Does Consensus Democracy Improve Economic Outcomes? · Democracy (2012), Lijphart recommends the...
Transcript of Does Consensus Democracy Improve Economic Outcomes? · Democracy (2012), Lijphart recommends the...
Does Consensus Democracy Improve Economic Outcomes?
Anna Colley
Worcester College, Oxford University
March 2018
Introduction
Of unparalleled importance to the framers of democratic constitutions is the relationship
obtaining between the nature of the institutions adopted and performance. In Patterns of
Democracy (2012), Lijphart recommends the consensus model of democracy over the
majoritarian model in virtue of delivering superior economic outcomes. Yet, in this essay, I
will reject Lijphart’s conclusion that consensus democracy improves economic outcomes by
arguing that the causal relationship he identifies is spurious. Firstly, I will address Lijphart’s
ad hoc identification of outliers and confounding variables. Once all outliers are excluded and
economic openness controlled for, the statistical significance he finds is undermined.
Subsequently, I will question the appropriateness of his inclusion of interest group
corporatism as a component of consensus democracy. And, lastly, I will suggest that
Lijphart’s case selection is to his advantage, for no robust relationship between consensus
democracy and economic outcomes is established amongst a sample of ten younger
democracies.
Definitions
Lijphart’s typology of democracies is simple: majoritarian democracies are governed by
majority rule, whereas consensus democracies seek to maximise agreement. Following
Lijphart, I will focus exclusively on the executives-parties dimension of his majoritarian-
consensus dichotomy1. To operationalise ‘economic outcomes’, I will ignore the growth rate
of Gross Domestic Product on the basis that Lijphart’s empirical analysis found no statistical
significance between this and the executives-parties dimension and will look solely at
inflation, as measured by the Consumer Price Index, and unemployment.
Theoretical Argument
The theoretical basis for Lijphart’s claim is far from implausible. Consensus democracies
tend to contain a greater number of veto players than majoritarian democracies; hence, their
economic policies are likely to find greater stability and broader support (Lijphart, 2012;
Tsebelis, 2002). Both of these attributes go some way to justify ascribing superior economic
outcomes to consensus democracies.
1 Lijphart’s executives-parties dimension is comprised of five variables. A consensus democracy has a dominant
legislative, a multiparty system, a proportional electoral system, corporatist interest groups, and unconcentrated
executive.
In the first place, it is not unreasonable to suppose that consensus democracies will
outperform majoritarian democracies in key indicators of economic performance in virtue of
having more stable economic policies. Sharp reversals of economic policy undeniably have
damaging repercussions on the rates of inflation and unemployment; and such reversals are
not uncommon amongst majoritarian democracies, where the two dominant parties take turns
as government and opposition. These reversals, or rather potential reversals, translate into
uncertainty as legislative elections draw near, too. And as Canes-Wrone and Ponce de Leon
note, uncertainty surrounding economic policy delays investment, particularly in capital and
other sunk costs (Canes-Wrone & Ponce de Leon, 2014). This is then associated both with a
reduction in the creation of new employment opportunities and with a reduction in the growth
of output that generally tempers price increases. Yet this is not the only way in which the
detrimental effect had on economic outcomes by the volatility of economic policy within
majoritarian democracies is exacerbated. Policymakers face a greater temptation to use the
fiscal tools at their disposal to stimulate a short term economic boom such that they stand a
higher chance of re-election, too. Myopically, suboptimal short term economic policies are
thus favoured (Aisen & Veiga, 2011).
In the second place, there is certainly some force behind Lijphart’s argument that the
decisions concerning economic policy reached by the governments of consensus democracies
are likely to be more successful than those of their majoritarian counterparts in virtue of
representing a greater section of society (Lijphart, 2012).
But there are strong counterarguments to consider.
The many veto players within consensus democracies indeed generate a demand for broad
agreement; but they do not ensure that this is supplied (Anderson, 2001). It is therefore quite
possible for consensus democracy to be altogether rather undeserving of its name. It is, after
all, obvious that the more veto players there are, the more difficult it becomes to reach an
agreement satisfactory to all. Less obvious, but no less important for economic outcomes, is
the decrease in the relative size of each political grouping. This increases their motive to
pursue particularist interests. Hence, consensus democracy need not necessary entail more
effective representation, nor more durable and acceptable economic policies.
