do the effects of corruption upon growth differ between democracies and autocracies?
TRANSCRIPT
Do the Effects of Corruption upon Growth DifferBetween Democracies and Autocracies?
Andreas Assiotis and Kevin Sylwester*
AbstractMany studies examining whether corruption lowers economic growth do not consider if the effects of cor-ruption differ across countries. Whether corruption produces the same effects everywhere or whether itseffects are conditional on some country characteristics are important questions. We investigate the associa-tion between corruption and growth, where the marginal impact of corruption is allowed to differ acrossdemocratic and nondemocratic regimes. Using cross-country, annual data from 1984 to 2007, we regressgrowth on corruption, democracy and their interaction. We find that decreases in corruption raise growthbut more so in authoritarian regimes. Possible reasons are that in autocracies corruption causes moreuncertainty, is of a more pernicious nature, or is less substitutable with other forms of rent seeking.
1. Introduction
Understanding the vast differences in income levels and economic growth rates hasgenerated many explanations for these differences. One explanation is that some gov-ernments are more corrupt as public officials abuse their power in order to extractbribes. Such abuse results in personal gain for such officials at the expense of thepopulace. Since such practices dissuade productive activities, they could lower growth(Svensson, 2005). Although corruption has not always been viewed negatively, recentstudies find that corruption lowers income (Shleifer and Vishny, 1993; Mo, 2001;Mauro, 1995). We also consider the effects of corruption upon growth but we allowthe effects of corruption to differ across political regimes. Corruption might affectgrowth differently in democracies relative to autocracies as we explain in section 3.Past research has considered links between political regime and corruption, but suchresearch has often considered whether democratization leads to more or less corrup-tion. Instead, we consider whether the type of political regime influences the effects ofcorruption upon growth.
2. Literature Review
Scholars have long debated how corruption affects economic growth. Leff (1964) andHuntington (1968) argue that corruption increases growth by helping agents avoidbureaucratic delays. In contrast, Tanzi and Davoodi (2000) view corruption as lower-ing growth. Mauro (1995) explains the lower growth through corruption’s negativeeffect on investment.1 More recent studies allow the effects of corruption to differacross country characteristics. Méon and Sekkat (2005) allow corruption to affect
* Assiotis: Department of Economics, University of Cyprus, 1678 Nicosia, Cyprus. Tel: +357-2250-0807;Fax: +357-2250-0082; E-mail: [email protected]. Sylwester: Southern Illinois UniversityCarbondale, Carbondale, IL 62901, USA. The authors would like to thank Elias Papaioannou for muchhelpful feedback and also thank one referee for comments that have improved the article. Of course, allerrors are the responsibility of the authors.
Review of Development Economics, 18(3), 581–594, 2014DOI:10.1111/rode.12104
© 2014 John Wiley & Sons Ltd
growth differently depending upon the quality of governing institutions. Corruptionlowers growth more in bad governance countries. Aidt et al. (2008) report oppositeresults using a different empirical specification. Swaleheen and Stansel (2007) findthat corruption raises growth in countries with high economic freedom but lowers itwith low economic freedom.
Mendez and Sepulveda (2006) also consider whether the effects of corruption uponeconomic growth differ between democracies and autocracies. They split their sampleinto free and not-free regimes and run separate regressions. For free countries, cor-ruption increases growth but then further levels of corruption decrease it. No strongassociation is found for the not-free group. However, some concerns with their meth-odology arise. For one, they use ordinal indices to measure corruption and so takingtheir square to fit a quadratic relationship might not be appropriate. Second, countriescan move between groups as political events unfold. Running separate regressionsdoes not directly take into account this transitioning between groups. Instead, twoobservations from a country that is free at one point in time but not free at anotherare treated as distinct. Third, their methodology does not address concerns thatgrowth can influence corruption or that time-varying latent factors could drive both.To address some of these concerns, we employ a panel dataset where countries areallowed to switch from a nondemocratic to a democratic regime and so we can betterexploit the within country variation in the sample. We also employ dynamic general-ized method of moments (GMM) estimation methodologies as robustness checks tobetter address endogeneity concerns.
3. Economic Framework
Why might the effects of corruption depend upon political regime?The first possibility from Murphy et al. (1993) and Harstad and Svensson (2011)
recognizes that corruption is a form of rent seeking that is perhaps substitutable withother types such as lobbying. While corruption is often perceived as bribes to policyenforcers, lobbying is usually associated with political campaign activities that aim toinfluence policy makers (Campos and Giovannoni, 2007). A firm that successfullylobbies to change the rules to its interests presumably finds it less necessary to bribeother officials. In contrast, firms that can easily bribe officials might not then lobby fora change in the rules. If opportunities for lobbying are less available in authoritarianregimes with fewer decision makers, then the degree of substitutability between thetwo is lower. Perhaps lowering corruption in authoritarian regimes could have greaterbenefits for economic growth because of the lower substitutability between corruptionand lobbying in these countries.
A second reason focuses upon the type of corruption. Assuming that democraciesare more transparent, corruption that exists in democracies could be more benignsince corrupt activities that greatly hurt the majority are more likely to be reportedand combated. This is not to say that the frequency of corrupt activities is lower indemocracies, only that its detrimental effects are smaller.2 Therefore, a fall in corrup-tion in an authoritarian regime could then have a more positive effect upon economicgrowth than in a democratic regime.
A third reason centers upon the uncertainty that corruption creates.3 This uncer-tainty could be greater under autocracies where the rules of the game could moregreatly change from regime to regime. Regulations certainly change when democraticadministrations change, but the process is likely less capricious and more transparent.
