TL;DR: the Canadian National Policy of 1879-1895 was a successful instance of deliberate infant industry protection (IIP). Selective targeting of industries with close domestic substitutes led to rapid output expansion, productivity growth, and price declines with little cost to consumer welfare. Nineteenth-century Canada is a good example of how a heavily agricultural, resource-dependent economy can use trade policy to develop a native manufacturing sector despite pressures toward primary product specialization. Main points:
Canadian protectionism worked -- it created internationally competitive industries at little cost to the domestic economy and forced US manufacturing firms to relocate inside the tariff wall.
It's possible that the result is idiosyncratic to the Belle Epoque world, in that Britain served as a "free trade sink" for Canada's export-facing economy.
IIP did not promote overall structural transformation, but mostly because it had to keep up with resource booms in the mineral and agricultural sectors.
Protection did not make Canada a less resource-dependent economy, but rather succeeded in onshoring more raw material processing
Two special announcements before we dive in. First, I’ve renamed (well, actually just removed the old placeholder title) the newsletter Great Transformations, which contains at least two book references and sort of reflects the general topics that I like to write about.
Second, I’ve added a paid subscription feature so that you can contribute to this project. The essays are staying in front of the paywall, but you’ll get a bit of additional content, including some enormous reading lists that I’m curating and weekly research posts. Maybe some other fun things in the future. The main idea, though, is that I’d like to derive some income from this project to make the time commitment worthwhile. If you can find the funds to support me, I’d really appreciate it.
Also, since this is the only time I’m asking for money in the body of the newsletter, I’d be grateful if you shared around this particular edition.
A quick note: if you are receiving this post for the second time, I apologize. Substack failed to send out this issue originally and is not responding to my support requests, so I’ve decided to cut my losses and re-send it.
The basic idea behind infant industry protection is that you, as a nascent manufacturer, need the state to tax international competitors until you reach an efficient scale and productivity level. It’s a seductive argument. Who doesn’t like protecting infants? But there aren’t many great empirical studies that show it working, and there is a lot of anecdotal evidence for it failing catastrophically.
In my last post, I used that argument in making the case that Russian economic growth prior to World War I was sustainable. I contested Bob Allen's view that Russia's natural resource exports and protectionism doomed her to a "Latin American" development path. One thing I suggested was that Canada, which both had a "wheat boom" and hiked tariffs at about the same time as Russia, might present a better counterfactual model.
What I didn't know for certain, however, was whether Canadian protection really worked. So I decided to figure that out.
Canada was a relative latecomer to protection. In the US, a fractious Republican Congress raised average tariff rates from 15 percent in 1859 to 45 percent in 1870, and per Irwin (2010) they remained above 30 percent through 1890. Strenuous opposition by the Maritime provinces (New Brunswick, Nova Scotia, and Prince Edward Island), good economic performance, and politically marginal manufacturers coupled with a revenue-based interpretation of tariffs' role meant that little change in the tariff structure occurred up to 1879. Raw materials and semi-finished intermediate inputs entered the country duty-free, while a 15 percent ad valorem tax was placed on manufactured imports, raised slightly to 17.5 percent in 1874 (the Mackenzie Tariff) for revenue purposes.
In 1873, however, global markets slumped and by 1876 an outright "depression" had set in. By 1879, prices had dropped by 20 percent and timber exports by half. Bank failures spiked and government revenues tumbled. After winning the election of 1878, the Conservative Party, led by Prime Minister John A. Macdonald, launched the "National Policy," a development program designed to expand and integrate the Canadian economy. The National Policy had three main planks: the construction of the Canadian Pacific Railroad (completed in 1885); the encouragement of immigration to offset emigration to the US and settle the West; and the adoption of a strongly protectionist tariff regime to bolster domestic industry. I'm writing about the third component, and treat "National Policy" and "protectionism" synonymously. I include both the initial 1879 hike, when average tariff rates weighted by import value rose immediately from under 14 percent to 21 percent, and the Tupper Tariff of 1887.
There were three compelling motivations for the National Policy. First, higher tariffs offered a way to augment the country's revenues, of which duties constituted 70 percent until 1887. Second, the depression of the '70s had hammered domestic industry, with manufactured prices falling further than raw materials. Industrialists' fears were exacerbated by falling transport costs, which rendered producers more vulnerable to foreign competition, and the failure to win reciprocal openness from the US. The Conservatives' main supporters were the industrialists and city businessmen who'd suffered most from the trade depression and would benefit most from protection. Finally, the National Policy appears to have been designed to "balance" Canadian development by making the country more self-sufficient in manufacturing, which would serve an East-West domestic market populated by immigrants, integrated by railroads, and protected by high tariffs.
There was high variation in the tariff structure, though. For non-manufactured raw materials, tariffs only increased from 5.6 to 9.5 percent, and those on exotics remained virtually stable at 38-9 percent. Duties on cotton textiles inicreased from 17.5 to 30 percent and primary iron and steel from 0-5 to 12.5-17.5 percent. Beaulieu and Cherniwchan (2014) derive trade restrictiveness indexes, or the general tariff level that produces the same welfare loss as the actual tariff schedule. As the narrative evidence suggests, goods with close domestic substitutes—i.e. which Canadians were already making—were primary targets. So the Tupper tariff, which strategically increased rates on specific products, actually boosted restrictions by about the same amount as the original National Policy.
Did Canadian protectionism work? There are basically three questions here:
How much of a welfare loss did protectionism cause?
Did protectionism promote long-term industrial growth?
Did (2) outweigh (1)?
I'll take the answers in turn.
Welfare Costs
Early evaluations of the National Policy, especially by Canadians, tended to be positive. A select committee ordered by the government to "inquire into the operation of the tariff" sent two agents, Alfred H. Blackeby and Edward Willis, to survey hundreds of manufacturing firms in Ontario, Quebec, and the Maritime provinces. Blackeby reported:
The value of the product in money does not represent the whole of the increase which has taken place ... Prices are so much lower now in most cases that a like production in value would mean a 10 to 15 percent difference in bulk ... Production has increased in a greater ratio than the number of hands, showing by means of better appliances and facilities brought into use, by reason of the larger trade to be done, each man produces more now than six years ago ... That the general result of that change [the 1879 tariff] has proved decidedly beneficial to manufacturing industries there is now no dispute, and is fully borne out by the figures herewith submitted (Canada, House of Commons, 1883: 4, 6, quoted in Harris et al. 2015).
