<p><em>Stephen Yeo interviews John Quiggin for Vox</em></p>
<p><em>November 2010</em></p>
<p><em>Transcription of an VoxEU audio interview [http://www.voxeu.org/index.php?q=node/5912]</em></p>
<p><strong>Viv Davies</strong>: Hello, and welcome to Vox Talks, a series of audio interviews with leading economists from around the world. I'm Viv Davies from the Center for Economic Policy Research. It's the 3rd of December, 2010. Earlier this week, Professor Robert Townsend of the Massachusetts Institute of Technology was in London to present the Gorman Lectures at the University College London. The title of the lectures was &quot;The Evaluation of Financial Systems in Developing Countries.&quot; Stephen Yeo from CEPR met up with Professor Townsend immediately after the lecture. He begins the interview by asking Professor Townsend to explain the role of risk sharing and networks in village economies.</p>
<p><strong>Stephen Ye</strong>o: One of the things that seemed to come out of what you were talking about was the thing that was in your 1994 Econometric paper on risk sharing in Indian villages, I think.</p>
<p><strong>Professor Robert Townsend</strong>: That's right.</p>
<p><strong>Stephen</strong>: You did make a point there about the ability of networks to spread and share risks, and I thought part of the story of the lecture was about how far you could go with networks and then how far the financial system had to step in to complete the job. Is that a fair..?</p>
<p><strong>Professor Townsend</strong>: Yes, it's very fair. At the local level, say, within the village, we're asking how well gifts and small loans and some sort of informality is doing to smooth adverse shocks among the households within the village, particularly so among households who are related to one another.</p>
<p><strong>Stephen</strong>: That's what you mean by network, or is that a broader concept?</p>
<p><strong>Professor Townsend</strong>: Well, network could mean that. It means, literally, just drawing lines among households who are related by blood or marriage. It begs the issue of, how does the network really work in practice? But it's an approximation, a hoped for approximation, of some measure of the willingness of people to help each other. Networks could also mean just, literally, who's helping whom in the data as we observe it, small gifts and loans and so forth. Actually, that turns out to be a different measure, but they are correlated with each other.</p>
<p><strong>Stephen</strong>: In terms of what you actually found - this goes way back to your 1994 paper, I suppose, in some sense - you found that networks did some pooling of risk.</p>
<p><strong>Professor Townsend</strong>: In the ICRISAT Indian villages, we had measures of caste relationships, but we didn't have a measure of who was related to whom, so we couldn't really do what we've now managed to do in the Thai data. Initially, when we first started interviewing, we&rsquo;d do an extensive census of everyone in the village, and we&rsquo;d find out who are the cousins and aunts and grandparents and so on, so we know all the connections among the sampled households, and then we can watch what happens over time. So the story is, basically, that networks are helping quite a lot. But not everyone in the village has family around, and those people are the vulnerable ones.</p>
<p><strong>Stephen</strong>: The non-networked.</p>
<p><strong>Professor Townsend</strong>: The non-networked people. So other things equal, being poor is not helpful. You're more likely to be suffering from adverse shocks and the consequences of that if you're poor. But if you then take into account that they may be connected to other households in the village, you can compensate for that wealth effect. So the networks consist of family related people. But some people in the same kinship group are wealthy. Some are poor. They all seem to be helping each other to some degree. We worry a lot that this is a bit of a fluke, or a very special finding, so we keep doing it different ways. We look at the relationship between consumption and income.</p>
<p>We then modify the model and look at more data that allows for labor supply and earned wages. We also look at the rate-of-return on assets, the way someone in financial economics would look at market portfolios on the New York Stock Exchange.</p>
<p>We look at households as engaged in various occupations and enterprise, and we look at their returns on assets and see if that's consistent with efficient markets. And every time we do this, using different data, we come up with the same conclusions - that these networks are doing pretty well at allowing households to smooth adverse shocks.</p>
<p>But the flip side is that relatively poor people in the same villages who don't have family around, they are the ones that are suffering. When income drops, their consumption drops, or they have to do more on the labor market to compensate. And so if one were looking for a guide to policy, we would think about targeting them as a group in need.</p>
<p>But we wouldn't do blind targeting by wealth. We would want to take into account whether or not they have family around.</p>
<p>Another way to say this is, if we introduced an insurance program within a village, we would expect the take up, other things equal, to be higher among the low wealth people who don't have family around, because the others wouldn't need it.</p>
<p><strong>Stephen</strong>: Would you expect that to be true in a developed country as well, but maybe a bit more attenuated?</p>
