Economic Analysis & Policy, Vol. 42 No. 1, march 2012
Corruption: Democracy, Autocracy, and
Political Stability
Kanybek Nur-tegin1
Wilkes Honors College, Florida Atlantic University,
5353 Parkside Drive, Jupiter, FL 33410,
U.S.A.
(Email: [email protected])
and
Hans J. Czap
University of Michigan–Dearborn,
FCS 121A, 19000 Hubbard Drive, Dearborn, MI 48126,
U.S.A.
(Email: [email protected])
Abstract: The recent empirical literature on corruption has identified a long list of variables that
correlate significantly with corruption but only five were distinguished by Leamer’s
Extreme Bounds Analysis as robust to various samples, measures of corruption, and
regression specifications. Among these five factors that were found to reduce corruption
are decades-long tradition of democracy and political stability. In today’s world,
however, there are many countries that combine one of these two robust determinants of
corruption with the opposite of the other: politically stable autocracies or newly formed
and unstable democracies. The central question raised in this paper is: Is it worth, in
terms of corruption, for a country to trade stability with autocratic rule for political
freedoms but with transitional instability? We find that the answer to this question is in
the affirmative – the level of corruption is indeed lower in unstable democracies than in
stable dictatorships. Our results are robust to various measures of corruption, alternative
regressor indices, and regression specifications.
“Power tends to corrupt, and absolute power corrupts absolutely. Great men
are almost always bad men…” (Sir John Acton, 18872)
I. Introduction
The world in 2010 and 2011 was shaken by an unprecedented wave of mass protests around
the globe against established and seemingly unshakable authoritarian regimes. Reportedly,
one of the main triggers was rampant corruption. As a result of these uprisings, significant
1Corresponding
2
author; primary authorship is not established.
http://en.wikipedia.org/wiki/John_Dalberg-Acton,_1st_Baron_Acton.
51
Corruption: Democracy, Autocracy, and Political Stability
political concessions were given in Algeria, Kuwait, Lebanon, Jordan, Oman, Morocco, Iraq,
and Bahrain. Libya plunged into civil war, Syria is on the brink of it, and protests still continue
in Thailand. Kyrgyzstan, Egypt, and Tunisia experienced complete regime changes3, moving
from autocracies to newly formed democracies. While the regime changes may be welcomed
by most citizens, there is a cost involved: long-lasting dictatorships provide political and
economic stability that new democracies initially cannot offer.
A move from dictatorship to democracy has many benefits for the great majority of a
country’s population. So it is arguably desirable in general, but perhaps not in every respect.
A good proportion of citizens in some countries appear to favor a strong authoritarian rule
over “excessive” political dialogs and freedoms. Russian politicians, for example, are used to
openly defending consolidation of power. Russian president Dmitry Medvedev told reporters
in a 2 July 2008 interview that his country requires “a strong executive leader” and that a
parliamentary method of governance “would mean the death of Russia as a country;” he
further added that “Russia must remain a presidential republic for decades or even hundreds
of years to come in order to stay united.”4 Perhaps even more striking were the results of an
online poll conducted by the Russian Television Channel, the Institute of Russian History of
the Russian Academy of Sciences, and the Public Opinion Fund, which in an effort to establish
the public’s opinion on “the most outstanding persona in the history of the nation” ended up
with preliminary results that placed Joseph Stalin in the lead with 160,000 votes.5
In comparing the merits of democracy and autocracy for a country’s development,
corruption takes an important place. It is a crippling problem in less developed, and often
autocratic, countries. However, established democracies are not free from it either. The rigged
bidding process for the construction of Terminal 2 at Germany’s Frankfurt Airport,6 frequent
corruption scandals in Italy,7 and the involvement of Detroit’s former mayor Kilpatrick in
illicit transactions are just a few recent examples.8
The relationship between regime types and corruption has received considerable research
interest, especially in the recent decades. The majority of theoretical inquiries find that democracy
lowers corruption in a country. In a typical principal-agent framework, citizens charge public
officials with a mandate to act in their interests, but are unable to perfectly align incentives
(Adsera et al, 2003). When information is imperfect, public officials may engage in corrupt
activities. Barro (1973) and Ferejohn (1986) suggest that regular voting provides a solution
to this problem through the increase in control. Thus, democratic voting provides incentives
for public officials to be less corrupt and to even actively fight corruption in order to increase
their chances for reelection (Persson and Tabellini 2000, Chowdury 2004, Treisman 2000).
3
4
5
6
7
8
52
See List of Revolutions and Rebellions 2010s; http://en.wikipedia.org/wiki/List_of revolutions_and_
rebellions#2010s
See AFP news report “Medvedev Eyes ‘Rational’ Democracy for Russia” online at afp.google.com/article/
ALeqM5isDarT_wZ2zTl-1ALq9iqY0CIBrg.
See “Surprisingly Enough, Many Russians Still Miss Stalin’s Strong Hand” online at english.pravda.ru/russia/
history/10-07-2008/105747-stalin-0/.
See The Financial Times (July 2, 1996), “German Airport Corruption Probe Deepens: Five Jailed and 20
Companies Under Investigation” and Reuters Business Report (September 25, 1996), “German Corruption
Wave Prompts Action,” as reported in Rose-Ackerman (1999, p.29).
