Who Benefits from the Decline of American Manufacturing? Evidence from 142,663 Foreign and Domestic Entries in China 1

Transcription

1 Who Benefits from the Decline of American Manufacturing? Evidence from 142,663 Foreign and Domestic Entries in China 1 March 12, 2018 Minwen Li School of Economics and Management, Tsinghua University Tanakorn Makaew U.S. Securities and Exchange Commission Vojislav Maksimovic Robert H. Smith School of Business, University of Maryland Abstract Using the establishment of U.S.-China Permanent Normal Trade Relation as a plausibly exogenous shock we study the effect of trade liberalization on an emerging market. We find almost half of new exporters in the Chinese Census are foreign owned. The rate of foreign entrants is more responsive to trade liberalization. This differential response is concentrated in financially underdeveloped localities. In these areas, foreign entrants are better capitalized, more liquid, and pay lower taxes. They tend to survive and grow faster than domestic entrants. Our results suggest that in underdeveloped markets, trade globalization may benefit foreign investors rather than domestic firms and entrepreneurs, who are financially constrained. JEL Classification: F13, F23, F61, G15, G32 Key Words: Globalization, International Trade, Export, Financial Development, Financial Constraint 1 Minwen Li can be reached at liminwen@sem.tsinghua.edu.cn. Tanakorn Makaew can be reached at tanakornm@hotmail.com. Vojislav Maksimovic can be reached at vmaksimovic@rhsmith.umd.edu. The Securities and Exchange Commission disclaims responsibility for any private publication or statement of any SEC employee or Commissioner. This article expresses the authors views and does not necessarily reflect those of the Commission, the Commissioners, or members of the staff. 1

2 1. Introduction Prior literature documents many serious consequences of international trade for the manufacturing sector in the United States. Import competition from China, in particular, is blamed for various problems in American manufacturing sector, including plant closures (Bernard, Jensen, and Schott, 2006), unemployment (Pierce and Schott, 2016), reduced innovative activities (Autor, Dorn, Hanson, Pisano, and Shu, 2016), and reduced sales growth and profitability (Hombert and Matray, 2017). Recent work shows that the economic impacts of Chinese import competition are also highly asymmetric many of the negative effects fall on working-class Americans (e.g., Pierce and Schott, 2015) and entrepreneurial firms (Ayyagari and Maksimovic, 2017). The question of who benefits at the expense of American workers and small business owners has drawn considerable attention from the public. 2 Despite detailed evidence on the downside of free trade in the United States, very little is known about the effects of free trade in China at the micro level. In this paper, we study exporting activities of new plants in China and identify which types of firms benefit the most from trade globalization. Overall, our results indicate that in underdeveloped financial markets, such as China s, trade liberalization may benefit foreign investors more than domestic firms, which lack financial resources to support exporting activities. Using an establishment-level dataset with 142,663 manufacturing entries and 670,049 plant-year observations from the Chinese National Bureau of Statistics (Chinese Census), our major findings are as follows: First, foreign-owned plants are prevalent in Chinese manufacturing, especially in the exporting sector. Over 40% of new exporters in the Chinese Census are foreign-owned. A majority (60%) of new foreign plants are exporters. 2 Consistent with popular view, Steve Bannon, former White House Chief Strategist, stated The globalists gutted the American working class and created a middle class in Asia [Politico Magazine, August 18, 2017]. 2

3 Second, we use the formation of U.S.-China Permanent Normal Trade Relation (PNTR), which is a change in U.S. trade policy that eliminated potential tariff increases on Chinese imports, as a plausibly exogenous trade shock. 3 We find that entry rate and export propensity of new foreign plants are more responsive to trade liberalization than those of domestic plants. Foreign plants with controlling foreign stake are more responsive to the trade shock than plants with minority foreign stake. Third, export propensity is more responsive to the trade shock in financially developed areas. We base our city-level measure of financial development on the financial market access score from the World Bank Enterprise Investment Climate Survey. This finding is consistent with the notion that establishments in financially developed areas have greater ability to raise external capital to finance costly export activities since overseas sales tend to require more financial resources than domestic sales. 4 Fourth, the differential response to the trade shock between developed and undeveloped areas is more pronounced among domestic plants. A potential explanation is that domestic entrants rely more on local financial markets. In financially developed areas, domestic plants are more likely to have access to external funds to finance export. In underdeveloped markets, domestic plants may not be able to raise the necessary amount of external capital so they are at a disadvantage relative to foreign plants. Fifth, foreign entrants are larger, better capitalized, less taxed, and more liquid than domestic entrants. These results confirm that foreign plants are less financially constrained than domestic plants. Foreign plants are also more likely to survive. Among plants that survive, 3 A number of papers use this shock to study the effects of free trade in the U.S. For example, see Pierce and Schott (2015, 2016). 4 Beck (2002, 2003), Zia (2008), and Minetti and Zhu (2011) show that financial constraints play an important role in determining the level of foreign sales. 3

