The Impact of Input Trade Liberalization on the Entry of Foreign Firms: Evidence from a Quasi-Natural Experiment in China

Abstract This paper integrates trade policy and foreign direct investment into a unified analytical framework, and investigates the effects of input trade liberalization on the entry of foreign firms. To identify the causal effects, we utilize China’s accession to the WTO in 2001 as a quasi-natural experiment, and perform difference-in-difference estimation. The results show that input trade liberalization significantly increases foreign entry. We also find that input trade liberalization not only promotes the entry of new foreign firms, but also restrains the exit of existing foreign firms, thereby contributing to the net growth of the number of foreign firms. The mechanism tests show that increasing variety as well as quality of intermediate input and reduction in marginal cost are the potential channels through which input trade liberalization promotes foreign entry. This paper further demonstrates that institutional environment strengths the positive effect of input trade liberalization on foreign entry, and the promotive effect of input trade liberalization on foreign entry increases with industry import intensity, additionally, input trade liberalization is also conducive to improving the quality of foreign investment.


Introduction
With its rapid economic growth and increasingly open investment environment, China has attracted a large number of foreign-funded enterprises to carry out international direct investment activities since the reform and opening-up in 1978.According to the '2010-2012 World Investment Prospects Survey Report' issued by UNCTAD, China ranks first in the list of the top 15 most attractive investment destinations in the world; and in 2021, China's actual use of foreign capital exceeded trillion yuan for the first time, reaching 1.1 trillion yuan, an increase of 14.9%, and the absorption of foreign capital remained the second in the world.There is entry.More importantly, we further demonstrate the heterogeneous impacts of input trade liberalization on the entry of different types of foreign capital, and the results show that input trade liberalization significantly promotes the entry of OECD foreign firms, whereas does not have any effect on the entry of foreign firms from Hong Kong, Macao and Taiwan, which indicates that input trade liberalization is conducive to improving the quality of foreign investment.
Taken together, the significance of this study is that, by examining the impact of input trade liberalization on the entry of foreign firms and the mechanism of its effect, it provides new explanations for the dynamic changes in China's FDI attraction after WTO accession, and to a certain extent enriches the research on the determinants of FDI in China.More importantly, it can also provide important guidance to China and other developing countries in formulating policies to attract foreign investment in a larger scale and higher quality manner.

Literature review
The main goal of this paper is to investigate the impact of input trade liberalization on foreign entry.The paper is related to at least two strands of the growth literature.The first related strand of the literature is the economic effects of input trade liberalization, and most of the studies have focused on the effects of trade liberalization on firm productivity (Ahsan, 2013;Amiti & Konings, 2007;Brandt, Van Biesebroeck, Wang, & Zhang, 2017;Defever, Imbruno, & Kneller, 2020;Fiorini, Sanfilippo, & Sundaram, 2021;Hu & Liu, 2014;Topalova & Khandelwal, 2011;Yu, 2015).For instance, Amiti and Konings (2007) analyze Indonesia manufacturing firms from 1991 to 2001, and find that firms' productivity gains from reduction of input tariffs are at least twice as much as those from reduction of output tariffs.Subsequently, an empirical study of Indian enterprises by Topalova and Khandelwal (2011) also confirmed the conclusion of Amiti and Konings (2007).Further, Ahsan (2013) uses Indian firm-level data to examine the complementarities between the speed of contract enforcement and the productivity gains from input tariff reductions, and demonstrates that the positive impact of input tariff reductions on firm productivity is normally stronger in those state with higher judicial efficiency.Hu and Liu (2014) investigate the effect of trade liberalization on Chinese manufacturing firms' productivity, and demonstrate that the overall productivity gain from the trade liberalization is a net result of a productivity enhancing effect of input tariff cut and a productivity depressing effect of output tariff cut.Yu (2015) studies the impact of trade liberalization on firm productivity in combination with the typical characteristics of China's processing trade, and finds that both of input and output tariff reductions increase firm productivity, and the effects are weaker as firms' share of processing imports grows.Brandt et al. (2017) not only investigate the impact of trade liberalization on the productivity of Chinese enterprises, but also explore whether trade liberalization can improve the overall performance of manufacturing industry through resource reallocation effect at the industry level.Recently, Defever et al. (2020) explore the role of wholesalers in mediating the productivity effects of trade liberalization, and find that firms that do not directly import experience productivity gains from input tariff reduction if trade intermediation of foreign inputs within their sector is high.Fiorini et al. (2021) take the further step to investigate the role of road infrastructure in determining the impact of a reduction in input tariffs on firm productivity, and find that the productivity effect is larger for the firm in areas where better roads improve access to other intranational markets.
The relationship between trade liberalization and firms' export has also attracted the attention of some scholars.Tian and Yu (2017) show that input trade liberalization is positively correlated with both of firms' export participation & export intensity by using Chinese firm-level data for 2000-2006.In addition, some researchers are also interested in the relationship between trade liberalization and firms' export quality.Bas and Strauss-Kahn (2015) investigate the effects of input trade liberalization on imported input and exported product prices, and find that firms exploit the input trade liberalization to upgrade the quality of their inputs in order to The impact of input trade liberalization on the entry of foreign firms 269 upgrade the quality of their exported products.Likewise, Fan, Li, and Yeaple (2015) argue that tariff reductions induce an incumbent firm to increase the quality of its exports based on Chinese firm-level data.Besides, studies by other scholars have also found that trade liberalization has an important impact on firms' innovation (Liu & Qiu, 2016;Liu et al., 2021) as well as firms' profit (Breinlich, 2016).However, although there are many studies investigating the economic effect of trade liberalization, there has been little empirical evidence on whether and how input trade liberalization affects foreign entry.
This paper is also related to the literature on the determinants of foreign direct investment.Kang and Hong (2007) investigate the determinants of location choice for south Korean multinational companies using firm level data for south Korean foreign affiliates in China, and find that market size and policies of economic development zones play a positive role in attracting South Korean foreign investment, while labor costs show negative and significant coefficients.Amiti and Javorcik (2008) also examine the determinants of entry by foreign firms, using information on 515 Chinese industries at the provincial level during 1998-2001, they find that market and supplier access are the most important factors affecting foreign entry. 2 Cai, Lu,  Wu, and Yu (2016) explore whether environmental regulation affects inbound foreign direct investment, and demonstrate that tougher environmental regulation leads to less foreign direct investment, and such an effect only exists in foreign multinationals from countries with worse environmental protections than China.Recently, Crescenzi, Cataldo, and Giua (2021) investigate the impact of Investment Promotion Agencies (IPAs) on FDI attraction, and find that IPAs significantly promote the entry of foreign capital, and such an effect is over and above other policies targeting the general economic improvement of the host economies.Although each of the above-mentioned studies focuses on the determinants of foreign entry from multiple dimensions, none of them directly touches on the influence of input trade liberalization, which is the focus of the present study and distinguishes the present study from the existent literature.