In their classic textbook, Canadian Economic History, Easterbrook and Aitken argue that the National Policy was a "necessary evil." Protectionism augmented the market power and profits of manufacturers at the expense of consumers, but successfully sheltered infant industry, drove import substitution, and increased the scale of production. Aitken also saw the tariff, which raised aggregate labor demand and promoted immigration, as part of a "defensive expansionism" program designed to make Canada too big to swallow for the US.
Soon afterward, however, a series of revisionist studies of the Cliometric Revolution era set out to debunk the traditional view. On the basis of a neoclassical trade model, the Canadian economist John Dales (1966: 109–10) called the tariff “the price we pay for our protected manufacturing industry," arguing that "protection fosters inefficient, oligopolistic forms of market organization in Canada" and that "Canadian growth has been distorted by the National Policy.” Worse, he thought that the tariff encouraged the emigration of skilled workers and the immigration of unskilled labor, increasing GNP and lowering wages.
Easton et al. (1988) tested the Dales model and found that the extent to which the tariff promoted immigration was contingent on its welfare consequences (unemployment, real wages, and per capita income), calling for further research. Richard Pomfret, in The Economic Development of Canada (1993), argued that the tariffs lowered competition, allowing manufacturers to charge well above their marginal costs. This would have lowered consumer surplus and thus aggregate welfare. He found that the National Policy knocked a fairly enormous 4-8 percent off Canada's GDP.
But the pendulum is swinging the other way again. Beaulieu and Cherniwchan (2014) point out two pretty gaping holes in Pomfret's analysis. First, he relies on the import-weighted average tariff (AWT), which is (per Irwin (2010)) downward-biased, ignores dispersion in rates among goods, and has zero economic interpretation (no good is actually charged the AWT). Second (and this mildly shocked me), his welfare analysis is inferred from later welfare estimates (from after 1950), assuming that since the AWT was higher over 1870-1910, the deadweight loss must have been greater.
The authors instead use a Trade Restrictiveness Index (TRI), which as mentioned above spits out the tariff rate which, if applied to all goods, would produce the same welfare loss as actually-existing rate structure. It's basically the sum of the (squared) ratio between expenditures with and without the tariff for each good. They also can compute the static deadweight loss directly using a similar formula. Using Canadian trade data from 1870-1910, they get a much lower, but still negative, estimate of the GDP loss, at about 0.7-1.5%. Targeting goods with close domestic substitutes, such that inefficient Canadian producers replaced foreign, increased the welfare cost of protection.
The problem with the partial equilibrium results is that the deadweight loss measure requires completely unrealistic assumptions: perfect competition, market clearing, no externalities, and constant returns to scale. All of these are manifestly untrue of 19th century Canadian industry. Partial equilibrium frameworks also don't account for changes in prices and productivity. Costinot and Rodriguez-Clare (2014) show that for small open economies like Canada (was) that rely heavily on international markets, even a modest influence over the terms of trade (the price of exports relative to imports) can have a big welfare effect. Partial equilibrium models, which are functions of industries' consumption shares, trade elasticities, and average weighted tariffs, ignore this channel.
Using product-specific trade data for Canada, the US, and Britain, Alexander and Keay (2018a) take a multi-sector general equilibrium approach. Surprisingly, they find that adopting protectionism actually increased GDP by 0.14-0.16 percent, as tariffs increased the domestic terms of trade in favor of manufacturing and augmented government revenues. Unilaterally adopting free trade, by contrast, would have lowered it by .23-.68 percent. Of course, the best-case scenario would have been for everyone to adopt free trade, but since the US was refusing reciprocality this wasn't a likely outcome.
In sum, early estimates of the short-term distortions imposed by the National Policy tariff were exaggerated. But did it succeed in nurturing competitive domestic manufacturing industries? I turn to this next.
Did Infant Industries Grow Up?
Perhaps the central purpose of the National Policy was to encourage domestic industrialization—basically an infant industry argument. The idea, often attributed to Alexander Hamilton in his 1791 Report on Manufactures, is that foreign producers with established industries will initially have a near-insurmountable advantage over a nascent one; both have to pay the hefty fixed costs of building a factory, buying machinery, and setting up infrastructure, but companies in the older industry can amortize the expense across more units of output (lower average total cost).
Plus, firms in existing industries have access to pools of skilled labor, knowledge spillovers, and technologies that are available only locally—what economists call Marshallian externalities. You need to offset the cost advantage with a tariff until your domestic manufacturers reach are efficient enough to be internationally competitive; otherwise, you can end up in a sub-optimal equilibrium specialized in a dead-end product.
Does this actually work in practice? Not always! Harrison and Rodriguez-Clare (2010) is an excellent survey of infant-industry protection and industrial policy more generally, and they write that "[c]ross-industry studies usually show that the removal of protection generates both intrafirm and intraindustry productivity gains." They make the (much stronger!) claim that "[t]here is no evidence to suggest that intervention for IP reasons in trade even exists," because, as in the Canadian case, revenue and special interest concerns seem to predominate.
More recently, however, Reka Juhasz has demonstrated that the Napoleonic blockade effectively protected French industry from British competition and let to growth in cotton textile factories. Per Pseudoerasmus, it's the first study to rigorously show that temporary protection for a "fledgling industry" can have positive long-run consequences.
The Canadian government was clearly trying to do IIP. Alexander and Keay (2019) find that the National Policy was highly selective, targeting finished goods over intermediate inputs. Transport equipment and petroleum and coal products experienced large increases, while leather and tobacco, say, were all but ignored. They show that the new tariffs disproportionately hit imports with close domestic substitutes (measured as higher trade elasticities, or the extent to which changes in price lead to changes in import quantities) produced by sizeable domestic industries if they had political clout.
That said, Ian Keay and a revolving cast of coauthors have produced a series of papers arguing that the National Policy worked. Inwood and Keay (2013) found that the tariffs of 1879 and 1887 were correlated with higher output of pig iron, increased industry turnover, and led to the transition from small, rural charcoal blast furnances to larger coke-burning plants. Domestic demand sharply boosted output while labor costs fell. They conclude that "tariff protection was required to trigger entry and investment in larger, technologically advanced blast furnaces, and only then could increasing domestic demand be satisfied by Canadian producers."
Harris et al. (2015) are more ambitious. Their analysis is motivated by the "New Trade Theory" models associated with Paul Krugman, Elhanan Helpman, and Marc Melitz. The main idea is that imperfect competition and increasing returns to scale change the calculus of neoclassical trade model, in which differences in productivity or factor endowments determined trade patterns. The Dales (1966) and Easton et al. (1988) papers that predicted slower growth were of the latter type.
