The impact of tariff reductions on firm dynamics and productivity in China: Does market-oriented transition matter?

The impact of tariff reductions on firm dynamics and productivity in China: Does market-oriented transition matter?

Accepted Manuscript The impact of tariff reductions on firm dynamics and productivity in China: Does market-oriented transition matter? Qilin Mao, Bi...

800KB Sizes 0 Downloads 5 Views

Accepted Manuscript The impact of tariff reductions on firm dynamics and productivity in China: Does market-oriented transition matter?

Qilin Mao, Bin Sheng PII: DOI: Reference:

S1043-951X(17)30102-5 doi: 10.1016/j.chieco.2017.07.011 CHIECO 1083

To appear in:

China Economic Review

Received date: Revised date: Accepted date:

19 October 2016 27 June 2017 24 July 2017

Please cite this article as: Qilin Mao, Bin Sheng , The impact of tariff reductions on firm dynamics and productivity in China: Does market-oriented transition matter?, China Economic Review (2017), doi: 10.1016/j.chieco.2017.07.011

This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

ACCEPTED MANUSCRIPT The Impact of Tariff Reductions on Firm Dynamics and Productivity in China: Does Market-oriented Transition Matter? Qilin Mao

PT

(Institute of International Economics & Collaborative Innovation Center for China Economy, Nankai University, Tianjin, China) e-mail: [email protected]

RI

Bin Sheng*

NU

SC

(Institute of International Economics & Collaborative Innovation Center for China Economy, Nankai University, Tianjin, China) e-mail: [email protected]

MA

Qilin Mao: Institute of International Economics, Collaborative Innovation Center for China China.

Tel.:

+86-13820006204;

E-mail:

D

Economy, Nankai University, Tianjin 30071, [email protected]. No. 94, Weijin Road, Nankai, Tianjin, P.R. China.

Bin Sheng (corresponding author): Institute of International Economics, Collaborative Innovation

CE

PT E

Center for China Economy, Nankai University, Tianjin 30071, China. Tel.: +86-13920055136; E-mail: [email protected]; [email protected]. No. 94, Weijin Road, Nankai, Tianjin, P.R. China.

AC

We would like to thank the editor and three anonymous referees for their very helpful comments and suggestions which substantially improve the paper. Qilin Mao acknowledges financial support from Major Projects of the National Social Science Foundation of China (Grant No. 15ZDA057), National Natural Science Foundation of China (Grant No. 71403135), Humanities and Social Science of Ministry of Education of the People’s Republic of China (Grant No. 16YJC790114), and China Postdoctoral Science Foundation (Grant No. 2017T100056). All remaining errors are our own.

1

ACCEPTED MANUSCRIPT The Impact of Tariff Reductions on Firm Dynamics and Productivity in China: Does Market-oriented Transition Matter?

Abstract:This paper investigates the impact of trade liberalization on firm dynamics and productivity

PT

in the context of dramatic tariff reductions after China’s accession to the WTO, and how this impact varies across regions with different marketization levels. Our results show that (a) on average, output

RI

tariff reductions tend to reduce firm entry rate and increase firm exit rate, while input tariff reductions help to increase both firm entry rate and exit rate, furthermore, regional marketization strengthens the

SC

impact of trade liberalization on firm dynamics; (b) trade liberalization exerts greater impact on the likelihood of exit for the least productive firms while it tends to reduce the probability of exit for the

NU

more productive firms, with regional marketization strengthening such a reallocation process of trade liberalization; (c) firm dynamics effect contributes approximately 43% of the growth of productivity,

MA

and it (especially the firm exit effect) is an important channel through which trade liberalization fosters productivity growth, and domestic market reform is found to strengthen such an impact.

PT E

D

Key Words: Tariff Reductions; Firm Dynamics; Aggregate Productivity; Marketization

AC

CE

JEL Classification: F13, O24, O47, P30

2

ACCEPTED MANUSCRIPT 1. Introduction Firm dynamics (or firm entry and exit) are of paramount importance for developing countries, and particularly for a market oriented economy. As early as in the 1930s, Schumpeter proposed an innovative view of “creative destruction”, which stated that innovation competition would force those plants with low efficiency to exit, while the plants with high productivity survived. Indeed, Chinese

PT

manufacturing firms have experienced a high ratio of entry and exit. During the period of 1998 to 2007, the annual average entry and exit rates of Chinese manufacturing firms are 25% and 15%,

RI

respectively1. During the same period of time, China successfully accessed to the World Trade

SC

Organization (in December 2001) and fulfilled the commitments, which made China enter a new phase of rapid trade liberalization. In particular, China’s average tariff rate decreased from 19.7% in

NU

1998 to 10.3% in 2007. Trade liberalization has long been regarded as an essential policy to achieve the goal of fostering economic development and improving living standards (Robertson, 1940; Lin

MA

and Sim, 2013). Using the Least Developed Countries (LDCs) data, Lin and Sim (2013) demonstrate that trade significantly raises GDP per capita; Lin (2015) and Hu et al. (2016) empirically study the

D

impact of exports on firm total factor productivity (TFP) based on Chinese firm-level data, both of

PT E

them show that exports increase firm TFP causally. However, although there are many studies estimating the productivity effect of trade liberalization, there are relatively few studies focusing on the relationship between trade liberalization and firm dynamics. An important question that arises

CE

against this backdrop is: how does trade liberalization affect Chinese manufacturing firm dynamics (i.e., the entry and exit) after China’s accession to the WTO?

AC

Some previous studies point out that dynamics of firm entry and exit constitute the source and driving force of aggregate productivity as well as economic growth, as the market selection process may involve the entry of plants with higher relative productivity while the exit of plants with lower relative productivity (Bartelsman et al., 2004; Blyde and Iberti, 2012). Thus, the current paper also attempts to quantify the contribution of firm entry and exit to Chinese manufacturing productivity growth and examine whether it is an important channel through which trade liberalization affects Chinese manufacturing productivity. Besides, we note that China’s accession to the WTO is also

1

Calculated based on the Chinese Annual Survey of Industrial Enterprises over the period 1998-2007. 3

ACCEPTED MANUSCRIPT accompanied by the development of domestic market institutions. For example, the marketization index increased from 4.12 in 1998 to 7.50 in 2007.2 Aghion et al. (2008) show that domestic institutions may affect the outcomes of market liberalization, as successful domestic market reforms are usually complemented with other supporting policies that would facilitate the reallocation of resource towards more efficient uses (Cheng, 2015). Inspired by these earlier works, this paper takes a step further to investigate how the impact of trade liberalization on firm dynamics and productivity

PT

varies across regions with different marketization levels.3

RI

The current paper is related to at least three strands of the growth literature. The first related

SC

strand of the literature is on the determinants of firm dynamics. For instance, Lay (2003) investigates the impact of firm-specific factors on the entry and exit rate using Taiwan’s manufacturing sector

NU

level data, and finds that the entry or exit sunk costs introduced the instantaneous movement of entry and exit rates. Using the industrial sectors level data in Sweden over the period 1997-2001, Nyström

MA

(2007) first explores the pattern of entry and exit, and then investigates the importance of profitability, industrial market growth, and economies of scale for entry and exit. In addition, instead of focusing on sector level like the above studies, Bernard and Jenson (2002) conduct research on the exit of the

D

U.S. manufacturing firms by using firm-level data from 1977 to 1997 and reveal that the competition

PT E

from low-income countries significantly induces firms to exit. In particular, the small, old and labor-intensive firms face a higher probability of exiting. Gu et al. (2003) examine the impact of tariff

CE

reductions on firm entry and exit during the post period of CUFTA and show that the tariff reductions increase entry and exit rate of Canadian manufacturing firms by significantly enhancing exit rate but

AC

has a smaller effect on entry rate. Baggs et al. (2009) study the connection between exchange rate movement and firm survival based on Canadian firm-level data from 1986 to 1997, they demonstrate that the appreciations in the Canadian dollar exert negative impacts on firm survival and sales. More recently, Blyde and Iberti (2012) use the plant-level data from Chile to examine the relationship between trade cost and resource reallocation, the results show that trade costs tend to protect inefficient producers, lower the likelihood of exiting, deter the expansion of efficient plants and slow down the export growth of current exporters, and thus affect the reallocation process across plants with different productivity levels. Although there are rich studies focusing on the determinants of firm 2 3

See Fan et al. (2010) for more details. It is worth noting that “region” corresponds to “province” in this paper. 4

ACCEPTED MANUSCRIPT dynamics, however, there are relatively few studies empirically investigate the impact of trade liberalization on firm dynamics. Our work advances the literature on trade liberalization and firm dynamics by introducing firm productivity heterogeneity and examining how the impact of trade liberalization on firm dynamics varies across regions with different marketization levels using detailed firm-level data from China.

PT

Our study is also related to the literature on the nexus between trade liberalization and firm productivity. Schor (2004) analyzes Brazilian manufacturing firms from 1986 to 1998, and finds that

RI

input tariff reductions significantly increase firm productivity. Using firm-level data on Indonesia

SC

from 1991 to 2001, Amiti and Konings (2007) show that firm productivity gains from reduction of input tariffs are at least twice as much as those from reduction of output tariffs. They argue that the

NU

primary reason for this result is that access to better intermediate inputs through the reduction of input tariffs is more important than the pro-competitive effect arising from lower output tariffs. Topalova

MA

and Khandelwal (2011) empirically investigate the impact of changes in tariffs on firm productivity using Indian firm-level data, and also demonstrate that input tariff liberalization has a stronger positive impact on firm productivity than that of output tariff liberalization. In contrast, Hu and Liu

D

(2014) show that output tariff reductions lead to a decrease in productivity for Chinese manufacturing

PT E

firms, while input tariff reductions exert positive impact on firm productivity. They further argue that both effects have diminished in magnitude over the years after China joined the WTO. More recently,

CE

Yu (2015) studies the connection between tariff reductions and firm productivity, based on Chinese firm-level data from 2002 to 2006. He finds that both the input and output tariff reductions increase

AC

firm productivity, and the effects are weaker as firms’ share of processing imports grows. Our paper is distinguished from the existing literature in that we investigate how trade liberalization affects manufacturing aggregate productivity growth rather than firm productivity, in particular, our work provides a new perspective by exploring the complementarities between trade liberalization and domestic market reform. Besides, some studies have focused on the heterogeneous repercussions of trade liberalization depending on firm ownership, internal economic geography and/or institutions (Wakasugi and Zhang, 2015; Keller et al., 2016; Ramondo et al., 2016; Aghion et al., 2015). In particular, by using the firm-level panel data of Chinese electric machinery, electronics equipment, and telecommunications 5

ACCEPTED MANUSCRIPT equipment industries, Wakasugi and Zhang (2015) find that China’s accession to the WTO had different effects on export decisions among foreign-invested enterprises, private domestic firms, and SOEs. Keller et al. (2016) use a new dataset on trade between fifteen Chinese treaty ports to study the role of domestic frictions in shaping the welfare gains of trade, and demonstrate that trade frictions which determined by China’s internal economic geography are found to be prominent in the

PT

distribution of the welfare gains of trade. Similarly, Ramondo et al. (2016) stress the importance of domestic frictions in reconciling the data and the trade theory, they show that domestic frictions can

RI

partially offset the scale effects present in the trade model, and hence lead to a better match with the data. Further, Aghion et al. (2015) exploit Chinese firm-level data over the period 1998-2007 to

SC

investigate the interaction between industrial policy and domestic market competition, and they find that there is a complementarity between product market competition and industrial policy aimed at

NU

fostering innovation and growth, and more interestingly, the imposition of tariffs in more competitive

MA

sectors is more likely to result in higher firm performance.

Among others, there are two papers which are closely related to our empirical study. One is Khandelwal et al. (2013) that studies productivity growth of Chinese exporters after removing

D

externally imposed quotas. They find that the elimination of quota misallocation contributes to 71

PT E

percent of the overall growth of productivity, while the removal of quota itself contributes to the 29 percent. Our study differs from the study of Khandelwal et al. (2013) in two important aspects. First,

CE

Khandelwal et al. (2013) focus predominately on a specific form of trade liberalization in China, e.g. the removal of export quotas on Chinese textile and clothing to the United States, European Union, and Canada, while we study the impact of both output and input tariff reductions in all manufacturing

AC

sectors due to China’s accession the WTO. Second, rather than conducting numerical solutions to uncover the role of domestic institutional reform, we directly introduce the interaction terms between regional marketization index and tariffs into our econometric model. Another study is Brandt et al. (2012), which examines the impact of tariff reductions on the performance of Chinese manufacturing industry at both firm and sector level. They find that at the firm level, both output tariff reductions and reduced protection of the domestic industry lead to higher firm productivity, and at the sector level, the sectors that liberalized most tend to attract especially productive entrants, private firms in particular, and thus raise aggregate productivity. Our paper is 6

ACCEPTED MANUSCRIPT distinguished from Brandt et al. (2012) in the following four grounds. First, instead of focusing on the impact of tariff reductions on firm productivity, this paper focuses predominately on the effect of tariff reductions on firm dynamics, and explores whether firm dynamics is an important way through which trade liberalization fosters aggregate productivity growth. Second, we directly examine the heterogeneous effects of trade liberalization on firm dynamics among firms with different productivity,

PT

which offers a fuller and more nuanced picture of the potential reallocation effect of trade liberalization. Third, this paper adopts both instrumental variable (IV) approach

and

RI

difference-in-differences (DID) technique to address the endogeneity of tariffs. Last but not least, we study the heterogeneous impact of trade liberalization on firm dynamics in different regions by

SC

considering the level of marketization. Specifically, using the firm-level production data obtained from the National Bureau of Statistics (NBS) of China over the period 1998-2007, the paper reaches

NU

the following findings. First, we find that on average, output tariff reductions tend to reduce firm entry rate and increase firm exit rate, while input tariff reductions help to increase both firm entry rate and

MA

exit rate, and regional marketization strengthens the impact of trade liberalization on firm dynamics. Second, trade liberalization has a greater impact on the likelihood of exit for the least productive firms

D

while it tends to reduce the probability exit for more productive firms, and such reallocation process

PT E

of trade liberalization is increasing significantly as the marketization level of the regions increases. Third, firm dynamics effect contributes approximately 43% of the growth of aggregate productivity, and it (especially the firm exit effect) is an important channel through which trade liberalization

CE

fosters aggregate productivity growth, and market reform helps to strengthen such an impact.

