University of Pretoria Department of Economics Working Paper Series The Relationship between Population Growth and Economic Growth Over 1870- 2013: Evidence from a Bootstrapped Panel-Granger Causality Test Tsangyao Chang Feng Chia University Hsiao-Ping Chu Ling-Tung University Frederick W. Deale University of Pretoria Rangan Gupta University of Pretoria Working Paper: 2014-31 June 2014 __________________________________________________________ Department of Economics University of Pretoria 0002, Pretoria South Africa Tel: +27 12 420 2413
31
Embed
University of Pretoria Department of Economics …...University of Pretoria Department of Economics Working Paper Series The Relationship between Population Growth and Economic Growth
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
University of Pretoria
Department of Economics Working Paper Series
The Relationship between Population Growth and Economic Growth Over 1870-
2013: Evidence from a Bootstrapped Panel-Granger Causality Test
This study applies the bootstrap panel causality test proposed by Kónya (2006), which
accounts for both dependency and heterogeneity across countries, to test the causal link
between population growth and economic growth in 21 countries over the period of
1870-2013. With regards to the direction of population growth-economic growth nexus, we
found one-way Granger causality running from population growth to economic growth for
Finland, France, Portugal, and Sweden, one-way Granger causality running from economic
growth to population growth for Canada, Germany, Japan, Norway and Switzerland, and no
causal relationship between population growth and economic growth is found in Belgium,
Brazil, Denmark, Netherlands, New Zealand, Spain, Sri Lanka, the UK, the USA and
Uruguay. Furthermore, we found feedback between population growth and economic growth
for Austria and Italy. Dividing the sample into two subsamples due to a structural break
yielded different results in that for the first period of 1871-1951 we found that population
growth Granger cause economic growth only for Finland and France, economic growth
Granger cause population growth for Denmark, Japan, and Norway and that there is
bidirectional causality between population growth and economic growth for both Austria and
Italy. For the period of 1952-2013 we found that population growth Granger cause economic
growth only for Sri Lanka, economic growth Granger cause population growth for Belgium,
Denmark, France, Germany, New Zealand, Spain, Switzerland, and Uruguay and that found
bidirectional causality between population growth and economic growth only for Japan. Our
empirical results have important policy implications for these 21 countries under study as the
directions of causality tend to differ across countries and depending on the time period under
question.
Keywords: Population Growth; Economic Growth; Dependency and Heterogeneity;
Bootstrap Panel Causality Test
JEL Classification: C32, C33, O40, Q56
3
1. Introduction
It remains very important whether there is any causal link between population growth
and economic growth, not only for demographers and economists but also for policy makers.
However, this relationship has long been contentious. For an excellent survey of the
relationship between population growth and economic growth, see Cassen (1976). Numerous
studies have found a negative association between these two variables (e.g. Galor and Weil
(2000) and Li and Zhang (2007), to name a few). Galor and Weil (1996) claimed that, given
economic growth increases women’s relative wages, the opportunity costs of raising children
increase simultaneously with economic growth, reducing fertility. In contrast, contradictory
results also exist in the previous studies (i.e., Dasgupta, 2000; Drèze and Murthi, 2001; Huang
and Xie, 2013; Yao et al. 2013). McNicoll (1984) stressed the causal effect of economic
growth on population growth which means that strong economic growth causes population
growth either through increased birth rates or migration. Thus, it is very important for
empirical researchers to formulate the causal link between economic growth and population
growth as two-equations modeled simultaneously (see, Darrat and Al-Yousif, 1999; Thornton,
2001; Huang and Xie, 2013; Yao et al. 2013).
This study represents our first attempt to study the causal link between population
growth and economic growth using a long historical time series data for the period 1870-2013.
Our recent experience in economic dynamics shows that turbulence in a region may easily be
transmitted to other regions through international trade and economic and financial
integration, two of the basic activities amongst regions implying the importance of taking into
account cross-section dependency in empirical analysis. Previous studies examining the
correlation between population growth and economic growth failed to examine their two-way
inter-relationship. Even though there is strong dependence between countries, it is well known
that each country sustains its own dynamics in the developmental process. This fact calls
4
attention to also controlling cross-country heterogeneity when initiating an empirical
modeling strategy. Taking the above into account, the panel causality method utilized in this
paper is good enough to control for dependency across countries as well as country-specific
characteristics following a systematic modeling strategy. When examining causal linkages
between the variables in concern, we separately test for both cross-section dependence and
cross-region heterogeneity by using recently developed and statistically powerful tests instead
of assuming the existence of these dynamics in the panel data set. This contributes to existing
literature by addressing these two concerns jointly.
