The guys at AdamSmithee have been busy of late, and there's lots of good stuff there worth reading by anyone concerned about international development, but there's one particular post I'd like to highlight, especially in light of recent poorly-reasoned efforts to tear down Hernando de Soto as an advocate of ineffective policies (a rebuttal of which can be found here).
David Dollar, author of much cited, much debated work on aid effectiveness, is back with more. This time, he's using the World Bank's Operations Evaluation Department ranking of project success to see if it links with country characteristics. Echoing earlier results, he finds that World Bank financed investments do far better in countries with stronger property and individual rights.
Surprising? Only to those who are ever-eager to replace individual rights with those of "collectives." Those who are willing to put increasing the welfare of the poor ahead of their ideological obsessions would do better to look closely at why it is property rights and individual freedom fare so poorly in the LDCs. Hint: it has nothing to do with an excessively zealous commitment to "neo liberal" mantras or free-trade.
What is this, "Comedy Econometrics Week"? Abiola, mate, have a look at the scatterplots at the end of the piece. They represent what I can only call the "plane through a cloud" school of regression analysis. (I also note that out in the top right corner, having the best project effectiveness and one of the highest scores for rule of law is a point labelled "VNM"; I can only think of "Vietnam" as a developing country containing those letters in that order).
The regressions don't fit the data particularly well. The R^2 figures are really quite low. This is often a sign that the significance coefficients shouldn't be trusted, as it is likely that there is a missing variable which is correlated with some of the regressors. Given that the regressors include two subjective assessments (the Freedom House and the IGLC indices) and they are meant to be explaining another subjectively assigned score, it's not sound at all. The position that "property rights matter" is entirely sensible. But so is the proposition that private gun ownership deters crime, and that has been quite badly damaged by people attempting to draw out the thread of their statistical modelling finer than the staple of their data.
Posted by: dsquared | February 17, 2005 at 06:58 PM
"The regressions don't fit the data particularly well. The R^2 figures are really quite low."
Now that I look at it, you're right; advantage to you on this one.
Posted by: Abiola Lapite | February 17, 2005 at 07:11 PM
"The regressions don't fit the data particularly well. The R^2 figures are really quite low."
That's quite interesting.
Actually the R^2s in that paper are not that low when compared to R^2s one sees in many empirical papers when working micro data.
Posted by: Sam | February 17, 2005 at 10:09 PM
That is true, but it is a bd thing about the rest of the microeconometrics literature, not a good thing about this paper. R-square is not the be all and end all, but a glance at the scatterplots (fair do's, good on Dollar for printing them) shows that the number is accurately summarising the goodness of fit here; it's not good. Pick a specification test and I doubt this model would pass it; any simple linear model is likely to poorly summarise something as complicated as the relationship between social structure and development. The position of the Vietnam datapoint on its own ought to alert us that there is something else going on here.
Posted by: dsquared | February 17, 2005 at 10:18 PM
I agree that the results aren't all that strong but just looking at the scatter plot and calling it a "plane through a cloud" isn't much of a criticism on its own. The scatter plot illustrates the unconditional correlation, but the regression estimates conditionals. Looking at the regression results its seems that including initial income (to control for "convergence effects" or whatever - it'd also explain Vietnam and other countries, China, Malaysia, up in that corner) is what makes the fit better than what you'd expect just by eyeballing the scatterplot. Drop that initial income and I bet you get all zeros.
The R-sq and the t-stats are low enough though to make one uncomfortable with believing this as conclusive. It's more like in "interesting, maybe even suggestive, more research required" category.
Posted by: radek | February 17, 2005 at 10:37 PM
The scatter plot illustrates the unconditional correlation, but the regression estimates conditionals.
But the sum of squares shows that conditioning on all the other variables doesn't help all that much either. I'd tend to disagree on the general point; my experience has typically been that if an effect is there, it's usually not too difficult to generate a scatterplot that has at least some sort of qualitative shape to it. I also still think it's fishy that Vietnam is defined as a high rule of law country given the interpretation the author wants to put on "rule of law". (I suppose it is at least possible that I have misunderstood the chart).