AC

The remainder of this paper is organized as follows. Section 2 describes the institutional background. Section 3 introduces our estimation strategy and the data. Section 4 presents and discusses our basic empirical results. Section 5 further investigates the nexus between trade liberalization, firm dynamics and productivity evolution. Section 6 concludes.

2. Institutional background China accession to WTO.—The Chinese government began its economic reform and adopted opening policy in 1978. Before 1978, however, Chinese economy was very closed and almost insulated from the world economy (Naughton, 2006). In order to open up its economy to the rest of 7

ACCEPTED MANUSCRIPT the world, Chinese government adopted a series of trade liberalization policies. Specially, Chinese government set up four special economic zones (i.e., Shenzhen, Xiamen, Zhuhai, and Shantou) in the early 1980s, and then opened up the economic and technical development zones (ETDZs) in fourteen coastal cities in the early 1990s. Besides, Chinese government also started to apply for WTO accession in the late 1980s to fully integrate to the world economy and proceed with its trade

PT

liberalization. In fact, Chinese government started negotiations to join the General Agreement on Trade and Tariffs (GATT) contracting party in 1986; then it experienced several rounds of bilateral

RI

negotiations between China and WTO members, and finally in December 2001, China became a

SC

member of the WTO.

In reality, China embarked on widespread unilaterally tariff reduction during the negotiation

NU

period of its WTO accession. Specifically, China’s average nominal tariff decreased from 43% in 1992 to 17% in 1997. After accession to the WTO, China committed to a broad range of reforms to

MA

open up its economy, including fulfilling its tariff reduction responsibilities. As shown in Fig.1, the average output tariff rate decreased from 19.7% in 1998 to 10.3% in 2007, and the average input tariff rate decreased from 10.8% in 1998 to 5.9% in 2007. For both output and input tariff rates, the most

D

substantial reductions took place in 2002, with the tariff rates dropping by 22.3% and 24.3%,

PT E

respectively. Furthermore, the standard deviation of both output and input tariff rates decreased steadily, reflecting greater uniformity of tariff rates across industries.

CE

[Fig.1 inserted here]

Development of domestic market institutions.—In addition to the “open-door” policy, the

AC

“deepened economic reform” policy is another fundamental doctrine of the Chinese government after 1978, which means that a centrally-planned economy is gradually transformed into a market-oriented one. The share of private entities was very small in the beginning of China’s economic transition, while state-owned entities accounted for large component of the economy. During the period 1980-1983, China’s economic reform focused predominately on transforming state-owned entities from administrative units to independent economic identities. And in 1984, Chinese government introduced dual track prices for state-owned entities, allowing them to go beyond the state planning system and trade their products in the market at prices within a 20 percent variation around the planned prices (Wen, 2007). Further, the “dual track” was replaced by market prices and most 8

ACCEPTED MANUSCRIPT intermediate goods opened to market competition since 1992. As a result, China’s marketization level and institutional quality has been systematically improving across all regions. The marketization and institutional development not only promote the development of the private sector, but also lead to a reduction in the level of government involvement in enterprise operations and therefore improve the institutional quality of local governments.

PT

More importantly, China’s accession to the WTO is also accompanied by the development of domestic market institutions. The marketization index increased from 4.12 in 1998 to 7.50 in 2007.4

RI

However, the marketization level and institutional environment vary significantly across regions in

SC

China. Fig.2 shows that there is a strong positive correlation between the marketization index in 2007 and the initial marketization index in 1998. Meanwhile, the East China had a much higher

NU

marketization level than Central China, while the level of West China is the lowest. It also indicates that the marketization index increased in almost all regions between 1998 and 2007, and Shanghai,

MA

Beijing and Zhejiang are the regions with the largest increase, while Hainan, Guangxi, and Gansu are those with the smallest growth.

[Fig.2 inserted here]

PT E

D

In this paper, we exploit the dramatic tariff reductions arising from China’s accession to the WTO to investigate the impact of trade liberalization on firm dynamics and productivity. Especially, we attempt to disentangle the impact of lower output tariffs from that of lower tariffs of intermediate

CE

inputs. It is important to make such a distinction, as the former leads to an increase in domestic competition while the latter enables firms to access more and cheaper imported inputs. Recently, some

AC

researchers also pay attention to the role of institutional environment when exploring the nexus between trade liberalization and firm performance, for instance, Ahsan (2013) shows that, the positive impact of input tariff reductions on firm productivity is normally stronger in those state with better institutional environment (i.e., the higher judicial efficiency). An important question that arises is whether the impact of trade liberalization on firm dynamics depends on institutional environment where the firms are located? Fortunately, the marketization level and institutional environment differ significantly across regions in China, which provides an excellent setting to answer this question.

4

See section 3.1 for details on the marketization index. 9

ACCEPTED MANUSCRIPT 3. Empirical strategy and data 3.1. Econometric model One of the main goals of this paper is to study the response of firm dynamics to trade liberalization. It should be pointed out that it is difficult to examine firm entry at the firm level due to

PT

the fact that we can only observe the firms that did enter but can hardly observe those that didn’t do so (Baggs et al., 2014). This means we fail to investigate the impact of trade liberalization on the

RI

probability of entry at the firm level. Instead, we conduct empirical analysis at the region-industry level using entry rate as the dependent variable. For symmetry, we can also analyze firm exit at the

SC

region-industry level using exit rate as the dependent variable. In particular, we consider the following

NU

regression:

Yjkt  0  1OutputTariff jt 1  2 InputTariff jt 1  X jkt 1   j   kt   t   jkt

(1)

MA

where j, k, and t denote industry (4-digit Chinese Industrial Classification level), region and year, respectively. Y jkt denotes firm dynamics (i.e., entry rate or exit rate depending on the specification),

PT E

D

We define entry rate (denoted by EenRate jkt ) as the number of new entering firms in industry j in region k at time t over the number of incumbent firms in industry j in region k at time t following Baggs et al. (2014), likewise, exit rate (denoted by ExRate jkt ) is measured by the ratio of the number

CE

of firms exit in industry j in region k between time t and t+1 to the number of incumbent firms in

AC

industry j in region k at time t. OutputTariff jt 1 and InputTariff jt 1 are the industry specific output tariffs and input tariffs, respectively; we use the one-year lagged term to control for potential simultaneity and reverse causality. In order to isolate the effect of trade liberalization, we control for several time varying industry-region specific characteristics ( X jkt 1 ) that may affect firm dynamics. Again, all these control variables are lagged in order to mitigate potential simultaneity and reverse causality. The control variables include: average sales growth rate (measured by the average growth of total sales from year t-1 to t and deflated using PPI), average profit rate (measured by the ratio of total profit to total sales value at the industry-region level), average size (defined by the logarithm of ratio of employees to the number of firms at the industry-region level), average capital intensity 10

ACCEPTED MANUSCRIPT (measured by the logarithm of the ratio of fixed assets to employees at the industry-region level), and herfindahl index (produced by the sum of each firm’s squared market share, for each industry-region pair and year t, with a lower value indicating a higher degree of competition).  j is the industry fixed effects, used to control for all time-invariant differences across industries;  kt is the region-year interaction fixed effects, controlling for time varying, unobservable region characteristics;

PT

 t denotes the year fixed effects.  jkt is the error term. We cluster standard errors at the

RI

industry-region level.

SC

In addition, we are also interested in whether the impact of trade liberalization on firm dynamics depends on regional marketization. To investigate the differential impacts of trade liberalization

NU

across regions, we interact the output tariffs and the input tariffs with the regional marketization index, and add them to our basic specification (1), and get the following empirical specification:

MA

Y jkt  0  1OutputTariff jt 1   2 InputTariff jt 1  3OutputTariff jt 1  Marketkt 1   4 InputTariff jt 1  Marketkt 1  X jkt 1   j   kt   t   jkt

(2)

D

where Marketkt 1 denotes the regional marketization index (or NERI index), which is obtained from the study of Fan et al. (2010), which includes annual indices from 1997 to 2007 on the provincial

PT E

basis. It is a weighted average of five sub-category indices,5 including the reduction of government intervention6, the development of non-state-owned enterprises7, the development of product market8, the development of factor market9, as well as the development of intermediaries and efficiency

CE

improvement of legal system10. According to Fan et al. (2010), each sub-category indice of NERI is a relative index which measures the relative position of a region compared to other regions, that is

AC

computed as ( xit  x0min ) ( x0max  x0min ) , where xit is the absolute value of index for region i in year t, x0min and x0max are the minimum and maximum value of index for a region in base year, respectively.

In order to make the NERI index comparable in different years, Fan et al. (2010) use 2001 as the base 5

It is important to note that the weight is calculated using arithmetic mean method in Fan et al. (2010), the advantage of this method is that it could ensure the comparability of NERI index in different years. 6 It includes the proportion of resource allocation by market, the reduction of extra-financial burden on farmers, the reduction of government interference, the reduction of extra-financial burden on enterprises, and the shrinking the size of government. 7 It includes the share of non-state sectors in GDP, the share of non-state sectors in total fixed investment, and the share of non-state sectors in urban employment. 8 It includes the degree of price controlled by the market, and the reduction of local protection in commodity markets. 9 It includes the structure of banking sector, the allocation of financial resource in state vs. non-state sectors, the environment for foreign direct investment, labor mobility, and the marketization of technological achievements. 10 It includes the development of intermediate institutions, the protection of the lawful rights and interests of producers, the intellectual property protection, and the protection of the rights and interests of consumers. 11

ACCEPTED MANUSCRIPT year and adjust the annual province level indices over the period 1997-2007. Therefore, the NERI index used in this paper reflects not only the relative position of marketization progress for different regions, but also the variation of marketization level for a certain region over time (Fan et al., 2010; Fan et al., 2011).11 According to Wu et al. (2013), the regional marketization index can capture the institutional progress in the transition period and therefore reflects the quality of China’s market-oriented reforms. In Eq. (2), the two interaction terms, OutputTariff jt 1  Marketkt 1 and InputTariff jt 1  Marketkt 1 , are our major interest. A negative and statistically significant  3 is an

PT

indication of complementarities between output trade liberalization and regional marketization, which means that the positive effect of output trade liberalization on firm dynamics is stronger in regions

RI

with higher marketization level than in regions with lower marketization level. Similarly, if  4 is

SC

negative and significant, then it reflects that there exists complementarities between input trade liberalization and regional marketization. The positive effect of input trade liberalization on firm dynamics is stronger in regions with higher marketization level than in regions with lower

NU

marketization level.

MA

For firm exit, we can also conduct empirical analysis at the firm level using firm exit dummy (denoted by exitijkt ) as the dependent variable. The impact of trade liberalization on firm exit is

D

estimated by the following probit model:

(3)

PT E

Pr(exitijkt )  (0  1OutputTariff jt 1  2 InputTariff jt 1  X ijk ,t 1   j   kt  t   ijkt )

where exitijkt is the firm exit dummy, which takes the value one if firm i exit between the periods t

CE

and t+1, and zero otherwise. ( ) denotes the standard normal cumulative density function. X ijk ,t 1

AC

is a vector of firm characteristics that may affect firm exit, such as firm sales growth rate (measured by the growth of firm total sales from year t-1 to t and deflated using PPI), firm profit rate (measured by the ratio of firm profit to its total sales value), firm size (measured by the logarithm of the number of firm’s employees), firm capital intensity (measured by the logarithm of the ratio of firm’s fixed assets to its employees). In addition, we also include herfindahl index to control for the impact of market structure on firm exit. Indeed, one important advantage of the firm level probit estimation is that it allows us to investigate the heterogeneous impacts of trade liberalization on the exit behavior across firms with 11

For this reason, the marketization index computed by Fan et al. (2010) is widely used in panel data empirical studies, such as Li et al. (2010), Wu et al. (2013), Ding et al. (2016), etc. 12

ACCEPTED MANUSCRIPT different productivity.12 To this end, we introduce interaction terms between the tariffs and the firm relative productivity into Eq. (3), and get the following extended empirical model: Pr(exitijkt )  ( 0  1OutputTariff jt 1   2 InputTariff jt 1  3OutputTariff jt 1  RelTFPijk ,t 1   4 InputTariff jt 1  RelTFPijk ,t 1  X ijk ,t 1   j   kt   t   ijkt )

(4)

PT

where RelTFPijk ,t 1 is a firm’s relative productivity, which is defined as the productivity of firm i relative to its peers in industry j in region k. More formally, it can be expressed as

RI

RELTFPijkt  tfpijkt tfp jkt , where tfpijkt is firm i’s productivity in year t,13 and tfp jkt denotes the

SC

average productivity for each industry-region pair in year t. The variables of interest are the two interaction terms, OutputTariff jt 1  RelTFPijk ,t 1 and InputTariff jt 1  RelTFPijk ,t 1 . In particular, if

NU

 3 (or  4 ) is positive and significant, then it indicates that the probability of firm exit upon output (or input) trade liberalization is relatively lower for more productive firms. For simplicity, we denote