As indicated by Afzal (2009), cross-national evidence on the relationship between
population growth and economic growth is inconsistent because the underlying parameters
and assumptions vary across countries. In recognition of this we apply the bootstrap panel
causality test proposed by Kónya (2006) to discover the dynamic and causal relationships
between population growth and economic growth for 21 countries over the period 1870-2013
since this test accounts for both dependency and heterogeneity across countries. By utilizing
the panel Granger causality approach instead of time series methods the panel data sets
include information not only from a time series dimension, but also from a cross-sectional
dimension and will undoubtedly allow for country-specific effects to be more readily
uncovered. Based on this advantage, non-stationary panel tests (unit root, cointegration, and
causality) have become a more powerful econometric methodology in recent years.
To the best of our knowledge, this is the first study that uses a bootstrap panel Granger
causality test to study the relationship between population and economic growth in the 21
countries using such a long time data series. One advantage of the econometric methodology
proposed by Kónya (2006) is that it allows for contemporaneous correlation across regions.
We are therefore utilizing a more meaningful and effective analysis methodology, because the
interaction between economic sectors across countries usually exists, as compared to a
cross-sectional analysis or time series analysis on a country-by-country basis. Hopefully, this
5
study can fill the void in current literature regarding population and economic growth.
The outline of this paper is organized as follows: Section 2 reviews some previous
literature, Section 3 presents the data used in this study and Section 4 describes the bootstrap
panel Granger causality test proposed by Kónya (2006). Section 5 presents our empirical
results while Section 6 concludes the paper.
2. Literature of Population and Economic Growth Nexus
In previous literature there are many views and schools of thought regarding the
relationship between population and economic growth. According to Luigi et al. (2010), these
can be grouped together in three main groups depending on their evaluation of population
growth and economic results namely Negative Effects, Positive Effects, and No Effects.
Based on previous literature, the first theorist who became well-known for his
population theory is Thomas R. Malthus (1798). According to Malthus (1798), population
growth is supposed to decrease the per capita output, because output growth cannot keep up at
the same pace as population growt. In order to keep the natural balance in the population,
especially that of food- and consumption, preventive checks (i.e. fertility reduction) and
positive checks (i.e. mortality increase) on population growth are necessary (Malthus, 1798).
According to Easterlin (1967), the main assumption of the Malthus theorem is the limited
availability of natural resources that constrains both population and economic growth.
In the neoclassic model of growth, Solow (1956) treated population as an exogenous
variable and he thought population growth naturally follow an arithmetical pattern instead of
a geometrical one. Based on this, Solow (1956) built his model using the population growth
rate and assuming that a constant and natural population growth is independent on economic
dynamics. According to Solow (1956), there are two distinct effects of population growth rate
change on the output growth. In his opinion, on the one hand, an increase in the population
6
growth rate will increase the amount of labor and thus both the absolute level of output and
the steady state output growth rate. On the other hand, it will also reduce physical capital
stock per worker; therefore, a decrease in productivity and in the steady state output per
worker. To simplify the explanation, it means that higher population growth per se would be
detrimental for economic development.
Mason (1988) also demonstrated from theoretical and empirical point of view that
population growth may reduce saving propensity, lower potential investment and this all leads
to a further decrease in physical capital per worker, and thus in per capita steady state output.
According to Easterlin (1967), the main assumption of his study is on the limited availability
of physical capital which does not affect population growth (which is exogenously
determined) but constrains economic growth.
In conclusion, both schools of thought implied higher population growth will be
detrimental to economic growth, therefore, they support population control policies,
especially in developing countries. Decreasing population is a necessary and important step to
living conditions, because it would raise per capita resources availability (see Easterlin, 1967).
According to Toney et al. (1981), Malthusian and neo-Malthusian position receive a wide
consensus, with very few exceptions.
The second group of social scientists who were known to challenge Malthusian theories
from an economic point of view were Kuznets, Quandt, and Frideman (Kuznets, Quandt, and
Frideman, 1960). They highlight the positive effectiveness of population growth on economic
cycles. They consider that three major activities conducted by people namely production,
consumption, and saving will contribute to economic growth. Kuznets (1976) provided more
empirical evidence on the beneficial effects related to population growth which is called a
deeper analysis and critique of Malthusian theories. Kremer (1993) has empirically confirmed
that larger population was associated with higher population growth rates and faster
technology improvement, which is a consequence of population growth, and leads to an
7
increase of labor productivity, per capita income and improvement in living standards. The
main focus of this school of thought has shifted from natural and reproducible physical capital
to knowledge. Therefore, production was theorized to be free from the diminishing returns to
scale that characterized the previous economic analysis. According to Espenshade (1978),
policy advice derived from this school of thought including support of fertility and
immigration in countries with declining or stationary population.