Beer’s maxim ‘representative government must not only represent, it must also govern’
captures the traditional hypothesis that majoritarian democracies deliver more effective
policies than consensus democracies (Beer, 1998). Much can be said in favour of this
hypothesis; indeed, much literature has, with Lowell going so far as to state his ‘axiom’ that
one-party cabinets are required for effective policy (Lowell, 1896). One supportive point
meriting mention here is that majoritarian democracies can respond more quickly to
exogenous aggregate demand and supply shocks.
There are, then, strong theoretical arguments both for and against Lijphart’s claim. Whether
or not consensus democracies really do outperform majoritarian democracies in their control
of inflation and unemployment consequently becomes a purely empirical matter. Prima facie,
the statistically significant relationship Lijphart finds between these variables appears to
settle the dispute. But this can be doubted. I propose to do just this by probing his treatment
of outliers, control variables, components, and cases in turn.
Empirical Analysis
A: Lijphart’s Findings
Lijphart’s empirical analysis finds a statistically significant relationship between both
inflation and unemployment and consensus democracy. For every one percentage point
increase in the extent to which a country fits the consensus model of democracy, his
multivariate regression indicates a fall in inflation by almost one-and-a-half percent and a fall
in unemployment by nearly two percent.
B: Identifying and Excluding Outliers
However, Lijphart gives no procedural justification for identifying Israel and Uruguay as
outliers with respect to inflation and seemingly omits to consider whether any democracy’s
unemployment ought to be treated as an outlier.
As figure 1 demonstrates, in fact a third country, namely, Jamaica, has an average annual
inflation rate that should be excluded from Lijphart’s analysis; and, regarding unemployment,
Jamaica and Spain are outliers and as such should be excluded.
Figure 1: Boxplots depicting inflation and unemployment for Lijphart’s thirty-six
democracies
Table 1: Lijphart's multivariate regression
Dependent variable:
CPI 1981-2009 Unemployment 1981-2009 (1) (2)
Executive-parties dimension 1981-2009 -1.480** -1.794* (0.607) (0.929)
Logged population in thousands 2009 -0.800** -0.468 (0.356) (0.676)
HDI 2010 -22.557*** -11.290 (6.057) (15.227)
Constant 32.348*** 22.792** (6.010) (10.518)
Observations 26 20
R2 0.592 0.369
Adjusted R2 0.537 0.251
Residual Std. Error 2.952 (df = 22) 3.247 (df = 16)
F Statistic 10.659*** (df = 3; 22) 3.124* (df = 3; 16)
Note: *p<0.1; **p<0.05; ***p<0.01
Removing these outliers, the relationship between consensus democracy and unemployment
is rendered statistically insignificant. Lijphart’s claim that consensus democracy improves
economic outcomes is no longer valid; at best, he can claim that consensus democracy
improves the control of one economic outcome, inflation. And, while still meaningful, the
impact seemingly had by consensus democracy on inflation is deflated.
Table 2: Regression results having excluded outliers
Dependent variable:
CPI 1981-2009 Unemployment 1981-2009 (1) (2)
Executives-parties dimension 1981-
2009 -0.943** -1.105
(0.374) (0.778)
Logged population in thousands 2009 -0.440* -0.618 (0.218) (0.546)
HDI 2010 -15.285*** -4.251 (3.736) (13.059)
Constant 22.110*** 17.414* (3.886) (9.695)
Observations 24 18
R2 0.595 0.230
Adjusted R2 0.535 0.065
Residual Std. Error 1.735 (df = 20) 2.584 (df = 14)
F Statistic 9.808*** (df = 3;
20) 1.393 (df = 3; 14)
Note: *p<0.1; **p<0.05; ***p<0.001
C: Economic Openness
One obvious potential explanatory variable for a country’s macroeconomic performance is
its level of economic development; another is population. Both of these Lijphart
acknowledges, controlling for HDI and the natural logarithm of the population in thousands.
