DRAFT The Contributions of Trade, Technology, and Consumption to Canadian Jobs and Output, 1986 to 2004 Ziad Ghanem Anick Johnson Industry Accounts Division Statistics Canada April, 2008 Table of Contents 1 Introduction........................................................................................................................... 3 2 Methodology ......................................................................................................................... 4 3 The sources of changes in employment.................................................................................. 9 4 The sources of changes in output ......................................................................................... 16 5 Trade by geographic area..................................................................................................... 16 6 Conclusion .......................................................................................................................... 19 Appendix 1: Data sources............................................................................................................ 22 2 1 Introduction The growing liberalization of trade, the emergence of the newly industrialising economies, and the continuing rapid pace of technological change are raising questions about the evolving role of the Canadian economy in the globalization of production processes. Two major manifestations of these developments within Canada are the continuing shifts in production and employment from the goods-producing to the services industries and an important growth in the education level of the employed workforce. The purpose of this paper is to examine these structural changes through an empirical measure of the distinct impacts of changes in trade, technology, and domestic consumption expenditures on industrial growth and employment in Canada. Our analysis is guided by separate theoretical frameworks that provide some overlapping if not completely integrated explanations. Ricardo’s comparative advantage and the Heckscher-Ohlin theorem provide the foundations of the analysis of international trade. Generally speaking, these theories postulate that countries trade in order to benefit from the international division of labour and that they should be net exporters of the commodities that embody their relatively abundant factors and net importers of the commodities that embody their relatively scarce factors. A second explanation of developments relates to the changing nature of demand for labour due to the inherent skill bias of technological change. Many authors believe that new technologies are intrinsically linked to an upgrading of skills. Bound and Johnson (1992) find such evidence for the U.S. Berman, Bound, and Machin (1998) find similar evidence across several advanced economies, thereby allowing them to generalize their finding that technological change has favoured the more skilled workers at the expense of the less skilled. The final argument is provided by Clark (1940) who has contended that with growing income levels, consumer demand will shift to services. He based his analysis on the hierarchy of needs hypothesis, which states that services satisfy higher needs than goods, and that, as income grows, a higher share of incomes will be used for the purchase of services. Clark also recognized that the resulting shift in employment will be compounded by the higher rates of productivity growth in the manufacturing sector as compared to the services sector. While recognizing that, in reality, the Canadian economy presents a complex web of interactions between these three different phenomena, this paper, nonetheless, adopts a framework that aims to assign a distinct empirical measure to each source of change in production and employment. The foundations of this method are anchored in the open Leontief input-output model. The specific model is based on the work of Gregory, Zissimos, and Greenhalgh (2001) which combines the methods developed by Chenery, Shishido, and Watanbe (1962) for the study of structural change and economic growth and the pioneering work by Leontief (1951) on the factor content of trade. Our analysis relies on several Statistics Canada data sources. The latter include the annual Canadian current and constant price input-output (IO) tables1 and the labour input data from the 1 The constant price input-output tables in year t are expressed in terms of the prices of year t-1 and weighted by the current price values of year t-1. 3 KLEMS database 2 produced by the Canadian Productivity Accounts. The study covers the years from 1986 to 2004. For the period from 1997 to 2004, primary source data from the International Merchandise Trade and the Balance of Payments programs are used to supplement the international trade vectors of the IO tables with the information required for analyzing the geographic patterns of trade. The major structural changes that we seek to explain are the following. According to the KLEMS database, 4.1 million full-time equivalent net jobs were added to the Canadian economy over the period from 1986 to 2004. Of these, the services industries accounted for an increase of 3.9 million while the goods-producing industries accounted for a mere 202 thousand jobs. At the same time, the share of high-education employment increased from 49% to 66% of the total employed workforce. The IO tables also show that over this same period, the services industries continued their historical ascendancy, increasing their share of GDP from 64% to 67%. Our main findings are that over the period under study, growth in employment in Canada has been led by a relatively constant growth in domestic consumption expenditures, especially of services. International exports played a growing role in job creation, reaching 25% of total employment by 2004 thereby overshadowing growing import intensities, which contributed to the loss of 751 thousand jobs. The most important factor in job shedding, however, was labour-saving technological change, which displaced over 3 million jobs. In general, domestic consumption and net trade have favoured skilled labour. Technological change, however, was the most important force in displacing low-skill employment both in the goods and services industries. Nonetheless, technological change was also a significant factor in job creation, promoting the addition of a large number of high skill jobs in the services industries. Exports contributed relatively more to GDP than to employment and the trend appears to be reinforced by the growing share of primary commodities in exports. An examination of the geographic nature of Canadian trade over the period from 1997 to 2004 confirms the continued dominant role of the U.S. export market for Canadian jobs, but an emerging trend points toward a switch in offshoring activities away from the U.S. toward all other major trading regions. The paper is organized in the following manner. The next section develops the model used in explaining the sources of output and employment change. Sections 3 and 4 present the results of the model in terms of employment and output respectively. Section 5 presents the results by geographic area for the period covering 1997-2004. Section 6 provides concluding remarks. 