4 foreign plants grow faster than domestic plants. These foreign-domestic differences are greater in underdeveloped areas, in accord with prior results. We also find that large plants are more responsive to the trade shock than small plants and that the differential responses between large and small plants are more pronounced in underdeveloped areas. These results are consistent with the notion that smaller plants, especially the ones in underdeveloped areas, are subject to financial constraint related to export. Double sorting by size and ownership verifies that the effects of size and foreign ownership are not confounded. In addition to financial development, we examine two other dimensions of city-level characteristics: government bureaucracy, proxied by the average number of days it takes to clear customs, and the general level of economic development, proxied by an infrastructure development index, both from the World Bank Enterprise Investment Climate Survey. Similar to financial development, plants in areas with less bureaucracy and better infrastructure are more responsive to the trade shocks. However, the inter-city differences in shock responses do not differ between foreign and domestic plants, suggesting that the competitive advantage of foreign plants does not arise from superior abilities to overcome bureaucratic or infrastructure obstacles. We find some evidence of congestion: conditional on their own shocks, plants in areas exposed to positive trade shocks are less likely to export. The congestion effect is more pronounced among domestic plants, indicating that the trade shock not only benefits domestic plants less than foreign plants, but may also hurt domestic plants due to congestion. The foreigndomestic difference in congestion effect is more pronounced in underdeveloped areas. These results are in accord with the view that domestic firms are more dependent on local resources and more sensitive to local economic conditions. This paper contributes to several strands of literature. A number of papers examine the effects of import competition in the U.S. On the financial side, import competition is shown to affect cash policy (Fresard, 2010), investment policy (Fresard and Valta, 2016), and household 4

5 leverage (Barrot, Loualiche, Plosser, and Saivagnat, 2017). On the production side, import competition is shown to increase unemployment and plant closures in manufacturing industries (Bernard, Jensen, and Schott, 2006; Pierce and Schott, 2016). The rapid decline of manufacturing industries also leads to unfavorable social outcomes ranging from criminal activities (Feler and Senses, 2015) to political polarization (Autor, Dorn, Hanson, and Majlesi, 2016). While these papers focus on the consequences of trade shocks in the U.S. or other developed countries 5, we study the behavior of plants in China where the import competition shock originates. Our work is also closely related to the papers on corporate inequality. In recent years, the profits, productivity, and pay gaps between top firms and other firms have been widening (Song et al., 2015; Bloom, 2017). Given that prior papers establish that free trade is specifically detrimental to entrepreneurs in the U.S. (Ayyagari and Maksimovic, 2017), one may conjecture that their counterparts in China would benefit more from free trade corporate inequality in China might be reduced if the trade shock creates new opportunities for small Chinese firms. Our paper shows that this conjecture is not true. In China, foreign plants, large plants, and plants in financially developed areas have greater ability to respond to the trade shock. China s underdeveloped financial markets inhibit their small domestic plants from taking advantage of export opportunities. Free trade seems to worsen corporate inequality problems both in the U.S. and in China. The financial development literature shows that well-functioning financial markets can alleviate agency and asymmetric information problems and promote economic growth (e.g., Demirguc-Kunt and Maksimovic, 1998; Rajan and Zingales, 1998; Beck, Demirguc-Kunt, and Maksimovic, 2005, 2008; Ayyagari, Demirguc-Kunt, and Maksimovic, 2010). Our paper shows 5 Bloom, Draca, and Van Reenen (2016) examine the effects of Chinese import shock on R&D in twelve European countries. 5

6 that financial development also affects how the gains from free trade are distributed across firms, thereby linking the financial development literature with the literature on international trade. 6 Given our focus on new plants, we also contribute to the literature on entrepreneurship and new firm creation. Prior papers document the importance of financing on entrepreneurial activities in the U.S. For example, Hellmann and Puri (2000) show that venture capital financing is related to product market strategies and outcomes of new firms. Babina, Ouimet, and Zarutskie (2017) find that wealth gains from successful IPOs allow employees to depart to start-ups. Our paper shows that regional financial development affects entry and export decisions of new plants, confirming the importance of financing. By examining the roles of financial development, government bureaucracy, and infrastructure development in China, our paper joins the growing literature that identifies institutional factors driving entrepreneurship in international context, such as culture (Guiso, Sapienza, and Zingales, 2006) and entry regulation (Klapper, Laeven, and Rajan, 2006). The rest of the paper is organized as follows: Section 2 documents the prevalence of foreign plants in China. Section 3 describes our data sources and the construction of variables from the matched Census-tariff dataset and presents the summary statistics. Section 4 discusses the difference-in-difference methodology and presents the effects of trade shock on entry rate and export propensity. Section 5 examines the types of plants that are more responsive to the trade shock. Section 6 examines the role of financing in export entry. Sections 7 and 8 compare the financial conditions and performance of foreign and domestic entrants. Section 9 provides additional findings on size, other measures of development, and the local general equilibrium effects, and Section 10 concludes. 2. Foreign Plants in China 6 See Bernard, Jensen, Redding, and Schott (2007) for a survey on the roles of firms in international trade. 6