Lijphart does not, however, control for the possible confounding variable which is the level
of exposure a country faces to external influences. How successfully a country achieves a
low rate of unemployment and inflation depends, to some extent, on the monetary, fiscal,
trade, and social support policies implemented by others. Economic openness, then, should
be controlled for. In operationalising economic openness, I have used the World Bank’s data
concerning trade as a percentage of GDP for 1996, as the median year for this time period.
Controlling for economic openness has two important effects on the relationship between
consensus democracy and economic outcomes. Firstly, the fit of the model is improved:
58% of the variation in inflation and 8.8% of the variation in unemployment can be
explained, whereas previously the respective R-squared figures were 0.535 and 0.065.
Secondly, no statistical significance reappears for inflation and, again, while the relationship
between consensus democracy and inflation remains significant at the 5% level, it becomes
weaker.
Table 3: Regression results having excluded outliers and controlling for economic openness
Dependent variable:
CPI 1981-2009 Unemployment
1981-2009 (1) (2)
Executives-parties dimension 1981-2009 -0.803** -1.065 (0.366) (0.769)
Logged population in thousands 2009 -0.832** -0.035 (0.307) (0.737)
HDI 2010 -16.472*** -5.954 (3.629) (12.983)
Economic openness 1996 -0.029* 0.037 (0.017) (0.031)
Constant 28.771*** 10.990 (5.344) (11.060)
Observations 24 18
R2 0.650 0.302
Adjusted R2 0.577 0.088
Residual Std. Error 1.655 (df = 19) 2.553 (df = 13)
F Statistic 8.836*** (df = 4; 19) 1.408 (df = 4; 13)
Note: *p<0.1; **p<0.05; ***p<0.001
D: Interest Group Corporatism
One or more of the features associated with consensus democracy rather than the concept
itself may be responsible for Lijphart’s findings. Indeed, just one component of the
executive-parties dimension, interest group corporatism, drives the connection Lijphart finds
between consensus democracy and improved economic outcomes (Anderson, 2001). The
mechanism for this connection is well known (Cameron, 1984). Within countries that have
corporatist interest groups, business and labour interests are brought together. This means that
the costs of pursuing particularist goals are acknowledged, such that wage and price increases
are willingly sacrificed in favour of full employment and price stability (Anderson, 2001).
But the theoretical connection between interest group corporatism and consensus democracy
is tenuous. Interest group corporatism seems to go equally well with the majoritarian model
as it does with the consensus model, for the centralisation typical of majoritarian democracies
may well be thought to extend to interest groups (Taagepera, 2003). There therefore seems to
be no persuasive reason for the inclusion of the degree of interest group corporatism within
his executives-parties dimension.
Given the conspicuous lack of a theoretical basis for including interest group corporatism as a
feature of consensus democracy, the executives-parties dimension ought to be recreated
without this.
Given also the strong correlation between the original and reformulated executives-parties
dimensions (Pearson’s product moment correlation coefficient takes a value of -0.98 with a p-
value less than 2.2e-16; see figure 2), if consensus democracy and not merely interest group
corporatism is responsible for superior economic outcomes, then the relationship between the
new executives-parties dimension and both inflation and unemployment should be
statistically significant.
But this is not the case; eliminating the degree of interest group corporatism eliminates all
statistical significance.
Figure 2: The relationship between the original and reformulated executive-parties
dimensions
Table 4: Regression results having excluded outliers and controlling for economic openness
with the new executive-parties dimension which lacks the degree of interest group pluralism
Dependent variable:
CPI 1981-2009 Unemployment 1981-2009 (1) (2)
Executives-parties dimension 1981-
2009 0.753 0.764
(0.454) (0.936)
Logged population in thousands 2009 -0.819** 0.145 (0.324) (0.750)
HDI 2010 -16.880*** -9.603 (3.789) (13.138)
Economic openness 1996 -0.030 0.039 (0.018) (0.033)
Constant 29.053*** 12.140 (5.624) (11.517)
Observations 24 18
R2 0.617 0.238
Adjusted R2 0.536 0.004
Residual Std. Error 1.732 (df = 19) 2.667 (df = 13)
F Statistic 7.651*** (df = 4; 19) 1.017 (df = 4; 13)
Note: *p<0.1; **p<0.05; ***p<0.01
E: Ten Younger Democracies
Lijphart’s prescription that constitutional engineers should favour the consensus model of
democracy to yield the best economic outcomes may be queried for another reason. Setting
aside any concerns about outliers, controls, and conceptualisation, the conclusion Lijphart
draws from his sample of democracies may not be generalisable across the wider population.