582 Andreas Assiotis and Kevin Sylwester
© 2014 John Wiley & Sons Ltd
Therefore, if corruption creates greater uncertainty in autocracies and if this uncer-tainty lowers growth then lowering corruption will have more positive effects upongrowth in autocracies.
Unlike the first three, a final possibility predicts that decreasing corruption shouldhave more positive effects upon economic growth in democracies. Murphy et al.(1993) argue that centralized corruption reduces income less than does decentralizedcorruption because of a “tragedy of the commons.” A dishonest official wants toextract the highest bribe possible without driving a firm out of business, but if corrup-tion is decentralized with numerous officials independently demanding bribes, then asa group bureaucrats set total bribe amounts too high and so drive firms out of busi-ness (or at least into informal sectors of the economy). A centralized bureaucracyoperating like a monopolist will take this consideration into account, making totalbribes lower. Given that democracies often incorporate greater checks and balancesresulting in less centralized concentrations of power, corruption could be less central-ized under democracies. Decreases in corruption should then have greater growtheffects in democracies.
4. Description of the Data and Descriptive Statistics
We use annual data from 119 countries (listed in the Appendix) from 1984 to 2007.Although Ehrlich and Lui (1999) and Mendez and Sepulveda (2006) consider multi-year windows so as to better focus on growth phenomenon as opposed to businesscycle movements, we primarily use annual data to better pinpoint regime changes.4
Gross domestic product (GDP) per capita and its growth rate (GDP and GROWTH),the share of government expenditures in GDP (GOV), and the investment share ofGDP (INV) are taken from the Penn World Tables, version 6.3.5 Annual populationgrowth (GPOP) is from the World Bank’s 2009 World Development Indicators.
Democracy (DEM) is measured using the Freedom House indices. The FreedomHouse data begins in 1972. The two indices consider two components of politicalfreedom. The political rights index measures the extent of free and fair elections,political pluralism and the rights of political minorities. The civil liberties index meas-ures individual liberties such as the freedoms of speech, to practice one’s religion andto peaceably assemble. Both indices range from one to seven where lower numbersindicate higher levels of freedom. To derive DEM, we first take the average of thesetwo indices. We then rescale this average, transforming it from a one to seven intervalto a six to zero interval so that higher values of DEM denote more democratic free-doms.6 As a robustness check, we also consider the Polity measure of democracy. It isconstructed in unit intervals from −10 to 10 with higher values denoting strongerdemocracies. The Polity indicator builds upon the following components: competi-tiveness of political participation, competitiveness and openness of executive recruit-ment, and constraints on the chief executive (see Marshall and Jaggers, 2004).
In addition to using these indices, we consider the binary variable fromPapaioannou and Siourounis (2008). They create a dummy variable, DEM_PS, thattakes the value one for a democracy and zero otherwise. In their classification system,a country is only considered to have democratized if that democratization was sus-tained and so did not revert back to authoritarianism. Therefore, once DEM_PSbecomes “one” it retains this value throughout the remainder of the sample period.7
The corruption index comes from the International Country Risk Guide (ICRG).This index is based on the opinion of experts and captures the degree to which “highgovernment officials are likely to demand special payments” and to which “illegal
DEMOCRACY–AUTOCRACY CORRUPTION EFFECTS 583
© 2014 John Wiley & Sons Ltd
payments are generally expected throughout lower levels of government in the formof bribes connected with import and export licenses, exchange controls, tax assess-ments, police protection, or loans.” ICRG classifies countries on a scale from 0 to 6with 6 indicating low levels of corruption. We use the ICRG data since it is availablefor more years than other measures of corruption. Nevertheless, as an alternativemeasure for corruption, we consider the corruption indicator from the World Bank’sWorld Governance Indicators (WGI). Although it only begins in 1996, it is availablefor more countries than is the ICRG measure. The WGI corruption index rangesfrom −2.5 to +2.5 where higher numbers denote a better control of corruption. Wealso use the Corruption Perception Index (CPI) from Transparency International tocheck for the robustness of our findings. The CPI is available from 1995 and rangesfrom 0 to 10 with lower numbers indicating high levels of corruption.
Democratization often accompanies economic reforms and not controlling forthese could bias upward the estimated effects of democracy upon economic growth.Therefore, as a robustness check, we control for economic reforms utilizing the classi-fication developed by Sachs and Warner (1995) and updated by Wacziarg and Welch(2003). For countries with open trade policies, the variable REFORM takes the valueone. REFORM equals zero for countries with high trade barriers.8 Like Giavazzi andTabellini (2005), we presume that REFORM is associated with more widespread lib-eralizations within the country. As countries liberalize, REFORM goes from zero toone.
5. Methodology
The Model
We employ a panel to capture within-country variation. Consider the followingempirical specification which we adapt from Ehrlich and Lui (1999):
Yi t i t it it it i t i t, , ,= + + ( ) + ( ) + ×( ) + ′ +−α η δ ς θ εCO DEM CO DEM X 1G (1)
where i, t denote country and time respectively. Yi,t is the log growth rate of annualreal GDP per capita adjusted for PPP. The intercepts αi and ηt indicate country andyear fixed effects in order to control for time invariant factors specific to a country aswell as global shocks that influence all countries similarly. DEM is the FreedomHouse democracy index, CO is the ICRG control of corruption index, andDEM × CO is the interaction term between them. The vector X will later containcontrol variables such as the lag of the natural log of GDP per capita (GDP), thepopulation growth rate (GPOP) as well as government purchases, investment andtrade. Finally, ε denotes the error term where E (εi,t) = 0 for all i and t.