More recently, a new school of thought argues that the rise in population is neutral on
economic growth in that it may not determine economic growth, but the former variable does
not hamper the latter (Simon, 1987). It may be the problem of employment, development and
distribution of the increased population (Kuznets, 1955; Todaro and Smith, 2006) when the
population is large. However, to date, due to the one factor or the other, the issue remains
inconclusive (Birdsall, et al. 2001).
3. Data
This study applies annual population and per capita real GDP for 21 countries over the
period of 1870-2013. Both data sets (from 1870-2001) are from the accompanying data sets of
“Two Thousand Years of Economic Statistics World Population, GDP and PPP” by Alexander
V. Avakov (2010) and extended by OECD data source from 2001-2013. Due to data
availability, we only have 21 countries with such a long time series, 1870-2013.1 Tables 1, 2,
3, and 4 report both summary statistics of population, population growth, per capita real GDP,
and per capita real GDP growth, respectively.
Table 1 indicates that the USA and Brazil have the highest and lowest mean per capita real
1 21 countries include Austria, Belgium, Brazil, Canada, Denmark, Finland, France, Germany, Italy,
Japan, Netherlands, New Zealand, Norway, Portugal, Spain, Sri Lanka, Sweden, Switzerland, the U.K.,
the U.S. and Uruguay. Lack of continuous data for physical capital and labor has restricted our model
to a bivariate case rather than multivariate case.
8
GDP of $11,657 and $2,335.09, respectively. The USA also has the highest mean population
of 154, 902 thousand and Uruguay have the lowest mean population of 1,963.79 thousand as
indicated by Table 3. If we look at tables 2 and 4, we can see that the per capita real GDP
growth is higher than the population growth for all 21 countries over the sample period of
1870-2013. If we look at Table 2, Japan and Uruguay have the highest and lowest per capita
real GDP growth rate of 2.4% and 1.1%, respectively. Table 4 indicates that Brazil and France
have the highest and lowest mean population growth rate of 2.13% and 0.37%, respectively.
4. Methodology
a. Bootstrap Panel Granger Causality Test
In this study, we apply the bootstrap panel causality method proposed by Kónya (2006)
to measure the determinants of causality between population and economic growth. As
emphasized by Kónya (2006),2 the results of the bootstrap panel causality method unit root
test and cointegration test are all robust which implies that not all variables need to be tested
for stationary series properties (Kónya, 2006).3 The robust feature of bootstrap panel
causality arises from the generation of country-specific critical values from the bootstrapping
method. It is important to note here that the variable levels used in empirical analysis play
crucial roles in determining causal linkages because differencing variables to make them
stationary (i.e., using the difference form of variables) may lead to a loss of trend dynamics in
the series.
The bootstrap panel causality approach of Kónya first requires estimating the described
system by SUR to impose zero restrictions for causality by the Wald principle, and then
requires generating bootstrap critical values. Since country specific Wald tests with country
2 The alternative panel Granger causality test was developed by Hurlin (2008). The method controls for unobservable heterogeneity in panel data, but not for heterogeneity problems in cross-sectional data. 3 We refer to Kónya (2006) for more details of the bootstrapping method and of country-specific critical values.
9
specific bootstrap critical values are used in the panel causality method, the Wald test does
not require a joint hypothesis for all countries in the panel.
