Advertisement

Journal of International Business Policy

, Volume 2, Issue 2, pp 142–166 | Cite as

Implications of Canada’s restrictive FDI policies on employment and productivity

  • Walid HejaziEmail author
  • Daniel Trefler
Article

Abstract

The merits of foreign direct investment (FDI) have been well documented, thus explaining the openness of many countries to foreign capital. There is, however, some debate on the importance of FDI into key basic business services sectors such as financial services, air transportation services, and telecom services, which remain protected in many countries, including Canada. These restrictions, together with the presence of a review mechanism for foreign investment, has resulted in Canada being ranked above average in terms of restrictiveness by the OECD. We estimate that a leveling down of Canadian FDI restrictions to average OECD levels would raise Canadian labor productivity by a very large 0.79%. It would also either raise employment by 137,400 jobs or raise annual earnings for each Canadian worker by CA$648, or $9.6 billion economy-wide. These results therefore highlight the merits of FDI to local economies.

Keywords

FDI protectionism productivity employment 

INTRODUCTION

There has been an active debate within media, policy and academic circles on the costs and benefits of inward FDI. Although it is generally accepted that inward FDI delivers significant benefits, there remains significant debate within the Canadian context around the benefits from liberalizing three basic business services sectors, namely financial services, telecommunication services and air transportation services. These sectors continue to be heavily restricted to foreign participation.1

According to the OECD’s FDI restrictiveness measure, Canada is considered to be a restrictive country, ranking well above average, owing to two main reasons. First is the mere presence of a review mechanism. Given that only four foreign investment proposals have been formally turned down by the Canadian government resulting from a review,2 some have argued that the OECD’s FDI restrictiveness measure is flawed. Nevertheless, the mere presence of a review mechanism is in effect a restriction on FDI, even though very few investments are formally turned down. There are many potential takeovers of Canadian assets that do not go forward for a myriad of reasons, including signals that the required “net benefit” test within the Investment Canada Act will not be met.3 The second is the fact that these three basic business services sectors have significant formal restrictions on foreign ownership.

Academics have argued extensively on the importance of these three critical basic business services sectors. Since all firms and organizations in the economy interact with these three sectors, inefficiencies therein will have a “magnified effect” on the overall economy. For example, it is argued that new technologies developed and deployed in these three sectors would have impacts across the entire economy given their significant linkages to the overall economy – i.e., they provide critical basic business services to all firms across the economy. As such, it is important that the conditions be set so that these sectors operate as efficiently, and are as globally competitive, as possible.

The objectives of this paper are as follows. First, we benchmark Canada’s FDI performance both at the aggregate and industry levels. There is relatively little information available on how open Canada is, measured as inward FDI to GDP, at the industry level and how Canada compares to other countries. We show that there are many sectors where Canada is quite open to inward FDI, sectors which have done quite well. Yet there are other sectors that remain quite closed to inward FDI. This comparison will therefore inform policymakers on how Canada compares to other countries at both the aggregate and industry levels.

Second, we review OECD measures of FDI restrictiveness at the aggregate and industry levels and across countries. Contrary to the perception that Canada is wide open to FDI, we show that in fact Canada remains quite restrictive. Third, we estimate the impact that restrictions on inward FDI into Canada have on the Canadian economy. To do this robustly, we analyze data on inward FDI into each of 18 OECD countries and across 48 industries used in our sample.4 By utilizing input–output matrices for each country, the impact of FDI restrictions can be measured on the own industry – the industry receiving the FDI, as well as on upstream and downstream industries.

The analysis on the effects of FDI is undertaken in three stages. The first stage develops and estimates an econometric model using the OECD’s measures of FDI restrictiveness. This analysis provides insights into how changing these restrictions would impact FDI into Canada. The second stage estimates the impact that FDI has on the Canadian economy, and more specifically, on employment and labor productivity. Furthermore, the estimated impact that FDI is having on the Canadian economy is broken down into the direct industry effect, the upstream effects, and the downstream effects. The third stage of the analysis integrates the first and second stages and measures the impact restrictions on FDI are having on the Canadian economy.

In order to measure the impact of FDI restrictions on the Canadian economy, we estimate the impact of reducing FDI restrictions to the average within the OECD. This is done first for all sectors, and then again limiting the liberalization to the three critical basic business services sectors (financial services, telecommunications services, and air transportation services). The estimated impact on the Canadian economy of liberalizing FDI across all sectors is very large. We estimate that a leveling down of Canadian FDI restrictions to average OECD levels would raise Canadian labor productivity by a very large 0.79%. It would also either raise employment by 137,400 jobs or raise annual earnings for each Canadian worker by CA$648, or $9.6 billion economy-wide. This $9.6 billion is a benefit that is reaped again and again in each year after liberalization. Approximately 28% of these gains would come from liberalizing the three critical basic business services sectors. These results therefore highlight the benefits of foreign direct investment to local economies, and the importance of ensuring that basic business services sectors are operating as efficiently as possible.

The format of this paper is as follows. “Literature review” provides a literature review on studies that consider the effects of trade liberalization and also measure the impact of inward FDI on local economies. “Canada’s experience with FDI” reviews Canada’s experience with FDI. “Measuring restrictions on FDI” provides a discussion of measures of FDI restrictiveness, and compares Canada to other developed economies. “Estimating the impact of inward FDI on labor productivity and employment” discusses upstream and downstream linkages across sectors in the Canadian economy and estimates the impact of FDI restrictions on the Canadian economy. “Conclusions” concludes.

LITERATURE REVIEW

While this paper focuses on the impact of FDI restrictions on inward FDI in the first stage, and the impact of inward FDI on productivity and employment within host economies in the second stage, it is helpful to also briefly consider the impact of tariff reductions and trade agreements on local economies. Trefler (2004) measures the impact of the Canada–US FTA on both employment and productivity. Those industries within Canada which saw the largest reductions in tariffs experienced a contraction of low-productivity plants and a reduction in employment of 12%. As a result of these changes, however, these industries experienced a 15% increase in labor productivity. Trefler therefore underscores how liberalized trade results in long run benefits reflected in improved productivity and more efficient manufacturing plants, while at the same time recognizing the short-term adjustment costs involved which are borne by “displaced workers and struggling plants”.5,6

The benefits which flow from trade liberalization and tariff reductions have been pursued by policy makers globally as a way to improve productivity and efficiency, and significant progress has been made in this regard. While further tariff reductions have stalled, there remains significant potential benefits that could flow from reductions in non-tariff barriers and the lifting of restrictions on FDI (Ahn et al. 2016).

Büthe and Milner (2008) focus on the commitment that comes with trade agreements as a mechanism to encourage FDI into local economies. It is argued that commitments which come with trade agreements are more credible than would flow from domestic policies which encourage FDI directly, as reneging on trade agreements would be far more costly to a country than simply backing out of a domestic policy. The evidence indicates that countries which are members of the WTO and party to preferential trading agreements attract more FDI, even after controlling for country-specific characteristics.

There have been many studies that identify benefits which are associated with inward FDI, including the following. Inward FDI is an important source of R&D diffusion (Coe & Helpman, 1995; Barrell & Pain, 1997; Eaton & Kortum, 1999; Gera, Gu, & Lee, 1999; Hejazi & Safarian, 1999; van Pottelsberghe & Lichtenberg, 2001). Foreign firms have also been found to have higher levels of productivity and trade propensities relative to Canadian firms (Baldwin & Saborin, 2001; Tang & Rao, 2001; Trefler, 1999). Inward FDI also contributes to domestic capital formation (Desai, Foley, & Hines, 2005; Hejazi & Pauly, 2003). In short, FDI has been shown to be important in many dimensions for host economies.

The benefits of inward FDI depend on the characteristics of local economies. Borensztein, Gregorio, & Lee (1998) find that inward FDI flows contribute to economic growth, but that the benefits depend on the host having a threshold level of human capital. The impact on inward FDI will also depend on host country trade policies. Bhagwati (1978) hypothesizes that the impact of inward FDI will be larger in countries that are pursuing export-promoting policies and smaller in countries which pursue import-substitution policies. This hypothesis finds support in the literature (Balasubramanyam, Salisu, & Sapsford, 1996; Kohpaiboon, 2003).

The literature of the past decade is filled with evidence relating to the positive productivity spillovers associated with inward FDI. A fundamental underlying assumption within the literature is that foreign-controlled firms are, on average, more productive than domestically-owned firms. It has been demonstrated by Helpman, Melitz, and Yeaple (2004), in their seminal heterogeneous firm theory, that low-productivity firms produce domestically, firms with higher productivity export and firms with the highest productivity carry out outward investment. Empirically, Rao and Tang (2005) found that the total factor productivity of multinationals in Canada is on average about 20% higher than Canadian-controlled firms. Consequently, it is widely recognized that there is room for Canadian-controlled firms to learn from foreign firms, especially from those which are operating at the technological frontier. Although the geographical source of Canada’s inward FDI has become increasingly diversified over the past two decades, the United States is still the dominant investor in Canada, accounting for more than half of Canada’s inward FDI stock.

Technological transfers can take place both “horizontally” and “vertically”. A horizontal transfer is known as the knowledge spillover effect among firms in the same industry, and occurs when domestically-owned firms are able to imitate innovations which come from foreign-owned firms, thus allowing them to increase productivity without having to incur the R&D costs. Being able to realize these spillover benefits depends heavily on the absorptive capacity of domestically-owned firms, as well as the willingness of foreign firms operating at the technological frontier to collaborate with competitors (Lileeva, 2010). In Canada, only 1% of foreign-owned firms have Canadian competitors as their R&D partners (Baldwin & Hanel, 2000, 2003). The empirical evidence concludes that the negative pro-competitive own-industry effects indeed dominate the positive intra-industry knowledge spillover effects (Aitken & Harrison, 1999; Lileeva, 2010).

A vertical linkage in the current context is the benefit a domestically-owned firm captures from their upstream foreign suppliers or downstream foreign customers. Forward linkages relate to the impact of FDI into a firm’s upstream suppliers’ industries, and capture both the increased competition in that industry as well as the diffusion of advanced technologies. By lowering the cost and improving the quality of intermediate inputs, these linkages thus promote efficiency of domestically-owned firms. On the one hand, some foreign firms work to limit any leakage of their technology to domestic competitors. On the other hand, however, they may have a strong incentive to share their proprietary technology with upstream suppliers so as to improve both quality and productivity. Furthermore, by making their technology much more widely available, the foreign firm can mitigate any hold-up situations that can arise when they have a single supplier (Blalock & Gertler, 2008).

Domestic firms can also benefit from FDI into downstream industries, either because foreign customers demand higher quality, or when these technologically advanced foreign customers provide the blueprints needed in production of intermediate goods to domestically-owned firms (Rodriguez-Clare, 1996). FDI into downstream industries can increase productivity of domestically-owned firms even if domestic firms ultimately are not initially chosen as suppliers for the foreign firms. This can occur when the resulting competitive pressures spur domestically-owned firms to improve their performance in order to win downstream foreign customers. Chung, Mitchell, and Yeung (2003) found that downstream Japanese FDI into automotive industries had a positive effect on the US automotive components industry. Baldwin and Hanel (2000) have reported evidence in favor of vertical transfers of technologies using Canadian data: 23% multinationals are reported as having R&D collaborations with suppliers, 29% report customers, and 43% report Canadian universities.

