Technology and Investment, 2013, 4, 67-72
Published Online Febr uary 2013 (http://www.SciRP.org/journal/ti)
Copyright © 2013 SciRes. TI
Do Redistributive Policies Affect Economic Growth?
——An Empirical Study based on Canadian data
Zheng Liu
Department of Economics, Concordia University, Montreal, Canada
Email: jacquesliu@ gmail.com
Received 2012
ABSTRACT
To reduce income inequality, redistributive policies are widely adopted by both federal and provincial governments in
Canada. Quebec and Canada have a fairer society in OECD countries. However, their economic growth is slower than
many other countries. This paper studies how these redistributive policies affect economic growth based on Canadian
data for the first time. The growth model is based on standard augmented Solow model and includes several different
self-defined policy indexes. Using high quality panel data spanning the period 1982 to 2009 calculated from statistic
Canada’s website and Arellano-Bond panel technique, empirical analyses show that redistributive policy is negatively
and significantly associated with economic growth. These findings are in accordance with many former literatures and
may have important policy significance.
Keywords: Income Inequality and Economic Growth
1. Introduction
Canada, especially Quebec, is one of the best areas
around the world on reducing income inequality. How-
ever, there is a debate whether Canadian benefit from a
more equal society.
The key of the debate is whether income inequality
will affect economic growth, which is a classical eco-
nomic question. Since Nobel laureate Simon Kuznets
introduced his Kuznet’s curve which shows inequality
first rises and later falls with economic growth, many
economists have made their contribution on this topic
theoretically or empirically. These studies either find a
negative or a positive coefficient on inequality. Alesina
and Rodrik (1994), Benabou (1996), and William Eas-
terly 2006, etc conclude inequality does cause underde-
velopment. However, Li and Zou (1998), Kristin Forbers
(2000), Mark Patridge (2004), Mark Frank( 2009) and
many authors argue that inequality is not harmful for
gro wth. Robert Barro (1999) fol lows Kuznet’s tradition
and concludes that inequality retards growth in poor
countries but encourages growth in richer places. Qual
(2001) also finds little relationship between inequality
and growth.
Why income inequality has negative or positive influ-
ence on economic growth? Many theories have been
proposed. On negative side, first, credit-market imper-
fection models suggest a transfer payment from rich to
poor raises the average productivity of investment. In the
presence of asymmetric information and limited access to
credit, investments favor people with more assets. Se-
condly, high income inequality in a society will often
bring social unrest even revolution. Poor people are easy
to engage in crime, riots and disruptive activities which
are a waste of social resources. Therefore, higher ineq ua-
lity means more social resource waste and lower eco-
nomic growth. Thirdly, under Alesina and Rodrik’s polit-
ical economy model, a system of majority voting tends to
redistributive policies from rich to poor and many public
expenditure programs. Poor people will lose interest to
work again and rich people will have to prevent this redi-
stributive policy through bribery. O ft e n there is no
enough investment in the economy and economic growth
slows down.
There are also many theories on positive relationship.
First, inequality brings higher level work effort and en-
trepreneurial energy and therefore promotes economic
growth. Secondly, under political economy model, if
public expenditure is mainly about public education
which will increase aggregate human capital, this will
increase economic growth. A third model argues that
inequality enhances mobility and the concentration of
high -ability workers in technologically advanced sectors;
therefore promote technological progress and economic
growth. Another model is about saving rates. Some
economists believe that individual saving rates rise with
the level of income. A transfer payment from rich to poor
means lower aggregate rate of saving in an economy and
Z. LIU
Copyright © 2013 SciRes. TI
reduce economic growth.
Clearly, no either of above theories can fully explain the
effect of income inequality on economic growth. This
point accords with the ambiguous empirical findings in
many literatures. Level of development, political institu-
tions, and many other factors will decide whether it is
positive or negative relation between income inequality
and economic growth.
This paper will take Canada as an example. Veall
(2012) in his most recent paper concludes that the surge
in top share incomes in Canada over the last 30 years is
clear. At the same time, redistributive policies namely,
mainly taxes and transfers are widely used by both
federal and provincial governments. The effect of redi-
stributive policies of Canadian governments on economic
growth is studied for the first time (according to our sur-
vey) in this paper. Section 2 explains our growth model
and dataset. Section 3 estimates this model based on the
panel technique developed by Allreno and Bond. Section
4 is some disc uss io n followed by a simple conclusion.