2 Methodology The model draws on the work of Gregory et al. (2001) for measuring the contributions of changes in domestic demand, technology, and trade on industrial output and employment. The starting point of the model is the basic input-output identity, which defines domestic production as equal to demand due to inter-industry transactions and final expenditures net of international imports. 2 The acronym KLEMS stands for “capital, labour, energy, materials, and services”. It is an experimental database of industry statistics for multifactor productivity analysis published by Statistics Canada. 4 Following the approach taken by Gregory et al. (2001), this basic identity is slightly modified in equation (1) to show capital expenditures as a function of industry gross outputs3. (1) g = (Dhˆ B)g + (Dhˆ J)g + Dhˆ f + Dx where g is the vector of total gross outputs by industry; B and J are the commodity by industry matrices of intermediate input and capital input coefficients per unit of gross output; f is a commodity vector of final domestic expenditures of households, non-profit institutions, and government; D is the industry by commodity market share matrix; x is a commodity vector of international exports; and h is a vector of home shares by commodity, where the symbol ^ indicates the conversion of a vector into a diagonal matrix. Defining capital expenditures as a function of gross outputs allows the model to endogenize capital expenditures as a function of all other final expenditures, in a sense, treating them similarly to intermediate inputs. This allows the model to better delineate the impact of exports on economic activity. Even though investment expenditures have an economic life that exceeds the one year accounting cycle of other expenditures in the IO tables and a direct link between investment expenditures and production for final consumption expenditures or exports cannot be established within a given year, it is nonetheless a reasonable allocation method given the relatively long period of analysis and the large number of industries covered by this paper. While recognizing its limitations, this method remains preferable to the alternative approach of treating investment expenditures as domestic expenditures thereby underestimating the true contribution of trade to domestic production. Solving equation (1) for gross outputs: (2) [ g = I − Dhˆ B − Dhˆ J ] −1 D(hˆ f + x) Equation (2) can be differenced to decompose variations from one year to the next (3) ( g = R 0 hˆ 0 f + hˆ f1 + x + hˆ B1g 1 + hˆ 0 Bg1 + hˆ J 1g 1 + hˆ 0 Jg 1 ) where the subscript 0 represents the starting period and the subscript 1 represents the subsequent −1 period, both valued in constant prices, and R = I − Dhˆ B − Dhˆ J D . [ ] Equation (3) expresses total changes in industry gross outputs as a function of changes in demand and technology. The impact of final domestic consumption is expressed in R 0 hˆ 0 f ; exports in R 0 x ; the impact of the changing share of imports related to final consumption, intermediate inputs, and capital expenditures are expressed in R 0 hˆ f1 + hˆ B1g 1 + hˆ J 1g1 ; and finally the impact of technological change is captured in R hˆ Bg + hˆ Jg . Technological change in this 0 ( [ 0 1 0 1 ) ] context has a very broad meaning—it refers to changes in industry demand that are due to changes 3 This equation is adapted from the standard industry by industry formulation to the Canadian industry by commodity Supply and Use tables. 5 in the intermediate input and capital expenditure coefficients, regardless of what may be at the root of such changes, whether they be due to changes in relative prices, economies of scale, firm entry or exit from the industry, changes in organization, technological innovation, change in output product mix, outsourcing of activities, etc. Impacts on employment can be derived from the employment by industry coefficients (4) L = Lr • g where L is an industry by skill level matrix of employment, Lr is a matrix of similar dimensions of ratios of employment to gross output by industry, and the operator • indicates an element-byelement multiplication. Yearly variations in employment can be decomposed into variations that are due to changes in the levels of industry output and to variations that are due to the labour input coefficient: (5) ∆L = Lr0 • ∆g + ∆Lr • g 1 By substitution of equation (3) into (5) (6) [ ] ∆L = Lr0 • R 0 hˆ 0 f + hˆ f1 + x + hˆ B 1 g 1 + hˆ J 1 g 1 + hˆ 0 Bg 1 + hˆ 0 Jg 1 + Lr • g 1 The last term in equation (6) adds a new element to the measure of technological change expressed in equation (3). Changes in employment that are due to changes in the demand for labour embodied in intermediate and capital inputs are complemented by changes in the direct employment coefficient of each industry. From the perspective of empirical analysis, the formulation in equation (6), as developed in Gregory (2001), is extremely rich and can be further extended in several ways to examine the sources of variations in GDP, the impacts of the geographic nature of trade, and the causality links between exports and domestic demand on the one hand and imports and technological change on the other. A further decomposition of the impact of domestic and export demand on imports and technological change is possible from an expansion of the gross output term g 1 in equation (6) by substitution from equation (2): (7) ( ) ∆L = Lr0 • R 0 hˆ 0 ∆f + ∆x + Lr0 • R 0 ∆hˆ f1 + ∆hˆ B1 R 1 + ∆hˆ J 1 R 1 + Lr • R hˆ R ( B + J ) + Lr • R [ [ 0 0 0 ( 1 1 )(hˆ f ](hˆ f 1 1 1 1 )] +x ) + x1 1 The first line in equation (7) again defines the impacts of changes in the levels of consumption expenditures and exports on employment; the second line regroups the impacts on imports of changes in the home shares of final domestic expenditures (in the first term) as well as the home shares of intermediate and capital expenditures used in the production for domestic consumption 6 and exports; the third line expresses the impacts of technological changes reflected in the changes in intermediate, capital, and direct employment coefficients related to production for domestic demand and exports. A study of impacts on GDP variations similar to the impact on employment can be operated based on three equations that are symmetric to equations (4)-(6), where the GDP coefficients by industry replace the employment coefficients. (8) gdp = gdp r • g where gdp is a vector of GDP levels by industry and gdpr is a vector of ratios of GDP to gross output by industry. gdp = gdp r0 • g + gdp r • g 1 (9) (10) ∆gdp = gdp r0 • R 0 hˆ 0 f + hˆ f1 + x + hˆ B1g 1 + hˆ J 1g 1 + hˆ 0 Bg1 + hˆ 0 Jg 1 + (gdp r ) • g 1 A further decomposition of international trade by geography is also possible. Thus, exports can be decomposed by geographic