Lijphart’s case selection may be to the advantage of his hypothesis. His thirty-six
democracies represent the rather uniform experience of Latin America and southern Europe;
of his cases, just ten are outside these geographical regions and only six are presidential.
There is then a real concern that the choice of political institutions by the countries within his
sample and their economic performance depends upon a cultural factor. This makes sense of
Fortin’s (2008) inability to replicate Lijphart’s findings for nineteen Eastern European
countries, and the fact that Croissant and Schächter (2009) again fail to replicate Lijphart’s
findings, this time across nine Asian countries between the 1980s and 2005.
Neither of these studies adopt Lijphart’s definition of democracy or operationalisation of the
components of the executives-parties dimension exactly. So, to test whether the positive
relationship he finds between consensus democracy and economic outcomes within his
sample is merely a result of his selection of cases according to constructive consensual
culture, a new sample of democracies meeting his definition and operationalisation must be
taken.
Following Liphart in using Freedom House’s rating of a country as free to determine whether
it qualifies as a democracy, provided that its population exceeds a quarter of a million, a
sample of ten countries newly democratic between the years 1998 and 2015 can be drawn.
These are: Belize, Benin, Bulgaria, Cape Verde, El Salvador, Estonia, Guyana, Hungary,
Latvia, and Lithuania2.
In calculating the executives-parties index for these countries, I refrained from including a
measure of interest group corporatism, partly due to a lack of data and partly due to the lack
of a justification for doing so. I followed Lijphart in measuring the effective number of
parties according to Laakso and Taagepera’s index (Laakso and Taagepera, 1979)3. I
operationalised the concentration of power within cabinets as the percentage of minimal
winning one-party cabinets. To measure executive dominance, I stuck to Lijphart’s choice of
average cabinet duration. For electoral system disproportionality I calculated Gallagher’s
index for each legislative election within the time period and took the mean. To then collate
these four variables into a single index, I adjusted each to make the mean zero and standard
deviation one after making taking the negative of the effective number of parties4.
The components of the executive-parties index for the ten new democracies show similar
interrelations to that of Lijphart’s thirty-six. Hence, there is good reason to expect that, if
Lijphart’s sample is duly representative, then his finding will be reproduced.
2 Chile also was democratic throughout the period 1998 to 2015 and was not included in Lijphart’s analysis.
However, I was unable to follow Lijphart’s method of personally adjusting his measure of executive dominance-
average cabinet duration-for presidential systems and as such deemed it better to omit Chile from my sample. 3 This I did for each legislative election held within the years 1998 to 2015 inclusive and then found the average. 4 So that a higher value for each component indicates a more consensus system.
Figure 3: Correlation matrix for the four components of the executive-parties dimension for
the new sample of democracies
Scaled
effective
number of
parties
Scaled
percentage of
minimal winning
one-party
cabinets
Scaled index
of executive
dominance
Scaled index of
electoral
disproportionality
Scaled effective
number of parties 1 0.734 0.714 0.381
Scaled percentage of
minimal winning one-
party cabinets
0.734 1 0.906 0.322
Scaled index of
executive dominance 0.714 0.906 1 0.421
Scaled index of
electoral
disproportionality
0.381 0.322 0.421 1
Figure 4: Correlation matrix for the first four components of the executives-parties
dimension for Lijphart’s 36 democracies
Effective
number of
parties
Percentage of
minimal winning
one-party
cabinets
Index of
executive
dominance
Index of electoral
disproportionality
Effective number of
parties 1 -0.801 -0.652 -0.603
Percentage of minimal
winning one-party
cabinets
-0.801 1 0.733 0.549
Index of executive
dominance -0.652 0.733 1 0.561
Index of electoral
disproportionality -0.603 0.549 0.561 1
However, there is no significant relationship between consensus democracy and economic
outcomes: both when following his original controls and when also controlling for economic
openness.
