The Bairoch conjecture on tariffs & growth

{ Note: This post describes and summarises a literature on 19th century growth & trade. I do not necessarily endorse its findings. This post is intended as largely descriptive. }

There is a vast cross-country literature which finds a positive correlation between economic growth and various measures of openness to international trade in the post-1945 period. Despite intense methodological bickering amongst researchers, nonetheless maybe 50 studies (maybe more?), using a variety of methods and approaches, come to the same conclusion: trade openness was associated with growth after 1945. (This amazing critical survey lists most of those studies.)

This huge body of research does have some quite compelling critics, the most prominent being Rodríguez & Rodrik (2000). This widely cited paper argues — amongst many other things — that there is no necessary relationship between trade and growth, either way. It depends on the global context as well as domestic economic conditions. I think that view is correct.

There is also a smaller literature on the “19th century growth-tariff paradox” associated with the historian Paul Bairoch. He argued informally that European countries with higher tariffs grew faster in the half century before the Great War.

Bairoch’s rough eyeball correlation was confirmed econometrically by O’Rourke (2000) for a sample of 10 rich countries (Australia, Canada, Denmark, France, Germany, Italy, Norway, Sweden, the UK, and the USA) in the period 1875-1913. This finding was supported by several other studies, including Clemens & Williamson (2001, 2004), but was disputed by Irwin (2002) on the grounds that the correlation was driven by rapidly growing settler economies with high land-labour ratios which relied on tariffs for revenue.

Lehmann & O’Rourke (2008, 2011) then countered by disaggregating tariffs of those 10 rich countries into revenue, agricultural, and industrial components, reporting that duties specifically protecting the manufacturing sector were indeed correlated with growth.

But these 19th century studies take the 20th century findings as a valid point of departure — there is a contrast between the positive tariff-growth correlation for 1870-1914 and the negative correlation after 1945. The ‘paradox’ is therefore in keeping with the criticisms of Rodriguez & Rodrik (2000), one of whose major points is that there is not, even in principle, any necessary relationship between openness and growth. It depends on the global environment, domestic conditions, complementary domestic non-trade policies, what actually gets protected, etc.

Yet even the small 19th century literature is mixed — more mixed than the 20th century literature. Clemens & Williamson (2001, 2004) confirms the overall positive correlation between tariffs and growth found by O’Rourke (2000). But C&W also tests the proposition with a larger sample of 35 countries in 1870-1914 that includes many from the poor periphery. The positive growth-tariff relationship for the rich countries is large; much smaller for the non-European periphery, and negative for the European periphery (e.g., Spain, Russia, etc.) So obviously even with the same global conditions there’s a lot of heterogeneity.

According to Clemens & Williamson (2001, 2004) the reason there was an overall positive correlation in the 19th century, may be that countries with higher tariffs tended to export to countries with lower tariffs:

“[E]very non-core region faced lower tariff rates in their main export markets than they themselves erected against competitors in their own markets. The explanation, of course, is that the main export markets were located in the Core, where tariffs were much lower.”

This is how I personally interpret it: Great Britain and others acted as free-trade sinks (my phrase, not C&W’s) for exporting countries such as the United States (and Wilhelmine Germany) which protected their steel and other industries. (Echoes of East Asia which benefited from US policy during the Cold War? It’s nice if there are countries willing to indulge your export-led development strategy without reciprocal openness.)

( Edit: Yes, yes, yes, as many have pointed out on Twitter, Britain in the 19th century famously settled its trade deficits with America, Canada, Europe, etc. through its surpluses with India, etc.

But as you can see in the graphic above, it’s a little bit more complicated: India also had surpluses with the USA, Japan, and Europe. )

(Clemens & Williamson appeal to a prisoner’s dilemma model in which trade coordination between two countries is the best outcome, but if one country does defect, i.e., imposes tariffs, then the other is better off retaliating. So in the 19th century, the non-retaliating party might have been worse off in terms of growth.)

The correlation reverses after 1945 because “tariff barriers faced by the average exporting country have fallen to their lowest levels in a century-and-a-half”, and rich countries in particular were much more open. So the international environment does matter for the relationship between trade and growth.

Jacks (2006) — using the Frankel-Romer gravity model approach — both replicates the positive correlation between growth & tariffs, and supports the free-trade-sink view. The reason higher-tariff countries grew faster was that “an increase of 1 percentage point in the level of tariffs led to an increase of roughly 0.7 percentage points in the balance-of-trade to GDP ratio”. In other words, the more protectionist countries could generate trade surpluses (which add to GDP), apparently because the more free-trading countries did not retaliate against them. Again, the global environment matters.

Tena-Junguito (2010) focuses on industrial tariffs and supports the other aspect of the Clemens & Williamson finding: the tariff-growth correlation applies only to the “rich country club”. For Latin America and the European periphery, the correlation was negative. Unfortunately this is a semi-cross-sectional view, relating tariffs in 1875 with cumulative growth in GDP per capita between 1875-1913, which obviously misses the change over time in tariff schedules after 1875. Most other studies use panel data, relating 5-year chunks of growth rates and average tariffs.

