Dani Rodrik: doing industrial policy right

Good morning. Traders are braced for today’s annual revisions to the consumer price index’s seasonal factors. Some think it could affect stocks, perhaps by making inflation look somewhat scarier. We at Unhedged are in disbelief that seasonal adjustment methodology could move markets, and are instead braced for the weekend. Email us: robert.armstrong@ft.com and ethan.wu@ft.com.

Friday interview: Dani Rodrik

Dani Rodrik, the Harvard economist, is best known for his early critiques of unfettered globalisation. But he has also long been one of the economics profession’s most careful proponents of industrial policy, the deliberate government steering of industry. Just as these sorts of interventions were coming into vogue in the US and elsewhere, Rodrik and his co-authors published an influential August 2023 paper called The New Economics of Industrial Policy. Below, Rodrik discusses what the old economics of industrial policy gets wrong, lessons from east Asia and Joe Biden’s record.

Unhedged: The conventional wisdom about industrial policy says it’s inefficient and vulnerable to both rent-seeking and regulatory capture. Maybe the space race works, but for the most part “picking winners” isn’t a good idea. What does that miss?

Dani Rodrik: I don’t think the evidence justifies such a broad-brush rejection of industrial policy. Industrial policy is just like any other policy. Yes, it can be inefficient. It can be captured. But that’s not unique. Education policy can be captured by powerful teachers’ unions. Macroeconomic policy can be captured by financial interests. Infrastructure or health policy can be captured by special interests. So that sort of broad-brush generalisation isn’t based on either the economics of industrial policy or the evidence. I think it’s more an ideological predisposition about the role of government. 

Unhedged: But aren’t some of those, like health policy, about providing public goods? The pushback here might be that industrial policy is stepping into the domain of private markets.

Rodrik: That’s an excellent question. The economic rationale for industrial policy is based on the internalisation of externalities: capturing learning and innovation externalities, offsetting externalities having to do with the disappearance of local jobs, the externalities of the green transition — environmental or carbon externalities. A lot of industrial policies are interventions to provide firms with customised services, such as in workforce training, infrastructure or land. These are all public goods. 

So whether it’s fixing externalities or providing public goods, the economic rationale for industrial policy is very strong. No growth economist or development economist or labour market economist would deny the prevalence and significance of these market failures that result in economic underperformance. So the debate is not really about whether the underlying rationale is solid or not. The debate has always been in practice about loose political economy arguments — about claims like “the government can’t pick winners”, or “governments get captured”. The rationale for industrial policy is really no different from the rationale for [interventions in] health or macroeconomic stabilisation or education. 

Unhedged: What does the evidence say about what kind of industrial policy works and what kind does not? 

Rodrik: Quite frankly, the evidence has not been able to tease out those distinctions. We have a fair amount of evidence that shows that industrial policy has worked in a number of settings, ranging from regional policies in Europe to R&D policies in the United States to export promotion and industrial diversification policies in South Korea and China. But has the literature gotten to the point where it tells us exactly why it worked in those settings, where in others it didn’t? I don’t think the evidence is strong enough to say for sure. I have my own views. But they are largely based on anecdotal and speculative evidence rather than systematic evidence. 

Unhedged: Happily, you’re talking to journalists. We love anecdotes and speculation.

Rodrik: Earlier I said people are making broad generalisations about industrial policy not working, so I want to be careful about what the evidence shows and doesn’t show. If you ask my high-level opinion about what successful industrial policy requires, it’s a few things. 

One is being clear-sighted about what the objectives are. The more things you try to achieve, the less likely you are to get them. So if you have a single-minded, regional policy focused on employment generation, you can do that; if you have a single-minded policy of generating productive capacity in some export-oriented manufacturing sector, you can do that. But if you’re trying to create jobs and further the green transition and promote innovation all at once, you’re likely to miss some of the targets. 

Second, industrial policy cannot be formulated and implemented in a top-down, arm’s-length manner. You can’t just leave it to bureaucrats and policymakers. Successful industrial policy typically operates in what a sociologist would call an “embedded” manner: the policymaking process is co-ordinated around information moving between the private sector, policy entrepreneurs and other local stakeholders. You need to base policy on input, information, iteration and learning. You have to practise industrial policy in a way where the government is constantly interacting with the private sector to understand where the opportunities are. Otherwise, it suffers from a lack of information.

