Skip to content

How ethnic profiling explains Dani Rodrik’s fondness for industrial policy

Airline passengers recently ejected an innocent Muslim family from an airplane because they were afraid the family were terrorists. Similar reasoning explains why Dani Rodrik favors industrial policy as a key to success.

Before getting overly critical of Dani, whom I admire a lot, let me confess I have frequently committed the same type of reasoning error myself, and so does virtually everyone else. But it’s still wrong.

All of us are making the amazingly common mistake of REVERSING CONDITIONAL PROBABILITIES. The airline passengers perceived from media coverage that the probability that IF you are a terrorist, THEN you are a Muslim is high. Unfortunately for the poor family, the passengers confused this with the relevant probability, which is the chance that IF you are a Muslim, THEN you are a terrorist (which is extremely low even if the first probability really is high, because terrorists are very rare).

So here is Dani Rodrik on success and industrial policy: “the countries that have produced steady, long-term growth during the last six decades are those that relied on a different strategy: promoting diversification into manufactured … goods” (cited in Economist’s View).

So Dani concludes, “What matters [for growth in developing countries] is their output of modern industrial goods” and that developing countries will have to get busy with “real industrial policies.” Finally, “external policy actors (for example, the World Trade Organization) will have to be more tolerant of these policies.”

Unfortunately, Dani is also REVERSING CONDITIONAL PROBABILITIES. Dani’s evidence is based on what he believes is the high probability that IF you have had steady growth for six decades, THEN you had industrial policy. This is interesting, but this is not the right probability in deciding whether to choose industrial policy, which is “IF you have industrial policy, THEN what is your chance of steady growth for six decades?”

This second, correct, probability would seem to be pretty low, since many other countries — especially African and Latin American — extensively tried industrial policies over the past six decades with low and erratic growth as a result. Attempts at forcing investments into industrialization led to a huge pileup of debt in Latin America in the 1970s, which erupted into a debt crisis in 1982 and subsequent lost decades, when the productivity of the investments proved to be low. The more extreme results in Africa include the Ajaokuta steel mill in Nigeria which went through $6 billion but never produced a bar of steel, or Tanzanian manufacturing, which had NEGATIVE growth of output per worker despite heavy capital investments. (For more on this see this paper.)

I am really going through a MAJOR Mlodinow slash Kahneman phase about how economists (present company included) misinterpret data. To all of you who I am tormenting with this stuff, I promise to move on to something more constructive soon, like maybe another edition of our popular series Notes from the Field.

This entry was posted in Cognitive biases and tagged . Bookmark the permalink. Follow any comments here with the RSS feed for this post. Both comments and trackbacks are currently closed.


  1. Alanna wrote:

    I really like this series. Notes from the field are the kind of stuff most of us have heard already. Serious thinking about causality is rarer.

    Posted May 14, 2009 at 1:10 am | Permalink
  2. Doug Barnes wrote:

    I quite like this series so far. It’s funny and engaging (because of the author), and it’s useful and informative to the non-economist.

    I’ve dealt with this particular cognitive bias both in the days when I was a security consultant, and also now as a lawyer.

    With respect to this article by Dani in particular, I think you’re right to call him on it. I came away from “One Economics, Many Recipies,” however, with more of an overall take of “industrial policy is not always bad, if carefully employed in connection with other necessary institutions and policies, adapted to local purposes.” This is not obviously wrong, but is a certainly a harder thesis to market in a newspaper column.

    Also, it’s been a while, so some of that may have been his book passing through my mental digestive tract and imposing some of my own biases on it.

    Posted May 14, 2009 at 6:58 am | Permalink
  3. Clay Wescott wrote:

    The same argument may apply to decentralization in developing countries. Some argue that we need more decentralization, pointing to a few cases of successful local government. This may be based on a belief that if you have strong development outcomes, then you had decentralization. This is not the right probability for deciding whether to support decentralization.

    Instead, you should ask: if you had decentralization, what is the chance of strong development outcomes? In the second case, the probability is low in many developing countries, where there may be thousands of local jurisdictions languishing under corrupt, under-resourced, ineffective administrations, and only a few doing well (and getting all the publicity). Some go on to argue that the reason the thousands are languishing is that we aren’t doing decentralization the right way; but if it’s so hard to do it right in developing countries, how can it be a good thing?

