Does corruption create economic incentives that improve efficiency? Sometimes it can. For example, bribing your professor to grade exams and homeworks quickly rather than waiting two or three weeks might improve efficiency, but bribing them to give you a higher grade you don't deserve would not. In this article examining corruption in the equivalent of the DMV in India, the negative effect - bribing for a higher score on the driving test - outweighs the positive effect of reducing processing time:
Driving in New Delhi: Don't complain about standing in line at the DMV., by Joel Waldfogel, Slate: At least in principle, some kinds of government corruption are not so bad because they promote efficiency in how regulations are administered. ... But a new study of driver's license examinations in New Delhi, India, confirms what most international policy wonks have long said: The benefits of corruption are not worth the costs.
The Department of Motor Vehicles, here and in many foreign countries, is a place of long lines, sour bureaucrats (think Patty and Selma Bouvier, Marge Simpson's chain-smoking spinster sisters), and bleak interior decorating. ... Since access to government clerks is normally allocated on a first-come, first-served basis, people pay with their time rather than their money. This is inefficient: Suppose you're in a big hurry and would be willing to pay a lot to avoid waiting, while I don't mind waiting. Then you could go ahead of me, making you a lot better off and me only a little worse off, which reduces our collective frustration. One way to achieve this efficiency would be to charge a higher price for expedited service. Yet, an expedited government service option typically does not exist. So, in some countries, the offer of a bribe in exchange for quicker processing is a common form of corruption—reducing the social cost of waiting in line.
But DMVs exist for a purpose... They're supposed to reduce unsafe driving. ... Do the clerks in fact do so? To study the process of getting a driver's license in New Delhi, the authors ... recruited 822 Indians who wanted a driver's license, randomly assigning them to three groups. ...
One of the groups in the Indian study was offered a cash bonus for getting a license within 30 days. These subjects had an incentive do whatever was necessary (offer bribes) to get a license quickly. ... A second group was given driving lessons. If the licensing process accurately screens out unprepared ... drivers, then these applicants should be more likely to succeed in getting a license. Both of these applicant groups were compared to a third control group who received neither lessons nor a speedy completion bonus. ... Eventually, they gave their subjects a follow-up written exam to gauge their driving skills.
So, what happened? More than a third (37 percent) of the control group got a license, compared to 45 percent of the subjects who took driving lessons and 65 percent of the people who got paid for getting a license quickly. Subjects in the cash bonus group were most likely to hire "agents" to help them navigate the bureaucracy, spending an average of 1,280 rupees to get a license, compared with 560 rupees for those without an agent. And applicants using agents got their licenses 15 percent faster, making an average of a quarter fewer trips to the Indian DMV... They spent about three hours of their own time, as opposed to five hours for those who did not hire an agent.
The agents saved applicants time by, for example, standing in line for them. But the extra cost of using an agent dwarfs the benefit of saving two hours for a typical Indian, who makes 40 rupees per hour, raising the suspicion that the agent's fee purchased something other than time. And indeed, 88 percent of the applicants who hired an agent did not have to take the driver's exam before getting a license, while almost all of the other applicants did.
Perhaps the aspiring drivers who hired agents were better-qualified...? Dream on. The drivers who used an agent had much lower scores on the follow-up drivers exam given by the researchers.
This study confirms the view of the World Bank, which "has identified corruption as among the greatest obstacles to economic and social development." Payoffs at the Indian DMV may save some qualified drivers some time. But it has the bad direct effect of allowing unsafe drivers on the road. ... The lesson applies beyond the DMV. If bribing a procurement officer works better than building a solid airplane or bridge, why bother with safety checks? Dealing with by-the-book Pattys and Selmas is pretty unpleasant. But a world without them is much worse.
See Is Corruption in Iraq Economically Beneficial? for more on the economics of corruption. Also see When is Corruption Good from the Private Sector Development blog which includes links to topics such as how to measure the costs of corruption. For even more, see Becker and Posner: Economics of Corruption-Posner, Becker's comments, Posner's response, and Becker's response. This quote from The New Yorker article highlighted in the first link is worth repeating:
[E]ven if corruption can be a useful means of bypassing inefficiencies in the short term, in the long term it tends to create inefficiencies of its own. Bribing, it turns out, doesn’t always speed things up ... a vast study of twenty-four hundred companies in fifty-eight countries ... found that the more a company had to bribe, the more time it spent tied up in negotiations with bureaucrats. Graft also encourages government officials to keep complicated procedures in place... So corruption isn’t just a product of bad institutions and policies; it also helps cause them. Almost every study done in the past ten years has found that, on the whole, corrupt countries grow more slowly and have a much harder time attracting foreign investment.
Update: Truck and Barter has more.