destination, so that we can express [ (11) x= xn , ] for n = 1,...N . n where x n is a commodity vector of international exports to the country n and N is the total number of countries. By substitution of equation (11) into the exports-related term in the first line of equation (7), it is possible to isolate the impact of changes in exports by geography on employment by industry and education level: (12) ∆L Xn = Lr0 • R 0 x n , for n = 1,...N . where LXn is an industry by education matrix of employment levels related to exports and N is the number of regions. Home shares can be related to the changes in import shares by country of origin (13) hˆ n = (I − ˆ n ), for n = 1,..., N . where ĥ n is a commodity vector of supply shares that exclude country n and vector of import shares from the country n. n is a commodity Similarly to exports, from equations (13) and the terms related to changes in import shares in equation (6), it is possible to isolate the impact of changes in imports by geography on employment by industry by education level. (14) [ ( )] ∆LMn = Lr0 • R 0 ∆ (I − ˆ n )f1 + (∆ (I − ˆ n )B1 R 1 + ∆ (I − ˆ n )J 1 R 1 ) hˆ 1f1 + x1 , for n = 1,...N . 7 where LMn are industry by education by region matrices of employment levels for each region n. Up to this point, the presentation of the model has proceeded with the assumption that the comparison between the base and end periods will be in constant prices. This is because in the present context, price variations can distort the measurement of the relationship between industry gross outputs and employment. In the short term, this relationship can be expected to exhibit relative stability in relation to volumes but certainly not to values of output. However, removing the distortions caused by price variations creates a contradiction between a static measurement of the most current information and the dynamic analysis of the changes that have brought about these results. Removing price variations creates a problem of non-additivity between the levels measured in the base period, the cumulative effect of yearly changes reflected in the analysis of the sources of the variations, both of which are measured for example, in base year prices, and the observed levels at the end period, if the latter are measured in the prices of the end period. The assumption that in a given year the best way to define a volume-based Leontief production function is by normalizing prices to one, will inevitably lead to a loss of additivity if those measures are to be integrated with an analysis of the sources of change that are net of price variations. It is therefore necessary to remove the role of prices in distorting the measurement of the sources of change in the short term but at the same time account for their role in reallocating inputs over the longer term. This role of prices in re-weighting costs can be measured through an application of equation (2) to the differences between the current and constant price IO tables for the same year and by explicitly introducing the role of prices: (15) ( ) ( ∆g KC = R C PtF • hˆ f + PtX • x − R K PtF−1 • hˆ f + Pt X−1 • x ) where R C and R K are respectively the current and constant price versions of the R matrix; P F and P X are commodity vectors of the prices of domestic final consumption and exports with the subscripts t and t-1 representing respectively current and base-year prices; and the f and x commodity vectors, in this context, are measured in volumes. The term ∆g KC represents the differences by industry between current and constant price values of gross output. (The difference between the ending year and the sum of the starting year and the cumulated changes due to changes in the levels of demand, technology, and import propensity will reflect the impact of price variations.) Distinguishing impacts on industry inputs, such as employment by type of demand can be obtained from equations (15) and (5) (16) [ ( ) ( )] ( ∆L KC = LrK • R K ∆P F • f + ∆P X x + ∆R PtF • hˆ f + PtX • x + ∆Lr • R C ∆P F • f + ∆P X x ) Equation (16) measures the impact of changes in relative prices on the allocation of employment by industry. As with equation (5), it reflects the embedded and direct within-industry impacts. It also reflects the changes in prices of domestic demand or exports. The total effect of this term 8 across demand categories must sum to zero as it can only act to re-allocate employment numbers between domestic demand and exports. Overall, the main advantage of the model is its capacity to explain the net changes in output and employment for each industry, from one period to the next, as a function of the distinct impacts of changes in the levels of domestic final demand, exports, import propensities, technology, and relative prices. One of its main weaknesses, however, is its incapacity to explain any changes in behaviour linking the different sources of change. For example, a decline in the price of an imported commodity may lead to a substitution effect in consumption expenditures; the model, however, would only attribute this change to two independent developments: an increase in the import propensity of the product and a change in expenditure levels, thereby shedding no light on any causality links between the two related developments. Two clarifications are necessary before proceeding to an analysis of the empirical results of the model. First, it has been assumed that short term price fluctuations do not impact production functions but that once they occur they become the new basis of allocations in the subsequent accounting period. The variations between 1986 and 2004 that are reported in this paper represent the sum of annual variations from base year current price to subsequent year constant price IO tables. This method of cumulating annual changes avoids the problem of large shifts in the relative weights of detailed components between the base and current years that would be associated with deflating and directly comparing the 2004 and 1986 tables. The added benefit is that the detailed annual data also allow a closer examination of the evolution of these variations over time. The second issue is that while one of the main objectives of this paper is the study of the skill content of employment, the only readily available pertinent data source, which is consistent with the industrial classification of the IO tables, the KLEMS database, contains information solely on formal education levels and not overall skills related to the job. The education level is therefore used as a proxy for the analysis of qualifications required to fulfill a job despite recognizing its limitations as a measure due to on the one hand the roles of on the job training and experience in enhancing purely formal qualifications and on the other hand the inherent ambiguity of the relevance of formal education to the accomplishment of job-related tasks. Furthermore, while the KLEMS database contains up to six education levels, for ease of analysis, they are grouped in this study into the two categories of lower education for all those that have at most completed their secondary education and higher education for those with some or completed post-secondary education. 