The authors suggest two ways (models) that you can think about infant industry protection working. The first presents manufacturing as an oligopoly (relatively small number of sellers with some market power) with free entry in which prices are marked up above marginal cost. Slapping a tariff on manufactured imports increases aggregate demand, causing the entry of new firms and the expansion of employment in industry. Increased profitability and larger market size (from firm entry) leads firms increase scale and thereby exploit internal economies, allowing them to cut prices. In the neoclassical model, domestic and foreign manufactures are perfect subsitutes, so in a small open economy prices increase by the size of the tariff. But in the "industrial organization" model, there aren't always close Canadian substitutes for foreign goods, and the tariff's positive effect on firm size reduces markups, which raises productivity and lowers prices.
The second model proposes that learning-by-doing is a function of "industry experience," proxied by cumulative output. Tariffs increase output, accelerating the growth of accumulated experience and raising productivity, lowering costs and prices. In the figure below, a tariff permanently raises output in protected industry and allows it to move faster along the experience curve. The prediction is basically the same as above, but here productivity increases because of learning-by-doing—adapting machinery, training labor, etc.
To assess the effects of the National Policy, Harris et al. exploit variation in effective protection across fifteen Canadian industries as a natural experiment. Industries are classified as "target," "broad," and "unaffected" based on the size of the tariff increase. They run difference-in-differences and treatment intensity regressions to compare perfomance changes between treatment groups. Targeted industries included tobaco, paper, transportation, petroleum; the broad group was mostly metals; and food, rubber, leather, and textiles were considered "unaffected." In the treatment intensity regression, elasticity of output and TFP growth to treatment intensity were 3.6 and 4.6 percent respectively, with an elasticity of -4.6 percent on the price change. In the DiD, they found that output rose by 6.7 percent and productivity by 10.6 percent in the target group (5.2 and 8.0 in the broad group). Those are big numbers!
You should be worried about endogeneity concerns, of course. That's the whole reason why the Juhasz paper is so groundbreaking. Did the Canadian government pick winners for treatment? DiD specifications also demand parallel pre-trends between targeted and non-targeted industries, which could easily have been violated. Karacaovali (2011) adds that “more productive sectors have ... more to gain from lobbying and can potentially generate more protection.” It doesn't look like the Macdonald government explicitly targeted the most promising sectors, though, but rather to "restore prosperity" to "struggling industries, now so sadly depressed." Treated industries were also performing much worse on prices, productivity, and output beforehand.
To address endogeneity concerns, they instrument for intensity using political, locational, and market structure characteristics, including whether the industry's value-added came from districts with close electoral races in 1878 (or, confusingly, where the Conservatives won handsomely). The estimates are about the same. They also do a placebo test with the treatement year in 1875. I would've liked to see randomized assignment of treatment intensity, but this isn't really a true causal inference paper.
Finally, Keay (2018) shows that the Canadian government followed through on the other side of the infant industry bargain—removing tariffs from mature industries once they had become competitive. He finds a strong negative correlation between net export performance and the level of tariff protection after 1890 which substantially reduced the deadweight losses from the National Policy. So industries were allowed to "graduate" from state protection if they showed promise as exporters, permitting consumers to reap the welfare gains from lower prices.
To summarize: the National Policy tariff program appears to have succeeded in nursing several key export industries to international competitiveness and backing off afterward. But was it worthwhile?
Protection and Development
How do we know if IIP worked? It's not enough to check a box if protected sectors expanded their output—if you wipe out an industry's competitors, then yeah, it's going to produce more. Harrison and Rodriguez-Clare (2010) propose two tests for whether IIP worked: the Mill test, which requires that a protected industry become internationally competitive, and the Bastable test, which requires that the aggregate benefits outweight the costs of protection (deadweight loss + enforcement costs). I haven't found any study that has truly assessed the aggregate dynamic effects of the National Policy. But some stylized facts about post-1879 Canadian growth can help to illustrate the point.
Economic historians often associate Canadian growth with the railway-fuelled wheat boom, which took off after 1896. But GDP per capita grew by 2.2% per annum during 1879-95 and manufacturing value-added per capita by 2.4%, the latter rising at 4.2% in the first decade of the National Policy. By contrast, the 1870s had seen essentially zero growth. I'm not one for chart-squinting, but the difference between 0.1% and 4.2% is pretty big!
Targeted industries, which had been struggling during the '70s, grew rapidly after 1879. Paper output fell by 4.3% per annum over 1870-7, but grew by 13% per year over the next decade; iron and steel jumped from -2.4 to 8.4%. Altogether growth rates in the targeted and broad groups rose by 12 pp and 9 pp respectively, while the unaffected groups only increased by 0.8 pp.
Inwood and Keay (2013) discuss the case study of the iron and steel industry, which received strong tariff protection during the 1880s, with the AWT rising from 13.7% in 1870 to nearly 30% in 1893. By 1900, advanced blast furnaces had been installed at Hamilton, Deseronto, and Midland in Ontario; and at Pictou and Sydney in Nova Scotia. Pig iron production was 120 times greater in 1913 than it had been in 1870 and 50 times greater than in 1890. McInnis (2000) also suggests that the tariff encouraged the utilization of empty plant capacity, leading to spurts immediately after the 1879 and 1887 tariff shocks.
You can take these descriptive statistics, along with the causal results from Harris et al. (2015), as small-sample and imperfect. You could argue that some of the growth of the 1880s was just rebounding from the previous decade. But it seems implausible to say that industrial growth, specifically in targeted sectors, would have been even faster had Canada not attempted IIP. With the United States remaining staunchly protectionist, even renouncing the "Reciprocity Treaty (1854)" in 1866, some tariff response was probably warranted even in the short run.
Did Canada have latent comparative advantages in protected industries? Looks like it. Exports of paper, mineral products, and petroleum boomed. Table 2 below shows that Canada began to substantially diversify its exports during the National Policy years. In 1851, Canada basically was a timber economy with some wheat and cattle thrown in. By 1890 and even more so by 1900, as some industries were starting to mature, Canada was selling machinery, agricultural implements, drugs, clothing, and furniture abroad. The total value of wood product exports grew markedly despite the waning of the timber trade, thanks to the explosion of the protected paper industry. Canada's imports were increasingly dominated by capital equipment and parts for manufacturing plants, as well as coal with which to power them.
