Legal Corruption
Daniel Kaufmann
Pedro C. Vicente
Oxford University
November 9, 2005
Conventional Definition of Corruption
• Abuse (usually taken as an illegal act)…
• of public office (a public-sector centered
definition)…
• for private gain.
But what about…
• A private individual buying a law (legal in a
number of countries, through allowed
lobbying), or…
• A private-sector firm employing a well-
known former politician to its advisory
council.
Question of the Paper
• Challenge conventional definition of corruption,
instead positing a broader notion to understand…
• What are the main simple determinants of the
world pattern of legal and illegal corruption
– Original political economy model of corruption,
presenting a simple explanation
– Use of broad range of different empirical counterparts
for testing, namely
• Newly available empirical measures of legal corruption,
from a worldwide firms’ survey in 104 countries
• Other external data sources
The Model
• Infinitely repeated complete information game
• 3 agents with individual ability a
• Every period an auction of a “favor” takes place
– Initial auctioneer is pre-defined
– Winner is next period auctioneer
– Bidders submit contract offers
• Auctioneer raises a at a given period; others
raise 0 (let’s call the loser population)
• Corruption
Date 0 Date 1 Date 2 Date 3
Auctioneer Bidders Auctioneer Bidders Auctioneer Bidders Auctioneer Bidders
1 2 2 1 1 2 2 1
…
3 3 3 3
• Auctioneer may choose to buy Legal Barriers
(proportion ϖ of transferred earnings)
– Legal Barriers inflict “myopia” on the loser, i.e.
her utility horizon is the present period
• (collective action undermined)
• Population may start an insurrection at the
end of period
– Insurrection is successful if λ(.)> λ ; o/w prob. ε
– λ(.) function depending on:
• General ability a (idea: potential for destruction)
• Raised amounts α by pop in last 3 periods (idea:
potential to “arm an army” by pop)
• Stage Game
Auctioneer
Decides
Bidders
Auctioneer Whether to Population Decides for
Submit
Decides Winner Build Legal Insurrection or Peace
Contract
Barriers
Offers
Time Line (Stage Game)
• Payoffs: ∞
– For all agents: ∑ [ δ i (1 − s ) w i ] , where we is
t e
t=0
end-of-period wealth
• For the Auctioneer: w = w + τ + a − ϖ
e b
• For the Winner: w = w + τ
e b
• For the Loser: w = w
e b
• Contract Incompleteness (CI):
– in the sense that (1-s) cannot be negotiated – it
has to be consumed
Equilibrium
• Inequality high (1/α) / Ability (a) low
⇒Insurrections started, corruption, no legal barriers
(unstable)
Intuition: CI=>power is best; pop almost (unstable) no threat;
• Inequality low / Ability high
– Accountability (ϖ) low
⇒No Insurrections started, corruption, legal barriers
(unstable?)
Intuition: pop is threat; worth spending ϖ; inequality increases
(possible instability in the long run)
– Accountability (ϖ) high
⇒No Insurrections started, no corruption, no legal barriers
(stable)
Intuition: pop threat, legal barriers too expensive; pop offered
insurrection payoff=>rotation of all agents (stable)
Testable Implications
Table 1:
Ability (Productivity) Low High
Exogenous /Equality
Accountability Low High Low High
Legal Corruption No No Yes No
Endogenous Illegal Corruption Yes Yes No No
Insurrections Yes Yes No No
• Three patterns of interaction of exogenous vs.
endogenous variables
Data
• New Dataset with proxies for Legal Corruption:
– Executive Opinion Survey (EOS) – Global Competitiveness
Report 2004-2005, World Economic Forum
• mail-based survey, 8729 firms, 104 countries
• Broad range of proxies for other variables (both from
EOS and other databases)
• Endogenous Variables
• Illegal Corruption:
– EOS Financial Honesty of Politicians
– EOS Frequency of Illegal Political Contributions
– EOS Frequency of Diversion of Public Funds Due to Corruption
– EOS Frequency of Bribery as State Capture
– KKM Control of Corruption (2002)
• Legal Corruption:
– EOS Favoritism in Policy and Procurement
– EOS Frequency of Legal Political Contributions
– EOS Influence in Laws and Regulations
» Adjust by Rule of Law: EOS Frequency of Bribery in Judicial
Decisions and KKM Rule of Law (2002)
• Insurrections:
– EOS Common Crime
– iJET Risk of Travel (2004)
– EIU - Armed Conflict, Violent Demonstrations, Violent Crime, Social
Unrest (2003)
– Civil War Dummy - constructed from Gleditsch et al, 2001 (1990-01)
• Parameters (Exogenous Variables)
• Ability (Productivity):
– Lagged logGDP per capita (1984)
• Equality:
– 100-Gini Coefficient (2002)
– EOS Equality in Healthcare
• Accountability:
– EOS Freedom of Press
– Freedom House:
» Civil Liberties (2003)
» Press Freedom (2004)
– KKM Voice and Accountability (2002)
– Government Fractionalization (2000) from DPI - Database of Political
Institutions, Beck et al, 2001
Simple Empirical Tests: Averages
• From first two rows in Table 1
– differences lower-upper quartiles/halves in terms
of income p/c or equality higher for illegal
corruption than for legal corruption
• From third row in Table 1
– Insurrection proxies are higher in lower income
p/c or equality groups of countries
• Note that differences are higher using 1st vs.