Third, it requires a certain amount of government discipline. Not “discipline” in the sense that the government must learn to pick winners. I don’t think any government can systematically pick winners. But neither can the private sector. The kind of discipline that’s required is the discipline of monitoring, figuring out whether what you’re doing is working, and being able to move away from mistakes when things aren’t working. Successful industrial policy is not about picking winners, it’s about letting the losers go. Some of the worst cases of industrial policy are when you keep putting good money after bad.

Unhedged: I want to ask you about methodology. You’ve made the point that one of the problems with the industrial policy literature is it tries to do systematic evaluations of non-random phenomena. That makes it hard to tell how effective the “average” industrial policy is. How does the best work get around that? 

Rodrik: Until recently, if you were to ask the sceptics for evidence, they might point to papers that use essentially correlational approaches. [They might] correlate the degree of intervention across Japanese or South Korean industries with some measure of productivity or exports and so forth, and notice that there was either no correlation or sometimes even negative correlation.

Those correlational studies have a fundamental problem, which is not well recognised: they don’t specify an appropriate counterfactual. When an industry receives support, we want to know how that industry would’ve done in the absence of that intervention — not compared to other industries that might not have faced similar challenges. 

I think the new literature is much more careful about drawing causal inference. It doesn’t entirely avoid the problem, though, because it’s difficult to make statements about essentially non-random phenomena just exploiting random variation in the data. But what you can exploit is, for example, historical accidents. Trade gets cut off for reasons having nothing to do with economic policies, such as a war. Or you look at discontinuities in the data, such as some firms receiving help while other firms don’t, for reasons not having to do with policy. From that kind of evidence, we can see, for example, that trade restrictions caused by trade blockades produce lasting changes in the structure of economic production, in a way mimicking the effect of an import substitution policy.

Unhedged: Can you give us an example?

Rodrik: The war example I mentioned comes from a great paper by Réka Juhász looking at the effects of the Napoleonic blockade on French industry. There’s also a nice paper on local industrial policies in Britain, which we don’t think of as a country with a successful history of industrial policy. The paper exploits differences in EU rules about what kind of regions were eligible for subsidies — disadvantaged regions could be subsidised. But there were changes in the rules as to which regions counted as disadvantaged, which from the perspective of the British government were reasonably exogenous. The authors show, using that exogenous variation in subsidies, that there were significant investment and employment benefits, particularly among smaller firms, and the cost per job was quite modest.

Unhedged: The East Asian economic miracle gets a lot of attention. It’s what a lot of people think of when they think of industrial policy. How was East Asian industrial policy rolled out, and what were the hallmarks of its success?

Rodrik: Essentially all the successful countries in east Asia, with the exception of Hong Kong, used a very heavy dose of industrial policy. The burden of proof is on the sceptic to show it did not help those countries. But leaving that aside, industrial policies differed significantly across those countries, and also changed over time. 

To take one country, South Korea, there were roughly two different periods. In the early 1960s through the mid-1970s, industrial policy took the form of heavy credit subsidies to firms, in exchange for export commitments. These were essentially export subsidies. The firms were very closely monitored by the government to ensure that their export targets were being met, that they were making the requisite investments, and so forth. We can see all the good design elements I talked about earlier — a clear and pointed focus on new export industries, close interaction with firms, and close monitoring to see that exports and investments were being undertaken. If they weren’t, firms would be disciplined; a tax inspector might show up at your door to have a closer look at your accounts. 

The second important period is covered in a paper by one of my co-authors, Nathan Lane. That period in the late 1970s focused on heavy chemical industries (HCI). The government was explicitly targeting capital-intensive upstream industries, where South Korea was thought to not have a comparative advantage. Some thought export success would be harder to generate in HCI, because it was unlike the labour-intensive light industries successfully developed in prior decades. Many people presumed the HCI period was not very successful, because the immediate benefits to exports were not closely linked to subsidy policies. What the Nathan Lane paper shows is these industries had strong spillovers through linkages to downstream industries. It shows that the industries that were heavy beneficiaries of industrial policy expanded investment and output quite significantly, if you take the whole supply chain into account.

Unhedged: In China, there’s been heavy credit subsidisation of manufacture and export industries, which for a long time generated an incredible amount of growth. But you could also argue that it has left the country with a very unbalanced economy and financial system. What lessons can we draw from the subsidisation of export industries in China? 

Rodrik: It’s a mixed picture, but given the incredible economic transformation of that country, you have to say that the whole complex of policies was successful. Not just for China’s own sake — for the sake of poverty reduction and employment generation — but also for the rest of the world. It created a much richer Chinese economy, and a much larger market for investors and exporters. 