MA

this result as the reallocation process of trade liberalization following Blyde and Iberit (2012). For the sake of completeness, we further introduce interaction terms between the tariffs, the firm

D

relative productivity and the regional marketization index as well as the triple interaction between

PT E

these covariates into the firm exit probit model, which enables us to explore whether the reallocation process of trade liberalization depends on regional marketization level. In particular, we consider the

CE

following specification:

Pr(exitijkt )  (  0  1OutputTariff jt 1   2 InputTariff jt 1  3OutputTariff jt 1  RelTFPijk ,t 1   4 InputTariff jt 1  RelTFPijk ,t 1  5OutputTariff jt 1  RelTFPijk ,t 1  Marketkt 1

AC

  6 InputTariff jt 1  RelTFPijk ,t 1  Marketkt 1   7OutputTariff jt 1  Marketkt 1

(5)

 8 InputTariff jt 1  Marketkt 1  X ijk ,t 1   j   kt   t   ijkt )

12

It is important to note that firm heterogeneity has been widely studied in the trade literature, most of which show that exporters are more productive than non-exporters. However, Lu (2010) finds that the opposite result for China, namely, exporters are less productive when compared to non-exporters, especially in labor intensive sectors. Further, Dai et al. (2016) demonstrate that processing exporters are less productive than non-processing exporters as well as non-exporters, and importantly, the abnormality in exporter performance in China would be disappeared once the processing exporters are controlled. Here, we instead focus on whether firm heterogeneity plays an important role in how the tariff reductions have affected Chinese firm dynamics. 13 In this paper, we measure firm productivity using the semi-parametric approach developed by Olley and Pakes (1996), the essence of this approach is to use firm investments as proxies for unobserved firm specific productivity shocks. Details of construction of TFP using Olley-Pakes method are provided in Appendix A, and Appendix Table A1 presents the estimated coefficients of the production function by industry. Besides, we also adopt the approach proposed by Levinsohn and Petrin (2003)—which uses the intermediate inputs as proxies for productivity shock—to calculate firm productivity, and find that the main estimation results are robust. 13

ACCEPTED MANUSCRIPT OutputTariff jt 1  RelTFPijk ,t 1  Marketkt 1

In Eq. (5), the triple interaction terms

and

InputTariff jt 1  RelTFPijk ,t 1  Marketkt 1 are our major interest, which capture the differential

reallocation process of trade liberalization across regions with different marketization levels.

3.2. Measure of tariff

PT

The earlier studies predominately used import penetration ratio as a proxy for trade liberalization (see, e.g., Harrison, 1994; Beyer, Rojas and Vergara, 1999). However, many researchers have pointed

RI

out that this indicator couldn’t reflect the degree of trade liberalization precisely, especially for those

SC

countries which are experiencing radical trade reforms14 (Amiti and Konings, 2007). To take this into account, we resort to other indicators to reflect the true changes of specific trade policy instruments.

NU

Inspired by Amiti and Konings (2007), we construct both the output tariffs and input tariffs to measure trade liberalization, this allows us to distinguish the pro-competitive effect of output tariff

MA

reductions from the intermediate inputs channel of input tariff reductions. We first construct output tariffs as follows:

 

p j

n pt   pt

p j

n pt

(6)

PT E

D

OutputTariff jt 

where j, p, and t denote the industry (4-digit CIC level), Harmonised System (HS) 6-digit product, and year, respectively.  j denotes a product set of sector j, n pt is the total number of tariff lines of

CE

product p,  pt denotes the tariff rate of HS6 product p in year t. In other words, our industry specific

AC

output tariff is the simple average of the tariffs in the HS 6-digit codes within each 4-digit CIC industry code.

We then construct the industry specific input tariffs following Amiti and Konings (2007), which uses an input cost-weighted average of output tariffs: InputTariff jt   w jw  OutputTariff wt

14

(7)

Harrison (1994) pointed out this drawback. For example, the trade reform of Cote d’Ivoire in 1985 reduced average tariff

by 30%, but the import penetration ratio maintained unchanged. 14

ACCEPTED MANUSCRIPT where InputTariff jt denotes the industry specific input tariffs in year t; OutputTariff wt is the tariff on industry w in year t, and  jw is the cost share of input w in the production of a good in industry j, we adopt Chinese Input-Output Table for year 200215 to calculate the weight  jw . Specially, we construct the industry specific output and input tariffs in four steps. First, we make a concordance

PT

between the Input-Output (IO) code and the CIC 4-digit code. Second, we match the CIC 4-digit code with the International Standard Industrial Classification (ISIC hereafter). Third, we link the HS 6-digit

RI

product code and the ISIC to find the corresponding tariffs from the WTO and the trade analysis and information system (TRAINS). Finally, we calculate the industry specific tariffs which are aggregated

SC

to the 4-digit CIC industry level.

NU

3.3. Data

This paper mainly relies on the following two disaggregated data sets: tariff data and firm-level

MA

production data. The tariff data of China is downloaded from the WTO website, which provides detailed information on the number of tariff lines, and average, minimum and maximum ad valorem

D

tariff duties for each product defined at HS 6-digit level. Since tariff information on the WTO website

PT E

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

CE

codes and the 2007 HS codes to the 2002 HS code using the standard HS concordance table. Consistent with Yu (2015), we use the average ad valorem duties at the 6-digit level to measure trade

AC

liberalization. Our second data source is the Annual Survey of Industrial Enterprises (ASIE) over the period 1998-2007, which is maintained by the NBS of China. The data set covers all state-owned firms and private firms with sales greater than RMB 5 million. This dataset provides detailed information on each firm, including complete information on the three major accounting statements (i.e., balance sheets, loss and benefit sheet, and cash flow statements), and covers all the required variables used in the current study, such as employment, wage, sales, capital, fixed assets, value added, and so on. We follow Brandt et al. (2012) to link firms over time using the originally assigned ID and

15

It is worth noting that we use the year 2002 because the input-output table is only available for every five-years and 2002 is the middle year of our sample period. 15

ACCEPTED MANUSCRIPT other additional information such as name, industry, addresses, etc. By doing so, we can account for the changes in ID due to changes in ownership, merger and acquisition, or restructuring. Since NBS issued a new Chinese Industrial Classification (CIC) system in 2002, we employ the concordance developed by Brandt et al. (2012) to achieve consistency in the industry codes before and after 2002. Similar to the existing literature, we focus on firms in the manufacturing sector. We drop firms in

PT

mining, tobacco, and public utility industries according to the CIC System. However, some samples are still noisy and are therefore misleading, we thus further follow Feenstra et al. (2014) to clean this

RI

dataset by deleting observations according to the basic rules of the Generally Accepted Accounting

SC

Principles (GAAP). To be concrete, we drop a firm from the data if any of the following is observed: (1) liquid assets are greater than total assets, (2) total fixed assets are greater than total assets, (3) the

NU

net value of fixed assets is greater than total assets, and (4) the firm’s identification number is missing. Besides, similar to Brandt et al. (2012), we also discard firms with less than eight workers since they

MA

may fall under a different legal regime.

4. Empirical analysis

PT E

D

4.1. Baseline results16

Columns (1)-(3) of Table 1 present the baseline results regarding the responsiveness of firm entry rate to trade liberalization.17 In column (1), we regress firm entry rate on only output tariff and input

CE

tariff, as a benchmark. We find that the coefficient on output tariff is positive and significant at the 1% level, indicating that output tariff reductions discourage firm entry, while the input tariff term is

AC

negative and statistically significant, suggesting that input tariff liberalization tends to increase firm entry rate. In column (2), we include a set of time varying industry-region characteristics (e.g., average sales growth rate, average profit rate, average size, etc.) that may influence firm entry rate. Evidently, our results on trade liberalization affecting firm entry rate are insensitive to those time 16

In the baseline regression, the industry is defined at the 4-digit CIC level, we also conduct a robustness check at the 2-digit CIC industry level (accordingly, there are fewer observations within each industry-region cell). The regression results are reported in Table A2. Evidently, our main results are robust to this more aggregate industry definition. 17 It is important to stress that most of the existing literature (e.g., Amiti and Konings, 2007; Bustos, 2011; Topalova and Khandelwal, 2011; Bas, 2012; Hu and Liu, 2014; Yu, 2015; Fan et al., 2015; Tian and Yu, 2015; Lu and Yu, 2015; Ludema and Yu, 2016; Fan et al., 2017; Brandt and Morrow, 2017) on trade liberalization and firm performance have focused predominately on tariffs reductions, and didn’t take non-tariff barriers into consideration due to data restrictions. As a robustness check, we however control for non-tariff barriers by including a dummy (NTBsDum, which equals to 1 if the industry has imposed any non-tariff barriers for at least one 8-digit HS product) in the regressions, following Cheng (2012). As shown in Table A3, we find that this additional control is statistically insignificant but our regressors of interest barely change in their significance and magnitude. These results indicate that our findings are not driven by non-tariff barriers. 16

ACCEPTED MANUSCRIPT varying industry-region characteristics. Furthermore, we add industry fixed effects, region-year fixed effects as well as year fixed effects to the baseline specification, as shown in column (3), our results are found to be robust to these additional controls. Specially, column (3) of Table 1 shows that the coefficient on output tariff is 0.0004 and statistically significant, indicating that a fall in output tariffs of 10 percentage points leads to a decrease in entry rate by 0.4 percent. A possible interpretation for

PT

this finding is that output tariff reductions lead to an increase in competition, which in turn discourages firm entry. In contrast, the coefficient on input tariff is statistically significant and

RI

negative. The point estimate has a higher absolute magnitude than that of output tariff, implying that a 10 percentage point fall in input tariffs leads to an increase in entry rate by 2.8 percent. The reasoning

SC

is that lower input tariffs tend to reduce price and increase the variety of intermediate inputs available for firms, and that industry profits increase due to the lower cost of production, which in turn attracts

NU

new entrants. Clearly, our results show that the impact of input trade liberalization on firm entry rate

MA

is much larger than that of output trade liberalization.

[Table 1 inserted here]

We now turn to analyze the effect of trade liberalization on firm exit rate. Column (1) of Table 2

D

presents the estimation results regarding the impact of trade liberalization on firm exit rate without

PT E

controlling for other time varying industry-region characteristics. The negative and significant coefficient of the output tariff suggests that the pro-competitive effect arising from output tariff reductions leads to an increase in exit rate; somewhat surprisingly, the coefficient on input tariff is

CE

statistically negative, indicating that input trade liberalization also increases exit rate on average. For the sake of robustness, we include time varying industry-region characteristics in column (2) of Table

AC

2, while in column (3), we further add industry fixed effects, region-year fixed effects and year fixed effects. All estimations show that both output tariff reductions and input tariff reductions lead to an increase in exit rate. Once again, the impact of input tariff reductions on firm exit rate is much higher than that of output tariff reductions. The result that output tariff reductions increase firm exit rate is perhaps not surprising given that lower output tariffs lead to more severe competition and hence force some low efficient firms to exit. Somewhat surprisingly, input tariff reductions are associated with higher exit rates. A possible interpretation for this finding is that the responses to lower input tariffs vary across firms with different efficiency. For instance, following input tariff reductions, firms can 17

ACCEPTED MANUSCRIPT import intermediate goods at a lower price and reduce their costs of production, which is followed by a decrease in the price of the final goods and an intensification of competition in domestic market. Faced with tougher domestic competition, some less productive firms might exit. [Table 2 inserted here]

PT

4.2. Heterogeneous impacts across regions with different marketization levels The aforementioned analyses investigate the average impact of trade liberalization on firm entry

RI

rate and exit rate. In this subsection, we intend to explore the heterogeneous impacts of trade liberalization on firm dynamics across regions with different marketization levels. Columns (4)-(7) of

SC

Table 1 present the heterogeneous impacts of trade liberalization on firm entry rate. In column (4), we focus only on the interactions between the output trade liberalization and the regional marketization.

NU

We find that the coefficient of interest on the interaction term between the output tariff and the marketization index is positive and significant at the 5% level, and the coefficient on output tariff is

MA

statistically positive but with a smaller magnitude compared to the baseline results reported in column (3) of Table 1. This provides strong evidence of complementarities between output trade liberalization

D

and regional marketization. To be more precise, output tariff reductions tend to decrease firm entry

PT E

rate, and this impact is stronger in regions with higher marketization level. A plausible explanation is that regional marketization and thus institutional environment is negatively associated with the degree of government intervention and development of the state-owned enterprises.18 In particular, firm

CE

entry is determined by a mix of political selection and market selection in the regions with lower marketization level, and therefore, the competition effect arising from lower output tariffs on firm

AC

entry is weaker in these regions. In contrast, the selection rule would become more market-based in the regions with higher marketization level, and hence output tariff reductions could exert a stronger effect on firm entry.

