A general equilibrium analysis of Canada’s national policy

JID: YEXEH

ARTICLE IN PRESS

[m3GeSsc;November 24, 2017;19:55]

Explorations in Economic History 000 (2017) 1–15

Contents lists available at ScienceDirect

Explorations in Economic History journal homepage: www.elsevier.com/locate/eeh

A general equilibrium analysis of Canada’s national policy Patrick D. Alexander a, Ian Keay b,∗ a b

International Economic Analysis Department, Bank of Canada Economics Department, Queen’s University, Canada

a b s t r a c t In this paper, we study the impact of Canada’s adoption of protectionist trade policy in 1879 on Canadian welfare. Under the National Policy, the Canadian average weighted tariff increased from 14% to 21%. The conventional view is that this was a distortionary policy that negatively affected Canadian welfare. We argue that this view is incomplete because it ignores general equilibrium effects. Using a multi-industry general equilibrium model with differentiated goods, we show that tariffs’ impact on welfare can potentially be positive, even for small open economies, due to their impact on domestic terms of trade and government revenues. We apply these theoretical insights in a reassessment of the static impact that the National Policy had on Canadian welfare in 1879, using newly compiled granular trade and production data, and newly estimated historically contemporaneous import demand elasticities. Our results suggest that, although a multilateral move to free trade would have resulted in the best welfare outcome for Canadians, the National Policy’s tariff increases actually improved Canadian welfare by approximately 0.15% of GDP in 1879 – an amount equivalent to approximately $2.3 billion 2016 USD. © 2017 Elsevier Inc. All rights reserved.

1. Introduction Powerful market integration and globalization forces were at work during the late nineteenth and early twentieth centuries. Global demand was rising sharply, transport costs were falling, and manufactured goods, raw materials, people and capital were moving across international borders with unprecedented fluidity as markets integrated along both intensive and extensive margins (Betran and Huberman, 2016; Estevadeordal et al., 2003; Huberman et al., 2017). Policy responses during this first era of globalization varied widely across countries. Britain maintained predominantly free trade policies that had originally been introduced during the mid-nineteenth century, while simultaneously pressuring their Commonwealth trade partners to resist domestic calls for protection (Bairoch, 1972; 1989; Irwin, 1994). The United States, in contrast, maintained persistently high tariff levels, particularly on manufactured imports, throughout the 1870–1913 period (Irwin, 2000a; 2000b; 2010; Taussig, 1931). Meanwhile, a group of rapidly converging small-open-economies, which included Argentina, Australia, the Scandinavian countries, and Canada, struggled to balance their need for tariff revenue against demands for protection from domestic producers who were facing growing import volumes and a shift in the industrial intensity of their imports (Beaulieu and Cherniwchan, 2014; Pomfret, 1995; Salmon, 1997). In Canada, the response to the domestic and international pressures brought on by globalization was decisive. In part inspired by Henry Clay’s early nineteenth century American System and informed by Britain’s experience with free trade, John A. Macdonald’s Conservative government introduced the National Policy as part of the federal budget on March 14, 1879. This policy had three broad objectives: the promotion of immigration, the building of a trans-continental rail line, and the protection of domestic manufacturers from US (and to a lesser extent British) competition. The National Policy’s tariff changes signalled a clean and abrupt discontinuity

∗

Corresponding author. E-mail addresses: [email protected] (P.D. Alexander), [email protected] (I. Keay).

https://doi.org/10.1016/j.eeh.2017.11.002 Received 2 March 2017; Received in revised form 8 November 2017; Accepted 15 November 2017 Available online xxx 0014-4983/© 2017 Elsevier Inc. All rights reserved.

Please cite this article as: P.D. Alexander, I. Keay, Explorations in Economic History (2017), https://doi.org/10.1016/j.eeh.2017. 11.002

JID: YEXEH

ARTICLE IN PRESS

P.D. Alexander, I. Keay

[m3GeSsc;November 24, 2017;19:55] Explorations in Economic History 000 (2017) 1–15

in the primary objective of Canadian trade policy, explicitly favoring protectionism for the first time.1 Virtually every line of the Canadian tariff schedule was rewritten, specific industries were narrowly targeted for protection, and average tariff rates imposed on manufactured imports rose from 14.2% in 1877 to 21.5% in 1880 (Alexander and Keay, 2017).2 Further tariff increases were imposed by Macdonald’s Finance Minister Charles Tupper in 1887, and even with the reintroduction of imperial preferences under the Fielding tariffs in 1897, and vigorous opposition to protectionism by Wilfrid Laurier’s Liberal Party in the ‘free trade elections’ of 1891 and 1911, a protectionist trajectory which remained in place over the next 110 years, was firmly established for Canadian trade policy.3 In this paper we use a static, general equilibrium trade model with multiple industries, similar to a version presented in Costinot and Rodriguez-Clare (2014), to show that during the first era of globalization, even for small open economies like Canada, increases in tariff protection were not necessarily associated with reductions in domestic welfare. Our general equilibrium approach allows us to capture not only the partial equilibrium distortionary effect of the National Policy’s tariffs on Canadian prices, but also the policy’s positive impact on domestic terms of trade, and government revenue.4 The conventional view among economic historians is that Canada’s National Policy may have fostered long run-dynamic welfare improvements by, for example, sheltering infant industries, promoting agglomeration and encouraging technological change, but the policy was costly for Canadians in terms of static reductions in welfare.5 Pomfret (1993), for example, suggests that the policy’s tariff increases reduced welfare by more than 4% of gross domestic product (GDP) in 1879. Beaulieu and Cherniwchan (2014) use Anderson and Neary (2005) partial equilibrium measure of static deadweight loss (DWL) to show that the National Policy’s welfare effects may have been considerably lower than Pomfret claimed, but they were still negative, and in the neighborhood of 0.7%– 1.5% of Canadian GDP. We believe that these conventional measures are incomplete because they only consider partial equilibrium distortionary price effects. The reasons for the adoption of an incomplete approach are twofold. First, any broader general equilibrium assessment of the impact of protective tariffs requires the measurement of terms of trade and government revenue effects, which are challenging to quantify in a historical context because the data requirements are more demanding than partial equilibrium assessments (Feenstra, 1995). Second, terms of trade effects, in particular, can be completely dismissed for small open economies like Canada in 1879, if we believe that these economies lacked any measurable international market power.6 The measurement of general equilibrium welfare effects in our model requires information from the year of the policy change on total income; import penetration ratios; expenditure shares; and industry-specific trade elasticities, for Canada and its trade partners – the United States, Britain and the rest-of-the-world (RoW); as well as industry-level evidence on the change in average weighted tariffs (AWT) under the National Policy. We construct or, where appropriate, estimate these variables and parameters using newly compiled granular, annual trade and production data. Our results suggest that the increase in Canadian tariffs under the National Policy actually improved Canadian welfare in the year in which the policy was introduced by between 0.14% and 0.16% of real domestic GDP. In 2016 USD, this aggregate static welfare effect is equivalent to approximately $2.3 billion.7 Our assessment is sensitive to the trade elasticity estimates used, but our finding of a small, positive welfare effect is robust across a wide range of modern and historically contemporaneous elasticity choices. We also show that although Canadian welfare was higher under this policy than it would have been had Canada unilaterally lowered its tariffs to zero in 1879, a multilateral move to free trade would have resulted in an even better welfare outcome for Canadians. We conclude that from a general equilibrium perspective, in the economic environment that characterized the first era of globalization, Canada’s unilateral adoption of protectionist trade policy was a welfare-enhancing policy change. The modern parallels to this historical episode are clear, if not perfect. Today, complexities associated with product and policy differentiation, global value chains, and multilateral and bilateral variation in trade costs, limit our ability to identify the conditions under which the adoption of protectionist objectives can significantly affect a country’s aggregate welfare, either positively or negatively (Caliendo and Parro, 2015; Costinot and Rodriguez-Clare, 2014; Feenstra, 1995; Ossa, 2014). Simplicity in the Canadian trading environment in 1879 – few trading partners, a relatively small number of products exchanged, uniform tariffs across partners, and newly available, finely detailed trade, price and production data – relax some of these modern limitations. The introduction of the National Policy in Canada, therefore, provides a clean, well-defined policy break that lends itself to a detailed general equilibrium assessment of the welfare impact of a country’s unilateral adoption of protectionism, relative to unilateral and multilateral free trade policy options. We do not only learn about the welfare impact of the National Policy – our assessment of this change in policy objectives reveals the need in both modern and historical contexts to adopt a general equilibrium perspective, and we can clearly

1 Revenue objectives were never completely abandoned. Tariffs on imported exotics, for example, which include goods with little or no domestic production and relatively low import demand elasticities, remained high throughout the post-1870 era (Beaulieu and Cherniwchan, 2014, 161). 2 During this period, Canadian trade, price and production data are reported over fiscal years that end June 30. As a result, the introduction of the National Policy spans the fiscal years 1878 and 1879. 3 For a detailed assessment of the structure of the tariff changes under the National Policy, see Alexander and Keay (2017), or Beaulieu and Cherniwchan (2014). Canada did not fully engage in more open trade until the Canada-US Free Trade Agreement came into effect on January 1, 1989. 4 Both our general equilibrium and the more conventional partial equilibrium approaches are static, which means that measured welfare changes apply only to the year in which the policy discontinuity occurred. 5 Inwood and Keay (2013) show that the National Policy induced investment and technological innovation in Canadian iron and steel industries. Harris et al. (2015) argue that the most protected industries under the National Policy experienced disproportionately large output increases, productivity improvements, and price declines. 6 As we discuss in Section 3, newer developments in trade theory provide a basis for accounting for the degree of market power, which, both according to these models and empirical evidence, is positive even for small open economies. 7 This $2.3 billion estimate is equal to 0.15% of Canada’s approximately $1.53 trillion USD 2016 GDP.

2

JID: YEXEH

ARTICLE IN PRESS

P.D. Alexander, I. Keay

[m3GeSsc;November 24, 2017;19:55] Explorations in Economic History 000 (2017) 1–15

Fig. 1. Canadian average weighted tariffs.

see how trade elasticity choices that vary across structural specifications, levels of aggregation, and country and period-specificity, impact our welfare measures.

2. Canadian trade policy, 1870–1913 Under the British North America Act, the provinces of Ontario, Quebec, Nova Scotia and New Brunswick formed the Dominion of Canada on July 1, 1867. One of the new federal government’s constitutional powers was the regulation of international trade. Because tariff revenues formed a substantial share of the federal budget – averaging over 55% from 1867 to 1878 – trade policy was a constant, often bitter, source of dispute amongst the federal political parties. Minor changes were made to the Canadian tariff schedule on an annual basis throughout the first era of globalization, but only six revisions affected the full schedule: the Mackenzie Tariff in 1874; the National Policy in 1879; the Tupper Tariff of 1887; the Fielding Tariff of 1897; and the Tariff Acts of 1894 and 1907 (Gillespie, 1991; McDiarmid, 1946). We measure Canadian average weighted tariffs with newly-compiled product-level data from the Canadian Trade and Navigation Tables, which were published in the federal government’s annual sessional papers for each fiscal year starting in 1868. These trade tables report the total value of imports for home consumption and the total value of all duties collected at the product level for all manufactured and non-manufactured import goods.8 Product-level tariff rates are simply total duties collected, divided by total import values. Following Beaulieu and Cherniwchan (2014), we assign each product a commonly identified six-digit Harmonized Commodity Description and Coding System (HS6) number, which facilitates the aggregation of individual products into 16 two-digit Standard Industrial Classification (SIC2) manufacturing industries, and an aggregate non-manufacturing industry.9 Non-manufactured products include unprocessed raw materials and foodstuffs, such as sand and gravel; logs in the rough; live animals; or raw sugar. In Fig. 1 we illustrate the Canadian import value weighted average ad valorem tariff on all manufactured and non-manufactured import products for each fiscal year from 1870 to 1913. The six broad tariff revisions can be easily identified in the series included in Fig. 1.