By far, studies examining all three channels – own-industry competition channel, horizontal-linkages channel, and the vertical-linkages channel – are altogether a small number. Working with Lithuanian firm-level data, Javorcik (2004) was the first to find significant positive productivity effects through backward linkages, that is, between foreign-owned firms and their local suppliers. In contrast, there was no evidence found in favor of both horizontal spillovers or forward linkages. Blalock and Gertler (2008) found welfare gains from FDI through technology transfer to local suppliers, using plant-level panel data for Indonesian manufacturers. Their estimates suggest a 2% productivity gain to local suppliers from downstream foreign-owned firms. As in the previous study, there was no clear evidence supporting horizontal spillovers.

Blalock and Simon (2010) went further to examine how Indonesian firms’ production capacities and absorptive capacities influenced the propensity to benefit from FDI into downstream industries. They argue that, due to firm heterogeneity, some firms are better able than others to benefit from the presence of foreign firms. Their results indicate that, while production capacities weaken the benefits, absorptive capacities (measured by human capital and R&D expenditures) strengthen these downstream benefits. There was no evidence found of any significant productivity effects moderated by a firm’s complementary capacities.

Liu, Wang, and Wei (2009) studied the impacts of inward FDI on firm-level productivity of Chinese manufacturers. They found that both state-owned and non-state-owned domestically-owned plants are negatively impacted by competition from foreign firms operating in the same industries. All else constant, they document that a one-standard-deviation (14.5%) increase in their measure of backward linkages results in a 23% increase in TFP growth of the local Chinese manufacturing firm, and a one-standard-deviation (5.3%) increase in their forward linkage variable leads to an 8.3% increase in TFP growth.

Lileeva (2010) estimated the effects of US FDI into Canada on productivity of domestically-owned manufacturing plants, with a panel dataset at the level of 145 manufacturing industries. She found that the downstream effects are most pronounced for firms which purchased science-based intermediate goods. US FDI into upstream industries also increased productivity of Canadian manufacturing firms. The negative intra-industry pro-competitive effects are more than offset by the inter-industry transfers of technologies. She also found that a 1% increase in the share of production by US-controlled plants leads to 0.128 percentage points of productivity growth reduction for Canadian-controlled plants in the same industry. However, a 1 percentage point change in upstream inward FDI increases productivity of downstream plants by 0.625 percentage points. No significant backward linkages were documented.

The Current Study

The distinguishing features of the current study relative to those reviewed are as follows. First, it explicitly assesses the impacts of FDI restrictions on domestic economies using a two-stage procedure. This is the first study to do both stages in a single analysis. Second, this study extends this line of analysis to include 18 OECD countries and 48 manufacturing and services industries, thus highlighting the important role played by service industries in modern developed economies. It also highlights the prevalence of FDI restrictions within service industries. As a result, the study is much broader in country and industry coverage. Third, the current study measures the impacts of inward FDI on employment, and productivity, thus providing more complete insights on the overall effects of inward FDI. The analysis therefore better informs the policy making process.

CANADA’S EXPERIENCE WITH FDI

Canada’s Inward FDI Shares

The share of the World’s FDI stock locating in Canada has fallen from 15% in 1970 to 7.8% in 1980, 5.4% in 1990, and then to below 3% in the late 1990s.7 Canada is the only G7 country that experienced such a persistent fall in its share of FDI over this period. This result continues to hold even when Canada is compared to a group of smaller OECD countries.

This trend reversed and Canada’s share began to increase over the 2000–2009 period, but fell again thereafter. As will be noted below, this increase in Canada’s share is coincident with the fall in Canada’s FDI restrictiveness, as reflected in the OECD’s restrictiveness index and the growth in FDI in Canada’s natural resources sector, which correlated closely with global energy prices.

Sectoral Considerations

The share of FDI locating in Canada within Manufacturing has been falling: although in 1999, 43% of the FDI in Canada was in Manufacturing, this has fallen to 30% in 2010, and further to 21% in 2017 (Figure 1). In contrast, Management of Companies and Enterprise8 has seen its share rise from 10% in 1999 to over 19% in 2010, and increasing slightly through 2017 to 21%. Similarly, Mining and Oil and Gas extraction has seen its share rise from approximately 8.5% to 19% over the 1999–2010 period, and rising through 2017 to 20%.
Figure 1

Distribution of inward FDI stocks within Canada, 1999–2017 (ranked by 2017 shares). Source: DFAIT/DFATD website (http://www.international.gc.ca/economist-economiste/statistics-statistiques/investments-investissements.aspx).

The sectors of most concern vis-à-vis the analysis in this paper are financial services, telecom services and air transportation services. Finance and Insurance have seen their shares remain in the 13–17% range: in 1999 accounting for 14% of all FDI in Canada, 13.2% in 2010, and 16.6% in 2017. The sectors which together are classified as Information and Communications Technology (ICT)9 have seen their share fall steadily, from around 8.6% in 1999 to 2.9% in 2012, and 1.9% in 2017. Information and Cultural Industries, and Transportation and Warehousing have small shares, combined at 2.4% in 2017. Therefore, FDI into the three critical sectors are clearly lagging. In the empirical analysis below, the focus will only consider the provision of services within the three basic business services sectors of interest, and excludes production of computers and telecom equipment.

Comparing Canada to Other Countries: Openness at the Industry Level

In this section, we compare the openness across countries for the three key sectors of interest, namely financial services, telecom services and air transportation services. We define openness to FDI for any given industry as the ratio of inward FDI into an industry in a country relative to industry output for the year 2005. Since I/O tables are not available for countries through time over the time period used in this analysis, it is not possible to get such FDI openness at the detailed industry level measured as a time series.

These industry-level data have allowed us to measure openness of countries to FDI at the industry level and hence to compare how open Canada is relative to other countries. These industry level comparisons have been for the most part absent in the public debate on FDI as well as in much of the academic literature, and hence our effort in this respect fills an important void. Each panel of Figure 2 provides openness measures for each of the three critical infrastructure industries across the reporting countries, with Canada’s openness highlighted.
Figure 2

Openness to FDI (industry FDI/industry output), 2005. Source: Authors’ calculations.

Openness in Financial Intermediation is highlighted in Panel A. It highlights how low Canada ranks in terms of openness for FDI: on this measure, Canada ranks 19th among 29 developed countries reporting. Furthermore, the gap between Canada and the most open economies is quite large – the most open economies in this sector are Estonia, Finland, Switzerland, and the Netherlands. This ranking is quite consistent with the general perception that Canada’s financial markets are heavily restricted to foreign participation.

The ranking for Auxiliary Financial Services is similar to that in Financial Intermediation: Canada ranks 10th out of 29 countries ranked. Furthermore, Canada’s openness in this sector is very low relative to the most open economies, namely Sweden, Ireland, the UK and Norway.

In the case of Post and Telecommunications, Canada ranks 19th, in Water Transport, Canada ranks 14th, and in Air Transport, Canada ranks 10th. The authors have also constructed measures of openness for a detailed set of industries.10 The most open industry in Canada is Office, Accounting & Computing Machinery, with a measured openness of 129%, resulting in a ranking of 4th. Mining and quarrying (energy) has an openness measure of 45%, giving Canada a rank of 6th. Mining and quarrying (non-energy) has an openness measure of 30%, giving Canada a rank of 10th. This analysis also shows tremendous heterogeneity – some Canadian industries are very open to FDI, whereas others are much less open. Many of the sectors that are quite open to FDI have done quite well in the presence of foreign participation. This of course raises the questions, why do there remain restrictions on foreign entry into these most restricted sectors, and why does there remain such resistance to reducing or even eliminating these restrictions? It is clear from the experience of other sectors that inward FDI brings many benefits and, furthermore, these other sectors have survived liberalization.

MEASURING RESTRICTIONS ON FDI

The OECD’s Measure of FDI Restrictiveness

The OECD’s 2010 measures of FDI restrictiveness is an update to its previous 2006 measures. It is available for all OECD countries and the number of sectors covered was expanded. It is composed of four components: equity restrictions; screening and approval requirements; restrictions on foreign key personnel; and other operational restrictions (such as limits on the purchase of land or on repatriation of profits and capital). These restrictions are scored whether or not they are discriminatory towards foreign investment. The reason for this is that to the extent any non-discriminatory measures are burdensome to foreign investors, the net impact will be that it will limit foreign investment. An example noted in the OECD’s report is rules around the nationality of directors. Such requirements would be tantamount to a restriction on FDI as it raises the costs associated with investing into any particular economy.

In reviewing the OECD’s restrictiveness measures for Canada, two notable points emerge. First, Canada’s measure of restrictiveness has fallen dramatically over this period. The second observation is that, even with this fall, Canada remains slightly more restrictive than the average within the OECD, and is more restrictive than all other G7 countries except Japan. The most restrictive countries ranked are Iceland, Japan, New Zealand, and Mexico.

Foreign investment into Canada that exceeds a given threshold must comply with the Investment Canada Act (ICA). Although Canada has traditionally been regarded as quite restrictive, there have been some recent changes that have reduced Canada’s level of restrictiveness. For example, the threshold for the investment that would trigger an automatic review under the ICA has been increased, meaning that fewer investments would face such a review.

Canada does, however, continue to have in place many restrictions that are sector-specific, such as equity restrictions in banking and finance, and it also has the authority to review investments made by foreign governments or related parties, as well as any other foreign investment that can threaten Canada’s national security. In short, therefore, Canada continues to rely on both sector-specific restrictions and the automatic policy review for investments above the threshold value.

The national security test screens investments that may have implications for Canada’s national security, and hence is subject to Ministerial and potentially Cabinet review, although no definition of what constitutes “national security” is given in the ICA. However, it is possible that investments impacting Canada’s sovereignty, national defense and potentially strategic sectors of the economy (such as natural resources), and investments by state-owned enterprises, may be considered under the national security test (Ram, Rao, Bhattacharjee, & Wright, 2010). Given the nature of these regulations, there is some concern that they could be used by the government for protectionist or politically-driven goals. This is in line with the theme that the mere presence of a review mechanism can limit foreign investment. However, it is unclear what impact such rules have been on FDI into Canada.

It is also important to understand what explains the fall in Canada’s ranking on the OECD’s FDI restrictiveness index. As noted in the OECD’s report, “the score for Canada is significantly reduced by its elimination of exceptions to national treatment in financial services.” (Kalinova, Palerm, & Thomsen, 2010).