Figure 5: Regression results for ten new democracies
Dependent variable:
CPI 1998-2015 Unemployment 1998-2015 (1) (2)
Executives-parties dimension 1998-2015 0.379 0.386 (0.913) (1.192)
Logged population in thousands 2007 0.718 -1.039 (0.545) (0.711)
HDI 1998-2015 4.828 22.330*** (4.469) (5.837)
Constant -5.349 1.503 (5.745) (7.504)
Observations 10 10
R2 0.355 0.767
Adjusted R2 0.033 0.650
Residual Std. Error (df = 6) 1.547 2.020
F Statistic (df = 3; 6) 1.101 6.579**
Note: *p<0.1; **p<0.05; ***p<0.01
Table 6: Regression results for ten new democracies controlling for economic openness
Dependent variable:
CPI 1998-2015 Unemployment 1998-2015 (1) (2)
Executives-parties dimension 1998-2015 0.001 0.317 (0.727) (1.336)
Logged population in thousands 2007 0.920* -1.002 (0.431) (0.793)
HDI 1998-2015 -1.736 21.132* (4.538) (8.343)
Economic openness 2007 0.039* 0.007 (0.018) (0.032)
Constant -6.529 1.288 (4.481) (8.238)
Observations 10 10
R2 0.678 0.769
Adjusted R2 0.420 0.584
Residual Std. Error (df = 5) 1.198 2.202
F Statistic (df = 4; 5) 2.627 4.165*
Note: *p<0.1; **p<0.05; ***p<0.01
Conclusion
In sum, then, there is substantial evidence against Lijphart’s contention that consensus
democracy improves economic outcomes. His empirical finding conceals both theoretical and
methodological weaknesses; in addition to inadequately dealing with outliers and controls,
his creation of the executive-parties dimension itself is misleading, and his case selection too
is not beyond suspicion. What would be welcome now to either resuscitate or decidedly put
down the notion that consensus democracy has a beneficial effect on the control of inflation
and unemployment would be further investigation of culture on the choice of political
institutions and economic performance.
Word count (excluding headings, tables, figures, footnotes, and appendices): 2,177
Appendix
A: References Used
Aisen, Ari & Veiga, Francisco Jose (2011). How Does Political Instability Affect Economic
Growth? IMF Working Paper. URL (cited 31st March 2018):
https://www.imf.org/external/pubs/ft/wp/2011/wp1112.pdf
Anderson, Liam (2001). The Implications of Institutional Design for Macroeconomic
Performance. Comparative Political Studies 34(4): 429-52, doi:
10.1177/0010414001034004004
Beer, Samuel (1998). The Roots of New Labour: Liberalism Rediscovered. Economist
Cameron, David R. (1984). Social democracy, corporatism, labour quiescence and the
representation of economic interests in advanced capitalist society. In J. H. Goldthorpe (Ed.),
Order and conflict in contemporary capitalism. Oxford, UK: Clarendon.
Canes-Wrone, Brandice & Ponce de Leon, Christian (2014). Elections, Uncertainty, and
Economic Outcomes. Working Paper. Stanford University.
Croissant, Aurel & Teresa Schächter (2009). Demokratiestrukturen in Asien – Befunde,
Determinanten und Konsequenzen. Zeitschrift für Politikwissenschaft
Fortin, Jessica (2008). Patterns of Democracy? Counterevidence from Nineteen Post-
Communist Countries. Zeitschrift für Vergleichende Politikwissenschaft, doi:
10.1007/s12286-008-0014-1
Lijphart, Arend (2012). Patterns of Democracy. New Haven: Yale University Press.
Lowell, A. Lawrence (1896). Governments and Parties in Continental Europe. Boston:
Houghton Mifflin.
Taagepera, Rein (2003). Arend Lijphart’s Dimensions of Democracy: Logical Connections
and Institutional Design. Political Studies 51(1): 1- 19, doi: 10.1111/1467-9248.00409
Tsebelis, George (2002). Veto Players: How Political Institutions Work. Princeton University
Press.