Of course, many of the specific policies were controversial, inflicting commercial costs on specific firms and investors. And, of course, Chinese trade imbalances have long been a problem. I don’t want to minimise the tensions that this strategy has created. But on balance, it’s been great for China and a good thing for the rest of the world.

In terms of subsidies, Chinese industrial policies have been very, very different. Most distinctively, a lot of it takes place at the regional and provincial level. The central government articulates certain priorities, such as electric vehicles or batteries, and then the provinces go on competing in stimulating local ecosystems. And in the end, often you do get overinvestment. China’s industrial policies have often been characterised by cycles of too much subsidy competition by provincial governments, forcing the national government to step in and consolidate the industry. It hearkens back to my point earlier about the need for government discipline.

From my experience in studying developing countries, I’d rather commit the error of over-promoting my industries and overinvesting in structural transformation in more productive areas than under-commit. 

Unhedged: How would you rate Joe Biden’s industrial policy push so far? 

Rodrik: The United States never abandoned industrial policy. It’s always been going on at the local level and also at the federal level, driven by defence-related needs and more recently by the Department of Energy. What’s different with Biden is that there’s been an explicit articulation of industrial policy as an essential component of federal government policy. So what used to fly under the radar, sometimes causing embarrassment with trade partners, has become something we’re shouting from the rooftops. 

On the whole, I like [Biden’s] policies. I think they will achieve some of their important aims, but there’s also some parts where they’re likely to fall short. I like the Inflation Reduction Act on the whole, because nothing you do to speed up the green transition at this point can be wrong. That’s a fundamental set of new investment priorities, and the early results are encouraging. In terms of the Chips Act and other investments in semiconductors, given the geopolitical goals of the United States, I do think they will produce some important benefits. 

Where I’m a little bit less enthusiastic is the additional objective of creating good jobs and of addressing the needs of lagging regions that gets tacked on to industrial policy. There’s a kind of disconnect between the design of these policies and the objective of good jobs. The reality is that advanced semiconductors don’t create many jobs. The green transition will create jobs in some places, but also will result in job losses in other places. And then we have the heritage of decades of erosion of good middle-class jobs in the US, which these policies do not directly target. 

Unhedged: But perhaps that’s a political inevitability, right? You have to have the word “jobs” in big letters on any policy in the US.

Rodrik: I think that’s right, and there’s another political inevitability. The Biden administration does have specific programmes that target good jobs or lagging regions. But the budget for them is like $100mn here, another few hundred million over there — orders of magnitude smaller than what is being spent on the IRA or Chips Act. The Biden administration could expand things like the Good Jobs Challenge run by the Department of Commerce, which I really like and think should be much, much larger. 

But it would run into opposition by Republicans. So where you can get agreement with congressional Republicans is when you’re addressing these geopolitical issues and framing these programmes as addressing competition with China. That’s the political reason why the Biden administration cannot move more frontally on the good jobs issue. 

Unhedged: Past instances of US industrial policy are often criticised as wasteful, like the stimulus programme in 2009 lending to failed rooftop solar start-up Solyndra. How do you think about these sorts of boondoggles?

Rodrik: Industrial policies have to be evaluated as a portfolio, just like venture capital. The typical rule is that a small number of successful investments will pay for all the failures. That’s by and large true of past programmes, of which Solyndra is an example. A very similar programme [as lent to Solyndra] was also responsible for Tesla’s survival, which received loans from the Department of Energy at a very critical point. Without that, it might have gone under, and not become the behemoth that it is today. 

Now, Solyndra was a case of mismanagement, because it violated one of the good design principles that I articulated earlier. The government kept supporting Solyndra long after it became evident that the specific technology it was using wasn’t going to pay off. Not only that, the government was fixated on making Solyndra a political showcase for the wider programme. So when Solyndra fails, it looks like the whole programme has failed.

The lesson from Solyndra is very important: don’t stick with the losers. And politically, don’t sell the programme on the basis of individual projects. You have to sell the programme as a portfolio, and be very explicit that you’re not expecting everything to be a roaring success. It’s not going to work out that way. 

One good read

The “enshittification” of Big Tech.

FT Unhedged podcast

Can’t get enough of Unhedged? Listen to our new podcast, hosted by Ethan Wu and Katie Martin, for a 15-minute dive into the latest markets news and financial headlines, twice a week. Catch up on past editions of the newsletter here.