3 The sources of changes in employment The sources of employment by educational attainment in 1986 and 2004 as well as the cumulative changes between the two periods, as formulated in the model, are presented in Table 1. The levels of employment for 1986 and 2004 are derived from the combination of equations (2) and (4). The sum of annual variations due to changes in demand, import shares, and technology are derived from equation (7) and those that are due to the impact of re-weighting are calculated from equation (16). 9 Table 1 Sources of changes in employment, by education, 1986 and 2004 Source Education 1986 Demand Import shares Technology Re-weighting 2004 Consumption expenditures Low High thousands of jobs 4,809 1,839 -320 -2,558 150 3,920 4,943 2,816 -300 662 115 8,235 Exports Low 1,320 1,456 -68 -1,111 -84 1,513 Total High 1,022 1,779 -63 -32 -180 2,525 12,094 7,889 -751 -3,039 0 16,193 An overview of Table 1 shows that the job market is dominated by the needs of consumption expenditures and that job creation, due to growth in the levels of demand by both the domestic and export sectors has largely outpaced job displacement by technological change and imports. At 4.7 million new jobs, the increase in the level of consumption expenditures has been the highest contributor to job creation, but despite the relatively smaller size of the export sector, growing export demand nonetheless generated an additional 3.2 million jobs. Thus, the 138% growth rate related to export demand substantially outpaced the 48% growth rate of employment linked to domestic consumption expenditures, pushing up the share of exports in total employment from 19% in 1986 to 25% by 2004. The impact of changing import propensities displaces 751 thousand jobs and does very little to counteract the overall role of export growth. Consequently, the overall contribution of net trade to job creation remains substantial with job creation due to international exports exceeding by almost four fold the total job losses due to growing import shares. Labour-saving technological change was the most important factor in counteracting the stimulative role of the growth in demand. It also appears to play a bigger role in the export sector. As a share of the gross demand for jobs by sector, that is, as a share of the sum of the 1986 level and the changes due to growth in demand, it represented 21% in the export sector as compared to 13% in the domestic sector. The yearly re-weighting of the constant price IO tables shows relatively minor effects. It contributes to the reallocation of 265 thousand jobs from the export to the domestic sector. This shift indicates that between 1986 and 2004, changes in relative prices appear to have generally favoured the domestic sector. The period under study witnessed the ascendance of high-skill employment to a position of dominance in the labour market. In 1986, low-skill jobs exceeded high-skill jobs by a small margin. This relationship was reversed by 2004 when there were almost twice as many high-skill jobs as compared to low-skill jobs. A closer examination of the origin of these jobs by type of demand shows that even though skilled labour only slightly exceeded non-skilled labour in the domestic sector and was lower than non-skilled labour for the export sector in 1986, it shifted to a solid supremacy in both sectors by 2004. The share of skilled labour grew to represent 68% in the domestic sector and 63% in the export sector. 10 In general, demand favoured growth in high-skill employment while technological change focused on the elimination of low-skill jobs even as it simultaneously promoted the creation of high-skill jobs. Imports had a relatively small effect but nonetheless still favoured by a small amount the displacement of low-education as compared to high-education jobs. Changes in the levels of demand led to the creation of almost 4.6 million high education jobs as compared to 3.3 million low education jobs. A breakdown by sector shows that in the domestic sector, the creation of high-education jobs outstripped low-education jobs by 53%, while in the export sector the same relationship persisted but at a more moderated level of 22%. Technological change was a very destructive force for low-education jobs. For the domestic sector, technological change brought about the elimination of approximately 2.6 million loweducation jobs, at the same time as it promoted the creation of 662 thousand high-education jobs. In the export sector, over 1.1 million low-education jobs were lost to technological change while the high-education jobs were barely touched with a loss of only 32 thousand. Decomposing these variations into their yearly elements highlights the role of two factors: the importance of the business cycle for all sources of change and the appearance of a structural break in the mid-1990s related to the demand for skilled labour. The main sources of yearly changes in employment by education level, covering 1986 to 2004 and as formulated in equation (6), are presented in Table 2. The variations sum to the aggregates presented in Table 1, however, for ease of presentation, changes in imports and technology are aggregated across types of demand. Table 2 Sources of employment change by education, 1986 to 2004 Source Education Consumption expenditures Low High 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 Total change 174 241 189 -22 -187 126 96 137 21 87 103 126 122 169 92 127 135 102 1,839 176 246 174 28 -82 137 87 147 36 105 112 219 229 268 194 249 268 225 2,816 Exports Import shares Low High Low High thousands of jobs 54 63 2 50 22 94 135 164 125 93 133 173 155 141 -6 20 -44 83 1,456 45 71 5 41 18 85 128 171 151 117 168 233 229 219 -27 30 -38 132 1,779 -2 -30 -10 -25 -59 -49 -57 -13 -26 -18 -47 -64 -9 50 22 -3 2 -48 -388 7 -28 0 -22 -41 -46 -56 -21 -23 -22 -46 -80 -15 66 43 -5 -3 -73 -363 Technological change Low High Total employment change Low High -132 -139 -118 -49 -54 -374 -374 -396 -221 -202 -385 -211 -119 -307 -222 -60 -202 -106 -3,669 95 136 63 -46 -278 -204 -199 -108 -100 -40 -196 24 149 54 -115 83 -109 30 -763 61 28 77 83 149 -69 177 69 160 -33 239 -80 -213 -273 57 22 211 -34 630 289 317 256 131 43 107 335 367 324 167 472 291 230 280 267 296 439 250 4,861 11 Over the period under coverage, growth in employment levels has been led by a relatively constant growth in domestic demand, which showed serious weaknesses only in 1990 and 1991. Net trade made most of its impact on job creation from 1992 to 2000, coinciding both with the introduction of the Canada-U.S. Free Trade Agreement and the growth period of the U.S. business cycle. Yearly variations for the domestic demand and trade categories are heavily influenced by the business cycle. The slowdowns in job creation around the North American recession in 1991 and the US recession in 2001 are clearly evident. However, while