See Reuters (February 17, 2010), “Corruption is surging in Italy, says state auditor.”
See The New York Times (December 16, 2010), “Kwame M. Kilpatrick.”
Kanybek Nur-tegin and Hans J. Czap
Gigliolo (1996) and da Silva (2000) further reason that democracies are better in protecting
the freedom of speech, which, through investigative journalism and public scrutiny, increases
the cost of corrupt behavior. Democracies also bring to power officials who are more likely to
work in the public’s interest (Zweifel and Navia 2000). Moreover, corruption may be lower
in democratic systems because public officials tend to change more frequently, which creates
uncertainty about whom to corrupt (Bohara et al. 2004). Lastly, greater civil liberties and a
more independent judiciary in democracies raise the effectiveness of anti-corruption measures
(Rose-Ackerman 1999, Schwartz 1999, Jamieson 2000, Moran 2001). In addition to these
direct effects, democracy has indirect ways of reducing corruption. Democratic reforms lead
to higher wages (Goldsmith 1995, Rodrik 1999), which has been shown to lower corruption
by decreasing the incentives for being corrupt (Sandholtz and Koetzle 2000, Van Rijckeghem
and Weder 2001).
There is also a great amount of empirical support for theoretical arguments that democratic
countries have less corruption (see, for example, Bohara et al. 2004, Chowdury 2004, Goldsmith
1999, and Sandholtz and Koetzle 2000). Adsera et al. (2003) find empirical evidence of a
negative, statistically and economically significant relationship between corruption and a wellinformed electorate. However, some studies, such as Ades and DiTella (1999) and Treisman
(2000), find no significant negative relationship between political rights and corruption.
According to Serra (2006), a negative relationship appears only when democracy is a decadesold established tradition.
Neither theoretical nor empirical evidence is unambiguous in supporting a negative
relationship between democracy and corruption. A number of studies exist that argue that
democracies are in fact not necessarily better in dealing with corruption. For example, Little
(1996), Johnston (1997), and della Porta and Vannucci (1999) have argued that democratic
voting systems may provide incentives for vote-buying or unlawful party financing. Montinola
and Jackman (2002) show that corruption is actually lower in dictatorships than in partial
democracies, although once democratization reaches a certain threshold, this relationship
changes and democratic regimes fare better. This result is supported by Sung (2004), who
accounts for nonlinearities in the relationship between democracy and corruption. His study
shows that starting from an authoritarian regime, democratization initially results in greater
corruption, before it eventually leads to a decline in corruption. Similar findings were established
by Mohtadi and Roe (2003) and Rock (2008).
If there is such an increase in corruption during democratization, it is very likely to be
related to the instability during the early years of transition from autocracy. As of December
2011, the situation even in those Arab Spring countries that had full regime changes, such
as Tunisia, Egypt, and Libya, remains unstable. Violence continues in Libya9 and new
demonstrations erupted in Egypt.10 The link between political instability and corruption has
been fairly well-established. According to Leite and Weidman (1999), instability means that
government officials do not have enough political clout to enact effective anti-corruption
measures. Campante et al. (2009) argue that instability shortens officials’ tenure, which in
turn is an incentive for increasing the rate of rent extraction. Furthermore, Treisman (2000)
9
10
http://topics.nytimes.com/top/news/international/countriesandterritories/libya/index.html
http://topics.nytimes.com/top/news/international/countriesandterritories/egypt/index.html
53
Corruption: Democracy, Autocracy, and Political Stability
and Persson and Tabellini (1999) suggest that political stability increases the value of building
a positive reputation and hence leads to less corruption.
To sum up, the consensus in the literature is that long-established democracies have less
corrupt governments. The picture is not as clear for countries with newly-acquired democratic
institutions, however, and corruption there may still be high due to, among other things, political
instability. In countries with recent or looming regime shakeups, such as Kyrgyzstan, Burma,
and Middle Eastern countries, an often-heard argument against changes is that new governments
may not be any better than the stable, albeit repressive, autocracies, especially in terms of
corruption. Thus, the question raised in this paper is: Does this argument have merit? That is,
how does corruption in new and politically unstable democracies compare to corruption in
countries with an autocratic but stable government? Before we proceed to the next sections, the
following remark is in order: we compare unstable democracies to stable dictatorships only in
terms of corruption. At the same time, despite this narrow focus, we believe that the findings
in this paper have broader implications as corruption has been shown to affect a number of
country performance indicators, such as economic growth (see Mauro, 1995).
II. Data and Estimation
Our goal in this paper is to examine how countries with secure and lasting dictatorships compare,
in terms of corruption, to countries with relatively recently acquired democratic regimes.
To do this we place countries along a continuum that increases from the most unstable and
most democratic regime to the most stable and most authoritarian regime. In order to ensure
robustness of our approach, we devise five versions of this continuum.
In the formulas listed below, A stands for concentration of power in the hands of one
individual (higher scores are associated with more authoritarian regimes). This variable was
created from the Polity2 variable in the Polity IV dataset of the Center for Systemic Peace. The
duration of the current regime is measured in years and rc­t is an operator that counts regime
changes. In other words, for every revolution or coup d’état within a specified time period rc­t
is equal to one for the years during which these regime changes occur. The time period during
which the number of regime changes is counted covers years 2000 – 2009.