<p><strong>Professor Townsend</strong>: I think people suspect that these family ties play less of a role, but I've been doing some work in Spain, in collaboration with the Bank of Spain, using data of firms, and it turns out that we know through this Bank of Spain database whether or not the firm is borrowing, and whether they're borrowing from one bank or several banks, or not borrowing from any bank at all. Then we made a bit of a surprising discovery - that the people without banks at all seemed to be doing better in smoothing their investment against cash flow fluctuations than the people who were borrowing from formal sector banks.</p>
<p>So then we're pondering what the mechanism is, and it turns out that this is correlated with them being in a family connected group of firms, perhaps informally, even at the level of looking at the phone number and the names of the boards of directors, not necessarily a formally incorporated conglomerate.</p>
<p>So it's painful and tedious to construct these family networks for an advanced industrialized country, but nevertheless, it does seem that this mechanism, the family and kinship groups, play a role, even as economies get bigger and much more developed.</p>
<p><strong>Stephen</strong>: So it's nice to have a bank, but it's even nicer to have a family.</p>
<p><strong>Professor Townsend</strong>: The message of the Gorman Lectures is, how connected are people from the village to the outside, say, across villages or to national level institutions?</p>
<p><strong>Stephen</strong>: Because you call them small open economies. So they're very small open economies, I suppose.</p>
<p><strong>Professor Townsend</strong>: Well, that's right. The typical village could have something like 200 households, although it could be as little as 25 or as many as 500, in Thailand. Some people incorrectly assume, when you talk about villages, that I mean villages as closed economies. And they're very hard to find. I mean, perhaps way back in medieval England, where there was a lot of problems in shipping grain around and so forth, you could take a village as approximately closed, but nothing in my arguments requires us to think of villages as closed.</p>
<p>Now, the degree of openness does vary a lot, depending on the region you're in. In some regions of my data, maybe as much as 50 percent of the consumption is homegrown in the village, whereas near Bangkok, 95 percent of it would be bought outside the village and imported.</p>
<p>So we're back to &quot;better to have a family than a bank.&quot; It turns out, things seem more complementary than substitutes. If one person in the household, say, is borrowing from a bank, they have the ability to get funding for investment projects that way, and also to potentially smooth consumption if they're having a bad income year. So they're directly linked. You could draw a line from the household to the government bank for agriculture, for example.</p>
<p>But back to the networks, it turns out being indirectly linked, as a customer, is just as good for smoothing consumption as being directly linked. In other words, that network is like a syndicate, and it's pooling the risk. Being connected to the outside allows better connections, better smoothing, not just against village level shocks but, again, shocks in the whole region, or even the whole nation.</p>
<p><strong>Stephen</strong>: Which you couldn't insure against.</p>
<p><strong>Professor Townsend</strong>: Which you couldn't insure against if it were only local. But even a household that's not connected as a member, who has no savings or credit with a financial institution, would show up as not having access to formal financial institutions in a survey, we know,from our data, is getting as much benefit, through the kinship network, as the direct member is getting. We did not anticipate that finding either.</p>
<p><strong>Stephen</strong>: So that's a new finding.</p>
<p><strong>Professor Townsend</strong>: Yes, that's relatively new.</p>
<p><strong>Stephen</strong>: And it's not specific to the Thai case. You've been doing the Thai studies for 10 years, haven't you, or gathering that data for quite a while?</p>
<p><strong>Professor Townsend</strong>: Well, we started in 1997, so we're up to 13 years. And the monthly data, we're at 12 years. We started that in 1998.</p>
<p><strong>Stephen</strong>: And you would expect that kinship/network phenomenon to exist in other places? It's not culture specific or country specific?</p>
<p><strong>Professor Townsend</strong>: I think that there is a great need for people to be gathering data. I wouldn't ask of them that they commit to gather it for 13 years, although I wish more people were doing it. I think this is the big risk of development economics. The data is still somewhat sparse, and so one tends to extrapolate and project from local experiences, from existing data, to other regions of the same country or to other countries. On the other hand, we know better what to look for. So knowing what we know now, we can provide advice to other people who might like to do a survey and suspect kinship networks may be playing a role. They don't necessarily have to gather 13 years of data to do that.</p>
<p><strong>Stephen</strong>: But still, they have to go after a kind of data that people, typically, probably didn't go after before you did these studies, which is to map these kind of associations between people.</p>