The equation system for panel causality analysis includes two sets of equations that can
be written as:
1 1
1 1
1 1
1, 1,1 1,1, 1, 1,1, 1, 1,1,
1 1
2, 1,2 1,2, 2, 1,2, 2, 1,2,
1 1
, 1, 1, , , 1, , 1, , 1, ,
1 1
ly lx
t i t i i t i t
i i
ly lx
t i t i i t i t
i i
ly lx
N t N N i N t i N i N t i N t
i i
PEG PEG POG
PEG PEG POG
PEG PEG POG
α β δ ε
α β δ ε
α β δ ε
− −= =
− −= =
− −= =
= + + +
= + + +
= + + +
∑ ∑
∑ ∑
∑ ∑
⋮
(1)
and
2 2
2 2
2 2
1, 2,1 2,1, 1, 2,1, 1, 2,1,
1 1
2, 2,2 2,2, 2, 2,2, 2, 2,2,
1 1
, 2, 2, , , 2, , , 2, ,
1 1
ly lx
t i t i i t i t
i i
ly lx
t i t i i t i t
i i
ly lx
N t N N i N t i N i N t i N t
i i
POG PEG POG
POG PEG POG
POG PEG POG
α β δ ε
α β δ ε
α β δ ε
− −= =
− −= =
− −= =
= + + +
= + + +
= + + +
∑ ∑
∑ ∑
∑ ∑
⋮
(2)
In the equation systems (1) and (2), PEG refers to the indicator of per capita economic growth,
POG denotes the indicator of population growth, N (=21) is the number of panel members, t
is the time period (t=1,…,T), and l is the lag length. In this regression system, each equation
has different predetermined variables and the error terms might be cross-sectionally correlated
hence, we can view these sets of equations as an SUR system. To test for Granger causality in
this system, alternative causal relations for each country are likely to be found: (i) there is
one-way Granger causality from POG to PEG if not all 1,iδ are zero, but all 2,iβ are zero; (ii)
there is one-way Granger causality from PEG to POG if all 1,iδ are zero, but not all 2,iβ are
zero; (iii) there is two-way Granger causality between POG and PEG if neither 1,iδ nor
2,iβ are zero; (iv) there is no Granger causality between POG and PEG if all 1,iδ and 2,iβ
10
are zero.
Before proceeding with the estimation, the optimal lag lengths must be determined.4
Since the results from the causality test may be sensitive to the lag structure, determining the
optimal lag length(s) is crucial for the robustness of the empirical foundings. In a large panel
system, lag lengths and numbers of independent variables can cause a substantial
computational burden. Following Kónya (2006), maximal lags are allowed to differ across
variables but need to be the same across equations. In our paper, the regression system is
estimated by each possible pair of 1ly , 1lx , 2ly , 2lx 1lz , and 2lz ; we assume 1 to 8 lags
exist, and then we choose the combinations that minimize the Schwarz Bayesian Criterion.5
b. Cross-Sectional Dependence Tests
One of the most important assumptions in the bootstrap panel causality is the existence
of cross-sectional dependence among the countries in the panel. In the case of
cross-sectionally correlated errors, the estimator from the regression system described with
the SUR is more efficient than the estimator with the pooled ordinary least squares (pooled
OLS) model because the country-by-country OLS approach does not consider cross-sectional
dependence. Therefore, testing for cross-sectional dependence is the most crucial issue for the
selection of an efficient estimator, and hence, for the panel causality results.
To test for cross-sectional dependence, the Lagrange multiplier (LM) test by Breusch
and Pagan (1980) has been extensively used in empirical studies. The LM procedure test
requires the estimation of the following panel data model:
it i i it ity x uα β ′= + + for 1,2,...,i N= ; 1,2,...,t T= (3)
4 As indicated by Kónya (2006), this is an important step because the causality test results may depend critically on the lag structure. In general, lag decisions may cause different estimation results. Too few lags means that some important variables are omitted from the model and this specification error will usually cause incorrect estimation in the retained regression coefficients, leading to biased results. On the other hand, too many lags will waste observations and this specification error will usually increase the standard errors of the estimated coefficients, leading to inefficient results. Based on Schwarz Bayesian Criterion, we found the optimal lag is 6 for our estimated model. 5 In order to save space, results from the lag selection procedure are not showed in the paper but are available upon the reader’s request.
11
In equation (3),ity is per capita economic growth (PEG), i is the cross-sectional dimension, t
is the time dimension, itx is 1k × vector of explanatory variable (such as Population growth
(POG)), and iα
and iβ are the individual intercepts and slope coefficients, respectively,
that are allowed to vary across countries. In the LM test, the null hypothesis of no-cross
sectional dependence, 0 : ( , ) 0it jtH Cov u u = , for all t and i j≠ - is tested against the
alternative hypothesis of cross-sectional dependence, 1 : ( , ) 0it jtH Cov u u ≠ , for at least one
pair of i j≠ . In order to test the null hypothesis, Breusch and Pagan (1980) developed the
LM test:
12
1 1
ˆN N
ij
i j i
LM T ρ−
= = +
= ∑∑ , (4)
where ijρ̂ is the sample estimate of the pair-wise correlation of the residuals from the
pooled OLS estimation of equation (3) for each i. Under the null hypothesis, the LM statistic
has an asymptotic chi-square with ( 1) / 2N N − degrees of freedom. It is important to note
that the LM test is valid for a relatively small N and a sufficiently large T. In the case of large
panels, for example, where ∞→T first and then ∞→N , Pesaran (2004) proposed a
scaled version of the LM test:
1/21
2
1 1
1ˆ( 1)
( 1)
N N
lm ij
i j i
CD TN N
ρ−
= = +
= −
− ∑∑ . (5)
Under the null hypothesis, the CDlm test converges to the standard normal distribution.