All formulas multiplicatively interact the variable measuring concentration of power, A,
with an index of stability. In Formula 1, the index of stability is simply the reciprocal of the
number of regime changes that took place in a given country. We add “1” to the denominator
(as well as in formulas 2, 4, and 5) to avoid divisions by zero.
Formula 1
Formula 2 is identical to Formula 1 with one crucial difference: more recent regime changes
are given greater weights.
Formula 2
54
Kanybek Nur-tegin and Hans J. Czap
The stability index in Formula 3 is simply the duration of the current regime in years.
Formula 3
The next two formulas are derived by integrating the stability element in Formula 3 into
formulas 1 and 2.
Formula 4
Formula 5
Notice that all formulas produce scores that increase both with concentration of power
and stability of the regime. In other words, newly established democracies will have low
scores, while entrenched autocracies will have high scores. Further technical details on the
construction of the indices are provided in Appendix A.
Figure 1 provides a rough visual illustration of how a few selected countries may be
placed with respect to the stable dictatorship – unstable democracy line using scores based
on the formulas above.
Figure 1: Selected Countries along the Stable Dictatorship –
Unstable Democracy Continuum*
* Countries on this graph were placed approximately based on recent actual data.
55
Corruption: Democracy, Autocracy, and Political Stability
High pair-wise correlations of our fairly diverse indices (see summary statistics in Table
2) shown in Table 1 indicate that they are consistent in describing the underlying relationship
they are devised to capture.
Table 1: Correlation of Unstable Democracy – Stable Dictatorship Indices
Formula 1
Formula 2
Formula 3
Formula 4
Formula 5
Mean
Formula 1
1.000
0.984
0.976
0.986
0.977
0.984
Formula 2
Formula 3
Formula 4
Formula 5
1.000
0.983
0.999
0.999
0.993
1.000
0.990
0.981
0.986
1.000
0.998
0.994
1.000
0.991
The constructed unstable democracy – stable dictatorship index is used as the independent
variable of interest in the following regression model:
Corri = ß0 + ß1Xi + ß2Zi + εi,
(1)
where Corri is corruption, Xi is one of our indices, and Zi is a vector of other determinants of
corruption.
Data for the dependent variable in equation 1 is periodically collected by a number
of sources. However, according to Serra (2006), aggregate corruption indices, such as the
Corruption Perceptions Index (CPI) produced by Transparency International and corruption
scores from the Worldwide Governance Indicators (WGI) by a World Bank team (Kaufmann,
Kraay, and Mastruzzi) are more reliable than simpler indices produced by individual sources.
Following Serra’s judgment and for robustness, we use both CPI and WGI for 2009 as our
dependent variable.
The recent empirical literature provides a long list of significant determinants, some of
which we include in vector Zi in equation 1. Among the most robust (Serra 2006) are decadeslong tradition of democratic institutions, political instability, GDP per capita, British colonial
heritage, and mainly Protestant population. We do not explicitly include the first two in our
analysis because these variables are already used in the composition of our main variable of
interest. Furthermore, they are highly correlated with our constructed indices. The level of
economic development, most often proxied by per capita GDP, is virtually always included
as one of the principal independent variables. For this paper, we have used the average of
2007 through 2009 GDP per capita based on purchasing power parity obtained from the World
Development Indicators (WDI) database. Most of the data on British colonial heritage was
taken from Price (2003) with missing values supplemented from CIA World Factbook and
other sources. La Porta et al. (1999) was our source for assembling the column of observations
on the percent of population following the Protestant religion.
We continue to pursue robustness by including other regressors found elsewhere in crosscountry analyses to significantly affect corruption. Greater openness to trade was found to
reduce corruption (see Ades and Di Tella 1999, Brunetti and Weder 2003, and Persson et al.
56
Kanybek Nur-tegin and Hans J. Czap
Table 2: Summary Statistics
Average
Std.
Deviation
Min
Max
Index 1
7
6.72
0
21
X2
Index 2
7
7.06
0
21
X3
Index 3
66
61.21
0
189
X4
Index 4
69
70.07
1
210
Name
Description
X1
X5
Index 5
67
71.49
0
210
Dur
Duration of regime (years)
17
15.45
0
83
Corr
Corruption (CPI 2009)
3.25
1.41
1.10
9.20
Corr
Corruption (WGI 2009)
-0.45
0.73
-1.75
2.26
LnGDPpc Log GDP per capita in PPP, mean of
2007-2009
8.22
1.23
5.66
10.77
Protest
Protestant (% of population)
7.6
13.52
0.00
66.00
BritCol
British colonial heritage (dummy: 1
if yes, 0 otherwise)
0.23
0.42
0
1
Trade
Imports + Exports, % of GDP, mean
of 2000-2009
84.21
38.79
0.67
200.46
Lit
Literacy rate (%)
79.67
20.51
26.18
99.80
LnPop
Log total population, 2009
16.22
1.45
13.34
21.01
Eng
Legal origin (dummy: 1 if English, 0
otherwise)
0.23
0.42
0
1
Soc
Legal origin (dummy: 1 if Socialist,
0 otherwise)
0.30
0.46
0
1
Fre
Legal origin (dummy: 1 if French, 0
otherwise)
0.45
0.50
0
1
Ger
Legal origin (dummy: 1 if German, 0
otherwise)
0.02
0.13
0
1
ELF1
Ethnolinguistic fractionalization
(ELF 1)
0.15
0.17
0.00
0.56
ELF8
Ethnolinguistic fractionalization
(ELF 8)
0.43
0.29
0.00
0.98
ELF15
Ethnolinguistic fractionalization
(ELF 15)
0.51
0.30
0.00
0.99
--
PolityIV*
2
6.32
-10
10
Total number of observations = 113; CPI ranges from 0 (highly corrupt) to 10 (very clean); WGI ranges from
-2.5 to 2.5 with higher number indicating less corruption; EFL(j), which ranges from 0 (very homogenous) to 1
(highly diverse), is a measure of ethnolinguistic fractionalization at the j’th level of aggregation constructed by
Desmet et al. (2011).