Empirical strategy
Our primary goal is to investigate the effects of input trade liberalization on foreign entry.To accurately examine the impacts of input trade liberalization, we take advantage of the fact that most of industries undergone substantial variations in input tariff reduction due to China's WTO accession, and adopt the DID technique to conduct the estimation.In the online Supplementary Appendix A, we describe the background regarding China's accession to WTO as well as some stylized facts.Specifically, after China joined WTO, industries that had higher initial input tariffs (i.e. more protected) would experience larger input tariff reductions under the WTO agreement, whereas industries with lower initial input tariffs would experience smaller input tariff cut after the WTO accession.We thus compare the number of foreign firms in industries experiencing larger input tariff reductions (i.e. the treatment group) to those in industries experiencing less input tariff reductions (i.e. the control group) before and after China's WTO accession.We consider the following specification for our benchmark DID estimation: where i, j, and t correspond to industry (three-digit Chinese Industrial Classification), region and year, respectively.ForEntry ijt denotes foreign entry, measured by the natural logarithm of '1þ NUM', where NUM is the number of foreign firms in industry i in region j in year t.InputTa i01 is the input tariff rate of industry i in 2001, we construct this variable following Amiti and Konings (2007) as well as Topalova and Khandelwal (2011), which uses an input cost-weighted average of output tariffs, that is 3 where h ik is the cost share of input k in the production of a good in industry i, we adopt Chinese Input-Output (IO) Table for year 1997 to calculate the weight h ik ; outputtar k, 2001 is the output tariff of industry j in year 2001, which is calculated as the simple average of the Harmonised System (HS) six-digit tariffs covered in industry k. Post02 t represents a post-WTO period, taking a value of 1 if it is year 2002 and onwards, and 0 otherwise.
The interaction term InpTar i01 Â Post02 t is our major interest, whose parameter b captures the average difference of the number of foreign firms in industries with higher initial input tariffs and lower initial input tariffs before and after China's accession to WTO, that is, the causal effect of input trade liberalization on foreign entry.In particular, a positive sign of b indicates that the number of foreign firms in industries with higher initial input tariffs would increase more, in other words, input trade liberalization helps to promote foreign entry.The vector k ij is a full set of region-industry fixed effects, controlling for all time-invariant differences across region-industry; the vector c t is a full set of year fixed effects to control for all yearly common shocks; the vector d jt is a full set of region-year interaction fixed effects, controlling for the influence of all time-varying regional factors; e ijt is the error term.We cluster standard errors at the industry level to address the potential serial autocorrelation and heteroskedasticity following Bertrand (2004).
In addition, we control for several characteristics X ijt that could affect foreign entry, the control variables include: capital intensity, export intensity, industry size, industrial output tariff, external tariff, industrial Herfindahl-Hirschman Index (HHI).Moreover, considering the choice of input tariffs in 2001 was nonrandom, to alleviate this identification concern, we follow Lu and Yu (2015) and add interactions between significant input tariff determinants (including SOEsh 01 , WAGEper 01 , and EXPint 01 ) and post-WTO indicator (Post02), to our DID regression.The definitions and selected reasons of these variables are outlined in online Supplementary Appendix B.

Data
The sample of our empirical analysis is derived from the following two disaggregated data sets: tariff data and firm-level production data.The first data source is from WTO website, which provides detailed information on the number of tariff lines, and average, minimum and maximum ad valorem tariff duties for each product defined at HS six-digit level.Since tariff information on the WTO website is available for 2001-2007 while missing for 1998-2000, we thus replenish the missing tariff data from the World Integrated Trade Solution (WITS) website maintained by the World Bank.It is worth noting that the HS codes used before and after 2002 are different, we therefore match the 1996 HS codes and the 2007 HS codes to the 2002 HS code using the standard HS concordance table.
Our second data source is the Annual Survey of Industrial Firms (ASIF) between 1998 and 2007, conducted by the National Bureau of Statistics (NBS) of China.The data set covers all state-owned firms and private firms with sales greater than RMB 5 million (approximately US$ 600,000).Since this paper studies the impact of input trade liberalization on foreign entry at the region-industry level, we need to use firm level information to construct regionindustry specific variables.In online Supplementary Appendix C, we process the data of Chinese ASIF data in several ways.We then follow Lu (2008) as well as Lu, Tao, and Zhu (2017) and define enterprises with foreign capital share above 25% as foreign firms, and on this basis, the region-industry level foreign entry as well as other control variables are constructed.Online Supplementary Appendix Table C1 presents the summary statistics of the main variables used in this study.
To preliminarily investigate the relationship between input trade liberalization and foreign entry, we first divide industries into two groups according to the median value of input tariffs: a group with lower initial input tariffs (i.e. the industries with input tariffs below the sample median in 2001, that is, the control group) and a group with higher initial input tariffs (i.e. the The impact of input trade liberalization on the entry of foreign firms 271 industries with input tariffs above the sample median in 2001, that is, the treatment group); we then plot the differential trends over time of average number of foreign firms for both the two groups.For the convenience of comparison, we normalize the number of foreign firms in 2001 for both groups to be one.As shown in Figure 1, the two types of industries have quite similar trends in the pre-WTO period, presenting some descriptive evidence in support of assumption that our treatment and control groups are ex ante comparable.However, they diverge in the post-WTO period.In particular, the treated group that experienced larger reduction in input tariffs exhibits more impressive increase in the number of foreign firms, and the divergence between the two groups is enlarging after China's accession to the WTO.This divergence suggests potential positive effects of input tariff reductions on foreign entry.