However, the CDlm test may be subject to substantial size distortions when N is large and T is
small. Pesaran (2004) developed a more general cross-sectional dependence test that is valid
for large panels; this CD test is:
1
1 1
2ˆ
( 1)
N N
ij
i j i
TCD
N Nρ
−
= = +
=
− ∑∑ . (6)
12
Under the null hypothesis, the CD test has an asymptotic standard normal distribution.
Pesaran (2004) indicated that the CD test has a mean that is exactly zero for fixed T and N,
and is robust for heterogeneous dynamic models that include multiple breaks in slope
coefficients and error variances, as long as the unconditional means of ity and
itx are
time-invariant and their innovations have symmetric distributions. However, the CD test will
lack power in certain situations in which the population average pair-wise correlations are
zero, but the underlying individual population pair-wise correlations are non-zero (Pesaran et
al., 2008). Pesaran et al. (2008) proposed a bias-adjusted test which is a modified version of
the LM test; it uses the exact mean and variance of the LM statistic. The bias-adjusted LM
test is as follows:
21
21 1
( )2ˆ
( 1)
N Nij Tij
adj ij
i j iTij
T kTLM
N N
ρ µρ
ν
−
= = +
− − =
− ∑∑
⌢
. (7)
In equation (7), Tijµ and
2
Tijν are the exact mean and variance of2( ) ijT k ρ−⌢
, respectively,
that are provided by Pesaran et al. (2008). Under the null hypothesis where first T→∞ and
then N→∞, the adjLM test is asymptotically distributed as a standard normal distribution.
c. Slope Homogeneity Tests
The second important aspect of the bootstrap panel causality approach is testing for
cross-country heterogeneity. Applying the Wald principal is to test the null hypothesis of slope
coefficient homogeneity against the alternative hypothesis. The Wald principle is valid for all
cases where the cross-sectional dimension (N) is relatively small and the time dimension (T)
of the panel is large6; the explanatory variables are strictly exogenous, and the error variances
are homoscedastic. Swamy (1970) developed the slope homogeneity test to detect
cross-sectional heteroscedasticity (Pesaran and Yamagata, 2008). Pesaran and Yamagata
(2008) proposed a standardized version of Swamy’s test (also called the ∆ɶ test) for testing
6 T > N is the basic requirement for our bootstrap panel causality test.
13
slope homogeneity in large panels. The ∆ɶ test is valid as ( , )N T → ∞ without any
restrictions on the relative expansion rates of N and T when the error terms are normally
distributed. In the ∆ɶ test approach, the first step is to compute the following modified
version of Swamy’s test:
( ) ( )21
Ni i
i WFE i WFE
i i
x M xS τβ β β β
σ=
′′= − −∑
⌢ ⌢ɶ ɶ ɶ
ɶ , (8)
where iβ⌢
is the estimator from the pooled OLS, WFEβɶ is the estimator from the weighted
fixed effect pooled estimation of the regression model in equation (3), Mτ is an identity
matrix, and 2
iσɶ is the estimator of 2
iσ .7 The standardized dispersion statistic is then defined
as
1
2
N S kN
k
− −∆ =
ɶɶ . (9)
Under the null hypothesis with the condition of ( , )N T → ∞ , so long as /N T → ∞ and
the error terms are normally distributed, the ∆ɶ test has an asymptotic standard normal
distribution. The small sample properties of the ∆ɶ test can be improved under normally
distributed errors by using the following bias-adjusted version:
1 ( )
var( )
itadj
it
N S E zN
z
− −∆ =
ɶ ɶɶ
ɶ, (10)
where the mean is ( )itE z k=ɶ and the variance is var( ) 2 ( 1) / 1itz k T k T= − − +ɶ .
5. Results and policy implications
a. Cross-sectional dependence and slope homogeneity
7 In order to save space, we refer to Pesaran and Yamagata (2008) for the details of Swamy’s test and the estimators described in equation (8).