both domestic consumption and net trade were contributing to job losses around 1991, the same was not the case for the years following 2001 when the domestic economy continued to generate a large number of jobs even as the export sector faltered. Starting in 1988 and up to 1999, increasing import intensity mildly but consistently contributes to job losses. This trend in job displacement slows down in 1999 and reverses directions in 2000 and 2001. It is worth noting that this reversal is not generalized in its character and is mostly attributable to the declining import intensity of intermediate inputs and not due to any major shifts in the trend related to final expenditures. Of the 181 thousand net jobs created over those two years, 198 thousand were in fact due to the decline in the import intensity of intermediate inputs while direct imports by final expenditures continued its contribution to job shedding even if at the mild difference of 17 thousand jobs. Until recently at least, the impact of net trade was also highly correlated with the movements of the Canadian exchange rate. Figure 1 shows that over the period from 1987 to 2000 the two variables show a high level of correlation. The year 2001, however, marks a decoupling of this relationship. The continued depreciation of the Canadian dollar in 2001 and 2002 did not stem the tide of the decline in the demand for exports. While the beginning of the ascent of the currency in tandem with the growing prices of primary commodities evidently did not hold back real growth in demand in 2004. Nonetheless, two factors are at play during this period: the Canada-U.S. Free Trade Agreement which boosted trade with the U.S. and the pull-back in export demand accompanying the U.S. economic slowdowns in 1991, 1995-1996, and 2001. 12 Figure 1 Employment change due to net trade and the Canada-US exchange rate, 1986-2004 1.6 600 500 400 300 1.4 200 1.3 100 Jobs (1000s) C$ per $US 1.5 0 1.2 -100 -200 19 87 19 88 19 89 19 90 19 91 19 92 19 93 19 94 19 95 19 96 19 97 19 98 19 99 20 00 20 01 20 02 20 03 20 04 1.1 Can-US exchange rate Employment change, net trade The biggest contributions by labour-saving technological change appear centered around the period of highest economic growth, from 1992 to 2001. This would be a reasonable development during periods of growing demand for labour. Conversely, the moderation in its impacts on labourshedding around the periods of economic slowdown may be partially related to the phenomenon of labour hoarding by firms during the troughs of the business cycle. While the overall growth in both domestic consumption and exports has favoured high-skill employment, the year 1994 appears to be a watershed in the evolution of this bias. Relatively small differences prior to 1994 evolve into large diverging trends. Up to 1993, domestic consumption created 766,000 high-skilled relative to 617,000 low-skilled jobs. In comparison, after 1993, this increases to 2.1 million high-skilled as compared to 1.2 million low-skilled jobs. This represents a growth in the share of the high skilled from to 55% to 63% between the two periods. For the export sector, which created an almost equivalent number for both categories at 393,000 for the high-skilled and 420,000 for the low-skilled during the first period, the relationship also shifts in the second period to 1.4 million high-skilled and one million low-skilled, to reach a share of 57% for the highly-skilled in job creation. Imports similarly evolve in two distinct phases. Up to 1993, imports favoured the displacement of low education jobs. Beginning in 1994, imports switch to competing more heavily with higheducation jobs. Not all years show a negative impact from increasing import shares. Most notably in 2000 and 2001, a reversal in offshoring activities actually favours job creation instead of job shedding, especially of relatively more skilled jobs. Overall net trade favours the higher relative to the less skilled jobs. Even if the focus is switched to the post-1993 period, the share of embodied skilled labour still maintains an overall superiority in exports, at 57% as compared to 54% for imports. 13 Technological change consistently contributes to the displacement of low-education jobs over the entire period. For most years, it also simultaneously favours the creation of high-education jobs; the most notable exception being from 1998 to 2000, at the height of the high-tech boom, when it actually contributed to the displacement of 566 thousand high-education jobs. A possible explanation could be that technological change reacted to the growing demand for skilled labour with the introduction of labour-saving transformations aimed at promoting substitution away from the growing relative scarcity of the factor. Table 3 basically shows the same information as Table 1 but presented by industry. The 1986 and 2004 levels as well as the sources of employment change between these two periods are categorized by industry4 and education level. Table 3 The sources of employment change by industry and education level, 1986 to 2004 Industry Primary Secondary Services Total Education Low High Low High Low High 1986 474 214 1,647 1,134 4,009 4,617 12,094 consumption expenditures 41 25 289 301 1,509 2,490 4,654 exports import shares thousands of jobs 148 -33 102 -23 675 -274 705 -247 633 -81 971 -94 3,235 -751 technology Reweighting 2004 -304 -48 -1,017 -17 -2,348 695 -3,039 -17 -2 -26 -75 109 11 0 308 268 1,293 1,802 3,831 8,690 16,193 Most benefits from the growing levels of demand accrued to the services industries which gained 4 million new jobs. Services accounted for 86% of the total new jobs related to domestic consumption and 50% of new jobs related to exports. Job losses due to imports had a much larger impact on the goods industries at 577 thousand and 77% of the overall impact by this source of change. Technological change slightly favours net job losses in the services industries at 1.65 million and 54% of the sector total. For both education categories, the biggest level increases have been in the services sector and associated with the growth in domestic consumption. The large job losses for low-skill employment associated with technological change were centered in services at 2.3 million and secondary industries at 1 million. A comparison of the beginning and end period levels shows the share of skilled employment increasing for all three industrial sectors. Between 1986 and 2004, it increased from 31% to 47% in the primary industries, from 41% to 58% in the secondary industries, and from 54% to 69% in services. 4 Primary industry includes agriculture, forestry, mining, and oil and gas extraction; secondary industry includes manufacturing, construction, and utilities; and services includes all other industries, including the non-profit and government sectors. 