Baseline results
Table 1 presents the baseline results.Column 1 only controls for region-industry fixed effects and year fixed effects, we find that the interaction term InpTar01ÂPost02 is positive and significant at the 1% level, suggesting that after China's WTO accession, the number of foreign firms increases more in industries experiencing larger input tariff reductions (i.e. the treatment group) compared with that in industries experiencing less input tariff reductions (i.e. the control group), in other words, input trade liberalization promotes foreign entry significantly.In column 2, we include a set of time-varying characteristics (e.g.capital intensity, export intensity, industry size) that may influence foreign entry.Clearly, the interaction term InpTar01ÂPost02 remains positive and statistically significant, indicating that input trade liberalization helps to promote foreign entry after controlling for the above-mentioned characteristics.Considering China's accession to the WTO, except the reduction of input tariff rate, the output tariff rate also declines, and the latter may increase import competition, thus affects the entry of foreign firms as well.In addition, during the sample period, the external tariff rate has also been reduced, which means that the export cost of Chinese enterprises tends to decline, which may also affect foreign entry.We thus include OutputTa 01 ÂPost02 and external tariff in column 3 of Table 1, we see that the interaction term InpTar01ÂPost02 is still statistically significant and  The impact of input trade liberalization on the entry of foreign firms 273 positive after controlling for the impacts of output tariff and external tariff reductions.Considering that the availability of higher quality inputs and the reduction in production costs resulting from input tariff reductions may affect the entry of domestic firms into the market, leading to the intensification of market competition, which may further affect the entry of foreign firms.To this end, we control for overall industrial competition by including the Herfindahl-Hirschman Index (HHI) in column 4 of Table 1, clearly, our result is robust to including this competition variable.
One may concern that the choice of input tariffs in 2001 was nonrandom, and, hence, our treatment and control groups could be systematically different ex ante.To alleviate this identification concern, we follow Lu and Yu (2015) and add interactions between significant input tariff determinants and post-WTO indicator (including SOEsh 01 ÂPost02, WAGEper 01 ÂPost02, and EXPint 01 ÂPost02), to our DID regression.Column 5 of Table 1 presents the regression results, evidently, our coefficient of interest on the interaction term, InpTar01ÂPost02, remains robust and stable with a similar magnitude.In reality, three important ongoing policy reforms (i.e. the SOE reform, the relaxation of FDI entry regulations and the reduction of trade policy uncertainty) took place around the time of the WTO accession.Considering that these policies may influence foreign entry, and hence the DID approach would be contaminated, we add three additional control variables to control for the impact of these policy effects.As for the SOE reform, we measure it by calculating the ratio of the number of SOEs over the total number of domestic firms following Lu and Yu (2015).For the relaxation of FDI entry regulations, we compare the 1997 and 2002 versions of the Catalogue for the Guidance of Foreign Investment Industries following Lu et al. (2017).In particular, we can identify the possible changes (i.e.supported, restricted, prohibited, and permitted) in product categories as well as their corresponding classifications in the changes in FDI regulations (i.e.encouraged, discouraged, or no change), by comparing the 1997 and 2002 versions of the Catalogue.Next, we aggregate the changes in FDI regulations from the Catalogue product level to the industry level (four-digit CIC), and construct the dummy variable of FDI deregulation industry (i.e.FDILib), FDILib equals one if an industry belongs to the encouraged industries, and zero if an industry belongs to the no-change industries.Finally, we construct the interaction between the dummy of FDI deregulation industry (FDILib) and the post-WTO period dummy (Post02), to control for the influence of the relaxation of FDI entry regulations.As for the reduction of trade policy uncertainty, we first follow Handley & Limão (2017) as well as Liu and Ma (2020) and use the difference between the Column 2 tariff and the MFN tariff in 2000 to measure industrial NTR gaps (TPU), and then construct the interaction between TPU and the post-WTO period dummy (Post02).Column 6 of Table 1 presents the results after controlling for the three policy reforms.As expected, the interaction term FDILibÂPost02 is significantly positive, indicating that the relaxation of FDI entry regulations does promote foreign entry; however, there is no clear evidence that the SOEs reform and TPU reduction affects foreign entry significantly.More importantly, we note that after controlling for three simultaneous policy reforms, the coefficient for InpTar 01 ÂPost02 is still positive and statistically significant, once again indicating that input trade liberalization helps to promote foreign entry.
In column 7 of Table 1, we include region-year interaction fixed effects to control for the impact of regional time varying factors on foreign entry.We see that after controlling for the region-year interaction fixed effects, our regressor of interest, InpTar01ÂPost02 is still statistically significant and positive, verifying that input trade liberalization does promote foreign entry.So far, all the estimations have used the interaction of input tariff in 2001 (InpTar01) and the post-WTO indicator (Post02) as our regressor of interest.For the sake of completeness as well as robustness, we also employ the actual input tariff changes (denoted by DInpTar) as an alternative measure of the degrees of input trade liberalization from China's WTO accession following Liu and Qiu (2016), one important advantage of this measure is that it is more convenient to quantify the economic significance of input trade liberalization on foreign entry.Specifically, the actual input tariff changes are defined as DInpTar i ¼ InpTar i, 1998À2001 À InpTar i, 2002À2007 , where InpTar i, 1998À2001 and InpTar i, 2002À2007 denote the average input tariff over the period 1998-2001 and 2002-2007, respectively.The last column of Table 1 reports the regression results with DInpTarÂPost02 as the core explanatory variable.We find that the interaction term DInpTarÂPost02 is positive and statistically significant, suggesting that input trade liberalization significantly promotes foreign entry, evidently, our main findings are robust to various input tariff measures.
Finally, according to the regression results in column 8 of Table 1, we can further calculate the economic significance of input trade liberalization on foreign entry.In particular, with the estimated coefficient (2.1836) in the specification with actual tariff cut as key regressor, the average 4.59% input tariff cut because of China's WTO accession increases the number of foreign firms (in log) at the region-industry level by 0.1002; in addition, since the average number of foreign firms (in log) at the region-industry level of the entire sample is 1.031, the number of foreign firms increases by 9.7% as a result of input tariff liberalization, therefore, the effect of input tariff liberalization on foreign entry is also economically significant.
4.2.Checks on the identifying assumption 4.2.1.Expectation effect.To confirm the validity of our specifications, we start by checking whether foreign investors adjust their entry decisions in anticipation of the coming WTO accession.In particular, we introduce an additional control, the interaction between the input tariff in 2001 (InpTar 01 ) and the one year before WTO accession indicator (OneYearBefore), to the DID specification (1), and the estimate is reported in column 1 of Supplementary Appendix Table D1.We find that the coefficient on InpTar 01 ÂOneYearBefore is statistically insignificant, indicating that foreign investors didn't adjust their entry decisions in anticipation of the coming WTO accession, in other words, the expectation effect is less of a concern.4.2.2.Flexible estimations.We proceed to investigate the relationship between input tariff reductions and foreign entry year by year.More specifically, we replace the interaction term InputTa i01 Â Post02 t in the DID specification (1) with a battery of interaction terms between InputTa i01 and the year dummies.Column 2 of Supplementary Appendix Table D1 reports the regression results based on flexible specification (3).We find that the estimates on the interactions for 1998-2000 are statistically insignificant, suggesting that relative to 2001, the number of foreign firms in industries with higher initial input tariffs did not experience diverse relative to those in industries with lower initial input tariffs, this provides strong evidence that our treatment and control groups are ex ante comparable.In addition, Supplementary Appendix Figure D1 presents more straightforward illustration of the quantitative effects of input tariff reductions relying on the regression results in column 2 of Supplementary Appendix Table D1.As shown in Supplementary Appendix Figure D1, there are no significant differences in the pre-trends for treatment and control groups for all years before China's WTO accession, while the significantly positive impacts of input tariff reductions occurred from 2002 onwards, and the magnitudes also become larger.