other quartiles than using halves
⇒we focus on the first
Simple Empirical Tests: Correlations
• From first two rows and first three columns in
Table 1
– If we take differences legal-illegal corruption and
take out lowest legal corruption countries
⇒ Clear positive correlations across countries should
arise with income p/c and equality
• From third row in Table 1
⇒ Clear negative correlations should arise with
income p/c and equality
• From first row in Table 1
– If we take legal corruption vs. accountability:
⇒ Clear negative correlations should arise for high
income p/c or equality first quartiles
⇒ Over all countries, we should see higher correlations
(in the sense of less negative)
An Econometric Model for
Testing
LK = i1 + aDGDP0 + bDEQUAL + cDACC + dDACC * DGDP0 + eDACC * DEQUAL (1)
IK = i4 + fDGDP0 + gDEQUAL + hDACC + iDACC * DGDP0 + jDACC * DEQUAL (2)
INSURR = i7 + kDGDP0 + lDEQUAL + mDACC + nDACC * DGDP0 + oDACC * DEQUAL (3)
• Restrictions on coefficients:
Eq(1): ⎡ a ⎤ ⎡+ ⎤ Eq(2): ⎡ f ⎤ ⎡−⎤ Eq(3): ⎡ k ⎤ ⎡− ⎤
⎢ b ⎥ ⎢+ ⎥ ⎢ g ⎥ ⎢− ⎥ ⎢ l ⎥ ⎢− ⎥
⎢ ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ ⎥
⎢ c ⎥ = ⎢0⎥ ⎢ h ⎥ = ⎢0⎥ ⎢m⎥ = ⎢ 0 ⎥
⎢ ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ ⎥
⎢d ⎥ ⎢− ⎥ ⎢ i ⎥ ⎢0⎥ ⎢ n ⎥ ⎢0⎥
⎢ e ⎥ ⎢−⎥
⎣ ⎦ ⎣ ⎦ ⎢ j ⎥ ⎢0⎥
⎣ ⎦ ⎣ ⎦ ⎢ o ⎥ ⎢0⎥
⎣ ⎦ ⎣ ⎦
• Eq(1) - Legal corruption proxies: we take diffs to
illegal corruption and lack of rule of law
Equation (1) - Legal Corruption (using differences)
Dependent Variable -------> Legal Corruption
Accountability Freedom of Press EOS Q5.06
Choice of Equality Gini
Empirical EOS Legal
Legal Corruption EOS Q4.12 EOS Q4.14 EOS Q5.14D
Measures Corruption
(Difference) EOS EOS EOS EOS
Adjustment Q5.12E Q5.12G
KKMRL KKMCC
Q5.12G
KKMRL KKMCC
Q5.12G
KKMRL KKMCC
coef. 0.55** 0.88** 0.13* 0.11 0.73 0.13 0.12 1.05** 0.10 0.09
DGDP
std. err. 0.24 0.42 0.07 0.08 0.48 0.10 0.10 0.41 0.09 0.10
coef. 0.34* 1.05*** 0.31*** 0.30*** 1.01** 0.26*** 0.25*** 0.67** 0.16** 0.15*
DEQUAL
std. err. 0.20 0.34 0.06 0.06 0.39 0.08 0.08 0.33 0.07 0.08
Explanatory coef. 0.22 0.04 0.03 0.05 0.26 0.08 0.10 0.17 0.03 0.04
DACC
Variables std. err. 0.22 0.39 0.07 0.07 0.44 0.09 0.10 0.38 0.08 0.09
coef.
DGDPDACC
std. err.
coef. -0.50 -1.28** -0.35*** -0.36*** -1.30* -0.33** -0.34** -0.63 -0.14 -0.15
DEQUALDACC
std. err. 0.33 0.58 0.10 0.11 0.66 0.13 0.14 0.56 0.13 0.13
Number of Observations 85 85 85 85 85 85 85 85 85 85
R2Adjusted 0.15 0.13 0.25 0.21 0.09 0.12 0.10 0.18 0.05 0.03
Fit (Testable Implications) +- OK EXACT EXACT EXACT EXACT EXACT EXACT +- OK +- OK +- OK
Notes: All regressions have GDP as GDP pc 1984; dummies take value 1 for the first quartile of the corresponding variable. *, **, ***, correspond to the
levesl of statistical significance 10%, 5%, and 1%, respectively.
