Lobbying and Corruption Dr James Tremewan - - PowerPoint PPT Presentation

Lobbying and Corruption

Dr James Tremewan (james.tremewan@univie.ac.at) Corruption Laboratory Experiments

Introduction

Design of Laboratory Experiments

• Today we will discuss three issues related to laboratory

experiments on corruption:

• Whether to repeat interactions, and if so, whether the groups

stay the same or change.

• Framing (placing the strategic decision in a particular context

through labeling players or actions, e.g “Politician” instead of “Player A”).

• Do laboratory experiments really evoke attitudes towards

corruption?

2/31

Introduction

Repeated Games

• In many experiments subjects play a game only once

(”one-shot”).

• Much experimental theory is about a one-shot interaction.
• Abstracts from complications with repeated interactions.
• However, we may want to have subjects play more than once:
• Learning:
• To allow subjects to become familiar with the environment

(understanding the game, or learning about the behaviour of

• thers). May want to compare experienced and novice subjects.
• Studying the learning process itself.
• To study issues like reputation and reciprocity.
• Lobbying and corruption can occur in both one-shot and

repeated environments:

• One-shot: city competing to hold the Olympic games,

application for electoral roll.

• Repeated: health inspections of restaurants, defence contracts.

3/31

Introduction

Stranger or Partner Matching?

• There are two basic alternatives when subjects play repeatedly:
• Stranger matching: each subject plays with a different person

each time.

• Partner matching: subjects play repeatedly with the same

person.

• Stranger matching:
• Often regarded as multiple one-shot games, so used for studying

pure learning/experience effects.

• Subjects may be re-matched with people they have played

before, but anonymity means they will not know for sure. ”Absolute/Perfect stranger matching”: subjects only play each

• ther once (and know this).
• Some evidence that subjects may not view random stranger

matching as one-shot (e.g. Botelho et al, 2005).

• Partner matching:
• Studying role of reputation or reciprocity.

4/31

Staff rotation as an anti-corruption policy

Staff rotation as an anti-corruption policy

• “Staff rotation as an anti-corruption policy: an experimental

study”, Klaus Abbink (2004), compares stranger and partner matching treatments of a corruption game to examine the effectiveness of staff rotation in reducing corrupt behaviour.

• Staff rotation has been used to reduce corruption (e.g. in public

procurement for the German federal government).

• Advantages:
• Public officials may be less influenced by gifts if they are unlikely

to interact with the giver again (no chance of further gifts).

• Potential bribers anticipate this, so are less likely to bribe.
• Potential bribers may be more worried about being reported if

they have not had experience with a particular official before.

• Problems:
• Reciprocity may mean that officials are influenced by gifts even

in one-shot intereactions.

• Costs: officials would frequently need to learn new routines, new

cases, new clients, etc. Loss of specific knowledge.

5/31

Staff rotation as an anti-corruption policy

Corruption Game

• Scenario: a firm wants to run an industrial plant which causes

negative consequences to the public (e.g. pollution). A public

• fficial must decide every period whether to give permission.
• Two-player game (framed neutrally in instructions):
• Firm chooses whether or not to offer a transfer (there is a small

cost for making an offer).

• Public official chooses whether or not to accept, and also

whether to refuse or grant permission. The transfer is multiplied by three: small cost to the firm, but big money to public official.

• If permission is granted, all other subjects in the experiment

lose some points (social cost of corruption).

• If corruption takes place there is a small probability that the two

players get caught and lose all money from game.

• Subjects play 30 rounds with either partner or stranger matching.
• Subgame equilibrium: no bribe ever offered (In last round, public
• fficial will always refuse permission, so no reason to offer

bribe.... unravels as with repeated prisoners’ dilemma).