Sectoral Restrictions

The OECD report also provides measures of restrictions for each of the key sectors. For the total FDI index (i.e., aggregating across all sectors), among G7 countries, only Japan is more restrictive than Canada. In contrast, Germany, France, the UK and Italy are far more open. In considering how restrictive Canada is at the sectoral level, Canada is ranked as the most restrictive in eight industries: Manufacturing, Construction, Distribution, Surface Transport, Hotels & Restaurants, Media, Telecom, Financial Services and Business Services.

In terms of the three basic business services sectors, the US and Japan are more restrictive in Transport than Canada, whereas Germany is slightly less restrictive. In the case of Financial Services and Telecom, Canada is more restrictive than all the other G7 countries.

A Note on Manufacturing

The openness of manufacturing must be highlighted at this point. Manufacturing in Canada is heavily dependent on foreign investment. Let’s consider the reported shares for Canadian manufacturing controlled by foreigners. Specifically, in the year 2005, the share of production in Canada accounted for by MNEs was 51.2% and the shares of MNEs gross operating surplus was 55.2%. In 2007, 54.5% of all assets in manufacturing were under foreign control, 54.1% of operating revenue, and 52.4% of operating profits. In other words, manufacturing in Canada is more than 50% foreign owned or controlled.11 This openness in manufacturing is in sharp contrast to the OECD’s FDI restrictiveness ranking, which puts manufacturing within Canada as the most restrictive among G7 countries, and one of the most restrictive within the OECD. This discrepancy is due to the presence of a review mechanism in Canada, and thus an increase in the OECD’s restrictiveness score across all sectors. However, it would be inappropriate to treat Canada’s manufacturing sector that restrictive given it is 50% foreign-owned/-controlled.12 As such, we adjust the restrictiveness measure for Canada’s manufacturing so that it is equal to the average restrictiveness measure for manufacturing within the OECD. This qualification will be discussed further in the “Empirical framework” section.

ESTIMATING THE IMPACT OF INWARD FDI ON LABOR PRODUCTIVITY AND EMPLOYMENT

Inward FDI creates both pressure and support for incumbents in an industry. It pressures incumbent firms by forcing them to improve operational efficiency and rethink value creation. It supports incumbent firms by bringing to Canada sophisticated foreign suppliers and demanding foreign customers. Sophisticated suppliers provide existing services less expensively and introduces new services that were previously unavailable in Canada. Demanding customers leverage their experience as users to help Canadian producers do better.

This is illustrated in Figure 3, which describes, as an example, a small Canadian firm in the web design sector. The firm purchases services from the ‘upstream’ telecom sector and sells its designs to the ‘downstream’ retail apparel sector. FDI in the telecom sector creates competition that allows web designers to purchase a wider variety of innovative services and to do so at a lower cost. In this way, it frees up the web designer’s resources so that the firm can devote more energy to what it does best – innovative design at a low cost. Further, FDI in the retail apparel sector brings sophisticated demanders of web design services to Canada. These experienced retailers insist on unique and innovative web designs, which supports Canadian web designers in creating value.
Figure 3

Upstream and downstream support.

From this discussion, we can identify three channels through which FDI affects labor productivity and employment.
  1. A.

    The ‘Upstream-Industry’ Channel: FDI brings to Canada sophisticated suppliers of services that support the industry under study. For example, web designers are helped by FDI in the upstream telecom sector.

     
  2. B.

    The ‘Downstream-Industry’ Channel: FDI brings to Canada demanding customers that drive firms in the industry under study to efficiently produce more innovative products. For example, web designers are helped by FDI in the downstream retail apparel sector.

     
  3. C.

    The ‘Own-Industry’ or ‘Head-On Competition’ Channel: FDI into the industry under study intensifies head-on competition in the industry. This occurs, for example, when a foreign web design firm sets up in Canada and competes head-on with the Canadian web designer. It is ‘own-industry’ in the sense that the FDI occurs in the web designer’s own industry rather than in an upstream or downstream industry.

     

In this section, we assess the relative importance of these three channels for Canada and the impact a liberalization of the FDI regime will have on the ability of Canadian firms to improve efficiency and create value.

Empirical Framework

Let Fic be the stock of inward FDI into industry i in country c. Let yic be an outcome of interest such as labor productivity or employment. Stated simply, if our interest were exclusively on the own-industry channel then we would consider a log-linear regression of the form:
$$\ln y_{ic} = \alpha_{i} + \beta \ln F_{ic} + \gamma X_{c} + \varepsilon_{ic}$$
(1)
where αi is an industry fixed effect and Xc is a set of covariates that control for country characteristics. β captures the effect of FDI on labor productivity or employment. By including an industry fixed effect, we are examining how cross-country differences in FDI within an industry lead to cross-country differences in labor productivity or employment. This is precisely the type of conceptual experiment that is our focus.
For β to have a causal interpretation – FDI causes improvements in labor productivity or employment – we must address the issue of endogeneity. The simplest cause of endogeneity is omitted variable bias. Suppose that a country has a unique characteristic that both increases its FDI and its productivity. For example, the UK may have wonderful institutions that both protect foreign investors and raise productivity. Such a characteristic would ideally be captured by the Xc, which in practice will include GDP and GDP per capita; however, if the characteristic were not fully captured by these Xc, then it would enter the residual ɛic and lead to a correlation between the residual and lnFic that would bias the estimate of β. To mitigate this bias, one could include country fixed effects αc:
$$\ln y_{ic} = \alpha_{i} + \alpha_{c} + \beta \ln F_{ic} + \varepsilon_{ic} .$$
(2)

Below, we will also outline and implement a more rigorous instrumental variables strategy for dealing with the endogeneity of FDI.

The Upstream-industry Channel

In order to examine the upstream-industry channel, we need to link FDI in each upstream industry j (e.g., telecom) to productivity and employment outcomes in industry i (e.g., web design). Specifically, for each industry i, we need to create a weighted average of the FDI in industries that are upstream from i:
$${\text{Upstream}}_{ic} = \sum\nolimits_{j} {\omega_{ijc} \ln F_{jc} }$$
(3)
where the weights sum to 1 (∑jωijc = 1) and are proportional to how much industry i buys from upstream industry j. The size of ωijc measures the ‘upstream-ness’ of j from i’s perspective.13,14
Let Mijc be the dollar amount of intermediate inputs that sector i buys from its upstream sector j. The Mijc are data available from the OECD input–output tables. Let aijc be the share of intermediate inputs that i buys from its upstream sector j. Then,
$$a_{ijc} = \frac{{M_{ijc } }}{{\mathop \sum \nolimits_{{j^{'} \ne i}} M_{{ij^{\prime}c}} }}$$
(4)
where we exclude i’s purchases from itself when calculating the share. As in Eq. (3), the weighted average of FDI into sectors that are upstream from i’s perspective is:
$${\text{Upstream}}_{ic} = \mathop \sum \limits_{j \ne i} \frac{{a_{ijc} }}{{\mathop \sum \nolimits_{j' \ne i} a_{ij'c} }}\ln F_{jc} .$$
(3′)

This is just Eq. (3) with ωijc = aijc/∑j′≠iaijc. The denominator ∑j′≠iaijc ensures that the weights sum to unity given that we are excluding i’s purchases from itself and that there exist non-comparable imports.15,16

The Downstream-industry Channel

In order to examine the downstream-industry channel, we need to link FDI in each downstream industry j (e.g., Retail) to productivity and employment outcomes in industry i (e.g., Web Design). In particular, we define Downstreamic as the weighted average of FDI in industries that are downstream from i’s persepective:
$${\text{Downstream}}_{ic} = \mathop \sum \limits_{j \ne i} \theta_{jic} \ln F_{jc}$$
(5)
where the weights sum to one (∑jθjic = 1) and are proportional to how much industry i sells to downstream industry j. The size of θjic measures the ‘downstream-ness’ of j from i’s perspective.

To understand the weights, return to Figure 3, Panel B. Web Design sells $2 of services to Telecom and $8 of services to Retail. It thus sells downstream a share of 0.2 [= 2/(2 + 8)] to Telecom and a share of 0.8 [= 2/(2 + 8)] to Retail. Again, letting Telecom, Web Design, and Retail be sectors 1, 2, and 3, respectively, we have θ12c = 0.2 and θ32c = 0.8.

A more formal definition of the θjic will now be developed. Let Mjic be the dollar amount of intermediate inputs that sector i sells to downstream sector j. Let bjic be the share of industry i’s output that is sold to industry j:
$$b_{jic} = \frac{{M_{jic} }}{{\mathop \sum \nolimits_{{j^{' \ne i} }} M_{j'ic} }}$$
(6)
where we exlcude own-industry sales Miic.17 From Eq. (5), the downstream channel is the weighted average of FDI into downstream sectors with weights θjic = bjic/ ∑ j’≠ibjic. The weights exclude a sector’s sales to itself and include a denominator to ensure that the weights sum to unity. Equation (5) thus becomes:
$${\text{Downstream}}_{ic} = \mathop \sum \limits_{j \ne i} \frac{{b_{jic} }}{{\mathop \sum \nolimits_{j' \ne i} b_{j'ic} }}\ln F_{jc}$$
(5′)
These general equilibrium variables allow us to estimate an extended version of Eq. (2), one that includes not only the own-industry channel but also the upstream- and downstream-industry channels:
$$\ln y_{ic} = \alpha_{i} + \alpha_{c} + \beta_{O} \ln F_{ic} + \beta_{U} {\text{Upstream}}_{ic} + \beta_{D} {\text{Downstream}}_{ic} + \gamma X_{c} + \varepsilon_{ic}$$
(7)

The Data

For each country, input–output tables as well as industry-level FDI stocks, employment, value-added, and output are from the OECD. The OECD attempts to ensure comparability of industrial classifications across countries. However, the comparability is not sufficient for our specific purposes. We therefore invested heavily in improving the cross-country comparability of the industrial classifications by choosing levels of industrial aggregation that are more appropriate given our focus on FDI and our focus on sectors of public policy relevance to Canada. Our starting point was the industry classification used in the 2005 input–output tables. Our final classification system has 48 industries.

One problem with input–output tables is that data on inter-industry purchases do not distinguish between purchases from domestic producers and purchases from foreign producers. We know total imports, but we do not know which sectors purchase them. Various methods have been proposed to impute data that allow one to distinguish between the two (see Trefler & Zhu, 2010 for a review). Unfortunately, a small-scale survey that was administered in order to check the reliability of existing imputations shows that they are not at all reliable (Puzzello, 2012).

Approximately 50% of Canada’s imports are imported by wholesalers and retailers who then sell them to final users. Unfortunately, once a good enters Canada and is sold from one Canadian to another, we can no longer identify whether or not it is a foreign good. Why does this matter? If sectors with high levels of FDI tend to import more (less), it will bias upwards (downwards) Upstreamic and Downstreamic. However, it is not clear how large this bias is or whether it will have any impact on our estimates.

The input–output tables also contain information on the value of sales and the value of output, which are needed to compute the ajic and the bjic. Value added are from the input–output tables. Employment data are from the OECD’s STAN database. We use what the OECD refers to as “number of workers engaged” rather than “number of employees” because the former is much more complete in its coverage. We define labor productivity as the log of value added divided by employment.