8

In total our data includes over 1450 product-year observations. We use the 1948 SIC classification to match our trade data with Urquart’s (1993) manufacturing industry output data. For a more detailed discussion of the derivation of our AWT see Alexander and Keay (2017). 9

3

JID: YEXEH

ARTICLE IN PRESS

P.D. Alexander, I. Keay

[m3GeSsc;November 24, 2017;19:55] Explorations in Economic History 000 (2017) 1–15

In response to increased demands for federal revenue, the Mackenzie Tariff was passed by John A. Macdonald’s Conservative government in 1874. This revision altered rates relatively uniformly across a wide range of products, but the average increases were small: only 2.5% for manufactured products and just over 1.5% for non-manufactured products. These modest changes did little to satisfy the calls for more explicitly protectionist policy among the growing industrial and urban elites, and their workers, who made up much of the Conservatives’ political base. One of the main planks in the Conservatives’ platform during the 1878 federal election campaign was a promise to foster an East-West domestic market for central Canadian industrial products. Protectionism was to play a key role in this market building policy. The Conservatives won the 1878 election, and Macdonald’s government established a budget sub-committee almost immediately to hear petitions from stake-holders seeking protection from foreign competition. The revision of the tariff schedule that resulted from this process reflected a dramatic shift in the country’s long standing trade policy objectives by explicitly embracing protection. Under Macdonald’s new National Policy, ad valorem tariff rates averaged over all import products rose from 14.2% in the fiscal year ending June 30, 1878, to 20.8% in the year ending June 30, 1880 (see Appendix Table A1). However, even more than the overall increase in average rates, the narrow targeting of these tariff increases really reveals the National Policy as an abrupt, clean policy discontinuity. Some industries and some product-types received large tariff increases, while others saw virtually no increases in protection. Protection for non-manufactured raw materials, for example, rose only modestly in 1879, from 5.6% to 9.5%, and tariffs on exotics – products with little domestic production and relatively low import demand elasticities, such as tea, spices or wine – were virtually unchanged, increasing from 38.2% to 38.9%. The AWT on manufactured products, in contrast, rose sharply, from 14.2% to 21.5%. The differential between manufactured products’ AWT and raw materials’ AWT increased by 2.6 percentage points in 1879, while the differential between manufactured and exotic products’ AWT decreased by more than 6.4 percentage points. Even within the manufacturing sector, tariff increases were not uniform – Leather, Wood, and Chemicals, saw relatively small tariff increases under the National Policy, while the average tariff imposed on products produced by Printing and Publishing, Non-Ferrous Metals, and Transport Equipment, more than doubled. In Fig. 1 we can see less dramatic, but still highly differentiated tariff changes following a second round of protectionist revisions under the Tupper Tariffs in 1887, and smaller, more uniform adjustments in the face of slowly declining AWT with the Tariff Act of 1894. The Fielding Tariff in 1897 and the Tariff Act of 1907 slightly lowered average tariff rates, particularly for manufactured goods, by reintroducing preferential tariffs for imported products originating within the British Empire. In 1913, on the eve of World War 1, the Canadian average weighted tariff over all import products remained nearly 4 percentage points higher than it had been in 1870, with non-manufactured goods’ AWT falling over the period, from 15% to 7%, and manufactured output products’ AWT rising from 14% to 20%. The protectionist objectives that became entrenched in Canada through virtually the entire twentieth century, first appear in the tariff schedule with the introduction of the National Policy in 1879. The policy break was clean, the Canadian trading environment was relatively uncluttered, and highly granular trade and production data is available for us to assess the welfare consequences of this move towards protectionism.

3. Measuring welfare: a general equilibrium approach Our measure of the general equilibrium welfare effects of the adoption of protectionist trade policy in Canada in 1879 is derived from a version of the basic Armington model with multiple sectors, as described in Arkolakis et al. (2012), and Costinot and RodriguezClare (2014). The main distinction between our approach – described in detail in the online supplementary materials’ Technical Appendix - and these other models is that they rely on an endowment economy. Goods in our environment are produced with wage labor, which in our view, provides a more intuitive micro-foundation for the derivation of general equilibrium welfare effects. In the end, the measurement of welfare changes in all three models relies on functionally identical structural representations of the change in real consumption. The model in Ossa (2015) also shares similar features with our approach, in that it includes wage labor, but Ossa also includes intermediates as inputs into the production of tradable products. As revealed in Costinot and Rodriguez-Clare (2014, Table 4.2), the inclusion of intermediates potentially increases the gains from unilateral protection (and the losses for Canada’s trade partners), which suggests that our measure of the potential welfare gains for Canada from the National Policy could be conservative due to the absence of intermediates in our model. We cannot incorporate intermediates because we have very limited information about the distinction between intermediates and final consumption goods for Canada, and no information about intermediates for the US, UK or RoW.10 Building monopolistic competition and firm heterogeneity into our model would allow us to introduce firm entry and firm selection, which could affect welfare changes resulting from the adoption of protectionism. However, Costinot and Rodriguez-Clare (2014) show that these extensions tend to affect gains from trade in ambiguous ways. It is, therefore, unclear how an augmented version of our model would impact our qualitative conclusions, either positively or negatively. The general equilibrium approach we develop is relatively parsimonious and it yields closed-form, empirically tractable equations for the impact of tariff changes on welfare.

10 Caliendo and Parro (2015, 28 and Table 11) note that the inclusion of intermediate goods in their model amplifies the welfare effects of NAFTA’s tariff reductions, increasing both the welfare gains enjoyed by the US and Mexico, and the welfare loses borne by Canada. Costinot and Rodriguez-Clare (2014, 229–231 and Table 4.2) show that a model with intermediates predicts larger welfare gains for the US, and larger welfare loses for its trade partners, following a counterfactual unilateral 40% increase in US tariffs, relative to a model without intermediates. These findings suggest that the inclusion of intermediates increases measured welfare effects, both positive and negative, in response to unilateral tariff changes.

4

ARTICLE IN PRESS

JID: YEXEH P.D. Alexander, I. Keay

[m3GeSsc;November 24, 2017;19:55] Explorations in Economic History 000 (2017) 1–15

For ease of exposition, we use what is commonly referred to as ‘exact hat algebra’ to represent static counterfactual scenarios, where 𝑧̂ = 𝑧′ ∕𝑧 denotes the ratio of counterfactual value to the initial value of any given variable z (Dekle et al., 2007). In the Technical Appendix, we derive the following expression for the general equilibrium change in welfare for country n following a change in ad valorem tariffs: ) ) 𝑠 𝑆 𝑆 (( ∏ ∏ 1 − 𝜆𝑛 ( 𝑠 ) −1𝑠 𝛼 𝑠 𝑊̂ 𝑛 = 𝐶̂𝑛𝑠 𝛼 = (1) 𝜋̂ 𝑛𝑛 𝜃 ′ 1 − 𝜆𝑛 𝑠=1 𝑠=1 Where 𝐶̂𝑛𝑠 denotes the change in real consumption in industry s of country n; 𝛼 s are industry-specific expenditure shares; 𝜃 s are 𝑠 is the share of country n expenditure industry-specific trade elasticities; 𝜆n is the share of tariff revenue in total expenditure; and 𝜋̂ 𝑛𝑛 in industry s, sourced from country n. Our model also yields closed-form equations for counterfactual tariff revenue shares, the change in real labor income (𝑌̂𝑛 ), and industry-specific trade shares. 𝜆′𝑖

=

𝑁 ∑ 𝑆 ∑ 𝑖=1 𝑠=1

𝑌̂𝑛 𝑌𝑛 =

𝑠 𝜋̂ 𝑛𝑖 =

𝑡𝑠𝑖 ′

1 + 𝑡𝑠𝑖 ′

𝑁 ∑ 𝑆 ∑ 𝑖=1 𝑠=1

( )− 𝜃 𝑠 𝑠 𝜏̂𝑠 𝑌̂ 𝛼𝑖𝑠 𝜋𝑖𝑛 𝑖𝑛 𝑗 ( )−𝜃 𝑠 ∑𝐽 𝜋 𝑠 𝜏̂𝑠 𝑌̂𝑗 𝑗=1

𝑖𝑗

(2)

𝑖𝑗

( )−𝜃 𝑠 𝑠 𝜏̂𝑠 𝑌̂ 𝛼𝑖𝑠 𝜋𝑖𝑛 𝑖𝑛 𝑛 𝑌̂𝑖 𝑌𝑖 1 ( ) 𝑠 𝑠 ′ − 𝜃 1 + 𝑡𝑛 ∑𝐽 1 − 𝜆′𝑖 𝑠 ̂𝑠 ̂ 𝑗=1 𝜋𝑖𝑗 𝜏𝑖𝑗 𝑌𝑗

(3)

( )−𝜃 𝑠 𝜏̂𝑠 𝑌̂ 𝑛𝑖 𝑗

∑𝐽

𝑗=1

(4)

( )− 𝜃 𝑠 𝑠 𝜏̂𝑠 𝑌̂ 𝜋𝑛𝑗 𝑛𝑗 𝑗

𝑠 captures changes in industry-specific total trade costs that fall In Eqs. (2)–(4), ts denotes industry-specific ad valorem tariffs, and 𝜏̂𝑖𝑛 on imports from i to n. In our model, the impact of a change in tariffs on welfare depends on how the change affects Eq. (1). Intuitively, an increase in tariffs in industry s of country n will raise welfare to the extent that the change positively affects the terms of trade in n, increases tariff revenues, and as a result, raises total real consumption in n. These are exactly the effects that are missed in standard partialequilibrium deadweight loss calculations, which depend only on industries’ home consumption shares, trade elasticities, and average weighted tariffs (Beaulieu and Cherniwchan, 2014: 152; Irwin, 2010: 121). Like partial-equilibrium deadweight loss measures, (1) also 𝑠 ). From captures the reduction in welfare that results from the distortionary effects of higher tariffs on home consumption shares (𝜋𝑛𝑛 Eqs. (1)–(4), we can see that the balance of these competing forces depends crucially on the trade elasticities parameters, 𝜃 𝑠 . All else equal, when the magnitude of 𝜃 𝑠 is larger, imports and home production are more substitutable, tariffs are more distortionary, and gains from unilateral tariffs are reduced. The key distinguishing feature of our general equilibrium approach, relative to partial equilibrium welfare measures, is the assumption that countries lie on a spectrum from smaller to larger economies, which exercise relatively little to relatively more international market power. The conventional partial equilibrium approaches assume that small open economies exert no influence on international prices. Costinot and Rodriguez-Clare (2014, 227–228) show that this assumption can be very restrictive, because even though smaller countries may exert little influence over their terms of trade, their greater reliance on international markets means that the welfare effect of any influence that does exist can be substantial. Our general equilibrium model, with Costinot and Rodriguez-Clare’s flexible interpretation of the impact of even relatively small amounts of market power, allows for heterogeneity along the country size-market power spectrum, and therefore, tariff changes can have other effects on welfare in addition to their distortionary impact on prices.11 We also note that our model, like the more conventional partial equilibrium welfare measures, is static. With Eq. (1) we measure the change in welfare that occurred due to tariff changes between 1877 – the last fiscal year before the introduction of the National Policy – and 1880 – the first fiscal year following the Policy’s introduction. The general equilibrium effects we report do not depend on dynamic changes in technology, learning-by-doing, or the exploitation of scale economies, and they are not cumulative over our period of study.