Equation (2) - Illegal Corruption
Dependent Variable -------> Illegal Corruption
Accountability Freedom of Press EOS Q5.06
Choice of
Equality Gini
Empirical
Measures EOS EOS EOS
Illegal Corruption EOS Q4.02 EOS Q4.13
Q5.11 Q5.12E Q5.12F
coef. -1.22*** -1.24** -1.26*** -1.00* -1.44*** -1.23*** -0.96***
DGDP
std. err. 0.34 0.51 0.39 0.59 0.34 0.32 0.28
coef. 0.42 0.14 0.25 -0.02 0.05 0.08 -0.04
DEQUAL
std. err. 0.27 0.23 0.32 0.27 0.27 0.26 0.23
Explanatory coef. -0.52 -0.60 -0.33 -0.26 -0.97*** -0.69** -0.69**
DACC
Variables std. err. 0.31 0.37 0.37 0.43 0.31 0.29 0.26
coef. -0.27 -0.75
DGDPDACC
std. err. 0.66 0.76
coef. -0.86* -0.91* -0.33 -0.40 -0.31
DEQUALDACC
std. err. 0.46 0.54 0.46 0.44 0.39
Number of Observations 85 85 85 85 85 85 85
R2Adjusted 0.50 0.48 0.40 0.39 0.61 0.53 0.52
Fit (Testable Implications) EXACT EXACT - OK
Notes: All regressions have GDP as GDP pc 1984; dummies take value 1 for the first quartile of the corresponding variable. *,
**, ***, correspond to the levesl of statistical significance 10%, 5%, and 1%, respectively.
Equation (3) - Insurrections
Dep. Variable --> Insurrections
Accountability Freedom of Press EOS Q5.06
Choice of
Equality Gini
Empirical
Measures EOS EIU EIU EIU
Insurrections iJET EIU 3003
Q5.09 3001 3002 3005
coef. -1.30*** -1.46*** -0.98*** -0.49 -0.96* -0.73* -0.67 -0.80*
DGDP
std. err. 0.44 0.38 0.26 0.67 0.58 0.40 0.59 0.44
coef. 0.11* -0.33* -0.39* -0.94*** -0.73*** -0.90*** -0.79*** -0.14
DEQUAL
std. err. 0.35 0.17 0.21 0.30 0.26 0.32 0.27 0.35
Explan. coef. -0.23 -0.40 -0.17 -0.80 -0.87** -0.67* -0.61 -0.77*
DACC
Variables std. err. 0.40 0.27 0.24 0.49 0.42 0.37 0.43 0.40
coef. 0.92* 0.67 0.72 0.02
DGDPDACC
std. err. 0.48 0.86 0.74 0.77
coef. -0.56 0.31 0.34 -0.44
DEQUALDACC
std. err. 0.60 0.35 0.56 0.61
Nr. of Observations 85 85 85 80 80 80 80 80
R2Adjusted 0.30 0.37 0.35 0.16 0.28 0.31 0.31 0.31
Fit (Test. Imp.) EXACT - OK EXACT EXACT - OK - OK EXACT - OK
Notes: All regressions have GDP as GDP pc 1984; dummies take value 1 for the first quartile of the corresponding variable.
*, **, ***, correspond to the levesl of statistical significance 10%, 5%, and 1%, respectively.
Main Findings
• Some accounting:
– Eq(1): 14 EXACT, 51 OK (exc. EXACT), out of 100
– Eq(2): 17 EXACT, 19 OK, out of 50
– Eq(3): 22 EXACT, 38 OK, out of 60
• Eq(1):
• Favoritism in Procurement and Legal Political Contributions;
Gini; Freedom of Press
• Eq(2):
• All except Diversion of Public Funds Due to Corruption; DPI
Fractionalization of Government
• Eq(3):
• EOS Common Crime; DPI Fractionalization of Gov
Concluding Remarks
• Simple model to explain basic pattern of legal/illegal
corruption
– General notion of corruption, founded at the micro level
– Income/equality determine basic political threat
– Accountability determines ability to undermine collective action by
population
• General validity of the model from testing with wide
range of empirical measures for the relevant
concepts
– Newly available legal corruption measures
• Possible Policy Messages:
• “It takes two to tango”
• We may be forgetting that some high income (namely G7)
countries have high (legal) corruption
• Accountability: may be key in determining development