The Impact of Emerging Market Competition on Innovation and Business Strategy: Evidence from Canada

Mu-Jeung Yang1 (Eccles School of Business, University of Utah) Nicholas Li (University of Toronto)

Lorenz Kueng (Kellogg School of Management & NBER)

This Version2: August 2019

Abstract How do firms in high-income countries adjust to emerging market competition? We empirically investigate how Canadian firms perform in the wake of China’s entry into the WTO and find evidence of competitive failure risk, which is defined as higher likelihood of bankruptcy in response to competition. Firms with process innovation strategies have higher profits ex-post, but are ex-ante more likely to exit in response to competition. In contrast, firms with product innovation strategies have higher profits if they survive, without significant impact on exit. Both empirical patterns are consistent with a view in which innovation strategies may not insulate firms from competitive shocks. As a consequence, higher profits of surviving innovators can be understood as risk compensation.

Keywords: Chinese competition, innovation, business strategy, commitment, exit, risk

JEL classifications: D2, F1, L2, L6, M2

1 Corresponding author. 1731 Campus Center Drive, Salt Lake City, Utah, 84112. 2 We would like to thank Nick Bloom, John Van Reenen, Pian Shu, and the seminar participants at the NBER PIE Spring Meetings, the University of Toronto, the WEAI Summer meetings, and the Sumantra Ghoshal Conference for helpful comments. We would also like to thank Loren Brandt and Peter Morrow for sharing data on processing trade with us. All remaining errors are our own.

2

1. Introduction What is the impact of trade integration with low-income countries on firm dynamics in high-income countries, including innovation activities and business strategy? A large empirical literature has documented that low-cost competition in the wake of China’s entry into the WTO has led firms in North American and Europe to cut jobs, lose market share, or shut down altogether and has correspondingly highlighted the labor market consequences of Chinese competition; see Bernard, Jensen, and Schott (2006), Iacovone, Rauch and Winters (2013) and Autor, Dorn, and Hanson (2013). Moreover, recent work has emphasized the importance of innovation responses by firms in developed countries, which add an important dynamic dimension to the impact of Chinese competition on profits and therefore net job creation in developed economies. Innovation might add significant dynamic costs for North American trading partners in response to Chinese competition, as in traditional R&D-based models of endogenous innovation3 where competition tends to lower innovation and therefore profits and net job creation. Alternatively, innovation might lead to significant dynamic gains in response to Chinese competition, as in models of quality differentiation such as Sutton (2012) and Amiti and Khandelwal (2013) or trapped factor models such as Bloom, Romer, Terry, and Van Reenen (2014). In this case, this additional innovation allows firms in mature economies to increase long-run profits and create more jobs. Previous empirical studies have found mixed results, with Bloom, Draca, and Van Reenen (2015) documenting positive innovation responses to Chinese competition in Europe, while Autor, Dorn, Hanson, Pisano, and Shu (2016) find that US public firms systematically reduced innovation. Much of this current theoretical and empirical literature has focused on innovation investments and outcomes without considering the business strategy choices of firms. We define business strategy as a long-term plan to pursue specific performance advantages based on novel products, better quality, or lower costs, and we focus on innovation strategies. Such strategies are often costly to reverse, and manufacturing firms such as Intel, General Electric, Nucor, and Ford rarely decide to change innovation investments on a year-to-year basis but rather view such investments as part of a long-term commitment. But such long-term strategic commitments in turn imply the possibility of considerable risk, as firms that are stuck with an irreversible innovation strategy might suffer losses when the business environment changes rapidly, particularly if their innovation attempts fail. In other words, the choice of an innovation strategy and its interaction with rising Chinese competition involves important risk-return tradeoffs that have been ignored in the current literature. In this context, large sample empirical evidence on the performance effects of innovation strategy commitments in response to intensifying competition is sparse. There are at least three challenges that have prevented the empirical literature from making progress in analyzing the performance impact of increased competition on firms with innovation strategy commitments. First, the identification of innovation strategy commitments is demanding as it requires at a minimum panel data on strategic choices over time that are typically not available for a large sample of firms. Second, identifying the costs of strategic commitments and inflexibility requires large competitive

3 Leading examples include Romer (1990), Grossman and Helpman (1991), Aghion and Howitt (1992), Klette and Kortum (2004), and Atkeson and Burstein (2010).

3

shocks as small or gradual changes in the competitive environment would be unlikely to produce large differences in performance for firms with and without these commitments. Third, market selection will lead to a survivorship bias in any analysis focused on firms that performed successfully in the face of Chinese competition.4 Utilizing rich Canadian firm-level panel data on strategy choices, exit and performance from 1999 – 2006, we add value relative to case studies of individual firms that have weathered intensified competition by explicitly taking this survivorship bias into account and interpreting its economic importance in our context. This study seeks to address these three challenges and offers, to our knowledge, the first large-sample empirical analysis of the relative performance effects of firms with innovation strategy commitments in response to intensifying competition. This analysis is made possible by the availability of representative panel data on intended strategic choices of Canadian firms from 1999-2006. To address the first challenge, we follow (Ghemawat, 1991) who states that “a strategy embodies commitment to the extent that, if adopted, it is likely to persist.”. We show that our measures of innovation strategy commitments indeed persist even in the wake of large competitive shocks. Additionally, we provide evidence that this persistence of innovation strategies is not driven by lack of managerial attention to competitive shocks nor by sub-optimal inertia. To address the second empirical challenge, we use China’s entry into the WTO in 2001 as a natural experiment for a large and persistent competitive shock. Previous empirical research has shown how low-cost competition due to China’s WTO entry has strongly intensified competition in the US and Canada leading firms in these regions to lose market share, downsize or shut down altogether; see (Bernard, Jensen, and Schott, 2006) and (Autor, Dorn, and Hanson, 2013) for the US and (Murray, 2017) for Canada. Given this large and salient competitive shock, the persistence of innovation strategies is unlikely to be driven by the lack of high stakes. Addressing the third challenge requires us to pay particular attention to the selection effects of intensifying Chinese competition. To understand this, consider the case in which in the face of a competition shock, a commitment to innovation increases firm profits if innovations are successful but lowers firm profits if innovations fail. Failed innovators with lower profits will therefore exit at higher rates, while surviving innovators will on average exhibit higher profits than firms not pursuing innovation (non-innovators). A simple profit comparison of surviving innovators to non-innovators might then show higher profits for innovators than non-innovators in response to the competitive shock. But if innovation strategies offer higher profits, why wouldn’t all firms pursue innovation strategy committments, see (Barney, 1991)? Our conceptual insight here is that innovation strategy committments imply higher competitive failure risk, defined as higher likelihood of bankruptcy in response to intensifying competition. In our example, competitive failure risk would manifest itself in higher overall exit rates of innovators compared to non-innovators in response to a competition shock. As a consequence, the higher profit of surviving innovators can be understood as risk compensation, which explains the limited imitation of innovation strategy choices in response to competitive pressure. Our analysis is also informative regarding the importance of the specific mechanism through which innovation affects firm performance, see Shapiro (2018). First, innovations – especially in the form

4 See for instance Bowal (2014), who discusses the successful introduction of mass customization of a Canadian furniture manufacturer in response to Chinese competition.

