Imagine that you’re talking to a friend, and you mention that smoking shortens average life expectancy, and that smokers should therefore be encouraged to quit. Suppose your friend replies, “Well, but my uncle Fred smoked every day, and he lived into his 80s.” If your friend means this either (a) as a serious challenge to your empirical claim that smoking is bad for you, or (b) as a critique of your prescriptive argument that smokers should therefore be encouraged to quit, then you would probably find his response absurd on its face. And if your friend were to say that he has posed a serious conceptual conundrum—say he calls it the “Uncle Fred Paradox”—you would probably laugh at him. His argument might seem marginally less ridiculous if he pointed not to his Uncle Fred but to, say, France—which has relatively high smoking rates and relatively high life expectancy—but we probably still wouldn’t view this as a serious challenge to the view that smoking is bad for you, nor would we spend a lot of time wringing our hands worrying about the “France Paradox” in the smoking-health relationship.
Yet for some reason, in serious discussions about the relationship between corruption and economic development, people seem to make precisely this sort of specious argument, and the argument gets taken very seriously by people who should know better. The form the argument takes in this context goes something like this: “It may be true that high corruption seems to be correlated with lower levels of economic development on average. However, many countries in East and Southeast Asia—such as China, South Korea, Japan, Taiwan, Thailand, and Indonesia—either achieved or currently are achieving impressively rapid economic growth despite widespread corruption.” This is the so-called “East Asian Paradox” (a term coined, as far as I know, by Professor Andrew Wedeman — see also his recent book). The somewhat more sophisticated version of the argument, developed most prominently in an article by Professor Michael Rock and Heidi Bonnett, notes that although perceived corruption has a negative relationship with growth and investment in most countries (especially small developing countries), this relationship becomes positive in a subsample consisting of five large, newly-industrializing Asian countries (China, Indonesia, South Korea, Thailand, and Japan), using data drawn from the early 1980s through the mid-1990s.
One encounters more-or-less sophisticated versions of the “East Asian Paradox” argument all the time when talking about the adverse impact of corruption on development. When someone says something like, “Corruption is a major threat to economic development,” someone almost invariably responds with something like, “But what about China? It has achieved impressive economic growth despite widespread corruption.” As far as I’m concerned, this is equivalent to saying, “But what about my Uncle Fred, the lifelong smoker who lived into his 80s?” But in case this is not completely obvious, let me explain why I think the “East Asian Paradox” argument, at least in its usual crude form, is mostly bogus.
- First, this may be a banal truism, but the world is a complicated place, development is a complicated process, and lots of different factors affect economic performance. The claim that corruption is bad for development is not some kind of deterministic statement that countries with high corruption will always realize bad development outcomes, any more than the claim that smoking shortens life expectancy is an assertion that smokers will always die younger than non-smokers.
- Second, and closely related to the first point: What we care about, when asking about the impact of corruption (or any other proposed explanatory variable) on development is the comparison of what actually happened to what would have happened if that variable had taken a different value (say, if corruption had been higher or lower). We don’t care how long your pack-a-day-smoking uncle lived, so much as we care about how long he would have lived if he hadn’t smoked. We don’t care about China’s current level or pace of development so much as we care about what that level or pace would have been if China’s corruption problem were much less (or much more) severe. Because we can never observe that counterfactual, answering the latter type of question is incredibly difficult in practice, at least in a context where randomized trials are impossible. But conceptually, the point should be clear.
- Third, the fact that several of the countries that appear to defy the usual correlation of high corruption with poor development outcomes are concentrated in a particular region (East/Southeast Asia) may be a statistical illusion. Consider the following thought experiment. Suppose you have 100 countries sorted, completely arbitrarily, into ten “regions,” with ten countries per region; suppose each country has a 50% chance of being a “high corruption” country and a 50% chance of being a “low corruption” country. Suppose that countries in the former group have a 60% chance of being “high development” and a 40% chance of being “low development”; for countries in the latter group the probabilities are reversed. In expectation, each region would have three low-corruption/high-development countries, three high-corruption/low-development countries, two high-corruption/high-development countries, and two low-corruption/low-development countries. But because of the inherent randomness of the process, there’s no guarantee that each region will have exactly this distribution—indeed, it’s extremely unlikely that such a uniform distribution would occur. It’s quite possible that in at least one region, there will be more high-corruption/high-development countries and low-corruption/low development countries than there are high-corruption/low-development and low-corruption/high-development countries. If that occurs, would we say that this region presents a difficult paradox? Would we need to undertake a deep, searching inquiry into the special culture and historical circumstances of, say, arbitrarily-designated “Region 6”? Clearly not — the suggestion is plainly silly. It’s never been clear to me that the alleged “East Asian Paradox” is all that different.