14 In the case of the primary goods industries, low-skilled employment benefited the most from the growth in export demand at 148 thousand jobs created, but this was more than offset by a drop of 304 thousand jobs brought about by technological change. In general, despite the larger level increases for low-skill jobs, growth rates favoured the higher skilled jobs and technological change disproportionately targeted the elimination of lower-skill jobs. High-skilled employment in the secondary industries benefited both from higher level increases and growth rates in demand. Imports were slightly biased against low-skill employment in terms of levels but this order is reversed if the data are examined in terms of growth rates. Over a million low-skill jobs were displaced by technological change, while high-skill jobs were barely touched with a loss of 17 thousand jobs. Together, imports and technological change displaced more lowskill jobs than was added by growth in the demand levels whereas it was the opposite case for high-skill jobs. In the services sector, high-skill jobs due to growing demand significantly outpaced the growing demand for low-skill jobs. The largest factor in shifting the sector’s skill distributions, however, was due to technological change. The latter had a substantial impact on reducing low-skill jobs by displacing 2.3 million at the same time as it actually increased the demand for high skill jobs by a non-negligible 695 thousand. Imports played a minor role in displacing jobs with only a slight bias against higher skill jobs. A more differentiated view emerges from an analysis of the industrial breakout of the skill content of Canadian trade. For the primary industries, the embodiment of high vs. low skill jobs shows an equivalent proportion for exports as compared to imports. Secondary industry shows a slightly higher relative proportion for skilled labour in exports. While services show the highest relative proportion for skilled labour embodied in exports as compared to imports. Other studies of the Canadian economy, using different methods, have found no links between imports and employment levels. Baldwin and Gu (2007) report no impact on employment from goods or services offshoring over the period from 1961 to 2003. In a similar vein, Morissette and Johnson (2007) also find no evidence of a correlation between services offshoring and the evolution of employment rates over the period from 1987 to 2006. These results are not very surprising since our model shows that due to their small relative weight the impacts of changes in imports have been swamped by other factors. Contradicting some of our findings though, Baldwin and Gu (2007) do find a relationship between the decrease in the share of university-educated workers and the increase in the material offshoring of the manufacturing sector. But they find no relationship between the offshoring of services activities across all industries and the share of university-educated workers. Researchers are also not unanimous when it comes to assessing the reasons for the growing role of skilled labour. Lee (1996) finds that over the period from 1970 to 1990, the relative employment of non-production to production workers in the manufacturing sector is negatively related to technical progress. Gera, Gu, and Lin (2001), however, contradict Lee’s findings. Basing their analysis on the share of skilled workers in the total wage bill, they find that in the 1980s and early 1990s technological change was biased toward the use of skilled workers in Canadian industries. 15 More in line with our results that both technological change and offshoring have exhibited a skill bias, Yan (2006) finds that in the manufacturing sector, over 1981 to 1996, both expenditures on information and communication technologies and offshoring have had an impact on the use of more skilled workers. 4 The sources of changes in output Table 4 shows the sources of average annual growth rates of real GDP by industry based on equation (10)5. The yearly averages demonstrate that, as with employment, the largest contribution is from domestic final consumption. Exports play a secondary but substantial role which, as with employment, significantly outpaces the impact of imports. The contribution of technical change to GDP growth is negative but very small. This visibly contrasts with its important impact on job displacement. However, in this context, the technology term is not very meaningful as it basically reflects the impact of shifts in the ratios of industries’ GDP to total inputs (current, primary, and capital). Since these ratios tend to decrease mildly over the long term, their impact will be manifested in equivalently small negative values. Table 4 Sources of GDP growth by industry, average annual rates, 1987 to 2004 Domestic Final Consumption Primary Secondary Services Total 0.02 0.27 1.50 1.78 Exports Percent 0.21 0.70 0.59 1.50 Import shares Technological change Total GDP growth -0.05 -0.24 -0.08 -0.37 -0.05 -0.12 0.06 -0.11 0.14 0.61 2.06 2.80 The relative order of industries in total GDP growth shows the dominant role of services in overall growth, followed by secondary, and finally by primary industry. Increasing domestic consumption levels make their biggest contribution to growth in the services industries, with a much smaller impact on secondary industry and an almost negligible role for primary industry. The export sector, however, reverses roles, with most of its impact being attributed to the goods producing industries. Nonetheless, services still benefit substantially from the export sector, even if a large portion of the activity is in fact attributable to the distribution and transportation of goods. Imports are heavily tilted towards secondary industry, which accounts for two-thirds of the total drag on GDP growth by imports. 5 Trade by geographic area The total impacts of changes in trade by geographic area on employment as expressed in equations (12) for exports and (14) for imports, are reported in Table 5. The analysis is restricted to the 5 The “Total GDP growth” rate is based on the growth rate between the constant price IO table in year t and the current price IO table in year t-1. 