4

of novel and differentiated products – can shield firms from competition, and therefore provide stronger innovation incentives as in Aghion, Harris, Howitt and Vickers (2001). In this context, stronger innovation incentives in response to competition have been called “Arrow Effects”, following Arrow (1962) and Shapiro (2018). Second, innovations – especially in the form of process innovations – can directly increase profitability without insulating firms from their competitors. In this context, many models of innovation, such as Romer (1990), Aghion and Howitt (1992), among others, have the property that more intense competition reduces expected profits from innovation and therefore innovation incentives. This effect of weaked innovation incentives in response to intensifying competition has been called “Schumpeterian Effects” by Homberg and Matray (2018) and Shapiro (2018). This naturally raises the question of how important these two innovation mechanisms are in influencing firm performance. Our empirical results show that firms adopting these different innovation strategies have systematically different performance responses to Chinese competition and that these can be understood in terms of the interaction of innovation incentives and competitive failure risk. Firms that initially pursue process innovation strategies have higher profits conditional on survival, but are also more likely to exit than firms without innovation strategies. Firms that initially pursue product innovation strategies have comparably higher profits conditional on survival, but in contrast are no more likely to exit than firms without innovation strategies. We show that both patterns are consistent with our hypothesis that innovation strategy commitments do not ensure insulation from competitive shocks but instead increase systematic risk. 2. Hypothesis Development In this section we offer a framework, that enables us to interpret our empirical results. Our discussion will follow along the path of four components and the empirical hypotheses that are implied by these components. In the online appendix, we provide a simple model that illustrates some of the following hypotheses. 2.1 Strategic Commitment We start our analysis with the assumption that the choice to pursue an innovation-strategy is irreversible. This assumption is motivated by theoretical considerations such as Ghemawat (1991), who argued that strategic choices fundamentally involve a high degree of commitment and a low degree of reversibility and who offered the following definition of commitment: “a strategy embodies commitment to the extent that, if adopted, it is likely to persist.” These considerations seem to apply very well in the context of the type of innovation strategies we are empirically analyzing here. In our context of manufacturing, innovation strategies by major manufacturing firms seem to offer a variety of examples that require strong long-term commitments that are very costly to reverse. Such examples include GE’s innovation strategy of pursuing connected devices, Ford’s innovation strategy to pursue autonomous vehicles or Genentech’s pursuit of personalized medicine. Indeed, in separate empirical work, Yang, Kueng and Hong, (2015) provide evidence consistent with the view that different strategic choices are

5

related to systematic differences in internal organizational choices and management practices and are therefore costly to change, especially in the short-run. A key implication of the costly reversibility of innovation strategies is that innovation strategy choices could unresponsive even in the face of large competitive shocks. Hypothesis 1: If innovation strategies are costly to reverse, innovation strategy choices will be unresponsive to competitive shocks. Although unresponsiveness of strategic choices to competitive shocks is implied by highly irreversible strategic commitments, such commitment effects are of course not the only possible explanation. Therefore, we consider two additional falsifiable hypotheses here that will help support our view that strategic commitments play a crucial role in our data. The first set of alternative explanations for non-response of strategic choices to competition is related to the failure of managers to perceive competitive shocks. From this perspective, firms that do not correctly perceive the intensity of competition are also unable to correctly formulate best responses. This line of reasoning is broadly consistent with the recent literature on behavioral inattention as discussed in Gabaix (2019), and with recent studies of limited perception of competition, as in Goldfarb and Xiao (2011) and Hortascu, Luco, Puller and Zhu (2019). If the failure to respond to competitive shocks is driven by such inattention to competition, then the competitive shocks should not be salient to managers. Since we instead suspect that non-response to competitive shocks is driven by commitment effects, we formulate the following hypothesis: Hypothesis 1A: Chinese competitive shocks are salient to managers. The second set of alternative explanations recognizes that even if competitive shocks are salient to managers, executives still might fail to respond, but this non-response is not optimal. As a result, non-response or delayed response should be interpreted as lack of flexibility. In fact, firms that switch into the same strategies but after the competitive shock would be expected to perform better than firms that are “stuck” with the same strategic choices. In contrast to this alternative perspective, the logic of strategic commitments implies that non-response can be optimal and that firms that are switching strategies are not necessarily performing better. Hypothesis 1B: Firms that select into innovation strategies after the competitive shock are not necessarily performing better. 2.2 Reversible Innovation Investments

We now focus on different innovation types. In particular, we will assume that innovation investments are reversible and innovations can be separated into product and process innovations. It is plausible that once firms have chosen an innovation strategy, they can increase the probability of a successful innovation by choosing the level of innovation investments. This is a standard assumption in the literature and has been used for example in models by Aghion and Howitt (1992), Kortum (1997), Klette and Kortum (2004), Homberg and Matray (2018) and others. As an

6

immediate consequence, the probability and therefore the number of successful innovations can be responsive to competitive shocks, even if innovation strategy choices are not. But in order to more precisely characterize how the number of successful innovations will respond to competitive shocks, we will need to separate two different types of innovations. In particular, we conjecture that successful product innovations allow firms to escape from product market competition, particularly from low-cost competitors, for example by being able to reduce product substitutability. The effect of product innovations is similar to the shielding effect that frontier firms enjoy in models such as Aghion, Bloom, Blundell, Griffith and Howitt (2005). In contrast, process innovations increase firm efficiency but do not shield firms against competitors. The difference between these two types of innovations then implies different predictions for the response of innovation to competitive shocks. In the following we will call the response of the probability or the number of innovations to a competition shock, the “innovation effect”. In the case of process innovations, the innovation effect will be negative: competition reduces the marginal expected profits of successful innovations, therefore reducing innovation incentives and ultimately the number of successful process innovations. In contrast, for product innovations, the innovation effect should be positive: the fact that successful product innovations can shield firms from competition increases innovation incentives, so the number of successful product innovations should rise in response to competition. Summarizing our discussion yields: Hypothesis 2: If innovation investments are reversible, the number of successful innovations responds to large competitive shocks. The number of process innovations should fall in response to competition, while the number of product innovations should rise or at least fall less in response to competition. 2.3 Innovator Performance in Response to Intensifying Competition We next turn to the consideration of performance effects, conditional on strategic choices. In this context, it is conceptually useful to separate three different types of firms based on their strategy choice and innovation outcomes. In particular, the data can be understood as consisting of firms that do not pursue an innovation strategy (henceforth “non-innovators”) as well as firms with an innovation strategy, which in turn can be separated into successful and failed innovators. Through this lens, one can understand previous empirical studies such as Homberg and Matray (2018) as focusing on the differential impact of competition on successful vs failed innovators. A novel feature of our analysis is idea that the strategic choice to pursue an innovation strategy might involve taking on an additional risk especially in the context of intensifying competition. In particular, failed innovations might reduce firm profits disproportionally in the face of more competition. This additional “competitive failure risk” captures possibly significant additional costs of failed innovations, such as delayed implementation on other projects as well as shutdown costs of projects and has previously been used in theoretical studies such as Atkeson and Burstein (2010). In the context of comparing the performance of innovators to non-innovators, the competitive failure risk and lower profit of failed innovators will generate a selection effect: exit of failed innovators will be higher than exit of non-innovators. At the same time, the failed innovators with the lowest profits will exit, so that profits of surviving failed innovators will be higher than profits