2. Growth Model and Dataset
This paper estimates growth as a function of redistribu-
tive policy index, lagge d per capita income, investment
share, education, and working population growth rate
using a model similar to the augmented Solow model in
N.Gregory Mankiw, David Romer and David N. Weil’s
(MRW ) 1992 paper. Besides using per capita GDP
growth rate instead of per capita income as the dependant
variable, the only other change from MRW’s original
model is the addition of policy index.
The growth model central to this paper is
it1i,t 12i,t
3i,t 4i,t
5i,t i,t
GrowthβIncomeβEducation
βInvestmentβPopulation
βRedistributive_Policyu
= +
++
++
(1)
where represents each province in Canada and t
represents each time period; is average an-
nual growth rate of per capita GDP for province dur-
ing period t; is per capita GDP for prov-
ince in time period t-1; Unlike MRW’s paper using the
percentage of the working-age population that is in sec-
ondary school as education index, this paper will use a
new education index. In developed country such as Can-
ada, there is very little difference on secondary school
enrollment rate for each province. Therefore, it is not a
good education index again. We decide to use school
board expenditure as the new education index in this pa-
per because it is easy to acquire long span data .
As in MRW’s paper, this paper uses average share of
real investment (including government investment) in
real GDP as investment index and average rate of growth
of the working-age population (15-64) as population in-
dex. The same as in MRW’s paper, we use ngδ
++
(Assu m ing
gδ0.05+=
) to be the population item.
Several redistributive policy indexes are used in this
paper. The first one is change of Gini coefficient before
and after execution of redistributive policies. Gini coeffi-
cient is widely used in many inequality-growth literatures.
Traditionally, redistribution could be achieved by many
methods like progressive income tax, capital tax, mini-
mum wage laws, public expenditure of government and
others. Gini coefficient, as a good indicator of income
distribution, reflects result of redistributive policies di-
rectly, especially the change of Gini coefficient before
and after execution of redistributive policies.
When there is no accurate Gini coefficient data availa-
ble, which is possible in many developing countries, two
other policy indexes are defined based on income distri-
bution change. Figure 1 gives an illustration of income
distribution change using data from Alberta, Ontario and
Queb ec in three different periods. Y-axis shows the ratio
of change and X-axis shows quintile. The figure clearly
shows Quebec has much stronger redistributive policy
than Alberta. Based on figure 1, the second and the third
policy index are the average slope and the regression slop
(from polyfit operation) of the income distribution
change line respectively. All three policy indexes are set
to be positive.
This paper will use all these different policy indexes
and give a comparison. At last, is the error term of
the model.
-.5 0.5 11.5-.5 0.5 11.5
20 40 60 80 1 00
20406080 100
QC ON
AB
1981 1995
2006
quntile
Graphs by Province
Figure 1: Illustration of income distribution change
Empirical studies of income inequality/economic growth
problem are often limited by the available data. Most of
the inequality/economic growth literatures use nations’
data to do regression. Patridge (1997) and Frank (2009)
analyze data from states in the U.S.A. Several advantages
to using states to examine inequality/growth issues are
believed by the authors. First, there appears to be suffi-
68
Z. LIU
Copyright © 2013 SciRes. TI
cient variation in income distribution to produce diffe-
rential economic outcomes. Secondly, large factor flows
between states should magnify how small disparities in
initial conditions affect economic growth. Thirdly, the
great homogeneity of state-level data helps mitigate the
possible omitted variable bias. Similarly, this paper uses
data from 10 provinces from 1982 to 2009 in Canada. All
data are from website of Statistics Canada (CANSIM).
Clearly, high quality is another important advantage to
use data from CANSIM.
3. Empirical Results
Empirical study of effect of inequality on economic
growth often bases on different regression methods.
There are either OLS regressions over a cross-section of
nations or panel techniques over both periods and geo-
graphic regions. Forbes(2000) once pointed out to eva-
luate which technique is optimal, there are three factors:
(1) the relationship between the region-specific effect
and the regressors; (2) the presence of a lagged endo-
genous variable; (3) and the potential endogeneity of the
other regressors. The same as Forbes, this paper will use
Arellano-Bond method to study the income inequali-
ty/economic growth effect. Standard panel techniques
like fixed effects and random effects have an important
problem that there is a lagged endogenous income varia-
ble in the growth model. Manuel Arellano and Stephen
R.Bond (1991) developed a special panel technique
which corrects not only for the bias introduced by the
lagged endogenous variable, but also permits a certain
degree of endogeneity in the other regressors. This me-
thod can eliminate the region-specific effect and then
uses all possible lagged values of each of the variables as
instruments through first-difference. Table 1 lists the
estimation result.