Potential Endogeneity
Income, democracy and corruption could all be endogenous. However, Acemoglu etal. (2008) argue that income does not lead to democratization once one controls fortime invariant factors that could drive both. They estimate:
DEM DEM GDPit i t it it it= + + + +− −β β β β ε0 1 2 1 3 1 . (2)
We take a similar approach so as to better justify taking corruption and democracy asexogenous. From column (1) of Table 1, the coefficient estimate of β3 on lagged
584 Andreas Assiotis and Kevin Sylwester
© 2014 John Wiley & Sons Ltd
income is statistically insignificant and so no strong evidence arises of an associationbetween income and corruption. In column (2), we replace DEM with the control ofcorruption (CO). Again, the results suggest that income is not strongly associatedwith corruption once one controls for time invariant factors. Column (3) also includesthe natural log of the number of per capita cellular phone subscriptions. Perhaps thereason that income was not significant is due to offsetting effects. Rising income couldmollify a restive populace but rising income could also be associated with greateraccess to technologies that could promote democratic change. Greater cellular phoneaccess could allow for greater access to information as well as allow dissenters tobetter coordinate. This greater access to information could also deter corruption asfound in Goel et al. (2012). By explicitly controlling for cell phone subscriptions, wehope that the coefficient on income better captures direct effects from rising incomeupon political change. Nevertheless, coefficients upon lagged income and cell phonesubscriptions are insignificant.9
Murtin and Wacziarg (2011) argue that the inclusion of fixed effects is ill-suited totest for whether income causes democracy as measurement error in fixed effectsmodels with persistent variables can lead to large biases in the coefficient estimates.They estimate (2) by GMM. Columns (4), (5), and (6) of Table 1 show results ofGMM estimation using two lags of the endogenous variables as instruments. No evi-dence arises that income causes democracy or the control of corruption (although thespecification in column (5) might be inappropriate due the presence of second orderserial correlation.)
Unfortunately, other endogeneity concerns arise. Several papers have consideredhow democracy affects corruption. Shen and Williamson (2005) suggest that democ-
Table 1. Income Regressions (Annual), 1984–2007
Estimation method
(1) (2) (3) (4) (5) (6)Fixedeffects
Fixedeffects
Fixedeffects
SYSGMM
SYSGMM
SYSGMM
Dependent variable DEM CO DEM DEM CO DEMConstant 1.31 0.57 1.28 0.25 0.02 0.11
(0.44)*** (0.57) (0.47)*** (0.34) (0.29) (0.22)GDP(−1) −0.07 −0.01 −0.06 −0.01 −0.003 −0.01
(0.04) (0.06) (0.05) (0.04) (0.03) (0.04)DEM(−1) 0.82 0.82 0.97 0.96
(0.01)*** (0.01)*** (0.03)*** (0.02)***CO(−1) 0.84 1.01
(0.01)*** (0.02)***Cellular −0.01 0.03
(0.01) (0.01)
Observations 2732 2705 2696 2732 2705 2696Number of countries 119 119 119 119 119 119Sargan test (p-value) — — — 0.16 0.23 0.94AR (2) test
(p-value)— — — 0.71 0.00 0.81
Notes: Standard errors in parentheses: *,**,***Significant at 10%, 5%, and 1%, respectively. The value forAR(2) denotes the p-value of a serial correlation test taking the null of no serial correlation in the first dif-ferences of the dependent variable. The value for Sargan denotes the p-value of the Sarganoveridentification test taking the null that the model is correctly identified.
DEMOCRACY–AUTOCRACY CORRUPTION EFFECTS 585
© 2014 John Wiley & Sons Ltd
racy has a positive effect on the level of corruption control. Ali and Isse (2003) alsopresent evidence that political freedom and transparency are positively correlatedwith corruption control. Conversely, Ehrlich and Lui (1999) affirm that autocraticregimes could achieve growth rates equal to or higher than decentralized democraciesbecause corruption is more constrained in autocracies.10 Rivera-Batiz (2002) developsa theoretical model showing stronger democratic institutions influence governance byconstraining the actions of corrupt executives. Musila J. (2007; unpublished manu-script) suggests a more nuanced relationship. Mid-level democracies are more proneto corruption than either authoritarian regimes or strong democracies. In contrast,our work does not consider democracy as a causal factor of corruption.
To consider whether democracy systematically influences corruption we createdtwo groups of countries. Group A countries remained nondemocratic during thesample period. Group B countries were initially autocratic but democratized (asdetermined by Papaioannou and Siourounis (2008)) within our sample period. Wethen take the average change in corruption for each group.11 For the countriesremaining nondemocratic, CO increased on average by 1.18. For group A, COincreased by 1.28. Thus, corruption on average changed nearly the same betweendemocratizing countries and those that remained nondemocratic. We also consideredall the countries that democratized during the sample period and compared theaverage corruption score for the five years before and after democratization. For 24 ofthese 36 countries, the absolute change in the corruption score was less than one, sug-gesting that becoming democratic did not greatly change the extent of corruption.
Despite these arguments, corruption and democracy could still be endogenous andso we will also employ dynamic GMM methodologies described below.