- Fourth, another possible explanation for the apparent East Asian Paradox has to do with the decisions as to which countries to include, and which to exclude, when performing the analysis. I’ll continue to focus on the Rock & Bonnet study, since that’s the best one making the case for an East Asian Paradox. As noted earlier, their subsample—the one that creates the paradox—is limited to five large, newly-industrializing (though at different points), East Asian economies: China, Japan, South Korea, Indonesia, and Thailand. Why just those countries? If the claim is that this has something to do with the region (and even if one wants to exclude the city-states Hong Kong and Singapore), why not include Malaysia, North Korea, Myanmar, Vietnam, Cambodia, Laos, the Philippines, Taiwan, and Mongolia? (Even if we’re using size as a cutoff – and I’m not sure why we are – Vietnam and the Philippines have larger populations than Thailand and South Korea.) And why are we limiting the sample to “newly-industrializing” countries? Doesn’t that risk introducing bias (if, for example, some of the “non-industrializing” economies are not industrializing because of excessive corruption)? And why exactly are we counting Japan and South Korea as “newly industrializing” as of the 1980s and 1990s?
- One more thing I think deserves mention. The Rock & Bonnett study – as I’ve said, the most sophisticated attempt to statistically establish the alleged East Asian Paradox of which I’m aware – uses a dataset that stops in 1996. And the five countries that produce the paradox – the countries where we find a positive rather than a negative relationship between perceived corruption and growth – are Japan, South Korea, China, Thailand, and Indonesia. Note the last two countries on that list—both of which were rated as high-corruption, and both of which were experiencing very high growth rates up through 1996. Did anything happen in those two countries, oh, I don’t know, maybe a year later? Perhaps something like a devastating financial crisis? Perhaps a financial crisis that, while it had numerous complex causes, may have been at least exacerbated by cronyism and conflicts of interest? Would the relationship that Rock & Bonnett find hold up if we used data from 1998 instead of 1996? I don’t know, and perhaps someone out there with the time and inclinations could try to replicate the Rock & Bonnett study using different and later time periods, and a larger sample of countries. But I’m a little skeptical of a conclusion that says, “Hey, in East Asia high corruption countries have no serious economic problems! Look at Thailand and Indonesia in 1996!” Much like the Titanic didn’t have any buoyancy problems before it hit the iceberg. (Even if corruption played no causal role in the crisis, it is still the case that much of the pre-1997 rapid growth in Southeast Asia was the result of a bubble, which subsequently burst — meaning that using those years of rapid growth as a benchmark may be highly misleading.)
Now, I want to be clear about a couple of things in closing. First, I’ve been assuming for the sake of argument that the statistical data shows a clear, robust, negative correlation between corruption and economic growth. However, although the negative correlation between perceived corruption and per capita income, and between corruption and a range of other development outcomes, is quite well-established, the evidence of a negative impact of corruption on growth is much more equivocal. I think the weight of the evidence does suggest that corruption has adverse growth effects, but I have to concede that the data here are not as clear. (And, indeed, I have criticized some of the leading studies that in my view overstate the evidence for the link.) The evidence may eventually undermine the claim that corruption is bad for growth—but that doesn’t change the fact that one counter-example, or a set of counter-examples that happen to be clustered in one geographic region, does not provide much of a refutation of the statistical evidence. Second, I don’t want to dismiss the idea that the impact of corruption on development is complex, that it is mediated through a range of formal and informal norms and practices, or that different forms of corruption may have quite different implications for economic development. Indeed, I don’t even want to dismiss the idea that there might – might – be something special and different about the East Asian economies. And some qualitative case studies have indeed produced some interesting and valuable evidence that the typically adverse impact of corruption may often be mitigated by other factors. But I haven’t yet seen any convincing statistical evidence that this is so as a general matter in any particular region, and the claims of an East Asian Paradox that I’ve encountered do not seem to me much more convincing than the claim that smoking isn’t necessarily bad for you because of someone’s Uncle Fred.