Canadian industries protected by the National Policy not only expanded their output, but saw rapid productivity growth, lowered their prices, and exported more. Targeted sectors displayed increasing returns to scale and large learning-by-doing effects, which together contributed to nearly 40 percent of their improvement in productivity. So it seems like the tariff passed the Mill test. And since the costs of the tariff might actually have been zero, it seems pretty likely that protection actually worked as an infant industry policy. The National Policy interacted with Canada's vast resource endowments and newly-integrated home market to produce competitive manufacturing industries, some of which would become the basis of the non-wheat economy of the twentieth century.
International Comparisons
By now, you're probably wondering why protection worked so well in Canada, but failed so frequently in other places and at other times. After all, we're familiar with how ISI in Latin America, especially Argentina (Canada's Spanish-speaking sister), went badly awry after World War II. But did protection work for Canada's contemporary rivals? The simplest justification for my argument would be a universal tendency for IIP to promote growth in the pre-1914 globalization era.
Coincidentally, Pseudoerasmus has a great post from 2016 on the "tariff-growth paradox"—Paul Bairoch's finding that countries with higher tariffs grew faster during the late 19th century. O'Rourke (2000), using a small panel of rich countries (Australia, Canada, Denmark, France, Germany, Italy, Norway, Sweden, Britain, and the USA) over 1875-1913, did find a correlation between average tariff rates and growth.
Clemens and Williamson (2001) confirmed this on a 35-country sample for 1865-1908 and 1919-1934. Both papers hypothesize that protection accelerated the expansion of "emerging" sectors (i.e. industry) at the expense of agriculture. The new sectors, characterized by Marshallian/learning externalities and scale economies, were supposed to head the transformation to developed-country status. This looks quite a bit like Canada, with the proviso that the wheat boom led to accelerated agricultural-sector growth after 1896.
Irwin (2002), however, countered that the correlation was driven by the use of revenue tariffs in fast-growing, land-abundant countries—namely the US, Argentina, and Canada. But as we've seen, revenue- and IIP-based tariffs can't be so easily differentiated. Lehmann and O'Rourke (2011), in any event, address this critique by cutting out revenue tariffs and find the positive relationship again. Industrial tariffs lead to positive growth over 1875-1913, and agricultural tariffs to negative. Canada's pattern of protection generally looks skewed toward IRS sectors, though it should be noted that some were declining outright and might have warranted sunsetting.
There are a couple of more recent studies here, like Schularick and Solomou (2011), who find that the 1875-1879 depression, which caused many countries (including Canada!) to adopt protectionism during the subsequent recovery, drove the tariff-growth correlation. This could be a damning indictment of my argument; as we've seen, the targeted sectors were doing worse during the depression. But I don't think that a "bounce-back" explanation really suffices to cover the structural expansion of protected industries in the Canadian economy or the new lines of production (e.g. machinery and farm tools) adopted.
I think Pseudo's interpretation of C&W—that the UK was a free-trade sink for protectionist countries—checks out for Canada. Exports to the UK rose from $22 million in 1870 to $43 million in 1880 and again to $93 million; exports to the protectionist US, which had been higher ($29 million) in 1870, increased only to $34 million in 1880 and $68 million in 1900. Indeed, many American firms located inside Canada during the National Policy era in order to take advantage of preferential intra-imperial duties. Altogether the UK was taking half of Canada's exports over 1880-1900. C&W's "prisoner's dilemma" model, in which coordination is the best option but protection is superior to unilateral free trade, is validated by the general equilibrium results that we discussed above.[1]
Protectionism in One Country
If the cross-country studies are "crude" and ambiguous, it's worth making straightforward comparisons with one-country case studies. Canada's mirror here is obviously the US, which as we mentioned ratcheted up protection after 1870. Head (1994) attributes the growth of America's steel rail industry to protection; it was uncompetitive in the 1860s, became a world leader under IIP, and later had its tariff successfully removed.
But Irwin (2000) finds that there's little room for a positive role for the US tariff, since "import tariffs may have raised the price of imported capital goods, thereby discouraging capital accumulation." He argues that "productivity growth in nontraded sectors, rather than in manufacturing, was the driving force behind the United States’s overtaking of the United Kingdom in per capita GDP during this period." That's not true of Canada, however, which largely exempted capital goods and intermediate inputs from its developmentalist program.
A more relevant point is made in Irwin's counterfactual simulation of the US tinplate industry, which was protected in 1890 (when there were no domestic producers) and subsequently became self-sufficient. He found that the tariff accelerated the industry's growth by 10 years, but that it would have emerged anyway thanks to falling ore prices and that the welfare cost outweighed the benefits. However, a smaller tariff (50% rather than 70%) might actually have improved welfare.
Clemens and Williamson (2005) suggest that Latin America's high pre-1914 tariffs–the world's highest–should be linked with the region's superior growth performance, at least relative to Asia (whose levels were four times lower). Uruguay had tariff barriers 2.5x Canada's in 1905. French tariffs were at 10.1% and Germany's at 9.1% in 1890, while the Latin American average was 34%. Labor productivity growth rates in Argentina, Brazil, and Chile all exceeded America's over 1900-1919, and Brazil's industrial GDP grew at a staggering rate of 9.8% per annum.
Why did protection fail so spectacularly in Latin America after World War II? The most likely reason appears to be the insertion of price distortions that prevented regional market integration, forcing producers to operate far below the minimum efficient scale. Argentina had over a dozen auto companies serving a home market that could barely fit one. Indeed, this was what happened in Canada until it signed the Automotive Agreement with the US, which assigned particular production lines to each country to ensure an MES. But why did IIP work in the (much smaller) settler economies of the Belle Epoque? Perhaps, as Pseudo suggests, it's the initial boost that comes from transferring unproductive labor out of agriculture that counted; once all available peasants had been mobilized, the growth gains went away. It also could be that Canada, unlike Argentina, discriminated in favor of intermediate inputs—unfortunately, meaning better institutions and policy choices, an annoying black box!
That doesn't quite explain how Canada ended the National Policy era with about the same manufacturing share with which she began it. Here, though, I think it's important to note that Canada had several resource booms during 1879-1914: minerals, metals, gold, wheat, wood pulp, etc. The discovery of new deposits and improvement in Canada's terms of trade might well have proved deindustrializing forces in the absence of industrial policies—for example, Ontario's ban on pulp wood exports—intended to encourage domestic processing of the bounty. With intervention (and American tech), however, Canada was able to channel cheap resources into a modern economy.
Vincent Geloso is rightly critical of this literature. He argues that the quality of data is bad even for Canada, and that it’s worse for other countries in the (tiny) sample. The Canadian numbers in particular are understated for the pre-1879 period and overstated for the Atlantic provinces adversely affected by the National Policy, which would damp down the shock from the tariff. Oddly, this actually helps to alleviate my concern that the post-tariff growth was just a rebound from the depression years.