The input–output tables, labor productivity, and employment data are from 2005. Inward FDI stock data at the industry level are from 2005–2010. The FDI restrictiveness measures, at the industry level, are from 2006 to 2010, as there is no 2005 measure available.18 In cases of missing data for FDI stock, the average for the years 2003–2007 was used for missing 2005 FDI data, and the average for the years 2008–2009 was used for missing 2010 FDI data.19

Upstreamic and Downstreamic capture complicated general equilibrium interactions within the economy. They can therefore only be calculated for countries with (1) relatively complete data on FDI by industry and (2) relatively disaggregated input output tables. Eighteen countries have sufficiently complete data that we can reliably calculate Upstreamic and Downstreamic. These are Austria, Canada, Czech, Denmark, Estonia, Finland, France, Greece, Hungary, Netherlands, Norway, Poland, Slovakia, Slovenia, Spain, Sweden, UK, and USA. In addition, data on labor productivity and employment are not available for all industries in all countries. This leaves us with 649 observations, each corresponding to a unique industry–country (i,c) pair.

Data on FDI restrictions are from the OECD FDI Restrictiveness Index, which in turn is a component of the OECD’s Product Market Regulation database. The FDI Restrictiveness Index calculates industry-level barriers in four areas: equity restrictions, screening and approval requirements, restriction on foreign personnel, and other operational restrictions. Data are available for 2006 and 2010.20 The analysis here uses the 2005 and 2010 FDI data and the 2006 and 2010 OECD restrictiveness data.

Results for the ‘Own-industry’ Channel

Table 1 reports the results of estimating Eqs. (1) and (2). The dependent variable is either the log of labor employment (columns 1 and 2) or the log of productivity (columns 3 and 4). For each dependent variable, there are two columns corresponding to Eq. (1) with industry fixed effects and Eq. (2) with country and industry fixed effects. Consider the results for log employment. In column 1, we see that FDI in industry i has a statistically significant and positive impact on employment in industry i. The two controls for country characteristics, GDP and GDP per capita, are also statistically significant. Interestingly, their coefficients are almost identical but of opposite sign. This means that all that matters is country size as measured by population.21 In column 2, country fixed effects are added to the specification. This means that we can no longer include any country characteristics because such characteristics are ‘absorbed’ into the country fixed effects. Specifically, we no longer include GDP or GDP per capita. The resulting estimate of 0.0782 is very similar to what we saw in column 1, which is to say that it is statistically significant.
Table 1

Estimates of the own-industry channel

 

Employment

Labor productivity

(1)

(2)

(3)

(4)

ln(FDIic)

0.0640*** (6.40)

0.0782*** (6.36)

0.351*** (18.68)

0.0670*** (6.78)

ln(GDPc)

0.937*** (45.07)

 

− 0.438*** (− 11.25)

 

ln(GDPc/Lc) (GDP per capita)

− 0.922*** (− 11.22)

 

− 0.150 (− 0.99)

 

Country fixed effects

No

Yes

No

Yes

Industry fixed effects

Yes

Yes

Yes

Yes

Observations

649

649

649

649

Adjusted R2

0.79

0.79

0.37

0.89

t statistics in parentheses.

***statistical significance at 99%.

Columns 3 and 4 of Table 1 report results for labor productivity. As is clear, there are highly significant impacts of own-industry FDI on labor productivity. The results with country and industry fixed effects (column 4) are much smaller than with just industry fixed effects (column 3). An important question deals with what feature of the data leads to this discrepancy. One explanation, offered above, is that the industry fixed effect specification does not purge the estimates of a spurious correlation between productivity and FDI caused by unobserved country characteristics. However, there is an empirically more plausible explanation.

We are imposing the restriction that FDI has the same impact in all countries, when, in fact, the data want there to be heterogeneous responses; specifically, the data want the response of productivity to FDI to vary by country and to be higher in countries with higher levels of FDI. To see this, note that the standard formula for the asymptotic consistency of our OLS estimator is the following. For simplicity, assume that all variables are measured in deviations from their country and industry means. Our estimating equation can then be written as ln yic = β ln Fic + ɛic. However, suppose that the true model has heterogeneous responses, which means that the β vary across countries: ln yic = βc ln Fic + ɛic. Then, the standard formula for the consistency of our OLS estimator becomes22:
$$p\lim \beta^{\text{OLS}} = \beta + p\lim \frac{{\mathop \sum \nolimits_{i,c} (\beta_{c} - \beta )(\ln F_{ic} )^{2} }}{{\mathop \sum \nolimits_{i,c} (\ln F_{ic} )^{2} }} = p\lim \mathop \sum \limits_{i,c} \beta_{c} \frac{{(\ln F_{ic} )^{2} }}{{\mathop \sum \nolimits_{i,c} (\ln F_{ic} )^{2} }}.$$
(8)

If there are no heterogeneous effects (βc = β for all c), then plim βOLS = β. However, if this is not the case, then βOLS estimates a weighted average of the true effects, and this weighted average will be larger the larger is the correlation between the βc and the ( ln Fic)2. In words, this weighted average will be ‘too’ large if countries with large impacts of FDI on productivity are countries with large levels of FDI.

To investigate, Table 2 reports estimates of the model:
Table 2

Country-level regressions: OLS

Labor productivity

Employment

Country

ln(FDIic)

t statistic

Country

ln(FDIic)

t statistic

Canada

0.195*

(2.35)

Canada

0.320*

(2.25)

Netherlands

0.098*

(2.70)

Slovenia

0.188

(1.64)

Slovakia

0.086**

(3.37)

Austria

0.084

(0.88)

France

0.081**

(3.15)

Greece

0.064

(0.68)

Norway

0.069*

(2.25)

Finland

0.052

(0.70)

Spain

0.062*

(2.04)

Denmark

0.049

(0.75)

Sweden

0.050*

(2.39)

Estonia

0.048

(0.59)

Greece

0.092

(1.62)

Norway

0.041

(0.61)

UK

0.081

(1.77)

Czech

0.0078

(0.13)

Slovenia

0.080

(1.75)

Poland

0.0027

(0.03)

Denmark

0.066

(1.97)

Sweden

− 0.016

(− 0.31)

USA

0.066

(1.74)

Spain

− 0.020

(− 0.26)

Hungary

0.064

(1.97)

Hungary

− 0.029

(− 0.57)

Finland

0.052

(1.86)

Slovakia

− 0.061

(− 1.02)

Czech

0.049

(1.83)

Netherlands

− 0.069

(− 1.01)

Austria

0.048

(1.07)

USA

− 0.071

(− 1.13)

Estonia

0.035

(1.03)

France

− 0.079

(− 1.19)

Poland

0.027

(0.54)

UK

− 0.120

(− 1.24)

Each row represents output for one country: a separate OLS regression for labor productivity and a separate regression for employment. Given there are 48 industries in the analysis, the sample size for each regression would be 48 if data were available for FDI and labor productivity/employment for all 48 industries for that country.

*statistical significance at 90%, **at 95%.

$$\ln y_{ic} = \alpha_{c} + \beta_{c} \ln F_{ic} + \eta_{ic} .$$
(9)

That is, we estimate the model separately for each country. The left-hand panel of the table reports results for log labor productivity, the right-hand panel for log employment. The rows are sorted from the largest to the smallest values of βc. Interestingly, Canada sits at the top of both columns, suggesting that it is a country that benefits most from FDI.23

Results for the ‘Own-industry’ Channel: Instrumental Variables Estimation

We have been reporting OLS estimates. Of course, there are reasons to expect that FDI ( ln Fic) is endogenous. We now turn to identifying a policy-relevant instrument for FDI. We tackle this in two steps. We first argue that changes in the OECD measure of restrictiveness can explain changes in FDI. To this end, we regress 2005–2010 changes in FDI on 2006–2010 changes in the OECD restrictiveness index (recall that there are no 2005 restrictivesnss data.) We then argue that this validates the use of the 2006 OECD restrictiveness index as an instrument for the level of the 2005 FDI stock.

Consider the first step, which involves a regression of the form:
$$\ln F_{ic,2010} - \ln F_{ic,2005} = \alpha^{\prime}_{i} + \rho \left( {R_{ic,2010} - R_{ic,2006} } \right) + \gamma^{\prime}X_{c} + \varepsilon^{\prime}_{ic}$$
(10)
where \(R_{ict} (t = 2006, 2010)\) is the OECD restrictiveness index in year t. We also estimate the comparable regression with country fixed effects in place of γXc. Table 3 reports the results. Column 1 is the specification in Eq. (10). The coefficient on the change in restrictiveness is − 2.118 and is statistically significant (t = − 3.19). The specification with country and industry fixed effects appears in column 2 and is similar. Models such as Eq. (10) in which the dependent variable is a log change often display mean reversion. That is, if FDI were unusually high in 2005, we might expect it to revert back to its steady-state level in subsequent years. To deal with mean reversion, it is common to include ln Fic,2005 as an explanatory variable and expect its coefficient to be negative. This is what appears in columns 3 and 4. As expected, its coefficient is negative. Further, the coefficient on the change in restrictiveness shrinks to approximately − 1.5 but remains statistically significant. Summarizing, the negative and statistically significant results of columns 1–4 mean that FDI can be explained by government restrictions: that is, restrictiveness can be used as an instrument for FDI.
Table 3

First stage regressions

 

Log change in FDI, 2005–2010

Log FDI in 2005: ln(FDIic)

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

Change in FDI restrictiveness 2006–2010

− 2.118** (− 3.19)

− 2.329** (− 3.44)

− 1.542* (− 2.34)

− 1.540* (− 2.33)

    

FDI restrictiveness, 2006

    

− 0.962 (− 1.78)

− 1.430* (− 2.51)

− 0.992 (− 1.82)

− 1.335* (− 2.38)

ln(FDIic): FDI in 2005

  

− 0.142*** (− 5.59)

− 0.216*** (− 7.10)

    

θi ln(Σi FDIic): industry share of world FDI

    

0.850*** (16.35)

 

0.751*** (13.80)

 

ln(GDPc)

0.193*** (3.78)

 

0.278*** (5.34)

 

0.175** (2.60)

 

0.222*** (2.82)

 

ln(GDPc/Lc) (GDP per capita)

− 0.235 (− 1.14)

 

− 0.439* (− 2.13)

 

− 0.418 (− 1.58)

 

− 0.457 (− 1.49)

 

Country fixed effects

No

Yes

No

Yes

No

Yes

No

Yes

Industry fixed effects

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Observations

649

649

649

649

649

649

649

649

Adjusted R2

0.05

0.09

0.09

0.16

0.74

0.76

0.74

0.76

Instrument set type

OLS

OLS

OLS

OLS

R,Z first stage

R,Z first stage

R,Z, δiDc first stage

R,Z, δiDc first stage

t statistics in parentheses.

*statistical significance at 90%, **at 95%, and ***at 99%.