4. Data To measure general equilibrium changes in welfare that were the result of the adoption of protectionist trade policy in Canada under the National Policy in 1879, we must solve the system of equations described by (1)–(4). Doing so requires information on 𝑠 and 𝜋 𝑠 ), and industry-specific expenditure national income (𝑌𝑛 ), industry-specific home consumption and external trade shares (𝜋𝑛𝑛 𝑛𝑖

11 The predictions from this model are generalizable across other environments, including monopolistic competition (Gros, 1987; Ossa, 2014), producer heterogeneity (Demidova and Rodriguez-Clare, 2009; Felbermayr et al., 2013; Alvarez and Lucas, 2007), and variable mark-ups (Demidova, 2015). In all of these environments, as long as goods are not perfect substitutes, in which case 𝜃 s approaches ∞, welfare will be maximized at unilateral tariff levels greater than zero for all countries.

5

JID: YEXEH

ARTICLE IN PRESS

P.D. Alexander, I. Keay

[m3GeSsc;November 24, 2017;19:55] Explorations in Economic History 000 (2017) 1–15

shares (𝛼𝑛𝑠 ), for Canada and its trade partners. We also need information on Canada’s trade partners’ late nineteenth century industryspecific tariff rates, Canadian average weighted tariffs from before and after the policy change (𝑡𝑠𝑛 and 𝑡𝑠𝑛′ ), and estimates of industryspecific trade elasticities (−𝜃 𝑠 ).12 Because trade with Britain and the United States accounted for over 90% of Canadian imports and exports during the late nineteenth century, the relevant bilateral trade relationships are limited in our historical context – we focus on Canada, the United States, Britain, and the rest-of-the-world.13 Canada’s highly concentrated trade patterns ease the relatively extensive data requirements that must be fulfilled to measure welfare changes in a general equilibrium setting. 4.1. Measuring average weighted tariffs We aggregate individual products’ average weighted tariffs, derived from the Canadian Trade and Navigation Tables, up to the SIC2 industry level using both pre- and post-policy import value weights. Products with relatively large tariff increases under the National Policy have larger post-policy reductions in their import weights (conditional on the product’s import demand elasticity). Using post-policy weights, therefore, results in a more conservative measure of industry-specific tariff changes. Columns (1a), (1b) and (1c) in Appendix Table A1 report Canadian average weighted tariffs (𝑡𝑠𝑛 ) by industry in 1877 – the last full fiscal year before the National Policy was introduced – and 1880 – the first fiscal year after the policy was implemented. Post-policy import value weights are used to derive 1880 AWT in column (1b), while pre-policy weights are used in column (1c). On average, the manufacturing sector’s AWT rose by more than half – from 14.2% to 21.5% – with the introduction of the National Policy. Tariff rates for all 16 of the manufacturing industries increased, but the margin by which these rates rose was clearly not uniform across industries. The non-manufacturing industry’s tariff levels were slightly lower than many of the manufacturing industries’ tariffs, but the change in the non-manufacturers’ tariffs in 1879 was not significantly different than the manufacturing sector, as a whole.14 The use of prepolicy weights has a strong effect on our measure of 1880 AWT for Tobacco – dramatically raising measured tariffs – and Coal and Petroleum, and Chemicals – significantly lowering measured tariffs. The net effect of our choice between pre- and post-policy import weights on our welfare measures is small, and none of our qualitative conclusions are affected by this choice. With product-specific data on import values and duties collected from the Commerce and Navigation Tables of the United States, and the United Kingdom’s Overseas Trade Tables (supplemented with additional information from both countries’ annual statistical abstracts), we also calculate industry-specific average weighted tariffs for the US and Britain in 1880 – Appendix Table A1: columns (2) and (3). As closely as possible, our assignment of US and British HS6 codes, and our product aggregation (using 1880 import weights), mirrors the approach taken with the Canadian trade data. We assume that rest-of-the-world AWT can be approximated by an import weighted average of US and UK rates – column (4).15 In general, the US rates confirm that Canada’s fastest growing trade partner during our period of study was firmly protectionist, with US manufacturing tariffs averaging more than 37% in 1880. Even after the introduction of the National Policy, US AWT were higher than Canadian (and British) rates for all industries except Tobacco, Chemicals and non-manufactured products. Irwin (2010, Table A1) reports that on average over all import products, US tariffs fell slightly as the National Policy was introduced in Canada – from 29.2% in 1877 to 29.1% in 1880. In Britain, outside of revenue generating tariffs applied to cigars and cigarettes, roasted coffee, alcoholic beverages, and some paper products, we see an economy that was essentially entirely open to trade – no duties were collected for 14 of the 17 industries in 1880. Trade figures in Mitchell and Deane (1962: 282 and 394) reveal that over all import products, British tariff rates were much lower than American and Canadian rates throughout the late nineteenth century globalization period and, like US average rates, British AWT fell slightly in 1879 – from 5.1% to 4.7%. The increase in Canadian tariffs under the National Policy appears to have moved Canadian rates closer to American levels of protection, but the Canadian policy change was clearly not part of a broad, uniform global increase in rates in 1879. Together, the patterns that are evident from the tariff rates reported in Appendix Table A1 suggest that the National Policy marked a significant, but highly selective policy shift that was unique among Canada’s primary trade partners. 4.2. Measuring industry-specific bilateral trade, expenditure and tariff revenue shares 𝑠 ) are measured as expenditure by country n on industry Canadian, US, British, and RoW industry-specific bilateral trade shares (𝜋𝑛𝑖 s output, produced by each trade partner i (including home produced goods), as a share of total expenditure by n, where expenditure is defined as the gross value of domestic production for each industry in each year, less the value of industry exports, plus the value of industry imports. Expenditure shares (𝛼𝑛𝑠 ) are measured as domestic expenditure on each industry’s output goods as a share of

12 We note that Eq. (1) does not require time series evidence on production and trade shares. Realizations of the required variables over time reflect other shocks in addition to the National Policy. Therefore, looking at evidence that extends well beyond the policy change complicates identification of the effect of the policy. Instead, following Caliendo and Parro (2015), our preferred approach is to calculate welfare effects according to our general equilibrium model, with data on economic conditions for Canada and its trade partners at the time of the policy change, combined with data on tariffs from before and after the policy was introduced. 13 In 1880, for example, 90% of Canadian exports went to the US or UK – 38% and 52%, respectively – and 89% of Canadian imports originated in the US or UK – 41% and 48%, respectively. After the US and Britain, France was the next largest Canadian trade partner in 1880, accounting for 0.9% of exports and 1.6% of imports. 14 We note that after similar tariff increases in 1879, manufactured goods’ tariffs grew much faster than non-manufactured AWT through the 1880s, and the two sectors’ tariff levels diverged well into the twentieth century. Because non-manufacturing import goods are typically used as inputs into domestic manufacturing, this divergence is another indication of the newly selective protectionist objectives in Canadian policy. 15 Because RoW accounts for such a small share of Canadian trade, our qualitative conclusion regarding the welfare impact of the National Policy is not dependent on our assumption about RoW tariff rates. We still find small positive welfare effects even if we assume RoW AWT in 1880 were equal to US or British rates.

6

JID: YEXEH

ARTICLE IN PRESS

P.D. Alexander, I. Keay

[m3GeSsc;November 24, 2017;19:55] Explorations in Economic History 000 (2017) 1–15

aggregate domestic expenditure. Tariff revenue shares (𝜆𝑠𝑛 ) are measured as the total value of all duties collected on import goods produced (or potentially produced) by each industry as a share of domestic expenditure on each industry’s output. Industry-specific gross output data for Canada are taken from Urquart (1993), and Harris et al. (2015). US production data has been collected from the 1880 Census of Manufacturing, Jeremy Atack and Fred Bateman’s 1880 census manuscript samples (described in Atack and Bateman (1999)), and the Historical Statistics of the United States, Volume 4 (2006, Series Dd13-231, Dd309-365, Db157 and Db166). British industry-specific output figures for 1880 have been derived by linking industrial production indexes produced by Hoffmann (1965), Lewis (1967) and Feinstein (1972), to gross output values recorded in the first Census of the Production of the United Kingdom (1907). Guided by the gross output produced by a wide range of European nations for several key industrial commodities (Mitchell 2003: Europe, Series D8-26), we assume that rest-of-the-world industry-specific output matches the total value of US and UK production combined.16 RoW trade figures are calculated as residuals, after accounting for all bilateral trade between Canada, the United States, and Britain. Because high home consumption shares reflect low import penetration, one might reasonably predict that there should be a positive relationship between a country and industry’s AWT, and its home consumption share. In general, this prediction holds for Canada and its trade partners during our period of study. In Appendix Table A2: columns (1a), (2b), (3c) and (4d), we see that Canada’s home consumption share, averaged over all products, was just under 90% in 1880, Britain had the lowest home consumption share at just 83%, while the US had the highest at 96%. Across industries within Canada, we also see a correlation between high tariffs and high home consumption – Transport Equipment, Coal and Petroleum, and Tobacco, for example, all had relatively high tariff rates and/or large tariff increases under the National Policy, and all had home consumption shares well above the average. Similar patterns are apparent across US industries. We can also see the relatively high concentration of Canadian trade patterns in Appendix Table A2, with 94% of Canadian imports originating in either the US or UK – columns (1b), (1c), (1d) – and over 90% of Canadian exports destined for just these two markets – columns (2a), (3a), (4a). All trade partners’ import penetration ratios for non-manufactured goods were relatively low, although, perhaps not surprisingly given the low transport costs between Canada and the US, intra-continental raw material trade in North America appears to have been significant. Again, the heterogeneity across countries’ and industries’ trade shares motivates the adoption of a multi-sector general equilibrium model in our historic setting. In Appendix Table A3, columns (1a)–(1d) report industry expenditure shares for Canada and its trade partners in 1880. As we might expect, some manufacturing industries were considerably more important as a share of total expenditure than others. For example, Food and Beverages, Textiles, and Iron and Steel were large industries in all four partners’ economies, while Leather and Wood were very large in Canada, but less so in the US, UK and RoW. Rubber, Tobacco, Coal and Petroleum, Miscellaneous Products, and Paper consistently make up less than 1% of total domestic expenditure. Tariff changes that affect the largest industries will be relatively influential in our assessment of the impact of the National Policy on Canadian welfare, while tariff changes that affect the smallest industries will be relatively inconsequential. We also note that manufactured goods as a whole account for just over 44% of aggregate expenditure in Canada – considerably larger than the manufacturing shares in the US, UK and RoW, which range from just 17%–31%. Large expenditure on manufactured goods, particularly Iron and Steel, Transport Equipment, and Wood, reflects Canada’s intensive late nineteenth century accumulation of urban and transportation infrastructure (Green and Urquhart, 1994). Turning to tariff revenue as a share of total expenditures we see that even after significant tariff increases under the National Policy, Canadian tariff revenue shares were larger than its trade partners’ shares in 1880, particularly for Food and Beverages, Rubber, Clothing, Paper, and Non-Ferrous Metals (Appendix Table A3: columns (2a)–(2d)). In the US, high revenue shares are found in Textiles, Iron and Steel, and Non-Metallic Minerals. The impact of changes in tariff revenue on aggregate welfare following the adoption of protectionism will be larger in countries and industries where revenues are a larger share of expenditure. In general, Food and Beverages, Tobacco, and Miscellaneous Products have relatively large tariff revenue shares in Canada, in part reflecting the persistent presence of revenue objectives in Canadian policy. Revenue is relatively high when tariffs are applied to low trade elasticity products, such as alcoholic beverages, spices, or cigars and cigarettes. We also note that over all traded products, Canadian revenue shares were virtually unchanged by the introduction of the National Policy – 2.0% in 1877 versus 2.3% in 1880. This suggests that there is a rather limited role for tariff revenue effects in our measure of the impact of the National Policy on Canadian welfare. 4.3. Measuring trade elasticities The identification of trade elasticity parameters (𝜃 𝑠 ) poses a significant empirical challenge for the measurement of welfare changes in the wake of Canada’s adoption of protectionist trade policy in 1879. Others who have studied the welfare implications of late nineteenth century trade policy have typically relied on disaggregate elasticity estimates that are derived from more recent panel or time-series data sources.17 The use of modern estimates in our historical context requires that we assume that the degree of substitutability between domestic and foreign products has been constant over the very long run. While this assumption is unlikely to hold, the estimation of elasticities of substitution that are appropriate to our period requires information from the late nineteenth