He also critiques the tariff restrictiveness measure used by C&W (import revenues / import volume), because a super high tariff that shut out all imports would get a zero. The Canada-specific papers solve this problem, however.
TL;DR: the Canadian National Policy of 1879-1895 was a successful instance of deliberate infant industry protection (IIP). Selective targeting of industries with close domestic substitutes led to rapid output expansion, productivity growth, and price declines with little cost to consumer welfare. Nineteenth-century Canada is a good example of how a heavily agricultural, resource-dependent economy can use trade policy to develop a native manufacturing sector despite pressures toward primary product specialization. Main points:
Canadian protectionism worked -- it created internationally competitive industries at little cost to the domestic economy and forced US manufacturing firms to relocate inside the tariff wall.
It's possible that the result is idiosyncratic to the Belle Epoque world, in that Britain served as a "free trade sink" for Canada's export-facing economy.
IIP did not promote overall structural transformation, but mostly because it had to keep up with resource booms in the mineral and agricultural sectors.
Protection did not make Canada a less resource-dependent economy, but rather succeeded in onshoring more raw material processing
Two special announcements before we dive in. First, I’ve renamed (well, actually just removed the old placeholder title) the newsletter Great Transformations, which contains at least two book references and sort of reflects the general topics that I like to write about.
Second, I’ve added a paid subscription feature so that you can contribute to this project. The essays are staying in front of the paywall, but you’ll get a bit of additional content, including some enormous reading lists that I’m curating and weekly research posts. Maybe some other fun things in the future. The main idea, though, is that I’d like to derive some income from this project to make the time commitment worthwhile. If you can find the funds to support me, I’d really appreciate it.
Subscribe now
Also, since this is the only time I’m asking for money in the body of the newsletter, I’d be grateful if you shared around this particular edition.
A quick note: if you are receiving this post for the second time, I apologize. Substack failed to send out this issue originally and is not responding to my support requests, so I’ve decided to cut my losses and re-send it.
The basic idea behind infant industry protection is that you, as a nascent manufacturer, need the state to tax international competitors until you reach an efficient scale and productivity level. It’s a seductive argument. Who doesn’t like protecting infants? But there aren’t many great empirical studies that show it working, and there is a lot of anecdotal evidence for it failing catastrophically.
In my last post, I used that argument in making the case that Russian economic growth prior to World War I was sustainable. I contested Bob Allen's view that Russia's natural resource exports and protectionism doomed her to a "Latin American" development path. One thing I suggested was that Canada, which both had a "wheat boom" and hiked tariffs at about the same time as Russia, might present a better counterfactual model.
What I didn't know for certain, however, was whether Canadian protection really worked. So I decided to figure that out.
Canada was a relative latecomer to protection. In the US, a fractious Republican Congress raised average tariff rates from 15 percent in 1859 to 45 percent in 1870, and per Irwin (2010) they remained above 30 percent through 1890. Strenuous opposition by the Maritime provinces (New Brunswick, Nova Scotia, and Prince Edward Island), good economic performance, and politically marginal manufacturers coupled with a revenue-based interpretation of tariffs' role meant that little change in the tariff structure occurred up to 1879. Raw materials and semi-finished intermediate inputs entered the country duty-free, while a 15 percent ad valorem tax was placed on manufactured imports, raised slightly to 17.5 percent in 1874 (the Mackenzie Tariff) for revenue purposes.
In 1873, however, global markets slumped and by 1876 an outright "depression" had set in. By 1879, prices had dropped by 20 percent and timber exports by half. Bank failures spiked and government revenues tumbled. After winning the election of 1878, the Conservative Party, led by Prime Minister John A. Macdonald, launched the "National Policy," a development program designed to expand and integrate the Canadian economy. The National Policy had three main planks: the construction of the Canadian Pacific Railroad (completed in 1885); the encouragement of immigration to offset emigration to the US and settle the West; and the adoption of a strongly protectionist tariff regime to bolster domestic industry. I'm writing about the third component, and treat "National Policy" and "protectionism" synonymously. I include both the initial 1879 hike, when average tariff rates weighted by import value rose immediately from under 14 percent to 21 percent, and the Tupper Tariff of 1887.
There were three compelling motivations for the National Policy. First, higher tariffs offered a way to augment the country's revenues, of which duties constituted 70 percent until 1887. Second, the depression of the '70s had hammered domestic industry, with manufactured prices falling further than raw materials. Industrialists' fears were exacerbated by falling transport costs, which rendered producers more vulnerable to foreign competition, and the failure to win reciprocal openness from the US. The Conservatives' main supporters were the industrialists and city businessmen who'd suffered most from the trade depression and would benefit most from protection. Finally, the National Policy appears to have been designed to "balance" Canadian development by making the country more self-sufficient in manufacturing, which would serve an East-West domestic market populated by immigrants, integrated by railroads, and protected by high tariffs.
There was high variation in the tariff structure, though. For non-manufactured raw materials, tariffs only increased from 5.6 to 9.5 percent, and those on exotics remained virtually stable at 38-9 percent. Duties on cotton textiles inicreased from 17.5 to 30 percent and primary iron and steel from 0-5 to 12.5-17.5 percent. Beaulieu and Cherniwchan (2014) derive trade restrictiveness indexes, or the general tariff level that produces the same welfare loss as the actual tariff schedule. As the narrative evidence suggests, goods with close domestic substitutes—i.e. which Canadians were already making—were primary targets. So the Tupper tariff, which strategically increased rates on specific products, actually boosted restrictions by about the same amount as the original National Policy.
Did Canadian protectionism work? There are basically three questions here:
How much of a welfare loss did protectionism cause?
Did protectionism promote long-term industrial growth?
Did (2) outweigh (1)?
I'll take the answers in turn.
Welfare Costs
Early evaluations of the National Policy, especially by Canadians, tended to be positive. A select committee ordered by the government to "inquire into the operation of the tariff" sent two agents, Alfred H. Blackeby and Edward Willis, to survey hundreds of manufacturing firms in Ontario, Quebec, and the Maritime provinces. Blackeby reported:
In their classic textbook, Canadian Economic History, Easterbrook and Aitken argue that the National Policy was a "necessary evil." Protectionism augmented the market power and profits of manufacturers at the expense of consumers, but successfully sheltered infant industry, drove import substitution, and increased the scale of production. Aitken also saw the tariff, which raised aggregate labor demand and promoted immigration, as part of a "defensive expansionism" program designed to make Canada too big to swallow for the US.