In column (5), we pay attention to the interaction between the input trade liberalization and the regional marketization. The results show that the coefficient on input tariff is negative but losing its significance, while the coefficient on the interaction term between the input tariff and the marketization index is statistically negative, indicating that in regions with lowest marketization level, 18

In section 4.7, we will explore the role of subcomponents of the marketization index in shaping the effects of tariff reductions on firm dynamics. 18

ACCEPTED MANUSCRIPT input trade liberalization does not increase firm entry rate, but it has a positive and stronger impact on firm entry rate as the marketization level of the regions increases. We provide one explanation as follows. Theoretically, input tariff reductions will be conductive to increase access to new varieties that were previously unavailable for importers (Grossman and Helpman, 1991; Klenow and Rodriguez-Clare, 1997), and reduce the cost of imported inputs, which tend to attract more firms to

PT

enter. However, some imported intermediate inputs require relationship-specific investment in their production process. In the context of contract incompleteness, foreign suppliers may under-invest in

RI

the production of relationship-specific inputs since buyers may back out at any moment (Ahsan, 2013). This is the well-known issue of “holdup problem” (Hart and Moore, 1990). One plausible way to

SC

solve this problem is to let foreign input sellers and domestic buyers agree on contracts, but whether such contracts are credible depends on the local institutional environment (i.e., the marketization level

NU

in this paper). In principle, it is better to sign the necessary contracts to access these intermediate inputs for the firms in regions with better institutional environment when compared to the firms in

MA

regions with weaker institutional environment. As a result, the firms located in regions with better institutional environment could access more varieties of intermediate inputs as well as import at a

D

lower cost following input trade liberalization. Furthermore, the selection rule in regions with higher

PT E

marketization level is more market-oriented, and therefore firm entry would be subject more to input tariff reductions. Consequently, input tariff reductions significantly increase firm entry rate in the regions with higher marketization level, yet there is a weaker impact of input trade liberalization on

CE

firm entry rate in the regions with lower marketization level. For the sake of robustness, we estimate Eq. (2) by including interactions of the output tariff and the input tariff with the marketization index.

AC

As shown in column (6) of Table 1, the coefficients of interest on the interaction terms between the tariffs and the marketization index remain robust and stable with a similar magnitude when compared to their counterparts reported in the columns (4)-(5) of Table 1. Furthermore, we include industry-year fixed effects in the last column of Table 1, we find that the coefficients of interaction terms between tariffs and marketization index barely change in their significance and magnitude. The coefficients of output tariff and input tariff fail to be estimated when the industry-year fixed effects are included, as both of them are time-varying industry characteristics, and they would be absorbed by industry-year fixed effects.

19

ACCEPTED MANUSCRIPT Next, we proceed to analyze the heterogeneous impacts of trade liberalization on firm exit rate across regions with different marketization levels. The last four columns of Table 2 report the estimation results for Eq. (2) with firm exit rate as the dependent variable. Columns (4) and (5) focus on interactions between the output trade liberalization and the regional marketization, and between the input trade liberalization and the regional marketization, respectively; while in column (6), we run Eq.

PT

(2) by including interactions of the output tariff and the input tariff with the marketization index simultaneously. Clearly, our coefficients of interest on the interaction terms between the tariffs and the

RI

marketization index are robust and stable with a similar magnitude across these estimations. In particular, according to column (6) of Table 2, we find that the coefficient on the interaction term

SC

between the output tariff and the marketization index is negative and significant at the 1% level, which indicates the existence of complementarities between output trade liberalization and regional

NU

marketization. Put differently, regional marketization strengthens the positive impact of output tariff reductions on firm exit rate. As pointed out earlier, output tariff reductions yield intensified

MA

competition, which in turn forces firm to exit. In addition, since the exit barriers are lower and the selection rule is more market based in regions with higher marketization level, thus the impact of

D

output tariff reductions on firm exit rate is stronger in these regions. Further, the interaction between

PT E

the input tariff and the marketization index also has a negative and significant impact, whereas the coefficient on the input tariff is statistically insignificant. The implication is that input tariff reductions do not affect firm exit rate in regions with the lowest marketization level, but it could lead to an

CE

increase in exit rate in regions with a higher marketization level, and such an impact is increasing as the marketization index of the regions increases. This is not too surprising given that the exit barriers

AC

are lower and the selection rule is more market based in regions with higher marketization level, and below we will further demonstrate that input tariff reductions mainly force the least productive firms to exit. As a robustness check, we add industry-year fixed effects to Eq. (2), the estimation result is reported in column (7) of Table 2, clearly, our regressors of interest (i.e., Output tariff×marketization index and Input tariff×marketization index) barely change in their significance and magnitude.

4.3. The Impact of trade liberalization on firm exit1920

19

There is a point worth to mention that Eq. (3) with firm fixed effects fail to be consistently estimated by probit, as it will potentially suffer from the incidental parameters problem (Greene, 2004) given the large number of fixed effects. Therefore, we don’t include firm fixed effects when conducting probit estimation, as in Bustos (2011) and Holloway (2017). 20

ACCEPTED MANUSCRIPT Thus far, we have estimated the impact of trade liberalization on firm dynamics using firm entry rate and exit rate as dependent variables. However, one potential drawback of the above analyses is that it assumes homogeneous impact on all firms, and thus may mask the heterogeneous effects of trade liberalization on firm dynamics among firms with different productivity. In this subsection, we proceed to investigate the impact of trade liberalization on firm exit by estimating a probit model at

PT

the firm level,21 which enables us to examine the potential reallocation effect of trade liberalization. We start by estimating Eq. (3) including both output tariff and input tariff and present the

RI

regression results in column (1) of Table 322. We find that the coefficient on output tariff is negative

SC

and significant at the 1% level. This suggests that the pro-competitive effect arising from output tariff reductions increases firms’ exiting likelihood, which is in line with the earlier findings (see, e.g.,

NU

Blyde and Iberti, 2012; Bernard et al., 2006). In addition, the coefficient on input tariff is negative but statistically insignificant, indicating that there is no evidence that input tariff reductions lead to a

MA

substantial increase in the average likelihood of firm exit. In column (2), we add an interaction term between the output tariff and the relative productivity to examine whether output trade liberalization exerts differential effects on the likelihood exit for firms with different productivity. We find that the

D

interaction term between the output tariff and the relative productivity is positive and statistically

PT E

significant, which means that the likelihood of exiting is lower for more productive firms in the face of output tariff reductions. We also note that the magnitude of the output tariff coefficient is slightly

CE

larger compared to its counterpart in column (1) of Table 3. This indicates that output tariff reductions do exert a stronger impact on the likelihood of exit for the least productive firms. Likewise, we

AC

examine whether input trade liberalization exerts differential impacts on the likelihood of exit for firms with different productivity by introducing an interaction term between the input tariff and the relative productivity. As shown in column (3) of Table 3, the coefficient on the interaction term between the input tariff and the relative productivity is statistically positive, suggesting that the probability of firm exit is relatively lower for more productive firms in the face of input tariff

20

As a robustness check, we estimate the impact of trade liberalization on firm exit using both linear probability model (LPM) and logit with firm fixed effects. As shown in Table A4, we find very similar results from these regressions, as compared to the baseline results using probit method here. We thank an anonymous referee for this suggestion. 21 As pointed out earlier, since we can only observe which firms did enter but hardly observe which potential firms that didn’t enter, therefore we fail to examine the impact of trade liberalization on firm entry at the firm level. 22 Here, we focus mainly on the direction and significance of the impact of trade liberalization on firm exit rather than economic magnitude, we thus report the original coefficients of probit regression in Table 3 rather than the marginal effects of each regressor, as in Greenaway et al. (2007), Minetti and Zhu (2011) and Li et al. (2015). 21

ACCEPTED MANUSCRIPT reductions. Interestingly, the coefficient on input tariff is now statistically negative, implying that input tariff reductions lead to a substantial increases in the likelihood of exit for least productive firms. Also, the results in column (1) of Table 3 provides a possible interpretation for the findings that input tariff reductions do not have significant impact on the likelihood of firm exit on the whole. On the one hand, input tariff liberalization leads to an increase in the likelihood of exit among the less productive

PT

firms, on the other hand, it reduces the probability of exit for the more productive firms. These two opposite effects may offset each other, resulting in an insignificant impact on the average. For the sake

RI

of robustness, column (4) of Table 3 presents the regression results following Eq. (4) which includes the interaction terms between the two tariff measures and the relative productivity. We find that the

NU

counterparts presented in columns (2) to (3) of Table 3.

SC

coefficients of interest remain robust and stable with a similar magnitude compared to their

[Table 3 inserted here]

MA

Besides, it is also interesting to examine whether the reallocation process of trade liberalization is related to regional marketization. We first investigate how the reallocation process of output tariff

D

reductions responds to regional marketization. In particular, we add a triple interaction term among

PT E

the output tariff, the relative productivity and the marketization index, as well as an interaction term between the output tariff and the marketization index to the regression. As shown in column (5) of Table 3, the coefficient on the triple interaction term, output tariff×relative productivity×

CE

marketization index, is positive and significant at the 5% level, and the coefficient on the interaction term between the output tariff and the relative productivity is statistically positive but with a smaller

AC

magnitude compared to its counterpart in column (2) of Table 3. This indicates that regional marketization magnifies the reallocation process of output tariff reductions. To be more precise, the reallocation process of output tariff reductions is relatively weak in regions with a lower marketization level, but it increases significantly as the marketization level of the regions increases. Next, we turn to analyze how the reallocation process of input tariff reductions responds to regional marketization. To this end, we introduce a triple interaction term among the input tariff, the relative productivity and the marketization index, as well as an interaction term between the input tariff and the marketization index to the regression. Column (6) of Table 3 presents the regression results. We find that the triple interaction, input tariff×relative productivity×marketization index, has a positive and significant 22

ACCEPTED MANUSCRIPT impact, while the coefficient on the interaction term between the input tariff and the relative productivity is now losing its significance. This suggests the reallocation process of input tariff reductions does not exist in regions with the lowest marketization level, but it becomes significant and stronger as the marketization index of the regions increases. In addition, we attempt to examine whether the reallocation process of trade liberalization (including both the output and input tariff

PT

reductions) is related to regional marketization by estimating Eq. (5). The result is reported in column (7) of Table 3. Evidently, our coefficients of interest remain robust and stable with a similar

RI

magnitude compared to their counterparts presented in columns (5) and (6) of Table 3. As a robustness check, we add firm export status23 to Eq. (3), column (8) of Table 3 presents the results. We find that

SC

the coefficient on the firm export indicator is statistically negative, indicating that exporters have a lower exit probability compared with non-exporters. More importantly, our regressors of interest

NU

barely change in their significance and magnitude. Finally, we further control for industry-year fixed effects and add the interaction terms between tariffs and firm export indicator in the last column of

MA

Table 324. An important advantage of this specification is that it could ensure that the conditioning role of initial productivity was not simply proxying for exposure to trade liberalization. As shown in

D

column (9) of Table 3, we find that our results are not sensitive to these additional controls, and it

PT E

once again demonstrates that regional marketization strengthens the reallocation process of trade liberalization.

CE

4.4. Endogeneity of trade policy and 2SLS estimation One concern with the estimation strategy in the current paper is that the trade policy may itself be

AC

endogenous, as is often emphasized in the trade literature. First and foremost, firm dynamics could have a reverse impact on tariff changes. More specifically, in the declining or lagging sectors—where the entry rate is lower while the exit rate is higher—the incumbent firms may have more incentives to lobby the government for lower tariffs on the products they sell. At the same time, they also have more incentives to lobby for lower tariffs on their upstream industries as this will help to raise their profits (Hu and Liu, 2014). Second, some unobservable heterogeneity as well as common year shocks (e.g., macroeconomic fluctuation) may affect both the level of tariffs and firm dynamics

23 24

Firm export indicator equals one if a firm exports and zero otherwise. We thank an anonymous referee for this suggestion. 23

ACCEPTED MANUSCRIPT simultaneously, if this is the case, trade policy may also be endogenous. In the previous sections, we have accounted for the potential simultaneity and reverse causality by using the one-year lagged output tariff and input tariff. For the sake of completeness and robustness, we further adopt the instrumental variables approach to control for the endogeneity issues. Admittedly, it is challenging to find an ideal instrument for tariffs. Inspired by Gaston and Trefler

PT

(1997), Beaulieu (2000) as well as Baggs (2005), we construct the instruments for tariffs in two steps. First, we conduct a cross-section regression of 2001 output tariffs on 1998-2000 industry output

RI

growth, 1998-2000 wage growth, 1998-2000 import growth, 1998-2000 employment growth,

SC

1998-2000 profit growth, and 1998-2000 sales growth. Second, after the fitted values of the 2001 output tariffs were estimated, we then obtain the fitted values of output tariffs in the years after 2001

NU

by applying a common phase-out rule25 for all industries to the fitted values of the 2001 output tariffs output following Beaulieu (2000). The constructed instruments for output tariffs (denoted by iv jt )

MA

consist of these fitted output tariffs after 2001 and the actual output tariffs before 2001. Similar to Eq. input output (7), the instruments for input tariffs can be calculated as iv jt   w jw  iv jt , where  jw is the

D

cost share of input w in the production of a good in industry j.

PT E

Table 4 reports the two-stage least squares (2SLS) estimates regarding the effect of trade output liberalization and marketization on firm entry and exit rate. In particular, we use iv jt and

CE

iv input as instruments for Output tariff and Input tariff, respectively. Accordingly, the interactions jt

AC

output input between iv jt (or iv jt ) and the marketization index are adopted as additional instruments for

the interactions between Output tariff (or Input tariff ) and the marketization index. As shown in column (1) of Table 4, the coefficient on the interaction term between the output tariff and the marketization index is positive and significant at the 5% level, and the coefficient on output tariff is also significantly positive. This once again suggests that output tariff reductions lead to a decrease in firm entry rate, and that regional marketization strengthens such an impact. For input trade liberalization, we find that both the coefficients on input tariff and its interaction with marketization index are negative and significant at the conventional statistical level. In addition, they have a much

25

This is calculated as the annual average tariff cut rates since 2001 as China committed when accessing to the WTO. 24

ACCEPTED MANUSCRIPT larger magnitude compared to their counterparts in column (6) of Table 1. This confirms the previous findings that input tariff reductions exert positive and stronger impact on firm entry rate as the marketization level of the regions increases. In column (3) of Table 4, we conduct a robustness check by further controlling for industry-year fixed effects, and find that our results are not sensitive to these additional controls.