7

of non-innovators. These two key performance moments – exit and profits conditional on survival – are thus critical for measuring competitive failure risk. Additionally, innovation effects will influence the relative performance of innovators compared to non-innovators, but these innovation effects will differ by type of innovation. Since process innovations will exhibit a negative innovation effect, the implied lower number of successful innovations will tend to reduce average profits and increase exit rates. Combining innovation effect and competitive failure risk, we therefore predict that process innovators will have higher exit rates than non-innovators in response to competition, while the profit impact can be ambiguous. Hypothesis 3A: More competition increases exit rates of firms pursuing process-innovation strategies relative to non-innovators, but might have an ambiguous effect on average profits of surviving process-innovators compared to surviving non-innovators. In contrast, since product innovations will exhibit a positive innovation effect, the higher number of successful innovators will tend to raise average profits while reducing exit rates. Combining this innovation effect with the competitive failure risk effect then implies that surviving product innovators are predicted to exhibit higher average profits compared to surviving non-innovators, while the effect on exit rates can be ambiguous. This is important, since one might have thought the combination of product innovators not exiting at higher rates than non-innovators in response to competition, while being systematically more profitable, implies that product innovations fully shield firms against competitive shocks. In contrast, our reasoning suggests that without competitive failure risk, exit rates of product-innovators relative to non-innovators should systematically decline in response to competition. The degree to which exit rates of product innovators do not decline relative to non-innovators is therefore indicative of competitive failure risk. Hypothesis 3B: More competition can have an ambiguous effect on exit rates of product-innovators relative to non-innovators, but will have a positive effect on average profits of surviving product-innovators compared to surviving non-innovators. 3. Data and Methodology

3.1 Data overview Our confidential firm-level data come from Canada’s Workplace and Employment Survey (WES), a random stratified sample conducted by Statistics Canada with the universe of Canadian firms as the sampling frame.5 The survey is stratified by industry, firm size, and region, and we use the population weights provided for all summary statistics and regressions. We use data from the 1999, 2001, 2003, and 2005 waves of the survey.6 The data are a panel with re-sampling to replenish the sample after firm exit or attrition. We restrict our attention to manufacturing firms (NAICS industry codes with 3 as the first digit) since Chinese exports are heavily concentrated in manufacturing, with over 80% of exports in the manufacturing sector during our sample period; see (Autor, Dorn, and Hanson, 2016). This gives us a starting sample of 1,370 firms, of which

5 See \url{http://www23.statcan.gc.ca/imdb/p2SV.pl?Function=getSurvey&SDDS=2615} for the WES questionnaire. 6 The survey was conducted every year from 1999 to 2006. Information about business strategies was asked every other year.

8

about 900 survive until the end of the sample period depending on which outcomes we examine. We apply sampling weights to these firms to make them representative for all manufacturing employer firms – firms with at least one employee – in total more than 300,000 firms. Our consideration of the universe of firms is important, as many previous studies of innovation responses to competition focus on the largest public and private firms, which are typically less likely to exit in response to competition shocks. A unique aspect of the WES data set is that it contains detailed measures of firms’ [ex-ante] intended strategies to deal with competition as well as firms’ [ex-post] outcomes, such as realized innovations and current performance. Table 1 presents summary statistics for our main variables, which we now describe in detail. Note that the sample contains a good mix of small, medium, and large firms. Firms’ business strategies are measured in Section G of the WES. Firms are asked to rate the importance of 15 different strategies on a five-point scale from “Not important'” to “Crucial,'” with strategies ranging from expansion to new markets, new products, quality management, and cost reductions. We focus on three sets of strategies. We are mainly interested in two types of innovation strategies, but we also consider low-cost strategies as controls. Innovation strategies differ by whether they pursue product or process innovations. These strategies are measured as follows. First, the product innovation strategy corresponds to the two factors of “Undertaking research and development” and “Developing new products/services,'” while the process innovation strategy corresponds to “Undertaking research and development” and “Developing new production/operating techniques.” Low-cost strategy corresponds to two different questions: “Reducing labor costs” and “Reducing other operating costs.” An important measurement issue we face is that respondents are asked to assign a numerical value from 1 to 5 to the importance of factors like “improving quality” or “lowering cost,” with higher values reflecting higher strategic importance. These numerical values by themselves seem problematic, especially when comparing responses across respondents. Specifically, it seems that some respondents systematically rate all strategic factors higher on average, considering everything as important, while others rate all factors particularly low. These different reference points make a direct comparison of numerical Likert-scores across respondents – and therefore across firms – potentially problematic. To deal with this issue, we construct two different strategy measures, which capture the essence of the specific research questions we aim to answer. First, for cross-firm comparisons, such as differences in the competitive response as functions of initial strategy, we construct top strategic priorities. These are defined as indicator variables equal to one if the firm considers the factors to be more important than, or at least as important as, any other strategic factors listed. We also require a strategic factor be considered at least “important” (a score of 3) to be considered a strategic priority. This strategic priority variable has the advantage that it extracts information on the relative priorities of the firm. We use it for comparing differential responses to Chinese competition across firms, as a function of these priorities. Second, we construct a more continuous measure to analyze within-firm changes in strategies as a response to the increase in competition. Our rationale is that when we ask “How much did the

9

importance of this strategy change over time?” we ideally want to rely on within-firm variation. However, even when we only use within-firm variation, survey respondents may change over time within the same firm. To deal with this issue, we adjust these measures of strategic change by subtracting the average importance of the relevant strategy relative to other strategic factors. Specifically, we first subtract the mean importance score of all 15 strategy questions for each firm (thereby normalizing the importance of a particular strategy relative to the others within a firm) and then average across the relevant questions for the innovation strategies or low-cost strategy listed above. Table 1, panel A, reveals that innovation strategies were rarely a top strategic priority in Canadian manufacturing in 1999. Section G of the WES also contains several questions measuring perceptions of competition. Firms are asked “To what extent do these firms offer significant competition to your business?” and respond based on a similar five-point importance scale (with “don’t know” as an additional category), with separate items for locally-owned firms, Canadian-owned firms, US-owned firms, and other internationally-owned firms. This allows us to assess whether the increase in Chinese import competition we measure in the data is salient to Canadian firms, which is useful to empirically test hypothesis 1A. As revealed in table 1, panel B, among firms that survived from 1999–2005, the increase in perceived importance of competition from “Other internationally-owned” firms is over three times as large as for US firms (16% vs. 5%). The WES asks detailed questions about innovation outcomes and technology expenditures. Section H asks whether the firm introduced new or improved products during the previous year and whether it introduced new or improved processes. Based on the response, we construct distinct measures of product versus process innovation for each firm by taking the cumulative number of years the firm innovated either an incremental or radical innovation over the period we examine (1999–2005). In constructing our innovation measures, we count incremental as well as radical innovations, since doing so increases the correlation of our innovation measures with performance, as we show below.