B: Background References
Borman, Nils-Christian (2010). Patterns of Democracy and Its Critics. Centre for
International and Comparative Studies. URL (cited on 31 March 2018): https://www.ethz.ch/
content/dam/ethz/special-interest/gess/cis/cis-dam/CIS_DAM_2015/WorkingPapers/Living_
Reviews_Democracy/Bormann.pdf
Hlavac, Marek (2018). stargazer: Well-Formatted Regression and Summary Statistics Tables.
R package version 5.2.1. https://CRAN.R-project.org/package=stargazer
C: Code
data<-read.csv("http://andy.egge.rs/data/L.csv")
library(stargazer)
#ascertaining which countries are already excluded from Lijphart's dataset
data$country
data$cpi_1981_2009
data$unemployment_1981_2009
#the five smallest states and the three states not yet democratic in 1981 have already been
excluded (of the two inflationary ouliers Lijphart identifies, only Uruguay has been excluded)
#identifying outliers
boxplot(data$cpi_1981_2009,ylab="average annual CPI inflation for the period 1981 to
2009")
boxplot(data$unemployment_1981_2009,ylab="average annual unemployment for the period
1981 to 2009")
#Israel, Jamaica and Costa Rica are outliers with respect to CPI inflation, and Jamaica and
Spain are outliers with respect to unemployment
#excluding Israel from cpi_1981_2009 in line with Lijphart's empirical analysis
data$cpi_1981_2009[18]<-NA
#recreating Lijphart's multivariate regressions
model1<-
lm(data$cpi_1981_2009~data$exec_parties_1981_2010+log(data$pop_in_thousands_2009)+
data$hdi_2010)
summary(model1)
model2<-
lm(data$unemployment_1981_2009~data$exec_parties_1981_2010+log(data$pop_in_thousa
nds_2009)+data$hdi_2010)
summary(model2)
#creating a multivariate regression table
stargazer(model1,model2,type="html",out="lijphart.doc",title = "Lijphart's multivariate
regression",dep.var.labels = c("CPI 1981-2009","Unemployment 1981-
2009"),covariate.labels = c("Executive-parties dimension 1981-2009","Logged population in
thousands 2009","HDI 2010"))
#excluding additional outliers identified
data$cpi_1981_2009[9]<-NA
data$cpi_1981_2009[20]<-NA
data$unemployment_1981_2009[20]<-NA
data$unemployment_1981_2009[30]<-NA
#re-running the regressions without outliers
model3<-
lm(data$cpi_1981_2009~data$exec_parties_1981_2010+log(data$pop_in_thousands_2009)+
data$hdi_2010)
summary(model3)
model4<-
lm(data$unemployment_1981_2009~data$exec_parties_1981_2010+log(data$pop_in_thousa
nds_2009)+data$hdi_2010)
summary(model4)
#CPi remains significant but unemployment does not
#creating a regression table
stargazer(model3,model4,type = "html",out = "outliers.doc",title = "Regression results having
excluded outliers",dep.var.labels = c("CPI 1981-2009","Unemployment 1981-
2009"),covariate.labels = c("Executives-parties dimension 1981-2009","Logged population in
thousands 2009","HDI 2010"))
#creating a new control variable, economic openness
economic_openness<-
c(12.506,38.172,70.05,105.857,91.548,96.775,90.956,70.3,84.492,68.471,65.737,43.937,44.9
97,37.496,70.378,22.167,138.704,61.286,42.983,99.068,18.525,53.404,190.432,237.57,127.8
58,109.129,71.122,55.116,60.208,46.328,67.489,79.67,86.896,51.305,39.528,22.611)
data1<-cbind(data,economic_openness)
#regressing now controling for economic openness
model5<-
lm(data1$cpi_1981_2009~data1$exec_parties_1981_2010+log(data1$pop_in_thousands_200
9)+data1$hdi_2010+data1$economic_openness)
summary(model5)
model6<-
lm(data1$unemployment_1981_2009~data1$exec_parties_1981_2010+log(data1$pop_in_th
ousands_2009)+data1$hdi_2010+data1$economic_openness)
summary(model6)