A second reasonable instrument for FDI is the following. Recall that we are concerned that there may be a feature of industry i in country c that both raises FDI and raises either productivity or employment. To be more specific, consider the case of reverse causality in which high productivity in Canada attracts foreign investors (who perhaps wish to learn from Canadians). In this case, our estimate of the impact of FDI on productivity will be upward-biased because it captures both the direct effect of FDI on productivity and the reverse effect of productivity on FDI. Alternatively, if high productivity in Canada discourages foreigners from competing here, then high productivity reduces FDI. In this case, our estimate of the impact of FDI on productivity will be downward-biased because it captures both the positive direct effect of FDI on productivity and the reverse negative effect of productivity on FDI. Clearly, the OLS bias can go in either direction, but our main point is that these examples suggest an instrument.

As the examples illustrate, we need to purge FDI of location decisions, i.e., of FDI that is either high or low because of local productivity. To this end, let δi be the average share of world FDI that is accounted for by industry i:
$$\delta_{i} = \frac{1}{18}\mathop \sum \limits_{c = 1}^{18} \frac{{F_{ic} }}{{\mathop \sum \nolimits_{k = 1}^{18} F_{ik} }}$$
where 18 is the number of countries in the sample. Further, let Fc = ∑ iFic be the total inward FDI stock of country c, i.e., the sum of FDI across all industries i. Then,
$$Z_{ic} = \delta_{i} F_{c}$$
is what country i’s FDI would be if it had a ‘typical’ distribution δi of FDI across industries. That is, Zic is an instrument that purges FDI of the country–industry-specific location decisions.
While δi is exogenous, Fc may be endogenous. To move to an exogenous replacement for Fc, we can replace it with a country dummy. Specifically, let D c k be a dummy variable that equals 1 if k = c and 0 otherwise. With 18 countries, there will be 18 such dummy variables, D c k for k = 1, … 18. With these we can create a set of 18 instruments of the form
$$W_{ic}^{k} = \delta_{i} D_{c}^{k} ,\quad k = 1, \ldots ,18.$$

One major drawback of using the W ic k is that it means using 18 instruments. Given the properties of IV, this will likely create a weak-instruments problem. Restated, this set of instruments solves one problem while creating another.

Columns 5–8 of Table 3 report the first stage of the IV regressions. Columns 5 and 6 use restrictiveness Ric,2006 and ln Zic as the two instruments. Columns 7 and 8 use Ric,2006 and the W ic k as the 19 instruments. Columns 5 and 7 have industry fixed effects, and columns 6 and 8 have industry and country fixed effects. The coefficients on restrictiveness in columns 5–8 are reassuringly similar to those in columns 3 and 4 that were obtained from estimating equations in differences. This is a very good sign that we are correctly interpreting these regressions.

Table 4 reports the instrumental variables (IV) estimates. We consider results which deal with labor productivity first. Column 4 is our OLS result from column 3 of Table 1. Column 5 uses Ric,2006 and ln Zic as instruments while column 6 uses Ric,2006 and the 18 W ic k as instruments. The IV estimates are greater than the OLS estimate. That is, the OLS estimate is downward-biased. Economically, this means that foreign investors are attracted to industry i in country c when productivity is low in that industry relative to what it is on average in other countries. Restated, foreign investors are attracted when a country’s industry has low productivity relative to international standards.
Table 4

Instrumental variables estimation

 

Employment

Labor productivity

(1)

(2)

(3)

(4)

(5)

(6)

Employ

Employ

Employ

Lab. prod.

Lab. prod.

Lab. prod.

OLS

IV

IV

OLS

IV

IV

2005 FDI

0.0640*** (6.40)

0.0298 (1.73)

0.0350 (1.46)

0.351*** (18.68)

0.906*** (18.49)

0.676*** (12.50)

GDP

0.937*** (45.07)

0.958*** (42.31)

0.954*** (38.49)

− 0.438*** (− 11.25)

− 0.778*** (− 12.07)

− 0.637*** (− 11.41)

GDP per capita

− 0.922*** (− 11.22)

− 0.973*** (− 11.38)

− 0.965*** (− 10.85)

− 0.150 (− 0.99)

0.690** (2.83)

0.342 (1.71)

Country fixed effects

No

No

No

No

No

No

 

OLS

R,Z

R,W

OLS

R,Z

R,W

Hausman–Wu (test for endogenous regressor) [Hausman p values]

 

6.144 [0.0132]

1.794 [0.1804]

 

460.697 [0.00]

64.212 [0.00]

Sargan–Hansen statistic for over identification [p value]

 

0.802 [0.37]

14.744 [0.61]

 

2.818 [0.09]

26.244 [0.07]

Cragg–Donald Wald F statistic (test for weak instruments)

 

157.838

6.941

 

157.838

6.941

Observations

649

649

649

649

649

649

Adjusted R2

0.790

0.786

0.787

0.371

0.478

0.080

t statistics are in () and p-values are in [].

***statiscal significance at 99%.

Table 4 also reports the Hausman–Wu endogeneity tests. The large test statistics and the low p values mean that we can reject the null hypothesis of exogeneity. That is, we find that FDI is endogenous, which motivates the use of an IV strategy. The table also reports the Sargan–Hansen J statistic, which is an over-identification test for the validity of the instruments. Essentially, it tests whether the instruments have a direct impact on productivity, in which case the instruments are not valid. The large p values (> 0.05) indicate that the instruments are valid. We also test for weak instruments. Specifically, we report the Cragg-Donald Wald F statistic. Rejection of weak instruments requires a statistic that is larger than 10 or even 20. The column 5 specifications has a value well above 20, but this is not so for column 6.24

Table 4 reports also reports IV results for the log of employment. There is mixed evidence of endogeneity; i.e., endogeneity is rejected at the 1% level in column (3) but not in column (2). We therefore conclude that the OLS estimates of the impact of FDI on employment do not suffer from endogeneity bias.

Summarizing, we have found three results in this subsection. First, the OECD restrictiveness measure is correlated with FDI both in levels and changes. The OLS impact of restrictiveness on FDI varied between − 0.962 (Table 3, column 5) and − 2.329 (Table 3, column 2), but with a strong central tendency of about − 1.5. Second, the use of IV estimators suggests that investors target industries in countries where those industries are underperforming vis-à-vis productivity. That is, there is a strong pro-competitive effect of FDI on labor productivity. This implies that the OLS estimate of the impact of FDI on labor productivity is downward-biased. Third, there is no strong evidence that FDI location decisions target employment; i.e., the OLS estimate of the impact of FDI on employment is likely unbiased.

The ‘Upstream-industry’ and ‘Downstream-industry’ Channels

So far we have concentrated on the ‘own-industry’ channel through which FDI impacts productivity and employment. This was necessarily the most difficult channel to examine because of endogeneity associated with FDI locational choices. Fortunately, endogeneity is less of a concern when dealing with the upstream- and downstream-industry channels, because, for example, the decision of foreigners to invest in a Canadian telecom company is unlikely to be influenced by the productivity of a Canadian web designer who uses the telecom’s services.25 We now turn to estimating the upstream-industry and downstream-industry channels.

Table 5 presents a number of different results for labor productivity and employment, always with industry fixed effects only. Column 1 repeats the familiar specification of Table 1, column 3. Columns 2–5 combine the three channels (own-industry, downstream-industry, and upstream-industry) in various ways. As is apparent, the downstream and upstream channels are always very significant both statistically and economically. Consider column 5, which includes all three channels. Recall that FDI in downstream industries brings demanding buyers to Canada; e.g., FDI in retail creates demanding buyers for web design services. A 10% increase in FDI in every one of an industry’s downstream buyers is associated with a 2.3% increase in labor productivity. Also, recall that FDI in upstream industries brings sophisticated suppliers to Canada; e.g., FDI in telecom creates sophisticated suppliers of services used in web design. A 10% increase in FDI in every one of an industry’s upstream suppliers is associated with a 3.08% increase in labor productivity. These are large effects.
Table 5

Estimates of the own-industry, upstream and downstream channels 

 

Labor productivity

Employment

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

ln(FDIic)

0.351*** (18.68)

   

0.155*** (9.05)

0.0640*** (6.40)

 

0.0595*** (4.95)

Downstream-industry channel (demanding buyers)

 

0.560*** (24.72)

 

0.293*** (9.15)

0.230*** (7.47)

 

0.0516*** (3.66)

0.0122 (0.76)

Upstream-industry channel (sophisticated suppliers)

  

0.594*** (25.99)

0.363*** (10.96)

0.308*** (9.72)

   

ln(GDPc)

− 0.438*** (− 11.25)

− 0.427*** (− 12.30)

− 0.566*** (− 15.97)

− 0.539*** (− 16.17)

− 0.580*** (− 18.33)

0.937*** (45.07)

0.957*** (44.18)

0.935*** (43.07)

ln(GDPc/Lc) (GDP per capita)

− 0.15 (− 0.99)

0.288* (2.02)

0.266 (1.92)

0.404** (3.09)

0.443*** (3.61)

− 0.922*** (− 11.22)

− 0.929*** (− 10.44)

− 0.907*** (− 10.38)

Country fixed effects

No

No

No

No

No

No

No

No

Industry fixed effects

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Observations

649

649

649

649

649

649

649

649

Adjusted R2

0.37

0.60

0.62

0.67

0.71

0.79

0.90

0.90

t statistics in parentheses.

*statistical significance at 90%, **at 95%, and ***at 99%.

Comparing columns 1 and 5, one sees that the own-industry channel has shrunk from 0.351 to 0.155; i.e., part of the own-industry effect in column 1 is actually best attributed to the downstream- and upstream-industry channels.

Looking at all three channels together as in column 5, if FDI increased by 10% in all industries, then labor productivity would rise by 1.55% due to the own-industry channel, plus 2.30% due to the downstream-industry channel, plus 3.08% due to the upstream-industry channel. Summing these, labor productivity would grow by 6.93%. This is an extremely large number.

It is quite possible that these estimates are too large. When country fixed effects are added to the specification, the downstream- and upstream-industry channels become statistically and economically insignificant.26 It is not immediately clear that one should include country fixed effects. Nor is it obvious why including country fixed effects leads to insignificance. To investigate, we consider a semi-parametric estimator of the specification in column 5. Specifically, we estimate the upstream-industry channel non-parametrically. To do so, we first regress labor productivity on all the column 5 regressors other than Upstreamic (i.e., ln FDIic, Downstreamic, the two GDP regressors, and the industry fixed effects) and recover the residuals ɛ ic Prod . We then regress Upstreamic on all the other variables (i.e., ln FDIic, Downstreamic, the two GDP regressors, and the industry fixed effects) and recover the residuals ɛ ic Up ). Finally, we non-parametrically regress ɛ ic Prod on ɛ ic Up . The results appear in Figure 4. The figure shows that the upstream effect only kicks in when Upstreamic is above its conditional mean, which is 0 in the figure. Thus, there appears to be an important non-linearity in the effect of Upstreamic that is missed by the linearity of the column 5 specification with country fixed effects. For clarity in the display, we have not used the 5% of observations with the largest |ɛ ic Up |. This does not and cannot affect the non-parametric results. Also, we implement the non-parametrics using the Lowess smoother.
Figure 4

Upstream effects.

