Rent(s) Asunder: Rent(s) Asunder: Sectoral Rent Extraction & Sectoral Rent Extraction & Bribery by Multi-National Bribery by Multi-National Corporations Corporations Dimitar Georguiev (UCSD) Edmund Malesky (UCSD) Nate Jensen (Washington University)
Jan 14, 2016
Rent(s) Asunder:Rent(s) Asunder:Sectoral Rent Extraction & Bribery Sectoral Rent Extraction & Bribery
by Multi-National Corporationsby Multi-National Corporations
Dimitar Georguiev (UCSD)Edmund Malesky (UCSD)
Nate Jensen (Washington University)
MotivationMotivation• Ngo Vinh Long (2010): “Huge concentrations of
money, especially from foreign sources, have been at the roots of many arbitrary decisions of the Vietnamese state.”
• Vu Quang Viet (2010): “This has helped make Vietnam more dynamic, capable of attracting more foreign direct investment (FDI), opening up the economy outwardly and generating much more wealth, and thus offering more spoils for abuse and bribery which have reached an unprecedented scale under the current regime.”
The PuzzleThe Puzzle• The prevailing theory in the literature
contradicts the view of VN experts: economic openness should reduce corruption through two distinct but not mutually exclusive channels. Mechanism – competition drives down
monopoly rents and therefore cost of bribe schedule (Sandholtz and Gray 2003, Rose-Ackerman 1978, 1999)
Mechanism – learning/diffusion of good practices, lobbying, impact of FCPA ((Sandholtz and Koetzle 2000, 40, Gerring and Thacker 2005, Kwok and Tadesse 2006)
Alternative Views in the LiteratureAlternative Views in the Literature
1. Corruption Deters Investment Directly Indirectly through Entry Strategy
2. Openness increases corruption Ades and di Tella also claim that corruption could worsen in high rent
sectors. Foreign firms use to solve a strategic disadvantage
3.Conditional relationship between openness and corruption
Quadratic effects, whereby corruption increases initially with FDI entry, but decreases as international pressure is brought to bear.
Impact of FDI on corruption depends on quality of local institutions, increasing corruption in poorly governed areas
Measurement Problems in Corruption Measurement Problems in Corruption Data Obscure FindingsData Obscure Findings
4 Biases1.Perception biases of respondents in how they respond to Likert scales. 2.Anchoring bias in the way corruption and bribes are understood. 3.Biases caused by the respondent’s confidence that the information they reveal will not be used to punish them.4.Question wording which invites respondents to answer about others’ experience with corruption and not their own, leading to exaggeration of the true bribe schedule.
Implications of Error for InferenceImplications of Error for Inference
• Perceptions of corruption tend to be highly correlated with economic development and institutional quality, but actual corruption is not (Triesman 2007, Olken 2008).
• Strong possibility of “systematic, variable measurement error.” Errors in the measurement of corruption are correlated with factors that are attractive to foreign investors. Some are measurable, many are not.
• Tell-tale signs. Institutions and wealth explain 90% of the variation in corruptions, but have no correlation with actual activity.
• Aggregate data predicts less reductions in corruption, but firm-level data shows foreign investors more likely to admit complicity.
Secondary Issue with Corruption DataSecondary Issue with Corruption Data• Corruption manifests itself in a number of ways
and impacts individuals differently. o Petty corruption – small bribes during regulatory procedures,
inspections, or policing.o Grand corruption – bribes during procurement, kick-backs on
government contracts, abuse of government resources.
• General questions ask individuals to discuss issues of corruption that they may have never experienced.
• General questions are not useful for understanding the implications of particular policy choices.
Our Three Theoretical RefinementsOur Three Theoretical Refinements1. Allow for possibility that foreign firms are complicit in
bribing.2. Focus on two particular form of economic – openness:
• FDI vs Domestic Firms• The removal of administrative and regulatory restrictions
3. Acknowledge that not all gatekeepers have the same level of extractive power, where restrictions create greater opportunities for rents, we should expect higher level of corruption.