16 Mitchell (2003) reports that the US and UK produced 75% of Europe’s pig iron, 60% of the continent’s steel, 63% of the raw cotton consumed, 93% of the sulphuric acid, 92% of the lumber, and 47% of the beer produced. The relatively small amount of trade passing between Canada and the RoW makes our conclusions about the welfare effects of the National Policy insensitive to our assumption about production levels in countries outside of Canada, the United States, and Britain. Assuming that the RoW produces only half the total value of US and British industry-specific production does not affect any of our qualitative conclusions. 17 For example, Beaulieu and Cherniwchan (2014) use Kee et al. (2008) panel estimates, and Irwin (2010) uses four sets of time-series estimates, all derived from data spanning the post-1970 period.

7

JID: YEXEH

ARTICLE IN PRESS

P.D. Alexander, I. Keay

[m3GeSsc;November 24, 2017;19:55] Explorations in Economic History 000 (2017) 1–15

and early twentieth centuries about domestic prices, import unit values, tariffs and other trade costs, bilateral trade volumes, and domestic expenditure. Typically this information either does not exist, or it is only available for very few countries, across short time periods, at national or sectoral levels of aggregation. The use of limited cross-sections, or panels of highly aggregated data, is problematic because it is unlikely to result in precise elasticity estimates, and aggregate international evidence yields results that do not reflect variation in substitutability among products within a particular country, or a particular industry.18 For our welfare measures, we use both modern, highly disaggregated elasticity estimates derived from panel data sources that allow for the inclusion of explicit controls for multilateral resistance and historically contemporaneous, but relatively aggregate import demand elasticities that we estimate with newly compiled time-series evidence on Canadian prices, production, and trade flows from the 1880–1913 period. Kee et al. (2008) estimate country-specific import demand elasticities at the six-digit product level using international panel data from 1988 to 2001, with an approach that nets out multilateral resistance. Ossa (2015) also estimates modern, disaggregate elasticities, using Feenstra (2010) method which relies on variation in demand and supply shocks across countries from 1994 to 2008.19 The advantage of using these modern values for 𝜃 𝑠 in our historical context is that they are estimated using cross-country international data, which reduces the likelihood that they are affected by country-specific shocks, they are based on structural estimation approaches that are theory consistent, incorporating controls for the effects of multilateral resistance, and they are product-specific, which allows for an investigation of the welfare effects of tariff changes at a fine, granular level of detail.20 The disadvantage of using these modern estimates is that differentiation and substitutability among products was likely to have been quite different at the end of the nineteenth century relative to the beginning of the twenty-first century, and Canadian-specific elasticities are likely to deviate from those estimated with international panel data (Hillberry and Hummels, 2013: 1245–1246).21 We report the Kee et al. (2008) and Ossa (2015) elasticity parameters, aggregated up to SIC2 manufacturing industries using product-specific import shares as weights, in Appendix Table A4: columns (4) and (5). 15 of the 17 Kee-Nicita-Olarreaga modern industry-level estimates lie within the [2.8, 0.2] interval (in absolute value) identified by Marquez (1999: 102) as the range of Canadian elasticities reported in 19 published sources between 1946 and 1994. The weighted average Kee-Nicita-Olarreaga elasticity parameter is 1.87 over all manufactured products, and 1.82 over all traded products. Ossa’s estimates are considerably higher, with all industries’ elasticity parameters lying above 2.8, and the weighted average over all manufactured products equal to 4.84 (4.64 over all traded products). As an alternative to these modern 𝜃 𝑠 estimates, we use time-series evidence to estimate historically contemporaneous, but more aggregate home-foreign import demand elasticities derived from Shiells et al. (1986) flexibe, log-linear Armington (1969) import demand functions.22 Irwin (2000a) and Inwood and Keay (2013) use this specification to estimate historical elasticities of substitution for late nineteenth century US and Canadian iron and steel producers. Because we do not have HS6 product-specific production or price data for Canada during our period, we estimate import demand functions using more aggregated annual information from 1880 to 1913 for each of our 17 manufacturing and non-manufacturing industries. For each industry (s) we estimate an import demand function that takes the form: ( 𝑠( )) ( 𝑠) 𝑃𝑛𝑖 1 + 𝑡𝑠𝑖 𝑃𝑛 ( 𝑠) ( ) ( ) 𝑠 𝑙𝑛 𝑋𝑛𝑖 = 𝛽0 + 𝛽1 𝑙𝑛 + 𝛽2 𝑙 𝑛 + 𝛽3 𝑙𝑛 𝑋𝑛𝑠 + 𝛽4 𝑙𝑛 𝐿.𝑋𝑛𝑖 (5) + 𝜖𝑠 𝑃𝑛 𝑃𝑛 𝑠 Where time subscripts are suppressed, country n is Canada, i denotes an import weighted average of the US, UK and RoW, 𝑋𝑛𝑖 denotes the nominal value of imports, 𝑃𝑛𝑠 is a domestic industry price index, Pn is a domestic aggregate wholesale price index for all products, 𝑃𝑛𝑖𝑠 is the import unit value (pre-tariff value of imports divided by quantity of imports), 𝑡𝑠𝑛 denotes the Canadian AWT, 𝑋𝑛𝑠 𝑠 denotes lagged imports, and 𝜖 s is a regression residual. Trade figures are taken from the is aggregate domestic expenditure, 𝐿.𝑋𝑛𝑖 Canadian Trade and Navigation Tables, and gross output comes from Urquart (1993). The sources for domestic prices, and the aggregate

18 Elasticity estimates derived from disaggregate data are also less prone to aggregation bias (Imbs and Mejean, 2015). For more on the impact of elasticity choice in a historical context, see Federico and Vasta (2015). 19 Other recent estimates that include explicit controls for multilateral resistance are provided by Simonovska and Waugh (2014), who estimate that the aggregate international trade elasticity parameter in 2004 was approximately 4.0, and Caliendo and Parro (2015), who report modern trade elasticity parameters derived from measures of import sensitivity to variation in tariffs across countries and time, that range from 8.62 for manufactured products to 4.55 for all traded products. Because Simonovska and Waugh do not report elasticities by industry, and because we cannot match Caliendo and Parro’s disaggregate estimates to our SIC2 classifications, we do not report welfare effects that rely on these estimates. 20 Variation in general equilibrium welfare measures due to aggregation are not reported in this paper, however, we find that aggregation from the HS6 product level to the SIC2 industry level reduces measured partial equilibrium DWL/GDP from 1.2% to 0.9% in 1880. 21 Klasing et al. (2015, Table 1) show that international foreign-foreign elasticities of substitution were lower (closer to zero) during the 1870–1913 period, relative to the inter-war and post-WW2 periods. Kee-Nicita-Olarreaga and Ossa’s estimates are generally higher than those reported by, for example, Stern et al. (1976), which are estimated for the US with data from the late 1960s and early 1970s, and Shiells, Stern and Deardorff’s (1986) elasticities, that use data for the US from 1962 to 1978. The estimates from the 1960s and 1970s are again higher than Irwin (2000a) historical estimates for US iron and steel from the late nineteenth and early twentieth centuries, and our OLS estimates for Canada from 1880 to 1913. 22 By “historically contemporaneous” we mean estimated with historical data and hence contemporaneous with our period of study. Hillberry and Hummels (2013, 1224–1227) provide a detailed discussion of the problems associated with using time-series evidence to estimate trade elasticities from import demand functions. They emphasize the importance of controlling for multilateral resistance, which requires sufficient cross-section or panel data for precise estimation of foreign-foreign elasticities of substitution. In most historical contexts, bilateral data and cross-section variation in trade costs are limited, which constrains our ability to control for multilateral resistance. Hillberry and Hummels (2013: 1240–1242) suggest that import demand specifications estimated with time-series data are most problematic when endogeneity is ignored, long run substitutability is constrained, or when trade elasticities derived from these specifications are used to measure persistent, long run welfare effects. Our import demand specifications and estimation approaches have been adopted in an effort to mitigate some of the concerns articulated in the Hillberry–Hummels critique.