Finally, turning to the employment results in Table 5 (columns 6–8), there is weak evidence for a downstream-industry channel. There is no evidence for an employment-related upstream-industry channel, results that we do not report.

To summarize, there is evidence of important downstream- and upstream-industry channels in operation for labor productivity, but not for employment.

The Impact of Liberalization of FDI Restrictions

We can use our estimated coefficients to calculate what would happen if Canada liberalized its foreign investment regime so that FDI restrictions fell to average levels observed in the OECD. Recall that Rict is the level of FDI restrictions in industry i in country c in year t. In what follows, we always use the most recent year of data and suppress the time subscript. Let Ri,OECD,2010 be the 2010 level of restrictions in industry i averaged across OECD countries. We exclude Canada when calculating this mean.27 Let Ri,CAN be the 2010 level of Canadian restrictions in industry i. Define
$$\Delta R_{i} = \hbox{max} \left( {R_{{i,{\text{CAN}}}} - R_{{i,{\text{OECD}}}} ,0} \right),$$
which is the change in Canadian restrictions in industry i that would be required to bring Canadian restrictions in line with the OECD average. Where Canada’s FDI regime is already less restrictive than the OECD average, so that Ri,CAN − Ri,OECD < 0, we set the change to zero (ΔRi = 0).
To examine what impact an FDI liberalization ΔRi would have on labor productivity and employment, we turn to Eq. (1) and difference it with respect to FDI to obtain:
$$\Delta \ln \hat{y}_{{i,{\text{Can}}}} = \hat{\beta }\Delta \ln \hat{F}_{{i,{\text{Can}}}} .$$
Further, to obtain an expression for \(\Delta \ln \hat{F}_{{i,{\text{Can}}}}\) , we turn to the equations estimated in columns 3 and 4 of Table 3, which here become:
$$\Delta \ln \hat{F}_{{i,{\text{Can}}}} = - 1.54\Delta R_{i} .$$
Plugging this into the previous equation, we obtain:
$$\Delta \ln \hat{y}_{{i,{\text{Can}}}} = - 1.54\hat{\beta }\Delta R_{i} .$$
(11)
We have choices for our estimate of \(\hat{\beta }\). We begin by focusing on specifications with both industry and country fixed effects. That is, we let \(\hat{\beta } = 0.0670\) when examining changes in labor productivity and we let \(\hat{\beta } = 0.0782\) when examining log changes in employment. These changes in productivity and log employment appear in Table 6. We consider several different experiments.
Table 6

Estimated impacts of levelling down Canadian FDI restrictions to average OECD levels

 

Labor productivity

Employment

Equivalent in increased jobs

Equivalent in increased earnings per year

Percent

Percent

Number

Per worker

Economy-wide ($ millions)

(1) (Percent)

(2) (percent)

(3)

(4)

(5)

All restrictions lowered to OECD levels

0.79

0.93

137,400

$648

$9613

Non-manufacturing restrictions lowered to OECD levels

0.42

0.49

72,559

$342

$5,077

Air transportation restrictions lowered to OECD levels

0.06

0.07

10,620

$50

$743

Banking restrictions lowered to OECD levels

0.04

0.04

6177

$29

$432

Telecom restrictions lowered to OECD levels

0.13

0.15

22,200

$105

$1553

Air, banking and telecom restrictions lowered to OECD levels

0.23

0.26

38,997

$184

$2728

Column 1 reports the impacts on labor productivity. Columns 2–5 report the impacts on labor markets either in terms of jobs (columns 2 and 3) or earnings (4 and 5). Note that labor market gains are expressed either in terms of jobs or earnings; i.e., the two gains should not be summed as this would lead to double counting. Gains are expressed as a permanent percentage increase (columns 1 and 2), as a permanent increase in the number of jobs (column 3), as a permanent increase in average annual earnings per worker (column 4), or as a permanent increase in the aggregate earnings of all workers in the economy (column 5). See the text in “The impact of liberalization of FDI restrictions” for an explanation of these calculations.

All Restrictions Lowered to OECD Levels

Consider first the average impact of liberalizing Canada’s FDI regime. To calculate this, we look at the average change in restrictiveness, ∑ i=1 N ΔRi/N (N is the number of industries), and use this in place of ΔRi in Eq. (11). Consider the first entry in columns 1 and 2 of Table 6. They state that such an FDI liberalization would raise labor productivity by 0.79%.28 It would also raise employment by 0.93%. Since there were 14,834,157 employees in Canada in 2012 (based on the SEPH survey), 0.93% translates into 137,400 jobs. This is reported in column 3. Needless to say, these are very large numbers and support a policy of liberalization.

Economists are often reticent to speak about the job or employment implications of a policy because the people to fill these jobs must come from somewhere and, if these people are not unemployed, then the job-creation numbers are overstated. At the extreme, when these people all move from other jobs then the only effect is that wages are bid up. As a result economists often speak about the equivalent gain in earnings. Earnings gains are reported in columns 4 and 5. To understand the calculations, first note that the current estimate of the elasticity of labor supply is 1.5 (Keane & Rogerson, 2012). Thus, a 0.93% increase in labor demand increases equilibrium labor supply by 0.93% and hence increases earnings by 1.39% (= 0.93 × 1.5). Weekly earnings in Canada averaged $897 in 2012, so that annual earnings averaged $46,644. A 1.39% increase in earnings amounts to a $648 increase in average annual earnings (= $46,644 × 1.39%). This is the number that appears in column 4. Multiplying this by the number of employees in Canada in 2012 yields $9613 million, the number reported in column 5. In short, the labor-market benefits of FDI liberalization amount to either an increase of 137,400 jobs or an increase of $9.6 billion in incomes.29

Non-manufacturing Restrictions Lowered to OECD Levels

In Table 6, we also consider other liberalizations. As discussed, the OECD restrictiveness measures inappropriately score the foreign investment review process in Canadian manufacturing as being highly restrictive. We therefore re-assess the impact of eliminating all restrictions under the assumption that ΔRi = 0 in manufacturing; i.e., under the assumption that Canada’s FDI regime in manufacturing is at least as liberal as the OECD manufacturing average. We then re-calculate the now-lower average change in restrictiveness ∑ i=1 N ΔRi/N = 0.041 and use this in place of ΔRi in Eq. (11). The results appear in the second row of Table 6. The impact of liberalization remains large: it would raise labor productivity by 0.42%. It would also raise employment by 0.49% (72,559 jobs) or, alternatively, raise earnings by $342 per year per worker ($5.1 billion per year for the economy as a whole).

Air Transportation, Banking and Telecom Restrictions Lowered to OECD Levels

We are also able to examine the effect of Canada’s FDI restrictions in three key ‘upstream’ sectors, namely air transportation, banking, and telecom. We first consider each sector separately. Using air transportation as an example, we set ΔRi = 0 in every sector except air transportation, re-calculate the now-even-lower average change in restrictiveness ∑ i=1 N ΔRi/N and use this in place of ΔRi in Eq. (11).30

Such a change would raise productivity by a mere 0.06%, and either create 10,620 jobs or raise annual earnings by $50 per worker. Repeating the exercise for the banking sector, eliminating FDI restrictions in banking would raise productivity by a mere 0.04%, and either create 6177 jobs or raise annual earnings by $29 per worker. Finally, a reduction in FDI restrictions covering Canadian telecom would have much more dramatic effects. For one, it would raise Canadian productivity by 0.13%. For another, it would either create 22,200 jobs or raise annual earnings by $105 per worker for an annual $1.6 billion economy-wide increase in earnings). These are very significant gains from liberalization of a single sector. Finally, liberalization in these three sectors together would raise productivity by a very substantial 0.23% and would either create 38,977 jobs or raise annual earnings by $184 per Canadian worker (which implies an annual increase in economy-wide earnings of $2.7 billion).

All the impacts that we have described so far operate through the own-industry channel. We saw in Table 5 that there may be very substantial downstream- and upstream-industry impacts. However, these calculations are considerably more complex, because we must calculate how changes in FDI restrictions translate not only into changes in ln Fi,Can (see the discussion leading up to Eq. 11), but also changes in Upstreami,Can and Downstreami,Can. This is done by calculating Δln Fi,Can as above and then using Eqs. (3′) and (5′) to calculate ΔUpstreami,Can and ΔDownstreami,Can. Using the coefficients from column 5 of Table 5, we can calculate the impact on labor productivity of eliminating FDI restrictions through the own-industry, downstream-industry, and upstream-industry impacts in air transportation, banking, and telecom impacts on productivity as follows:
$$\Delta \ln (LP_{{i,{\text{Can}}}} ) = 0.155\Delta \ln F_{{i,{\text{Can}}}} + 0.230\Delta {\text{Downstream}}_{{i,{\text{Can}}}} + 0.308\Delta {\text{Upstream}}_{{i,{\text{Can}}}} .$$

Only productivity impacts are reported because there are no statistically significant downstream- or upstream-industry impacts on employment.

The productivity impacts are very large. Removal of FDI restrictions in these sectors would raise productivity by 0.31% in downstream industries (e.g., web designers benefiting from more FDI in telecom) and raise productivity by a further 0.53% in upstream industries (e.g., web designers benefiting from more FDI in retail). The total effect is 1.36%, a dry number for non-specialists, but one that is nevertheless very large. The largest impacts come from eliminating restrictions in telecom. These results are reported in Table 7.
Table 7

Estimated impacts on labor productivity of leveling down Canadian FDI restrictions to average OECD levels (percent)

Own-, downstream- and upstream-industry channels (percent)

Air transportation restrictions lowered to OECD levels

 Own-industry channel

0.14

 Downstream-industry channel (demanding buyers)

0.13

 Upstream-industry channel (sophisticated suppliers)

0.11

 Total

0.38

Banking restrictions lowered to OECD levels

 Own-industry channel

0.08

 Downstream-industry channel (demanding buyers)

0.11

 Upstream-industry channel (sophisticated suppliers)

0.21

 Total

0.40

Telecom restrictions lowered to OECD levels

 Own-industry channel

0.30

 Downstream-industry channel (demanding buyers)

0.08

 Upstream-industry channel (sophisticated suppliers)

0.20

 Total

0.58

Air, banking and telecom restrictions lowered

 Own-industry channel

0.52

 Downstream-industry channel (demanding buyers)

0.31

 Upstream-industry channel (sophisticated suppliers)

0.53

 Total

1.36

See the text for a detailed explanation of the calculations.

CONCLUSIONS

Canada’s economy is above average among OECD countries in terms of its restrictiveness to FDI. These restrictions have reduced the amount of FDI into the Canadian economy. Given the benefits associated with inward FDI, these restrictions have impacted the Canadian economy.