PT

[Table 4 inserted here] The 2SLS estimates regarding the effect of trade liberalization and marketization on firm exit

RI

rate are presented in column (2) of Table 4. It turns out that the coefficients of interest on the

SC

interaction term between the output tariff and the marketization index, and the interaction term between the input tariff and the marketization index are significantly negative, which are consistent

NU

with their counterparts reported in column (3) of Table 2. And in column (4) of Table 4, we further control for industry-year fixed effects and still get qualitatively similar results. This once again

MA

confirms and reinforces our previous findings that trade liberalization complements regional marketization. To be more precise, the regional marketization strengthens the positive impact of tariff reductions on firm exit rate. In addition, we perform several tests to verify the quality of the

PT E

D

instruments. We start by using the Kleibergen-Paap LM  2 statistic to check whether the excluded instruments are correlated with the endogenous regressors. As shown in the bottom module of Table 4, the null hypothesis that the model is under-identified is rejected at the 1% significance level. We then

CE

perform the Kleibergen and Paap (2006) Wald statistic to check whether the instrument is weakly correlated with the endogenous variable, and find that the null hypothesis that the first stage is weakly

AC

identified is rejected at the 1% significance level. Finally, the Anderson and Rubin  2 statistics reject the null hypothesis that the coefficient of the endogenous regressor is equal to zero. All the above tests confirm that our instruments are valid and the specifications are well justified. We now turn to analyze the IV probit estimates regarding the effect of trade liberalization on firm output input exit, which are reported in Table 5. Likewise, we use iv jt , iv jt and the relevant interactions as

instruments for output tariff, input tariff and the interactions, respectively. As shown in column (1) of Table 5, the coefficient on output tariff is significantly negative, while the coefficient on the interaction term between the output tariff and the relative productivity is statistically positive The 25

ACCEPTED MANUSCRIPT results suggest that output tariff reductions lead to a significant increase in the likelihood of exit for the least productive firms, but such an impact is lower for the more productive firms. Furthermore, we see that our coefficient of interest on the interaction term between the input tariff and the relative productivity is positive and significant at the 5% level, while the coefficient on input tariff is statistically negative. These results are in line with their counterparts presented in column (4) of Table

PT

3. As a robustness check, we further control for industry-year fixed effects. As shown in column (3) of Table 5, we find that our regressors of interest (i.e., Output tariff×relative productivity and Input

RI

tariff×relative productivity) barely change in their significance and magnitude. These results once again confirm that input trade liberalization exerts heterogeneous impacts on firm exit among those

SC

firms with different productivity. In particular, input tariff reductions lead to a substantial increase in the likelihood of exit for the least productive firms, while the probability of exitis lower among the

NU

more productive firms in the face of input tariff reductions.

MA

[Table 5 inserted here] Next, we proceed to explore whether the reallocation process of trade liberalization is related to regional marketization by conducting an IV probit estimation. As shown in column (2) of Table 5,

D

once we add the triple interaction terms (i.e., output tariff×relative productivity×marketization index

PT E

and input tariff×relative productivity×marketization index), the coefficient on the interaction term between the output tariff and the relative productivity loses its significance, while the coefficient on

CE

the triple interaction term, output tariff×relative productivity×marketization index, is significantly positive, which are consistent with their counterparts reported in column (7) of Table 3. This suggests

AC

that the reallocation process of output tariff reductions is weaker in regions with a lower marketization level, but it is increases significantly as the marketization level increases. Similarly, the triple interaction, input tariff×relative productivity×marketization index, is positive and significant at the conventional statistical level, while the coefficient on the interaction term between the input tariff and the relative productivity is statistically insignificant. This once again confirms that there is not an obvious reallocation process of input tariff reductions in regions with the lowest marketization level, but the proecess becomes significant and stronger as the marketization index increases. Last but not least, we further control for industry-year fixed effects in the last column of Table 5, and find that our results are not sensitive to these additional controls. 26

ACCEPTED MANUSCRIPT 4.5. Difference-in-differences (DID) estimation In section 4.4, we have adopted the instrumental variables (IV) approach to deal with the endogeneity

of

trade

liberalization,

for

the

sake

of

robustness,

here,

we

employ

difference-in-differences technique to further address the endogeneity concern. The rationale for DID estimation is that the industries that had higher initial tariff level (i.e., more protected) witnessed

PT

larger tariff reductions under the WTO agreement, and thus undergone higher degrees of trade liberalization, whereas it is opposite for industries that had lower initial tariff level (Lu and Yu, 2015;

RI

Liu and Qiu, 2016), we thus treat the firms in industries with previously higher tariff level as treatment group, and the firms in industries with previously lower tariff level as control group. In

SC

particular, following Lu and Yu (2015) as well as Liu and Qiu (2016), we use the DID specifications (8) and (9) to investigate the impact of output tariff and input tariff liberalization on firm dynamics,

NU

respectively:26

MA

Yjkt   jk  1OutTarj 01  Post 02t   2 InputTariff jt 1  X jkt 1  t  kt   jkt Yjkt   jk  1InTarj 01  Post 02t  2OutputTariff jt 1  X jkt 1  t  kt   jkt

(8) (9)

D

where j, k, and t denote industry, region and year, respectively; Y jkt denotes firm dynamics (i.e.,

PT E

entry rate or exit rate depending on the specification); OutTarj 01 and InTarj 01 are the output tariff rate and input tariff rate of industry j in 2001, respectively; Post 02t represents a post-WTO period,

CE

taking a value of 1 if it is year 2002 and onwards, and 0 otherwise. In specification (8), the interaction term between the output tariff in 2001 ( OutTarj 01 ) and the post-WTO indicator ( Post 02t ) is our

AC

major interest, whose parameter 1 captures the causal effect of output tariff liberalization on firm dynamics; similarly, in specification (9), our variable of interest is the interaction term between the input tariff in 2001 ( InTarj 01 ) and the post-WTO indicator ( Post 02t ), and its parameter 1 captures the causal effect of input tariff liberalization on firm dynamics.  jk is the region-industry fixed effects,

t is the year fixed effects, kt is the region-year interaction fixed effects, and  jkt is the error 26

It is important to note that it’s irrelevant to contain both the interaction terms OutTarj 01  Post 02t

and

InTarj 01  Post 02t in one DID specification, because of this, Lu and Yu (2015) adopt a similar specification as Eq. (8) to explore the impact of output tariff liberalization on markup dispersion, and Liu and Qiu (2016) employ a similar specification as Eq. (9) to empirically investigate the effect of input tariff liberalization on firm innovation. 27

ACCEPTED MANUSCRIPT term. Columns (1) and (2) of Table 6 present results for the DID specifications (8) and (9) with firm entry rate as dependent variable, respectively. We find that the interaction term, OutTat01×Post02, is negative and statistically significant, suggesting that the firm entry rate decreased more for firms in industries which face a larger reduction in output tariffs (i.e., higher OutTat01) after China’s WTO

PT

accession, namely, output tariff reductions discourage firm entry. As shown in column (2), our regressor of interest, InTat01×Post02, is positive and significant at the 1% level, implying that the firm

RI

entry rate increased more after 2002 in industries with higher input tariffs in 2001 than in industries with lower input tariffs in 2001, that is, input tariff liberalization tends to increase firm entry rate. In

SC

columns (5) and (6) of Table 6, we report the results for the DID specifications (8) and (9) with firm exit rate as dependent variable, respectively. We see that both the interaction terms, OutTat01×Post02

NU

and InTat01×Post02, are positive and statistically significant, indicating that both of output tariff liberalization and input tariff liberalization increase exit rate on average.

MA

[Table 6 inserted here]

To investigate the heterogeneous impacts of trade liberalization on firm entry rate across regions different

marketization

levels,

we

add

the

triple

interaction

terms,

D

with

PT E

OutTat01×Post02×marketization index and InTat01×Post02×marketization index, to the baseline DID specifications (8) and (9), respectively. As shown in column (3) of Table 6, the coefficient of interest on the triple interaction term, OutTat01×Post02×marketization index, is negative and statistically

CE

significant, suggesting that the discouraging effect of output tariff liberalization on firm entry is stronger in regions with higher marketization level. Relying on column (4) of Table 6, the coefficient

AC

on the triple interaction term, InTat01×Post02×marketization index, is positive and statistically significant, whereas the coefficient on InTat01×Post02 is statistically insignificant and small in magnitude, this implies that in regions with lowest marketization level, input tariff liberalization could not increase firm entry rate, but would cause a positive and stronger impact on firm entry rate as the marketization level of the regions increases. We now turn to explore the heterogeneous impacts of trade liberalization on firm exit rate across regions with different marketization levels, to this end, we add the triple interaction terms, OutTat01×Post02×marketization index and InTat01×Post02×marketization index, to the baseline DID specifications (8) and (9) with firm exit rate as dependent variable, and the results are reported in the 28

ACCEPTED MANUSCRIPT last

two

columns

of

Table

6.

Evidently,

both

of

the

triple

interaction

terms,

OutTat01×Post02×marketization index and InTat01×Post02×marketization index, are positive and statistically significant, indicating that there does indeed exist a strong complementarity between trade liberalization and regional marketization. Taken together, these results are consistent with the previous findings in sections 4.1 and 4.2.

4.6. Further robustness checks of controlling for other regional characteristics27

PT

There may still be a concern that some other regional characteristics could correlate with

RI

marketization index, and if this is true, the conditioning effect of marketization found above would be biased. To mitigate this identification concern related to omitted variable bias, here we augment the

SC

basic empirical specification (2) with a couple of controls, i.e., the interactions between tariffs and other indicators, including GDP per capita, trade openness, FDI inflows, share of processing trade,

NU

pro-competition indicator at the region level. Specifically, GDP per capita (denoted by pgdpkt ) is measured by the logarithm of GDP per capita of region k in year t; regional trade openness (denoted

MA

by openkt ) is calculated as the ratio of export value to the GDP of region k in year t; FDI inflows (denoted by fdikt ) is measured by the logarithm of inbound FDI amount of region k in year t; share of processing trade (denoted by processsharekt ) is measured as a proportion of processing import in

D

total imports of region k in year t28; and pro-competition indicator (denoted by competitionkt ) is

PT E

defined as 1-Lerner Index following Aghion et al. (2015), where Lerner Index is calculated as the ratio of operating profits less capital costs to sales of region k in year t. Table 7 reports the estimation results after controlling for the interactions between tariffs and the

CE

above-mentioned regional characteristics. In columns (1)-(5), the dependent variable is firm entry rate, while in columns (6)-(10), the dependent variable is firm exit rate. As shown in columns (1)-(5), we

AC

find that the coefficient of Output tariff × open is statistically positive, suggesting that output tariff reductions exert stronger depressing impact on firm entry rate in more opened regions; in addition, the coefficient of Input tariff × processshare is statistically positive, implying that the encouraging effect of input tariff reduction on firm entry rate is decreasing as the share of processing trade increases; and more importantly, our regressors of interest (i.e., the interactions between tariffs and marketization index) barely change in their significance and magnitude compared with column (7) of Table 1. [Table 7 inserted here] 27

We thank an anonymous referee for this suggestion. Since the ASIE data does not contain the information of processing trade, we use the ASIE data merged with the Chinese Customs Trade Statistics data to calculate the share of processing trade in region level. 29 28

ACCEPTED MANUSCRIPT Now turning to analyze the results with firm exit rate as dependent variable. As shown in columns (6)-(10), we find that the coefficient of Output tariff × open is negative and significant at conventional levels, implying that the positive impact of output tariff reductions on firm exit rate is stronger in more opened regions; additionally, both the interaction terms, Output tariff × processshare and Input tariff × processshare, are positive and statistically significant, indicating that the positive impact of tariff reductions on firm exit rate is less profound as the processing trade share increases,

PT

thus, by and large, the impact of tariff liberalization on firm dynamics becomes weaker in the regions with higher share of processing trade; and once again, our regressors of interest (i.e., the interactions

RI

between tariffs and marketization index) are pretty close to their counterparts in column (7) of Table 2.

SC

All in all, our main findings regarding regional marketization strengthens the impact of trade liberalization on firm dynamics are not driven by these other regional characteristics.

NU

4.7. The role of subcomponents of the marketization index29

So far, all estimations have used the overall regional marketization index, as in Li et al. (2010),

MA

Fan et al. (2011), and Ding et al. (2016). However, a new interesting question is aroused that whether five subcomponents of marketization index30 play the same role in the effects of tariff reductions on firm dynamics, if not, which subcomponent of marketization index is more important? In this

D

subsection, we will explore the potential heterogeneous role of marketization by interacting tariffs

PT E

with the reduction of government intervention (MI1), the development of non-state-owned enterprises (MI2), the development of product market (MI3), the development of factor market (MI4), and the

CE

development of intermediaries and efficiency improvement of legal system (MI5), respectively. Table 8 presents the estimation results with subcomponents of marketization index as the key

AC

variables of interest. In columns (1)-(5), the dependent variable is firm entry rate, while in columns (6)-(10), the dependent variable is firm exit rate. We start by analyzing the heterogeneous role of subcomponents of marketization index on firm entry rate. As shown in column (1) of Table 8, we find that the coefficient of Output tariff × MI1 is statistically positive, while the coefficient of Input tariff × MI1 is statistically negative, suggesting that reduction of government intervention strengthens the depressing impact of output tariff liberalization on firm entry rate, and in regions with less government intervention, the encouraging effect of input tariff reduction on firm entry rate is stronger,

29

We thank an anonymous referee for this suggestion. It includes reduction of government intervention, development of non-state-owned enterprises, development of product market, development of factor market, and development of intermediaries and efficiency improvement of legal system. 30 30

ACCEPTED MANUSCRIPT this is consistent with our aforementioned results using overall marketization index as the key variable of interest. From column (2), we see that in regions with better development of non-state-owned enterprises, the depressing impact of output tariff liberalization on firm entry rate is stronger, whereas the moderating role of development of non-state-owned enterprises in shaping the effects of input trade liberalization is insignificant. In columns (3) and (4), we interact tariffs with the development of

PT

product market (MI3), and the development of factor market (MI4), respectively. The results show that the coefficients on the interaction terms are statistically insignificant across both specifications. In

RI

column (5) of Table 8, we interact tariffs with the development of intermediaries and efficiency improvement of legal system (MI5), evidently, the results are similar to column (1) of Table 8,

SC

implying that development of intermediaries and efficiency improvement of legal system tends to strengthen the depressing impact of output tariff liberalization on firm entry rate, and enhance

NU

encouraging effect of input tariff reductions on firm entry rate.