Placebo tests.
We conduct placebo test to confirm our main DID estimation results in two ways.First, we conduct placebo test in looking at the impact of input tariffs on foreign entry in the pre-WTO accession period (i.e. 1999-2001).The rationale for this approach is that we should not expect any significant impacts of input tariffs on foreign entry as it did not change much during this period; if the result is the contrary, it may suggest the existence of some underlying confounding factors-other than the China's WTO accession-drive foreign entry and thereby bias our results.In particular, we replace InputTa i01 Â Post02 t in the specification (1) with Input tarif f it : The results are given in column 3 of Supplementary Appendix The impact of input trade liberalization on the entry of foreign firms 275 Table D1, we see that the coefficient on Input tariff is statistically insignificant, indicating that input tariffs have almost zero effect on foreign entry in the pre-WTO accession period.
In our second placebo test, we run the DID specification (1) using the subsample of processingtrade firms.The intuition behind this approach is that according to China's trade policy, processing-trade firms are allowed to import materials by enjoying zero input tariff rate, and hence the estimation using the subsample of processing traders should yield an insignificant liberalization impact on foreign entry.As shown in column 4 of Supplementary Appendix Table D1, we find that the coefficients on the regressor of interest, InpTar 01 ÂPost02, are small in magnitude and highly insignificant confirming the conjecture mentioned above.
4.2.4.Industry time trend.In the DID estimation, another important identifying assumption is that conditional on a list of controls (X ijt , k ij , c t ), the regressor of interest, InpTar i01 Â Post02 t , is uncorrelated with the error term e ijt : In other words, foreign entry of the treatment and control groups would follow the same time trend if there had been no China's WTO accession in the end of 2001.However, considering foreign entry trends in different industries could be different as they may be affected by industry-specific confounding factors, if this is the case, the identifying assumption mentioned above would be violated.To check whether unobserved industry-specific factors would contaminate our estimates, we add an industry-specific linear time trend (i.e. a i Á t) as an additional control in our DID specification (1), following Liu and Qiu (2016).Column 5 of Supplementary Appendix Table D1 reports the new estimation results using this approach, we find that our coefficients of interest on the interaction term, InpTar 01 ÂPost02, remain robust and stable with a similar magnitude, as compared to the baseline results reported in column 7 of Table 1.Therefore, there is strong evidence that our estimation results are not driven by any unobserved industry trends.

Two periods estimation.
There is a concern that DID estimation and resulting statistical inference depend on how to accurately calculate standard errors.Thus far, we have followed Bertrand (2004) to cluster standard errors at the industry level.As a robustness check, we use another approach also suggested by Bertrand (2004) to calculate standard errors.More specifically, we first divide our whole panel structure into two periods (pre-and after post-WTO accession), and then take the mean average of each variable in the two periods to perform the first-difference estimations using White-robust standard errors.By this way, we can also address the potential serial correlation due to some unobservable macroeconomic factors, which pointed in Bertrand (2004).The estimation results using this approach are presented in column 6 of Supplementary Appendix Table D1 and show qualitatively similar results as the benchmark.

Firm level regression and other robustness checks
4.3.1.Firm level regression.Thus far, we have empirically examined the effects of input trade liberalization on foreign entry at the region-industry level, which is consistent with Amiti & Javorcik (2008) and Cai et al. (2016).For the sake of robustness, we now proceed to investigate the relationship between input trade liberalization and foreign entry at the firm level, and set the following regression specification: where f, i, j, and t correspond to firm, industry (three-digit Chinese Industrial Classification), region and year, respectively.Y fijt is a firm-level indicator of foreign entry, specifically, we measure foreign entry at the firm level in two ways: one is a dummy variable for whether Chinese firms have received FDI (denoted as ForEntryDum fijt ), and the other is the firm-level foreign share of equity (denoted as ForEntryShare fijt ).We follow Wang and Wang (2015), and add a set of firm-level characteristics (Z ft ) to the regression model: Firm TFP, calculated using the semi-parametric approach proposed by Olley and Pakes (1996); Firm size, measured by the logarithm of firm's total sales; Firm wage, measured by the logarithm of the ratio of firm's payrolls to employment; Firm age, defined as the difference between the current year and the opening year; Firm capital intensity, measured by the logarithm of capital stock over total employment; Firm export status, equals one if a firm exports and zero otherwise; Firm leverage ratio, calculated by the ratio of firm's total liabilities to total assets; Dummy of state/collectively owned, equals 1 if the firm is a state or collectively owned enterprise and 0 otherwise.k f is a full set of firm fixed effects, and the other variables are set as in Equation (1).Table 2 reports the firm level regression results.Column 1 of Table 2 presents the regression results with ForEntryDum fijt (it equals one if a firm receives any FDI and zero otherwise) as the dependent variable.We find that the estimate on the interaction InpTar01ÂPost02 is positive and significant at the 5% level, indicating that after China's WTO accession, the probability of firm receiving FDI increases more in industries experiencing larger input tariff reductions compared with that in industries experiencing less input tariff reductions, that is, input trade liberalization significantly increases the probability of foreign entry.For robustness, we redefine ForEntryDum fijt , specifically, it is taken to be one if the firm receives a foreign capital share above 25% and zero otherwise.As shown in column 2 of Table 2, the estimate on the interaction InpTar01ÂPost02 is significantly positive, once again suggesting that input trade liberalization tends to increase the probability of foreign entry.In column 3 of Table 2, we run Equation (2) using ForEntryShare fijt as the dependent variable.We find that the coefficient on the interaction term InpTar 01 ÂPost02 is significantly positive, implying that input trade liberalization significantly increases firm's foreign share of equity.Overall, the above empirical regressions also suggest that input trade liberalization significantly promotes foreign entry, which is consistent with the previous findings based on the region-industry level regressions.