8

JID: YEXEH

ARTICLE IN PRESS

P.D. Alexander, I. Keay

[m3GeSsc;November 24, 2017;19:55] Explorations in Economic History 000 (2017) 1–15

manufacturing wholesale price index are described in Harris et al. (2015). Import unit values are calculated from the Canadian trade tables and indexes reported in Taylor (1931).23 Lagged imports are included in Eq. (5) to allow for the possibility that import demand responses adjust to price changes over more than one period (Hillberry and Hummels, 2013: 1241). Estimates of 𝛽 2 , therefore, only capture immediate, short run responses in Canadian import demand. Longer run substitutability and adjustment is allowed for when we measure import demand elasticity paramters as 𝛽2 ∕(1 − 𝛽4 ).24 We estimate Eq. (5) by ordinary least squares (OLS) with robust standard errors for each industry, and for the manufacturing sector as a whole (including industry fixed effects, clustering standard errors by industry – see Supplementary Materials Table S1). The long run import demand elasticity parameters (and standard errors) derived from these estimates are reported in Appendix Table A4: column (1). 15 of the 17 industry-level OLS elasticity paramter estimates are statistically distinguishable from zero, and all but one (Paper) lie within Marquez’s [2.8, 0.2] interval. Our OLS import demand elasticity paramter estimate of 𝜃 𝑠 is 1.10 for the manufacturing sector as a whole, and 2.18 for all traded products. When we compare our OLS estimates to the Kee-Nicita-Olarreaga and Ossa modern estimates (columns (4) and (5) in Appendix Table A4), we see that although the rank correlations across industries are high, all 17 of the Ossa industry-level estimates, and eight of the 17 Kee-Nicita-Olarreaga estimates, are larger than our historical estimates, and our OLS estimate for the manufacturing sector as a whole is statistically significantly lower than even the Kee-Nicita-Olarreaga sectoral estimate (1.10 versus 1.87). This is perhaps not surprising if we believe that there were fewer domestic substitutes for manufactured goods being imported into Canada in the late nineteenth and early twentieth centuries, than there were in the late twentieth and early twenty-first centuries. However, it is also possible that the lower OLS estimates could reflect the presence of simultaneity and/or selection bias (Kee et al., 2008, 672; Hillberry and Hummels, 2013, 1224–1227). More specifically, our OLS estimates could be biased downwards (closer to zero) due to the supply-side relationship between import volumes and prices, or because of the possible selection of inherently inelastic goods into the export market.25 To address these identification issues, we adopt a two-stage least squares instrumental variable (IV) approach, instrumenting for domestic prices (IV1), and both domestic prices and import unit values (IV2), in our import demand functions. Following Irwin (2000b, 285), we use Canadian and US raw material prices, fuel prices, unskilled wage indexes, and a user cost of capital as instruments. The plausible exogeneity of these instruments is based on the assumption that international markets for raw materials, coal, labor, and capital were well integrated over the 1880–1913 period (Coe and Emery, 2004; Emery and Levitt, 2002; Inwood and Keay, 2015; Keay and Redish, 2004). Input market integration implies that industries face relatively elastic input supply curves, such that industry-specific domestic import volumes would be inconsequential in the determination of the input price instruments. Our claim, therefore, is that Canadian and US input prices were important determinants of industry-specific domestic output prices and import unit values, but they were not themselves strongly affected by the volume of Canadian industry-specific imports.26 In the online supplementary materials (Table S1) we report both the OLS and second stage IV estimates for the parameters from the import demand Eq. (5), estimated with data from 1880 to 1913, over all 17 industries.27 This table also includes first stage Shea’s partial R2 ; p-values from Hansen valid instrument tests and Angrist–Pischke weak instrument tests, which reveal that in our panel regressions the instruments are both statistically strong and valid; and p-values from Hausman exogeneity tests, which reveal that exogeneity cannot be rejected for import unit values (IV2).28 The import demand parameter estimates across all three specifications are consistent with Armington (1969) theoretical predictions, and as expected, the domestic price and import unit value estimates are larger (farther from zero) when we instrument. In Appendix Table A4: columns (2) and (3), we report the industry-specific long run import demand elasticity parameters (and standard errors) derived from our two-stage IV estimates of Eq. (5). Relative to the OLS estimates, when we instrument for domestic prices alone (IV1), our long-run elasticity paramters are larger for all traded goods (2.69 versus 2.18), and for 12 of the 17 industries. 14 of the 17 IV1 estimates are statistically distinguishable from zero, all lie within the [2.8, 0.2] interval, and 10 of the industry-level IV1 estimates are larger than the Kee-Nicita-Olarreaga modern estimates. When we instrument for both domestic prices and import unit values (IV2), the manufacturing elasticity parameters fall slightly relative to the IV1 estimates, but they are still larger than the OLS estimates on average (2.83 versus 2.18), and for 11 of the 17 industries they are larger than both the OLS and Kee-NicitaOlarreaga estimates. We note that both our IV1 and IV2 estimates are still significantly smaller than Ossa’s modern 𝜃 𝑠 figures and 23 Sensitivity tests using alternate aggregate price indexes (CPI and GDP deflator, for example); including a linear time trend; or dropping lagged imports, do not have any qualitative impact on our elasticity estimates. 24 𝑠 ) + 𝛽3 (Irwin 2000b: 285). Armington’s elasticity of substitution parameter can be derived from (5) as: 𝜎 𝑠 = (𝛽1 ∕𝜋𝑛𝑛 25 Imbs and Mejean (2015) discuss the selection bias issue in detail, and provide a possible correction. 26 In addition to the market integration references provided in the text, we note that the time-series correlation between Canadian, US and British iron ore and coal prices, unskilled manufacturing wage indexes, and industrial bond yields between 1870–1913 are, in almost all cases, strongly statistically significant and well over 0.80 (Inwood and Keay, 2015: data appendix). We also note that the terms of trade effects that contribute to our measure of Canadian welfare changes are associated with Canadian import products that were disproportionately finished manufactured goods with relatively low elasticities of substitution, rather than raw material inputs (Alexander and Keay, 2017). Costinot and Rodriguez-Clare (2014) show that for smaller countries that are active on international markets, even small price effects for these import products can lead to large changes in welfare. 27 Use of data from the pre-policy (1870–1877) period does not have a qualitative effect on our elasticity estimates, but there is evidence of a structural break in some of the parameters in 1880. Industry fixed effects are included and robust standard errors are clustered by industry in the estimates reported in online appendix Table 1. We also report the long run aggregate import demand elasticities and long run Armington elasticities of substitution implied by the reported parameter estimates. 28 Given Canada’s relatively small size in international markets, it is not surprising that Canadian import volumes do not appear to strongly affect import unit values. We report import demand elasticities from specification IV2, which treats domestic prices and import unit values as endogenous, even though our panel estimates cannot reject exogeneity for import unit values, because for some individual industries the results from these tests are marginal. The elasticity estimates from IV2, therefore, can be treated as robustness checks for our preferred IV1 specifications.

9

ARTICLE IN PRESS

JID: YEXEH P.D. Alexander, I. Keay

[m3GeSsc;November 24, 2017;19:55] Explorations in Economic History 000 (2017) 1–15

Table 1 Impact of the National Policy on Canadian welfare. IV1 (2)

IV2 (3)

Kee et al. (2008) (4)

Ossa (2015) (5)

𝜃 𝑠 = 5.5 (6)

Panel A: general equilibrium %Δ welfare/GDP National Policy 0.161 Unilateral free trade −0.678 Global free trade 0.165

0.152 −0.655 0.174

0.137 −0.607 0.207

0.141 −0.609 0.248

0.032 −0.332 0.330

0.000 −0.232 0.378

Panel B: partial equilibrium %Δ DWL/GDP National Policy −0.461

−0.569

−0.599

−0.386

−0.983

−1.164

Panel C: optimal tariff (%) Optimal tariff

52.37

49.54

56.96

23.33

16.67

OLS (1)

82.08

Note: Detailed description of welfare and optimal tariff measurement provided in text. Values in column (1) are constructed using OLS estimates of historic import demand elasticities. Values in column (2) are constructed using IV estimates of historic import demand elasticities, instrumenting for domestic prices. Values in column (3) are constructed using IV estimates of historic import demand elasticities, instrumenting for domestic prices and import unit values. Values in column (4) are constructed using import-weighted industry average modern import demand elasticities from Kee et al. (2008). Values in column (5) are constructed using import-weighted industry average modern import demand elasticities from Ossa (2015). Values in column (6) are constructed using import demand elasticity parameters set equal to 5.5 for all industries. In Panel A, rows 1, 2, and 3 correspond to Canadian tariff changes associated with the 1879 National Policy, unilateral free trade, and global free trade, respectively. Panel B reports partial equilibrium welfare losses derived from a standard DWL measure (Beaulieu and Cherniwchan, 2014; Irwin, 2010), constructed with the same trade shares, AWT, and import demand elasticities used in the general equilibrium calculations. Panel C reports an approximation of the welfare maximizing optimal tariff implied by our general equilibrium model (Costinot and Rodriguez-Clare, 2014, 227)

some of the more aggregate, international trade elasticity paramaters reported in the recent empirical trade literature (Caliendo and Parro, 2015; Klasing et al., 2015; Simonovska and Waugh, 2014). This may raise some concern that our Canadian time-series import demand IV specifications are not allowing for sufficient long run adjustment, or addressing all potential sources of bias. For example, our historical estimates could reflect inelastic import responses to cyclical price dynamics that may be imbedded in our late nineteenth and early twentieth century Canadian data.29 Because the more structural, theory consistent Kee-Nicita-Olarreaga and Ossa estimates are derived from international panel data, they allow for explicit controls for multilateral resistance, they reflect more generalized foreign-foreign elasticities of substitution, and they capture longer-run adjustments associated with gravity-type indicators. Both as a test of the sensitivity of our welfare results, and because the modern, OLS, and IV historical estimates are capturing trade responses from slightly different perspectives (reflecting differences in the characterization of trade costs, and differences in perceived product differentiation and substitutability), we measure the welfare effects of Canada’s move to protectionist trade policy in 1879 using all five sets of elasticity parameters reported in Appendix Table A4.30 5. Results We calculate the change in Canadian welfare that resulted from a unilateral increase in domestic tariffs in 1879 by solving the system of equations derived from our general equilibrium model, using the tariff rates, elasticities, and trade, expenditure and tariff 𝑠 , 𝜋 𝑠 , 𝑡𝑠 and 𝑡𝑠′ into (3) to solve for values of 𝑌̂ . shares reported in Appendix Tables A1–A4. We first substitute values for 𝑌𝑛 , 𝛼𝑛𝑠 , 𝜋𝑛𝑛 𝑛 𝑛 𝑛𝑖 𝑛 These measures are then substituted into (1)–(4) to find 𝑊̂ 𝑛 . With this approach we are assuming that all other international market conditions evolve as observed in the data. Therefore, for example, our welfare calculation relies on the assumption that the transition to protectionism in Canada under the National Policy did not significantly affect the evolution of non-tariff trade costs, or trade policy in other countries.31 In Table 1: Panel A we report our general equilibrium measure of the change in Canadian welfare resulting from the introduction of protectionist tariffs in 1879, the change in Canadian welfare that would have resulted from a unilateral reduction in Canadian

29

For a detailed discussion of this potential source of bias, and its implications, see Ruhl (2008). Data limitations constrain our ability to employ more structural estimation approaches in our historical context. However, using the data from our welfare calculations, we can compile a single cross-section of bilateral trade flows for Canada and its main trade partners (US, UK and RoW) in 1880. Combining these bilateral trade flows with information on importer and industry specific ad valorem tariff rates from 1880, and partner-specific ad valorem shipping costs for pig iron (Inwood and Keay, 2015), we can estimate a historically contemporaneous, aggregate foreign-foreign elasticity of substitution. Our theory consistent cross-section specification, taken from Hillberry and Hummels (2013, Eq. (18.13)) and Head and Mayer (2014: Eq. (22)), includes a full set of importer-exporter (or bilateral pair) fixed effects to explicitly control for multilateral resistance. Even when we use all available bilateral flows, and estimate an aggregate elasticity over all 17 SIC2 industries, we only have 204 observations available for estimation, and as a result, our cross-section estimates - reported in the online supplementary materials (Supplementary Materials Table S2) - are considerably less precise than our time-series import demand elasticities. Our preferred estimate, which uses all available data and importer-exporter FE, yields a historically contemporaneous trade elasticity parameter for Canada and its main trade partners equal to 3.51 (with a robust standard error of 2.06). This estimate is smaller than Ossa’s modern aggregate elasticity of substitution parameter (4.64), but it is larger than, although statistically indistinguishable from, either of our IV1 or IV2 aggregate import demand estimates (2.68 and 2.33, respectively). We do not use the aggregate historical elasticity of substitution estimates in our welfare calculations due to the imprecision in their estimation, and because we cannot use this approach to estimate industry-specific elasticities. 31 In support of this assumption, we note that aggregate AWT for the US and Britain, the two countries most affected by Canada’s adoption of protectionist policy objectives, fell by less than one half of one percentage point in 1879; over the full 1870–1913 period, the time-series correlation between Canadian-US and CanadianUK AWT is just −0.23 and −0.32, respectively; and there are no significant changes in the linear time trend in British or global freight rates associated with any of the six major tariff schedule revisions undertaken by Canada through the 1870–1913 period (Jacks and Pendakur, 2010; Mohammed and Williamson, 2004). 30