On the other hand, Schularick & Solomou (2011) find no evidence for the tariff-growth correlation for a sample of 30 countries, rich and poor. It turns out to be spurious once you control for the business cycle (which was transmitted internationally via the gold standard). The time trend which drives the tariff-growth relationship is apparently the 1875-79 depression. Countries became more protectionist during the recession, but the higher tariffs remained in place after the global economy recovered, driving the spurious result.

However, Schularick & Solomou do not look specifically at manufacturing tariffs, so their findings do not overturn Lehmann & O’Rourke (2008, 2011), which did find a correlation between manufacturing tariffs and overall growth for the rich countries. It’s possible Lehmann & O’Rourke’s results might be duplicated for a larger sample which includes poor countries.

Personally I find that unlikely because, if anything, the duties levied by primary commodities-exporting countries such as those in Latin America would surely have been on manufactured goods.

In the final analysis, we shouldn’t make too much out of this literature, either way. Cross-country regressions — especially in a sample ranging from 10 to 30 countries — are a pretty crude and blunt tool for assessing infant-industry arguments.

Edit: I prefer single-country examinations of tariffs and development. For the USA in 1870-1913, some great examples are Yoon; DeLong; and Irwin, all of which argue that tariff protection likely did not play an important role in US economic development in the post-civil war period. Also see my post on the Napoleonic blockade & the infant industry argument.

Addendum: One major reason (amongst many many!) that cross-country regressions are a poor tool for assessing the infant industry argument is that many industries are often protected for the ‘bad’ reasons. (But Nunn & Trefler address this issue successfully in my opinion.)

Another major reason is this: it’s actually quite easy for poor countries to temporarily increase growth rates “through protectionism” as long as you can do technological upgrading by importing it from more advanced countries. If tariff policy or state subsidies distort investment away from agriculture and toward industry by increasing the rates of return in manufacturing, you will get structural transformation (movement of labour from lower-productivity to higher productivity sectors).

Traditional agriculture is so unproductive and so full of underemployed labour that any diversion of resources toward any ‘modern’ sector can raise growth rates, all else equal. If a government decided to import some machines, close off the country to global trade, round up peasants at gun point, and force them to work in factories, you will get growth. If this were not true, Stalinist industrialisation would have been impossible ! (I ignore the welfare considerations, of course.)

In fact, even if productivity growth were zero in both the traditional and modern sectors, you can still get positive economy-wide productivity growth just by moving resources out of the traditional sector and into the ‘modern’ sector. All it takes is movement out of one stagnant sector to another stagnant (but ‘better’) sector. You can continue with this process until you run out of peasants — or run out of idle labour in agriculture (because, at some point, agriculture itself will need productivity growth to release more labour — unless you start importing food, in which case you will need to start export something).

Postscript: By the way, the spectacular historical vulgarian Ha Joon Chang has a habit of cherry-picking from this literature. HJC gleefully cites any study, especially O’Rourke (2000), which confirms his priors, but fails to report nuanced or inconsistent results.

Of course, I do not fault Chang for failing to reference research which did not exist at the time of writing his popular Kicking Away the Ladder (2002), but he doesn’t cite any of the follow-ups in Bad Samaritans (2007), either. And every time he writes an article, he always appeals to the same references and fails to cite stuff inconsistent with his priors (e.g., HJC 2010 and 2013, as well as Chang’s reply to Easterly’s review of Bad Samaritans.)

To the best of my knowledge the only time he has come close to nuance is in the 2010 proceedings of the Annual World Bank Conference on Development Economics, but the modicum of subtlety is relegated to the footnotes:

At least for the 1870–1913 period, there is even evidence of a positive correlation between tariff rate and rate of growth (O’Rourke 2000; Vamvakidis 2002; Clemens and Williamson 2004)”

“5. Irwin (2002) argues that this correlation was driven by high tariffs imposed for revenue reasons in the New World countries (the United States, Canada, and Argentina in his sample) that were growing quickly for other reasons (such as rich natural resource endowments). However, the United States was the home of infant industry protection at the time, and many of its tariffs were not for revenue reasons. Moreover, O’Rourke (2000) and Lehmann and O’Rourke (2008) show that the positive tariff-growth statistical correlation is not driven primarily by the New World countries.”

“6. Clemens and Williamson (2001) argue, on the basis of an econometric analysis, that around a third of this growth differential between Asia and Latin America during 1870–1913 can be explained by the differences in tariff autonomy.”

Notice that even when HJC does mention Clemens & Williamson (2001), he omits details which do not suit his priors !

Chang (2005, 2010, 2013) loves to cite Rodríguez-Rodrik (2000) and argues against the feasibility of econometrically validating any trade-growth relationship. But then he cites those selective bits of the econometric literature on the Bairoch conjecture anyway, and all the time! Apparently the Rodríguez-Rodrik warnings about the non-universality, historical contingency, and context-specificity of the trade-growth relationship apply only to the 20th century findings!

Edit 26 December 2016: See Vincent Geloso’s remarks on some of the papers mentioned in this post. He’s also posted below in the comments section.