    Some countries may pursue decentralization for various political reasons, which is their prerogative. However, we should question the investment of billions of aid dollars on decentralization in the pursuit of improved service delivery, based on weak logic and evidence. For more, see Treisman, Daniel, 2007. The Architecture of Government: Rethinking Political Decentralization. UK: Cambridge University Press.

    Posted May 14, 2009 at 10:12 am | Permalink
  4. Anonymous wrote:

    In the end, isn´t it the case with industrial policy, once again, that some economists point out that growth was caused by a certain policy and others claim that growth happened in spite of such policy? I guess we don´t have this problem in micro experiments…

    By the way, I couldn´t agree more with Alanna. Keep these gifts of serious thinking to your daily audience, Mr Easterly and Mrs. Freschi.

    Posted May 14, 2009 at 12:37 pm | Permalink
  5. Luis Enrique wrote:

    I am not sure that your analogy is apt. The following is probably terribly offensive, but all I’m doing is trying to follow the analogy through:

    All poor countries are, presumably, trying to get richer. Not all religions are busy trying to produce terrorists. If all religions were busy trying to produce terrorists, but only the Muslims succeeding*, one might conclude that if you wanted to be a terrorist then converting to Islam might help. This would be true, even if there was only 1 terrorist for every 100 million Muslims.

    If it is true that all the countries that have achieved sustained growth practised activist industrial policy, the interesting information here is that no country pursuing alternatve policies has achieved sustained growth.

    Perhaps you are wrong when you say the right question, when deciding whether to choose industrial policy is: “IF you have industrial policy, THEN what is your chance of steady growth for six decades?” Because, however low that probability is, it appears activist industrial policy offers the highest probability of achieving that goal. The right question to ask is rather: “IF you have industrial policy, THEN what is your expected growth rate over the next six decades?”

    To illustrate what I mean, consider the case in which for every 10 countries choosing industrial policy, 1 achieves sustained growth and 9 fare disasterously, so expected growth is negative, whereas for every 10 countries chooosing hands-off policy, none achieve steady growth but all 10 countries achieve sporadic but on average respectable growth.

    Perhaps you should really be arguing that because of the probability distribution of outcomes, Dani is setting the wrong goal for policy makers in developing countries – maybe he is doing the equivalent of asking, what is the best chance of winning a million dollars? It’s higher when buying a lottery ticket than when not, but that doesn’t mean buying lottery tickets is a smart idea.

    But thinking about the distribution of outcomes, and expectations, is a subtly different thing from asking which policies are capable of delivering six decades of sustained growth. If that is your question, I don’t think Dani’s reasoning is fallacious.

    I can’t off the top of my head think of any research that shows the distribution of growth outcomes for poor countries pursuing industrial policy, and for those doing otherwise. Can anybody offer a reference?

    * I know this is not the case; it’s a hypothetic argument.

    Posted May 14, 2009 at 1:13 pm | Permalink
  6. happyjuggler0 wrote:

    Personally I have long been of the opinion that for an underdeveloped country to grow at a high rate, the head of state has to have that as an objective, indeed probably has that as the highest objective in economic policy.

    Furthermore, it is not enough for a current head of state to have that objective, it has to be sustained by whomever future heads of state are. Still further, it probably needs to be credibly seen as an enduring goal, or you won’t get the long term investments that you need for such growth, by either locals or especially Foreign Direct Investment.

    If the country’s head of state doesn’t have that objective and that credibility, the country won’t grow at a high rate.

    But if the country’s head of state does have that objective and credibility, I strongly suspect that he could choose from a range of “recipes” that would yield a tasty growth rate, anywhere from something that Milton Friedman would swoon over to something that Dani Rodrik would swoon over.

    Sadly, most such underdeveloped countries are in fact severely underdeveloped precisely because their collective heads of state had other objectives that contradicted the goal of wealth creation. Namely kleptocracy or benign-intent wealth redistribution from the productive to the unproductive.

    By the way, it is worth stressing that sustained “high economic growth” in a country is a euphemism for sustained “massive wealth creation” in a country. If one views capitalists or capitalism as evil or something to be avoided for whatever reason, such as the Rodrik blog post and Easterly response here, then that clearly radically diminishes the chances of such wealth creation from happening.