I think, sadly, I'd put this one in the "interesting, perhaps suggestive, almost certainly not a problem that is amenable to regression analysis so not worth bothering doing more research of this kind, try a fundamentally different approach" category.
Posted by: dsquared | February 17, 2005 at 11:23 PM
btw, I've finally managed to read DeSoto and my assessment is that most of what's valuable in what he's saying is originally in "The Nature of Mass Poverty" by JK Galbraith, a book which really, really, shouldn't be as hard to get hold of as it is. The Slate hit-piece is a bit unfair in so far as it is directed at DeSoto rather than his wannabes (although Keynes, Marx and Milton Friedman have swallowed buckets of this treatment, so why should DeSoto be special), but I do think it's indicative that someone so beloved of a certain kind of development commentator doesn't really have a single concrete success story to point to.
Posted by: dsquared | February 17, 2005 at 11:29 PM
""try a fundamentally different approach" category."
Like what? Case studies? The usual problems are even worse with these. Or do you mean some high fallutin' statistical approach?
Regression might not be much but it might be all that
can be done at this point. Better data and (as you imply) better data definitions would probably advance the research the most at the margin right now.
As an aside, I also happen to think De Soto is a bit overrated (though I only read the Capital). Underneath all the anecdotes etc. he's basically saying that better defined property rights could ... raise the saving/investment rate. But that accounts for only a small portion of the difference between rich and poor countries.
Posted by: radek | February 18, 2005 at 01:20 AM
"But that accounts for only a small portion of the difference between rich and poor countries."
Only if one fails to take into account the *quality* of investment as well as its quantity. Even small amounts of private FDI that acts to shake up productivity in local industries can be far more effective than the World Bank throwing a few more billions at yet another dam project, but if most investment consists of the latter - as it will in most scenarios in which property rights are poorly defined - its hardly suprising that the results can be meager; but then again, that was the whole point of this post to begin with.
I'll give one example from my personal experience of how big a difference investment quality can make. Nigeria's mobile operators have invested far less in the country than the various transnational aid agencies have handed out over the years, but their impact on the lives of ordinary people in the country has been dramatically greater; it's hard to describe how much of a double-take it is to see ordinary farmers, mechanics, hawkers, market women, errand boys and the like blithely chatting away on cell phones as if it were the most mundane thing in the world, with 10 million subscribers after just 5 years, in a country in which the state monopolist NITEL managed to install only 400,000 lines in 30 years. Telephones used to be something the lower classes would come to beg the well-to-do to allow them to use, and then only in emergencies, but not anymore. You can imagine how life-changing this is for so many people, and this is why the whole business of property rights matters so much: I think that Hernando de Soto is, if anything, *underrated* rather than the opposite.
Posted by: Abiola Lapite | February 18, 2005 at 01:38 AM
Dollar's earlier results were also criticised for not being robust (see the second set of links in the extract from AdamSmithee), and some of the reported regressions in this paper again suggest the relationship isn't all that solid, which is why the (partial, I think) scatter plots are no great surprise. Although, to be fair to Dollar, pretty much any cross-country regression result can be shown to be fragile without too much effort... development is apparently a complex business.
The dependent variable (OED success measures) isn't necessarily a perfect one, either --it is hardly a transparent and completely objective measure.
But despite all of that it does seem likely that better institutions lead to better project outcomes. Surely a project where all of the aid financing ends up in a Swiss bank account is less likely to be successful than one where some money, at least, is spent on the intended activities...
Posted by: AdamSmithee | February 18, 2005 at 04:37 AM
" You can imagine how life-changing this is for so many people, and this is why the whole business of property rights matters so much"
This looks like a bit of a non-sequitur to me. I really don't see the connection between DeSoto's small-is-beautiful capitalism and mobile phone masts. Vietnam has mobile phones now too, but it also has collective agriculture (which has worked wonders for living standards in the country over the last ten years, at the price of knackering the world coffee price). This is what I meant by "a certain kind of development commentator" earlier, btw; my experience is that when people start quoting DeSoto at you, you have to watch them like a hawk because they start off by talking about title and legal form, and then ten minutes later they're yammering on about removing capital controls and getting rid of food subsidies as if it's all the same thing.