14
As we outlined earlier, testing for the cross-sectional dependence and slope
homogeneity in the bootstrap panel causality analysis is crucial for selecting the appropriate
estimator and for imposing restrictions for causality because countries are highly integrated
due to a high degree of globalization in economic or financial relations. Therefore, our
empirical study starts with examining the existence of cross-sectional dependency and
heterogeneity across the countries under concern. To investigate the existence of
cross-sectional dependence, we carry out four different tests (BPCD ,
lmCD , CD , and
adjLM ), of which the results are presented in Table 5. The null of no cross-sectional
dependence is rejected at the conventional levels of significance, implying that the SUR
method is more appropriate than country-by-country OLS estimation, which is assumed in the
bootstrap panel causality approach. This founding implies that a shock occurring in one
country seems to be transmitted to other countries. The cross-sectional dependency
furthermore implies that examining causal linkages between population and economic growth
in these countries requires taking into account this information in estimations of causality
regressions. In the presence of cross-sectional dependency, the SUR approach is more
efficient than the country-by-country ordinary least-squares (OLS) method (Zellner, 1962).
Therefore, the causality results obtained from the SUR estimator developed by Zellner (1962)
will be more reliable than those obtained from the country-specific OLS estimations.
In Table 5 we also report the results from the slope homogeneity tests of Pesaran and
Yamagata (2008). Three tests ( ∆ɶ , adj∆ɶ , and Swamy Shat) all reject the null hypothesis of the
slope homogeneity hypothesis, supporting country-specific heterogeneity, with the exception
of the test of adj∆ɶ . This rejection implies that a panel causality analysis which imposes a
homogeneity restriction on the variable of interest results in misleading inferences. Therefore,
the direction of causal linkages between population growth and economic growth may differ
15
across the selected countries.
Both the cross-sectional dependency and the heterogeneity across the 21 countries
provide evidence for the suitability of the bootstrap panel causality approach.
b. Causality
The empirical results from the bootstrap panel Granger causality analysis are reported in
Table 6 and 78. These empirical findings have four major policy implications. First of all, in 5
out of 21 countries, i.e., Canada, Germany, Japan, Norway, and Switzerland, we found
evidence of one-way Granger causality running from economic growth to population growth,
implying that economic growth is of great importance for population growth in these four
countries. If we look at the sign of the coefficients, we found that three countries (i.e., Canada,
Norway, and Switzerland) with positive coefficients, indicating that for these three countries
economic growth has a positive impact on population growth. This argument implies that the
function of economic system may exert a large impact on the population development in these
three countries. On the other hand, we found the sign of the coefficients of the other two
countries (i.e., Germany and Japan) negativ. These results indicate that for these two countries,
economic growth has negative impact on population growth. One explanation for these results
could be that due to higher economic growth people get use to enjoying a wealthier life and
do not want to have more children causing fertility reductions.
Secondly, evidence shows one-way Granger causality running from population growth
to economic growth in Finland, France, Portugal, and Sweden indicating that population
growth does have an effect on economic growth. However, if we look at the sign of the
coefficients, we found that Finland, Portugal, and Sweden with negative coefficients. These
results indicate that for these three countries, population growth has negative impact on
economic growth. The negative impact of population growth on economic growth seems to
8 We refer to Kónya (2006) for explanations of the bootstrap procedure and how the country-specific
critical values are generated.
16
support the arguments of the Malthus (1798) where population growth is supposed to
decrease the per capita output, because output growth cannot keep up the at the same pace. In
order to keep the natural balance between population, food and consumption, preventive
checks (i.e. fertility reduction) and positive checks (i.e. mortality increase) on population
growth are necessary (Malthus, 1798). However, on the other hand, we found the sign of the
coefficient for France is positive. This result indicates that for France, population growth has
positive impact on economic growth. The positive impact of population growth on economic
growth seems to support the arguments of the Kremer (1993). Kremer (1993) has empirically
confirmed that a larger population was associated with higher population growth rates and
faster technological improvement. This technological development is a consequence of
population growth which leads to an increase of labor productivity, per capita income and
improvement in living standards.
Thirdly, we found bidirectional Granger causality between population growth and
economic growth in both Austria and Italy. These results suggest that for these two countries
the population growth and economic growth both are endogenous, indicating that they
mutually influence each other. Their mutual reinforcement has important implications for the
conduct of economic or population policies in both Austria and Italy. If we look at the sign of
the coefficients in both two equations, we found that for Italy population growth has a
positive affect on economic growth; however, economic growth has negative impact on
population growth. The positive impact of population growth on economic growth further
supports the arguments of the Kremer (1993). However, if we look at the sign of the
coefficients in both two equations for Austria, we found that population growth has a negative
affect economic growth, and economic growth also has negative impact on population growth.
The negative impact of population growth on economic growth further supports the
arguments of the Malthus (1798). These results demonstrate that rapid population growth is a
real problem in Austria because it contributes to lower investment growth and diminishes the
17
savings rate. Policy makers in Austria can address these serious economic consequences of
rapid population growth by investing in family planning services. Development of
independent media and liberal education in educational institutions will in time also help by
encouraging a smaller family size.