6/31

Staff rotation as an anti-corruption policy 7/31

Staff rotation as an anti-corruption policy 8/31

Staff rotation as an anti-corruption policy 9/31

Staff rotation as an anti-corruption policy

Evolution of permission frequencies

10/31

Staff rotation as an anti-corruption policy 11/31

Staff rotation as an anti-corruption policy

Permission frequencies depending on transfer

12/31

Staff rotation as an anti-corruption policy

Conclusions

• In this experimental environment, staff rotation was effective in

reducing corrupt behaviour:

• Bribes less likely to be offered.
• Bribes of all sizes less likely to influence the official.
• The average bribe was reduced by half.
• Average frequency of inefficient decision reduced by 2/3.

13/31

Framing

Framing

• Most economic experiments are done without context (a

”neutral frame”), e.g. choose option A or B, rather than ”cooperate” or ”defect.” Remember: lab experiments are not about re-creating reality!

• This is done to retain generality, avoid inducing particular

behaviour because of connotations of words, or ”role-playing.” Traditionally economic theory has held the view that arbitrary labels should not matter.

• However, researchers may use framed experiments if they are

interested specifically in the effects of context.

• Also, in some instances, not framing experiments can lead to a

loss of control: subjects may apply their own context to help understand an abstract situation, and if different subjects apply different contexts there will be uncontrolled heterogeneity.

14/31

Framing

Framing

• Sometimes framing matters and sometimes it doesn’t. We will

look in detail at two papers.

• Neutral versus loaded instructions in a bribery experiment,

Abbink and Hennig-Schmidt (2006).

• Finds no effect of framing in corruption experiment.
• The effects of externalities and framing on bribery in a petty

corruption experiment, Barr and Serra (2009).

• Finds a significant effect of framing in a similar environment,

and suggests why the result differs from the previous paper (role-playing).

15/31

Framing

Abbink and Hennig-Schmidt (2006)

• Same game as Abbink (2004).
• Two-player game, framed (neutral):
• ”Firm” (”player 1”) chooses whether or not to offer a ”private

payment” (”transfer”).

• ”Public official” (”player 2”) chooses whether or not to accept,

and also whether to ”grant permission” (”choose Y”) or ”not grant permission” (”chose X”).

• If permission is granted, all other subjects in the experiment

lose some points (social cost of corruption).

• If corruption takes place there is a small probability that the two

players lose all money they could have earnt from the experiment.

• Subjects play 30 rounds with the same partner (long-term

relationship).

16/31

Framing

• Hypothesis: framing in morally negative context of corruption

highlights negative consequences of decisions and should reduce corrupt behaviour.

17/31

Framing 18/31

Framing 19/31

Framing

Results

• Average bribe offer: 2.56 (framed), 2.93 (neutral). Not

significantly different (Mann-Whitney-U test applied to average

• ffer in each pair, 18 in each treatment, one-sided

p-value=0.39.)

• Permissions given: 32.4% (framed), 43.3% (neutral). Not

significantly different (Mann-Whitney-U test applied to average in each pair, 18 in each treatment, one-sided p-value=0.27.)

• Bribes accepted: 21.3% (framed), 12% (neutral). Not

significantly different (Mann-Whitney-U test applied to average in each pair, 18 in each treatment, one-sided p-value=0.34.)

• Authors’ conclusion: no evidence of strong framing effect. Game

design already transmits all essential features of bribery situation.

20/31

Framing

Barr and Serra (2009)

• Similar set-up to Abbink and Hennig-Schmidt.
• Major differences:
• Briber is labeled ”private citizen” instead of ”firm.”
• If bribe is accepted, service is given automatically (only one

choice for public official).

• Negative impact of corruption affects inactive players.
• One-shot game.
• Game designed to represent ”petty” corruption, e.g. bribing

doctor to get treated immediately.

• 2x2 design: framed/neutral and low/high negative externality on

”other members of society.”

• sL = neutral treatment, sH = framed treatment.

21/31

Framing

Corruption Game

• Two active players:
• Private Citizen chooses whether or not to offer a transfer (there

is a small cost for making an offer).