Other robustness checks.
In this section, we perform robustness checks in the following ways: (1) Alternative measure of foreign entry.First, we follow the standard definition of FDI according to IMF and regard enterprises with foreign capital share above 10% as foreign firms, then we construct a new foreign entry variable; second, we regard enterprises with a foreign capital greater than 0 as foreign firms, and construct another foreign entry variable; third, we use the foreign share of employment in region-industry to describe the degree of foreign entry; in The impact of input trade liberalization on the entry of foreign firms 277 addition, we also adopt foreign share holding and foreign share of total sales in region-industry to measure the degree of foreign entry, respectively.(2) Alternative measure of the regressor of interest.First, we employ the average input tariff rate over 1998-2001 as an alternative measure of the degrees of input trade liberalization from China's WTO accession; second, we replace InpTar 01 ÂPost02 in the specification (1) with yearly input tariff levels (i.e.Inputtariff it ).( 3) Alternative definition of processing-trade firms.First, enterprises whose ratios of processing exports over their total exports higher than 0.9 are regarded as processing trade enterprises; second, we treat pure export enterprises as processing-trade firms.(4) Alternative estimation methods.We run a fixed effects Poisson regression, a random-effect Poisson regression, as well as a Tobit regression, respectively.(5) Industry level estimation.We investigate the nexus between input trade liberalization and foreign entry at the industry level.Online Supplementary Appendix E shows the detailed results of robustness checks.