HypothesesHypothesesH1: The propensity of foreign firms to bribe increases with economic rents.
H2: The removal of investment restrictions will:a) reduce economic rents in a sector.b) reduce the willingness of foreign firms to bribe.
Toward a Precise Method of Toward a Precise Method of Measuring CorruptionMeasuring Corruption
• Unmatched Count Technique (UCT, LIST) Question (Raghavarao and Federer, 1979).
• On one form, respondents are asked how many innocuous but relatively infrequent activities they participated in over a clearly defined time period and location.
• The second group of respondents receives the same list of innocuous items as well as an additional sensitive item.
• Respondent only must recall the number of activities they engaged in and must never admit to sensitive behavior.
• Aggregation protects privacy and shields complicity.• The difference in means between the two forms provides the
proportion of respondents participating in the sensitive action. • The technique has been shown to provide the most accurate
and confidential assessments of sensitive activity in repeated testing (Couts and Jann 2009).
• Estimate (Probability of Bribery) = Mean (Form A) – Mean (Form B)
Laboratory Results For Sensitive Laboratory Results For Sensitive Question TechniquesQuestion Techniques
Our UCT, Designed to Measure Our UCT, Designed to Measure Corruption During Business EntryCorruption During Business Entry
Sample and MethodSample and Method• 7,300 Domestic, Private Enterprises AND 1,155
Foreign Direct Enterprises in Vietnam.• Survey from Summer of 2010, as part of Vietnam
Provincial Competitiveness Index Survey.• LIST questions embedded into survey.• 35% response (high for business surveys), but
problematic – detailed comparison to underlying population to make sure not biased (Appendix 2)
• 30% item non-response to LIST question. We use multiple-imputation, but make sure results are biased by procedure.
Table 3: Balance Test of Key Indicators of Control and Treatment Group
Mean Std. Dev. Mean Std. Dev. P-Value T-Score Labor Size Category (1 - 8) 1.029 0.169 1.030 0.169 0.937 -0.045Capital_Size Category (1 - 8) 2.327 1.202 2.358 1.177 0.261 -1.124Manufacturing (0 - 1) 0.208 0.406 0.222 0.415 0.172 -1.367Construction (0 - 1) 0.235 0.424 0.252 0.434 0.103 -1.632Services and Trade (0 - 1) 0.612 0.487 0.597 0.491 0.193 1.303Agriculture (0 - 1) 0.063 0.244 0.069 0.253 0.379 -0.879Mining (0 - 1) 0.028 0.164 0.023 0.151 0.263 1.119SOE History (0 - 1) 0.063 0.243 0.070 0.255 0.247 -1.158National Cities (0 - 1) 0.150 0.357 0.159 0.365 0.336 -0.962Industrial Zone (0 - 1) 0.096 0.295 0.089 0.284 0.278 1.0851StopShop (0 - 1) 0.101 0.094 0.301 0.292 0.336 0.962Telephones Per Capita 0.280 0.357 0.275 0.355 0.607 0.515
Mean Std. Dev. Mean Std. Dev. P-Value T-Score Labor Size Category (1 - 8) 3.909 2.248 3.681 2.119 0.119 1.561Capital_Size Category (1 - 8) 5.343 1.411 5.316 1.494 0.809 0.242Manufacturing (0 - 1) 0.686 0.465 0.729 0.445 0.152 -1.433Construction (0 - 1) 0.064 0.246 0.037 0.190 0.055 1.917Services and Trade (0 - 1) 0.213 0.410 0.209 0.407 0.888 0.141Agriculture (0 - 1) 0.074 0.263 0.051 0.220 0.133 1.504Mining (0 - 1) 0.020 0.141 0.014 0.120 0.497 0.679SOE History (0 - 1) 0.014 0.116 0.007 0.085 0.325 0.985National Cities (0 - 1) 0.338 0.474 0.385 0.487 0.148 -1.448Industrial Zone (0 - 1) 0.490 0.501 0.481 0.500 0.271 0.7861StopShop (0 - 1) 0.358 0.480 0.377 0.485 0.568 -0.571Telephones Per Capita 0.218 0.210 0.262 0.317 0.027 -2.218
Domestic Private Entrprises (N = 7138)
Covariates
Foreign Invested Enterprises (N = 1124)
Covariates
Difference in Means
Difference in Means
Control (N = 3512) Treatment (N = 3626)
Control (N = 296) Treatment (N = 828)
Propensity to Bribe During Propensity to Bribe During
RegistrationRegistration
By Foreign and Domestic FirmsBy Foreign and Domestic Firms
Ceiling/Floor Effects are Negligible.Ceiling/Floor Effects are Negligible.