I also wish you'd knock it off with the "Swiss bank accounts" speech. The level of corruption in the Abacha government was not at all typical of aid-dependent countries, and it's just not the case that "all the aid financing", or even a meaningful proportion of it, gets diverted in most projects. As I mentioned on an earlier thread, if I was less scrupulously ethical, I'd accuse you of rather glorying in the worst examples corruption because they confirm your worldview and allow you to feel smug at the expense of the aid community.
Finally, there's no one-to-one correlation between "better institutions" and better outcomes, as the Eastern European cases show us. Better institutions, starting from the wrong base, can just mean better and more efficient ways to steal the cash. And widespread private ownership does not map one-to-one onto "better institutions"; the experiment with allowing foreign companies to become massively important political players has been tried in the fruit-growing countries of Central America and it wasn't exactly a roaring success. I suspect that this development model ends up looking a lot more like Honduras than Botswana.
Radek: I'm afraid it probably does come down to case studies in the end (for microeconomic work like this at least; I remain convinced that macro modelling of things like exchange rate regimes is worthwhile). They're massively flawed, but at least they don't represent themselves with spurious precision. I don't think that any statistical technique is going to get over the basic problem that when you reduce something as complicated as the legal and social system of a country, to a single number (or short vector of numbers), you are throwing away more or less all of your information.
Posted by: dsquared | February 18, 2005 at 07:30 AM
I've just noticed that Abiola didn't make the Swiss bank account comment and therefore apologise unreservedly.
Posted by: dsquared | February 18, 2005 at 10:26 AM
"This looks like a bit of a non-sequitur to me. I really don't see the connection between DeSoto's small-is-beautiful capitalism and mobile phone masts."
What non-sequitur? The connection isn't between "small-is-beautiful capitalism and mobile phone masts", but between *secure property rights* and higher quality investment, a connection I explicitly spelled out in what I was writing.
Posted by: Abiola Lapite | February 18, 2005 at 10:56 AM
"...but it also has collective agriculture (which has worked wonders for living standards in the country over the last ten years, at the price of knackering the world coffee price)."
Actually, I attended a job-market seminar 2 weeks ago where the persenter (an LSE PhD candidate of French-Vietnamese origin - Ms Ngo is her name) stated that collective agriculture was phased out in Vietnam, a little over ten years ago.
Posted by: Jeffery | February 18, 2005 at 08:50 PM
"Only if one fails to take into account the *quality* of investment as well as its quantity. Even small amounts of private FDI that acts to shake up productivity in local industries can be far more effective than the World Bank throwing a few more billions at yet another dam project"
Well, I totally agree that the quality of investment matters (and that simple analysis which use only saving/investment rates miss a lot - thought it's a hard thing to control for empirically) and that private FDI is a great thing. But DeSoto is not saying "follow policies which encourage more FDI" but rather "give poor people property rights which they can use to mobilize saving". Of course the two approaches can be complementary. But I tend to believe that, at the margin, encouraging more FDI would probably do a lot more for development than assigning property rights.
Or to put it another way, DeSoto's proposal's are sensible and good but they are not the panacea they are sometimes made out to be - they're not enough to turn poor countries into rich or even middle income ones.
(both links in the argument "prop. rights --> collateral increases saving --> more and higher quality investment" can be pretty weak). Basically, I think that to get governments to listen to him DeSoto oversold his ideas (not that I blame him).
Anyway, one can support his ideas just from a, ummm, "social justice", point of view. I'd rather the squatters had the property right than the government.
Posted by: radek | February 18, 2005 at 09:28 PM
stated that collective agriculture was phased out in Vietnam, a little over ten years ago.
No, she was right, my phrasing was sloppy. Farming rice and food crops through village collectives, which is what "collective agriculture" really means, bit the dust in the doi moi reforms. But the coffee plantations around Buon Ma Thuot, which has provided a lot of the foreign exchange for Vietnam's economic recovery, are big government owned agroindustrial complexes, which is what I meant.
Posted by: dsquared | February 19, 2005 at 08:23 PM