As Table 1 shows, the regression coefficients of all
three policy indexes are negative and significant. In all
three cases, the higher policy-index means more transfer
payment for the poor and higher tax for the rich.
This result is consistent with the simple theory model in
Li and Zou (1998)’s paper. When public consumption
enters utility function, income inequality has a positive
relationship with economic growth. In Canada, especially
in Quebec, government spends a lot on public services
like education, public health, pension plan etc. It is easy
for residents to lose incentive to work harder and it is
hard for government to find enough capital to invest for
economic growth.
Comparing three policy indexes from column 2 to
column 4, no much difference is found although PC1 has
a better Sargan test result. This shows Gini change is a
better policy index compared to others in our regression.
This result is also consistent with former panel data lite-
ratures where income inequality has a positive relation-
ship with economic growth. As a comparison, column 5
gives the positive regression result on Gini coefficient in
our model.
For the other explanatory variables, all the regression
coefficients are significant. The coefficients of lagged
per capita GDP and population growth rate are negative,
and the coefficients of education and investment are pos-
itive. These results are consistent with results of aug-
mented Solow model and other researchers’ findings.
In Allreno-Bond estimation, instruments setup is very
important in order to get the ideal result. To improve
efficiency of instruments, first is to choose endogenous
and exogenous variables. In this paper we choose lagged
per capita GDP and investment share to be endogenous
and policy index as exogenous variable. Secondly, “sys-
tem GMM” is introduced to further improvement instru-
ments. The “system GMM” estimator uses the levels
equation to obtain a system of two equations: one diffe-
renced and one in levels. By adding the second equation
additional instruments can be obtained. Third, there is a
rule of thumb in Allreno-Bond method which is to keep
the number of instruments less than or equal to the num-
ber of groups. In this paper, because we have only 10
provinces (groups), we must be very careful on selection
of the instruments. Here “collapse” inside gmm() option
is used to reduce the number of instrument s.
Sensitivity analysis is usually an important step to
examine the robustness of baseline regression results. To
do sensitivity analysis, more variables are required to be
added into the growth model. In this paper, only two
more sensitive variables, international trade share and
R&D expenditure share over GDP share, are considered.
In step 1, both two sensitive variables are included. In
step 2 and 3, only one sensitive variable is added each
time.
Table 2 shows results of the simple sensitivity analy-
sis. When sensitive variables are added, first, better Sar-
gan test value shows that instruments setup improves.
Secondly, more variables including sensitive variables
become insignificant. Thirdly, the regression coefficients
of policy index keeps negative and significant. This find-
ing seems to confirm that the negative relationship between
income inequality and economic growth is robust.
4. Discussion
It is important to note that there are still several questions
for further discussion. These will be explained in this
section.
(1) Measurement-error and omitted variable
bias
Forbes (2000) pointed out that both measurement error
and omitted variable bias will bring direction bias on the
estimated policy index. Measurement error will also re-
69
Z. LIU
Copyright © 2013 SciRes. TI
Table 1: Baseline Regression Results
Number of obs = 270 Number of groups = 10 Number of instruments = 10
---------------------------------------------------------------------------------------------------------------------------------
Growth Rate| PC1 PC2 PC3 Gini
-------------+------------------------------------------------------------------------------------------------------------------
GDP(-1) | -.2383283(0.000) -.2064298(0.000) -.2037915(0.000) -.3968854(0.000)
In(School) | .0269738(0.006) .0226871(0.017) .0219274(0.023) .0376422(0.000)
In(I/GDP) | .1230517 (0.000) .0968017(0.003) .0945082(0.003) .2088573(0.000)
In(n+ g+ δ)| -.2021602(0.009) -.1073335(0.055) -.1108057(0.050) -.1986347(0.004)
In(PC) | -.1314319(0.001) -.0548954(0.000) -.0572575(0.000) 1.398316(0.001)
_cons | 1.530564(0.000) 1.677499(0.000) 1.734619(0.000) 2.97497(0.000)
Sargan test| 0.207 0.116 0.107 0.948
---------------------------------------------------------------------------------------------------------------------------------
Note: P-values are In parentheses. g+δis assumed to be 0.05.