7 China adopted the open-door policy and started its market-oriented economic transformation in Since then, Mainland China is among the top destinations for foreign direct investment. The exporting sector has also become a crucial contributor to the Chinese economy. 8 Using data from the Chinese Census, we assess the role of foreign establishments in the economy and the exporting sector in our sample. Table 1 presents the results. Insert Table 1 Here Table 1 shows that foreign plants are prevalent in manufacturing industries. The census panel contains 670,049 plant-year observations and 142,663 entries. Around 19.57% of these manufacturing plants and 16.5% of entrants receive investment from foreign countries or Hong Kong, Macau, and Taiwan. 9 Compared to the U.S., 10 export sector plays a more prominent role in China: 25.76% of manufacturing plants and 21.36% of manufacturing entrants export. Foreign presence in the export sector is especially strong. Majority of foreign plants (60%) in our sample export. Overall, 45% of manufacturing exporters receive foreign investment. When we turn to manufacturing entries, foreign presence is similarly crucial: 41% of entrants to the export sector are foreign in our sample. 7 The open-door policy is an economic policy announced by Xiaoping Deng in 1978 that welcomed foreign businesses. Special Economic Zones (SEZ) were set up with favorable tax and regulatory terms to attract capital and business from overseas and promote exports. It was a turning point when China s economic policy shifted to encouraging foreign investment and international trade. 8 In 2016, Mainland China received around 1.46 trillion dollars of foreign direct investment, ranked the third worldwide (only after the U.S. and the U.K.). Around 19.5% of 2016 GDP can be attributed to export. Information on FDI inflows, GDP, and export in China and FDI inflows of other countries are from the World Fact Book by Central Intelligent Agency (2016). 9 Since Hong Kong, Macau, and Taiwan adopt different political and economic systems than Mainland China, we include investment from these regions in foreign investment unless notified otherwise. In Appendix Table 6, we show that our main results hold if we exclude investment from Hong Kong, Macau, and Taiwan. 10 Bernard, Jensen, Redding, and Schott (2007) document that the overall share of U.S. manufacturing firms that export is 18%. 7

8 Prior literature shows that export requires more resources than domestic sales (e.g., Beck 2002, 2003; Zia, 2008; Minetti and Zhu, 2011), which might give foreign plants more competitive advantages in a less developed economy like China. To begin with, foreign plants can utilize technologies, brands, or know-hows of the parent companies (Fresard, Hege, and Phillips, 2017). They can also utilize distribution channels in parent countries. Foreign plants may also have tax or regulatory advantages due to China s open-door policy. More importantly, foreign plants can access the global financial markets and the financial resources of the parent companies, while local plants are more restrained to the domestic financial market which is underdeveloped and largely segmented (See, for example, Qian, Strahan, and Yang, 2015; Allen, Qian, Shan, and Zhu, 2017). 3. Data and Descriptive Statistics 3.1 The Sample Our sample consists of establishment-level manufacturing entries in the Chinese Census over the time period of 1999 to Plant-level data on entry, location, financial and ownership information, and exporting activity are from Chinese Census, which is Chinese National Bureau of Statistics (NBS) s annual survey of large and medium manufacturing establishments. This survey covers all Chinese industrial establishments with annual sales over 5 million RMBs during the period of 1998 to Recent papers (e.g., Brandt et al., 2012; Zhang et al., 2017) use this data to examine macroeconomic issues such as total factor productivities. In this paper, we choose our sample period around the formation of US-China PNTR, so that the sample spans two years before (i.e., year ) and three years after this trade shock (i.e., year ) is the first year that entry data can be derived from Chinese Census based on our definition of entry. Our sample ends in 2003 because NBS substantially increased the coverage of the survey in 2004, leading to a discontinuity in data collection practice. In Appendix Table 4, we show that our main results are robust to the extension of sample period to

9 City-level development data on financial market access, customs bureaucracy, and infrastructure development are constructed from World Bank Enterprise Investment Climate Survey (2004). Industry-level tariff data are from Pierce and Schott (2015). 3.2 Variable Construction Below we describe the construction of our main variables. Detailed descriptions of our variables are provided in Appendix Table 1. Our paper focuses on entrants in manufacturing industries. Based on the panel of plant-year observations in Chinese Census, we identify entry as a plant that did not exist in the previous year but enters in the current year. 12 We compute entry rate as the ratio of number of new entrants divided by total observations. A key dependent variable in this study is Export Indicator, which equals one if a plant exports in a specific year, and zero otherwise. We thus examine the propensity of export entry by computing the percentage of entrants that export at the year of entry. To investigate the financial conditions of the entrants, we use total assets to proxy for size, total liabilities divided by total assets to proxy for capital structure, corporate income taxes divided by net sales to proxy for tax liability, and cash divided by total assets to proxy for liquidity. To study the performance of entrants, we first compute one-year, two-year, and three-year survival rates, which are defined as the percentage of plants that exist throughout the time windows of one year, two years, and three years, respectively, after the entry. We then compute the growth rate of total assets for surviving plants during these horizons. Finally, as an all-inclusive performance measure, One-Year Performance is defined as the one-year growth rate of total assets for surviving plants, and -100% for plants that do not survive one year after the entry. Two-Year Performance and Three-Year 12 Alternatively, we identify entry as a plant with less than or equal to two years of age based on its founding date, and derive similar results. We do not base our main analysis on this alternative definition due to the large amount of missing information on founding date. 9