[Table 8 inserted here]

MA

Next, we proceed to investigate the heterogeneous role of subcomponents of marketization index on firm exit rate. As shown in columns (6)-(10) of Table 8, we find that the coefficients on the

D

interaction terms between tariffs and MI1, tariffs and MI2, are statistically negative, this suggests that

PT E

both reduction of government intervention and development of non-state-owned enterprises tend to strengthen the positive impact of trade liberalization on firm exit rate. Additionally, we see that the interaction between input tariff and MI5 also has a negative and significant term, suggesting that in

CE

regions with better development of intermediaries and efficiency improvement of legal system, the positive impact of input tariff reductions on firm exit rate is also stronger. Furthermore, the

AC

coefficients on the interaction terms between tariffs and MI3 (see column (8)), tariffs and MI4 (see column (9)), are either statistically insignificant or small in magnitude. To sum up, our results show that the five subcomponents of marketization index do indeed paly heterogeneous role in shaping the effects of tariff reductions on firm dynamics, of which the reduction of government intervention, the development of non-state-owned enterprises, and the development of intermediaries and efficiency improvement of legal system matter more.

5. Trade liberalization, marketization and aggregate productivity 5.1. Comparing average productivity across different groups and decomposition 31

ACCEPTED MANUSCRIPT In this subsection, we first compare the average productivity among surviving, entering, and exiting firms, which allows us to verify the existence of market selection; we then conduct a decomposition analysis using the dynamic Olley-Pakes decomposition method proposed by Melitz and Polanec (2015). Table 9 presents the differences of average productivity for all firms and by period as well as by

PT

region. It turns out that the surviving firms’ average productivity is significantly higher than that of entering firms as well as exiting firms, and the average productivity of entering firms is significantly

RI

higher than that of exiting firms when compared at the whole sample level. This suggests that the exiting firms are on average less productive than both the surviving and entering firms, which

SC

provides evidence of the existence of market selection process. The above results are in line with most

NU

of the previous literature, such as the careful analyses done by Hahn (2000) on Korea, Aw et al. (1997) on Taiwan and Gebreeyesus (2008) on Ethiopia, etc.

MA

In addition, we also compare the average productivity difference of these different groups of firms by period (i.e., the period before China’s WTO accession vs. the period after China’s WTO accession) and by region (i.e., the regions with a higher marketization level vs. the regions with a

D

lower marketization level). As shown in Table 9, the average productivity of surviving firms is

PT E

significantly higher than that of exiting firms for all of the four subgroups; and the average productivity of entering firms is also significantly higher than that of exiting firms for each subgroup; last but not least, the surviving firms are found to be more productive than entering firms in 3 out of 4

[Table 9 inserted here]

AC

CE

subgroups, the only exception being the case of the period after China’s WTO accession.

The findings above indicate that the entering firms are more productive than exiting firms, and thus firm dynamics may raises aggregate productivity. We now proceed to conduct a decomposition analysis using the dynamic Olley-Pakes method developed by Melitz and Polanec (2015). By doing so, we can quantify the contribution of firm entry and exit to aggregate productivity growth. We first define industrial aggregate productivity in a year as: TFPjt 



i j

32

it

 tfpit

(10)

ACCEPTED MANUSCRIPT where i, j , and t denote the firm, industry (CIC 3-digit)31, and year, respectively;  j is the set of firms in industry j;  it is a firm specific weight of firm i active in industry j in year t. Following Baldwin and Gu (2003), Disney et al. (2003) as well as De Loecker and Konings (2006), we use the ratio of firm i’s employment over the industry j’s total employment to proxy for  it , and tfpit is firm

expressed as



i( S , N )

it  tfpit 



i( S , X )

it 1  tfpit 1

RI

TFPjt 

PT

i’s productivity in year t. Then, the industrial aggregate productivity growth between years can be

(11)

SC

where S, N and X correspond to the incumbent, entering and exiting firms, respectively. We next decompose industrial aggregate productivity growth using the dynamic Olley-Pakes decomposition

NU

method proposed by Melitz and Polanec (2015) which can be expressed as TFPjt  TFP St  covS  it (TFPNt  TFPSt )  it 1 (TFPSt 1  TFPXt 1 ) iX

(12)

MA

iN

As shown in Eq. (12), a change in industrial aggregate productivity in Eq. (12) can be

D

decomposed into four parts. The first is the within term which measures the aggregate productivity

PT E

growth due to the average productivity improvements of the incumbent firms; the second is the covariance term, which is calculated by the change in covariance between labor share and productivity for incumbent firms. It can capture the contribution of resource allocation among

CE

incumbents to aggregate productivity growth. The third is the entry effect, which captures the aggregate productivity growth due to the entrants. It is positive only if the weighted average

AC

productivity level of entering firms is higher than the weighted average productivity level of continuing firms in the corresponding year. The fourth is the exit effect, which represents the gains in aggregate productivity growth from the exiters. It is positive if the weighted average productivity level of exiting firms is lower than the weighted average productivity level of continuing firms in the corresponding year. In this paper, we define the sum of the last two terms in decomposition Eq. (12) as the dynamics effect (or the net entry effect), which reflects the resource reallocation effect due to firm entry and exit. 31

It is worth noting that we define the industry at 3-digit CIC level here, as if we define it at 4-digit CIC level, then a lot of region-industry pairs have few observations or even no observation. We also define the industry at 4-digit CIC level, however, the results are not sensitive to this change. 33

ACCEPTED MANUSCRIPT Table 10 presents the results of a decomposition analysis. The values in the brackets indicate the contributions of decomposition terms to aggregate productivity growth (%), while the values outside the brackets indicate the decomposition terms. We start by analyzing the decomposition results for the whole sample (see the top panel of Table 10). We find that the aggregate productivity growth is as high as 13.4 per cent during our sample period. The second column shows the contribution of the

PT

within effect. It accounts for around 58.6% of aggregate productivity growth over the whole period. Evidently, it is the most important impetus for aggregate productivity growth. The next is the exit

RI

effect (see column (5)), which accounts for around 37% of aggregate productivity growth during the same period. The results show that exiters indeed have lower productivity. The fourth column shows

SC

that the entry effect, which accounts for approximately 6% of aggregate productivity growth, is also positive. Interestingly, the contribution of between effect is negative but small in absolute value. In

NU

addition, we see that the dynamics effect, which contributes approximately 43% of the growth of productivity, plays an important role. These results are in line with what has been found in the

MA

previous studies, particularly those documenting high firm dynamics (see, e.g., Disney et al., 2003; Bartelsman et al., 2004; Gebreeyesus, 2008).

D

The middle panel of Table 10 shows the results of the decomposition analysis by period (i.e., the

PT E

period before China’s WTO accession vs. the period after China’s WTO accession). Clearly, the productivity growth rate in the period after China’s WTO accession is much higher than that in the

CE

period before China’s WTO accession. More interestingly, the contribution of the dynamics effect is much larger in the period after China’s WTO accession compared with that in the period before

AC

China’s WTO accession. To be more precise, they contribute 53.1% and 31.9% of the growth of productivity, respectively. In the bottom panel of Table 10, we decompose aggregate productivity growth by region (i.e., the regions with a higher marketization level vs. the regions with a lower marketization level). We find that the productivity growth rate is much larger in regions with a higher marketization level In addition, the dynamics effect contributes more in regions with a higher marketization level (49.6%) than that in regions with a lower marketization level (29.3%). [Table 10 inserted here]

5.2. The impact of trade liberalization and marketization on aggregate productivity evolution 34

ACCEPTED MANUSCRIPT To further shed light on how trade liberalization and marketization affect aggregate productivity evolution, we decompose aggregate productivity for each industry-region pair, and consider the following empirical specification: Y jkt  0  1OutputTariff jt 1   2 InputTariff jt 1  3OutputTariff jt 1  Marketkt 1   4 InputTariff jt 1  Marketkt 1  5 Herfindahl jkt   j   kt   t   jkt

(13)

PT

where j, k, and t correspond to the industry, region and year, respectively. Y jkt denotes aggregate

RI

productivity growth as well as its components. OutputTariff jt 1 and InputTariff jt 1 are the one-year

SC

lagged industry specific output tariff and input tariff, respectively. Marketkt 1 denotes the one-year lagged regional marketization index. Herfindahl jkt denotes the herfindahl index, which is used as

NU

the control variable. Besides, we further control for the industry fixed effects (  j ), the region-year

MA

interaction fixed effects (  kt ), as well as the year fixed effects (  t ).  jkt is the error term. Table 11 presents the results from estimating Eq. (13).32 In column (1), we first run a regression

D

on TFPjkt of Output tariff, Input tariff as well as the Herfindahl index as a benchmark, and find that

PT E

the coefficients on Output tariff and Input tariff are both negative and significant at the conventional statistical level, suggesting that trade liberalization leads to a significant increase in aggregate

CE

productivity . In column (2), we add two interaction terms between the tariffs and the marketization index. It turns out that the coefficients on the interaction terms, Output tariff × marketization index

AC

and Input tariff × marketization index, are statistically negative, which suggests that there exists complementarities between the market reform and the productivity gains from trade liberalization. In other words, regional marketization strengthens the positive impact of trade liberalization on aggregate productivity growth. More interestingly, the coefficients on Output tariff and Input tariff are now losing their significance and the magnitude declines dramatically when compared to their counterparts in column (1) of Table 11. This in turn confirms the existence of complementarities between the trade liberalization and the market reform.

32

Here, we use firm TFP calculated using Olley and Pakes’s (1996) method to construct aggregate productivity. The main results are robust to alternative TFP estimations (e.g., LP method and simple OLS method) as well as to labor productivity, please see Table A5 in the Appendix for more details. We thank an anonymous referee for suggesting such a test. 35

ACCEPTED MANUSCRIPT Column (3) of Table 11 presents the estimates following Eq. (13) for within component. It can be seen that the coefficient on the interaction terms, Output tariff × marketization index and Input tariff × marketization index, are negative and significant at least the 5% level, indicating that regional marketization tends to strengthen the positive impact of trade liberalization on average productivity improvement. In Column (4), we report the estimated results for Eq. (13) with between effects as

PT

dependent variable. It turns out that the between effects are not an important channel through which trade liberalization and marketization foster aggregate productivity growth, since none of the variables

RI

of interest is significant at the conventional level. Column (5) of Table 11 reports the estimates following Eq. (13) on the entry effect. We note that only the coefficient on the interaction terms, Input

SC

tariff × marketization index, is significant at the conventional statistical level, suggesting that regional marketization tends to strengthen the positive impact of input tariff reductions on entry effect and thus

NU

foster aggregate productivity growth. Last but not least, Column (6) of Table 11 presents the estimates following Eq. (13) on the exit effect. As shown in Column (6), both the interactions, Output tariff ×

MA

marketization index and Input tariff × marketization index, have a negative and significant term, indicating that regional marketization strengthens the positive impact of trade liberalization on the exit

D

effect. To sum up, the above results demonstrate that firm dynamics especially the exit effect is an

PT E

important channel through which trade liberalization fosters productivity growth, and market reform strengthens such an impact.

AC

6. Conclusion

CE

[Table 11 inserted here]

In this paper, we investigate the impact of trade liberalization on firm dynamics and productivity, and how this impact varies across regions with different marketization levels. Following the accession to the WTO in 2001, China was required to cut tariffs dramatically, which provides an excellent setting to analyze the questions proposed in this paper. Using the firm-level production data obtained from the National Bureau of Statistics of China over the period 1998-2007, we find that on average, output tariff reductions tend to reduce firm entry rate and increase firm exit rate, while input tariff reductions help to increase both firm entry rate and exit rate. We also explore the heterogeneous impacts on firm dynamics across regions with different marketization levels, and demonstrate that 36

ACCEPTED MANUSCRIPT regional marketization strengthens the impact of trade liberalization on firm dynamics. To examine the potential reallocation effect of trade liberalization, we further investigate the impact of trade liberalization on firm exit across different productivity at the firm level. The results show that trade liberalization exerts a greater positive impact on the likelihood of firm exit for the least productive firms while it tends to reduce the probability of firm exit for the more productive

PT

firms. In addition, we also examine whether the reallocation process of trade liberalization depends on regional marketization level, and find that the reallocation process of trade liberalization is relatively

RI

weaker in regions with a lower marketization level, but it is increasing significantly as the marketization level of the regions increases. Last but not least, we conduct a decomposition analysis

SC

using the dynamic Olley-Pakes method proposed by Melitz and Polanec (2015). The decomposition

NU

results show that firm dynamics effect contributes approximately 43% of the growth of productivity. More interestingly, it (especially the firm exit effect) is an important channel through which trade

MA

liberalization fosters aggregate productivity growth, and market reform helps to strengthen such an impact.