[Table 1] Note that the average firm in our data innovates quite frequently based on this variable. Table 1 reveals that for the average firm that survived the seven-year period from 1999 to 2005, there were 2.2 years involving some product innovation and 1.7 years involving some process innovation. There are several reasons why the number of innovations is so high in our data. First, product and process innovation need not correspond to a patent or world-first innovation. The survey explicitly recognizes that an innovation could be a world-first but could also be a Canada first or a local market first, which may simply involve adoption of existing ideas and technologies. Second, firms often pursue product and process innovations together. Although the mean innovation is high, the standard deviation is also high, consistent with a wide variance of innovation outcomes across firms. Section I asks about the firm’s technology use, classified as computers, computer-controlled/computer-assisted technology (e.g. robotics, optical, or laser technology), and other major implementations of technologies or machineries. Table 1 reports the total cumulative expenditures on adopting any of these new technologies over the relevant period, normalized by initial revenue in 1999. The average surviving firm in our data spent around 8.4% of its 1999 revenue on technology adoption over the 1999-2005 period. The WES contains several variables that can be used to assess firm performance. Firms are asked to report their revenues, total employment, gross payroll, and operating profits (defined as

10

revenues minus operating expenses) from the previous year. 7 We use and report these variables in log changes except for profits (due to negative values), for which we calculate the change in operating profits normalized by initial revenues, i.e., the operating profit. The average Canadian manufacturing firm that survives from 1999 to 2005 sees substantial growth of revenue, payroll, and profits over the period (from 15–25% total over a six-year period) but very low employment growth (under 4% over a six-year period). It is worth emphasizing that special care was taken to ensure that our firm exit variable captures either bankruptcy or plant shutdown but not events such as non-response or M&A. In particular, the protocol that analysts at Statistics Canada followed in case of non-response was to first re-contact establishments and in case of persistent non-response to check in administrative tax data whether the firm had declared bankruptcy or the plant had shut down. Only in these circumstances is our variable recording an “exit.”8 3.2 Validating innovation measures Since our study relies on self-reported innovation measures, we first provide corroborating evidence that these potentially noisy measures affect firm performance. Because innovation outcomes are self-reported by firms and claims of novelty are not verified by outside observers such as patent officers, we take two steps to confirm validity. First, we offer additional evidence from a related innovation survey, in which respondents have been directly asked about the economic significance of the self-reported innovation outcomes. In particular, the Survey of Innovation and Business Strategy (SIBS), which is a repeated cross-section with data for 2009 and 2012, asked respondents who reported process or product innovation about the quantitative importance of these innovations for costs and sales. Firms reporting successful process innovations in the last 3 years claim that new or improved processes led to an average unit cost reduction of 7.3%. In contrast, firms reporting successful product innovations in the last 3 years claim that new or improved products account for an average of 5.2% of revenue. Second, while the WES survey does not directly ask respondents about the economic significance of reported innovation outcomes, the panel nature of our data does allow us to validate our innovation measures using other observable outcomes. In particular, the data include information on operating revenue and operating costs, which we cross checked with the corresponding revenue and cost reports of the same firms in administrative tax data. If our self-reported innovation measures are indeed related to real effects on revenues and costs, as claimed by companies responding to the SIBS, then we would expect them to significantly impact reported revenues and costs that we have ensured are consistent with administrative tax data.

[Table 2] Table 2 reports our results of regressing revenue and operating cost growth on our measures of

7 We cross checked the reported revenue and cost data from the WES against balance sheet and cash flow data from the General Index of Financial Indicators (GIFI), which itself is based on corporate tax disclosures. Additionally, we cross checked WES revenue data for all manufacturing firms against reported revenues in the Annual Survey of Manufacturing (ASM). 8 Appendix B provides additional supplemental information on strategic choices and innovation behavior across industries, documenting that all industries have a large degree of within-industry variation in strategic choices and innovation behavior.

11

product and process innovations, for the sample of continuing firms. As these regressions are based on within-firm differences, these results are not driven by the fact that larger firms are more likely to innovate. Consistent with our expectations, we see a significant impact of product innovations on revenue growth. Product innovations also significantly increase operating cost growth, presumably because they lead to an increase in the number of offered products and therefore increased demand for inputs to produce these new products. Process innovations have a significantly negative impact on operating cost growth, as expected.9 3.3 Identification and empirical strategy Our main objective is to estimate the causal effect of increases in Chinese import competition on strategy, innovation, and performance for Canadian firms. We measure the strength of Chinese import competition using the share of Chinese imports over total imports within a 4-digit NAICS industry. Between 1999 and 2005, the average 4-digit NAICS manufacturing sector experienced a rise in Chinese import share from 2.9% to 7.9%, but for some industries the increase was much larger. Figure 1 plots the initial share of Chinese imports in 1999 for each of the 85 4-digit NAICS industries against the subsequent change, revealing a wide dispersion across industries that serves as our main source of identifying variation. For instance, China’s contribution to Canadian imports in 1999 was particularly high in “apparel accessories” and “footwear,” with shares of about 25%. Accordingly, in the six-year period from 1999 to 2005, in which China’s exports increased dramatically, these shares increased by another 13–15%. On the other hand, industries like “dairy product manufacturing” or “printing” had low Chinese import shares in 1999 and experienced only modest increases over the subsequent six years.

[Figure 1] Our estimation strategy is a difference-in-difference strategy: we use cross-industry differences in the change in Chinese import shares to identify the effects of competition on Canadian firms, while at the same time including firm fixed effects. That is, we estimate specifications like equation (1):

𝑦𝑦𝑖𝑖,𝑘𝑘,𝑡𝑡 = 𝛼𝛼𝑡𝑡 + 𝛽𝛽 ⋅ 𝑐𝑐𝑘𝑘,𝑡𝑡 + 𝐷𝐷𝑖𝑖 + 𝜖𝜖𝑖𝑖,𝑘𝑘,𝑡𝑡 (1) where 𝑦𝑦𝑖𝑖,𝑘𝑘,𝑡𝑡 is the firm-level outcome of interest, and 𝑐𝑐𝑘𝑘,𝑡𝑡 is the Chinese import share in industry 𝑘𝑘 at time 𝑡𝑡 and 𝐷𝐷𝑖𝑖 are firm fixed effects. We would not expect the impact of Chinese competition on strategy, innovation and performance to be significant in the short run, and therefore we focus on long-run outcomes, similar to specifications of (Bloom, Draca, and Van Reenen, 2015) and (Autor, Dorn, Hanson, Pisano, and Shu, 2019). Our main specification uses a long-differenced version of equation (1) where we take differences from 1999 to 2005 within each firm for the set of firms that survive throughout the period. For regressions where firm exit is the outcome of interest, we simply use a dummy variable equal to one for firms that exited by 2005 and zero otherwise. One potential concern about estimating equation (1) by OLS is that the changes in Chinese import share that we observe are correlated with industry-level Canadian demand shocks or industry-level Canadian technology/supply-side shocks. For instance, Canadian demand for Chinese textiles

9 Ideally, we would have used unit costs, but these are not reported in the WES, which also lacks information on output quantities.