    Posted May 14, 2009 at 2:43 pm | Permalink
  7. Tord Steiro wrote:

    Firstly, I must say I enjoy your ranting about causality. It is unusual and highly educative for economists to read serious analysis of serious work. At least for me.

    Second, I think you are wrong in the way you dismiss industrial policy. Yes, I agree that having industrial policy is no guarantuee for development, in fact, it is very often a guarantee for corruption, mismanagement, and kleptocracy.

    But, certain countries managed to get their industrial policies right, even if so many got them all wrong. The question is why? If you are running industrial policies, which often gives incentives for theft, why did it, in certain countries, NOT encourage theft, but rather the opposite? Why did certain countries institutionalize anti-corruption boards and serious countermeasures to reduce the potential perverse incentives created by industrial policy?

    This is a most likely a question abut political economy. In Latin America, the short version would be that industrial policy where used to strengthen one particular group of society – hence it failed. In many African countries, industrial policies simply a sheer cover for stealing and theft – hence it failed. In some countries, as you point out, industrial policy worked – probably because in these countries, the industrial policy was genuine for development of productive capacity, and none of the above.

    This point is often missing in Rodrik’s work, which often seem to be a bit naive; let developing countries make their industrial policies through trial-and-error, and, over time, everything will be fine.

    I have said it before, though: Incentives matter. The power of icnentives, and how they may shape outcomes of policies, is beautifully explained in your own “The Elusive Quest for Growth”, the best book on development economics I have read. I would be thrilled if you could come up with an analysis of the incentives governments face in the context of managing a succesful industrial policy.

    Posted May 15, 2009 at 2:36 am | Permalink
  8. Dani Rodrik wrote:

    Dani Rodrik’s comment has been moved to the main post section of the blog

    Posted May 15, 2009 at 6:52 am | Permalink
  9. Below is my response to this post. I have also posted it on my blog and it is in the April 16th links of Economist’s View, titled, “Scientific does not mean simple-minded”:

    Here is what I think is the biggest overall way that economists misinterpret data: They think scientific means simple-minded, like you aren’t scientific unless you assume away (and really take these assumptions literally, or substantially overly literally, act as though they are actually true in making conclusions for the real world) all of the complexity, all of the non-formal evidence and information, and assume everything relevant all fits into an extremely unrealistically simple formula, turn off your high-level, high-dimensional flexible thinking (which can still be rock solidly logical, with each link in the logic chains completely solid, even if it’s not written in fancy looking, but really grossly over-simplified mathematics), and ignore any a priori information that’s not formal and published in an academic journal in your field, no matter how important and compelling it is.

    Harvard Economist Dani Rodrik is clearly an extremely proficient and regular user of high level intelligence. He is not saying that any industrial policy is a sufficient condition for strong movement out of poverty, he is just saying that the formal simple statistical data suggests that it can substantially increase the odds, and if you read his books, you see that he constructs very strong logic chains using less mathematical and formal, but also less simple, evidence to add a great deal of support to the simple statistical data.

    Yes, there are a lot of countries that used at least some form (not necessarily a good one) and some amount (not necessarily relatively large and sustained) of industrial policy and failed, but that doesn’t mean that industrial policy (good industrial policy) often cannot substantially increase the odds and amount of success, because there is a lot more to success than just having any industrial policy.

    Every ATP tour tennis pro plays and practices an average of more than 10 hours per week, but only a fraction of 1% of people who play and practice more than 10 hours per week achieve ATP tour status. So are you going to make the simple-minded argument that it doesn’t matter if you play and practice more or less than 10 hours per week, that it doesn’t matter how much you play and practice per week in your probability of becoming an ATP tour pro? Or that “it’s impossible to tell” (This is one of the lines I find most irritating. It’s usually highly inaccurate and misleading. Often the evidence, all of the evidence at hand, although it won’t tell you for sure, it will tell you that one possibility is much more likely than another, but people tend to feel so smart and level-headed saying “It’s impossible to tell”.)

