Much of the focus in combating corruption in government bureaucracies focuses on creating the right incentives for public servants after they’ve assumed their positions.  The goal is usually to create a system of rewards and punishments – and perhaps also a professional culture – that incentivizes honest behavior and deters wrongdoing.  Creating those incentives is obviously crucial, but it’s also important not to neglect the selection process – choosing who gets to become a civil servant or public official in the first place.  After all, it’s probably a lot easier to help a basically honest person to resist temptation than it is to discourage a venal opportunist from abusing her position.  Moreover, selecting the wrong people into public service can create a vicious cycle: a government agency with a reputation for corruption will tend to attract individuals who more interested in abusing their positions, while an agency with a reputation for probity will be more likely to attract individuals interested in serving the public good.

That’s the theory, anyway.  Gathering empirical evidence on the honesty (or lack thereof) of incoming civil servants is challenging, and so is assessing the impact of personal integrity on actual behavior on the job.  But some terrific experimental research in India, conducted by my Harvard colleague Rema Hanna, along with two different co-authors (Shing-Yi Wang and Iqbal Dhaliwal), suggests (1) that in India, people with a greater propensity to cheat also have a stronger interest in government jobs, and (2) civil servants with a greater propensity to cheat are also more likely to shirk on the job.  The paper is well worth reading.  Though the results should be interpreted cautiously, they raise a number of important issues and questions for anticorruption strategy.

I won’t try to summarize the details of the experiments here, but let me give the basic gist.  In the first set of experiments, Hanna and Wang had Indian college students play a game in which they privately rolled a die 42 times and were told to record the result of each roll; the students were then paid an amount proportional to the total sum.  Students had an incentive to cheat (by reporting higher numbers), and though the researchers could not directly observe cheating, they could infer which students were statistically likely to have cheated based on the improbability of their rolls (i.e., too many high numbers).  The researchers also asked the students a battery of survey questions, including future career ambitions.

In a separate experiment (also reported in Hanna & Wang’s paper), Hanna and Dhaliwal had government nurses in India play a variant of the same game, and later conducted random spot-checks of public healthcare clinics to see which nurses were absent on days when then claimed to be present (and collected a paycheck).

The results: (1) the dishonest students (those who are more likely to have cheated at the dice game) were more likely to express a preference for a government job; (2) the dishonest nurses (again, the likely cheaters on the dice game) were more likely to be absent – collecting a government paycheck without doing their jobs.

One always needs to be careful about generalizing the results from these experiments to make more sweeping conclusions. Among other things, one needs to be careful not to denigrate all Indian civil servants — many of whom are people of the utmost integrity who do their best to serve their country under trying conditions — based on a statistical average derived from an experiment. Nonetheless, the results are at least consistent with the hypothesis that India’s serious public corruption problem is due in part to the fact that government jobs are perceived by some as providing opportunities for corruption (whether in the form of bribe-taking or shirking or something else), and that as a result some honest would-be recruits are deterred from seeking government jobs, while some less scrupulous candidates seek out those jobs.

My suspicion – and I would encourage Rema or others to pursue this – is that if one were to do similar research in other countries, one would find very different patterns that might well be highly correlated with other measures of corruption in those countries.  (I suspect that in the United States, for example, all the cheaters would want to go into the financial sector, not public service.)  It might also be interesting to perform similar research to see if there are differences across public agencies within the same country: what sort of people, for example, want to become police officers, or customs inspectors, or forest managers?

Let’s suppose, as I think is plausible, that further research will confirm that personnel selection – picking honest people – is a crucial contributor to the success of anticorruption efforts in public service.  What follows from this?  That’s a bigger topic for another day, but let me throw out a couple of preliminary ideas.

• First, some public agencies might consider more seriously whether there are ways of screening candidates not just for competence and cognitive ability, but for ethics.  This is hard to do, because once people know they are being tested for integrity, their behavior is likely to change.  But it might not be impossible, and some anticorruption authorities – such as the KPK in Indonesia – already make efforts to screen candidates in this way, using a battery of computerized integrity tests developed by an outside human resources consultant.
• Second, more aggressive attempts to weed out – and not just to punish – officials who abuse their positions might be in order.  This is a tricky subject, because there’s a countervailing consideration in maintaining legitimate employment protections for civil servants, and there’s always the concern about politicized terminations.  But where corruption is endemic, the balance might tilt more toward getting dishonest people – even those whose dishonest behavior seems relatively mild – out of the civil service as quickly as possible.