6. Results
Table 2 presents results of the model in (1). Before presenting the baseline specifica-tion, column (1) first considers a specification without any explanatory variables butcorruption, democracy and the interactive term. Column (2) removes countries thatwere always democratic during the sample period. The control group of countries isnow those that remained nondemocratic (instead of those that remained nondemo-cratic or were democratic throughout the sample period). Column (3) revisits theconcern that changes in telecommunications drives growth, political reform and theextent of corruption. Although the coefficient upon the prevalence of cell phone sub-scriptions is significant, the coefficients upon CO, DEM, and CO × DEM remain sig-nificant. Column (4) considers other control variables whereas column (5) againremoves those countries that were always democratic. For all these cases, we find thatthe control of corruption and the level of democracy are positively associated witheconomic growth. Papaioannou and Siourounis (2008) and Rodrik and Wacziarg(2005) both find positive associations between democracy and growth.12 However, thecoefficient on the interactive term is negative. The association between corruptionand growth is less positive in democracies, suggesting that the benefits upon growth ofcontrolling corruption are actually greater in authoritarian regimes.
To explore the economic magnitude suggested by the coefficient estimates incolumn (3), consider three hypothetical countries where the level of democracy is low(DEM = 0) in country A, average (DEM = 3) in country B and high (DEM = 6) incountry C, respectively. Consider an increase of CO by 0.68 which is the mean of the119 within country standard deviations. For country A, growth increases by 0.44(= 0.65 × 0.68) percentage points. For country B, this same change in CO raises
586 Andreas Assiotis and Kevin Sylwester
© 2014 John Wiley & Sons Ltd
Tab
le2.
Pan
elD
ata
Reg
ress
ions
(Ann
ual)
,198
4–20
07
Cor
rupt
ion
inde
x(1
)(2
)(3
)(4
)(5
)(6
)(7
)(8
)(9
)IC
RG
ICR
GIC
RG
ICR
GIC
RG
WG
IW
GI
CP
IC
PI
Pan
elA
:Coe
ffici
ente
stim
ates
CO
0.65
0.94
0.63
0.85
1.02
4.30
3.85
1.36
1.29
(0.3
3)**
(0.4
8)**
*(0
.29)
**(0
.42)
**(0
.59)
*(0
.73)
***
(0.9
7)**
*(0
.67)
**(0
.59)
**D
EM
0.87
81.
400.
831.
011.
240.
130.
051.
481.
24(0
.29)
***
(0.4
7)**
*(0
.27)
***
(0.3
5)**
*(0
.54)
**(0
.41)
(0.3
6)(0
.52)
***
(0.4
1)**
*C
O×
DE
M−
0.14
−0.
41−
0.16
−0.
24−
0.38
−0.
92−
0.80
−0.
30−
0.25
(0.0
7)**
(0.1
3)**
*(0
.07)
**(0
.09)
***
(0.1
6)**
(0.3
2)**
(0.0
5)**
(0.1
4)**
(0.1
1)**
Cel
lula
r0.
12(0
.05)
**G
DP
(−1)
−8.
60−
4.65
−12
.06
−6.
81−
5.51
−4.
71(1
.75)
***
(1.1
9)**
*(1
.67)
***
(3.2
3)**
(1.4
9)**
*(1
.93)
***
GP
OP
−0.
030.
801.
241.
290.
600.
52(0
.50)
**(0
.59)
*(0
.30)
***
(0.4
1)**
(0.3
1)*
(0.2
4)**
Obs
erva
tion
s28
2318
0327
8127
1316
4194
612
5495
212
92N
umbe
rof
coun
trie
s11
976
119
119
7611
914
511
915
5R
2(w
ithi
n)0.
110.
170.
050.
170.
290.
390.
400.
400.
40
Pan
elB
:Est
imat
edef
fect
sof
aon
e-un
itch
ange
inco
rrup
tion
upon
grow
thfo
rdi
ffer
entv
alue
sof
DE
M
DE
M=
00.
65**
0.94
**0.
63**
0.85
***
1.02
**4.
30**
3.85
***
1.36
**1.
29**
*D
EM
=3
0.23
−0.
290.
150.
13−
0.12
1.54
1.45
0.46
0.54
DE
M=
6−
0.19
***
1.52
**−
0.33
−0.
59**
*−
1.26
**−
1.22
**−
0.95
**−
0.44
**−
0.29
**
Not
es:
Dep
ende
ntva
riab
leis
the
grow
thra
teof
real
GD
Ppe
rca
pita
(PP
P).
All
regr
essi
ons
cont
ain
tim
ean
dco
untr
yfix
edef
fect
s.St
anda
rder
rors
inpa
rent
hese
s:*,
**,*
**D
enot
esi
gnifi
canc
eat
10%
,5%
and
1%,r
espe
ctiv
ely.
Reg
ress
ions
perf
orm
edut
ilize
dW
hite
hete
rosk
edas
tic-
cons
iste
ntco
vari
ance
mat
rice
s.W
ald
coef
ficie
ntte
sts
used
tode
term
ine
stat
isti
cal
sign
ifica
nce
inP
anel
B.C
olum
ns(2
)an
d(4
)ex
clud
eal
lth
eco
untr
ies
from
our
sam
ple
that
wer
ede
moc
rati
cth
roug
hout
the
sam
ple
peri
od.
Col
umns
(5)–
(8)
cons
ider
diff
eren
tin
dica
tors
for
corr
upti
on.
