MethodLogical is now at

MethodLogical is now at

Due to some persistent technical issues we've been having with Blogger, we're now posting at Please update your RSS feeds, etc. For the time being, our new posts will automatically be mirrored here, but you'll have to visit the new site to comment.

Monday, November 8, 2010

Health insurance as a gateway drug?

This past week, I wrapped up a trip to East Africa as part of my work with a microcredit organization. I was initially inspired to enter the microfinance field based on the "give the poor a loan, grow a business, success abounds" story. Naive? Sure. I was one of those richly educated graduates new to development with little field experience hoping to find that one thing blocking the poor from unimaginable wealth and development. Silver bullets, as Danika noted, don't exist in development.

After a few years seeing microfinance in Africa, I largely believe microcredit helps a small subsection of the poor to grow their microbusinesses into small enterprises on to medium enterprises and eventually into major employers. For the rest, loans make businesses a little bigger or provide an income smoothing device and commitment mechanism. Don't get me wrong, going from $1/day to $2/day or even allowing the $1/day to be the same each day is still valuable. In meeting borrowers around Africa, I've continued to hear that loans allow families to pay for kids' school fees (either because of additional money earned from businesses or by allowing a family to access a lumpy sum of money difficult to otherwise save to pay for a semester's fees) and the downstream benefits to these children from continuous education can be critical when human capital in a country is low. In the least developed places (e.g., South Sudan), I've learned that profits from loans can allow a family to feed all the children each day.

The real lesson of microfinance is that profiled extremely well in Portfolios of the Poor: the poor use and need a wide range of financial services. Loans are one piece, often one of the smaller pieces, of the puzzle and complement savings and insurance that are critical mechanisms to handling shocks that hit the poor daily.

However, Timothy Ogden and a range of others have pointed out a strange problem with microinsurance:
Microinsurance is one of those maddening products in development circles because it makes perfect sense (so much so that its a bit surprising it hasn’t evolved on its own) but the existing offerings haven’t had much success—take-up rates are typically very low.
If you ever travel to Kenya or Uganda, ask someone about a SACCO (savings and credit cooperative) that went bust. The story often goes that a charismatic individual starts up a SACCO offering loans at great rates if you save X% upfront. Upfront compulsory savings is typical in Africa for legitimate microfinance or even bank loans. Back to the SACCO: You save over a number of weeks as it's difficult to plunk down all of the money in one go. Some people don't even want loans, they just save with the SACCO because the interest rate offered on savings is also good. The SACCO continues to grow rapidly and then the General Manager takes the savings and runs away. It's a ponzi scheme and the recoverability of savings is next to nil. Those with loans outstanding greater than their upfront savings win but the broad majority have net savings with the SACCO and lose it all.

We take for granted that our financial institutions rarely go under and when they do our deposits are insured. In the absence of that, it's a bit more difficult to trust a financial service provider and the bias is to obtain funding from the provider (read: loans), even at higher costs, rather than save or purchase a service.

An extension of this dilemma is the lower than expected real demand for microinsurance. Insurance is predicated on paying premiums before you get paid out as a way to hedge against the potential for real negative shocks (i.e., fire, car accident, etc.). The poor, as it happens, get hit with shocks more frequently. They are also wary, though of obtaining microinsurance when they don't know the provider (especially the provider's staying power) all that well. Furthermore, it's hard to sell someone with a time horizon largely focused on this week and that agricultural insurance will make sense over the 3 year horizon.

Health insurance, in the category of insurance, is a unique offering in that it hedges against major shocks (catastrophic health events including ER visits) and pays out somewhat frequently for routine health needs (primary doctor visits and, in the case of Africa: diarrhea in children, parasites, etc.). Purchasers of the policy get a payout in the near term providing some kind of assessment of the insurer and recouping some of the premium. Anecdotally, the uptake of health insurance sold by Kenyan microfinance institutions is pretty strong and some microfinance institutions even require purchase as part of the loan as health-related shocks are a leading cause of loan default for small traders in many places.

A Rwandan man shows off his health insurance card.

From Ogden's post:
Chris Udry of Yale presented an experiment in Ghana that tackles these questions. The experiment was based on a large question: why do farmers underinvest in their farms? Is it that they lack investment capital or that they are risk averse? To take this on, the program provided some farmers with rainfall insurance, some with cash grants, some with both and of course a control group. Bucking the trend of small impacts, the study found significant impacts on farm investment: farmers in the treatment group bought more fertilizer, planted more acreage, and hired more labor. Unsurprisingly come harvest time they had higher yields and more income. They missed fewer meals and their children missed less school.
With such large and positive impacts you would expect that the farmers would be interested in buying insurance for the following year—and they were. But subsidizing a year of insurance to generate clients isn’t a great way to scale microinsurance. So Udry’s team looked at demand among friends, neighbors and acquaintances of the participants to see if word spread and generated demand beyond the original participants. They found that demand did spread along social networks with those who knew multiple participants more likely to be interested in purchasing insurance, particularly if they knew someone who had gotten an insurance payout. But the really exciting thing about Udry’s research is the significant benefits participants achieved.
For farmers, agriculture shocks are among the most pernicious. They cause farmers, ex ante, to invest in lower yielding options that are more hardy (e.g., cassava grows everywhere and under very harsh conditions) but provide less income. For those that take a bet on higher yielding crops like coffee or tea, one bad season can cause a family to go hungry and curtail key development expenses.

Why mention agriculture and health insurance together? Insurance in its purest form, paying a premium to mitigate the highly negative impacts of rare events, may be a hard sell to the poor. But health insurance may be less so because of how visceral the impacts of not being able to go to the doctor are and the more frequent payouts. If insurance companies started there, built their brand and reputation among the poor, might they be able to sell and expose the poor to the value of products like agriculture insurance?


  1. Hey Ben- this is really neat. In American health policy discussion, people talk a lot about moral hazard- the concept that people will be compelled to seek additional (and possibly unnecessary) health care when they know they don't have to pay for it. How true a phenomenon it is is not totally clear, but it's an effective talking point. What's cool, is that this might actually be a good thing in the developing world where health care utilization is low. Additionally, transportation costs and other indirect costs of care may lead disincentivize seeking of unnecessary care, further preventing abuse of insurance.

  2. We actually know surprisingly little about the impact of microfinance. This new working paper by Banerjee, Duflo, Glennerster and Kinnan describes the results of the first-ever randomized controlled trial on microfinance:

    While they do find some benefits, uptake is low and certain households increase their consumption of "temptation" goods like tobacco and alcohol. If you believe people are vulnerable to making unhealthy choices when they're addicted, etc. then it's not clear whether a microcredit program is a good thing, at least in this sample. I may blog about this article in more detail later. It's fascinating but we need to work through what it actually means for microlenders.

  3. I slightly disagree with one comment made by Ogden that "subsidizing a year of insurance to generate clients isn’t a great way to scale microinsurance." In fact this is exactly what I have heard what large agricultural corporations like Monsanto are doing to attract small farmers to their seed and other products - give them free or subsidized to start, demonstrate success, build trust, do that for one year or so, and then for the life time charge unsubsidized rates. It's good marketing, and could be used to attract clients who otherwise are unfamiliar with for-profit development products. Even Ogden noted that the marketing effect did not just affect the client, but also trickled throughout their social network.

  4. Jason,

    The study on microcredit actually found a reduction in temptation goods ("Spending on temptation goods is reduced by Rs 9 per capita
    per month.").

    In addition, we have few RCTs on microcredit but the degree to which we "know" surprisingly little about microcredit is more debatable. I think the distinction is critical and, as much as I respect the randomistas, I also believe there is a need for humility and acknowledgment of other types of learning/knowledge.


  5. Ben -

    You're completely right that the mean treatment effect on temptation spending was negative. But as I said above, it is indeed the case that certain households increased their temptation spending, namely those without existing businesses and with low propensity to start one; see Table 6, Column 5. Banerjee et al. note that this increase is roughly proportional to the overall rise in consumption, but it's still problematic if we think that these goods are actively bad for you (e.g. cigarettes).

    I didn't mean to imply that an RCT is the only way to learn about microfinance, but I agree with the paper's authors that selection bias is a major issue for non-random evaluations of microfinance. However, it would be foolhardy for me to ignore the body of knowledge accumulated by microlenders. I can think of two ways in which that knowledge can be put to better use. First, publish what is already known; if academics are unaware of solid evidence, chances are other MFIs are as well. Second, improve the quality of future evidence by randomizing future program expansions (or using another technique) to mitigate selection bias.

    I want to emphasize that I have zero doubt that microcredit is valuable and has positive effects. The remaining questions are about how big the effects are, and how they are distributed. Banerjee et al. simply confirmed what I knew about the overall value of microcredit; their contribution is in terms of the magnitude & distribution questions.

    Incidentally Google scholar tends to link to an older version of the working paper; the latest one I'm aware of is located here:

    - Jason


Note: Only a member of this blog may post a comment.