#CPI still significant, but smaller coefficient and larger R-squared
#creating a regression table
stargazer(model5,model6,type = "html",out = "econopen.doc",title = "Regression results
having excluded outliers and controlling for economic openness",dep.var.labels = c("CPI
1981-2009","Unemployment 1981-2009"),covariate.labels = c("Executives-parties dimension
1981-2009","Logged population in thousands 2009","HDI 2010","Economic openness
1996"))
#creating a new executives-parties dimension excluding the degree of interest group
pluralism
parties<-scale(-data$eff_num_parl_parties_1981_2010)
cabinets<-scale(data$pct_minimal_winning_one_party_cabinet_1981_2010)
exec<-scale(data$index_of_exec_dominance_1981_2010)
disp<-scale(data$index_of_disproportionality_1981_2010)
ep<-(parties+cabinets+exec+disp)/4
#checking the new executives-parties dimension is relevantly similar to the original
cor.test(ep,data$exec_parties_1981_2010)
plot(ep,data$exec_parties_1981_2010,xlab = "Executive-parties dimension without degree of
interest group pluralism 1981-2009",ylab = "Executives-parties dimension 1981-2009")
abline(lm(data$exec_parties_1981_2010~ep))
#regressing CPI and unemployment against the executive-parties dimension without degree
of interest group pluralism
model7<-
lm(data1$cpi_1981_2009~ep+log(data1$pop_in_thousands_2009)+data1$hdi_2010+data1$e
conomic_openness)
summary(model7)
model8<-
lm(data1$unemployment_1981_2009~ep+log(data1$pop_in_thousands_2009)+data1$hdi_20
10+data1$economic_openness)
summary(model8)
#displaying this via a regression table
stargazer(model7,model8,type = "html",out = "ep.doc",title = "Regression results having
excluded outliers and controlling for economic openness with the new executive-parties
dimension which lacks the degree of interest group pluralism",dep.var.labels = c("CPI 1981-
2009","Unemployment 1981-2009"),covariate.labels = c("Executives-parties dimension
1981-2009","Logged population in thousands 2009","HDI 2010","Economic openness
1996"))
#looking at countries democratic from at least 1998 to today
#processing data
parties1<-c(-1.63,-4.43,-3.85,-2.07,-3.10,-4.62,-2.25,-2.53,-6.02,-5.88)
sparties<-scale(parties1)
cabinets1<-c(100,0,27.8,100,0,0,100,0,0,0)
scabinets<-scale(cabinets1)
exec1<-c(8.5,4,4,10,3,4,7,5.33,4,4)
sexec<-scale(exec1)
disp1<-c(17.79,8.57,7.37,6.85,3.78,3.85,0.87,11.9,4.12,10.71)
sdisp<-scale(disp1)
epnewcountries<-(sparties+scabinets+sexec+sdisp)/4
#loading data
newdata<-read.csv("newdata.csv",header=TRUE)
data.frame(newdata)
#comparing the correlation between components of the executives-parties dimension for the
new democracies with that of Lijphart's
correlation_matrix<-
cor(newdata[,c("Effective.number.of.parties","Percentage.of.minimal.winning.one.party.cabi
nets","Index.of.executive.dominance","Index.of.electoral.disproportionality")])
corr_matrix<-
cor(data[,c("eff_num_parl_parties_1981_2010","pct_minimal_winning_one_party_cabinet_1
981_2010","index_of_exec_dominance_1981_2010","index_of_disproportionality_1981_20
10")])
#creating a correlation matrix table
stargazer(correlation_matrix,type = "html",out = "correlation.doc")
corr_matrix<-
cor(data[,c("eff_num_parl_parties_1981_2010","pct_minimal_winning_one_party_cabinet_1
981_2010","index_of_exec_dominance_1981_2010","index_of_disproportionality_1981_20
10")])
stargazer(corr_matrix,type = "html",out = "corr.doc")
#regressing CPI and unemployment against the executives-parties dimension for the new ten
democracies
model9<-
