C. Ferraz & F. Finan

NBER Working Paper 14937

A Summary


In a Nutshell

Variation in electoral systems is thought to explain differences in corruption practices, so electoral rules that enhance accountability should constrain the ability of politicians to be corrupt. The model developed by Tim Besley states that where politicians seek re-election incentives in a two period setting, they face incentives not be corrupt in the first period in order that they might be re-elected to the second period. The lack of re-election constraint in the second period likewise indicates that the incentive not be corrupt is absent. Thus the paper tests empirically the hypothesis that flows from this i.e. that there will be on average more corruption in the second terms of elected officials relative to first terms even when controlling for other factors that may affect corruption such as experience of the individual politician. Using data from municipal governments drawn from anti-corruption audits undertaken by the Brazilian central government, they find that mayors with re-election incentives are significantly less corrupt that mayors without such incentives. The effect of these incentives varies according to local institutional settings that govern information that is available to voters, the likelihood and consequences of detection, and how competitive the political environment is in that locality. In other words, electoral accountability acts as a powerful mechanism to align the actions of politicians with the preferences of voters.

Besley Model (2006)

  • A political agency model whereby voters decide whether to re-elect an official but are unable to observe his type or actions leading to an information asymmetry which can be exploited by the corrupt such that they pose as good politicians in period 1 in order to maximise rent-seeking possibilities in period 2.
  • Based on state of the world S, which is only revealed to the politician, not the voters, a politician can set policy E such that S=E in which case the voters get a payoff, V. Alternatively, the politician can set policy such that S≠E in which case the voters get a payoff of 0 and the politician receives a payoff of R. After the policy is set voters observe their payoff and re-elect accordingly. All politicians that set policy such that S=E are re-elected.
  • In the second period clean politicians set the policy such that S=E and corrupt politicians set policy such that S≠E and he gets his payoff. Thus the voters are better off re-electing only clean politicians so they want to maximise the likelihood that only non-corrupt politicians are elected in the second period.
  • In the first period clean politician behave in accordance with voter preferences. Corrupt politicians face a trade off: be corrupt today and don’t get re-elected, or be good today and get re-elected to reap the benefits from corruption in the second period. Thus they will pretend to be good politicians if the discounted present value payoff from period 2 is greater than the payoff receivable in period 1.
  • When faced with the possibility of re-election corrupt politicians have the incentive to reduce rent seeking in period 1 and provide more public goods. Rents, then will tend to be higher in the second period, absent this incentive, and this is the testable proposition of this paper.

The Setting and Data

  • Municipal governments in Brazil are responsible for a great amount of public good provision and it is funded by transfers from the central government.
  • There is widespread corruption: phantom firms, uncompetitive bids, over-invoicing etc.
  • The central government does random audits on municipal governments in order to expose such practices. The data are drawn from these reports.
  • Corruption is defined as the resources diverted for corrupt activities expressed as a fraction of the total amount transferred. Other studies use subjective measures, but this measure is somewhat objective (absent bribery of the auditors).


  • Y = corruption level, X = 1st term, + municipal characterises + mayoral characteristics.
  • Political ability and experience could affect both re-election and corruption. They control for this using an RDD specification i.e. by comparing elections where the incumbents just lost, or just won, this controls for unobserved characteristics. They also compare second term mayors with first term mayors that had previously been elected who would be as politically able.
  • To control for experience (i.e. learning-by-doing) they make the same comparison as above, but also compare second term mayors who then went on to national politics. Theory suggests that they will have similar concerns as first term mayors and will act accordingly (i.e. lower corruption) which is inconsistent with the story whereby corruption increases as leaning-by-doing increases.


  • The basic results are that 1st term mayors are associated with a 27% decline in corruption. This is impervious to controls added. The result is also consistent with other specifications such as the RDD specification and also to different definitions of corruption.
  • This indicates that mayors who still face the possibility of re-election engage in less corrupt activities than mayors who have a shorter political horizon.
  • Experience control: first term mayors previously elected do not have similar corruption levels compared with second term mayor, indicating that corruption is not driven by experience (i.e. having well established networks of corruption).
  • Future careers: politicians that go on to future roles nationally etc. have lower 5% lower levels of corruption in the second terms than other second term mayors.

Local Context

  • The presence of media outlets in the locality may reduce the ability of the politician to exploit the information asymmetry that lies at the heart of the agency model, as voters are more likely to observe the type of the mayor if he is exposed by the press. As the likelihood that a corrupt politician is discovered in period 1, corrupt politicians will be less likely to pool with clean politicians so discipline will be reduced (i.e. corruption in period 1 will increase as they try to extract as much rent as possible). Conversely it also means that the such politicians are less likely to survive into period 2 so corruption will decrease in the second period. Thus the difference in the two periods should be smaller with local media. This is confirmed in the results.
  • The cost of corruption should increase if there is a local prosecutor who is willing to punish those caught being corrupt. This reduces the benefits of future rent seeking. This increased corruption in period 1 and reduces it in period 2. i.e. the gap should be smaller. This is found to be the case.
  • If political support is such that the mayor enjoys much legislative support then he can afford to be less disciplined, so again the difference between periods should decrease. This is found to be the case. Municipalities with low competition exhibit no differential effect between first and second term mayors.


  • The findings support a political agency model where mayors with re-election incentives refrain from rent extraction in order to increase their likelihood of re-election.
  • Assuming that absent the re-election incentive 2nd term mayor behaved like first term mayors US$160m would have been misappropriated.
  • Re-election has both discipline and selection effects.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

%d bloggers like this: