FAI Video: Economist Dean Karlan and Timothy Ogden Discuss Microfinance and Reaching the Ultra-Poor (Transcript)
Timothy Ogden : So, Dean one of the innovations in microfinance is trying to figure out a way to reach further down the poverty ladder—with programs generally referred to it as graduation or targeting the ultra-poor.
Would you tell us something about what those programs are and the process of figuring out how they work?
Dean Karlan: Sure. So, the first thing I think that’s important to note is the broad scope of the goals. BRAC in Bangladesh is a micro-lender who was an early group thinking about these issues—but in a very simple way this really had absolutely nothing to do with microfinance. This is about how you help the ultra-poor build up a livelihood. It’s recognizing that micro-lending despite its rhetoric has never really reached those people and so it has never been a satisfactory tool not only in reaching them but also lifting them from poverty when it does reach them. But in a sense this really is an issue which has nothing to do with microfinance or the way we should be thinking about it.
We should be thinking about it as: what do you do with the households that are so poor that there is no real sustainable livelihood going on? That they are having to live hand to mouth and not always hand to mouth because they don’t always have food, they don’t have the skills, they don’t have the information and they don’t have a way of getting out of this trap—this poverty trap basically—because there is really no reasonable access to credit that can reach them. So, really what this program did is started off by saying the way we are going to solve this is not about giving them a loan. First of all the risk of repayment is too high and odds are they probably don’t even want the loan. The thought of dealing with a formal institution like this is too scary and too outside of their normal context. So let’s take a big push approach, let’s give them 4 goats, let’s give them training on how to use those goats, let’s go and monitor to make sure that the goats are being taken care of, they are raising goats in the way we need to raise them. Let’s give them consumption support, a weekly bag of rice or monthly bag of rice and beans so that they use that to eat and not sell the goats in order to eat. Let’s give them a savings account, let’s provide them healthcare. So it’s really a big push which is wonderful from a program perspective, in the sense that there are lot of reasons and theories to suggest that all these components are necessary. It’s honestly horrible from a learning perspective and it’s quite reasonable to think that all of these components may not be necessary and there may be a couple of them which may be really expensive and you can do without and then service more people. So this is one of the struggles that these projects have had. And so we have set up seven randomized trials on this idea but six of them are all testing out the package versus a control— which is great for establishing the base case whether or not this is a successful idea but it’s horrible for understanding the ins and outs of why is it not going to work or why it does work. In Ghana, we have a setting which we are very excited about in which we are teasing out some of the components. So there is a set of people who are just getting the assets transferred. Here are the four goats which are going to bring you out of the poverty trap in one fell swoop but do nothing more, is that sufficient to get you out of the poverty trap? We don’t have results on them, yet, but when they are out we would be excited to come back and share them.
Ogden: So, one of the standard questions often about randomized control trials is internal validity versus external validity. Doing seven different sites in seven different places around the world would seem to go a long way to answer that external validity depending on how varied the results are. There is a curious
1
thing, too, on some level —if all the results seem the same in ...very very different contexts that would also be puzzling. For people this poor in Ethiopia versus Honduras...shouldn’t things work a little differently?
Karlan: Sure. Yes and no. First of all some of the results that are coming out are strikingly similar. The reason we are doing it in several places is not about external validity, the thing that is not well understood is that when we do randomized trials we are doing them in order to improve the internal validity of what we can learn in a particular situation. There is no harm done to external validity by doing randomized trials, we don’t lose external validity. You don’t gain anything, you gain internal validity. When you have stronger internal validity you have more confidence in the results there. And so, if you want to go somewhere new and ask the question whether the results in this other place should be used over here, you should be asking whether the results were internally valid. It’s the first starting point. RCTs help with that. And then you need to ask: what’s the theory as to why this works? What are the contextual factors that are important and if those apply, then they should apply over here. That’s the external validity question and you have that from the understanding of the context and having a theory as to why this is working. Having randomized variation in one particular way something was done neither helps nor hurts when it come to the external validity. The reason for doing seven is more about persuasion, is more about politics and is more about variety in terms of the organizations that are doing the work in order to able to come up with a preponderance of evidence across these different types of organization structures and seeing whether it systematically can have this type of impact. In that sense an external validity question is gained but it’s gained by being able to say that there is Government, here’s an NGO. But at the end of the day let’s say the NGO’s side works and the Government’s side doesn’t. I am going to be really hard pressed to sit back and tell you I have this theory which says that NGOs are doing better than the Government when there are other factors that are going to apply and drive success in one place versus another. So to get to those types of issues one of the first things we need to do is to have better monitoring data so that we can really say concretely, when one side has more success than another, how much if this was due to the actual differences in implementation of the different sides versus some underlying theory about the type of constraints the ultra-poor are facing
Ogden: Some of the other interesting things about the difference between the programs are that in some of them you are evaluating not just economic well-being, you are evaluating psychological well-being. How is that being evaluated?
Karlan: So, this has become fairly common place in a lot of impact evaluations that are particularly working with the ultra-poor to the extent that stress and depression are important manifestations of being poor. We want to understand how these things are changing alongside more traditional economic outcomes. We have been learning a lot from psychology literature about how to measure things like depression and stress. We are hoping in some sites to use bio-markers. We haven’t been able to do that yet but we are using surveys that are batteries of questions that are basically measures of depression and stress and locus of control and other types of well-being measures and some of the results so far are actually showing striking impacts positive on depression. So, these are very exciting results to see because ultimately there has to be a bit of realism in any sort of intervention. We are not going to ever take the ultra-poor and make them rich. And so when we think about what we are really trying to do in any sort of social policy—it’s increase utility, and so in that sense I am particularly a big fan at looking at the measures of depression because it’s a pretty clear way of looking at the left tail of utility which is what we ultimately care about. “Utility” being economist jargon for happiness.
Ogden: When will results start to come out?
Karlan : Some results have started to come out but it will probably be 2 years before we see the full set.