DEMOCRACY–AUTOCRACY CORRUPTION EFFECTS 587
© 2014 John Wiley & Sons Ltd
growth by 0.16 percentage points. For the fully democratic country C, growth falls by0.13 percentage points. These results indicate that the effects of corruption upongrowth could vary nontrivially across countries with different political regimes. Mostinterestingly, the results reveal that the control of corruption might even lowergrowth in strong democracies. With a coefficient upon CO of 0.65, a value of DEM of5 or 6 pushes the sum of the coefficients upon CO (0.65 − 0.14 × DEM) to becomenegative, suggesting that increases in the control of corruption actually lower growth.Perhaps corruption in these strong democracies more often occurs so as to “greasethe wheel” so as to facilitate productive activities. Panel B provides magnitudes of thepredicted change to growth of a one unit change in CO for different values of DEM.As above, an increase in CO is associated with faster growth in nondemocracies butslower growth in democracies.
Columns (6) and (7) replace the corruption variable from ICRG with that from theWorld Bank’s World Governance Indicators. The latter is available only after 1996but is available for more countries. Column (6) only uses the WGI corruption variablefor the 119 countries used in other specifications whereas column (7) considers alarger set of countries. The coefficient upon the interaction term remains negative.Similarly, the last two columns replace the corruption variable from ICRG with thatfrom Transparency International, denoted as CPI. Coefficient estimates remainrobust. The control of corruption raises growth in authoritarian countries but the pre-dicted association is negative in democratic countries.
Table 3 considers other democracy measures. The first three columns employDEM_PS from Papaioannou and Siourounis (2008). The latter three columns
Table 3. Robustness Checks using alternative measures for Democracy
(1) (2) (3) (4) (5) (6)
CO 0.38 0.37 0.37 0.29 0.22 0.28(0.30) (0.26)* (0.18)** (0.27)*** (0.28)** (0.12)**
DEM_PS 2.03 2.24 1.83(0.93)** (0.87)** (0.75)**
CO × DEM_PS −0.73 −0.67 −0.45(0.30)*** (0.30)** (0.22)**
GDP(−1) −4.63 −5.14 −5.26 −4.56(1.24)*** (0.56)*** (2.45)** (0.62)***
GPOP 0.88 0.69 0.74 0.36(0.30)* (0.14)*** (0.40)* (0.31)
POLITY 0.29 0.26 0.12(0.08)*** (0.11)** (0.04)***
CO × POLITY −0.07 −0.07 −0.03(0.02)*** (0.03)** (0.01)**
Observations 1803 1641 2703 1629 1564 2454Number of countries 76 76 119 76 76 119R2 (within) 0.18 0.28 0.25 0.17 0.26 0.21
Notes: Dependent variable is the growth rate of real GDP per capita (PPP). All regressions contain timeand country fixed effects. Standard errors in parentheses: *,**,***Denote significance at 10%, 5% and 1%,respectively. Regressions performed utilized White heteroskedastic-consistent covariance matrices.Columns (3) and (6) include all the countries from our sample. The remaining columns exclude countriesthat were democratic throughout the sample period.
588 Andreas Assiotis and Kevin Sylwester
© 2014 John Wiley & Sons Ltd
consider the Polity index. For both measures, the coefficient on the interactive termremains negative and significant. Table 4 includes further controls. The first threecolumns consider a quadratic specification for corruption. We include the square ofCO by itself and interacted with DEM. Column (1) takes the full sample whilecolumn (2) removes countries that were always democratic. Consider the same 0.68increase of CO as above and the same three values for DEM: 0, 3 and 6. In the lowdemocracy country, this increase in CO is predicted to raise growth by 2.95 percent-age points. In the mid-level democracy, growth increases by 0.60 percentage points.Growth decreases by 1.76 percentage points in the strong democracy. Although thesevalues are larger than those for the linear model, the same conclusion remains.Improvements in the control of corruption are less beneficial for economic growth asdemocracy strengthens. In accordance with the linear model, the quadratic model ofcolumn (1) also suggests that increasing CO in strongly democratic countries(DEM = 4, 5 or 6) lowers growth. A difference, however, is the presence ofnonmonotonicities for lower values of DEM. For DEM = 3, increases in CO raisegrowth when CO is less than 2.9 but lower growth otherwise. For DEM = 0, thisthreshold is 2.34.
Column (3) re-considers a linear specification but adds the lag of GOV as an addi-tional explanatory variable. Column (4) replaces GOV with one year lagged invest-ment (INV). Column 5 considers REFORM. Again, the results do not appear tochange even when we include GOV, lagged INV and REFORM in the same specifi-cation as in columns (6) and (7). Corruption, democracy, and the interaction termremain statistically significant and therefore, consistent with the results providedearlier.13 Column (8) includes all of the additional controls, including the quadraticterms.
As an additional robustness check we perform dynamic GMM. Table 5 presentsthese results for both difference-GMM (columns 1–3) and system-GMM (columns4–6) estimators considering different democracy measures. We run specifications withthe only regressors being the lagged dependent variable, corruption, democracy andthe interactive term between the two. To keep the instrument set parsimonious, weonly use the second period lags as instruments. The results of the GMM estimates arein agreement with those above for all three democracy measures. The coefficient esti-mates for control of corruption and democracy are significant and positive whereasthe coefficient upon the interaction term is negative.