* PolityIV is not included directly in the regressions, but was used for constructing our indices. PolityIV scores
range from -10 (fully institutionalized autocracy) to 10 (fully institutionalized democracy).
57
Corruption: Democracy, Autocracy, and Political Stability
2003). Our trade variable is an average of the sums of WDI imports and exports as percent of
GDP over 2000-2009. We have taken the average over such a long period of time to account
for high annual volatility of trade. Country size, proxied by total population, has also come up
as a potentially relevant variable (see Knack and Azfar 2003, and Fisman and Gatti 2002). Our
country size variable is created by transforming WDI total population data into logarithms.
Another often-used independent variable is the degree of ethnolinguistic fractionalization
(e.g. Mauro 1995, La Porta et al. 1999, and Cerqueti et al. 2009). Desmet et al. (2011) have
developed an advanced method of classifying countries by their ethnolinguistic fractionalization
(ELF) and produced fifteen ELF indices based on various degrees of linguistic aggregation.
We chose ELF1, ELF8, and ELF15 – highest, medium, and lowest levels of aggregation,
respectively – to include in our regressions. It would seem reasonable to assume that countries
with more educated citizens are more likely to have more accountable bureaucrats (see Persson
et al. 2003). To investigate this possibility, we include literacy rates in one of our regressions.
Regional factors may also play a role (Treisman 2000). To account for this, we use regional
dummy variables made available by Caselli and Coleman (2001), who split the world into
East Asia, East Europe, Arab, Latin America and Caribbean, Sub-Saharan Africa, and Other
Asia. Finally, we follow La Porta et al. (1999), Treisman (2000), and Serra (2006) by including
dummies for legal origin (English, German, French, Socialist, and Scandinavian) based on the
La Porta et al. dataset. More details on all variables are given in Table 2: Summary Statistics.
As indicated in Table 2, the number of countries in our final dataset was 113. In addition,
for ease of interpretation all our indices and both corruption proxies are measured in standard
deviations from their respective means.
iii. Results
We investigate the relationship described by equation (1) above using the ordinary least squares
method with White-corrected robust standard errors. The results are reported in Table 3.
The estimated coefficient on the unstable democracy – stable dictatorship index (in this
case X1), the main variable of interest, is negative, sizable, and highly significant across various
specifications of the regression model. It shows that a movement of one standard deviation in
the index from unstable democracy to stable dictatorship worsens the corruption index (i.e.,
more corruption) by approximately one-sixths of one standard deviation from the mean in
CPI. It is important to note that other independent variables are consistent with established
empirical results in the literature. The coefficients on economic development (LnGDPpc)
and the percent of population following the Protestant religion (Protest) are correct in sign
and significant across various specifications. While the coefficients on all used measures of
ethnolinguistic fractionalization (ELF(j)) coincide in sign, only ELF1 (Desmet et al.’s highest
level of aggregation11) is significant. It indicates, as expected a priori, that countries with
greater ethnic and linguistic diversity suffer from more corruption. One surprising result in
Table 3 is that the coefficient on the dummy for former British colonies, which was found to
be robust in other studies (see Serra 2006), is not significant.
11
58
At high levels of aggregation, countries’ ethnolinguistic diversity is evaluated at the level of language families,
while at low levels of aggregation, every language is taken separately.
-3.9476
(-8.99)
113
0.4495
19.24
0.0000
(1)
-0.1800
(-2.74)
0.4578
(8.30)
0.0104
(2.07)
0.0542
(0.40)
-3.9576
(-8.80)
113
0.4497
15.41
0.0000
(2)
-0.1788
(-2.76)
0.4619
(7.40)
0.0106
(2.14)
0.0560
(0.40)
-0.0003
(-0.18)
-3.1459
(-3.94)
113
0.4553
15.45
0.0000
-0.0468
(-1.07)
(3)
-0.1815
(-2.75)
0.4528
(8.31)
0.0095
(1.91)
0.0749
(0.56)
-3.8754
(-8.70)
113
0.4742
16.02
0.0000
-0.8246
(-2.32)
(4)
-0.1628
(-2.54)
0.4613
(8.26)
0.0132
(2.59)
0.0818
(0.62)
-3.9020
(-8.43)
113
0.4497
15.55
0.0000
-0.0469
(-0.22)
(5)
-0.1780
(-2.71)
0.4547
(8.80)
0.0104
(2.08)
0.0552
(0.41)
-3.6766
(-7.02)
113
0.4541
15.37
0.0000
-0.2270
(-0.99)
(6)
-0.1703
(-2.68)
0.4374
(7.44)
0.0115
(2.33)
0.0753
(0.56)
-4.0093
(-8.92)
113
0.4523
15.39
0.0000
-0.0032
(-0.89)
(7)
-0.1806
(-2.76)
0.4956
(6.90)
0.0107
(2.06)
0.0546
(0.39)
-4.0465
(-5.47)
113
0.4667
9.61
0.0000
No
(p=0.38)
(8)
-0.1805
(-2.53)
0.4658
(5.17)
0.0079
(1.30)
0.0710
(0.52)
Yes
(p=0.001)
-3.1169
(-4.58)
113
0.4639
200.45
0.0000
(9)
-0.1615
(-2.41)
0.4437
(6.82)
0.0093
(1.67)
-0.0623
(-0.24)
t-statistics in parentheses are based on robust standard errors; *p-value indicates joint significance of dummies; regional dummies are based on Caselli and Coleman
(2001) and include Arab World, Latin America and Caribbean, Sub-Saharan Africa, Eastern Europe, East Asia, and Other Asia; legal origin dummies are from La
Porta et al. (1999) and include English, French, German, Socialist, and Scandinavian.