The analysis in this paper has quantified the impact of these restrictions on the Canadian economy, and specifically on Canadian labor productivity, employment and earnings. We identified three potentially beneficial channels, the own-industry, downstream-industry, and upstream-industry channels. The own-industry channel, which captures the effects of increased head-on competition as new foreign players enter Canada, was statistically and economically large. We estimate that a leveling down of Canadian FDI restrictions to average OECD levels would raise Canadian labor productivity by a very large 0.79%. It would also either raise employment by 137,400 jobs or raise annual earnings for each Canadian worker by $648 (or $9.6 billion economy-wide). We also provided evidence of additional downstream- and upstream-industry benefits for Canadian labor productivity. These benefits are associated with the arrival in Canada of demanding foreign buyers and sophisticated foreign suppliers who pressure and support Canadian firms into being the best they can be.

Despite the significant economic benefits that are estimated to flow from the liberalization of FDI policies, there are obstacles to the implementation of these policies. As famously noted by Stigler (1971), protection is often by and for industry. In our context, this means that FDI restrictions are for the benefit of domestic incumbents rather than new innovators or consumers. These ideas are explored in detail by Grossman and Helpman (1994) for trade policy in general, and in Grossman and Helpman (2002, chapter 8) for FDI policy in particular. Empirical validation of some of these political economy impediments to first-best policies appear in Trefler (1993) and Goldberg and Maggi (1999).31

Finally, something must be said about the renegotiated North American Free Trade Agreement, renamed the USMCA. As noted in the literature review above, trade policy impacts trade, FDI, and will also influence the impact that FDI will have on host economies. As such, the renegotiated agreement is an important consideration for the results in this paper, with its ultimate impact on the Canadian economy very much an empirical question. This is a recommended area for future research.

NOTES

  1. 1

    The issue of critical basic business services sectors and the importance of FDI were discussed before the House of Commons Standing Committee on Industry, Science and Technology, May 6, 2010. A transcript is available here: http://www.parl.gc.ca/HousePublications/Publication.aspx?DocId=4505686&Language=E&Mode=1&Parl=40&Ses=3.

     
  2. 2

    MacDonald Dettwiler in 2008, Potash Corporation in 2010, Allstream in 2013, and Aecon Group Inc in 2018. After rejection of the MacDonald Dettwiler in 2008, the Canadian government introduced an explicit consideration for national security in reviewing foreign investments within the Investment Canada Act.

     
  3. 3

    Discussions between the authors and Industry Canada revealed that they do not systematically collect data on preliminary discussions with potential foreign investors, or their representatives, and on potential investments that do not materialize. As a result, there is little information available on investments that are not formally turned down but did not proceed because the investors believed the investment would have been turned down by Industry Canada. Note the name Industry Canada has now been changed to Innovation, Science and Economic Development Canada.

     
  4. 4

    The 18 countries in the sample are dictated by data availability. Furthermore, over 80% of Canada’s inward FDI comes from the sample of OECD countries in the sample, hence providing significant coverage of Canada’s FDI experience.

     
  5. 5

    As reported in Trefler (2004), earlier studies had focused on developing countries, including Tybout et al. (1991) and Pavcnik (2002) on Chile, Levinsohn (1993) on Turkey, Harrison (1994) on Coˆte d’Ivoire, Tybout and Westbrook (1995) on Mexico, and Krishna and Mitra (1998) on India.

     
  6. 6

    Another common approach in this area involves simulating the impact of tariff reductions using a general equilibrium model, such as the Global Trade Analysis Project (GTAP) model. Huff and Hertel (2000) used this model to simulate the impact of tariff reductions between the US, the EU and the Rest of the World. The European Commission and the Government of Canada used the GTAP model to estimate the impact of the Comprehensive Economic Trade Agreement between Canada and the EU, a trade agreement that came into effect in 2017 (Government of Canada and European Commission, 2008). Both authors of the current paper contributed to this Canada-EU report.

     
  7. 7

    Detailed figures highlighting these results are available upon request from the authors.

     
  8. 8

    Management of Companies and Enterprises is an industry classification within the North American Industry Classification System (NAICS) Canada 2012 which includes establishments that are “primarily engaged in managing companies and enterprises and/or holding the securities or financial assets of companies and enterprises, for the purpose of owning a controlling interest in them and/or influencing their management decisions.”

     
  9. 9

    ICT comprises all NAICS codes that relate to information and communication technology industries. This includes portions of the following sectors: 'Manufacturing' (NAICS 31-33), 'Wholesale trade' (NAICS 41), 'Information and cultural industries' (NAICS 51), 'Professional, scientific and technical services' (NAICS 54) and 'Other services (except public administration)' (NAICS 81).

     
  10. 10

    Tables reporting these openness measures are available upon request from the authors.

     
  11. 11

    The share of Canada’s manufacturing industry that was foreign owned has fallen from its high of 56% in 2008 to 47% in 2016. https://www150.statcan.gc.ca/t1/tbl1/en/tv.action?pid=3310003301.

     
  12. 12

    As noted in Lileeva (2010), “Foreign-controlled producers account for over 50% of Canadian manufacturing output, and this ratio has remained stable over the last 25 years. The majority of output is produced by affiliates of U.S. multinationals, who account for over 80% of output produced by foreign-controlled firms in Canada” (583). In other words, the significant openness of Canada’s manufacturing industry is driven by US investment in Canada.

     
  13. 13

    We include country subscripts \(c\) on the \(\omega_{ijc}\) because the upstream-ness of an industry will vary somewhat from country to country. Also, we set \(\omega_{iic} = 0\) since we are only interested here in upstream FDI.

     
  14. 14

    The calculation of the weights \(\omega_{ijc}\) is necessarily complicated as it involves information from the entire general equilibrium structure of the economy. A simple example may be helpful. Consider Figure 3, Panel B. There are three sectors of the economy: Telecom, Web Design, and Retail. The purchasing industries are listed across the top, and along the side are the producing industries. For example, Web Design firms buy $6 of services from Telecom, $2 of services from other web designers, and $4 of services from Retail. They thus buy $12 of services in total and $10 (= $6 + $4) from Telecom and Retail, sectors that are upstream from the perspective of Web Design.

    Relating this back to the \(\omega_{ijc}\) of Eq. (3), let Telecom, Web Design, and Retail be sectors 1, 2, and 3, respectively. Then, we have \(\omega_{21c} = 6/\left( {6 + 4} \right) = 0.6\) and \(\omega_{23c} = 4/\left( {6 + 4} \right) = 0.4\). Note that upstream-ness is always defined from the buyer’s perspective and changes from buyer to buyer. From the perspective of Retail, it buys $5 from Telecom and $8 from Web Design and so has Eq. (3) weights of 5/(5 + 8) for Telecom and 8/(5 + 8) for Web Design.

     
  15. 15

    In building input–output tables, an industry called ‘non-comparable imports’ is created to account for purchases of goods and services that are not produced in the home country, e.g., Canada imports but does not produce tropical fruit. Since there cannot be FDI into the non-comparable imports sector (e.g., foreigners cannot invest in Canada’s tropical fruit sector because it does not exist), we exclude non-comparable imports from the summation in the denominators of Eqs. (4) and (3).

     
  16. 16

    Excluding \(i\)’s purchases from itself is conceptually correct. However, note that, at the level of aggregation of all available input–output tables, it is not entirely correct empirically. For example, most input–output tables lump together autos and auto parts. Thus, when General Motors buys auto parts from Magna, the input–output tables treat this as \(i\)’s purchases from itself even though it is GM purchasing from its upstream supplier, Magna. In our input–output tables, \(a_{iic}\) averages 0.20 across sectors and countries. To investigate further, we have added the regressor \(a_{iic} \ln F_{ic}\) to our productivity and employment regressions. This regressor is often statistically significant. However, it is economically small. That is to say, it does not add any new insights beyond what one obtains from having \(\ln F_{ic}\) as a regressor. We therefore do not report results with \(a_{iic} \ln F_{ic}\).

     
  17. 17

    The issue of non-comparable imports does not arise here; e.g., Web Design cannot sell to the non-comparable imports sector. Using input–output terminology, sales to any foreign entity, regardless of sector, are classified as exports and appear as a final demand category. We are only dealing with sales to producers; i.e., sales of intermediate inputs.

     
  18. 18

    It is important to note that the time period for this analysis overlaps with the 2008 global financial crisis, which is not taken into account. More research is needed to understand what impact that crisis had on FDI patterns.

     
  19. 19

    While the current paper uses FDI stock data, other papers such as Lileeva (2010) have used FDI flow data. We use FDI stock data because they are more readily available for the large sample of countries and industries used.

     
  20. 20

    The OECD has created these FDI Restrictiveness data for the years 1997, 2003, 2006, 2010-2016. http://www.oecd.org/investment/fdiindex.htm.

     
  21. 21

    To see this, note that \(0.937\ln \left( {{\text{GDP}}_{c} } \right) - 0.922\ln \left( {\frac{{{\text{GDP}}_{c} }}{{L_{c} }}} \right) = 0.015\ln \left( {{\text{GDP}}_{c} } \right) + 0.922\ln \left( {L_{c} } \right) \approx \ln \left( {L_{c} } \right)\).

     
  22. 22

    In Eq. (8) there is also a term \(p\lim \sum\nolimits_{i,c} {(\ln F_{ic} ) \varepsilon_{ic} }\). Under the usual assumption that \(F_{ic}\) and \(\varepsilon_{ic}\) are uncorrelated this term vanishes.

     
  23. 23

    Of course, the estimates by country are not reliable. For one, they involve very small sample sizes: if data are available for FDI, productivity/employment for all industries in a particular country, then the sample size would be 48 in each regression. For another, they rely exclusively on cross-industry comparisons within a country, whereas we would prefer to focus on cross-country comparisons within an industry.

     
  24. 24

    In Table 4, we do not report specifications with country fixed effects. The inclusion of both industry and country fixed effects absorbs \(\ln Z_{ic}\) so that \(\ln Z_{ic}\) cannot be an instrument in column 5. The inclusion also results in an unacceptably small Cragg–Donald Wald F statistic (= 1.68) in column 6 (which is not reported).

     
  25. 25

    Lileeva (2010) argues that upstream FDI may be encouraged by the potential of serving downstream industries. The most prominent example of this that we can think of is autoparts FDI, which is upstream to auto assembly so that autoparts FDI is likely influenced by employment and productivity in auto assembly. A similar argument for the endogeneity of downstream FDI can also be made. While estimating an equation with three endogenous variables would strain the credibility of any instrument set, we can investigate possible endogeneity bias as follows. As noted, the industry where such concerns are the greatest is the auto industry because this is the industry where supply chains are most integrated and locations carefully thought out. We therefore re-estimate our model without the auto sector and check if the coefficients on the three FDI variables change. They do not. For example, in column 5 of Table 5, when autos are excluded from the sample, the coefficients (t statistics) on ln(FDIic), downstream, and upstream are, respectively, 0.157 (8.93), 0.241 (7.63), and 0.298 (9.26). That is, they are essentially unchanged. This suggests that the endogeneity of upstream and downstream variables is not a major issue.