Measuring Restrictions on FDI Entry Measuring Restrictions on FDI Entry We take advantage of the fact that Vietnamese investment laws require special approval from the Prime Minister’s Office for certain sectors (Group A).•We code these sectors at the four-digit ISIC level, over time.•Measure is nice, because it is less subject to simultaneity bias.
Excerpt of our coding scheme
Rents are Significantly Higher in Rents are Significantly Higher in Protected SectorsProtected Sectors
Do Restrictions Do Restrictions Generate Rents?Generate Rents?
Multiple Regression using LIST Multiple Regression using LIST
MethodMethod
Conditional model of predicted sensitive activities
Results for H1Results for H1• Confirmation of H1: When all variables are held
constant at their mean, foreign firms attempting to enter restricted sectors have a 31% predicted probability of engaging in bribery, 14% higher than their domestic competitors in restricted sectors and 19% more likely to bribe than foreign firms in non-restricted sectors.
• Robust to: Year FE, Province FE, Truncation, Glynn Piecewise Regression Non-Imputation.
• Holds for Type of investor: OECD v. Non-OECD; 100% FDI v. JV; Industrial Zone v. Externalo OECD investors in restricted sectors are less likely to bribe. o Firms in Izs (in non-restricted sector are less likely to bribe).
Figure 4: Differences in Bribe Frequency
(By Foreign Firms, Restricted Sectors, and WTO Entry)
19.0%
35.7%28.5%
-7.0%
-40
-20
020
40
60
All Years Pre-WTO Pre-WTO (3y) Post-WTO
Restricted vs. Non-Restricted Foreign Firms
14.0%22.9% 26.2%
-3.8%
-40
-20
020
40
60
All Years Pre-WTO Pre-WTO (3y) Post-WTO
Foreign vs. Domestic Firms in Restricted Sectors
Ma
rgin
al D
iffe
ren
ce
in
Brib
e F
req
ue
ncy
Results for H2Results for H2• Mixed conclusions for Hypothesis 2. While opening
up to foreign competition does reduce the propensity for MNCs to bribe, it compels their domestic competitors within the same sectors to increase their level of bribery.o In years where a product faced restrictions, MNCs were
more likely to pay bribes. Foreign firms facing Group A restrictions pay 12.5% more than firms in the same sector that did not face such restrictions.
o Reducing the protections for domestic firms had the opposite effect. The rate of bribery among domestic firms, now suddenly subject to increased foreign competition, actually increased by an average of 14.7%.
o Robust to truncation, piecewise, and non-imputation.o Helps explain why the Vietnamese analysts, who
primarily cite domestic incidents, have observed an uptick.
ConclusionsConclusions• More refined analysis of globalization and
corruption.o Openness can have differential effects on
corruption, even within the same country and the same institutional setting.
o Foreign firms are not necessarily victims, they appear willing to bribe if the rents justify the activity in both entry and procurement.
o In lower-rent industries, there is no positive or negative influence of “foreigness.”
o Removing restrictions and subjecting a country to globalization can impact corruption indirectly, by enhancing corruption in the domestic sectors.