Is bribery a regressive tax?

India's public services are riddled with bribery. While most other developing countries also face the same problem, India is in a special league. I have even seen the man in charge of the firewood at the cremation ground demanding a premium on dry wood.

The poor can't afford it. The rich can. And this paper* on bribery is precisely about that.

Jennifer Hunt and Sonia Laszlo of McGill University in Montreal have studied Peru in two surveys of bribery. They say they have tried to "provide a theoretical framework for understanding when an official angles for a bribe, when a client pays, and the payoffs to the client's decision."

When you read it, two similarities with India are obvious. The same sorts of public officials need to be bribed, namely those dealing with services such as municipal ones, social security, public sector banks, judiciary, water, telephone, electricity, state schools, arbitration, agriculture, industry, tax and customs, hospitals, police, development agencies and a few others.

The other similarity is that "the stigma associated with bribery in Peru is not large. Proetica, a Peruvian anti-corruption group, found that when asked to define the Peruvian slang for bribe ("coima"), less than half their survey respondents gave answers with a negative connotation."

Using extensive data - such data must be a first of sorts: more about it in a minute - they have come up with a theory with predictive power as well. It predicts that "bribery is more attractive to both parties when the client is richer, and that both bribery incidence and value increase with household income."

But it also turns out that almost two-thirds of the relationship between bribery and incomes is because richer people use public officials more than the non-rich. "This is likely the result of greater demand for public services by the non-poor." South Delhi residents, for instance.

Not just that, they identify the corrupt officials more, and they do it better. Sometimes the rich also bribe voluntarily in anticipation of bad service. This, they think, will bring the service up to the standard one may normally expect.

The authors then go on to argue that "bribery is not a regressive tax" presumably because it is mostly the rich who have to pay. I am not sure what the moral implications of this are. "The results suggest that the bribery tax is a progressive, or at worst flat, tax."

Be that as it may, the authors say "given this, we think it is unlikely that any differential deterrence effect of bribery on use by the poor and non-poor would be large enough to make the tax regressive for the whole population. The main distributional consequence of bribery appears to be a transfer from clients to officials in return for no net improvement in service."

Now to the data. There was one survey in 2002 and another stretching from May 2003 to April 2004. A quarter of the 2002 households were also interviewed in 2003-04. The bribery rate was highest for the police (37 per cent reporting). Fifty-two per cent of the respondents had bribed to obtain a driver's licence and reported having paid the bribe to an agent.

The authors admit that the "number of bribery episodes per household is understated owing to the fact that each respondent can only report one bribery episode per official per year." It is also probably under reported if dalals (agents) are used.

It would have been useful if the theory could predict a short run equilibrium price and how to get there. Unfortunately, it doesn't.

One last point: in Peru, a client who refuses to bribe has a probability (of getting his work done) 16 percentage points lower than one who pays. NCAER, with its strong survey division, should study the corresponding numbers here. Perhaps the PMO or the Planning Commission will give it a research grant.