Potential mechanisms
In this section, we conduct mechanism test to further reveal the internal relationship between input trade liberalization and the entry of foreign firms.
In the previous section, we measure foreign entry by the number of foreign firms at the region-industry level.In fact, in each year, some foreign firms enter while some foreign firms withdraw.Therefore, whether the number of foreign firms increases depends on the relative size of the number of foreign firms entering and exiting during last year, that is, whether the net entry number of foreign firms (denoted as NetEntry) is greater than zero.Following the above logic, we decompose the net entry number of foreign firms at the region-industry level into two parts: the number of new foreign firms entering and the number of old foreign firms existing, that is where i, j, and t correspond to industry, region and year, respectively.ENnum ijt is the number of new foreign firms, measure by the number of new foreign firm entering in industry i in region j at time t; EXnum ijt is the number of old foreign firms exiting, measure by the number of foreign firm withdrawn in industry i in region j at time t.The above decomposition allows us to test the promotion effect of input trade liberalization on foreign entry works more by increasing the number of new foreign firms entering or by reducing the number of existing foreign firms existing.
Column 1 of Table 3 reports the regression results using the net entry number of foreign firms (NetEntry) as the dependent variable.We see that the coefficient on the interaction term, InpTar01ÂPost02, is significantly positive, indicating that input trade liberalization significantly increases the net entry number of foreign firms, which provides further micro-evidence for the above findings that input trade liberalization promotes foreign entry.In order to further shed light on how input trade liberalization affects the net entry of foreign firms, we report the regression results using the number of new foreign firms entering (ENnum) and the number of old foreign firms exiting (EXnum) as the dependent variables in the columns 2 and 3 of the Table 3, respectively.We find that the coefficient on the interaction term, InpTar01ÂPost02, is significantly positive in column 2, while it is significantly negative in column 3, indicating that input trade liberalization promotes the entry of new foreign firms on the one hand, and restrains the exit of existing foreign firms on the other hand.By comparing the coefficient of the interaction term in columns 2 and 3, we can easily conclude that the promotion effect of input trade liberalization on foreign entry works more by increasing the number of new foreign firms entering.
In addition, a great deal of research has empirically confirmed that the reduction of input tariffs is conducive to increasing the varieties and quality of imported input (Amiti & Konings, 2007;Goldberg et al., 2009;Topalova & Khandelwal, 2011), while the availability of diversified and high-quality inputs is an important factor affecting the entry of foreign firms (Amiti & Javorcik, 2008).Hence, we expect both the varieties and quality of imported inputs may be important The impact of input trade liberalization on the entry of foreign firms 279 channels through which input trade liberalization promotes foreign entry.To test this speculation, we merge the ASIF data with the China Customs Trade Database from 2000 to 2007 following Yu (2015), and on this basis, the number of imported input (all goods) varieties at the region-industry level can be calculated (denoted as Variety ijt ); in addition, we measure the quality of imported input(all goods) following Khandelwal, Schott, and Wei (2013), and then average them at the region-industry level to obtain the region-industry imported input quality (denoted as Quality ijt ).Column 4 of Table 3 reports the regression results with the logarithm of the number of imported input varieties (denoted as lnVariety) as the dependent variable.We find that the estimate on the interaction InpTar01ÂPost02 is positive and significant at the 5% level, indicating that after China's WTO accession, the number of imported input varieties increases more in industries experiencing larger input tariff reductions compared with that in industries experiencing less input tariff reductions, that is, input trade liberalization significantly increases the varieties of imported input, which is consistent with the findings of Amiti and Konings (2007) on Indonesia and Goldberg et al. (2009) on India.In order to further confirm whether increasing variety is a potential channel through which input trade liberalization promotes foreign entry, we add an interaction between lnVariety and InpTar 01 ÂPost02 to our basic DID specification.As is shown in column 5 of Table 3, we find that the estimate on the triple interactive term InpTar 01 ÂPost02 Â lnVariety is significantly positive, indicating that for the regions and industries with more varieties of imported input, input trade liberalization has a greater promotion effect on foreign entry, in other words, increasing variety is an important channel via which input trade liberalization promotes foreign entry.We now turn to examine whether imported input quality is an important channel through which input trade liberalization promotes foreign entry.As shown in column 8 of Table 3, the estimate on the interaction InpTar 01 ÂPost02 is significantly positive, suggesting that input trade liberalization significantly raises the quality of imported input, this is strongly in line with the earlier findings.In addition, column 9 of Table 3 further examines the role of imported input quality in the impact of input trade liberalization on foreign entry.The positive and significant coefficient of triple interactive term InpTar 01 ÂPost02ÂQuality implies that imported input quality is a plausible channel via which input trade liberalization promotes foreign entry.What's more, we note that the estimate for InpTar 01 ÂPost02 decreases in absolute value compared with that in our baseline result, which further confirms that imported input quality is an important channel for input trade liberalization to promote foreign entry.For the sake of robustness, we further use the classification of BEC 4 to identify imported intermediate inputs, and then construct region-industry level imported intermediate varieties and imported intermediate quality on this basis.As shown in columns 6-7 and columns 10-11 of Table 3, the increasing variety as well as quality of intermediate input are the important channels through which input trade liberalization promotes foreign entry.
In fact, the reduction of input tariffs not only increases the varieties as well as quality of intermediate input imports, but also significantly reduces the cost of intermediate input imports (Amiti & Konings, 2007;Yu, 2015).Typically, the decline of the import cost of intermediate products will directly reduce the marginal production cost of enterprises, and the industries (or regions) with lower marginal production cost are more conducive to attracting the entry of foreign firms (Kang & Hong, 2007).Accordingly, it is reasonable to anticipate that the decline of marginal cost might be a potential channel through which input trade liberalization promotes foreign entry.Considering that there is no information on marginal cost in the ASIF data, we cannot use this data to calculate the indicator of firms' marginal cost directly.We note that the markup measure contains both price and cost information, therefore, we can use the firm markups to back out firms' marginal cost.To this end, we first calculate firm markups following Lu and Yu (2015), which is expressed as follows: where #m ft denotes the output elasticity on variable input (i.e.intermediate materials), which is estimated following Lu and Yu (2015); c m ft denotes the revenue share of the expenditure on intermediate input by firm f at time t, which is available from the ASIF data.Next, we use the firm identity to merge the product-level data 5 with the ASIF data over the period 2000-2006, and then we select the sample of single-product firms in the merged product-ASIF data, which contains information on output quantity and revenue, and on this basis we can calculate firms' product price (i.e.Price ft ).Finally, the logarithm of firms' marginal cost can be calculated according to the following formula: where ln Price ft denotes the logarithm of firms' product price, d mkp ft denotes the logarithm of firms' markup.After obtaining firms' marginal cost, we simply average it at the region-industry level to obtain the region-industry marginal cost (i.e.MarCost ijt ).Regression results using MarCost as the outcome variable are reported in column 12 of Table 3.We find that the coefficient on the interaction term InpTar 01 ÂPost02 is significantly negative, implying that input trade liberalization significantly reduces the marginal cost, this is understandable given the fact that input tariff reduction helps to reduce the import cost of intermediate inputs, and consequently reduce the marginal production cost of the firms.In order to test whether the reduction in marginal cost is a possible channel through which input trade liberalization promotes foreign entry, we interact MarCost with InpTar 01 ÂPost02, and then introduce the triple interaction into basic DID specification.As shown in column 13 of Table 3, the triple interaction term InpTar 01 ÂPost02 Â MarCost is significantly negative, indicating that for regions and industries with lower marginal cost, input trade liberalization has a greater promotion effect on foreign entry, in other words, reduction in marginal cost is an important channel through which input trade liberalization promotes foreign entry.

Heterogeneous effects
In previous sections, we estimate the average effect of input trade liberalization on foreign entry across regions and industries, and find that input trade liberalization significantly promotes foreign entry.However, regions and industries with different characteristics may respond differently, in addition, different types of foreign firms may also respond differently to input trade liberalization.In this subsection, we investigate the heterogeneous effects of input trade liberalization on the entry of foreign firms, to further shed light on how foreign entry is affected by input trade liberalization.