10

JID: YEXEH

ARTICLE IN PRESS

P.D. Alexander, I. Keay

[m3GeSsc;November 24, 2017;19:55] Explorations in Economic History 000 (2017) 1–15

tariffs to zero in 1879, and the change in welfare that would have resulted from a multilateral reduction in all global tariffs to zero in 1879. These welfare changes are calculated using our OLS and IV historically contemporaneous 𝜃 s estimates (columns 1, 2 and 3), KeeNicita-Olarreaga and Ossa’s modern elasticity of substitution parameters (columns 4 and 5), and a set of fixed elasticity parameters, equal to 5.5 for all industries (column 6). The fixed 𝜃 𝑠 = 5.5 is the ‘break even’ elasticity parameter at which our general equilibrium measure of the welfare impact of the National Policy falls to zero. Holding all else in the model constant, any average trade elasticity parameter larger than this break even value result in a reduction in Canadian welfare following the adoption of protectionism, while all values below 5.5 result in an increase in Canadian welfare.32 To emphasize the importance of adopting a general equilibrium approach to welfare measurement in our historical context, in Table 1: Panel B we also report the change in partial equilibrium deadweight loss resulting from the introduction of the National Policy. The DWL figures use the same elasticity estimates, trade shares, AWT, and expenditure values that we use in our general equilibrium model, but they are derived with a standard Anderson–Neary partial equilibrium approach (Beaulieu and Cherniwchan, 2014: 152). Because our general equilibrium model captures the terms of trade and tariff revenue effects of the National Policy and the partial equilibrium distortionary DWL effects, the differences between the results reported in Panels A and B reflect the positive welfare effects that the partial equilibrium approach misses. In the theoretical environment described in Section 3 (and the online Technical Appendix), we can derive an expression for the optimal welfare maximizing tariff that is decreasing in 𝜃 s and, because larger countries have a greater scope for influencing domestic and foreign demand, increasing in country size (Ln ). A close approximation of the optimal tariff for a given industry s in our model is: 1 𝑡𝑠𝑛 ∗ = 𝑠 𝑠 (6) 𝜃 𝜋𝑖𝑖 Where 0 < 𝜋𝑖𝑖𝑠 ≤ 1 denotes the share of all other countries’ spending on all other countries’ goods in industry s.33 For larger countries, 𝜋𝑖𝑖𝑠 is smaller and as a result, the optimal tariff will be higher. For Canada, which was a small open economy in the late nineteenth century, 𝜋𝑖𝑖𝑠 was close to 1. Since our estimates of 𝜃 s are lower than 10 and lie above 1 in almost all cases, the optimal tariff for Canada in our historical context will necessarily be strictly greater than zero, but less than 100%. This finding is not consistent with predictions from partial equilibrium approaches, which suggest that the optimal tariff for all small open economies should be zero because the minimization of DWL requires that 𝑡𝑠𝑛 ∗ = 0 in these models, even when tariffs generate revenue. The more conventional partial equilibrium results are based on the assumption that small economies exercise no market power, and hence have no influence on their terms of trade.34 In Table 1: Panel C, for each set of trade elasticity parameters, we report an approximation of the welfare maximizing optimal tariff implied by our general equilibrium model. Our preferred measure of the static general equilibrium welfare impact of the National Policy is reported in the first row of column (2) in Panel A, which uses the industry-specific historic IV elasticity estimates that instrument for domestic prices alone. We find that the National Policy tariff raised welfare in Canada in 1879 by 0.15% of GDP. This welfare effect is equivalent to $2.3 billion 2016 USD. The first row of columns (1), (3) and (4) report the impact of the National Policy using our OLS elasticities, our IV estimates that instrument for domestic and import prices, and the Kee-Nicita-Olarreaga modern estimates, respectively. In all three cases, we find that the National Policy improved Canadian welfare – by 0.16%, 0.14% and 0.14% of GDP. With Ossa’s import demand elasticity parameters, which are considerably larger than any of our other estimates, the partial equilibrium distortions are larger, the optimal tariff is lower, and the National Policy appears to improve Canadian welfare by only 0.03% of GDP (column 5), which is less than the other measures but, notably, still positive. In the first row of the last column in Table 1: Panel A, we show that our measure of the change in Canadian welfare following the adoption of the National Policy tariffs in 1879 falls to zero – implying that the positive general equilibrium terms of trade and tariff revenue effects of the National Policy were perfectly offset by the distortionary DWL effects – when average trade elasticity parameters reach 5.5 for all SIC2 industries.35 In the second row of Table 1: Panel A, we report the impact of a unilateral reduction in Canadian tariffs to zero in 1879, for each of our six sets of elasticity estimates. Again, the results are consistent with our intuition in a general equilibrium setting. Because the optimal tariff for Canada is above the observed AWT for most industries prior to the move to protectionism in 1879, welfare would have fallen if, rather than raising tariffs, Macdonald’s Conservative government had adopted a unilateral free trade policy. The welfare loss in this counterfactual case ranges from −0.68% to −0.23% of GDP, with higher trade elasticity parameters associated with smaller welfare losses. In the third row of Panel A, we report the welfare impact of a multilateral move to free trade in 1879. Under all six elasticity estimates, reducing Canadian, US, UK and RoW tariffs to zero in 1879 would have improved Canadian welfare substantially more than the change following the introduction of the National Policy, or unilateral Canadian liberalization. Although our model suggests that the optimal tariffs for Canadian industries were well above zero in 1879, this is conditional on the maintenance of observed tariff

32 We note that 5.5 far exceeds our historical estimates and most (but not all) of the trade elasticity parameters reported in the more recent empirical trade literature – figures in excess of 5.5 are reported in Caliendo and Parro (2015), and Eaton and Kortum (2002), for example. 33 A similar expression to (6) appears in Costinot and Rodriguez-Clare (2014). See also Felbermayr et al. (2013), Demidova and Rodriguez-Clare (2009), Alvarez and Lucas (2007), or Gros (1987) for more on small open economies’ optimal tariffs. 34 Broda et al. (2008) show that both large economies and, although to a lesser extent, small economies exercise international market power due to specialization in differentiated industries. 35 To provide some sense of the size of these welfare effects in a modern context, consider Caliendo and Parro (2015, Table 2), who report that Canadian welfare dropped by 0.06% of domestic GDP following NAFTA’s tariff reductions; or Costinot and Rodriguez-Clare (2014, Table 4.2), who report that US welfare would rise by 0.26% of GDP following a unliateral increase in protective tariffs to 40%.

11

JID: YEXEH

ARTICLE IN PRESS

P.D. Alexander, I. Keay

[m3GeSsc;November 24, 2017;19:55] Explorations in Economic History 000 (2017) 1–15

rates among Canada’s trade partners. For all countries in a general equilibrium setting, the negative impact of foreign tariffs will be significant. Global free trade removes all tariff distortions, such that, in our Canadian historical context, the reduction from observed global tariff rates in 1877 to 0, improves domestic welfare by between 0.38% and 0.17% of GDP, depending on how substitutable Canadian and foreign products were. These relatively large welfare effects reflect both high international (and Canadian) AWT, and high import penetration during our period. The welfare effects reported in Panel A are clearly not consistent with the more conventional partial equilibrium estimates that suggest that the welfare impact of the National Policy was strongly negative. (Beaulieu and Cherniwchan, 2014, 146), for example, report that, “...the static welfare losses arising from protectionism [in Canada during the first era of globalization] amounted to 0.7%–1.5% of GDP.” When we use our trade, production, and elasticity figures to measure partial equilibrium deadweight loss due to the National Policy, we also find reductions in welfare that range from −0.46% to −1.16% of GDP, with DWL rising as trade elasticity parameters rise (Table 1: Panel B). The differences between our general and partial equilibrium results imply that the positive terms of trade and tariff revenue effects of Canada’s adoption of protectionism (using our preferred elasticity estimates), amount to just over 0.72% of GDP (= 0.15% – (−0.57%)). On average over all manufactured imports, the National Policy raised Canadian tariffs from approximately 14% to just under 22%. In Table 1: Panel C, we report an approximation of the optimal welfare maximizing tariff implied by each of our trade elasticity estimates, in our general equilibrium theoretical environment. As our elasticity parameter estimates rise, the optimal tariff falls, from 82% using our OLS estimates, to 23% using Ossa’s figures. Our IV1 historical trade elasticity estimates suggest an optimal tariff of 50% for Canada in 1879, while a trade elasticity parameter of 5.5 for all industries pushes the optimal tariff down to 17%, which we note is still above the average tariff that was in place in Canada prior to the National Policy, and far above the 0% optimal tariff that partial equilibrium approaches recommend. Because the National Policy moved most industries’ tariff rates closer to their (general equilibrium) optimal levels, despite the tariffs’ distortionary price effects, welfare improved for Canadians as a result of this policy change.

6. Conclusion The late nineteenth and early twentieth centuries were a period of rapid international market integration and globalization. Canada, as a small open economy, faced falling international transport costs, rising industrial import competition, and rising import penetration during the 1870s. In this environment, John A. Macdonald’s Conservative government sought to protect domestic producers and improve Canadian welfare in the face of a highly protectionist southern neighbor, and rapidly integrating global economy, by prioritizing protectionist goals in a comprehensive revision of the Canadian tariff schedule. Under the National Policy, introduced in March 1879, the Canadian average weighted tariff on all import products rose from 14% to just over 21%, and it remained well above its 1877 level until at least World War 1. In this paper, we present evidence, based on a multi-industry, differentiated product general equilibrium trade model, and newly compiled, highly granular trade and production data, which suggests that the static welfare impact of the National Policy was in fact positive for Canadians. Specifically, our findings suggest that the tariff increases under the National Policy improved Canadian welfare in 1879 by approximately 0.15% of GDP – equivalent to $2.3 billion in 2016 USD. To be clear, our findings also support the view that global free trade would have led to the largest improvements in Canadian welfare – well above the gains from introducing the National Policy. However, given that outside of the UK, most of Canada’s trade partners were increasingly protectionist through this era, our theoretical approach, and our evidence, suggests that the National Policy was a welfare-enhancing policy for Canada. The National Policy was a clean, abrupt policy discontinuity, the Canadian trading environment during the late nineteenth century was uncomplicated, and highly granular trade and production data is available for Canada and its trade partners. These features not only allow us to illustrate the greater flexibility inherent in a general equilibrium approach to welfare measurement, relative to the more conventional partial equilibrium approaches, but we can also quantify the extent to which deviations from contemporaneous, industry and country-specific trade elasticity estimates affect our welfare assessments. The historical context clearly matters, and any modern parallels are not perfect, but our findings do more than simply reveal something new about the welfare impact of late nineteenth century Canadian trade policy – they also provide us with an example of an economic environment in which both unilateral increases in trade protection and multilateral free trade were potentially welfare improving policy choices.