    First, clearly, if playing and practicing heavy hours was unrelated to success it would be very unlikely that it would just purely by chance end up existing in every single success story, but be relatively rare in the non-success stories. Now, there is data mining, but to see whether this is a random artifact of data mining, you should look at more than only the simple statistical data. You should use your high level intelligence and look at all of the evidence, even if it doesn’t fit into a simple mathematical formula. You look at human biology, physiology, neurology; you look at evidence from sports in general, etc., etc. From this you can construct extremely compelling, rock solid logic chains, anchored to just very reasonable and mild assumptions (as opposed to the kind typically used by Chicago-style economists), showing that how much you play and practice matters greatly in your probability of achieving ATP pro status, even if those logic chains don’t consist of solely highly simplifying mathematical symbols.

    And, by the way, I do think advanced mathematics can be very useful, and use it myself, but only if it’s used and interpreted intelligently.

    You can come to some pretty ridiculous conclusions if you ignore the vast majority of the evidence because it does not fit some snobbish and simple-minded definition of what’s scientific. Here’s my definition of what’s scientific, or at least the most important part of it: It’s logical. The conclusions to the real world that these snobbish pseudo-scientific types make often don’t meet this definition (see Chicago School).

    Posted May 17, 2009 at 9:44 pm | Permalink
  10. anon wrote:

    So, wait, you’re saying that because every pregnant person is a woman, we can’t assume all women are pregnant? In fact, at any given moment, most women in the world are not pregnant? Not even a little bit?

    Seriously, people are dumb.

    Posted May 21, 2009 at 11:27 am | Permalink
  11. rochey wrote:

    But this still doesn’t escape the fact that NO country has succeeded WITHOUT industrial policy.

    And therefore, what’s missing from Rodrik’s analysis is, Why did industrial policy succeed in some cases and fail in others?

    If all terrorists are muslim, we need to look at what makes some muslims terrorists and others not? It seems Easterly is suggesting, by his own analogy, that Islam is not relevant to an understanding of terrorism.

    In my view, the answer to the question of the differential performance of industrial policy is that it depends on the politics. Some countries were able to make a success of industrial policy because the balance of power between different groups and the state allowed the rents created by the state to be used for productive purposes, while in others they were captured by non-productive groups.

    Posted May 24, 2009 at 12:58 pm | Permalink
  12. Avinash Dixit wrote:

    Apologies for barging into this dispute so late, but I am not a regular follower of blogs and came across this purely by accident.

    The dispute between Easterly and Rodrik cannot be resolved by looking at countries that have had an industrial policy and seeing whether they have enjoyed development success (as Easterly proposes), or looking at countries that have enjoyed development success and seeing whether they pursued an industrial policy (as Rodrik argues). We need a four-fold classification, with two possibilities along each of two dimensions: (1) Industrial policy versus No industrial policy, and (2) Development success versus no development success. Then a chi-square test can be applied to this “contingency table” to test whether countries with industrial policy are significantly more likely to enjoy development success. Even then we will have only a statistical correlation; going from that to concluding causation is a different, and much harder, task.

    Posted May 27, 2009 at 5:23 pm | Permalink
  13. Bill Easterly wrote:

    Avinash, you are of course correct about using all the data. I was illustrating the fallacy of reversing conditional probabilities by focusing on the “IF industrial policy” probability. When you address the probability “If you do industrial policy, what is chance of success?” you would also have to compare this to the probability in the control group, namely “If you don’t do industrial policy, what is chance of success”, and test whether one is significantly higher than the other, which your test would calculate. Thanks for your comments, Bill

    Posted June 10, 2009 at 10:53 am | Permalink
  14. I am a Dani fan too. Thank you for the post anyway.

    Posted September 17, 2009 at 2:29 am | Permalink
  15. Gold wrote:

    According to the first stanza, are Muslims being feared?

    Posted September 18, 2009 at 3:35 am | Permalink
  • About Aid Watch

    The Aid Watch blog is a project of New York University's Development Research Institute (DRI). This blog is principally written by William Easterly, author of "The Elusive Quest for Growth: Economists' Adventures and Misadventures in the Tropics" and "The White Man's Burden: Why the West's Efforts to Aid the Rest Have Done So Much Ill and So Little Good," and Professor of Economics at NYU. It is co-written by Laura Freschi and by occasional guest bloggers. Our work is based on the idea that more aid will reach the poor the more people are watching aid.

    "Conscience is the inner voice that warns us somebody may be looking." - H.L. Mencken

  • Archives