10 Performance are defined accordingly based on the time windows of two years and three years after the entry, respectively. The following serve as the explanatory variables in our study. (1) Variables related to the trade shock: We obtain NTR Gap from Pierce and Schott (2015). NTR Gap is defined as the gap between NTR tariff and non-ntr tariff for a specific industry (four-digit SIC). We then construct another industry-level indicator variable, High NTR Gap, based on NTR Gap. We first compute the median NTR Gap in each industry group (twodigit SIC). High NTR Gap equals one (zero) if NTR Gap of a specific industry is higher (lower) than the group median. By doing so, we ensure that there are industries with relatively high and low tariff gap within each industry group, effectively matching the treatment subsample (High NTR Gap=1) and the control subsample (High NTR Gap=0) within the same industry group. Given our definition, High NTR Gap is set to missing for industry groups with zero variations in NTR gap, for example, tobacco products and petroleum and coal products. Since the U.S. granted China PNTR status at the end of 2000, we define PstPNTR as an indicator variable that is equal to one for observations from or after 2001, and zero otherwise (Pierce and Schott, 2015). To conduct the difference-in-difference analysis, we construct the interaction term between PstPNTR and High NTR Gap (PstPNTR*High NTR Gap). (2) Ownership types: We examine the following ownership types based on the registration-type information of manufacturing plants in Chinese Census. First, foreign (domestic) plants refer to those with (without) investment from foreign countries or Hong Kong, Macau, and Taiwan. 13 Second, among domestic plants, we study two major ownership types: state-owned and privately-owned. State (privately)-owned plants refer to those whose controlling shareholder is state (private entities or individuals). We note that a small number of domestic plants also include 13 According to Bilateral FDI Statistics from United Nations Conference on Trade and Development, after the establishment of PNTR ( ), apart from Hong Kong, Macau, and Taiwan, the countries with the largest FDI in China are United States ($14.06 billion), Japan ($13.59 billion), South Korea ($9.36 billion), Singapore ($6.54 billion), and Germany ($3.00 billion). 10

11 other ownership types. For example, collectively-owned plants refer to those jointly owned by farmers in a village. (3) City-level variables on development: We construct variables on city development from the World Bank Enterprise Investment Climate Survey (2004). Our main development measure, Development, is computed by averaging the scores on financial market access across all enterprises that participated in the survey in each city. We use this measure to proxy for regional financial development. Similarly, we compute the average number of days customs clearance takes and the average scores on infrastructure development in each city, respectively, to obtain Customs Bureaucracy and Infrastructure Development. 3.3 Summary Statistics Table 2 presents summary statistics for our main variables. To mitigate the effect of outliers, all variable except categorical ones are winsorized at the 1 and 99 percent levels. As mentioned earlier, 21.36% of entrants in our sample export. Among all entrants, 16.5% are foreign, and (46.25%) are state- (privately)-owned. An average entrant operates in an industry with a NTR gap of 29.82%, and have million RMB assets, 170 employees, liabilities/assets of 57.44%, corporate taxes/sales of 0.57%, and cash/assets of 19.79%. Insert Table 2 Here On average, for an entrant that survives the first year, its asset growth rate is 29.6%. The average two-year and three-year asset growth rates for a surviving entrant are 61.2% and 98.44%, respectively. If we assume the asset growth rate of a non-surviving plant as -100%, the average one-year, two-year, and three-year asset growth rate is, -0.94%, 2.63%, and 7.36%, respectively. 4. The Effect of Trade Shock on Entry Rate and Export Propensity In this section, we describe the trade shock from Pierce and Schott (2015) and examine the effect of the trade shock on entry rate and export propensity of the entrants. 11

12 Since 1980, the U.S. has applied the Normal Trade Relations (NTR) tariff, the relatively low rates reserved for WTO members, to Chinese imports. However, these low NTR rates required annual renewals which were politically contentious. Without the renewal, the tariffs could revert to a much higher non-ntr tariff rates, originally set in 1930 under the Smoot- Hawley Tariff Act. In October 2000, the U.S. Congress passed the law granting Permanent Normal Trade Relations (PNTR) to China, thereby eliminating the possibility of sudden spikes on tariff. Pierce and Schott (2015) document a link between this tariff shock and a sharp decline in manufacturing unemployment in the U.S. As noted earlier, following Pierce and Schott (2015), we define NTR gap as the difference between the non-ntr rates and the NTR rates that were locked in by PNTR for each industry. A potential concern is that raw NTR gap might be correlated with some sectoral characteristics. Appendix Table 2 confirms that the raw tariff gap is correlated with export propensity: Industries with high NTR gap tend to have a high propensity to export in our sample. As an example, textile mill products, which represent 14% of total entries, have an average NTR gap of 49.39% and export propensity of 42.91%, compared with the average NTR gap of 29.82% and export propensity of 21.36% in the full sample. In contrast, chemicals and allied products, which represent 11.62% of total entries, have an average NTR gap of 22.70% and export propensity of 13.70%. To alleviate this concern, we strengthen the identification strategy in Pierce and Schott (2015) by matching the treatment with control from the same industry group, splitting each 2- digit SIC into high and low NTR gap subsamples. 14 By construction, we exclude industry groups with zero variation of NTR gap. We then use the standard difference-in-difference approach, comparing observations with high NTR gap and low NTR gap before and after the granting of 14 Export propensity is still higher in treatment than in control group in Table 4, but compared to Appendix Table 2, the industry matching does reduce the pre-treatment difference in export propensity between treatment and control groups. 12