# of observations
R-squared
F
Prob > F
Constant
Legal origin dummies*
Regional dummies*
Lit
ELF15
ELF8
ELF1
LnPop
Trade
BritCol
Protest
LnGDPpc
X1
Table 3: OLS estimates. Dependent variable: CPI 2009; index: X1.
Kanybek Nur-tegin and Hans J. Czap
59
60
-0.8406
(-2.33)
Yes
(p=0.01)
-3.1356
(-4.49)
ELF1
Legal origin
Constant
0.4795
182.31
0.0000
R-squared
F
Prob > F
113
0.0124
(2.19)
Protest
# of observations
0.4484
(6.70)
0.4461
(6.70)
0.4488
(6.69)
lnGDPpc
0.0000
189.49
0.4838
113
-3.1319
(-4.52)
Yes
(p=0.01)
-0.8051
(-2.31)
0.0000
185.34
0.4806
113
-3.1369
(-4.50)
Yes
(p=0.01)
-0.8297
(-2.32)
0.0123
(2.19)
-0.1309
(-1.97)
-0.1807
(-2.48)
-0.1274
(-1.93)
X(j)
0.0122
(2.19)
CPI, Index4
CPI, Index3
CPI,
Index2
0.0000
181.02
0.4783
113
-3.1304
(-4.48)
Yes
(p=0.01)
-0.8477
(-2.35)
0.0124
(2.19)
0.4487
(6.68)
-0.1232
(-1.85)
CPI, Index5
0.0000
103.30
0.4291
113
-3.1875
(-4.16)
No
(p=0.17)
-1.1868
(-3.35)
0.0113
(1.98)
0.4256
(5.77)
-0.1642
(-2.27)
WGI, Index1
0.0000
99.53
0.4225
113
-3.1750
(-4.11)
No
(p=0.16)
-1.2345
(-3.52)
0.0118
(2.04)
0.4273
(5.75)
-0.1455
(-1.97)
WGI, Index2
Table 4: OLS Estimates. Robustness Check.
0.0000
103.24
0.4238
113
-3.1578
(-4.09)
No
(p=0.16)
-1.2016
(-3.47)
0.0117
(2.04)
0.4237
(5.71)
-0.1495
(-1.97)
WGI, Index3
0.0000
100.53
0.4223
113
-3.1708
(-4.10)
No
(p=0.16)
-1.2245
(-3.51)
0.0117
(2.04)
0.4265
(5.74)
-0.1450
(-1.94)
WGI, Index 4
0.0000
98.80
0.4211
113
-3.1695
(-4.09)
No
(p=0.16)
-1.2425
(-3.55)
0.0118
(2.05)
0.4273
(5.74)
-0.1410
(-1.90)
WGI, Index5
Corruption: Democracy, Autocracy, and Political Stability
Kanybek Nur-tegin and Hans J. Czap
To ensure that our results are robust to variations of our index and alternative corruption
proxies, we ran nine more regressions, with results presented in Table 4. Control variables
include only those factors that were found significant in regressions reported in Table 3. The
estimated coefficients on our main independent variable across all X(j) – corruption proxy
combinations is negative, consistent in size, and statistically significant.
A couple of concerns may initially be raised by readers with regard to using OLS in this
context. First, one may argue that slope estimates may be biased due to possible endogeneity
of both our indices and other control variables. Second, since we are interested in comparing
stable dictatorships to fragile democracies, our dataset excludes the other two categories, namely,
decades-long democracies and unstable dictatorships. In terms of Figure 1, we keep countries
that belong to quadrants II and IV and ignore industrialized countries, such as the United States
and Japan, and countries in the third quadrant, such as Sudan. Our analysis, therefore, may be
subject to censoring or sample-selection problems. Breen (1996, p. 35) notes that
“If we think hard enough, we can probably find some sort of selection process underlying any piece
of social science data. A random sample of the adult population is actually only a random sample
of those members of the population who are listed in the sampling frame; so if the sampling frame
is, say, an electoral register, those adults who are not registered cannot be sampled.”
The censoring problem, even if it applies our data, is very difficult to deal with in our case.