7. Conclusion
Our finding that the association between corruption and growth is less positive indemocracies runs counter to that in Mendez and Sepulveda (2006). They find that cor-ruption in democracies first raises growth and then decreases it as corruption con-tinues to increase. They also find that corruption is not strongly associated witheconomic growth in nondemocracies. Given these differences, further work is war-ranted in exploring how associations between corruption and economic growth differacross political regimes. Our results also counter claims that corruption is less harmfulin authoritarian countries because it allows one to “grease the wheels” and avoidinstitutional obstacles dissuading productive activities. If anything, more evidence ofgreasing the wheels arises for democracies. Our results suggest that controlling cor-ruption in authoritarian regimes can produce greater benefits than limiting corruptionin democracies. Future work will attempt to uncover if this is the explanation or ifother reasons are more likely for why corruption’s influence upon growth depends
DEMOCRACY–AUTOCRACY CORRUPTION EFFECTS 589
© 2014 John Wiley & Sons Ltd
Tab
le4.
Rob
ustn
ess
Che
cks
usin
gA
dditi
onal
Con
trol
s
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
CO
4.90
5.12
1.10
1.03
1.00
0.91
0.76
2.71
(0.9
6)**
*(1
.29)
***
(0.5
7)**
(0.5
8)*
(0.5
7)*
(0.5
)*(0
.22)
***
(0.9
3)**
*D
EM
2.03
2.62
1.23
1.26
1.20
1.05
0.66
0.98
(0.4
0)**
*(0
.74)
***
(0.5
2)**
*(0
.53)
**(0
.53)
**(0
.45)
**(0
.20)
***
(0.4
0)**
CO
×D
EM
−1.
29−
1.58
−0.
35−
0.37
−0.
34−
0.29
−0.
15−
0.59
(0.2
5)**
*(0
.52)
***
(0.1
5)**
*(0
.15)
**(0
.15)
**(0
.13)
**(0
.05)
***
(0.5
2)**
CO
_SQ
−0.
82−
0.82
−0.
47(0
.17)
***
(0.2
4)**
*(0
.17)
***
CO
_SQ
×D
EM
0.20
0.21
0.09
(0.0
4)**
*(0
.08)
***
(0.0
4)**
GD
P(−
1)−
4.46
−4.
65−
4.73
2.82
−3.
58−
6.61
(0.9
5)**
*(1
.13)
***
(0.9
8)**
(0.8
3)**
*(0
.60)
***
(0.7
1)**
*G
PO
P0.
800.
760.
760.
400.
220.
38(0
.41)
**(0
.42)
*(0
.43)
*(0
.33)
(0.1
6)(0
.20)
*G
OV
(−1)
−0.
15−
0.11
−0.
07−
0.04
(0.0
7)**
(0.0
6)*
(0.0
2)**
(0.0
3)IN
V(−
1)0.
060.
030.
010.
05(0
.02)
**(0
.04)
(0.0
2)(0
.02)
**R
EF
OR
M1.
891.
441.
501.
10(0
.50)
***
(0.6
6)**
(0.3
8)**
*(0
.48)
**
Obs
erva
tion
s28
2318
0316
4116
4116
4116
4124
2024
20N
umbe
rof
coun
trie
s11
976
7676
7676
119
119
Cou
ntry
fixed
effe
cts
YE
SY
ES
YE
SY
ES
YE
SY
ES
YE
SY
ES
Tim
efix
edef
fect
sY
ES
YE
SY
ES
YE
SY
ES
YE
SY
esY
ES
R2
(wit
hin)
0.13
0.05
0.30
0.29
0.30
0.29
0.22
0.10
Not
es:
Dep
ende
ntva
riab
leis
the
grow
thra
teof
real
GD
Ppe
rca
pita
(PP
P).
All
regr
essi
ons
cont
ain
tim
ean
dco
untr
yfix
edef
fect
s.St
anda
rder
rors
inpa
rent
hese
s.*,
**,*
**D
enot
esi
gnifi
canc
eat
10%
,5%
and
1%,r
espe
ctiv
ely.
Reg
ress
ions
perf
orm
edut
ilize
dW
hite
hete
rosk
edas
tic-
cons
iste
ntco
vari
ance
mat
rice
s.C
O_S
Qde
note
sth
esq
uare
ofC
O.C
olum
ns(3
)–(6
)lim
itth
esa
mpl
eto
coun
trie
sth
atw
ere
init
ially
nond
emoc
rati
c.
590 Andreas Assiotis and Kevin Sylwester
© 2014 John Wiley & Sons Ltd
Tab
le5.
Dyn
amic
GM
MR
egre
ssio
ns(A
nnua
l),1
984–
2007
Est
imat
ion
met
hod
(1)
(2)
(3)
(4)
(5)
(6)
Dif
f-G
MM
Dif
f-G
MM
Dif
f-G
MM
Sys-
GM
MSy
s-G
MM
Sys-
GM
M
GD
P(−
1)0.
826
1.07
0.83
0.99
1.00
1.01
(0.0
06)*
**(0
.009
)***
(0.0
7)**
*(0
.001
)***
(0.0
01)*
**(0
.003
)***
CO
0.00
40.
018
0.01
0.02
10.
019
0.00
3(0
.002
)**
(0.0
03)*
**(0
.004
)***
(0.0
02)*
**(0
.001
)***
(0.0
01)*
*D
EM
0.00
70.
029
(0.0
02)*
**(0
.001
)***
DE
M_P
S0.
110.
15(0
.022
)***
(0.1
32)*
**P
OL
ITY
0.00
50.
001
(0.0
01)*
**(0
.000
)***
CO
×D
EM
−0.
003
−0.
006
(0.0
00)*
**(0
.000
)***
CO
×D
EM
_PS
−0.
025
−0.