13 PNTR to China. In regression analyses, we use the difference-in-difference shock term (PstPNTR*High NTR Gap) to assess the impact of trade shock. We begin our analysis by examining whether manufacturing entry increases for treatment industries following the trade shock. Table 3 presents the results. Panel A shows that entry rate of treatment industries does not increase more than control group: the increase in entry rate is 4.71% for the high NTR gap group but 5.43% for the low NTR gap group. The results, however, are opposite for the foreign subsample. Before and after PNTR, the entry rate among foreign establishments increases 1.35% for treatment group and 0.92% for control group. The difference in the increase of entry rate between the treatment versus control group is 0.42%, which is around 31% of the increase in entry rate for the treatment group itself. Insert Table 3 Here Panel B of Table 3 reports OLS regression results on the effect of the trade shock on entry propensity after controlling for industry, year and city fixed effects. Our results confirm the previous finding that only in the foreign subsample, the likelihood of entry is significantly positively affected by the shock. The coefficient on the shock term is estimated at , implying that before and after the shock, the propensity of entry increases 0.82% more for high NTR gap industries than low NTR gap industries. 15 The magnitude of the effect of shock (0.82%) almost doubles the previous comparative statistics result (0.42%) after controlling for the effects of industry, year and plant location on the likelihood of entry. In contrast, the coefficients on the shock term are not significant for either the overall sample or the domestic subsample. 15 In Appendix Table 3, we double sort the sample by ownership and development. We find that trade shock only affects entry rate among foreign plants in developed areas. This result differs from our later finding that the effect of trade shock on export propensity of foreign entrants is more pronounced in undeveloped areas. These findings suggest that market imperfections might prevent foreign businesses to enter, but if they manage to overcome these imperfections, foreign entrants have comparative advantages in undeveloped markets. 13

14 Next, we investigate among the entrants whether the likelihood of export during the first year of entry increases following the shock. Table 4 presents the results. Panel A shows that the propensity to export increases from 24.83% to 26.12% (an increase of 1.29%) for treatment group and from 18.17% to 18.27% (an increase of 0.10%) for control group, implying that entrants export propensity increases around 1.20% more for treatment group. This indicates a strong impact of the shock, as the difference between two groups (1.20%) is in a similar magnitude as the increase in export entry by treatment groups itself (1. 29%). We also consider an alternative measure export entry scaled by total observations, rather than by number of entries. We find similar results based on this alternative measure: export entry divided by total observations increases by 1.46% for treatment group and 1.01% for control group. Insert Table 4 Here In Panel B, we examine export entry propensity using three different regression models. In Column (1), we report OLS regression results with the following explanatory variables: the industry-level indicator High NTR Gap, the timing indicator related to the shock PstPNTR, and the interaction term between the first two. Our result indicates that the difference between two industry groups, as documented in Panel A, is statistically significant. In Column (2), we further control for industry, year and city fixed effects. The magnitude of the coefficient on the shock term increases substantially from 1.20% (Column 1) to 6.95% (Column 2), implying that the likelihood of export entry increases 6.95% more for industries with High NTR gap. We use OLS estimation to ease the interpretation of economic effects in the first two columns. But, the last column confirms that Probit estimation also yields the same outcome We adopt the specification in Column (2) as our benchmark specification in subsequent analyses, due to the complications regarding estimating and interpreting interaction terms in Probit models. In unreported analyses, we confirm that our main results are robust to using OLS regressions without controlling for fixed effects (as in Model 1) and Probit regressions (as in Model 3). 14

15 In conclusion, prior papers (Pierce and Schott, 2015, 2016) show that granting China PNTR affects firms in the U.S. Our findings reveal that this shock also affects Chinese establishments. After the shock, entry rate rises for foreign establishments, and the propensity to export greatly increases for all entrants. 5. Who Is More Responsive to Trade Shock? Given our finding that granting China PNTR significantly increases the propensity to export for entrants in China, in this section, we ask among different types of entrants who is more responsive to this shock. To answer this question, we divide the sample of entrants by ownership and location and repeat the previous analysis. Table 5 presents the results. Insert Table 5 Here The results in Table 5 show that foreign plants are more responsive to trade shock than domestic plants. Comparative statistics results in Panel A indicate that the propensity to export increases 2.19% (0.65%) more for foreign (domestic) plants in treatment group relative to the control group. In the OLS regression results in Panel B, the shock coefficient is estimated at for foreign plants and for domestic plants, suggesting a 2.9% difference between the foreign and domestic subsamples regarding the relative increase in export propensity. 17 As discussed earlier, foreign plants have competitive advantages in terms of technology, brand, know-how, financial resources, and distribution channels, allowing them to take advantage of the trade shock. Inspection of Table 5 also reveals that plants in financially developed cities are more responsive to trade shock than those in underdeveloped cities. Comparative statistics results in 17 In unreported analyses, we adopt an alternative specification with a triple interaction term between the shock variable (PstPNTR*High NTR Gap) and Foreign Indicator. The coefficient estimate of this term is positive and significant, confirming that foreign entrants are more likely to export than domestic entrants following the trade shock. Adopting the same approach, we also find that entrants in developed cities are significantly more likely to export than those in undeveloped cities following the shock. 15