Soon afterward, however, a series of revisionist studies of the Cliometric Revolution era set out to debunk the traditional view. On the basis of a neoclassical trade model, the Canadian economist John Dales (1966: 109–10) called the tariff “the price we pay for our protected manufacturing industry," arguing that "protection fosters inefficient, oligopolistic forms of market organization in Canada" and that "Canadian growth has been distorted by the National Policy.” Worse, he thought that the tariff encouraged the emigration of skilled workers and the immigration of unskilled labor, increasing GNP and lowering wages.
Easton et al. (1988) tested the Dales model and found that the extent to which the tariff promoted immigration was contingent on its welfare consequences (unemployment, real wages, and per capita income), calling for further research. Richard Pomfret, in The Economic Development of Canada (1993), argued that the tariffs lowered competition, allowing manufacturers to charge well above their marginal costs. This would have lowered consumer surplus and thus aggregate welfare. He found that the National Policy knocked a fairly enormous 4-8 percent off Canada's GDP.
But the pendulum is swinging the other way again. Beaulieu and Cherniwchan (2014) point out two pretty gaping holes in Pomfret's analysis. First, he relies on the import-weighted average tariff (AWT), which is (per Irwin (2010)) downward-biased, ignores dispersion in rates among goods, and has zero economic interpretation (no good is actually charged the AWT). Second (and this mildly shocked me), his welfare analysis is inferred from later welfare estimates (from after 1950), assuming that since the AWT was higher over 1870-1910, the deadweight loss must have been greater.
The authors instead use a Trade Restrictiveness Index (TRI), which as mentioned above spits out the tariff rate which, if applied to all goods, would produce the same welfare loss as actually-existing rate structure. It's basically the sum of the (squared) ratio between expenditures with and without the tariff for each good. They also can compute the static deadweight loss directly using a similar formula. Using Canadian trade data from 1870-1910, they get a much lower, but still negative, estimate of the GDP loss, at about 0.7-1.5%. Targeting goods with close domestic substitutes, such that inefficient Canadian producers replaced foreign, increased the welfare cost of protection.
The problem with the partial equilibrium results is that the deadweight loss measure requires completely unrealistic assumptions: perfect competition, market clearing, no externalities, and constant returns to scale. All of these are manifestly untrue of 19th century Canadian industry. Partial equilibrium frameworks also don't account for changes in prices and productivity. Costinot and Rodriguez-Clare (2014) show that for small open economies like Canada (was) that rely heavily on international markets, even a modest influence over the terms of trade (the price of exports relative to imports) can have a big welfare effect. Partial equilibrium models, which are functions of industries' consumption shares, trade elasticities, and average weighted tariffs, ignore this channel.
Using product-specific trade data for Canada, the US, and Britain, Alexander and Keay (2018a) take a multi-sector general equilibrium approach. Surprisingly, they find that adopting protectionism actually increased GDP by 0.14-0.16 percent, as tariffs increased the domestic terms of trade in favor of manufacturing and augmented government revenues. Unilaterally adopting free trade, by contrast, would have lowered it by .23-.68 percent. Of course, the best-case scenario would have been for everyone to adopt free trade, but since the US was refusing reciprocality this wasn't a likely outcome.
In sum, early estimates of the short-term distortions imposed by the National Policy tariff were exaggerated. But did it succeed in nurturing competitive domestic manufacturing industries? I turn to this next.
Did Infant Industries Grow Up?
Perhaps the central purpose of the National Policy was to encourage domestic industrialization—basically an infant industry argument. The idea, often attributed to Alexander Hamilton in his 1791 Report on Manufactures, is that foreign producers with established industries will initially have a near-insurmountable advantage over a nascent one; both have to pay the hefty fixed costs of building a factory, buying machinery, and setting up infrastructure, but companies in the older industry can amortize the expense across more units of output (lower average total cost).
Plus, firms in existing industries have access to pools of skilled labor, knowledge spillovers, and technologies that are available only locally—what economists call Marshallian externalities. You need to offset the cost advantage with a tariff until your domestic manufacturers reach are efficient enough to be internationally competitive; otherwise, you can end up in a sub-optimal equilibrium specialized in a dead-end product.
Does this actually work in practice? Not always! Harrison and Rodriguez-Clare (2010) is an excellent survey of infant-industry protection and industrial policy more generally, and they write that "[c]ross-industry studies usually show that the removal of protection generates both intrafirm and intraindustry productivity gains." They make the (much stronger!) claim that "[t]here is no evidence to suggest that intervention for IP reasons in trade even exists," because, as in the Canadian case, revenue and special interest concerns seem to predominate.
More recently, however, Reka Juhasz has demonstrated that the Napoleonic blockade effectively protected French industry from British competition and let to growth in cotton textile factories. Per Pseudoerasmus, it's the first study to rigorously show that temporary protection for a "fledgling industry" can have positive long-run consequences.
The Canadian government was clearly trying to do IIP. Alexander and Keay (2019) find that the National Policy was highly selective, targeting finished goods over intermediate inputs. Transport equipment and petroleum and coal products experienced large increases, while leather and tobacco, say, were all but ignored. They show that the new tariffs disproportionately hit imports with close domestic substitutes (measured as higher trade elasticities, or the extent to which changes in price lead to changes in import quantities) produced by sizeable domestic industries if they had political clout.
That said, Ian Keay and a revolving cast of coauthors have produced a series of papers arguing that the National Policy worked. Inwood and Keay (2013) found that the tariffs of 1879 and 1887 were correlated with higher output of pig iron, increased industry turnover, and led to the transition from small, rural charcoal blast furnances to larger coke-burning plants. Domestic demand sharply boosted output while labor costs fell. They conclude that "tariff protection was required to trigger entry and investment in larger, technologically advanced blast furnaces, and only then could increasing domestic demand be satisfied by Canadian producers."
Harris et al. (2015) are more ambitious. Their analysis is motivated by the "New Trade Theory" models associated with Paul Krugman, Elhanan Helpman, and Marc Melitz. The main idea is that imperfect competition and increasing returns to scale change the calculus of neoclassical trade model, in which differences in productivity or factor endowments determined trade patterns. The Dales (1966) and Easton et al. (1988) papers that predicted slower growth were of the latter type.