Partly, this is due to the values of the dependent variable associated with dropped observations
not being censored at one or both ends of the corruption continuum. In fact, even multiplethreshold censoring models do not apply here since our “unobserved” – or more precisely,
dropped – data is not censored at clearly specified intervals of the dependent variable, but are
discarded in a more complex process that involves a multiplicative interaction of two variables.
In addition, Breen (1996) and Wooldridge (2005) point out that the regressors of the outcome
equation must be a strict subset of the regressors of the selection equation.12 That is, we have
to have an exclusion restriction in the selection part of the two-step estimation procedure. An
exclusion restriction is a variable that affects the probability that an observation will manifest
itself in the main regression, but does not influence the dependent variable itself in the main
regression. Unfortunately, we were unable to find a variable that influences the probability that
a country belongs to the subset of stable autocracies and newly established democracies, but
has no effect on the level of corruption. Finally, it is likely that selection is not exogenously
determined from the level of corruption, which further complicates the matter by introducing
simultaneity (see Breen 1996, p. 55).
Despite these difficulties with running a full-scale Heckman or maximum likelihood
estimation for investigation of the possibility of sample selection, it is possible to get some
insight on the consequences of considering all types of countries, including the originally
ignored group. We do this by simply representing established dictatorships and unstable
democracies as dummy variables against the base of the remaining countries, i.e. the combined
group of countries with established democracies and unstable dictatorships. The results of
this estimation are presented in Table 5. Note that the number of countries in this regression
is 152, up from 113.
12
See Nur-tegin (2008) for more details.
61
Corruption: Democracy, Autocracy, and Political Stability
Table 5: OLS Estimates with Extended Sample of Countries.
Dependent variable: CPI
Unstable democracy dummy
Stable dictatorship dummy
lnGDPpc
Protest
ELF1
Legal origin dummies
Constant
# of observations
R-squared
F
Prob > F
Coefficient
t-statistic
-0.1661
-0.3089
0.4491
0.0101
-0.6578
Yes
-3.2168
152
0.7163
84.30
0.0000
(-1.76)
(-2.13)
(11.28)
(2.36)
(-2.48)
(p=0.0006)
(-6.63)
Both unstable democracy and stable dictatorship dummies are negative and significant,
which indicates that corruption levels are higher in these countries compared to the base
group. This is unsurprising because all observations of the base group, with the exception
of only Bhutan, Mauritania, Egypt, and Sudan, are long-established democracies (see Serra
2006). More importantly, the coefficient for entrenched autocracies is greater in magnitude
and more significant than the coefficient for unstable democracies. The absolute value of the
difference between the two coefficients is equal to 0.1428, which is strikingly close to 0.1469
– the average of the size of estimated coefficients on the constructed index, X(j), in all nine
regressions in Table 4 and the slope estimate on X1 in column (9) of Table 3.13 Therefore, the
results in Table 5 seem to support our main findings in Tables 3 and 4.
Turning to endogeneity, it does not appear to be a serious issue for our main variable of
interest. First, our index is constructed by multiplicatively combining two variables, autocracy
and stability, and it is difficult to make a convincing argument that the resulting variable will be
affected by corruption in any particular direction. Second, even when autocracy and stability are
taken separately, it is unclear how corruption changes these variables. Corruption may sustain
autocracy by providing financial resources to the ruling elite, or it may undermine autocracy by
engendering non-ruling but powerful political clans.14 The stability of an autocrat’s regime may
benefit from corruption by making it more powerful, or corruption may lead to a revolution.
Third, Pellegrini and Gerlagh (2008) reason against endogeneity of their persistent democracy
variable in the regression on corruption and confirm their judgment using the Hausman test.
Fourth, and most importantly, our index is lagged and averaged over ten years (2000-2009),
while the dependent variable is constructed from CPI and WGI scores for the last year, 2009.
As for the remaining independent variables, the only ones that may give reasons for concerns
about endogeneity are the level of economic development (lnGDPpc) and openness to trade
13Columns
(4) and (9) in Table 3 are the most comparable columns in terms of the number of significant
variables to regressions in Table 3. Replacing column (9)’s estimate with column (4)’s estimate changes the
average by a very negligible magnitude.
14
E.g., in Kyrgyzstan many of the most prominent opposition leaders have held a top government position at
some point in the past and made most of their fortunes while in office.
62
Kanybek Nur-tegin and Hans J. Czap
(Trade). Both of these variables are lagged and averaged over several years, which should be
sufficient to avoid problems due to simultaneity given that we are interested in these variables
only as control variables.
IV. Conclusion
In this paper, we find strong empirical evidence that democracies, even if they are politically
unstable, have public officials who are less corrupt than their counterparts in countries with
authoritarian regimes. This result is robust across various measures that combine the regime
type with political stability and across different regression specifications. An important
feature of our study is that we focus only on countries that lie along the stable dictatorship –
unstable democracy continuum (along the line in Figure 1) and exclude countries with shaky
autocracies and well-established democratic systems of governance. We do this because
existing research on corruption in the latter two groups of countries is fairly unambiguous.
However, to our knowledge, the question about the level of corruption in stable dictatorships
vis-à-vis unstable democracies has thus far remained unresolved. Our results fill this gap
and show that the global drive toward more open and better-represented societies should not
be diminished due to the unfounded fear that the new government will be more corrupt than
the overthrown autocrats.