lm(newdata$CPI~newdata$executives.parties+log(newdata$population)+newdata$HDI
summary(model9)
model10<-
lm(newdata$unemployment~newdata$executives.parties+log(newdata$population)+newdata
$HDI)
summary(model10)
model11<-
lm(newdata$CPI~newdata$executives.parties+log(newdata$population)+newdata$HDI+new
data$economic.openness)
summary(model11)
model12<-
lm(newdata$unemployment~newdata$executives.parties+log(newdata$population)+newdata
$HDI+newdata$economic.openness)
summary(model12)
#creating regression tables
stargazer(model9,model10,type = "html",out = "newcountries.doc",title = "Regression results
for ten new democracies",dep.var.labels = c("CPI 1998-2015","Unemployment 1998-
2015"),covariate.labels = c("Executives-parties dimension 1998-2015","Logged population in
thousands 2007","HDI 1998-2015")))
stargazer(model11,model12,type = "html",out = "newcountries1.doc",title = "Regression
results for ten new democracies",dep.var.labels = c("CPI 1998-2015","Unemployment 1998-
2015"),covariate.labels = c("Executives-parties dimension 1998-2015","Logged population in
thousands 2007","HDI 1998-2015","Economic openness 2007")))
D: Additional Data
The data used to calculate each of the four components of the executive-parties dimension for
the ten new democracies was sourced from a combination of Wikipedia, African election
database, caribbeanelections.com, and Knoema.
These were averaged across all legislative elections within the time period, as per Lijphart’s
methodology.5
All data concerning inflation, unemployment, population and economic openness comes from
the world bank. The United Nations development programme was used to find the Human
Development Indices.
Table 1: Components of the Executive-Parties Dimension for Ten Young Democracies
Country
Effective
number of
parties
Percentage of one-
party minimal
winning cabinets
Cabinet
Duration
Gallagher’s index of
electoral
disproportionality
Executive-
parties index
Belize 1.63 100 8.50 17.79 1.523
Benin 4.4 0 4 8.57 5.980
Bulgaria 3.85 27.8 4 7.37 12.500
Cape
Verde 2.07 100 10 6.85 0.840
El
Salvador 3.10 0 3 3.78 155.200
Estonia 4.62 0 4 3.85 10,055.780
Guyana 2.25 100 7 0.87 1.276
Hungary 2.53 0 5.33 11.9 1.420
Latvia 6.02 0 4 4.12 1.981
Lithuania 5.88 0 4 10.71 2.070
5 Belize held legislative elections in 1998, 2003, 2008, 2012, and 2015. Benin’s legislative elections took place
in 1999, 2003, 2007, and 2011. Bulgaria staged legislative elections in 2001, 2005, 2009, and 2014. Cape
Verde’s occurred in 2001, 2006, and 2011. El Salvador held legislative elections in 2000, 2003, 2006, 2009,
2012, and 2015. Estonia held legislative elections in 1999, 2003, 2007, 2011, and 2015. Guyana’s were in 2001,
2006, 2011, and 2015. Hungary’s took place in 1998, 2002, 2006, 2011, and 2014. Latvia’s occurred in 1998,
2002, 2006, 2010, 2011, 2014. Lithuania’s legislative elections were in 2000, 2004, 2008, and 2012.
Table 2: Inflation, unemployment, level of development, population, and economic openness
for ten young democracies
Country
CPI,
average
1998-2015
Unemployment,
average 1998-2015
HDI,
average
1998-2015
Population in
thousands,
2007
Economic
openness,
average 1998-
2015
Belize 1.58 11.2 0.70 298,407 14.6
Benin 2.82 1.02 0.46 8,454,791 57.6
Bulgaria 5.87 11.8 0.77 8,312,068 103.4
Cape
Verde 2.18 9.2 0.63 486,438 96.9
El
Salvador 2.64 6.8 0.67 6,038,475 69.2
Estonia 3.82 9.9 0.84 1,340,680 142.3
Guyana 5.05 11.2 0.63 747,869 155.2
Hungary 5.98 9.0 0.82 10,055,780 143.2
Latvia 4.29 12.5 0.81 2,200,325 100
Lithuania 2.49 11.8 0.83 3,231,294 119.4