Acknowledgements The authors would like to thank Kris Mitchener and three anonymous referees for valuable suggestions and improvements. Alex Chernoff, Kris Inwood, Oleksiy Kryvtsov, Césaire Meh, Rhys Mendes, participants at the 2016 Canadian Economic Association conference, the 2016 Carleton Macroeconomics and Finance conference, and seminar participants at the Bank of Canada provided helpful comments. Katrina Brazzell, Bill Dorval, Hannah Keay, and Xiangtao Ming contributed excellent research assistance. The views expressed in this paper are our own and do not necessarily reflect those of the Bank of Canada.

Supplementary material Supplementary material associated with this article can be found, in the online version, at 10.1016/j.eeh.2017.11.002 12

ARTICLE IN PRESS

JID: YEXEH

[m3GeSsc;November 24, 2017;19:55]

P.D. Alexander, I. Keay

Explorations in Economic History 000 (2017) 1–15

Appendix A. Tables

Table A1 Average weighted tariffs (𝑡𝑠𝑛 ), by country and industry (%). Canada

Food and Beverages Tobacco Rubber Leather Textile Clothing Wood Paper Printing and Publishing Iron and Steel Transport Equipment Non-Ferrous Metals Non-Metallic Minerals Coal and Petroleum Chemicals Miscellaneous Manufacturing Non-Manufacturing Total

US

UK

RoW

1877 (1a)

1880Post (1b)

1880Pre (1c)

1880 (2)

1880 (3)

1880 (4)

15.7 50.3 17.5 14.9 16.3 17.2 8.1 16.9 6.5 10.8 3.1 7.7 16.8 22.3 9.2 14.1 14.2 12.3 14.2

31.2 59.2 25.1 18.9 22.3 24.4 13.0 23.9 15.3 13.2 24.7 16.8 23.1 31.7 13.7 22.0 21.5 19.1 20.8

27.1 82.9 25.2 20.9 23.7 23.2 13.4 22.1 15.9 15.3 23.7 15.9 23.8 17.9 10.1 25.2 21.5 25.4 22.5

39.7 75.5 35.0 28.9 52.0 29.5 21.6 33.1 21.8 42.2 35.0 18.0 48.3 42.3 10.1 24.3 37.3 25.1 29.1

10.3 277.8 0.0 0.0 0.0 0.0 0.0 4.6 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 8.9 0.0 4.7

27.2 156.3 9.9 15.7 12.1 15.5 15.0 23.7 13.0 15.5 10.5 4.0 15.8 25.8 3.7 23.0 21.5 17.0 20.0

Note: Calculation of AWT described in text. Industries are defined by 1948 2-digit SIC classifications. HS6 products are aggregated up to SIC2 level using import value weights. (1b) uses post-policy weights (1880); (1c) uses pre-policy weights (1877). RoW AWT measured as import value weighted average of US and UK rates. Source: Import values and tariff revenues from Canadian Trade and Navigation Tables (1877–1880); Commerce and Navigation Tables of the United States (1880); United Kingdom’s Overseas Trade Tables (1880). Table A2 𝑠 ), by country and industry (%). Bilateral trade shares in 1880 (𝜋𝑛𝑖 Canada

Food and Beverages Tobacco Rubber Leather Textile Clothing Wood Paper Printing and Publishing Iron and Steel Transport Equipment Non-Ferrous Metals Non-Metallic Minerals Coal and Petroleum Chemicals Miscellaneous Manufacturing Non-Manufacturing Total

US

UK

RoW

Canada (1a)

US (1b)

UK (1c)

RoW (1d)

Canada (2a)

US (2b)

UK (2c)

RoW (2d)

Canada (3a)

US (3b)

UK (3c)

RoW (3d)

Canada (4a)

US (4b)

UK (4c)

RoW (4d)

88.9 92.6 62.3 95.7 67.7 63.6 95.9 74.9 89.2 75.9 97.0 61.6 80.3 92.8 66.4 60.4 82.1 95.1 89.3

7.9 5.1 29.0 2.2 9.0 6.3 3.9 13.2 5.8 11.7 0.8 14.0 7.8 6.3 12.6 21.6 7.4 3.8 5.4

1.7 0.6 8.6 1.3 23.1 29.3 0.1 11.2 4.2 11.9 2.2 21.0 8.3 0.9 13.7 14.4 9.5 0.9 4.7

1.4 1.7 0.1 0.7 0.2 0.8 0.1 0.7 0.8 0.5 0.0 3.5 3.6 0.0 7.3 3.6 1.1 0.2 0.6

0.6 0.0 0.0 0.4 0.0 0.0 3.1 0.0 0.0 0.1 0.0 0.0 0.2 0.1 0.1 0.1 0.5 0.1 0.2

96.1 96.8 94.4 92.4 73.4 86.1 96.3 96.0 95.8 81.9 99.1 75.7 51.7 97.9 46.6 75.4 85.6 97.5 96.3

0.1 0.0 4.1 0.0 12.1 0.4 0.1 1.1 3.6 12.7 0.0 20.0 9.9 0.1 22.1 3.1 5.3 0.0 0.8

3.2 3.2 1.4 7.2 14.5 13.5 0.4 2.9 0.5 5.4 0.9 4.2 38.2 1.9 31.2 21.4 8.7 2.4 2.8

0.7 0.0 0.0 0.0 0.0 0.0 3.9 0.0 0.0 0.0 0.1 0.0 0.0 0.0 0.2 2.2 0.4 0.1 0.3

6.6 0.4 0.0 1.0 0.1 0.0 1.5 0.6 0.3 0.5 0.1 0.1 0.1 16.2 2.9 4.9 2.3 0.7 2.2

71.1 89.3 68.7 85.7 93.6 88.1 72.2 76.5 99.7 97.3 99.8 99.6 92.7 80.5 49.5 13.3 83.8 93.7 83.1

21.5 10.3 31.3 13.2 6.4 11.9 22.5 22.9 0.0 2.2 0.0 0.3 7.2 3.3 47.4 79.6 13.5 5.5 14.4

0.4 0.0 0.0 0.0 0.0 0.0 0.8 0.0 0.0 0.0 0.1 0.0 0.0 0.0 0.0 0.3 0.1 0.0 0.0

6.1 1.4 1.1 0.5 0.6 0.1 4.2 0.6 0.3 1.4 0.7 2.4 0.8 25.0 24.1 13.9 2.9 2.9 1.9

3.4 0.6 32.0 3.8 36.8 9.1 1.1 15.4 2.8 20.2 0.9 12.1 10.4 2.3 57.4 24.8 17.7 0.4 3.0

90.1 98.1 66.9 95.7 62.6 90.8 94.0 84.0 96.9 78.4 98.2 85.5 88.8 72.6 18.5 61.0 79.2 96.6 95.1

Note: Calculation of bilateral trade shares described in text. Small adjustments have been made to ensure balanced trade for each country. 𝜋 nn reflect home consumption shares; 𝜋 ni reflect external trade shares between country n and i. RoW trade figures calculated as residuals after bilateral flows between Canada, US, and UK have been accounted for. Over 90% of Canadian trade flowed through US and UK in 1880. Source: See Table 1. Canadian output from Urquart (1993), and Harris et al. (2015). US output from Census of Manufacturing (1880), Jeremy Atack and Fred Bateman’s 1880 census manuscript samples (Atack and Bateman, 1999), and Historical Statistics of the United States, Volume 4 (2006). British output from Hoffmann (1965), Lewis (1967), Feinstein (1972), and Census of the Production of the United Kingdom (1907). RoW output based on Mitchell (2003: Europe).

13

ARTICLE IN PRESS

JID: YEXEH P.D. Alexander, I. Keay

[m3GeSsc;November 24, 2017;19:55] Explorations in Economic History 000 (2017) 1–15

Table A3 Expenditure and tariff revenue shares in 1880, by country and industry (%). Expenditure share (𝛼𝑛𝑠 )

Food and Beverages Tobacco Rubber Leather Textile Clothing Wood Paper Printing and Publishing Iron and Steel Transport Equipment Non-Ferrous Metals Non-Metallic Minerals Coal and Petroleum Chemicals Miscellaneous Manufacturing Non-Manufacturing Total

Tariff revenue share (𝜆𝑠𝑛 )

Canada (1a)

US (1b)

UK (1c)

RoW (1d)

Canada (2a)

US (2b)

UK (2c)

RoW (2d)

9.9 0.5 0.2 5.2 3.2 5.9 5.0 0.6 1.1 6.2 2.0 0.9 1.1 0.6 1.4 0.8 44.3 55.7 100.0

4.2 0.5 0.0 1.5 2.5 1.7 1.5 0.3 0.4 2.1 0.3 0.6 0.5 0.3 0.6 0.6 17.3 82.7 100.0

8.2 0.7 0.2 2.2 5.7 2.1 1.5 0.2 0.4 3.6 1.2 2.3 0.7 0.3 1.5 0.2 30.8 69.2 100.0

4.5 0.5 0.1 1.4 7.4 1.6 1.2 0.3 0.4 3.7 0.7 1.6 0.4 0.6 0.2 0.4 24.5 75.5 100.0

3.5 4.5 9.6 0.8 7.3 9.0 0.5 6.1 1.7 3.2 0.8 6.6 4.7 2.3 4.7 8.9 3.9 1.0 2.3

1.2 2.0 1.6 1.8 11.1 3.3 0.6 1.1 0.7 6.1 0.3 3.5 18.7 0.7 4.3 4.8 4.3 0.5 1.1

5.8 58.9 0.0 0.0 0.0 0.0 0.0 2.1 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 2.8 0.0 0.8

1.7 1.9 2.1 0.4 2.8 0.9 0.6 2.4 0.3 2.1 0.1 0.4 1.1 5.7 1.9 4.7 2.8 0.4 1.0

Note: Calculation of expenditure and tariff revenue shares described in text. See notes from Tables 1 and 2.

Table A4 Canadian import demand elasticity parameter estimates (𝜃 𝑠 ), by industry.