This paper is one of the first to empirically investigate the impact of trade liberalization on

D

Chinese manufacturing firm dynamics and aggregate productivity by considering the role of domestic

PT E

market reform. In addition, our study also contributes to the literature that explores the impact of trade reforms on efficient allocation of resources across firms. More importantly, the current paper also has rich policy implications. On the one hand, as trade liberalization not only leads to an increase in firm

CE

productivity but also improve the resource allocation efficiency and thus foster aggregate productivity growth, it is important for Chinese government to deeply engage in more multinational trade

AC

agreements to push further trade liberalization; on the other hand, further market-oriented economic transition and improvement of regional institutional environment are needed to maximize productivity gains from trade liberalization.

References Aghion, P., Burgess, R., Redding, S. J., & Zilibotti, F. (2008). The unequal effects of liberalization: evidence from dismantling the license raj in India. American Economic Review, 98(4), 1397-1412. 37

ACCEPTED MANUSCRIPT Aghion, P., Cai, J., Dewatripont, M., Du, L., Harrison, A., & Legros, P. (2015). Industrial policy and competition. American Economic Journal: Macroeconomics, 7(4), 1-32. Ahsan, R. N. (2013). Input tariffs, speed of contract enforcement, and the productivity of firms in India. Journal of International Economics, 901), 181-192. Amiti, M., & Konings, J. (2007). Trade liberalization, intermediate inputs, and productivity: evidence from Indonesia. American Economic Review, 97 (97), 1611-1638.

PT

Aw, B.Y., Chen, X., & Roberts, M. J. (1997). Firm-level evidence on productivity differentials, turnover, and exports in Taiwanese manufacturing. NBER working paper.

RI

Baggs, J. (2005). Firm survival and exit in response to trade liberalization. Canadian Journal of Economics, 38(4), 1364-1383.

SC

Baggs, J., Beaulieu, E., & Fung, L. (2014). Persistent effects of transitory exchange rate shocks on firm dynamics. Contemporary Economic Policy, 32(2), 334-350 Baggs, J., Beaulieu, E., & Fung, L. (2009). Firm survival, performance, and the exchange rates. Canadian Journal of Economics, 42(2), 393-491.

NU

Baldwin, J. R., & Gu, W. (2003). Plant turnover and productivity growth in Canadian manufacturing. Analytical Studies Branch Research Paper Series.

MA

Bartelsman, E., Haltiwanger, J., & Scarpetta, S. (2004). Microeconomic evidence of creative destruction in industrial and developing countries. Tinbergen Institute discussion paper.

D

Beaulieu, E. (2000). The Canada-U.S. free trade agreement and labor market adjustment in Canada. Canadian Journal of Economics, 33(2), 540-563.

PT E

Bernard, A. B., & Jensen, J. B. (2002). The deaths of manufacturing plants. NBER working paper. Bernard, A., Jensen, J. B., & Schott, P. (2006). Trade costs, firms and productivity. Journal of Monetary Economics, 53(5):917-937.

CE

Beyer, H., Rojas, P., & Vergara, R. (1999). Trade liberalization and wage inequality. Journal of Development Economics, 59(1), 103-123.

AC

Blyde, J., & Iberti, G. (2012). Trade costs, resource reallocation and productivity in developing countries. Review of International Economics, 20(5), 909-923. Brandt, L., Van Biesebroeck, J., & Zhang, Y. F. (2012). Creative accounting or creative destruction? Firm-level productivity growth in Chinese manufacturing. Journal of Development Economics, 97(2), 339-351. Brandt, L., Van Biesebroeck, J., Wang, L., & Zhang, Y. (2012). WTO accession and performance of Chinese manufacturing firms. CEPR discussion paper. Bustos, P. (2011). Trade liberalization, exports, and technology upgrading: Evidence on the impact of MERCOSUR on Argentinian firms. American Economic Review, 101(1), 304-340. Cheng, W. (2015). Regional variation in trade liberalization outcomes: Evidence from Chinese manufacturing industry. University of Manchester working paper. Dai, M., Maitra, M., & Yu, M. (2016). Unexceptional exporter performance in China? The role of 38

ACCEPTED MANUSCRIPT processing trade. Journal of Development Economics, 121, 177-189. De Loecker, J., & Konings, J. (2006). Job reallocation and productivity growth in a post-socialist economy: Evidence from Slovenian manufacturing. European Journal of Political Economy, 22(2), 388-408. Ding, S., Sun, P., & Jiang, W. (2016). The Effect of import competition on firm productivity and innovation: Does the distance to technology frontier matter. Oxford Bulletin of Economics and Statistics, 78(2), 197-227.

PT

Disney, R., Haskel, J., & Heden, Y. (2003). Restructuring and productivity growth in UK manufacturing. Economic Journal, 113(489), 666-694. Fan, G., Wang, X. L., & Ma, G. R. (2011). Contribution of marketization to China’s economic growth. Economic Research Journal, 9, 4-16.

SC

RI

Fan, G., Wang, X. L., & Zhu, H. P. (2010). China’s marketization index: 2009 Report of relative marketization progress for different regions, Beijing: Economic Science Press. Fan, H., Lai, E., & Li, Y. A. (2015). Credit constraints, quality, and export prices: Theory and evidence from China. Journal of Comparative Economics, 43(2), 390-416.

NU

Feenstra, R. C., Li, Z., & Yu, M. (2014). Exports and credit constraints under incomplete information: Theory and evidence from China. Review of Economics and Statistics, 96(4), 729-744.

MA

Gaston, N., & Trefler, D. (1997). The labor market consequences of the Canada-U.S. free trade agreement. Canadian Journal of Economics, 30(1), 18-41.

D

Gebreeyesus, M. (2008). Firm turnover and productivity differentials in Ethiopian manufacturing. Journal of Productivity Analysis, 29(2):113-129.

PT E

Greenaway, D., Guariglia, A., & Kneller, R. (2007). Financial factors and exporting decisions, Journal of International Economics, 73(2), 377-395. Greene, W. H. (2004). Econometric analysis, Prentice Hall, New Jersey.

CE

Grossman, G. M., & Helpman, E. (1991). Innovation and growth in the global economy, Cambridge, MA: MIT Press.

AC

Gu, W., Sawchuk, G., & Rennison, L. W. (2003). The effect of tariff reductions on firm size and firm turnover in Canadian manufacturing. Review of World Economics, 139(3), 440-459. Hahn, C. H. (2000). Entry, exit, and aggregate productivity growth: Micro evidence on Korean manufacturing. OECD Economics Department working paper. Harrison, A. E. (1994). Productivity, imperfect competition and trade reform: Theory and evidence. Journal of International Economics, 36(1-2), 53-73. Hart, O., & Moore, J. (1990). Property rights and nature of the firm. Journal of Political Economy, 98(6), 1119-1158. Holloway, I. R. (2017). Learning via sequential market entry: Evidence from international releases of U.S. movies. Journal of International Economics, 104, 104-121. Hu, A. G., & Liu, Z. (2014). Trade liberalization and firm productivity: Evidence from Chinese manufacturing industries. Review of International Economics, 22(3), 488-512. 39

ACCEPTED MANUSCRIPT Hu, C., Lin, F. Q., & Wang, X. S. (2016). Learning from exporting in China: A firm-specific instrumental approach. Economics of Transition, 24(2), 299-334. Keller, W., Andres Santiago, J., & Shiue, C. H. (2016). Trade in China during the treaty-port era. CEPR discussion paper. Khandelwal, A. K., Schott, P. K., & Wei, Shang-Jin. (2013). Trade liberalization and embedded institutional reform: Evidence from Chinese exporters. American Economic Review, 103(6), 2169-2195.

PT

Klenow, P., & Rodriguez-Clare, A. (1997). Quantifying variety gains from trade liberalization. Mimeo.

RI

Lay, T. (2003). The determinants of and interaction between entry and exit in Taiwan’s manufacturing sector. Small Business Economics Journal, 20(4), 319-334.

SC

Levinsohn, J., & Petrin, A. (2003). Estimating production functions using inputs to control for unobservables. Review of Economic Studies, 70(2), 317-341. Li, H., Ma, H., & Xu, Y. (2015). How do exchange rate movements affect Chinese exports: A firm-level investigation. Journal of International Economics, 97, 148-161.

NU

Li, K., Zhang, J., Yu, Y., & Liu, Z. (2010). Does market-oriented economic transition enhance enterprise productivity? Evidence from China’s enterprises. Pacific Economic Review, 15(5), 719-742.

MA

Lin, F. Q. (2015). Learning by exporting effect in China revisited: An instrumental approach. China Economic Review, 36, 1-13. Lin, F. Q., & Sim, N. C. (2013). Trade, income and the Baltic dry index. European Economic Review, 59(4), 1-18.

PT E

D

Liu, Q., & Qiu, L. D. (2016). Intermediate input imports and innovations: Evidence from Chinese firms’ patent filings. Journal of International Economics, 103, 166-183. Lu, D. (2010). Exceptional exporter performance? Evidence from Chinese manufacturing firms. Mimeo. Lu, Y., & Yu, L. (2015). Trade liberalization and markup dispersion: Evidence from China’s WTO accession. American Economic Journal: Applied Economics, 7(4), 221-253.

CE

Melitz, M. J., & Polanec, S. (2015). Dynamic olley-pakes productivity decomposition with entry and exit. Rand Journal of Economics, 46(2), 362-375.

AC

Minetti, R., & Zhu, S. C. (2011). Credit constraints and firm export: Microeconomic evidence from Italy”, Journal of International Economics, 83, 109-125. Naughton, B. (2006). The Chinese economy: Transition and growth, Cambridge, MA: MIT press. Nyström, K. (2007). Patterns and determinants of entry and exit in industrial sectors in Sweden. Journal of International Entrepreneurship, 5(3), 85-110. Olley, S., & Pakes, A. (1996). The dynamics of productivity in the telecommunications equipment industry. Econometrica, 64(6), 1263-1297. Ramondo, N., Rodriguez-Clare, A., & Saborio, M. (2016). Trade, domestic frictions, and scale effects. UC-Berkeley working paper. Schor, A. (2004). Heterogeneous productivity response to tariff reduction: Evidence from Brazilian manufacturing firms. Journal of Development Economics, 75(2) 373-396. 40

ACCEPTED MANUSCRIPT Topalova, P., & Khandelwal, A. (2011). Trade liberalization and firm productivity: The case of India. Review of Economics and Statistics, 93(3), 995-1009. Van Biesebroeck, J. (2005). Exporting raises productivity in Sub-Saharan African manufacturing firms. Journal of International Economics, 67, 373-391. Wakasugi, R., & Zhang, H. (2015). Impacts of the World Trade Organization on Chinese exports. RIETI discussion paper.

PT

Wen, M. (2007). Foreign direct investment, regional market conditions and regional development: A panel study on China. Economics of Transition, 15(1), 125-151.

RI

Wu, W., Rui, O. M., & Wu, C. (2013). Institutional environment, ownership and firm taxation: Evidence from China. Economics of Transition, 21(1), 17-51.

AC

CE

PT E

D

MA

NU

SC

Yu, M. (2015). Processing trade, tariff reductions and firm productivity: Evidence from Chinese Firms. Economic Journal, 125 (585), 943-988.

41

12

20

ACCEPTED MANUSCRIPT

B

0

1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 year SD of output tariffs

Mean of input tariffs (%)

SC

Mean of output tariffs (%)

1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 year

RI

0

2

2

PT

4

6

4

8

6

10

12

8

14

16

10

18

A

SD of input tariffs

AC

CE

PT E

D

MA

NU

Fig.1. Trends in the mean and standard deviation of tariffs during 1998-2007.

42

12

ACCEPTED MANUSCRIPT

24 30 5

10

15 26 2

8

31 1 25 16 12 8

14 11

22

6

19 7

4 23

SC

28

10 139

RI

6

18 29

21

PT

17

3

4

20

4 Marketization index in 1998

NU

2

6

MA

Fig.2. Marketization index by region in 1998 and 2007. Notes: The regions and associated codes are classified as follows: Anhui (1), Beijing (2), Fujian (3), Gansu (4), Guangdong (5), Guangxi (6), Guizhou (7), Hainan (8), Hebei (9), Henan (10), Heilongjiang (11), Hubei (12), Hunan (13), Jilin (14), Jiangsu (15), Jiangxi (16), Liaoning (17), Inner Mongoria (18), Ningxia (19), Qinghai (20), Shandong (21), Shanxi

D

(22), Shaanxi (23), Shanghai (24), Sichuan (25), Tianjin (26), Tibet (27), Xinjiang (28), Yunnan (29), Zhejiang (30),

AC

CE

PT E

Chongqing (31). The marketization index of Tibet in 1998 is missing.