12

might have increased in this six-year period, which could have led to an increase in China’s import share in this industry. Alternatively, suppose better value-chain management by Canadian firms makes it less costly to off-shore production to China. This better technology makes textiles cheaper and hence increases sales. At the same time, it also makes off-shoring to China more likely and thus increases import shipments of textiles from China to Canada.10 Our main solution for this problem is to use the initial Chinese share of imports in 1999 as an instrument for future Chinese import growth at the industry level, following (Bloom, Draca, and Van Reenen, 2015). The idea behind the IV strategy is that WTO accession and productivity growth in China were plausibly exogenous and unrelated to unobserved domestic Canadian industry-level shocks. At the same time, comparative advantage arguments suggest that a reduction in overall trade barriers towards China should increase Chinese export growth the most in sectors with the highest comparative advantage, see (Bloom, Draca, and Van Reenen, 2015). This initial comparative advantage in turn can be measured by initial industry exports from China to Canada, see (Balassa and Noland, 1965) and (Bloom, Draca, and Van Reenen, 2015). Consequently, the interaction between initial 1999 Chinese exports to Canada and the aggregate trade growth of the Chinese economy should be a valid instrument to predict sectoral Chinese exports to Canada, while at the same time being unrelated to unobserved domestic Canadian industry-level factors. Figure 1, which plots the growth of Chinese import shares against the initial Chinese import share for each NAICS 4-digit industry, shows that this correlation is high, and we generally find F-statistics above 10 in the first stage of our instrumental variable regressions. Our baseline specification (1) allows us to characterize average changes in the outcomes we analyze. To capture moments conditional on strategy, we use the following interaction regression:

𝑦𝑦𝑖𝑖,𝑘𝑘,𝑡𝑡 = 𝛼𝛼𝑡𝑡 + 𝛽𝛽 ⋅ 𝑐𝑐𝑘𝑘,𝑡𝑡 + 𝛾𝛾 ⋅ 𝑐𝑐𝑘𝑘,𝑡𝑡 × 𝑠𝑠𝑖𝑖,𝑡𝑡 + 𝐷𝐷𝑖𝑖 + 𝜖𝜖𝑖𝑖,𝑘𝑘,𝑡𝑡 (2) where 𝑠𝑠𝑖𝑖,𝑡𝑡 is an indicator for an initial strategy, such as product innovation strategy or process innovation strategy. With the dependent variable being either exit 𝑦𝑦 = 𝛿𝛿 or profits 𝑦𝑦 = lnΠ, the interaction coefficient corresponds to the moment

𝛾𝛾 =Δ�𝑦𝑦𝑖𝑖,𝑘𝑘,𝑡𝑡

1 − 𝑦𝑦𝑖𝑖,𝑘𝑘,𝑡𝑡0 �

Δ𝑐𝑐𝑘𝑘,𝑡𝑡

(3)

where 𝑦𝑦𝑖𝑖,𝑘𝑘,𝑡𝑡𝑠𝑠 is the outcome for firms with strategy 𝑠𝑠. It therefore identifies the performance

response to Chinese competition, contrasting firms with an innovation strategy to those without within the same 4-digit industry. We include the initial strategy variables to control for direct impacts of strategic choice on the outcomes and focus on the interaction between strategy and competition. We also include low-cost strategies and initial size as additional controls, on their own and interacted with the trade shocks, to be able to fully account for these factors. Firm size is a central determinant of innovation and exit responses to competition across a range of heterogeneous firm models; while firm size is correlated with strategic choice, we try to isolate the independent effect of strategic choice in these specifications and our results highlight that this is an important additional dimension of firm heterogeneity that affects firm performance in response to trade shocks. An important requirement for (3) to be correctly identified through the specification (2) is that

10 We explicitly analyze potential offshoring effects in section 5.2 below.

13

initial strategy choices are predetermined at the time of the shock. For example, if initial strategic choices anticipate future competitive shocks, this could potentially undermine identification. To ensure that this predetermination condition for strategic choices is met, we take two additional steps. First, we focus on initial strategy choices in 1999, at least two full years before China’s official entry into the WTO at the end of 2001. During 1999, uncertainty about China’s entry into the WTO was high, due to difficulties during negotiations as well as the accidental bombing of the Chinese embassy in Belgrade on May 7, 1999 by the US11. Second, we utilize data on perceptions of international competition, to ensure that the initial strategy choices are not systematically correlated with competitive perceptions. However, it should be emphasized that despite these efforts to minimize the influence of potential endogeneity on our measures of initial strategy, we cannot completely rule out potential selection effects without explicit randomization or quasi-randomization of strategic choices, see (Hamilton and Nickerson, 2003). It is therefore useful to explicitly state how selection concerns will affect our analysis and the interpretation of our results. As mentioned, the main endogeneity issue we are facing is the possibility that some firms forecasted the entry of China into the WTO and therefore selected into product- and process innovation strategies already in 1999. Taking this level of foresight as given, theory would suggest the firms that select into these innovation strategies are likely to either benefit the most from innovation strategies or to be insulated the most from Chinese competition. As a result, firms that self-selected in 1999 into innovation strategies should be less likely to exit and more likely to generate high profits after China’s WTO entry. Econometrically this translates into a downward bias in regression coefficients of exit on Chinese competition and an upward bias of the coefficient of profit on competition.12 As we will discuss in the context of our results, the direction of these biases cannot explain the combination of our performance results and make it therefore unlikely that our performance results are completely driven by this type of strategy self-selection. 4. Results 4.1 Average impact of Chinese competition and strategic response Table 3 presents the average effect of Chinese import competition on performance. Standard errors are clustered by NAICS 4-digit industries throughout, which is the level of variation for our shock. We find a large effect of Chinese competition on firm exit. The coefficient implies that the 5-percentage-point increase in industry-level Chinese import share between 1999 and 2005 for the average firm led to the exit of 4.2% of the firms sampled in 1999 over that period, which relative to the 17% overall exit rate of these firms means that exit increased by around 25%. This large effect of Chinese competition on firm exit is consistent with empirical work on the large employment impact of Chinese competition in Canada. For example, (Murray, 2017) finds that Chinese competition explains around 20% of the manufacturing job losses in Canada from 2001-

11 See: https://en.wikipedia.org/wiki/United_States_bombing_of_the_Chinese_embassy_in_Belgrade 12 To clarify, since the impact of Chinese competition on exit is positve, if strategy self-selected firms are less likely to exit, then the regression coefficient of exit on competition will be less positive and therefore downward biased. In contrast if strategy self-selected firms are to be more insulated from Chinese competition than the regression coefficient of profit on competition should be more positive than without this self-selection, which by definition is an upward bias.

14

2011. On the profit side, IV specifications are consistent with negative effects of rising import competition on the profits of surviving firms, even as the standard errors are too large to reject a zero effect. This is likely due to selection effects: survivors are likely to be the best-performing firms, which would lead to an upward bias that could partially offset the negative effect of increased Chinese import competition. It is important to keep these selection effects in mind as we further analyze the performance consequences of increased Chinese competition.

[Table 3]

The last columns of Table 3 report the impact of Chinese competition on strategy choices, in particular, process and product innovation strategies. Our baseline specification focuses on continuing firms from 1999 to 2005 in the WES. Consistent with our hypothesis of difficulty reversing innovation strategy choices, we do not find any statistically significant impact of Chinese competition on strategy. However, note that the signs of the average effects are consistent with firms on average switching away from process innovation strategies and on average switching towards product innovation strategies.13 4.2 Salience of Chinese competition We continue with an analysis of the impact of Chinese competition on perceptions of competition in Canadian firms. This analysis serves at least two purposes. First, it helps us test our hypothesis 1A regarding the salience of the competitive shock under consideration. Second, the fact that many countries were simultaneously affected by Chinese competition potentially poses a difficult identification issue. For a small open economy like Canada, it is possible that the main competitive effects of China’s WTO entry were not related to direct competitors from China, but were instead the consequence of an indirect effect through US competition. From this perspective, Chinese competition might affect US competition, which could in turn affect Canadian firms.

