There are not many game-theoretic models for the study of African corruption.
So I have been wondering whether, as part of an early effort of figuring out some foundations, corruption in these parts could be examined through the combined lens of: prospect theory and base error neglect; and whether that might yield any insight for something sorely needed: a more rigorous economic model for studying the phenomenon in these parts.
This tool is a useful alternative to the usual method of assuming that decision makers evaluate the gains and losses that may result from their decisions with equal scales.
Assume that the main decision makers whose choices of action and inaction might impact on the prosecution and deterrence of corruption are concerned about the following potential outcomes:
1. A prosecution should bolster the image of the government as competent and effective.
2. A prosecution should deter other party officials from corrupt acts and reduce the risks of embarrassment to the government.
3. A prosecution should endear the ruling party to swing voters who are responsive to signs of good government.
4. A failure to prosecute shall add a powerful negative campaign message to the arsenal of the Opposition.
5. A failure to prosecute shall embolden a wide range of lower level officials to behave in ways likely to embarrass the government.
6. A failure to prosecute shall alienate important foreign donors with the capability to influence global elite opinion.
In a classical ‘expected utility model’, one may perceive a symmetry across the potential outcomes and may conclude that the gains of actions outweigh the costs of action such as:
1. Alienation of the party factions of the prosecuted official.
2. A signal to lower level officials to acquire ‘insurance’ against prosecution for past conduct etc.
Symmetric cost-benefit may thus be perceived as the primary tool for weighing action in this classical model.
In a ‘prospect-theoretic model’ however this approach is considered suspect. The entire situation could be framed in the alternative, for instance, by first recognising that ‘prosecution’ could have two results: success and failure, and that the costs of failure might far outweigh the costs of inaction. They may include:
1. Accusations of collusion with the defence or of general incompetence, which may alienate swing voters.
2. Elevated impunity due to perceptions of weak capacity.
3. Perceptions of unfair persecution of the acquitted officials among party members etc.
This is a very crude treatment, but the essential point is that ‘losses aversion’ can be a stronger motivator than ultimate reward, and that decision-makers may amplify the impact of losses and downplay the reward of gains.
*Base Error Neglect*
Combining the ‘mental accounting’ results of the above method with base error analysis can be jarring but also interesting.
The crude and basic definition of this base error neglect idea is the tendency of observers served with two sets of facts to ignore the first set which constitutes the primary evidence base and rather to focus on the newer stream of information from the second set.
Consider the following stylised facts:
1. Of the more than ten thousand political appointees Ghana has employed in the 4th Republic, less than one hundred have actually been convicted of corruption by a Court of law or indicted by a commission of inquiry or government whitepaper.
2. The recent auditor general’s report into public agencies did not clear any agency of irregularities, and none of the Parliamentary reports on the irregularities have been acted upon.
3. Of the x<100 convictions and indictments for corruption, only about a fifth have been reversed by the courts.
4. The likely conviction rate for corruption in Ghana by a government with the political will is likely therefore to be in the region of 80%
Is the fallacy very obvious to you? It probably is, but that is because this framing is especially weak. Usually, if properly framed, a good understanding of Bayesian techniques is needed to tease out the fallacy, which is that by the 4th point, most people would have disregarded the baseline evidence stack in statement 1, which heavily constrains the other facts in the subsequent statements.
Nevertheless, the effect of base error fallacy in this example is a heightened expectation of a successful conviction thus aggressively raising the ‘fear of losing’ risk for the decision-maker contemplating whether or not to prosecute suspected corruption.
In this formulation of the model for analysing decision making factors associated with corruption prosecution, it can be shown, albeit very heuristically, that base error neglect reinforces prospect theory in making decisions to prosecute corruption very hard for political bosses.