Sargan test is used to test if the null hypothesis of “the instruments as a group are exogenous” is true. Therefore, the
higher the value the better.
Table 2: Sensitive Analysis
------------------------------------------------------------------------------------------------------------
Number of obs = 270 Number of groups = 10 Number of instruments = 10
------------------------------------------------------------------------------------------------------------
Growth Rate | step 1 step 2 step 3
-------------+----------------------------------------------------------------------------------------------
GDP(-1) | -.4765018(0.043) -.5322517(0.012) -.2352878(0.000)
In(Sc ho o l ) | .1402053(0.217) .0576037(0.022) .1543243(0.137)
In(I/GDP) | -.0393416(0.842) -.1069355(0.520) .1600455(0.003)
In(n+ g+ δ)| -.3579005(0.077) -.2326544 (0.030) -.3910543(0.034)
In(PC) | -.483134(0.044) -.4332232(0.046) -.2830407(0.035)
In(trade) | .4726888(0.292) .5782205(0.150)
In(R&D) | -.0862684(0.454) -.124573(0.216)
_cons | 1.536825(0.532) 3.099894(0.012) -.3129374(0.843)
Sargan Test | 0.837 0.79 0.619
Note: P-values are in parentheses. g+δis assumed to be 0.05.
duce the significance of results. This paper addresses
these issues by selection of high quality data and panel
technique.
All data used in our regression are calculated from ba-
sic data of Statistics Canada’s database. The homogenei-
ty of these province-level data helps mitigate the omitted
variable problem because provinces have relatively simi-
lar growth mechanisms and institutions. Also these data
can be more accurate than counterparts in other
cross-nation dataset which includes many data from de-
veloping countries.
Panel technique is also used to reduce omitted-variable
bias. A key advantage of our Allreno-Bond fixed-effects
model is to control for any time-invariant omitted va-
riables. Especially, as Forbes (2000) pointed out that
panel techniques can specifically estimate how redistri-
butive policies predict economic growth rate because
they use within-province time series variation.
(2) Small N Large T panel problem
The Arellano-Bond estimator was designed for small-T
large -N panels. In large-T panels, the correlation of the
lagged dependent variable with the error term will be
insignificant. (Roodman, 2006) However, the panel in
this paper is a typical small-N large T panel. This is the
reason that a “collapse” option has to be used in the Al-
lreno -Bond command. Otherwise, there will be too many
instruments in the regression which bring insignificant
resul t.
As Forbes (2000) and Patridge (2004) pointed out, one
reason for the conflicting inequality-growth results in the
literature is the different time periods in the study. Short
run (5-10 years) methods like fixed-effect panel estima-
tion and long run (25-30 years) methods like OLS can
have different response. This paper uses a short run tech-
nique to study inequality/growth relationship in a long
period. Further reassessment is needed to study if there
are any negative effects using this method. However,
because similar study is limited by only 10 provinces in
Canada, other small-N large-T panel technique like
SURE is worth being considered.
70
Z. LIU
Copyright © 2013 SciRes. TI
(3) Looking for negative factor
Equality has long been considered very important to the
pursuit of long-term prosperity in aggregate term for so-
ciety as a whole by World Bank. Easterly (2006) con-
cludes that structural inequality caused by conquest, co-
lonization, slavery and land distribution is unambi-
guously bad for economic growth. Only inequality made
by market forces may have positive relationship with
economic growth. Although we can say it is mainly
market inequality in a society like Canada, it is interest-
ing to find the negative factor of inequality on economic
growth.
One channel that suggests income equality can pro-
mote economic growth is human capital accumulation
through better education opportunities, more creativity.
However, inequality also encourages individuals to in-
vest more on education for a better future. In our growth
model, it is not easy to find a good proxy for human cap-
ital. This paper uses school board expenditure to
represent education. However, some better proxies like
share of university degree holders over population do not
support long span period. How to find a proxy to de-
scribe a worker’s education, creativity, incentive to work
and others related to human capital therefore is an im-
portant task for further study of this inequality/growth
relationship, especially for long run study.
5. Conclusion
Richard Wilkinson and Kate Pickett in their famous 2009
book ”The Spirit Level: Why More Equal Societies Al-
most Always Do Better” concluded that countries that
are most equal do best. This paper is motivated by the
desire to provide an answer to the question if a more
equal society does good to economic growth. In real life,
on the contrary, a fairer society like Quebec has lower
economic growth rate than societies with higher Gini
coefficient like Alberta. What is the reason behind this?
Many former literatures focus on cross-countries dataset
or U.S state level data. However, these data may either
have measurement error problem or are not suitable to
Canadian experience. This paper, for the first time as we
believe, constructs a Canadian income inequali-
ty/economic growth dataset and draws a negative rela-
tionship between redistributive policy and economic
growth. This result is consistent to other literatures and
the simple theory model introduced by Li and Zou
(1998).
The Allreno-Bond panel technique used in this paper
can help to mitigate omitted variable bias and is useful
for policy analysis in nature. However, the small-N
large -T panel in our regression will bring some potential
problems to us. Further reassessment is needed.
Income inequality and economic growth generally
have an ambiguous relationship. What we find in this
paper depends on our estimation technique, our dataset.
An important direction for further study is to find a good
proxy for human capital accumulation. This will be the
key to answer if a fairer society can promote economic
growth in long run.
REFERENCES
[1] Arellano, Manuel and Bond, Stephen R. (1991) “Some
Tests of Specification for Panel Data: Monte Carlo Evi-
dence and an Application to Employment Equations.”
Review of Economic Studies, 58(2),pp.277-97
[2] Barro, R. J.(2000,”Inequality and Growth in a Panel of
Countries”, Journal of Economic Growth, 5,5-32
[3] Benabou, Roland(1996), “Inequality and Growth”, in Ben
Bernanke and Julio Rotemberg(eds), National Bureau of
Economic Research Macroeconomics Annual, Cambridge:
MIT Press, pp.11-74
[4] Easterly, William(2002), “Inequality does Cause Under-
development: New Evidence,” Working Paper 1, Center
for Global Development, Washington DC.
[5] Forbes, K.J., (2000) “A Reassessment of the Relationship
Between Inequality and Growth”, American Economic
Review, Vol. 90, pp.869-887
[6] Frank, M.W.(2009). "Inequality And Growth In The
United States: Evidence From A New State-Level Panel
Of Income Inequality Measures," Economic Inquiry,
Western Economic Association International, vol. 47(1),
pages 55-68, 01.
[7] Frank, M.W.(2009). "Income Inequality, Human Capital,
and Income Growth: Evidence from a State-Level VAR
Analysis," Atlantic Economic Journal, International At-
lantic Economic Society, vol. 37(2), pages 173-185, June.
[8] Kuznets, S. (1955). “Economic Growth and Income In-
equalit”, American Economic Review, 45, pp. 1-28
[9] Li, Hongyi and Heng-fu Zou. (1998). “Income Inequality
is not Harmful for Growth: Theory and Evidence”, Re-
view of Development Economics, 2(3): 318-334.
[10] Mankiw, N.G, David Romer and David N. W,1992)“ A
Contribution to the Empirics of Economic Growth, The
Quarterly Journal of Economics, Vol. 107, No. 2.,
pp407-437
[11] Mileva, Elitza (2007), “Using Arellano-Bond Dynamic
Panel GMM Estimators in Stata”
[12] Patridge, M. D.(2005). “Does Income Distribution Affect
U.S.state Economic Growth”, Journal of Regional
Science, 45(2),pp. 363-394
[13] Quah, D. (2001), “Some simple arithmetic on how in-
come inequality and economic growth matter”, London
School of Economics Working Paper
[14] Ravallion, Martin, 2004. "Pro-poor growth : A primer,"
Policy Research Working Paper Series 3242, The World
Bank.
[15] Ravallion, Martin and Chen, SHaohua(2003), “Measuring
Pro-Poor Growth”, Economics Letters, 78(1), 93-99
71
Z. LIU
Copyright © 2013 SciRes. TI
[16] Rodrik, Dani, (2005) “Why We Learn Nothing from Re-
gressing Economic Growth on Policies”,
[17] Roodman, David (2006) “How to Do xtabond2: An In-
troduction to “Difference”and “System” GMM in Stata”,
[18] Sims, C.A. (1980). “Macroeconomics and reality”, Eco-
nometrica, 48(1), 1-48.
[19] Veall,Michael R. “Top income shares in Canada: recent
trends and policy implications”, Canadian Journal of
Economics, Vol 45, No.4 pp1247-1272
72