[Table 4]

To address both concerns, we utilize the unique perception data on competitors, by location. We measure changes in the perception of foreign (non-US) competition by taking the perceived importance of competition from “Other internationally-owned” firms and subtracting the mean importance of competition from all four sources (local, Canadian, US, non-US foreign), which like before normalizes our measure to capture changes in the relative importance of competition from this source within a firm. As table 4 shows, Chinese competition had a strong impact on perceived competition by continuing firms that remained in the sample from 1999 to 2005. These results are also reassuring regarding the absence of significant findings on process- and product innovation strategies in table 3. In both tables, the dependent variables were constructed in a similar manner to extract the relative importance, but only the perception of non-US competition shows a highly significant effect. If indeed the construction of the dependent variables would have been problematic, one would have

13 This is a prediction that is consistent with the formal model we develop in online appendix A.

15

expected to not find any significant results. Furthermore, no significant effect can be shown at any level of confidence for perceived US competition, confirming that our results reflect direct effects from Chinese competitors. 4.3 Benefits of moving later? In this section we follow hypothesis 1B and analyze whether firms that respond to intensifying Chinese competition more slowly in terms of strategic repositioning earned systematically lower profits. The WES data allows us to measure strategic priorities every 2 years, so that we can measure the strategic positioning choices for 1999, 2001, 2003 and 2005. From this panel data on strategic choices we derive strategy variables that measure how early a firm adopted a given strategic position, with higher values capturing earlier adopters of a strategic position. The key interaction variable we derive from this information is how long firms had adopted a specific strategic position. Higher values correspond to longer duration since the firm adopted a specific strategy such as a process innovation strategy. If the ability to switch strategies is crucial, one would expect interaction effects that are negative, as firms that switch into an innovation strategy later will do so because it is beneficial.

[Table 5] In contrast, under the perspective of strategic commitment and irreversible investments, summarized in hypothesis 1B, one would expect no significant difference between early and late adopters of a given strategic position. In table 5 we show interaction effects that are potentially informative for three strategic positions: product innovation strategies, process innovation strategies and low-cost strategies. The results indicate that late moves into product innovation strategies tend to be associated with higher profits in response to Chinese competition, while firms that moved into process innovation strategies tend to have higher profits. However, note that all the standard errors in this specification are very large, so that these results could be driven by random chance. These results are consistent with the view that firms that could change their strategy more flexibly were not necessarily performing better on average. In other words, commitment to a given strategic position was not necessarily suboptimal. 4.4 Impact of Chinese competition on number of innovations We now follow hypothesis 2 and analyze whether the number of product and process innovations was affected by Chinese competition, even if strategic choices were not. Our IV results for process innovation are consistent with hypothesis 2, in that they show a strongly negative response of process innovation. There are very strong and robust negative effects of Chinese import competition on process innovation of surviving firms. The average surviving firm innovated almost 4 times out of a possible 12 innovation counts in the 6 years between 1999 and 2005, but the average effect of Chinese import competition lowers this by about 0.6.

[Table 6] Furthermore, consistent with hypothesis 2, we find a much weaker effect of Chinese import competition on product innovation, where the Chinese import coefficient is not significant in the

16

IV specification and the magnitude of the effect is only about a tenth of the size for product innovation as compared to process innovation. This comparatively weaker response of the number of product innovations in comparison to process innovation is again consistent with hypothesis 2 and highlights the importance of distinguishing between different types of innovation that might be more or less complements or substitutes with product market competition from low-wage countries. 4.5 Initial strategies and selection concerns As described in section 3.3, identifying innovator performance moments in response to a competition shock requires that initial strategy choices be predetermined. Our main concern is that the firms that benefitted the most from an innovation strategy at the time of China’s WTO entry in 2001, were also able to forecast this WTO entry already in 1999. The consequence would be strategy self-selection, which would lead to a downward bias in the exit coefficient and an upward bias in the profit coefficient. To investigate the plausibility of this strategy self-selection, we again utilize the competition perception data we previously used in section 4.2 and table 4. As shown in section 4.2 this data on perceptions of competition by location of competitors did mirror the increase in Chinese competition from international trade data and therefore plausibly captures the salience of competition for firm executives.

[Table 7] Another key advantage of this data is that it is available at the firm level. If it is indeed the case that firms that more strongly believe in increasing Chinese competition did more systematically self-select into innovation strategies, then we should see a strong correlation of perceived non-US international competition and strategy choices. Table 7 shows that there is no significant relationship between perceptions of stronger non-US international competition in 1999, and innovation strategies. 4.6 Innovator performance effects in response to Chinese competition With respect to performance, conditional on strategy, hypotheses 3A and 3B made two unambiguous predictions. First, in response to competitive shocks, exit should increase for process innovators. Second, in response to competitive shocks, profits should increase for product innovators. Table 8 shows that these two unambiguous predictions hold in the data. Overall, the pattern of conditional performance moments is consistent with the predictions of hypotheses 3A and 3B.

[Table 8] But our framework is even more helpful to interpret the remaining coefficients, for which we made ambiguous predictions. To understand why, note that each of these ambiguous predictions was the result of two opposing forces. While the competitive failure risk effect would pull the predictions in one direction, the innovation effect would pull it in the opposite way. We can therefore interpret

17

the net effect on the remaining coefficients as reflecting the relative strength of competitive failure risk vs. innovation effects. Let us start with process innovation strategies and the impact of competition on profits for surviving firms. The theoretical prediction in hypothesis 3A is ambiguous, since the innovation effect would predict that firms reduce process innovation investment in response to competition, which will then reduce the probability of successful process innovations, thereby reducing average profits ex-post. But at the same time, the competitive failure risk effect will tend to increase observed average profits, conditional on survival, as the most unprofitable firms – often failed innovators – will tend to exit. Our empirical finding of higher average profits for surviving firms with initial process innovation strategy therefore tells us that competitive failure risk forces dominate the innovation effects for process innovation strategies. A similar logic applies to product innovation strategies, i.e. these strategies turn out to be risky, even as the impact of Chinese competition on exit rates is statistically insignificant. To see why, remember that the prediction of hypothesis 3B about exit rates in the product innovation strategy case is ambiguous. On the one hand, product innovators are expected to increase innovation to try to shield themselves against competition. If this were the only effect at work, one should observe a decline in exit rates of product innovators, as more innovation investments imply more successful innovations. However, in our empirical results there is no significant difference in exit rates of product innovators and non-innovators in response to Chinese competition. This empirical pattern is indicative of the competitive failure risk effect being at work: failed product innovators are more likely to exit in response to Chinese competition than non-innovators. Therefore, even the insignificant difference in exit rates for product innovators is indicative of the existence of higher bankruptcy risk from innovation commitments to a product innovation strategy. At this point it is also useful to return to our discussion of potential strategy self-section effects mentioned in section 3.3 and section 4.5. Recall that our concern is that firms may choose a strategy in anticipation of greater benefits in the face of future Chinese competition. This should tend to bias interactions of strategy and Chinese competition down for exit and up for profit. But these biases are not able to fully explain the results in table 8. For example, if this bias is at work, then the unbiased exit effect of Chinese competition for process innovators is likely even more positive as exit in response to Chinese competition is more likely for firms that are not self-selected in terms of their strategic choice. In other words, if we could take out the firms that selected into process innovation strategies because they anticipated not facing much bankruptcy risk, then the exit rates of the remaining firms must be even higher. But this would imply that the underlying bankruptcy risk effects are even stronger than indicated by our results. A similar argument can be made for product innovators. We therefore conclude that even if we cannot provide randomization of strategy choices, the biases from strategy self-selection are likely to work against us finding significant competitive failure risk effects. We also note that the signs on the interaction coefficients for low-cost strategies are consistent with the view that low-cost firms consistently underperformed in response to Chinese competition. This is important to note, as process innovations are sometimes argued to mainly reflect cost-saving innovations. Our results highlight that process innovation strategies have their own distinct effects.

