How do we know when index insurance payouts are right? It’s not an easy question to answer.

Tigray, Ethiopia during the rainy season. Insured farmers grow maize, wheat, teff, barley, and sorghum. Photo: Dan Osgood/IRI.

There are a bunch of cool ways to use things like cell phones to get massive scale reports from farmers around the world.

But can you trust that data?

When we gather data from farmers about the effectiveness of index insurance, there are reasons for farmers to answer questions strategically to protect themselves: If your insurance company asks you if you didn’t really need a payout, what should you tell them?

If farmers answer strategically, can we know if #indexinsurance payouts are right, or even if people understand what they bought?

This will be really critical to nail if insurance is to scale for real.

In 2016, we provided technical support for insurance reaching hundreds of thousands of farmers across Africa, including these farmers attending a product design and validation meeting in Senegal. Cell phones play a role in sales, marketing, and payouts for many of these projects. We hope phones will soon play a much more fundamental role in product design and validation as projects scale. Photo: Dan Osgood/IRI.

The solution could be #NAITIC (Noisy Audited Incentivized Truthtelling for Informed Clients), hopefully my most annoying acronym yet (sorry @climatesociety).

Unfortunately, it’s often tricky for grants to pay for this kind of work. If you’re interested in helping, click here or send me an email.

Back in ancient times (the 70s), folks like Stiglitz and Akerlof got the Nobel prize discovering this strategic behavior. The “Shepherd’s Dilemma” paper I co-authored with Glenn Sheriff a few years ago builds on this stuff through a livestock example. The theory is that you should get truth-telling if you have a contest for farmers to predict what an audit of themselves would find, and then you audit some of the lucky participants, giving a prize to those who were right, perhaps using satellite data as the audit (or part of the audit).

If you do it right, both the farmer and the audit can be “noisy” — meaning the data doesn’t have to be perfect. Keeping with the spirit of things, we did get a prize for our paper. In principle, you could have a contest in which farmers guess what years in the past a satellite would have identified insurance payouts for them. This should simultaneously improve the satellite calibration, inform the farmers about the product and inform the product about how well farmers understand things. That’s the theory.

If your insurance company asks you if you didn’t really need a payout, what should you tell them?

The big question is whether it works. And how big the reward should be. And how frequent the audit needs to be. And how accurate the audit needs to be. And if the satellite is good enough. And, of course, is strategic reporting from farmers a big enough issue to be a problem at all?

So, there is some work to be done, but we’re working on it. We are trying to test if prizes increase the accuracy of information reported by farmers, and if prizes work to encourage under-represented farmers to chime in.

Unfortunately, it’s often tricky for grants to pay for this kind of work. I’m actually using my own credit card for the fees of our most experimental prototypes. But donations can pay for this, and are always welcome. If you want to pitch in, click here.

If we succeed, the answers from a #NAITIC text message contest could automatically fix issues in index insurance by plugging directly into the communications tech-driven participatory index-designed systems we work with, like #SNIID (the Social Network for Index Insurance Design), my previously most annoying acronym.

As always, our page is at iri.columbia.edu/fi.