If you can bribe someone to avoid punishment, you probably will.

In many developing countries, the struggle for economic growth is set back by rampant corruption. According to figures in a new study of the issue, people in urban areas of Kenya typically pay bribes 16 times a month. That's a drain on the economy, and it adds a layer of complexity between citizens and essential government services.

While a variety of policy approaches have attempted to limit corruption, it's difficult to track their effectiveness. Now, an international team of researchers has developed a game-theory approach to teasing out the factors that contribute to corruption. Their results show that under the wrong circumstances, a common method of limiting corruption—government transparency—can actually make matters worse.

Cooperation vs. the freeloaders

The foundation of this work is what's called a "public goods game," which measures people's willingness to cooperate. In this game, everyone starts with a pool of cash and is given the opportunity to contribute to a common, public pool. The resulting pool is then multiplied, and its contents are distributed evenly among the players. The group as a whole works out best if everyone cooperates, contributing the maximum amount to the pool. But individuals do best if they freeload: contribute nothing, then take their share of the public pool.

Institutional punishment—essentially a government—can be added to the public goods game. In each round, a "leader" gets to see everyone's contributions and can levy fines against freeloaders. This tends to increase cooperation, but it can also degenerate into rounds of retribution if people take turns as the leader. (Some societies also apparently have issues with random punishment of people who are cooperating.) Past studies have shown that people are often willing to pay a personal cost in order to punish freeloaders.

To create a bribery game, the research team started with a variation on the institutional punishment game, one where everybody was expected to pay a tax to the public pool, and a leader was there to punish anyone who didn't. In the bribery game, participants were allowed to pay a bribe to the leader prior to any decisions on punishment. Presumably, the leader would view the bribe as a suggestion that punishment should be skipped.

To study different effects, the researchers varied the amount of money that could be levied as a fine, creating weak and strong leaders. By altering the starting money and the amount the pool was multiplied by, they could alter the economic growth potential of the group as a whole.

Corruption

To see how this worked out, the researchers recruited a bunch of people with a very low reputation for corruption: Canadians (some of the researchers were based at the University of British Columbia). But their Canadianness only got them so far. "Not surprisingly, when corruption could enter, it did," the authors write, "and cooperation deteriorated." When bribery was an option, the average contribution to the public pool dropped by a quarter.

The subjects did include people who had immigrated to Canada, including some from countries with a history of corruption, and these were somewhat more likely than native Canadians to engage in corruption. But the children of immigrants were even less likely to do so than people with a longer family history in Canada.

Among the controls, strong leadership (meaning a large ability to punish) significantly increased the average contribution to the public pool. But when bribery was introduced, having a strong leader actually lowered contributions, as more people were spending that money to bribe the leader—who quickly became relatively wealthy. Partial transparency, in which participants could see the leader's contribution to the public pool, made very little difference. Only full transparency, in which participants could see every transaction the leader engaged in (including bribes), restored contributions to something approaching the control situation.

When the economy of the game is strong (larger starting pots and/or higher multipliers), contributions go up in the controls, regardless of whether the leader is weak or strong. In both leadership cases, however, bribery still caused the typical contributions to drop. The percentage drop, however, was smaller than it was in the poor economy/strong leader situation.

Notably, partial transparency never made anything better. In fact, in a weak leader/weak economy group, it made matters worse. Only full transparency helped, in some cases returning contribution levels to those seen in the bribery-free controls.

If these trends also apply in the real world, there's a couple of results here that run counter to our expectations. One is that strong leadership—in the form of a strong central government that's able to mete out real punishments—doesn't stop corruption. In fact, it makes matters worse when bribery's an option and simply leads to a wealthy leader. The second issue is that government transparency doesn't seem to help much unless it includes transparency about all the bribery. Since that's overwhelmingly unlikely to happen, it's not clear whether government transparency does much good here.

What does seem to matter is a flourishing economy, with the players having a sense that the game is structured so they get something out of playing. Here, corruption still takes a toll, but it's a relatively small percentage of the total economic activity.