• Public Official chooses whether or not to accept.
• If bribe is accepted, five inactive subjects (called “Other

Members of Society” in framed sessions) in the experiment lose some points (social cost of corruption).

• If corruption takes place there is a small cost to the official,

representing the risk of getting caught (cost is deterministic to reduce uncontrolled heterogeneity due to risk aversion - less “realistic” but better experimental control!!!).

• Subgame perfect equilibrium:
• All Private Citizens offer slightly more that the cost to the

Public Officials from accepting.

• Public Officials accept.

22/31

Framing

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

① ①

PRIVATE CITIZEN PUBLIC OFFICIAL not offer bribe

• ffer bribe

b ∈ {1, ..., 20} not accept bribe accept bribe (35,35,25) (34,35,25) (40-b,30 + b,25 − Nch) Nc ∈ {0, ..., 5}: number of corrupt pairs; h ∈ {1, 4} (low/high externality)

23/31

Framing 24/31

Framing

• Private Citizens:
• Offered no bribe: 40% (framed), 10% (neutral). Significantly

different at 5% level (n=65).

• Average bribe: 8.57 (framed), 7.62 (neutral). Not significantly

different (n=48).

• Public Servants:
• Accepted no bribe: 26% (framed), 10% (neutral). Not

significantly different (n=65).

• Average bribe: 7.34 (framed), 7.74 (neutral). Not significantly

different (n=53).

• Why significant results here, but not in previous paper?
• In previous paper, victims of corruption could have been seen as

business competitors, and may also be expected to do harm, so moral argument less strong.

• Subjects playing as firm may have been ”role-playing” and may

have tried to act as they thought a firm would, whereas subjects are private citizens, so more likely to internalize moral cost of harming others. (Role-playing would occur for public officials in both experiments).

25/31

Culture and Corruption

Corruption and Culture

• Social norms regarding corruption vary from country to country,

e.g. Indian students demonstrating against measures infringing

• n their ”right to cheat” in university exams!

(http://www.bbc.com/news/magazine-29950843)

• Do social norms internalized during upbringing affect propensity

to act corruptly in a new environment? i.e. do people act corruptly because it is necessary in a given environmnent, or is corruption partly a cultural phenomenon that works through individuals’ norms?

• Can laboratory experiments replicate corrupt exchanges closely

enough to learn about true propensities to act corruptly?

• Corruption and culture: An experimental analysis, Abigail Barr

and Danila Serra, Journal of Public Economics (2010):

• Examines these questions using the corruption game from Barr

and Serra (2009) with students from many countries who were studying at Oxford university.

26/31

Culture and Corruption

Subjects

• Country of origin:
• 1/3 of participants were British.
• 2/3 from 33 other countries.
• Level of studies:
• 68% graduate students.
• 32% undergraduate students.

27/31

Culture and Corruption

Corruption levels of subjects’ country of origin based on Transparency International’s CPI (bigger number means more corrupt)

28/31

Culture and Corruption

Subjects’ perceptions of corruption at home

• CPI based on survey of experts, so may not reflect participants’

experience.

• Questionnaire completed after experiment suggests CPI

correlated strongly with experience (but not so much with attitudes):

29/31

Culture and Corruption

Results

30/31

Culture and Corruption

Conclusions

• Propensity to both offer and accept bribe in laboratory

corruption game is correlated with level of corruption in country

• f origin... but only for undergraduate subjects!
• Experiment was repeated with new sample, and results

replicated.

• Explanations for different effects for graduate/undergraduate

subjects:

• Graduate students have spent longer in UK and may have

changed attitudes.

• Selection: the type of student who engages in graduate studies

is less likely to conform to social norms.

• Experience of corruption in upbringing appears to affect

propensity to act corruptly, even in new environments.

• Correlation between behaviour in the lab and data regarding

levels of actual corruption is evidence of external validity of lab experiments.

31/31