16 Panel A show that the propensity to export increases 1.54% (0.09%) more for plants in developed (undeveloped) cities in treatment versus control industries. In Panel B, the shock coefficient is estimated at for plants in developed cities and in underdeveloped cities, suggesting a 2.3% difference in developed versus undeveloped cities regarding the relative increase in export propensity. A potential explanation is that exporting activities require more financial resources than domestic sales. Establishments in financially developed areas may have greater ability to raise external capital to finance costly exporting activities, consistent with the notion that financing is a key issue in the decision whether to export. Appendix Table 4 shows that our results in Tables 4 and 5 are robust to the extended sample period of 1999 to 2007, which spans seven years after the trade shock. We also break down our subsamples of foreign and domestic plants into more detailed ownership types. Among domestic plants, we examine the subsamples of state- and privately-owned plants. Among foreign plants, we separate plants with investment from Hong Kong, Macau, and Taiwan and those with investment from other countries. We also consider whether a plant has a controlling/noncontrolling foreign stake based on the percentage of capital stock contributed by foreign investors, using the 25% cutoff. The subsample regression results in Panel A of Appendix Table 5 confirm that all four types of foreign plants (those with investment from Hong Kong, Macau, Taiwan and from other countries and those with controlling or non-controlling foreign stake) are more responsive to the trade shock than privately-owned domestic plants, with state-owned domestic plants being least responsive to the shock. Taking together, our results in this section show that foreign plants and plants in financially developed areas are more responsive to trade shock: Specifically, foreign entrants and entrants in financially developed areas are more likely to export following the shock Foreign plants and plants in financially developed areas also have higher export propensity before the establishment of PNTR. (See Rows 1 and 4 of Table 5 Panel A for details.) This is consistent with 16

17 6. The Role of Financing in Export Entry Having shown that foreign entrants are more likely to take advantage of the trade shock than domestic entrants, we examine whether such differential responses are due to financing issues. As discussed earlier, previous literature has shown that export requires more financial resources than domestic sales (Beck, 2002, 2003; Zia, 2008; Minetti and Zhu, 2011). Domestic plants tend to rely more on local financial markets. In financially developed areas, domestic plants are more likely to have access to external funds to finance exporting activities. In underdeveloped capital markets, however, domestic plants may not be able to raise the optimal amount of external capital, so they are at a disadvantage relative to foreign plants. That is, if the differential responses are related to financing issues, the disparity between financially developed and undeveloped areas should be larger among domestic plants. We test this hypothesis directly by double-sorting the sample by ownership and financial development in this section. Table 6 presents the results. Insert Table 6 Here Both the comparative statistics (Panel A) and regression results (Panel B) confirm the above hypothesis. For foreign plants, the differential response between developed and undeveloped areas is 0.15% according to the regression analysis: The shock coefficient is 7.98% for developed areas and 7.83% for undeveloped areas. For domestic plants, however, the differential response is 1.76%: The shock coefficient is 6.07% for developed areas and 4.31% for undeveloped areas. Appendix Table 5 shows that our results are robust to an alternative specification where we use interaction terms instead of subsamples. Here, we interact Foreign Indicator with the shock term as well as time and industry fixed effects. The results confirm that foreign plants are expectation since factors that allow these plants to export more prior to the shock should allow them to be more responsive to the shock as well. 17

18 more responsive to shock than domestic plants and the difference is larger in financially underdeveloped area. In sum, the results in this section show that the difference between foreign and domestic plants is concentrated in financially underdeveloped areas. This finding is consistent with the notion that financing issues drive the cross-sectional difference in responses to the trade shock. We break down domestic and foreign plants into more detailed ownership types, and report subsample regression results in Appendix Table 6. Our results show that among domestic entrants, private plants are more responsive to the shock than state-owned plants. The developedundeveloped difference is greater among private plants. Since private plants are likely to rely more on local financing, this finding further highlights the role of financing in exporting activities. Our main result that foreign plants are more responsive to trade shock than domestic plants is also robust across different subgroups of foreign entrants: foreign plants with investment from Hong Kong, Macau, and Taiwan or from other countries, and foreign plants with controlling or non-controlling foreign stake. 7. Comparing Financial Conditions of Foreign versus Domestic Entrants To shed further light on the reason why foreign entrants are more capable of responding to the trade shock, in this section, we compare the financial conditions of foreign and domestic entrants using our full sample. Specifically, we test whether there exist differences in size, capital structure, tax liability, and liquidity between foreign and domestic entrants, and whether such differences vary by regional financial development. Table 7 reports the results. Insert Table 7 Here We show that foreign entrants are larger, better capitalized, less taxed, and more liquid than domestic entrants. The univariate results in Panel A of Table 7 indicate that an average foreign entrant is million RMB larger and the ratio of corporate income tax divided by sales is 0.35% lower for an average foreign entrant. The leverage ratio of foreign plant is 5.43% lower 18