16 period covering 1997 to 2004 due to limitations on data availability6. We organize the trade data by OECD and non-OECD membership as a proxy for the categorization of less vs. more developed economies. We also break out the detail for the most important trading partner in each grouping, the U.S. and China. Table 5 Employment effects due to changes in trade, by geography, 1998-2004 1998 1999 2000 2001 2002 2003 thousands of jobs Exports 5 4 8 5 -4 6 China Import -11 -6 -15 -16 -30 -29 Net -7 -2 -7 -11 -34 -23 Exports -5 3 23 -5 -2 27 Other non-OECD Imports -14 6 8 -6 -9 -23 Net -19 9 31 -10 -11 3 Exports 32 25 79 3 -3 12 Other OECD Imports -37 -22 -3 -21 -3 6 Net -5 3 76 -18 -6 19 Exports 373 352 250 -36 59 -127 United States Imports -81 -2 126 108 34 45 Net 292 350 376 72 93 -82 Exports 405 384 360 -33 50 -81 Total Imports -144 -24 116 65 -8 -1 Net 261 360 476 32 42 -83 2004 Total 23 -54 -31 24 -34 -10 29 -18 11 138 -15 123 214 -121 93 47 -162 -115 65 -73 -7 177 -99 78 1,010 215 1,225 1,299 -118 1,181 From 1998 to 2004, net trade is job creating with the exception of 2003, the unique occurrence of a negative net employment effect, mostly due to the impact of the slowdown in U.S. exports. As expected, table 5 shows that the U.S. is the main driving force behind changes in the trade-related component of the job market. Over this period, the U.S. represented 78% of total jobs created by exports and a smaller but still important component of the impact of imports on jobs. The trend points to the growing role of trade with all non-OECD relative to trade with other OECD countries, excluding the U.S. Overall, net trade shows that job creation is outpacing job losses with OECD countries while job losses are increasingly dominating job creation in trade with China and to a much lesser extent with the other non-OECD countries. Table 6 shows the impacts on employment of the changes in import intensities by geography and by type of demand as derived from equation (14). The results are summed over the period from 1998 to 2004. Table 6 Employment effects due to changes in import shares by geography and type of demand, 1998 to 2004 Domestic Export Total Source 6 Consumption Intermediate inputs Capital Intermediate inputs Capital Domestic Exports Appendix 1 provides an explanation of the data sources. 17 thousands of jobs China Other non-OECD Other OECD US Total -66 -26 -33 34 -91 -29 -13 -18 73 13 -23 -4 -19 16 -29 -34 -29 -16 82 3 -10 -1 -12 10 -14 -118 -43 -70 123 -108 -44 -30 -28 92 -10 Over this period, domestic final expenditures are the main source of the growing impact of increasing import intensities. Altogether, they contribute to the loss of 134 thousand jobs. While consumption expenditures are the most important factor, capital imports also make a noticeable contribution. Total imports into intermediate inputs show little impact, masking some changes in the geographic structure of offshoring activities. An examination of the geographic detail brings to light a growing dichotomy between the relative roles of the U.S. and all other countries. It appears that offshoring activities are shifting from the U.S. to all the other regions. Table 7 shows the impact of exports by geographic region7 on the level of Canadian employment and income-based GDP at basic prices for 1997 and 2004. Table 7 Sources of GDP and employment, total and by export region, 1997 and 2004 GDP Employment 1997 2004 1997 2004 billions of current dollars Thousands of jobs 2 6 34 71 China 21 25 334 330 Other non-OECD 39 52 622 655 Other OECD 180 280 2,568 2,983 United States 242 363 3,559 4,038 Total exports 817 1,201 14,025 16,193 Total economy This comparative static analysis shows total exports contributing $363 billion or 30.2% of Canada’s GDP in 2004. This is slightly up from its share of 29.6% in 1997. The data again confirm the importance of the U.S. export market for the Canadian economy. In 2004, 23% of Canada’s GDP was attributable to U.S. exports. The other OECD countries accounted for most of the remainder, at around 4%, with the ‘Other’ category representing around 2%. China’s negligible contribution at .5%, visibly contrasts with the previously noted growing impact of Chinese imports on the Canadian labour market. Over this period, the U.S. increases its share by a percentage point and China also makes a slight gain while both ‘other’ categories show a slight decline. 7 A measure of the levels of GDP by export country can be obtained from a substitution of equations (8) and (11) into the export component of equation (2) to obtain: GDPn = gdp • Rx n , equations (4) and (11) into equation (2) for the employment levels. X for n = 1,...N. And similarly for 18 Interestingly, the contribution of exports to employment shows both a lower level and a different evolution than its contribution to GDP. While exports initially accounted for 25.4% of total employment in 1997, at 4.2 percentage points lower than their share in GDP, that share was down to 24.9% by 2004, just as the share in GDP was increasing. The U.S. and China increasing their weight at the expense of the other regions. The higher share in GDP as compared to employment may be partially due to a superior level of productivity in the export sector relative to production for the domestic economy. This would be in line with other findings. For example, Baldwin and Gu (2004) report that export-market participants demonstrated a higher productivity growth among Canadian manufacturing plants from 1984 to 1996, during the lead up period to the estimates reported in Table 7. They attribute this superior productivity performance to increases in product specialization, access to foreign technology, and the greater intensity of competition. While differing productivities may explain some of the differences in the initial 1997 relative shares, the continued growth in the divergence is mostly due to the changing mix of exported commodities. The continuing spread is mostly due to the growing share in exports of the mining and oil and gas extraction industries, since the latter have, in general, much higher GDP to employment ratios than other exporting industries. 6 Conclusion The paper applied an input-output approach to measuring the links between trade, technology, and domestic final consumption and the observed changes in industry GDP and employment by skill level. The model lacks certain behavioural links and incorporates certain basic weaknesses that that are characteristic of input output modelling in general. Yearly variations are based on snapshots that do not fully articulate the inter-relationships and feedbacks that characterize the dynamic transition process. In a sense, the model ascribes independence and additivity to phenomena that are clearly subject to more complex relationships. Nonetheless, despite its simplifying assumptions, the model likely provides a reasonable approximation of the most important aspects of these relationships. We find that in Canada, over 1986 to 2004, employment in general, and skilled employment in particular, has benefited the most from growth in consumption expenditures on services. International exports have increased their share in employment and have substantially overshadowed the impact of growing import intensities. The most important factor in job shedding, however, was labour-saving technological change, which displaced over 3 million jobs. Domestic consumption, net trade, and technological change have all favoured skilled labour. Technological change, however, displaced a large number of low-skill jobs across all industries even as it simultaneously promoted the creation of a large number of high skill jobs in the services industries. Exports contributed relatively more to GDP than to employment. An examination of the more recent period from 1997 to 2004 shows that among export regions, the U.S. market continues to 19 play a dominant role. For imports, though, an emerging trend points toward a switch in offshoring activities away from the U.S. toward all other major trading regions. While much of this paper has concentrated on the sources of the large structural changes in the Canadian economy, the role of services employment would benefit from further exploration. The classification of industries according to primary, secondary, and tertiary is very limited and does not fully elaborate the dependent role of services in the distribution of goods (e.g., transportation, wholesale, and retail services), in the production of goods (e.g., financial and engineering services) or in the direct provision of services to final demand. Furthermore, an analysis based on occupations may also be useful as a means of avoiding some of the problems of classifying activities by industry as well as provide a more precise measure of the skill content of employment as compared to the education level criterion. 