devel-
opment: An Empirical Investigation. American Economic Review 91, 1369-1401 (2001).
5. Acemoglu, D., Johnson, S., Robinson, J.A. Reversal of Fortune: Geography and Institu-
tions in the Making of the Modern World Income Distribution. Quarterly Journal of Eco-
nomics 117, 1231-1294 (2002).
6. Sachs, J., Warner, A.: Economic Reform and the Process of Global Integration. Brookings
papers on Economic Activity 1, 1-95 (1995).
7. Chernyak, O., Stavytskyy, A., Bazhenova, O., Shebanina, O.: Econometrics. MNAU,
Mkolayiv (2014).
8. Kuznets, S.: Modern Economic Growth: Findings and Reflections. Lecture to the memory
of Alfred Nobel, December 11 1971, Homepage,
http://www.nobelprize.org/nobel_prizes/economic-sciences/ laureates/1971 / kuznets - lec-
ture.html, last accessed February, 19, 2018.
9. World Development Indicators. The World Bank. Homepage,
http://databank.worldbank.org/ data/reports.aspx? source=wdi-database-archives-
%28beta%29#, last accessed February, 22, 2018.
10