19 and the cash ratio of foreign plants is 1.77% higher than those of domestic plants. All the differences are significant at the 1 percent level and robust to the inclusion of controls for industry, year, and location fixed effects. These results confirm that foreign plants are less financially constrained than domestic plants. Next, we examine whether the differences between foreign and domestic plants are more pronounced in underdeveloped financial markets. In the even columns of Table 7 Panel B, we include an interaction between Foreign Indicator and Development. The results show that domestic establishments are smaller and less liquid than foreign establishments, particularly in underdeveloped areas. We also find evidence that domestic establishments in underdeveloped areas pay much more corporate taxes than foreign establishments. 19 In Appendix Table 7, we divide domestic entrants into state and private subsamples. State-owned entrants are larger and more liquid than private entrants but smaller and less liquid than foreign entrants. State entrants are more levered than both foreign and domestic entrants, perhaps due to their ability to obtain inexpensive loans from state-owned banks. State entrants also pay more taxes than foreign and private entrants. Overall, examination of entrant characteristics indicates that foreign entrants are less financially constrained than domestic entrants. The differences are more pronounced in underdeveloped areas. 8. Performance of Foreign versus Domestic Entrants 19 Following the open-door policy, the central government of China implemented a series of tax incentive programs for foreign plants, particularly for foreign exporting plants, during our sample period. As an example, foreign plants were exempted from corporate income taxes during the first two years after they turned profitable and granted 50% reductions in corporate income tax during the third to fifth year. Local governments are encouraged to initiate additional tax incentives for foreign plants. These incentive plans tend to be greater in underdeveloped areas for attracting foreign businesses. 19

20 Thus far, we have shown that foreign entrants have comparative advantages in term of financial resources, and therefore they can better take advantage of the trade shock. In this section, we assess whether these foreign entrants indeed perform better than their domestic peers. We measure the performance of entrants along two dimensions: the first is survival rates during the one-year, two-year, and three-year periods after the entry, and the second is one-year, twoyear, and three-year growth rates of surviving entrants. Table 8 presents the results. Insert Table 8 Here Panel A of Table 8 compares various descriptive statistics between the two groups. The results show that foreign plants are more likely to survive than domestic plants. The survival rates for foreign (domestic) plants are 84.35% (75.53%), 76.34% (62.64%), and 70.43% (53.84%) one, two, and three years after entry, respectively. Furthermore, among plants that survive, foreign plants grow faster than domestic plants. The medians of asset growth of surviving foreign (domestic) plants are 10.97% (9.03%), 25.41% (25.01%), and 44.05% (40.75%) one, two, and three years after entry, respectively. Finally, the cross-sectional dispersion of asset growth is higher among domestic plants. The standard deviations of growth rates among surviving foreign (domestic) plants are 67.58% (73.30%), 1.08 (1.18), and 1.51 (1.70) one, two, and three years after entry, respectively. 20 Inspections of Panel B of Table 8 reveal that such differences in performance are statistically significant, after considering the differences in industry, year, and location between the two groups. In addition, such differences are more pronounced in underdeveloped financial markets. The estimated coefficients on the interaction term between Foreign Indicator and Development are consistently negative whereas the coefficients on the Foreign Indicator are all 20 Appendix Table 8 presents the comparative statistics based on the developed and underdeveloped area subsamples. We find that the foreign-domestic differences in survival rates, growth rates, and growth dispersions are larger in underdeveloped areas. In fact, the growth rates of surviving foreign plants are larger than those of domestic plants only in underdeveloped areas. In developed areas, surviving domestic plants grow faster. 20

21 positive, confirming that the foreign-domestic differences are concentrated in underdeveloped financial markets. 21 In sum, foreign entrants are more likely to survive and grow faster than domestic entrants. In term of growth rates, there is less heterogeneity among surviving foreign entrants, compared to their domestic peers. The foreign-domestic differences are greater in underdeveloped areas, consistent with the prior results. 9. Additional Tests Our results thus far suggest that foreign entrants in China benefit more from trade liberalization since they are less financially constrained than their domestic peers. In this section, we explore alternative interpretations of our main results, and discuss additional test results. 9.1 The Effect of Size on the Responsiveness to Trade Shock Ayyagari and Maksimovic (2017) show that small businesses in the U.S. are hurt most by trade liberalization. It thus becomes a point of interest to investigate whether small or large plants in China benefit most from trade liberalization. Furthermore, given our previous finding that foreign entrants are typically larger, it is important to mitigate the concern that the responsiveness to the shock by foreign entrants is solely driven by the effect of their large size. We start with examining the effect of plant size on the responsiveness to trade shock. We divide the sample into four size groups based on number of employees. Following Ayyagari and Maksimovic (2017), the cutoffs of the number of employees are 50, 100, and 250 for the size groups. Table 9 presents the results. We show that large plants are more responsive to trade shock than small plants. Both the comparative statistics and regression results reveal that the effect of trade shock on export propensity increases monotonically with plant size. The economic 21 Appendix Table 9 presents the regressions of performance on state and private indicators. We find that the performance of state-owned entrants is worse than private entrants but both state- and privately-owned entrants perform worse than foreign entrants. 21