The authors suggest two ways (models) that you can think about infant industry protection working. The first presents manufacturing as an oligopoly (relatively small number of sellers with some market power) with free entry in which prices are marked up above marginal cost. Slapping a tariff on manufactured imports increases aggregate demand, causing the entry of new firms and the expansion of employment in industry. Increased profitability and larger market size (from firm entry) leads firms increase scale and thereby exploit internal economies, allowing them to cut prices. In the neoclassical model, domestic and foreign manufactures are perfect subsitutes, so in a small open economy prices increase by the size of the tariff. But in the "industrial organization" model, there aren't always close Canadian substitutes for foreign goods, and the tariff's positive effect on firm size reduces markups, which raises productivity and lowers prices.
The second model proposes that learning-by-doing is a function of "industry experience," proxied by cumulative output. Tariffs increase output, accelerating the growth of accumulated experience and raising productivity, lowering costs and prices. In the figure below, a tariff permanently raises output in protected industry and allows it to move faster along the experience curve. The prediction is basically the same as above, but here productivity increases because of learning-by-doing—adapting machinery, training labor, etc.
To assess the effects of the National Policy, Harris et al. exploit variation in effective protection across fifteen Canadian industries as a natural experiment. Industries are classified as "target," "broad," and "unaffected" based on the size of the tariff increase. They run difference-in-differences and treatment intensity regressions to compare perfomance changes between treatment groups. Targeted industries included tobaco, paper, transportation, petroleum; the broad group was mostly metals; and food, rubber, leather, and textiles were considered "unaffected." In the treatment intensity regression, elasticity of output and TFP growth to treatment intensity were 3.6 and 4.6 percent respectively, with an elasticity of -4.6 percent on the price change. In the DiD, they found that output rose by 6.7 percent and productivity by 10.6 percent in the target group (5.2 and 8.0 in the broad group). Those are big numbers!
You should be worried about endogeneity concerns, of course. That's the whole reason why the Juhasz paper is so groundbreaking. Did the Canadian government pick winners for treatment? DiD specifications also demand parallel pre-trends between targeted and non-targeted industries, which could easily have been violated. Karacaovali (2011) adds that “more productive sectors have ... more to gain from lobbying and can potentially generate more protection.” It doesn't look like the Macdonald government explicitly targeted the most promising sectors, though, but rather to "restore prosperity" to "struggling industries, now so sadly depressed." Treated industries were also performing much worse on prices, productivity, and output beforehand.
To address endogeneity concerns, they instrument for intensity using political, locational, and market structure characteristics, including whether the industry's value-added came from districts with close electoral races in 1878 (or, confusingly, where the Conservatives won handsomely). The estimates are about the same. They also do a placebo test with the treatement year in 1875. I would've liked to see randomized assignment of treatment intensity, but this isn't really a true causal inference paper.
Finally, Keay (2018) shows that the Canadian government followed through on the other side of the infant industry bargain—removing tariffs from mature industries once they had become competitive. He finds a strong negative correlation between net export performance and the level of tariff protection after 1890 which substantially reduced the deadweight losses from the National Policy. So industries were allowed to "graduate" from state protection if they showed promise as exporters, permitting consumers to reap the welfare gains from lower prices.
To summarize: the National Policy tariff program appears to have succeeded in nursing several key export industries to international competitiveness and backing off afterward. But was it worthwhile?
Protection and Development
How do we know if IIP worked? It's not enough to check a box if protected sectors expanded their output—if you wipe out an industry's competitors, then yeah, it's going to produce more. Harrison and Rodriguez-Clare (2010) propose two tests for whether IIP worked: the Mill test, which requires that a protected industry become internationally competitive, and the Bastable test, which requires that the aggregate benefits outweight the costs of protection (deadweight loss + enforcement costs). I haven't found any study that has truly assessed the aggregate dynamic effects of the National Policy. But some stylized facts about post-1879 Canadian growth can help to illustrate the point.
Economic historians often associate Canadian growth with the railway-fuelled wheat boom, which took off after 1896. But GDP per capita grew by 2.2% per annum during 1879-95 and manufacturing value-added per capita by 2.4%, the latter rising at 4.2% in the first decade of the National Policy. By contrast, the 1870s had seen essentially zero growth. I'm not one for chart-squinting, but the difference between 0.1% and 4.2% is pretty big!
Targeted industries, which had been struggling during the '70s, grew rapidly after 1879. Paper output fell by 4.3% per annum over 1870-7, but grew by 13% per year over the next decade; iron and steel jumped from -2.4 to 8.4%. Altogether growth rates in the targeted and broad groups rose by 12 pp and 9 pp respectively, while the unaffected groups only increased by 0.8 pp.
Inwood and Keay (2013) discuss the case study of the iron and steel industry, which received strong tariff protection during the 1880s, with the AWT rising from 13.7% in 1870 to nearly 30% in 1893. By 1900, advanced blast furnaces had been installed at Hamilton, Deseronto, and Midland in Ontario; and at Pictou and Sydney in Nova Scotia. Pig iron production was 120 times greater in 1913 than it had been in 1870 and 50 times greater than in 1890. McInnis (2000) also suggests that the tariff encouraged the utilization of empty plant capacity, leading to spurts immediately after the 1879 and 1887 tariff shocks.
You can take these descriptive statistics, along with the causal results from Harris et al. (2015), as small-sample and imperfect. You could argue that some of the growth of the 1880s was just rebounding from the previous decade. But it seems implausible to say that industrial growth, specifically in targeted sectors, would have been even faster had Canada not attempted IIP. With the United States remaining staunchly protectionist, even renouncing the "Reciprocity Treaty (1854)" in 1866, some tariff response was probably warranted even in the short run.
Did Canada have latent comparative advantages in protected industries? Looks like it. Exports of paper, mineral products, and petroleum boomed. Table 2 below shows that Canada began to substantially diversify its exports during the National Policy years. In 1851, Canada basically was a timber economy with some wheat and cattle thrown in. By 1890 and even more so by 1900, as some industries were starting to mature, Canada was selling machinery, agricultural implements, drugs, clothing, and furniture abroad. The total value of wood product exports grew markedly despite the waning of the timber trade, thanks to the explosion of the protected paper industry. Canada's imports were increasingly dominated by capital equipment and parts for manufacturing plants, as well as coal with which to power them.
Canadian industries protected by the National Policy not only expanded their output, but saw rapid productivity growth, lowered their prices, and exported more. Targeted sectors displayed increasing returns to scale and large learning-by-doing effects, which together contributed to nearly 40 percent of their improvement in productivity. So it seems like the tariff passed the Mill test. And since the costs of the tariff might actually have been zero, it seems pretty likely that protection actually worked as an infant industry policy. The National Policy interacted with Canada's vast resource endowments and newly-integrated home market to produce competitive manufacturing industries, some of which would become the basis of the non-wheat economy of the twentieth century.