041
(0.0
05)*
**(0
.003
)***
CO
×P
OL
ITY
−0.
002
−0.
002
(0.0
00)*
*(0
.001
)**
No.
ofco
untr
ies
119
119
119
119
119
119
No.
ofob
serv
atio
ns28
2328
2325
5628
2328
2325
56Sa
rgan
test
(p-v
alue
)0.
260.
530.
420.
100.
110.
14A
R(2
)te
st(p
-val
ue)
0.46
0.11
0.18
0.45
0.57
0.11
Not
es:
Dep
ende
ntva
riab
leis
real
GD
Ppe
rca
pita
(PP
P).
All
regr
essi
ons
cont
ain
peri
odfix
edef
fect
s.St
anda
rder
rors
inpa
rent
hese
s.*,
**,*
**D
enot
esi
gnifi
canc
eat
10%
,5%
and
1%,r
espe
ctiv
ely.
The
valu
efo
rA
R(2
)de
note
sth
ep-
valu
eof
ase
rial
corr
elat
ion
test
taki
ngth
enu
llof
nose
rial
corr
elat
ion
inth
efir
stdi
ffer
ence
sof
the
depe
nden
tva
riab
le.T
heva
lue
for
Sarg
ande
note
sth
ep-
valu
eof
the
Sarg
anov
erid
enti
ficat
ion
test
taki
ngth
enu
llth
atth
em
odel
isco
rrec
tly
iden
tifie
d.
DEMOCRACY–AUTOCRACY CORRUPTION EFFECTS 591
© 2014 John Wiley & Sons Ltd
upon the type of political regime. Despite the reasons as to why corruption anddemocracy are exogenous to growth, the concern remains that the two are endog-enous. Future work will also attempt to better address such concerns.
Appendix: Country Sample
Albania, Algeria, Angola, Argentina, Australia, Austria, Bahamas, Bahrain, Bangla-desh, Belgium, Bolivia, Botswana, Brazil, Brunei, Bulgaria, Burkina Faso, Cam-eroon, Canada, Chile, China, Columbia, Congo (Dem.), Congo (Rep.), Costa Rica,Cote d’Ivoire, Cuba, Cyprus, Denmark, Dominican Republic, Ecuador, Egypt, ElSalvador, Ethiopia, Finland, France, Gabon, The Gambia, Ghana, Greece, Guate-mala, Guinea, Guinea-Bissau, Guyana, Haiti, Honduras, Hungary, Iceland, India,Indonesia, Iran, Iraq, Ireland, Israel, Italy, Jamaica, Japan, Jordan, Kenya, Korea,Kuwait, Lebanon, Liberia, Libya, Luxembourg, Madagascar, Malawi, Malaysia,Mali, Malta, Mexico, Mongolia, Morocco, Mozambique, Namibia, Netherlands, NewZealand, Nicaragua, Niger, Nigeria, Norway, Oman, Pakistan, Panama, Papua NewGuinea, Paraguay, Peru, Philippines, Poland, Portugal, Qatar, Romania, SaudiArabia, Senegal, Sierra Leone, Singapore, Somalia, South Africa, Spain, Sri Lanka,Sudan, Suriname, Sweden, Switzerland, Syria, Tanzania, Thailand, Togo, Trinidadand Tobago, Tunisia, Turkey, Uganda, UAE, UK, USA, Uruguay, Venezuela,Vietnam, Zambia, Zimbabwe.
References
Acemoglu, Daron, Simon Johnson, James A. Robinson, and Pierre Yared, “Income andDemocracy,” American Economic Review 98 (2008):808–42.
Ades, Alberto and Rafael Di Tella, “Rents, Competition, and Corruption,” American Eco-nomic Review 89 (1999):982–93.
Aidt, Toke, Jayasri Dutta, and Vania Sena, “Governance Regimes, Corruption and Growth:Theory and Evidence,” Journal of Comparative Economics 36 (2008):195–220.
Ali, Abdiweli and Hodan Isse, “Determinants of Economic Corruption: A Cross-country Com-parison,” Cato Journal 22 (2003):449–69.
Campos, Nauro and Francesco Giovannoni, “Lobbying, Corruption and Political Influence,”Public Choice 131 (2007):1–21.
Campos, J. Edgardo, Donald Lien, and Sanjay Pradhan, “The Impact of Corruption on Invest-ment: Predictability Matters,” World Development 27 (1999):1059–67.
Ehrlich, Isaac and Francis Lui, “Bureaucratic Corruption and Endogenous Economic Growth,”Journal of Political Economy 107 (1999):270–93.
Giavazzi, Francesco and Guido Tabellini, “Economic and Political Liberalizations,” Journal ofMonetary Economics 52 (2005):1297–330.
Goel, Rajeev, Michael Nelson, and Michael Naretta, “The Internet as an Indicator of Corrup-tion Awareness,” European Journal of Political Economy 28 (2012):64–75.
Harstad, Bård and Jakob Svensson, “Bribes, Lobbying and Development,” American PoliticalScience Review 105 (2011):46–63.
Huntington, Samuel, Political Order in Changing Societies, New York: Yale University Press(1968).
Knack, Stephen and Phillip Keefer, “Institutions and Economic Performance: Cross-country Tests Using Alternative Institutional Measures,” Economics & Politics 7 (1995):207–27.
Leff, Nathaniel, “Economic Development through Bureaucratic Corruption,” AmericanBehavioral Scientist 8 (1964):8–14.