Fourth and finally, we found no causal relationship between population growth and
economic growth is found in Belgium, Brazil, Denmark, Netherlands, New Zealand, Spain,
Sri Lanka, the UK, the USA and Uruguay. These results support the neutrality hypothesis for
the population-income nexus, which indicates that population growth and economic growth
may not influence each other. For example, in these 10 countries, an economic policy may not
be effective for population growth, while a population policy may also have no impact on
economic growth since the results show no evidence of the relationship between population
growth and economic growth in these two countries.
c. Robustness check
Since our sample is quite long with economies having undergone tremendous transition
both in terms of economic growth and population growth, we decided to take cross-sectional
average for both economic and population growth rates and applied the CUSUM test to the
two time series of averages across the 21 countries. We found a structural break in 1952,
which is not surprising given the high growth rates in both population and GDP witnessed
after World War II. Therefore, we divided the total sample into two sub-sample periods,
1871-1951 and 1952-2013, as a robustness check. We reported the 1871-1951 results in
Tables 8 and 9. Based on the empirical results from Tables 8 and 9, we found that population
growth Granger cause economic growth for Finland and France. We also found a relationship
from economics growth to population growth for Denmark, Japan, and Norway and
bidirectional causality between population growth and economic growth for both Austria and
Italy. For the rest of 14 countries (i.e., Belgium, Brazil, Canada, Germany, Netherlands, New
18
Zealand, Portugal, Spain, Sri Lanka, Sweden, Switzerland, the UK, the USA, and Uruguay)
we found no causality between population growth and economic growth.
If we look at the sign of the coefficients from Table 8 for 1870-1951, we can see that
population growth has a significant negative effect on economic growth for Finland. For
France we found that population growth has a significant positive effect on economic growth.
We also found that economic growth has a significant negative impact on population growth
for both Denmark and Japan based on the negative sign of the coefficients. However, if we
look at Norway, we find that the sign of the coefficient of economic growth on population
growth is significantly positive. This indicates that for Norway, economic growth has positive
impact on population growth. For Austria and Italy there exist bidirectional causality between
population growth and economic growth, however the sign coefficients differ. On one hand,
we found that population growth has a positive effect on economic growth, however
economic growth has a negative impact on population growth in Italy. For Austria on the
other hand, we found that population growth has a negative effect on economic growth, and
economic growth also has a negative impact on population growth.
Results for 1952-2013 are reported in Tables 10 and 11. For this time period we found
that population growth Granger cause economic growth only for Sri Lanka and that economic
growth Granger cause population growth for Belgium, Denmark, France, Germany, New
Zealand, Spain, Switzerland, and Uruguay. We also found bidirectional causality between
population growth and economic growth only for Japan. For the rest of 11 countries (i.e.,
Austria, Brazil, Canada, Finland, Italy, Netherlands, Norway, Portugal, Sweden, the UK, and
the USA), we found no causality between population growth and economic growth.
If we look at the sign of the coefficient for Sri Lanka we see a negative impact from
population growth on economic growth. An opposite relationship from economic growth to
population growth was found for Belgium, Denmark, France, Germany, New Zealand, Spain,
Switzerland, and Uruguay, indicating that economic growth is of great importance for
19
population growth in these 8 countries. We found that the sign of the coefficients for all
countries are significantly positive, with the exception of Uruguay. Looking at the coefficients
of both equations for Japan we found that population growth has a positive affect on
economic growth and economic growth also has a positive impact on population growth.
6. Conclusions
This study applies the bootstrap panel causality test proposed by Kónya (2006) to test the
causal link between population growth and economic growth in 21 countries over the period
of 1870-2013. The bootstrap panel causality test, which accounts for dependency and
heterogeneity across countries, supports evidence on the direction of causality. Regarding the
direction of population growth-economic growth nexus, we found one-way Granger causality
running from population growth to economic growth for Finland, France, Portugal, and
Sweden and one-way Granger causality running from economic growth to population growth
for Canada, Germany, Japan, Norway, and Switzerland and no causal relationship between
population growth and economic growth is found in Belgium, Brazil, Denmark, Netherlands,
New Zealand, Spain, Sri Lanka, the UK, the USA and Uruguay. Furthermore, we found
feedback between population growth and economic growth for Austria and Italy. Due to a
structural break in 1952 we also divided the sample into two subsamples which provides
conflicting results. For the period of 1871-1951 we found that population growth Granger
cause economic growth for Finland and France and a relationship from economics growth to
population growth exists for Denmark, Japan, and Norway and that there is bidirectional
causality between population growth and economic growth present for both Austria and Italy.