     
  26. 26

    These results are not reported, but are available upon request.

     
  27. 27

    The OECD countries used are Australia, Austria, Chile, Czech, Denmark, Estonia, Finland, France, Germany, Greece, Hungary, Ireland, Israel, Italy, Japan, Korea, Netherlands, Norway, Poland, Portugal, Slovakia, Slovenia, Spain, Sweden, Switzerland, Turkey, UK, and USA.

     
  28. 28

    The mean change \(\sum\nolimits_{i}^{N} {\Delta R_{i} /N}\) is 0.077 and \(\hat{\beta } = 0.0670\). Thus, from Eq. (11), we have \(- 1.54 \times 0.067 \times 0.077 = 0.0079\) or 0.79%.

     
  29. 29

    This would generate tax revenues of about $3 billion.

     
  30. 30

    The mean changes \(\sum\nolimits_{i}^{N} {\Delta R_{i} /N}\) are 0.0059 for ‘Air’ restrictions, 0.012 for ‘Telecom’ restrictions, and 0.035 for ‘Banking’ restrictions.

     
  31. 31

    In 2013, as the Federal Government courted entry by foreign telecoms in order to increase competition, the Canadian telecom industry went on a successful public relations offensive to persuade Canadians that this policy would be harmful to Canada. See https://www.cbc.ca/news/business/verizon-pullback-a-blow-to-canadian-wireless-consumers-1.1346672.

     

Notes

ACKNOWLEDGEMENTS

We would like to thank Juan Ma and Umar Boodoo for excellent research assistance, and Andre Downs, Aaron Sydor and many others with the Government of Canada who provided constructive and helpful feedback. We thank an anonymous referee for this basic business services terminology, which replaces the “critical infrastructure sectors” label used previously.

References

  1. Ahn, J.B., Dabla-Norris, E., Duval, R., Hu, B., & Njie, L. 2016. Reassessing the productivity gains from trade liberalization. IMF Working Paper.Google Scholar
  2. Aitken, B.J., & Harrison, A. E. 1999. Do domestic firms benefit from direct foreign investment? Evidence from Venezuela. American Economic Review, 89, 605–618.Google Scholar
  3. Balasubramanyam, V.N., Salisu, M.A., & Sapsford, D. 1996. Foreign direct investment and growth in EP and IS countries. Economic Journal, 106(434): 92–105.Google Scholar
  4. Baldwin, J. R., & Hanel, P. 2000. Multinationals and the Canadian innovation process. Statistics Canada, Analytical Studies Branch, Research Paper 11F0019MPE No.151.Google Scholar
  5. Baldwin, J. R., & Hanel, P. 2003. Innovation and knowledge creation in an open economy: Canadian industry and international implications. Cambridge: Cambridge University Press.Google Scholar
  6. Baldwin, J., & Saborin, D. 2001. Impact of the Adoption of Advanced Information and Communication Technologies on Firm Performance in the Canadian Manufacturing Sector, Statistics Canada working paper.Google Scholar
  7. Barrell, R., & Pain, N. 1997. Foreign direct investment, technological change, and economic growth within Europe. Economic Journal, 107: 1770–1786.Google Scholar
  8. Blalock, G., & Gertler, P. J. 2008. Welfare gains from foreign direct investment through technology transfer to local suppliers. Journal of International Economics, 74: 402–421.Google Scholar
  9. Blalock, G., & Simon, D. H. 2010. Do all firm benefit equally from downstream FDI? The moderating effect of local supplier’s capabilities on productivity gains. Journal of International Business Studies, 40: 1095–1112.Google Scholar
  10. Bhagwati, J. N. 1978. Anatomy and consequences of exchange control regimes, Vol. I, studies in international economic relations, No. I0, New York: National Bureau of Economic Research.Google Scholar
  11. Borensztein, E., Gregorio, J. D., & Lee, J. W. 1998. How does foreign direct investment affect economic growth? Journal of International Economics, 45(1): 115–35.Google Scholar
  12. Büthe, T., & Milner, H. V. 2008. The politics of foreign direct investment into developing countries: Increasing FDI through international trade agreements? American Journal of Political Science 52(4): 741–762.Google Scholar
  13. Chung, W., Mitchell, W., & Yeung, B. 2003. Foreign direct investment and host country productivity: The American automotive component industry in the 1980s. Journal of International Business Studies, 34(2): 199–218.Google Scholar
  14. Coe, D. T., & Helpman, E. 1995. International R&D spillovers. European Economic Review, 39(5): 859–887.Google Scholar
  15. Desai, M. A., Foley, C. F., & Hines, Jr., J. R. 2005. Foreign direct investment and the domestic capital stock. American Economic Review, 95(2): 33–38.Google Scholar
  16. Eaton, J., & Kortum, S. 1999. International technology diffusion: theory and measurement. International Economic Review, 40: 537–570.Google Scholar
  17. Gera, S., Gu, W., & Lee, F. C. 1999. Foreign direct investment and productivity growth: the Canadian host-country experience. Industry Canada Working Paper No. 30.Google Scholar
  18. Goldberg, P. K., & Maggi, G. 1999. Protection for sale: An empirical investigation, American Economic Review, 89(5): 1135–1155.Google Scholar
  19. Government of Canada and European Commission. 2008. Assessing the costs and benefits of a closer EU-Canada economic partnership: A joint study by the European Commission and the Government of Canada. https://www.international.gc.ca/trade-agreements-accords-commerciaux/agr-acc/eu-ue/study-etude.aspx?lang=eng
  20. Grossman, G. M., & Helpman, E. 2002. Interest groups and trade policy. Princeton: Princeton University Press.Google Scholar
  21. Grossman, G. M. & Helpman, E. 1994. Protection for sale. American Economic Review, 84, 833–850.Google Scholar
  22. Harrison, A. E. 1994. Productivity, imperfect competition and trade reform: Theory and evidence. Journal of International Economics, 36(1–2): 53–73.Google Scholar
  23. Hejazi, W., & Pauly, P. 2003. Motivations for FDI and domestic capital formation. Journal of International Business Studies, 34: 282–289.Google Scholar
  24. Hejazi, W., & Safarian, A. E. 1999. Trade, foreign direct investment, and R&D spillovers. Journal of International Business Studies, 30(3): 491–511.Google Scholar
  25. Helpman, E., Melitz, M. J., & Yeaple, S. R. 2004. Export versus FDI with heterogeneous firms. American Economic Review, 94(1): 300–316.Google Scholar
  26. Huff, K., & Hertel, T. 2000. Decomposing welfare changes in the GTAP Model, GTAP technical paper no. 5. Center for Global Trade Analysis, Purdue University, West LafayetteGoogle Scholar
  27. Javorcik, S. B. 2004. Does foreign direct investment increase the productivity of domestic firms? In search of spillovers through backward linkages. American Economic Review, 94(3): 605–627.Google Scholar
  28. Kalinova, B., Palerm, A., & Thomsen, S. 2010. OECD’s FDI restrictiveness Index: 2010 update. OECD Working Papers on International Investment, 2010/03, OECD Publishing, Paris. https://www.oecd-ilibrary.org/docserver/5km91p02zj7g-en.pdf?expires=1553267912&id=id&accname=guest&checksum=CD53DEA90839C5FEDDF26CB7242E1CC4
  29. Keane, M., & Rogerson, R. 2012. Micro and macro labor supply elasticities: A reassessment of conventional wisdom. Journal of Economic Literature, 50(2): 464–476.Google Scholar
  30. Kohpaiboon, A. 2003. Foreign trade regimes and the FDI–growth nexus: a case study of Thailand. The Journal of Development Studies, 40(2): 55–69.Google Scholar
  31. Krishna, P., & Devashish, M. 1998. Trade liberalization, market discipline and productivity growth: New evidence from India. Journal of Development Economics, 56(2): 447–62.Google Scholar
  32. Levinsohn, J. 1993. Testing the imports-as-market-discipline hypothesis. Journal of International Economics, 35(1–2): 1–22.Google Scholar
  33. Lileeva, A. 2010. The benefits to domestically owned plants from inward direct investment: the role of vertical linkages. Canadian Journal of Economics, 43(2): 574–603.Google Scholar
  34. Liu, X., Wang, C., & Wei, Y. 2009. Do local manufacturing firms benefit from transactional linkages with multinational enterprises in China? Journal of International Business Studies, 40: 1113–1130.Google Scholar
  35. Pavcnik, N. 2002. Trade liberalization, exit, and productivity improvement: Evidence from Chilean plants. Review of Economic Studies, 69(1): 245–276.Google Scholar
  36. Puzzello, L. 2012. A proportionality assumption and measurement biases in the factor content of trade. Journal of International Economics, 87(1): 105–111.Google Scholar
  37. Ram C. A., Rao, S., Bhattacharjee, S., & Wright, L. 2010. Outward FDI from Canada and its policy context. Transnational Corporations Review 2(3): 28–43.Google Scholar
  38. Rao, S., & Tang, J. 2005. Foreign ownership and total factor productivity. In L. Eden, & W. Dobson (Eds), Governance, multinationals and growth: 100–121. Northampton, MA: Edward Elgar.Google Scholar
  39. Rodriguez-Clare, A. 1996. Multinationals, linkages, and economic development. American Economic Review, 86(4): 852–873.Google Scholar
  40. Stigler, George J. 1971. The theory of economic regulation. The Bell Journal of Economics and Management Science, 2(1): 3–21Google Scholar
  41. Tang, J., and S. Rao. 2001. R&D propensity and productivity performance of foreign-controlled firms in Canada. Industry Canada Research Publications Program, Working Paper No. 33.Google Scholar
  42. Trefler, D. 2004. The long and short of the Canada-US free trade agreement. American Economic Review, 94(4): 870–895.Google Scholar
  43. Trefler, D. 1999. Does Canada need a productivity budget? Policy Options, July-August: 66–71.Google Scholar
  44. Trefler, D. 1993. Trade liberalization and the theory of endogenous protection. Journal of Political Economy, 101(1): 138–160.Google Scholar
  45. Trefler, D., & Zhu, S. 2010. The structure of factor content predictions. Journal of International Economics, 82(3): 195–207.Google Scholar
  46. Tybout, J. R., de Melo, J., & Corbo, V. 1991. The effects of trade reforms on scale and technical efficiency. Journal of International Economics, 31(3–4): 231–250.Google Scholar
  47. Tybout, J. R., & Westbrook, D. M. 1995. Trade liberalization and the dimensions of efficiency change in mexican manufacturing industries. Journal of International Economics, 39(1–2): 53–78.Google Scholar
  48. van Pottelsberghe De La Potterie, B., & Lichtenberg, F. 2001. Does foreign direct investment transfer technology across borders? Review of Economics and Statistics, 83(3): 490–497.Google Scholar

Copyright information

© Academy of International Business 2019

Authors and Affiliations

  1. 1.Rotman School of ManagementUniversity of TorontoTorontoCanada

Personalised recommendations