<p><strong>Professor Townsend</strong>: I think there is an increasing recognition, and not just in my work, that these networks are quite important. I mean, it's also very hard to sample because, if you don't sample everyone, then you sever some of the links, and it may appear in the data that people are more isolated than they actually are. So there's some challenging data issues involved. So a stratified random sample doesn't quite work as well as we might hope.</p>
<p><strong>Stephen</strong>: You did a census, didn't you, in effect?</p>
<p><strong>Professor Townsend</strong>: We did an initial census. Even though we didn't end up serving everyone, at least we know all the relationships in the village. So we know even when they transact with someone who I'm not continuously sampling, we do know who those people are. We don't lose the thread.</p>
<p><strong>Stephen</strong>: In the development literature, there's a big debate - which you know well, I'm sure - between the people who are advocates of micro finance and all that, sort of nontraditional lending, and the people on the other side who say it doesn't scale, it's not as effective, it's fine in some circumstances, but you need a formal banking sector and a formal financial sector in order to grow the economy properly. Does your work connect with that debate at all, or is it just looking at slicing the problem in a completely different way?</p>
<p><strong>Professor Townsend</strong>: Well, it connects in the sense that, during the time we've been gathering the data, there have been some unanticipated government programs that are much like micro finance. The government of Thailand, under Prime Minister Thaksin, the former prime minister, introduced something like a village level savings and loan association in virtually every village in the country. The government provided the initial funds to capitalize the institution, at something like $25,000 per village. Which, when you add up over the 70,000 villages, made for one and a half percent of GDP, which is arguably one of the world's largest micro finance interventions anywhere.</p>
<p>And we were able to assess the impact of that program through a fluke, basically, in the way it was administered. It was the same amount of money per village, but the number of households in a village varies. So the per capita treatment, if you will, varied. And so we could look at villages where there were few households, then there was much more credit per person, and we can see what happened and assess the impact.</p>
<p><strong>Stephen</strong>: What happened?</p>
<p><strong>Professor Townsend</strong>: Consumption went way up. But there are, as well, other things on the production side. Investment in agriculture went up. Local wages in the village went up. Savings actually went down. Overall, credit went up, which is important because you might worry that this is just a substitute for something else they already had. That did not seem to be true. This was a definite augmentation of the amount of credit. And defaults rate went up within a year after the program, a bit.</p>
<p>So, some of those things seem interesting, others a bit more challenging. We actually have created a structural model to try to understand the story of the impact of the intervention. And the idea is that, basically, people build buffer stocks in liquid savings against future emergencies, and when these village funds came in, they actually facilitated access to credit. So savings dropped because one could rely on credit in the future and less on buffer stock savings.</p>
<p>Consumption went up because there were some low wealth segments of the population that were constrained in the amount that they were able to afford a liquidity constraint, essentially. Although, it's interesting - there's also some interesting variety. We have these investment opportunities that were random but large.</p>
<p>And some households actually cut back on consumption in order to invest, and it's an investment they would not have made if they had not had the resources from the village fund. So in other words, they're pooling their own resources, augmenting it by cutting down on consumption, and then borrowing to be able to get to this larger size investment that they would not have otherwise done.</p>
<p><strong>Stephen</strong>: And that led to some efficiency gains, then?</p>
<p><strong>Professor Townsend</strong>: We've measured the welfare gains. It turns out, it depends on who you are in this population. The people who were constrained in consumption and these would be investors gained a lot. They would be happier with the program than with just getting a lump sum transfer in equivalent amount, but the majority of the population actually would have preferred the money coming in as a lump sum transfer.</p>
<p><strong>Stephen</strong>: Those are the numbers where you said 10 percent benefited a lot, 10 percent benefited almost not at all?</p>
<p><strong>Professor Townsend</strong>: Exactly.</p>
<p><strong>Stephen</strong>: Then there's this great mass of people in the middle who...</p>
<p><strong>Professor Townsend</strong>: I think something like 28 percent found the program beneficial relative to a transfer, and the majority would not. So this is an important qualification. In any of these programs, not just the ones that happened during the time I was gathering the data, in other countries as well, the impact is often heterogeneous. It's not uniform. It's not uniformly a good thing. Not necessarily. There have been, surprisingly, relatively few rigorous evaluations of these micro credit programs.</p>
<p><strong>Stephen</strong>: Very good. Thank you very much.</p>
<p><strong>Professor Townsend</strong>: You're quite welcome.</p>