18

5. Robustness This section provides additional evidence to better connect our analysis to the literature in at least two respects. First, although our results on the negative impact of Chinese competition on process innovation are consistent with similar findings by Autor, Dorn, Hanson, Pisano, and Shu (2016), Bena and Simintzi (2015), Gong and Xu (2017), and Li and Zhou (2017) for the US, they are somewhat different from the findings of Bloom, Draca, and Van Reenen (2015) for Europe. The question here is whether there is evidence that could help us understand how and why the North American results differ from the European results of Chinese competition. Second, a number of studies, such as Bena and Simintzi (2015) and Branstetter, Chen, Glennon, Yang, and Zolas (2017), have used China’s entry into the WTO as an outsourcing shock, rather than as a direct competitive shock. The outsourcing channel has potentially different implications for firm performance and welfare, which is why separating competitive effects in product markets from outsourcing effects seems important. 5.1 Size effects of innovation One basic question about the negative impact of Chinese competition on process innovation is whether these effects might generally be driven by the contraction of firms. For example, if Chinese competition leads to a general contraction of firms and innovation expenditures are proportional to firm size, then in absolute terms innovation might fall, even as innovation intensity stays constant.

[Table 9] To investigate these size effects, we use revenue as a measure of firm size. A simple approach would be to scale our process innovation measures by revenues. However, since the innovation outcomes are not measured in expenditure terms, we prefer a more general approach. In particular, we flexibly control for a third-order polynomial of revenue growth to allow for a general relation between revenue and process innovations. Table 9 shows that despite this general strategy to control for size effects, Chinese competition has a strongly negative effect on process innovation. 5.2 Outsourcing and processing trade In this section, we analyze whether outsourcing effects are an alternative mechanism that can explain the negative process innovation effects in response to Chinese trade. Outsourcing would have different implications in terms of firm profits and welfare. If firms optimally outsource activities to China, then the fall in process innovation would not capture competitive effects, but would instead reflect the fact that outsourcing is a substitute for process innovations. The implications for firm profits under the outsourcing mechanism would be very different, as outsourcing firms would just replace cost-saving process innovation with cost-saving outsourcing practices. Hence, firm profits might still increase, even under intensive use of outsourcing. To analyze whether our results are indeed driven by outsourcing, we utilize aggregate trade data on processing trade. If outsourcing is indeed a major factor for manufacturing, then one would

19

expect this to be related to trade of intermediates, or processing trade. Before moving to the formal analysis, it is worthwhile to inspect the general time trends in overall Chinese trade to Canada, as well as processing trade, in figure 2.

[Figure 2] As figure 2 shows, overall Chinese exports to Canada accelerated sharply after China’s WTO entry at the end of 2001. The dashed line captures China’s processing trade, which also accelerated, but at a slower pace. In fact, the share of processing trade in China’s exports to Canada has been systematically falling since China’s entry into the WTO. These aggregate trends already foreshadow some of our empirical analysis.

[Table 10] Table 10 summarizes our analysis of the influence of processing trade, as captured by the processing trade share, on process innovation. As the table shows, controlling for processing trade share does not substantially affect the negative impact of Chinese competition on process innovation.

[Table 11] But even though processing trade and outsourcing does not explain the negative impact of Chinese trade on process innovation, it might have an impact on the performance impact of competition, conditional on strategy. To see this, consider low-cost strategies. Although these low-cost firms might be negatively affected by increased Chinese product market competition, they might at the same time benefit from lower costs of outsourcing. This is indeed what table 11 shows. Firms with low-cost and product innovation strategies seem to benefit from outsourcing, as captured by processing trade. On the other hand, our main result of increased exit and better performance conditional on survival for firms with process innovation strategies is robust to controlling for processing trade. 5.3 Domestic competition and product homogeneity In this section, we provide evidence that might explain why Chinese competition could have negative effects on innovation in the US, as shown by Autor, Dorn, Hanson, Pisano, and Shu (2016), while also leading to positive effects on innovation in Europe, as shown by Bloom, Draca, and Van Reenen (2015). The basic idea follows the insight of Aghion, Bloom, Blundell, Griffith, and Howitt (2005) that the impact of competition on innovation might follow an inverted U relationship. If the initial level of competition in Europe was relatively low, then one would expect increased Chinese competition to lead to increased innovation. In contrast, the initial level of competition in the US and Canada might be considered relatively high, so that a further competitive shock from China leads to a negative effect on innovation. A full cross-country investigation of this hypothesis is beyond the scope of this paper. However, we can analyze the implications of the inverted U hypothesis within Canada. Specifically, if correct, the inverted U hypothesis would predict that firms or sectors with more intense initial

20

competition should exhibit significantly more negative effects of Chinese competition on innovation. We investigate this hypothesis using three different measures of competition. First, we use the previously discussed measure of perceived competition. This has the advantage of being available at the firm level and therefore allows for variation within sectors. Second, we use a Herfindahl index of industry concentration, where low values reflect greater competition. Third, we use measures of product homogeneity, based on elasticities of substitution from Broda and Weinstein (2006). High homogeneity (high elasticity) may be related to greater competition.

[Table 12] Table 12 shows that all three of these measures of competition are potentially consistent with an inverted U hypothesis. The table shows that more initially competitive industries and firms that perceived higher initial levels of competition indeed exhibited more negative process innovation effects in response to Chinese competition. This evidence is only suggestive, as most of these coefficients are not statistically significant from each other. Since our results on product homogeneity might suggest that the impact of Chinese competition was stronger in low-product-homogeneity sectors, we also add results for firm exit for our product homogeneity sample splits. These show that firms in high-homogeneity sectors were more likely to exit in response to Chinese competition, suggesting that high initial product homogeneity resulted in a stronger competitive impact of Chinese competition.

[Table 13] Table 13 documents the results of sample splits according to product homogeneity for the performance effects of Chinese competition, conditional on initial strategy. We focus on the product homogeneity results due to space constraints, but results with perceived competition and Herfindahl indexes are similar. The table shows that our baseline results of the effects of Chinese competition on innovation strategy firms are driven by sectors with high product homogeneity. 5.4 Long-run analysis of innovation strategy choices Here we provide additional evidence on strategic choices of firms in the long run, beyond 2005, which is the end of our WES panel. For convenience, we start by reminding the reader of our baseline results, which are reported in the first two columns of table 14. In general, we do not find evidence of any statistically significant impact of Chinese competition on strategy for surviving firms. While the signs of the estimates are both consistent with our baseline theory, it seems that firms’ strategic responses to Chinese competition were too heterogeneous to lead to a common pattern.