Regional institutional environment
Since the reform and opening in 1978, China has undergone substantial progress in institutional transition, while the institutional environment is quite imbalance and different among different regions of China.Recently, some researchers pay attention to the role of institutional environment when exploring the nexus between input trade liberalization and firm performance, Ahsan (2013), for instance, shows that in the state with better institutional environment (i.e. the higher judicial efficiency), the stronger positive effect of input tariff reductions on firm productivity will be.An important question that arises is whether the impact of input trade liberalization on foreign entry depends on institutional environment where the firms located?
For the sake of robustness, we measure institutional environment (denoted by InstEnvment) in several ways: First, we use the regional marketization index-which is obtained from the study of Fan, Wang, and Zhu (2010)-to measure institutional environment (denoted by InstEnvment1); Second, we use the index of the development of intermediaries and efficiency The impact of input trade liberalization on the entry of foreign firms 281 improvement of legal system to measure institutional environment (denoted by InstEnvment2), which is also obtained from the study of Fan et al. (2010); Third, we use the efficiency of contract enforcement to measure institutional environment, specifically, it is defined as the inverse of 'the official cost of going through court procedures' (denoted by InstEnvment3) or the inverse of 'the time cost of going through court procedures' (denoted by InstEnvment4) 6 , a higher value of these variables indicate a higher efficiency of contract enforcement and hence the better quality of institutional environment.Next, we introduce interaction terms between the regional institutional environment index, InputTa i01 and Post02 t as well as the triple interaction among these three covariates into our baseline DID specifications, results are reported in columns 1-4 of Table 4.In column 1 of Table 4, we use the regional marketization index as a measure of institutional environment (denoted by InstEnvment1), and find a statistically significant and positive estimate for the triple interaction term, InpTar 01 ÂPost02ÂInstEnvment1, suggesting that in the regions with better institutional environment, the promotive effects of input trade liberalization on foreign entry is stronger compared to those in the regions with weaker institutional environment, in other words, a good institutional environment tends to strengthen the positive impact of input trade liberalization on foreign entry.In addition, we note that both the coefficient and significance of the interaction term InpTar 01 ÂPost02 is now showing a significant decline, indicating that in regions with poor institutional environment, the promotive effects of input trade liberalization on foreign entry is weaker.Columns 2-4 of Table 4 present the regression results using the index of the development of intermediaries and efficiency improvement of legal system, the inverse of 'the official cost of going through court procedures', and the inverse of 'the time cost of going through court procedures' as the measure of institutional environment, respectively.We find that the estimate on the triple interactive term is positive and significant, once again suggesting that the promotive effects of input trade liberalization on foreign entry is increasing as the institutional environment of the region improves.
A possible interpretation for the above findings is as follows.Theoretically, input tariff reductions help to increase access to new imported varieties that were previously unavailable (Klenow & Rodriguez-Clare, 1997) as well as reduce the cost of imported inputs, which are beneficial for foreign entry, however, some imported intermediate inputs require relationshipspecific investment in their production process, especially in the context of contract incompleteness, the foreign suppliers may under-invests in the production of relationship-specific inputs since the buyer may back out at any moment (Ahsan, 2013), this is well known holdup problem.One way to solve this problem is let foreign input sellers and domestic buyers agree on a contract, however, whether such contracts are credible depend on the regional institutional environment.In principle, it is better able to sign the contracts necessary to access these intermediate inputs for the firms in regions with better institutional environment when compared to firms in regions with weaker institutional environment, 7 and hence the firms located in regions with better institutional environment would access more varieties of intermediate inputs as well as import at a lower cost following input tariff reductions.Consequently, after the liberalization of input trade, regions with better institutional environment can obtain a wider range of imported intermediate products, and the import cost of intermediate products will be lower, therefore, these regions will attract foreign-invested enterprises to a greater extent.

Industry import intensity
The previous study found that the mechanisms for the input trade liberalization to promote the entry of foreign firms are that on the one hand, input tariff reduction helps to increase the varieties of imported intermediate goods, on the other hand, it helps to reduce the import cost of intermediate inputs, and then reduce the marginal production cost.Therefore, compared with industries with lower import intensity, industries with higher import intensity benefit more from input trade liberalization.We speculate that the promotion effect of input trade liberalization on The impact of input trade liberalization on the entry of foreign firms 283 foreign entry may be related to the import intensity of the industry.To verify this assumption, we first use the merged sample of China's ASIF data and China's Customs Trade data to construct the industry import intensity index, in particular, industry import intensity (denoted by IMPshare) is measured by the ratio of industry import volume to industry total sales.Next, we introduce interaction terms between IMPshare, InpTar 01 and Post02 as well as the triple interaction among these three covariates into our basic DID specification.It can be seen clearly in column 5 of Table 4 that the estimate on the triple interactive term (InpTar 01 ÂPost02ÂIMPshare) is positive and significant, indicating that the promotive effect of input trade liberalization on foreign entry increases with industry import intensity.Additionally, we note that both the coefficient and significance of the interaction term InpTar 01 ÂPost02 showing a significant decline compared with the baseline results in column 7 of Table 1, suggesting that the promotive effect of input trade liberalization on foreign entry is weaker in the industry with lower import intensity.Overall, the impact of input trade liberalization on the entry of foreign firms is indeed related to the import intensity of the industry, which is consistent with our expectation.

Types of foreign investment
There are different types of foreign investment in China, we can classify foreign investment into two categories according to the source and destination: one is foreign investment from Hong Kong, Macao and Taiwan, and the other is foreign investment from non-Hong Kong, Macao and Taiwan regions (or OECD foreign investment). 8Compared with foreign investment form Hong Kong, Macao and Taiwan, FDI coming from OECD usually has more advanced technologies and know-how, because of this, the proportion of OECD foreign investment can often reflect the quality of foreign investment in a country or region.Thus, an interesting research question arises: whether there is a difference in the impact of the input trade liberalization on the entry of different types of foreign firms.To this end, we first use the ASIF data to construct two variables: one is the logarithm of the number of foreign firms from OECD (denoted by ForEntry OECD ), and the other is the logarithm of the number of foreign firms from Hong Kong, Macao and Taiwan (denoted by ForEntry HMT ); and then run specification (1) with ForEntry OECD and ForEntry HMT as dependent variables.As shown in columns 6 and 7 of Table 4, the estimate on the interaction term (InpTar 01 ÂPost02) is positive and significant in column 6, indicating that input trade liberalization significantly promotes the entry of OECD foreign firms; by contrast, the coefficient on the interaction term is statistically insignificant, suggesting that input trade liberalization does not have any effect on the entry of foreign firms from Hong Kong, Macao and Taiwan.One possible reason for such a difference is that most foreign firms from Hong Kong, Macao and Taiwan are export-oriented and engaged in processing trade, mainly to make use of China's relatively cheap local labor resources (Lin, Liu, & Zhang, 2009), so the entry decision of this type of foreign firms is relatively weak due to the reduction of input tariffs; in contrast, the foreign firms from OECD often have advanced production technology and management experience, and their production and operation activities also have higher complexity, so high quality and diversity of intermediate inputs are particularly important for such firms, therefore, it is not difficult to understand why input trade liberalization plays a relatively stronger role in promoting the entry of foreign firms from OECD.