22 magnitude of the size effect is large. For example, in the regression analysis, the shock coefficient is estimated at for entrants in the smallest group and for entrants in the largest group, implying a 7% difference in the relative increase in export propensity for entrants in the largest versus smallest group. Prior literature identifies size as a key component of financial constraint measures (e.g., Hadlock and Pierce, 2010; Whited and Wu, 2006). So, a potential explanation is that large firms are able to overcome financial constraint related to exporting. Insert Table 9 Here Next, we investigate whether the differential responses between developed and undeveloped areas vary depending on plant size. In Appendix Table 10, we repeat similar analyses in Table 6 for size and find that the differential responses between developed and undeveloped areas are more pronounced among small plants. For plants with 1-49 employees, the differential response between developed and undeveloped areas is estimated at 3.22%, while, among plants larger than 250 employees, the differential response is estimated at 1.40%. Finally, the results in Appendix Table 11 confirm that the effects of size and foreign ownership are distinct. When we double-sort the observations by size and ownership structure, we find that the differential response between foreign and domestic plants are present in both large and small subsamples. The differences between state and private plants are also not driven by size. 9.2 Alternative Measures of Development In the previous sections, we examine the effect of financial development based on plant location. Here, we consider two other important aspects of regional development that may also affect export propensity. First, we examine the effect of government inefficiency in the form of customs bureaucracy. Government red tape may increase the costs of doing business and impede international trade. As noted earlier, we use the average number of days it takes to clear customs 22

23 in a particular city constructed from the World Bank Enterprise Investment Climate Survey (2014) as a proxy for bureaucracy. Second, we examine the effect of economic development in general using the average scores on infrastructure development in a particular city, also from the World Bank Enterprise Investment Climate Survey (2004). Well-developed infrastructure should increase business activities and facilitate export. Table 10 reports the results. Insert Table 10 Here Table 10 Panel A shows that similar to financial development, plants in areas with less bureaucracy and better infrastructure are more responsive to trade shock. The shock coefficients are estimated at in areas with less bureaucracy and in areas with more bureaucracy, implying a 3.2% difference in relative increases in export propensity between the two groups. The shock coefficients are estimated at in areas with better infrastructure and in areas with worse infrastructure, implying a 1.6% difference in relative increases in export propensity between the two groups. In Panel B of Table 10, we double-sort entrants by ownership and institutional measures. Unlike financial development, the developed-undeveloped differences in responses to shock do not vary across ownership types. From the first four columns, the difference between bureaucratic and non-bureaucratic areas is 1.79% for foreign plants and 1.45% for domestic plants. From the last four columns, the difference between areas with good and bad infrastructure is 1.44% for foreign plants and 1.41% for domestic plants. These findings suggest that the competitive advantage of foreign plants does not come from superior abilities to deal with government inefficiency or overcome poor infrastructure. In addition to the two measures from World Bank Enterprise Investment Climate Survey (2014) above, we examine the effects of trade shock across cities with different initial conditions based on the pre-shock distribution of local plants. We compute, for each city, the fraction of plants that are (1) private small businesses (private plants with fewer than 50 employees) and (2) foreign businesses. Here, we focus on 23

24 private plants rather than state plants since the presence of private plants is more likely to reflect a city s developmental conditions. (The measures constructed from all firms yield similar results.) We use the data from 1999 (the year prior to the trade shock) to compute the measures to ensure that they are not endogenously driven by the shock themselves. The results are reported in Table 11. Insert Table 11 Here In Panel A of Table 11, we find that the effects of trade shock on export propensity is greater in cities where exporters are more prevalent. The shock coefficients are estimated at in areas with more exporters and in areas with fewer exporters. Intuitively, a large presence of exporters signifies friendly business environment that encourages international trade so it is expected that entrants in these areas are more responsive to trade shock. The effects of trade shock on export propensity is greater in cities with larger plants. The shock coefficients are estimated at in areas with larger plants and in areas with smaller plants. We note that this result is different from Chinitz (1961). The Chinitz s hypothesis suggests that the prevalence of small firms is associated with local economic development because small firms have stronger ties with local suppliers and are more conducive to knowledge spillover through local labor forces. Perhaps, in our Chinese context, a large fraction of small businesses reflects a city s lack of opportunities for firms to grow rather than Chinitz s active entrepreneurial sectors. In Panel B of Table 11, we double-sort entrants by ownership and city conditions. The inter-city difference in trade shock response is concentrated in cities among domestic plants. In other words, similar to financial development, the city conditions proxied by the ex-ante fractions of exporters and small firms are more important for domestic firms than foreign firms. 9.3 Local General Equilibrium Effects of Trade Shock on Export Propensity 24