[Table 14] Even as we do not find any effects on strategic priorities for continuing firms from 1999 to 2005, there are two ways in which this might understate possible strategic effects. First, even if business strategies are hard to change in the short to medium run, given enough time the same firms might eventually change their strategies. Second, even if incumbent firms do not adjust their strategies, it is possible that new entrants choose optimal strategies when they entter. Together with selective

21

exit of firms with sub-optimal innovation strategies, this could lead to a change in strategic orientation within an industry without much of a change within continuing firms. To allow for both of these channels and to maximize our chances of detecting a strategic response, we move to a sector-level analysis, which allows us to add more data from the SIBS dataset on strategic behavior. Since the SIBS does not ask firms for all the detailed strategic dimensions of the WES, we had to adjust the WES strategy measures to focus on whether product or process innovation were considered important. With this industry-level variation in the strategy measure over the 2001–2012 period, we estimated whether the number of firms that considered process or product innovation important for their strategy was affected by Chinese competition.14 Columns 3 and 4 of table C document the results. Process and product innovation strategies both appear to have declined, with process innovation strategy falling almost twice as much with a decline that is statistically significant only at the 10% level. Overall, our results are consistent with strategic orientation changing through selection, including entrants adjusting their strategy, while the strategic choices of incumbent firms continue to persist, consistent with Ghemawat’s definition of strategic commitment. 6. Conclusion To our knowledge this is the first large sample multi-industry study that finds evidence for the effects of innovation strategy commitments. The limitations of this study provide several promising avenues for further research. First, while we find that the unresponsiveness of innovation strategies cannot be explained by a failure to perceive competition, we did not offer any direct evidence for the factors that determine the degree of commitment to specific innovation strategies. Furthermore, different types of innovation strategies might involve different types of adjustment costs, where horizontal differentiation involves different adjustment costs from vertical differentiation. More detailed direct evidence on the nature of such adjustment costs would further our understanding of the degree of strategic commitment for different types of innovation strategies. Second, our performance results indicate that superior performance of innovators in response to competitive shocks can be interpreted as compensation for increased competitive failure risk. It is worth re-emphasizing that the underlying shock – in the form of a competition shock – is common to all firms in an industry, but that firms can differ in terms of their exposure to this common or systematic risk. This mechanism is reminiscient of basic theories of finance, such as CAPM, where common or systematic risk in the stock market is the basis of risk compensation and individual firms can have different risk exposures, or different “betas”. Our results suggest that the “beta” for “competitive failure risk” is significantly higher for firms with innovation strategies. Our study therefore provides a natural link between strategic management and finance. For example, one question is whether this type of risk compensation might apply more broadly to other types of strategic commitments.

14 For this long-run specification, we used three non-overlapping time differences from 2001–2005, 2005–2009, and 2009–2012 and estimated effects with industry and year fixed effects.

22

References Aghion, P. and P. Howitt, 1992. A Model of Growth through Creative Destruction, Econometrica Aghion, P., Bloom, N., Blundell, R., Griffith, R. and P. Howitt, 2005. Competition and Innovation: An Inverted- U Relationship. Quarterly Journal of Economics Aghion, P., Harris, C., Howitt, P. and J. Vickers, 2001. Competition, Imitation and Growth with Step-by-Step Innovation, Review of Economic Studies Amiti, M. and A. Khandelwal 2013. Import Competition and Quality Upgrading, Review of Economics and Statistics Arrow, K., 1962. Economic Welfare and the Allocation of Resources to Invention. In: The Rate and Direction of Inventive Activity: Economic and Social Factors, edited by the Universities-National Bureau Committee for Economic Research and the Committee on Economic Growth of the Social Science Research Councils. Princeton, NJ: Princeton University Press Atkeson, A. and A. Burstein, 2010. Innovation, Firm Dynamics, and International Trade. Journal of Political Economy Autor, D., Dorn, D. and G. Hanson, 2013. The China Syndrome: Local Labor Market Effects of Import Competition in the United States. American Economic Review Autor, D., Dorn, D., Hanson, G., Pisano, G., and P. Shu, 2019. Foreign Competition and Domestic Innovation: Evidence from US Patents. American Economic Review: Insights. Balassa, B. 1965. Trade Liberalisation and "Revealed" Comparative Advantage, The Manchester School. Barney, J., 1991. Firm resources and sustained competitive advantage. Journal of Management Bena, J. and E. Simintzi, 2015. Labor-Induced Technological Change: Evidence from Doing Business in China, Working Paper Bernard, A., J. Jensen, and P. Schott, 2006. Survival of the Best Fit: Exposure to Low-Wage Countries and the (Uneven) Growth of US Manufacturing Plants, Journal of International Economics Bloom, N., Romer, P., Terry, S. and J. Van Reenen, Trapped Factors and China’s Impact on Global Growth, Working Paper, 2014. Bloom, N., Draca, M. and J. Van Reenen, 2015. Trade-Induced Technical Change? The Impact of Chinese Imports on Innovation, IT and Productivity. Review of Economic Studies

23

Bowal, P., 2014. In a sea of cheap imports, how can a high-end furniture store stay afloat? The Globe and Mail Branstetter, L., Chen, J., Glennon, B., Yang, C. and N. Zolas 2017. Does Offshoring Manufacturing Harm Innovation in the Home Country? Evidence from Taiwan and China, mimeo Carnegie Mellon University Gabaix, X. 2019. Behavioral Inattention. In: Handbook of Behavioral Economics, Vol 2, edited by Bernheim, D., DellaVigna, S. and D. Laibson, 2019 Ghemawat, P., 1991. Commitment: The dynamic of strategy. New York: The Free Press Goldfarb, A. and M. Xiao (2011). Who Thinks about the Competition? Managerial Ability and Strategic Entry in US Local Telephone Markets, American Economic Review Grossman, G. and E. Helpman, 1991. Quality Ladders and Product Cycles, Quarterly Journal of Economics Hombert, J. and A. Matray, 2018. Can Innovation Help US Manufacturing Firms Escape Import Competition from China?. Journal of Finance Hortascu, A., Luco, F., Puller, S. and D. Zhu (2019). Does Strategic Ability Affect Efficiency? Evidence from Electricity Markets, Working Paper, University of Chicago Klette, T. and S. Kortum, 2004. Innovating Firms and Aggregate Innovation, Journal of Political Economy Kortum, S. 1997. Research, Patenting and Technological Change, Econometrica Murray, A., 2017. The effect of import competition on employment in Canada: evidence from the “China Shock”, Working paper, Center for the Study of Living Standards Romer, P., 1990. Endogenous Technological Change, Journal of Political Economy Schumpeter, J., 1942. Capitalism, socialism and democracy. Harper and Brothers. Shapiro, C. 2012. Competition and Innovation: Did Arrow Hit the Bull's Eye? in: Lerner, J and S. Stern, 2012. The Rate and Direction of Inventive Activity Revisited, NBER. Sutton, J. 2012. Competing in Capabilities: The Globalization Process, Oxford University Press Yang, MJ, Kueng, L., and B. Hong, 2015. Business Strategy and the Management of Firms. NBER Working Paper

24

25

Figure 1: Graphical summary of first stage of instrumental variable for Chinese competition.

26

27

28

29

Figure 2: Time series of total Chinese exports to Canada, processing exports and the share of processing trade from 2000 to 2006.

30

31

32