Appendix A: Details of Constructing the Index
(1) We start with PolityIV (Polity2 column in PolityIV dataset), which ranges from -10 (most
autocratic) to +10 (most democratic) scores that span the time period from 1980 to 2009.
(2) These Polity scores are transformed into a new variable, which we call Autocracy, by
subtracting Polity from 11. The strength of autocracy increases as Autocracy goes from
1 (most democratic) to 21 (most autocratic). This variable is constructed from Polity for
the last year only, 2009, i.e. not an average of a number of recent years. The averaging
is not important because most countries have consistent Polity scores for a number of
recent years.
(3) The variable Durable is the number of years the current regime has been in place. Every
time a regime change occurs (as a result of a revolution, coup, etc.), the count drops to
zero.
StaDem is a variable, which includes countries that in year 2009 were categorized as Stable
Democracies. A country’s StaDem score is created by taking the average of Autocracy
scores for this country over the period of 25 years (1985-2009), if this average Autocracy
score is less than 5 (fairly democratic) and if the regime hasn’t changed over this period
(i.e. Durable doesn’t drop to zero at any point during this period). I.e. these are countries
that have consistently been democratic for the past 25 or more years.
63
Corruption: Democracy, Autocracy, and Political Stability
35 countries that belong to StaDem were excluded from the main analysis. These countries are:
Argentina, Australia, Austria, Belgium, Bolivia, Botswana, Canada, Colombia, Costa Rica,
Cyprus, Denmark, El Salvador, Finland, France, Germany, Greece, Honduras, India, Ireland,
Israel, Italy, Jamaica, Japan, Mauritius, Netherlands, New Zealand, Norway, Portugal, Spain,
Sweden, Switzerland, Trinidad, Turkey, United Kingdom, and United States.
(4) UnstDict1 was created to represent countries that in the year 2009 belonged to the Unstable
Dictatorship group. A country’s UnstDict1 score was created by taking the average of
Autocracy scores for this country over the period of 7 years (2003-2009), if this average
Autocracy score is greater than 17 (fairly dictatorial) and if the current regime’s duration
has lasted only for the past six years or less (unstable).
Countries that belong to UnstDict1 were excluded from the main analysis. There are no
countries in this group.
(5) UnstDict2 is an alternative index created to represent countries that in year 2009 belonged
to the Unstable Dictatorship group. A country’s UnstDict2 score was created by taking
the average of Autocracy scores for this country over the period of 10 years (2000-2009),
if this average Autocracy score is greater than 15 (fairly dictatorial) and if the current
regime’s duration has lasted only for the past nine years or less (unstable).
4 countries that belong to UnstDict2 were excluded from the main analysis. These countries
are: Bhutan, Egypt, Mauritania, and Sudan.
(6)Next, we take the sample of remaining countries (after excluding StaDem and UnstDict2)
and calculate the five indices at year 2009 (i.e. filtering away all other years), as shown
in Formulas 1-5, for time periods 1992-2009 and 2000-2009. That is in the Short time
period, Durable is truncated at 10, while in the Long it is truncated at 18. Also, the number
of regime changes are different in the two time periods; rc(Short) ≤ rc(Long).
Note that for the short period (2000-2009) in Formula 3 the variable Durable is truncated
at 9, i.e. nine years max (which actually should have been 10).
(7) Once we get the raw index scores, we transform them into standard deviations from the
mean.
64
Kanybek Nur-tegin and Hans J. Czap
REFERENCES
Ades, A. and R. Di Tella (1999). Rents, Competition, and Corruption. The American Economic Review.
89: 982-993.
Adsera, A, C. Boix and M. Payne (2003). Are You Being Served? Political Accountability and Quality
of Government. Journal of Law, Economics, & Organization. 19: 445-490.
AFP News Agency (2008). Medvedev Eyes ‘Rational’ Democracy for Russia. Published on July 2,
2008. Available on the web: afp.google.com/article/ALeqM5isDarT_wZ2zTl-1ALq9iqY0CIBrg,
accessed June 25, 2011.
Barro, R. (1973). The Control of Politicians. An Economic Model, Public Choice. 14: 19-42.
Bohara, A., N. Mitchell, and C. Mittendorff (2004). Compound Democracy and the Control of Corruption:
A Cross-Country Investigation, Policy Studies Journal. 32: 481-499.
Breen, R. (1996). Regression Models: Censored, Sample-Selected, or Truncated Data. Sage University
Paper series on Quantitative Applications in the Social Sciences, 07-111. Thousand Oaks, CA: Sage.
Campante, F.R., D. Chor, and Q-A. Do (2009). Instability and the Incentives for Corruption, Economics
& Politics. 21: 42-92.
Caselli, F. and W. Coleman (2001). Cross-Country Technology Diffusion: The Case of Computers.
NBER working paper.
Cerqueti, R., R. Coppier, and G. Piga (2009). Corruption, Growth and Ethnolinguistic Fractionalization:
A Theoretical Game Model. Unpublished manuscript.