Conclusion
In this paper, we investigate whether trade policy affects the entry of foreign firms, in particular, we focus predominately on the role of intermediate input trade liberalization since trade in intermediate inputs has become increasingly important in the world economy.We find that input trade liberalization significantly increases foreign entry, and this result is robust to a series of model specifications, and significant both statistically and economically.We also find strong evidence that input trade liberalization not only promotes the entry of new foreign firms, but also restrains the exit of existing foreign firms, thereby contributing to the net growth of the number of foreign firms.In addition, the mechanism tests show that increasing variety as well as quality of intermediate input and reduction in marginal cost are the potential channels through which input trade liberalization promotes foreign entry.
To further shed light on how foreign entry is affected by input trade liberalization, we take the further step to look at the heterogeneous effects of input trade liberalization on the entry of foreign firms.The results show that institutional environment strengths the positive effect of input trade liberalization on foreign entry, and the promotive effect of input trade liberalization on foreign entry increases with industry import intensity, additionally, input trade liberalization significantly promotes the entry of OECD foreign firms, whereas does not have any effect on the entry of foreign firms from Hong Kong, Macao and Taiwan, implying that input trade liberalization is also conducive to improving the quality of foreign investment.
This paper provides empirical evidence that trade policy can significantly affect the entry of foreign firms from the largest developing economy, which is helpful to understand the driving factors of China's attracting foreign investment in recent years.This study enriches the research on the determinants of China's foreign direct investment on the one hand, and it also contributes to the literature that evaluates the economic effects of input trade liberalization on the other hand.Our paper also has rich policy implications.Input trade liberalization not only promotes the entry of foreign firms on the whole, but also helps to improve the quality of foreign investment.Therefore, the Chinese government should continue to implement trade liberalization measures such as reducing tariffs on intermediate products as well as reducing non-tariff barriers, and encourage the firms to actively use tariff relief on intermediate products to import diversified and high-quality intermediate products and parts, this is not only help to directly improve the performance of importers, but also conducive to attracting high-quality foreign firms.

Notes
1.The number of foreign firms increased from 21,500 in 1998 to 58,000 in 2007, with an average annual growth rate of 12.4%; the actual use of foreign investment increased from $45.46 billion in 1998 to $74.77 billion in 2007, with an average annual growth rate of 6.1%.2. Although Amiti and Javorcik (2008) also examine the determinants of foreign entry, they focus mainly on the role of market and supplier access, while using input tariffs as a control variable.In contrast, the focus of this paper is on the impact of input trade liberalization on the entry of foreign firms, using China's WTO accession as a quasi-natural experiment.More importantly, this paper not only conducts empirical research in the regionindustry dimension, but also further examines the impact of input trade liberalization on the likelihood of Chinese firms to receive foreign investment as well as the foreign share of equity at the firm level, and the latter of which has not been considered in the current literature.3. It should be noted that import tariffs for all products (not just intermediate goods) within the industry are used to measure industrial output tariff, then weighted by the importance of that input in production, which in turn yields industry level input tariff.In other words, the 'intermediate' characteristics are portrayed through Input-Output table.Indeed, this is the most commonly used method for measuring industry level input tariff and has been widely used in the literature, e.g.Amiti and Konings (2007), Topalova andKhandelwal (2011), Yu (2015), Liu and Qiu (2016), Defever et al. (2020), among others.4. Similar to Feng, Li, and Swenson (2016), we regard products with BEC codes '111', '121', '21', '22', '31', '322', '42', '53' as intermediate input imports. 5.The product-level data is maintained by the National Bureau of Statistics (NBS) of China for the period [2000][2001][2002][2003][2004][2005][2006], which contains information on each product (defined at the 5-digit product level) produced by the firm.6.They are derived from the World Bank Report entitled 'Doing Business in China 2008'.7.This is mainly because the settlement of legal disputes usually requires a certain amount of time and material costs.The better institutional environment is, the lower the cost paid for it.Therefore, the foreign intermediate suppliers are more willing to cooperate with firms in better institutional environment.8.It is important to note that since the ASIF data does not further classify the capital of non-Hong Kong, Macao and Taiwan (hereafter referred to as non-HMT), we cannot separate the OECD foreign capital from the capital of non-HMT under current data conditions.For this, we approximate foreign investment in non-HMT regions to OECD foreign investment following Du, Harrison, and Jefferson (2011).
The impact of input trade liberalization on the entry of foreign firms 285

Figure 1 .
Figure 1.The number of foreign firms by high-and low-initial input tariff industries.Source: Authors' calculations based on the Annual Survey of Industrial Firms (ASIF) and tariff data.

Table 1 .
Basic results Notes:Robust t-values are in parentheses are corrected for clustering at the industry level.ÃÃÃ , ÃÃ , and Ã indicate statistical significance at the 1, 5, and 10 percent, respectively.Source: Authors collected tariff data and the Annual Survey of Industrial Firms (ASIF).

Table 2 .
Firm level regression Notes: Robust t-values are in parentheses are corrected for clustering at the industry level.ÃÃ indicate statistical significance at the 1, 5, and 10 percent, respectively.Firm level controls include firm TFP, firm size, firm wage, firm age, firm capital intensity, firm export status, firm leverage ratio, dummy of state/collectively owned.Other control variables correspond to the control variables as in column 7 of Table1.Source: Authors collected tariff data and the Annual Survey of Industrial Firms (ASIF).

Table 3 .
Estimates for channels Robust t-values are in parentheses are corrected for clustering at the industry level.ÃÃÃ , ÃÃ , and Ã indicate statistical significance at the 1, 5, and 10 percent, respectively.All regressions include control variables as in column 7 of Table1.The sample period used in columns 1-3 of Table3 is1999-2006, since both the years 1998 and 2007 are used to identify the entry and exit of foreign firms.The sample period used in columns 4-11 of Table 3 is 2000-2007, since both lnVariety and Quality are constructed using the merged ASIF data with the China Customs Trade Database from 2000 to 2007.The sample period used in columns 12-13 of Table 3 is 2000-2006, since MarCost is calculated using the subsample of single-product firms in the merged product-ASIF data from 2000 to 2006.Source: Authors collected tariff data, the Annual Survey of Industrial Firms (ASIF), the product-level data and China Customs Trade Database.

Table 4 .
Heterogeneous effects The estimates for the region level variables (e.g.InstEnvment1-InstEnvment4) are not reported in the table because these variables are absorbed by region-year interaction fixed effects.All regressions include control variables as in column 7 of Table1.Source: Authors collected tariff data, institutional environment data, the Annual Survey of Industrial Firms (ASIF), and China Customs Trade Database.