For the more recent time period of 1952-2013 we found that population growth Granger cause
economic growth only for Sri Lanka and that economic growth Granger cause population
growth for Belgium, Denmark, France, Germany, New Zealand, Spain, Switzerland, and
Uruguay and bidirectional causality between population growth and economic growth only
for Japan. Due to the differences in the existence and direction of causality between countries
and across time periods our results provides important policy implications for these 21
countries.
20
References
Afzal, M. (2009). Population growth and economic development in Pakistan, The Open
Demography Journal, 2, 1-17.
Avakov, Alexander V. (2010). Two Thousand Years of Economic Statistics: World Population,
GDP and PPP, Algora Publishing.
Birdsall, N.; Allen C. Kelley; and Steven W. Sinding (2001 eds.): Population Matters:
Demographic Change, Economic Growth, and Poverty in the Developing World. New
York: Oxford University Press.
Breusch, T. S., & Pagan, A. R. (1980). The Lagrange Multiplier test and its applications to
model specification in econometrics. The Review of Economic Studies, 47(1), 239-253.
Cassen, R. (1976) Population and development: a survey, World Development, 4, 785-830.
Darrat, A. F, Al-Yousif, Y. K. (1999). On the long-run relationship between population and
economic growth; some time series analysis evidence for developing countries, Eastern
Economic Journal, 25, 310-312.
Dasgupta, P. (2000). Population and Resources: An Exploration of Reproductive and
Environmental Externalities. Population and Development Review, 26: 643–689.
Drèze, J. and Murthi, M. (2001). Fertility, Education, and Development: Evidence from India.
Population and Development Review, 27: 33–63.
E a s t e r l i n , R i ch a r d A . ( 1 9 6 7 ) . E f f e c t s o f P o p u l a t i o n Gr o wt h o n t h e
E c o n o m i c D e v e l o p m e n t o f 18 Developing Countries. Annals of the American
Academy of Political and Social Science 369:98-108.
Espenshade, Thomas J. (1978). Zero Population Growth and the Economies of Developed
Nations. Population and Development Review 4, no. 4: 645-680.
Galor, O., Weil D. (1996). The gender gap, fertility, and growth, American Economic Review,
86, 374-387.
Galor, O., Weil D. (2000). Population, technology and growth: from Malthus stagnation to the
demographic transition and beyond, American Economic Review, 90, 806-828.
Huang, T-H., Xie Zixiong. (2013). Population and economic growth: a simultaneous equation
perspective, Applied Economics, 45, 3820-3826.
Hurlin, C. (2008). Testing for Granger Non Causality in Heterogeneous Panels, Mimeo.
Department of Economics: University of Orleans.
Kónya, L. (2006). Exports and growth: Granger causality analysis on OECD countries
with a panel data approach. Economic Modelling, 23, 978–992.
Kremer, Michael. (1993). Population Growth and Technological Change: One Million B.C. to
21
1990. The Quarterly Journal of Economics 108, no.3:681
Kuznets, S. (1955). Economic Growth and Income Inequality. American Economic Review.
(March). 45(1): 1-28
Kuznets, Simon. (1976). Demographic aspects of the size distribution ofincome. Economic
Development and Cultural Change 25, no. 1 (October):1-94.
Kuznets, Simon, Richard E. Quandt, and Milton Friedman. (1960). Population Change and
AggregateO u t p u t . In Demographic and Economic Change in Developed Countries,
Universities- National Bureau, I:340 - 367. New York: National Bureau of Economic
Research.
Li, H, Zhang, J. (2007). Do high birth rates hamper economic growth? Review of Economics
and Statistics, 30, 67-80.
Luigi Palumbo, Ferrara Alberto, Varra Pasquale. (2010). Population and Economic Growth:
Evidence from EU and Emerging Countries, Gottingen Summer School 2010, Relations
between the EU and the emerging global players.
Malthus, T.R. (1798). An Essay on the Principle of Population, J. Johnson, London.
Mason, Andrew. (1988). Saving, Economic Growth, and Demographic Change.
Population and Development Review 14, no. 1: 113-144.
McNicoll, G. (1984). Consequences of rapid population growth: an overview and assessment,
Population and Development Review, 10, 177-240.
Pesaran, M. H. (2004). General diagnostic tests for cross section dependence in panels.
Cambridge Working Papers in Economics No. 0435, Faculty of Economics, University
of Cambridge.
Pesaran, M. H., Ullah, A., & Yamagata, T. (2008). A bias-adjusted LM test of error