International Comparisons
By now, you're probably wondering why protection worked so well in Canada, but failed so frequently in other places and at other times. After all, we're familiar with how ISI in Latin America, especially Argentina (Canada's Spanish-speaking sister), went badly awry after World War II. But did protection work for Canada's contemporary rivals? The simplest justification for my argument would be a universal tendency for IIP to promote growth in the pre-1914 globalization era.
Coincidentally, Pseudoerasmus has a great post from 2016 on the "tariff-growth paradox"—Paul Bairoch's finding that countries with higher tariffs grew faster during the late 19th century. O'Rourke (2000), using a small panel of rich countries (Australia, Canada, Denmark, France, Germany, Italy, Norway, Sweden, Britain, and the USA) over 1875-1913, did find a correlation between average tariff rates and growth.
Clemens and Williamson (2001) confirmed this on a 35-country sample for 1865-1908 and 1919-1934. Both papers hypothesize that protection accelerated the expansion of "emerging" sectors (i.e. industry) at the expense of agriculture. The new sectors, characterized by Marshallian/learning externalities and scale economies, were supposed to head the transformation to developed-country status. This looks quite a bit like Canada, with the proviso that the wheat boom led to accelerated agricultural-sector growth after 1896.
Irwin (2002), however, countered that the correlation was driven by the use of revenue tariffs in fast-growing, land-abundant countries—namely the US, Argentina, and Canada. But as we've seen, revenue- and IIP-based tariffs can't be so easily differentiated. Lehmann and O'Rourke (2011), in any event, address this critique by cutting out revenue tariffs and find the positive relationship again. Industrial tariffs lead to positive growth over 1875-1913, and agricultural tariffs to negative. Canada's pattern of protection generally looks skewed toward IRS sectors, though it should be noted that some were declining outright and might have warranted sunsetting.
There are a couple of more recent studies here, like Schularick and Solomou (2011), who find that the 1875-1879 depression, which caused many countries (including Canada!) to adopt protectionism during the subsequent recovery, drove the tariff-growth correlation. This could be a damning indictment of my argument; as we've seen, the targeted sectors were doing worse during the depression. But I don't think that a "bounce-back" explanation really suffices to cover the structural expansion of protected industries in the Canadian economy or the new lines of production (e.g. machinery and farm tools) adopted.
I think Pseudo's interpretation of C&W—that the UK was a free-trade sink for protectionist countries—checks out for Canada. Exports to the UK rose from $22 million in 1870 to $43 million in 1880 and again to $93 million; exports to the protectionist US, which had been higher ($29 million) in 1870, increased only to $34 million in 1880 and $68 million in 1900. Indeed, many American firms located inside Canada during the National Policy era in order to take advantage of preferential intra-imperial duties. Altogether the UK was taking half of Canada's exports over 1880-1900. C&W's "prisoner's dilemma" model, in which coordination is the best option but protection is superior to unilateral free trade, is validated by the general equilibrium results that we discussed above.[1]
Protectionism in One Country
If the cross-country studies are "crude" and ambiguous, it's worth making straightforward comparisons with one-country case studies. Canada's mirror here is obviously the US, which as we mentioned ratcheted up protection after 1870. Head (1994) attributes the growth of America's steel rail industry to protection; it was uncompetitive in the 1860s, became a world leader under IIP, and later had its tariff successfully removed.
But Irwin (2000) finds that there's little room for a positive role for the US tariff, since "import tariffs may have raised the price of imported capital goods, thereby discouraging capital accumulation." He argues that "productivity growth in nontraded sectors, rather than in manufacturing, was the driving force behind the United States’s overtaking of the United Kingdom in per capita GDP during this period." That's not true of Canada, however, which largely exempted capital goods and intermediate inputs from its developmentalist program.
A more relevant point is made in Irwin's counterfactual simulation of the US tinplate industry, which was protected in 1890 (when there were no domestic producers) and subsequently became self-sufficient. He found that the tariff accelerated the industry's growth by 10 years, but that it would have emerged anyway thanks to falling ore prices and that the welfare cost outweighed the benefits. However, a smaller tariff (50% rather than 70%) might actually have improved welfare.
Clemens and Williamson (2005) suggest that Latin America's high pre-1914 tariffs–the world's highest–should be linked with the region's superior growth performance, at least relative to Asia (whose levels were four times lower). Uruguay had tariff barriers 2.5x Canada's in 1905. French tariffs were at 10.1% and Germany's at 9.1% in 1890, while the Latin American average was 34%. Labor productivity growth rates in Argentina, Brazil, and Chile all exceeded America's over 1900-1919, and Brazil's industrial GDP grew at a staggering rate of 9.8% per annum.
Why did protection fail so spectacularly in Latin America after World War II? The most likely reason appears to be the insertion of price distortions that prevented regional market integration, forcing producers to operate far below the minimum efficient scale. Argentina had over a dozen auto companies serving a home market that could barely fit one. Indeed, this was what happened in Canada until it signed the Automotive Agreement with the US, which assigned particular production lines to each country to ensure an MES. But why did IIP work in the (much smaller) settler economies of the Belle Epoque? Perhaps, as Pseudo suggests, it's the initial boost that comes from transferring unproductive labor out of agriculture that counted; once all available peasants had been mobilized, the growth gains went away. It also could be that Canada, unlike Argentina, discriminated in favor of intermediate inputs—unfortunately, meaning better institutions and policy choices, an annoying black box!
That doesn't quite explain how Canada ended the National Policy era with about the same manufacturing share with which she began it. Here, though, I think it's important to note that Canada had several resource booms during 1879-1914: minerals, metals, gold, wheat, wood pulp, etc. The discovery of new deposits and improvement in Canada's terms of trade might well have proved deindustrializing forces in the absence of industrial policies—for example, Ontario's ban on pulp wood exports—intended to encourage domestic processing of the bounty. With intervention (and American tech), however, Canada was able to channel cheap resources into a modern economy.
Vincent Geloso is rightly critical of this literature. He argues that the quality of data is bad even for Canada, and that it’s worse for other countries in the (tiny) sample. The Canadian numbers in particular are understated for the pre-1879 period and overstated for the Atlantic provinces adversely affected by the National Policy, which would damp down the shock from the tariff. Oddly, this actually helps to alleviate my concern that the post-tariff growth was just a rebound from the depression years.
He also critiques the tariff restrictiveness measure used by C&W (import revenues / import volume), because a super high tariff that shut out all imports would get a zero. The Canada-specific papers solve this problem, however.