Chowdury, S.K. (2004). The Effect of Democracy and Press Freedom on Corruption: An Empirical
Test, Economics Letters. 85: 93-101.
da Silva, C.E.L. (2000). Journalism and Corruption in Brazil, in: J.S. Tulchin and R.H. Espach (eds.),
Combating Corruption in Latin America. Washington, D.C.: Woodrow Wilson Center Press: 173-192.
della Porta, D.and A. Vannucci (1999). Corrupt Exchanges: Actors, Resources, and Mechanisms of
Political Corruption. New York: Aldine de Gruyter.
Desmet, K., L. Ortuno-Ortin, and R. Wacziarg (2011). The Political Economy of Linguistic Cleavage,
Journal of Development Economics. Forthcoming.
Ferejohn, J. (1986). Incumbent Performance and Electoral Control, Public Choice. 50: 5-25.
Gigliolo, P.P. (1996). Political Corruption and the Media: The Tangentopoli Affair, International Social
Science Journal. 48: 381-394.
Goldsmith, A.A. (1995). Democracy, Property Rights, and Economic Growth, Journal of Development
Studies. 32: 157-174.
Fisman, R. and R. Gatti (2002). Decentralization and Corruption: Evidence Across Countries, Journal
of Public Economics. 83: 325-345.
Jamieson, A. (2000). The Antimafia: Italy’s Fight Against Organized Crime. New York: Palgrave
MacMillan.
Johnston, M. (1997). Public Officials, Private Interests, and Sustainable Democracy: When Politics
and Corruption Meet, in: K.A. Elliott (ed.), Corruption and the Global Economy. Washington, DC:
Institute for International Economics: 61-82.
Knack, S. and O. Azfar (2003). Trade Intensity, Country Size and Corruption, Economics of Governance.
4: 1-18.
La Porta, R., F. Lopez-de-Silanes, A. Shleifer, and R. Vishny (1999). The Quality of Government,
Journal of Law, Economics, and Organization. 15: 222-279.
Leite, C. and J. Weidman (1999). Does Mother Nature Corrupt? Natural Resources, Corruption, and
Economic Growth. International Monetary Fund Working Paper, WP/99/85, pp. 1-34.
Little, W. (1996). Corruption and Democracy in Latin America, IDS Bulletin, 26, pp. 64-70.
Mauro, P. (1995). Corruption and Growth, Quarterly Journal of Economics. 110: 681-712.
Mohtadi, H. and T. Roe (2003). Democracy, Rent Seeking, Public Spending and Growth, Journal of
Public Economics. 87:.445-466.
65
Corruption: Democracy, Autocracy, and Political Stability
Montinola, G.R. and R.W. Jackman (2002). Sources of Corruption: A Cross-National Study, British
Journal of Political Science. 32: 147-170.
Moran, J. (2001). Democratic Transition and Forms of Corruption, Crime, Law, and Social Change.
36: 379-393.
Nur-tegin, K. (2008). Determinants of Business Tax Compliance, B. E. Journal of Economic Analysis
and Policy. 8: Topics.
Pellegrini, L. and R. Gerlagh (2008). Causes of Corruption: A Survey of Cross-Country Analyses and
Extended Results, Economics of Governance. 9: 245-263.
Persson, T., G. Tabellini, and F. Trebbi (2003). Electoral Rules and Corruption, Journal of the European
Economic Association. 1: 958-989.
Pravda.ru (2008). Surprisingly Enough, Many Russians Still Miss Stalin’s Strong Hand. Available on
the web: http://english.pravda.ru/history/10-07-2008/105747-stalin-0/, accessed June 25, 2011.
Rock, M. (2008). Corruption and Democracy, Journal of Development Studies. 45: 55-75.
Rodrik, D. (1999). Democracies Pay Higher Wages, Quarterly Journal of Economics. 114: 707-738.
Rose-Ackerman, S. (1999). Corruption and Government: Causes, Consequences, and Reform. New
York: Cambridge University Press
Sandholtz, W. and W. Koetzle (2000). Accounting for Corruption: Economic Structure, Democracy,
and Trade, International Studies Quarterly. 44, 31-50.
Schwartz, H. (1999). Surprising Success: The New Eastern European Constitutional Courts, in: A.
Scheller, L. Diamond, and M.F. Plattner (eds.), The Self-Restraining State: Power and Accountability
in New Democracies. Boulder, CO: Lynne Rienner Publisher: 195-214.
Serra, D. (2006). Empirical Determinants of Corruption: A Sensitivity Analysis, Public Choice. 126:
225-256.
Sung, H.E. (2004). Democracy and political corruption: A cross-national comparison, Crime, Law and
Social Change. 41: 179-194.
Treisman, D. (2000). The Causes of Corruption: A Cross-National Study, Journal of Public Economics.
76: 399-457.
Van Rijckeghem, C. and B. Weder (2001). Bureaucratic Corruption and the Rate of Temptation: Do
Wages in the Civil Service Affect Corruption, and By How Much?, Journal of Development
Economics. 65: 307-331.
Wikipedia (2010). The Free Encyclopedia. List of Revolutions and Rebellions 2010s. Available on the
web: http://en.wikipedia.org/wiki/List_of revolutions_and_rebellions#2010s, accessed July 3, 2011.
Wooldridge, J. (2005). Introductory Econometrics: A Modern Approach. Mason, Ohio: South-Western
College Publishing.
Zweifel, T. and P. Navia (2000). Democracy, Dictatorship, and Infant Mortality, Journal of Democracy.
11: 99-114.
66