Food and Beverages Tobacco Rubber Leather Textile Clothing Wood Paper Printing and Publishing Iron and Steel Transport Equipment Non-Ferrous Metals Non-Metallic Minerals Coal and Petroleum Chemicals Miscellaneous Manufacturing Non-Manufacturing Total

OLS

IV1

IV2

Kee et al. (2008)

Ossa (2015)

(1)

(2)

(3)

(4)

(5)

3.45 3.54 0.81 1.30 1.82 1.19 1.38 1.87 2.07 2.18 2.71 2.19 1.76 0.90 1.60 1.71 1.87 1.67 1.82

5.10 4.56 3.83 6.53 4.54 5.88 4.28 3.34 8.13 4.29 3.81 3.46 3.07 3.50 3.19 3.72 4.84 3.84 4.64

2.06 1.57 2.08 2.30 1.87 1.38 1.78 3.13 2.60 2.24 1.63 1.72 1.72 1.71 1.35 0.87 1.10 1.31 2.18

(0.836) (0.322) (0.214) (0.347) (0.533) (0.118) (0.588) (0.511) (0.507) (2.168) (0.775) (0.346) (0.444) (1.316) (0.172) (0.513) (1.062) (0.143) (0.652)

2.26 2.70 2.69 2.08 1.92 1.38 1.87 2.70 2.19 2.58 1.36 2.08 2.64 2.11 1.23 1.11 2.55 1.40 2.69

(1.054) (0.750) (1.169) (0.465) (1.060) (0.120) (0.752) (0.374) (0.631) (1.834) (0.923) (0.658) (0.886) (3.392) (0.145) (0.671) (1.233) (0.361) (1.216)

2.24 1.71 2.17 2.29 2.38 1.22 1.82 2.45 2.28 2.47 1.91 2.10 2.33 1.45 1.37 0.91 2.16 1.90 2.83

(1.057) (0.513) (0.355) (0.364) (0.979) (0.159) (0.677) (0.312) (0.303) (6.671) (0.898) (1.216) (3.220) (1.910) (0.291) (1.410) (3.270) (0.567) (1.413)

Note: Detailed description of estimating equation and derivation of long run elasticities provided in text. OLS, IV1 and IV2 estimated over 1880–1913 period. Robust standard errors reported in parentheses. IV1 treats domestic prices (Pnn ) as endogenous; IV2 treats domestic prices and import unit values (Pni ) as endogenous. Excluded instruments include Canadian and US raw material, fuel, labor, and capital input prices (Irwin (2000b), and Alexander and Keay (2017). Two-stage IV estimates pass standard weak instrument and instrument validity tests (see Appendix Table A1). Kee et al. (2008) HS6 elasticities estimated with international panel data from 1988 to 2001. Ossa (2015) HS6 elasticities estimated with international panel data from 1994 to 2008. Disaggregate modern are estimates aggregated up to SIC2 level using 1880 import value weights (Beaulieu and Cherniwchan, 2014). Source: See Tables 1 and 2. Import unit values from Taylor (1931). Canadian industry-specific prices from Harris et al. (2015).

14

JID: YEXEH

ARTICLE IN PRESS

P.D. Alexander, I. Keay

[m3GeSsc;November 24, 2017;19:55] Explorations in Economic History 000 (2017) 1–15

References Alexander, P. D., Keay, I., 2017. Responding to Globalization: The Causes and Consequences of Canadian Trade Policy, 1870–1913. Working Paper. TBD. Alvarez, F., Lucas, R.J., 2007. General equilibrium analysis of the Eaton-Kortum model of international trade. J. Monetary Econ. 54 (6), 1726–1768. Anderson, J.E., Neary, J.P., 2005. Measuring the Restrictiveness of International Trade Policy. MIT Press, New York, NY. Arkolakis, C., Costinot, A., Rodriguez-Clare, A., 2012. New trade models, same old gains? Am. Econ. Rev. 102 (1), 94–130. Armington, P., 1969. A theory of demand for products distinguished by place of production. Int. Monetrary Fund Staff Papers 16, 159–176. Atack, J., Bateman, F., 1999. Nineteenth century US industrial development through the eyes of the census of manufactures. Historical Methods 32 (4), 177–188. Bairoch, P., 1972. Free trade and european economic development in the nineteenth century. Eur. Econ. Rev. 3, 211–245. Bairoch, P., 1989. European Trade Policy, 1815–1914. In: Mathias, P., Pollard, S. (Eds.), Cambridge Economic History of Europe, 8. Cambridge University Press, pp. 1–160. Beaulieu, E., Cherniwchan, J., 2014. Tariff structure, trade expansion, and Canadian protectionism, 1870–1910. Can. J. Econ. 47 (1), 144–172. Betran, C., Huberman, M., 2016. International competition in the first wave of globalization: new evidence on the margins of trade. Econ. History Rev. 69, 258–287. Broda, C., Limao, N., Weinstein, D.E., 2008. Optimal tariffs and market power: the evidence. Am. Econ. Rev. 98 (5), 2032–2065. Caliendo, L., Parro, F., 2015. Estimates of the trade and welfare effects of NAFTA. Rev. Econ. Stud. 82 (1), 1–44. Coe, P., Emery, H.C., 2004. The disintegrating Canadian labour market? The extent of the market then and now. Can. J. Econ. 37 (4), 879–897. Costinot, A., Rodriguez-Clare, A., 2014. Trade theory with numbers: quantifying the consequences of globalization. In: Handbook of International Economics. In: Handbook of International Economics, 4. Elsevier, pp. 197–261. Dekle, R., Eaton, J., Kortum, S., 2007. Unbalanced trade. Am. Econ. Rev. 97 (2), 351–355. Demidova, S., 2015. Trade Policies, Firm Heterogeneity, and Variable Markups. Department of Economics Working Papers 2015-04. McMaster University. Demidova, S., Rodriguez-Clare, A., 2009. Trade policy under firm-Level heterogeneity in a small economy. J. Int. Econ. 78 (1), 100–112. Eaton, J., Kortum, S., 2002. Technology, geography, and trade. Econometrica 70 (5), 1741–1779. Emery, H.C., Levitt, C., 2002. Cost of living, real wages and real incomes in thirteen Canadian cities, 1900−1950. Can. J. Econ. 35 (1), 115–137. Estevadeordal, A., Frantz, B., Taylor, A.M., 2003. The rise and fall of world trade, 1870−1939. Q. J. Econ. 118 (2), 359–407. Federico, G., Vasta, M., 2015. What do we really know about protection before the great depression: evidence from Italy. J. Econ. History 75 (4), 993–1029. Feenstra, R., 1995. Estimating the effects of trade policy. In: Grossman, G.M., Rogoff, K. (Eds.), Handbook of International Economics. In: Handbook of International Economics. chapter 30, 3. Elsevier, pp. 1553–1595. Feenstra, R., 2010. Product Variety and Gains from International Trade. MIT Press, Cambridge, MA. Feinstein, C., 1972. National Income, Expenditure and Output of the United Kingdom, 1855–1965. Cambridge University Press, Cambridge, UK. Felbermayr, G., Jung, B., Larch, M., 2013. Optimal tariffs, retaliation, and the welfare loss from tariff wars in the melitz model. J. Int. Econ. 89 (1), 13–25. Gillespie, I., 1991. Tax, Borrow and Spend: Financian Federal Spending in Canada, 1867–1990. Carleton University Press, Ottawa, ON. Green, A., Urquhart, M., 1994. Estimates of output growth in Canada: measurement and interpretation. In: McCalla, D., Huberman, M. (Eds.), Perspectives on Canadian Economic History. Copp, Clark, Longman, pp. 158–176. Gros, D., 1987. A note on the optimal tariff, retaliation and the welfare loss from tariff wars in a framework with intra-Industry trade. J. Int. Econ. 23 (3–4), 357–367. Harris, R., Keay, I., Lewis, F., 2015. Protecting infant industries: canadian manufacturing and the national policy, 1870−1913. Explorations Econ. History 56 (C), 15–31. doi:10.1016/j.eeh.2015.01.001. Head, K., Mayer, T., 2014. Gravity Equations: Workhorse,Toolkit, and Cookbook. In: Handbook of International Economics, 4. Elsevier, pp. 131–195. Hillberry, R., Hummels, D., 2013. Trade Elasticity Parameters for a Computable General Equilibrium Model. In: Handbook of Computable General Equilbrium Modeling, 1. Elsevier, pp. 1213–1269. Hoffmann, W.G., 1965. British Industry, 1700–1950. Basil Blackwell, Oxford, UK. Huberman, M., Meissner, C., Oosterlinck, K., 2017. Technology and Geography in the Second Industrial Revolution: New Evidence from the Margins of Trade. J. Econ. History 77 (1), 39–89. Imbs, J., Mejean, I., 2015. Elasticity optimism. Am. Econ. J. 7 (3), 43–83. Inwood, K., Keay, I., 2013. Trade policy and industrial development: iron and steel in a small open economy, 1870−1913. Can. J. Econ. 46 (4), 1265–1294. Inwood, K., Keay, I., 2015. Transport costs and trade volumes: evidence from the trans-Atlantic iron trade, 1870−1913. J. Econ. History 75 (1), 95–124. Irwin, D., 1994. The political economy of free trade: voting in the british general election of 1906. J. Law Econ. 37, 75–108. Irwin, D., 2000. Could the united states iron industry have survived free trade after the civil war? Explor. Econ. History 37 (3), 278–299. Irwin, D., 2000. Did late nineteenth century US tariffs promote infant industries? evidence from the tinplate industry. J. Econ. History 60, 335–360. Irwin, D., 2010. Trade restrictiveness and deadweight losses from US tariffs. Am. Econ. J. 2 (3), 111–133. Jacks, D., Pendakur, K., 2010. Global trade and the maritime transport revolution. Rev. Econ. Stat. 92 (4), 745–755. Keay, I., Redish, A., 2004. Financial markets and industrialization: evidence from US and canadian steel producers, 1910−1990. Explor. Econ. History 41 (4), 377–403. Kee, H.L., Nicita, A., Olarreaga, M., 2008. Import demand elasticities and trade distortions. Rev. Econ. Stat. 90 (4), 666–682. Klasing, M., Milionis, P., Zymek, R., 2015. Gravity across space and time. Discussion Paper Series 265. Edinburgh School of Economics. Lewis, W. A., 1967. The Declaration of British Growth, 1873–1913. Department of Economics Discussion Paper 3. Princeton University. Marquez, J., 1999. Long-period trade elasticities for canada, japan, and the united states. Rev. Int. Econ. 7 (1), 102–116. McDiarmid, O., 1946. Commercial Policy in the Canadian Economy. Harvard University Press, Cambridge, MA. Mitchell, B., Deane, P., 1962. Abstract of British Historical Statistics. Cambridge University Press, Cambridge, UK. Mitchell, B., 2003. International Historical Statistics: Europe, 1750–2000, Fifth Edition Palgrave Macmillan, London, UK. Mohammed, S.S., Williamson, J., 2004. Freight rates and productivity gains in british tramp shipping, 1869−1950. Explor. Econ. History 41 (2), 172–203. Ossa, R., 2014. Trade wars and trade talks with data. Am. Econ. Rev. 104 (12), 4104–4146. Ossa, R., 2015. Why trade matters after all. J. Int. Economics 97 (2), 266–277. Pomfret, R., 1993. The Economic Development of Canada. Nelson, Toronto, ON. Pomfret, R., 1995. Australia’s Trade Policies, 8 Oxford University Press, Melbourne, AU. Ruhl, K. J., 2008. The International Elasticity Puzzle. Department of Economics Working Paper 08–30. New York University. Salmon, P., 1997. Scandinavia and the Great Powers, 1890–1940. Cambridge University Press, New York, NY. Shiells, C., Stern, R., Deardorff, A., 1986. Estimates of the elasticities of substitution between imports and home goods for the united states: reply. Rev. World Econ. 125 (2), 371–374. Simonovska, I., Waugh, M.E., 2014. The elasticity of trade: estimates and evidence. J. Int. Econ. 92 (1), 34–50. doi:10.1016/j.jinteco.2013.10. Stern, R., Francis, J., Schumacher, B., 1976. Price Elasticities in International Trade. Macmillan Press, London, UK. Taussig, F., 1931. Tariff History of the United States, 8 ed. Putnam, New York, NY. Taylor, K., 1931. Statistical Contributions to Canadian Economic History, Volume 2. MacMillan, Toronto, ON. Urquart, M., 1993. Gross National Product, Canada, 1870–1926: The Derivation of the Estimates. McGill-Queen’s Press, Monteal, QC.

15