Letters to a Young Actuary

Being as that I work on a randomized evaluation of an index-based rainfall insurance product in Ahmedabad, Gujarat, and seeing as that there haven't been any payouts from the policies over the last three years, it is perhaps fair to say that the promise of rainfall insurance -- at least among SEWA members in Gujarat -- appears to be swiftly on the wane.

In some ways, no payouts over three years is perhaps not too surprising, given that this insurance is meant to cover 'catastrophic situations' and it looks like the districts in which we work (Ahmedabad, Anand and Patan) haven't had too bad a run of late.

That said, to any actuaries designing such policies or organizations interested in implementing such a policy, I'd like to share a couple of observations that have come out of focus groups with farmers, field visits, analyzing the data and just thinking about the effectiveness of this product in aiding household risk management:

1.) Variability of rainfall across time: One really needs to factor in different activities in the cycle of agricultural productivity in to the policy, i.e. sowing, the growth period, harvesting and so forth. At present, policies tend to insure against cumulative amounts of rainfall which, of course, doesn't take in to account excess/deficits in the amount of rainfall during certain periods (e.g. too little water during sowing or the growth phase is undesirable, just as too much water during harvest is undesirable).

One farmer compared cumulative measurements of rainfall to receiving 5 liters of milk in one day – 5 liters is plenty of milk, but it’ll go bad in 3 days, so what’s the point? (thanks to Sakhi Vyas for translating this in light of my non-existent Gujarati)

2.) Variability of rainfall across space: At present, contracts are typically written contingent on district-level measurements of rainfall. There is perhaps a good actuarial reason for this -- or perhaps it is just the unavailability of data at a finer level -- but either way, the variability of rainfall within districts is nothing short of remarkable. For example, we work in two talukas in Ahmedabad (Sanand and Daskroi) in which there was a difference of 400mm of cumulative rainfall in the last monsoon. This, in turn is about the 10 year average rainfall for Santalpur taluka in Patan, another taluka in which we work.

If folks want to make this work, private industry should work with the government to set up more geographically precise measurements of rainfall, or else, the basis risk -- the risk that the data on which claims are contingent upon bears no semblance to the the reality experienced by the claimant -- exists to such an extent as to make the policy near farcical. Perhaps contracts can be written contingent on taluka/tehsil level measurements of rainfall?

3.) Crop-specific water requirements: It probably makes sense to have crop-specific policies http://www.blogger.com/img/blank.gif, because different crops have different water requirements, and these are published by the FAO. I suppose the competing tension would be to make the policy simpler, but I honestly think the difficulties with selling insurance to rural consumers are in how people think about probabilities and payoffs; farmers are very well acquainted with different water requirements for crops and I imagine this would be quite intuitive to explain.

4.)Keeping policy-holders informed: At present, I don't think we do a good enough job of keeping farmers informed of the amount of rainfall, or the amounts needed for a payout, resulting in confusion, apathy and in many cases, anger.

Moreover, most small and marginal farmers seem to take their cue for rainfall predictions from the Gujarati calenders that publish panchang/astrological based rainfall predictions, rather than the vacillations of the Indian Meteorological Department.

To counter this, I think we should find way (perhaps cellphone, tv and/or radio) to send accurate rainfall data (the same that is used to determine payouts), as well as find a way to publicly display information regarding payouts (i.e. amounts of rainfall required for payouts, and how to calculate payouts). Or else, we are left with word-of-mouth explanations of these which probably change over the course of a three month monsoon period and lead to unrealistic hopes and expectations.

If anyone has any further ideas, I'd be glad to get your views.