20 References Baldwin, J. R. and W. Gu (2004) Trade Liberalization: Export-market Participation, Productivity Growth and Innovation, Economic Analysis Research Paper Series, catalogue no. 11F0027MIE, Ottawa, Statistics Canada. Baldwin, J. R. and W. Gu (2007) Outsourcing and Offshoring in Canada: Trends, Causes and Economic Impact, Micro-Economic Analysis Division, Ottawa, Statistics Canada, Forthcoming. Berman, Eli, John Bound, and Stephen Machin (1998) “Implications of skill-biased technological change: international evidence,” Quarterly Journal of Economics, 113, 1245-79. Bound, Johnson and George Johnson (1992) “Changes in the structure of wages in the 1980s: an evaluation of alternative explanations,” American Economic Review, 82, 371-92. Chenery, Hollis B., Shuntaro Shishido, and Tsunehiko Watanbe (1962) “The Pattern of Japanese Growth, 1914-1954,” Econometrica, vol. 30 no. 1, 98-131. Clark, Colin (1940) The Conditions of Economic Progress. MacMillan & Co. Ltd. London. Gera, Surendra, Wulong Gu, and Zengxi Lin (2001) “Technology and the Demand for Skills in Canada: An industry-level Analysis,” Canadian Journal of Economics, vol. 34, no. 1, 132-148. Gregory, M., B. Zissimos, and C. Greenhalgh (2001) Jobs for the skilled: how technology, trade, and domestic demand changed the structure of UK employment, 1979-90, Oxford Economic Papers, 53, 20-46. Lee, Frank C. (1996) Implications of technology and imports on employment and wages in Canada. Working paper no. 12. Industry Canada. Leontief, Wassily (1951) The Structure of the American Economy 1919-1939: An Empirical Application of Equilibrium Analysis. Oxford University Press, New York. Morissette, René and Anick Johnson (2007) Offshoring and Employment in Canada: Some Basic Facts, Analytical Studies Branch Research Paper Series, catalogue no. 11F0019MIE, Ottawa, Statistics Canada. Yan, Beiling (2006) “Demand for skills in Canada: the role of foreign outsourcing and information-communication technology,” Canadian Journal of Economics, vol. 39, no. 1, 53-67. 21 Appendix 1: Data sources All data used in this paper are from Statistics Canada sources. The IO tables and KLEMS database are published by Statistics Canada. The geographic detail of trade consistent with the IO tables, however, is not a published product. It had to be constructed from source data on merchandise trade published by the International Trade Division and on services trade published by the Balance of Payments Division. Input-Output tables The Industry Accounts Division publishes annual IO tables both in current and constant prices. The constant price table for a given year is expressed in terms of the weights and prices of its preceding year. From 1986 to 1996, the IO tables are organized in 243 industries by 679 commodities. The industrial classification is according the 1980 Standard Industrial Classification (SIC). Beginning in 1997, the IO tables are organized in approximately 300 industries, based on the North American Industrial Classification System (NAICS), by 727 commodities. The two industrial classification systems are bridged by a 1997 table cross-classified to both systems. In the IO tables, capital expenditures are available at a higher aggregation level of approximately 50 industries for both classification systems. Since the model requires a consistent industrial classification for both current and capital expenditures in a given year, capital expenditures by industry are allocated to the lower level of detail based on industry gross outputs. Trade by geographic area International trade by geographic area was constructed based on source data on trade in merchandise and services. Due to difficulties in constructing this series, we limited our focus to the most recent period covering 1997 to 2003. The International Trade Division publishes the complete geographical detail for all merchandise trade. This information was used to generate the estimates of exports and imports of goods by geography that are consistent with the IO framework. The Balance of Payments (BOP) Division publishes adjustments to merchandise trade and trade in services between Canada and all other countries for the travel, commercial, transportation, and government services accounts. Commodity detail for these accounts is only published for six geographical areas: the United States, United Kingdom, Other European Union, Japan, Other OECD, and Other Countries. Given the growing importance of trade with China, we split out the geographic area of China based on the totals in the major accounts and the commodity detail from the category Other Countries. Since trade with China is dominated by goods and services represent a small value, this method is unlikely to have any serious consequences on the quality of the overall estimates of trade with China. Due to a difference in coverage, the IO estimates of total trade are higher than the BOP estimates of total trade in merchandise and services by the value of ‘financial intermediation’ services. The 22 value of ‘financial intermediation’ services was allocated to geographic regions based on the closest proxy, the BOP estimate of the value of the item 'financial services other than insurance'. Finally, the RAS method was used to ensure that the data on geographic areas match both the commodity dimension of the IO tables and the total trade by geographic area reported by the primary source data. The implicit price indices from the current and constant price IO tables were used to deflate the trade by geographic area. Employment The Canadian Productivity Accounts publish the KLEMS database, which provides estimates of capital, labour, energy, materials, and services that are consistent with the industrial classifications of the IO tables. Labour inputs by industry are classified by education level in terms of hours worked. Prior to their use in the model, the data are converted to full-time equivalent jobs based on average hours worked. The database, however, is only available at a higher level aggregation of 124 SIC industries and 89 NAICS industries. The results of the model were therefore summed to these higher aggregation levels prior to the application of the education coefficients by industry. 23