592 Andreas Assiotis and Kevin Sylwester
© 2014 John Wiley & Sons Ltd
Marshall, Monty G. and Keith J. Jaggers, “Polity IV Project: Political Regime Characteristicsand Transitions, 1800–2004 Dataset Users’ Manual,” Center for Global Policy, School ofPublic Policy, George Mason University, available at www.bsos.umd.edu/cidcm/polity/(2004).
Mauro, Paolo, “Corruption and Growth,” Quarterly Journal of Economics 110 (1995):681–712.———, “Corruption and the Composition of Government Expenditure,” Journal of Public
Economics 69 (1998):263–79.Mendez, Fabio and Facundo Sepulveda, “Corruption, Growth and Political Regimes: Cross
Country Evidence,” European Journal of Political Economy 22 (2006):82–98.Méon, Pierre-Guillaume and Khalid Sekkat, “Does Corruption Grease or Sand the Wheels of
Growth?” Public Choice 122 (2005):69–97.Mo, Pak Hung, “Corruption and Economic Growth,” Journal of Comparative Economics 29
(2001):66–79.Murphy, Kevin, Andrei Schleifer, and Robert Vishny, “Why is Rent-seeking So Costly to
Growth,” American Economic Review 83 (1993):409–14.Murtin, Fabrice and Romain Wacziarg, “The Democratic Transition,” NBER working paper
17432, Cambridge, MA (2011).Papaioannou, Elias and Gregorios Siourounis, “Democratization and Growth,” Economic
Journal 118 (2008):1520–51.Rivera-Batiz, Francisco, “Democracy, Governance and Economic Growth: Theory and Evi-
dence,” Review of Development Economics 6 (2002):225–47.Rock, Michael, “Corruption and Democracy,” Journal of Development Studies 45 (2009):55–75.Rodrik, Dani and Romain Wacziarg, “Do Democratic Transitions Produce Bad Economic Out-
comes?” American Economic Review 95 (2005):50–57.Sachs, Jeffrey and Andrew Warner, “Economic Reform and the Process of Global Integra-
tion,” Brookings Papers on Economic Activity 1 (1995):1–118.Shen, Ce and John Williamson, “Corruption, Democracy, Economic Freedom, and State
Strength: A Cross-national Analysis,” International Journal of Comparative Sociology 46(2005):327–45.
Shleifer, Andrei and Robert Vishny, “Corruption,” The Quarterly Journal of Economics 108(1993):599–617.
Svensson, Jakob, “Eight Questions about Corruption,” Journal of Economic Perspectives 19(2005):19–42.
Swaleheen, Mushfiq Us and Dean Stansel, “Economic Freedom, Corruption, and Growth,”Cato Journal 27 (2007):343–58.
Tanzi, Vito and Hamid Davoodi, “Corruption, Growth and Public Finance,” IMF workingpaper 116, Washington DC (2000).
Wacziarg, Romain and Karen Horn Welch, “Trade Liberalization and Growth: New Evi-dence,” NBER working paper 5416, Cambridge, MA (2003).
Notes
1. See Knack and Keefer (1995), Ades and Di Tella (1999), Mo (2001) and Mauro (1998).2. Ideally, country-level measures of corruption would account for both the frequency andseverity of corrupt activities but it is doubtful that measures perfectly capture such nuances.3. Campos et al. (1999) consider how uncertainty from corruption affects investment.4. Results (available upon request) are robust to using 5-year averages.5. We remove observations where the absolute value of the growth rate exceeds 20%.Although 20% is arbitrary, some observations are clear outliers. In one case, annual per capitaGDP growth exceeded 80%. However, results are robust to including these outliers.6. A weakness of the Freedom House index is that it has components that are not exactlymeasures of democracy. For instance, the power of the citizenry to exercise the right to ownproperty, to make free economic resource-allocation decisions and to enjoy the fruits of such
DEMOCRACY–AUTOCRACY CORRUPTION EFFECTS 593
© 2014 John Wiley & Sons Ltd
decisions are all included. Another potential problem is that it is an ordinal measure offreedom, not a cardinal one.7. Their dataset ends in 2003. Therefore, to extend DEM_PS to 2007, we follow their steps incategorizing countries. Of note is that we removed Thailand from the set of democracies giventhe events of 2007.8. REFORM is based on five criteria. A country is considered closed if one of the followingholds: (1) average tariff rates exceed 40%, (2) nontariff barriers apply to more than 40% ofimports, (3) it has a socialist economic system, (4) it has a state monopoly of major exports, and(5) the black market premium exceeds 20%.9. We also considered using other technologies that could promote communication acrossopposition groups such as internet measures, but such measures are not available at the begin-ning of our sample, 1984. Replacing cell phone subscriptions with these other measures did notchange results.10. See Rock (2009) where he claims an inverted-U relationship between the age of democracyand corruption.11. For each country we find the difference in corruption between the first year and the lastyear in our sample period. Then, we obtain the value of the average change in corruptionbetween these two years for each group.12. Although our coefficient estimates greatly differ from initial estimates in Mendez andSepulveda (2006), they first consider a cross-section of countries and so employ a much differ-ent specification.13. We also considered life expectancy as another and results are robust. Swaleheen andStansel (2007) report that corruption’s effect depends upon the degree of economic freedom.Since many democracies are economically free, perhaps the democracy variables are merelyproxies for economic freedom. Using the same measure of economic freedom as do Swaleheenand Stansel (2007), our findings remain robust.
594 Andreas Assiotis and Kevin Sylwester
© 2014 John Wiley & Sons Ltd