43

ACCEPTED MANUSCRIPT Table 1. The effect of trade liberalization and marketization on firm entry rate at the industry-region-year level. (1)

(2)

(3)

(4)

(5)

(6)

Output tariff

0.0005*** (3.93)

0.0003** (2.25)

0.0004*** (3.08)

0.0002* (1.80)

0.0004*** (3.11)

0.0001* (1.78)

Input tariff

-0.0029***

-0.0025***

-0.0028***

-0.0028***

-0.0009

-0.0007

(-7.46)

(-5.96)

(-7.65)

(-7.67)

(-1.60)

(-1.51)

Output tariff × marketization index

0.0002**

0.0002**

(2.06)

(2.01)

(1.89)

-0.0003**

-0.0006**

-0.0004** 0.0032

Average capital intensity Herfindahl index

(-0.56)

(-0.57)

0.0005***

0.0005***

(4.51)

(4.41)

-0.0198*** (-12.22) -0.0045***

-0.0032**

(-2.68)

(-1.99)

-0.0487*** (-7.84) No

No

Region-year fixed effects

No

No

Year fixed effects

No

No

Industry-year fixed effects R-squared

(-1.99)

(-2.38)

-0.0018

-0.0044

(-0.56)

(-0.57)

(-1.38)

0.0005***

0.0004***

(4.40)

(4.40)

(4.40)

(4.18)

-0.0214***

-0.0213***

-0.0215***

-0.0215***

-0.0164***

(-13.53)

(-13.50)

(-13.62)

(-13.58)

(-9.32)

-0.0032**

-0.0031*

-0.0031*

-0.0044**

(-2.01)

(-1.94)

(-1.96)

(-2.48)

-0.0185***

-0.0184***

-0.0178***

-0.0176***

-0.0064

(-2.92)

(-2.90)

(-2.79)

(-2.76)

(-0.90)

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes Yes

63870

53183

53183

53165

53165

53165

53165

0.14

0.18

0.25

0.25

0.25

0.25

0.28

D

Observations

(-2.03)

0.0005***

MA

Industry fixed effects

0.0001*

-0.0018

RI

(0.98) 0.0006***

SC

Average size

-0.0018

NU

Average profit rate

-0.0018

PT

Input tariff × marketization index Average sales growth rate

(7)

Note: Robust t-values are in parentheses are corrected for clustering at the industry-region level. ***, **, and * denote

PT E

significant at the 1%, 5%, and 10% level respectively. The explanatory variables are lagged by one year to control for

AC

CE

potential simultaneity and reverse causality.

44

ACCEPTED MANUSCRIPT Table 2. The effect of trade liberalization and marketization on firm exit rate at the industry-region-year level. (1)

(2)

(3)

(4)

(5)

(6)

Output tariff

-0.0012*** (-8.75)

-0.0011*** (-8.32)

-0.0010*** (-8.46)

-0.0001 (-0.46)

-0.0010*** (-8.41)

-0.0002 (-0.62)

Input tariff

-0.0025***

-0.0023***

-0.0018***

-0.0019***

-0.0004

-0.0003

(-6.32)

(-5.62)

(-5.31)

(-5.44)

(-1.13)

(-1.02)

Output tariff × marketization index

-0.0002***

-0.0003***

(-4.12)

(-4.41)

(-2.42)

-0.0002**

-0.0002**

-0.0003***

Average sales growth rate

-0.0330*** (-9.17)

(-7.74)

(-7.79)

Average profit rate

-0.0006**

-0.0006**

-0.0006**

(-2.00)

(-2.23)

-0.0295*** (-18.49) -0.0120*** (-7.44) 0.0533*** (8.67)

Herfindahl index No

No

Region-year fixed effects

No

No

Year fixed effects

No

No

Industry-year fixed effects R-squared

(-2.10) -0.0246*** (-7.15) -0.0006**

(-2.23)

(-2.23)

(-2.23)

(-2.47)

-0.0297***

-0.0294***

-0.0297***

-0.0295***

-0.0275***

(-19.15)

(-18.97)

(-19.13)

(-19.00)

(-15.43)

-0.0087***

-0.0089***

-0.0088***

-0.0089***

-0.0070***

(-5.59)

(-5.70)

(-5.62)

(-5.67)

(-3.87)

0.0164***

0.0165***

0.0164***

0.0170***

0.0179**

(2.61)

(2.64)

(2.61)

(2.69)

(2.49)

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

RI

(-7.78)

Yes

63870

53183

53183

53165

53165

53165

53165

0.10

0.14

0.21

0.21

0.21

0.21

0.24

D

Observations

(-2.24) -0.0267*** -0.0006**

MA

Industry fixed effects

(-3.07)

-0.0266*** (-7.76)

SC

Average capital intensity

-0.0002**

-0.0006**

NU

Average size

-0.0267***

PT

Input tariff × marketization index -0.0265***

(7)

Note: Robust t-values are in parentheses are corrected for clustering at the industry-region level. ***, **, and * denote

PT E

significant at the 1%, 5%, and 10% level respectively. The explanatory variables are lagged by one year to control for

AC

CE

potential simultaneity and reverse causality.

45

ACCEPTED MANUSCRIPT Table 3. The effect of trade liberalization on firm exit. Output tariff

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

-0.0039***

-0.0048***

-0.0038***

-0.0045***

-0.0041***

-0.0039***

-0.0043***

-0.0041***

(-5.19)

(-6.17)

(-5.19)

(-6.05)

(-5.58)

(-5.19)

(-5.79)

(-5.60)

-0.0023

-0.0024

-0.0054***

-0.0063***

-0.0023

-0.0043***

-0.0040***

(-1.50)

(-1.51)

(-4.63)

(-4.71)

(-1.50)

(-3.81)

0.0010**

0.0009**

0.0003*

(2.07)

(2.03)

(1.65)

Input tariff

Output tariff× relative productivity

Input tariff× relative productivity

0.0021***

0.0023***

(2.62)

(2.68)

× marketization index Input tariff× relative productivity × marketization index

D E

Output tariff× marketization index

PT

Input tariff× marketization index

E C

Relative productivity

Firm sales growth rate

Firm profit rate

Firm size

Firm capital intensity

-0.0005**

A M

-0.0007**

-0.0007**

I R

SC

0.0001

(-3.71)

(-3.55)

0.0002

0.0002

0.0002

(1.52)

(1.46)

(1.23)

0.0001

0.0002

0.0001

(1.50)

(1.55)

(1.06)

0.0003**

0.0002**

0.0004**

(2.06)

(2.52)

(2.42)

(2.44)

0.0003**

0.0003**

0.0003**

0.0002**

(2.13)

(2.54)

(2.48)

(2.16)

-0.0003

-0.0004*

-0.0005*

-0.0006**

(-1.60)

(-1.67)

(-1.94)

(-2.07)

-0.0004

-0.0004

-0.0003

-0.0004

(-0.86)

(-0.83)

(-0.80)

(-0.74)

-0.0000

-0.0003*

-0.0004*

-0.0011*

-0.0004*

(1.47)

-0.0037***

0.0002**

U N

Output tariff× relative productivity

T P

(9)

(-2.00)

(-2.26)

(-2.25)

(-1.80)

(-0.01)

(-1.73)

(-1.88)

(-1.68)

-0.1708***

-0.1707***

-0.1707***

-0.1707***

-0.1706***

-0.1702***

-0.1685***

-0.1670***

-0.1669***

(-35.12)

(-35.11)

(-35.12)

(-35.11)

(-35.06)

(-34.99)

(-34.30)

(-34.41)

(-34.40)

-0.0084*

-0.0084*

-0.0084*

-0.0084*

-0.0084*

-0.0084*

-0.0084*

-0.0082*

-0.0082*

(-1.73)

(-1.73)

(-1.73)

(-1.73)

(-1.72)

(-1.73)

(-1.73)

(-1.73)

(-1.73)

-0.1498***

-0.1498***

-0.1498***

-0.1498***

-0.1497***

-0.1497***

-0.1495***

-0.1323***

-0.1324***

(-46.79)

(-46.78)

(-46.78)

(-46.78)

(-46.60)

(-46.50)

(-46.17)

(-42.41)

(-42.25)

-0.0636***

-0.0636***

-0.0636***

-0.0636***

-0.0638***

-0.0638***

-0.0641***

-0.0619***

-0.0620***

(-29.50)

(-29.50)

(-29.50)

(-29.50)

(-29.55)

(-29.58)

(-29.64)

(-29.25)

(-29.26)

C A

46

ACCEPTED MANUSCRIPT Herfindahl index

0.0874***

0.0873***

0.0873***

0.0873***

0.0890***

0.0818***

0.0847***

0.0775***

0.0757***

(3.34)

(3.33)

(3.34)

(3.33)

(3.38)

(3.12)

(3.22)

(2.97)

(2.90)

-0.1743***

-0.1246***

(-25.59)

(-7.74)

Firm export indicator

Output tariff× firm export indicator

T P

I R

Input tariff× firm export indicator

Industry fixed effects

Yes

Yes

Yes

Yes

Yes

Region-year fixed effects

Yes

Yes

Yes

Yes

Yes

Year fixed effects

Yes

Yes

Yes

Yes

Pseudo R-squared Log likelihood

655180

655180

655180

0.09

0.09

0.09

-267699

-267696

-267697

D E

A M 655180 0.09

-267696

SC

U N

Industry-year fixed effects Observations

-0.0013

Yes

(-1.55) -0.0026* (-1.89)

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes Yes

655146

655146

655146

655146

655146

0.09

0.09

0.09

0.09

0.09

-267664

-267651

-267586

-266974

-266963

Note: Robust t-values are in parentheses are corrected for clustering at the industry-region level. ***, **, and * denote significant at the 1%, 5%, and 10% level respectively. The explanatory variables are lagged by one year to control for potential simultaneity and reverse causality.

T P E

C C

A

47

ACCEPTED MANUSCRIPT Table 4. The industry-region-year level estimates with controlling for endogeneity: 2SLS.

Input tariff × marketization index

Average sales growth rate

Average profit rate

(4)

Entry rate

Exit rate

Entry rate

Exit rate

0.0008**

-0.0003

(1.97)

(-0.76)

-0.0011*

-0.0006*

(-1.78)

(-1.66)

0.0009**

-0.0002***

0.0006**

-0.0003***

(2.56)

(-4.02)

(2.21)

(-4.13)

-0.0010***

-0.0003**

(-3.02)

(-2.37)

0.0005

-0.0289***

(0.16) 0.0005***

(-2.47)

(-2.27)

0.0016

-0.0343***

(-7.79)

(0.44)

(-8.63)

-0.0006**

0.0005***

-0.0006*

(-2.23)

(5.14)

(-1.78)

-0.0334***

-0.0171***

-0.0272***

(-6.97)

(-9.32)

-0.0049***

-0.0106***

-0.0052***

-0.0101***

(-2.74)

(-5.87)

(-2.95)

(-5.29)

-0.0383***

0.0148**

-0.0440***

0.0207*

(-4.62)

(2.33)

(-4.46)

(1.74)

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

NU

MA

Herfindahl index

Industry fixed effects

D

Region-year fixed effects

Industry-year fixed effects

PT E

Year fixed effects

-0.0003**

(-15.62)

-0.0158***

Average capital intensity

-0.0014**

(-7.94)

(6.02) Average size

PT

Output tariff × marketization index

(3)

RI

Input tariff

(2)

SC

Output tariff

(1)

Yes †



Yes †

30.49†

Kleibergen-Paap rank LM  statistic

33.65

Kleibergen-Paap rank Wald F statistic

38.44†

38.44†

37.56†

37.56†

Anderson-Rubin  2 statistic

20.28†

483.97†

20.67†

479.98†

53165

53165

53165

53165

0.17

0.11

0.18

0.12

2

R-squared

CE

Observations

33.65

30.50

Note: Robust t-values are in parentheses are corrected for clustering at the industry-region level. ***, **, and * denote

AC

significant at the 1%, 5%, and 10% level respectively. †indicates significance of p-value at the 1% level. The explanatory variables are lagged by one year to control for potential simultaneity and reverse causality.

48

ACCEPTED MANUSCRIPT Table 5. The firm level estimates with controlling for endogeneity: IV probit.

Input tariff

Output tariff × relative productivity

Input tariff × relative productivity

(2)

(3)

(4)

-0.0069**

-0.0074**

(-2.06)

(-2.32)

-0.0089**

-0.0050*

(-2.01)

(-1.67)

0.0013*

0.0004

0.0011*

0.0005

(1.91)

(1.12)

(1.88)

(1.21)

0.0030**

0.0011

0.0026*

0.0013

(1.98) Output tariff × relative productivity

(1.54) 0.0005** (2.23)

(1.91)

RI

× marketization index

PT

Output tariff

(1)

Input tariff × relative productivity

Output tariff × marketization index

0.0006**

(2.04)

(1.97)

-0.0009*

-0.0008

(-1.66)

(-1.63)

-0.0009

-0.0007

(-1.02)

(-0.93)

NU

Input tariff × marketization index

Relative productivity

-0.0007

(2.30)

0.0007**

SC

× marketization index

(1.59) 0.0006**

-0.0009

-0.0009

-0.0014*

(-1.38)

(-1.45)

(-1.72)

-0.1692***

-0.1700***

-0.1688***

(-35.19)

(-33.00)

(-34.28)

(-32.93)

-0.0084*

-0.0199**

-0.0098*

-0.0205**

(-1.72)

(-2.06)

(-1.89)

(-2.09)

-0.1502***

-0.1491***

-0.1394***

-0.1429***

(-46.00)

(-44.98)

(-39.27)

(-44.04)

-0.0630***

-0.0644***

-0.0689***

-0.0681***

(-28.26)

(-29.52)

(-30.62)

(-30.55)

0.0860***

0.0776***

0.3795***

0.1040***

(3.21)

(2.82)

(12.89)

(3.24)

Yes

Yes

Region-year fixed effects

Yes

Yes

Yes

Yes

Year fixed effects

Yes

Yes

Yes

Yes

Yes

Yes

MA

(-1.11)

-0.1708***

Firm sales growth rate

D

Firm profit rate

PT E

Firm size

Firm capital intensity

AC

Industry fixed effects

CE

Herfindahl index

Industry-year fixed effects Observations Log likelihood

655146

655146

655146

655146

-3558787

-3625403

-3469503

-2877061

Note: Robust t-values are in parentheses are corrected for clustering at the industry-region level. ***, **, and * denote significant at the 1%, 5%, and 10% level respectively. The explanatory variables are lagged by one year to control for potential simultaneity and reverse causality.

49

ACCEPTED MANUSCRIPT Table 6. The DID estimation results. (2)

(3)

(4)

(5)

(6)

(7)

(8)

Entry rate

Entry rate

Entry rate

Entry rate

Exit rate

Exit rate

Exit rate

Exit rate

-0.0006*** (-3.26)

InTat01×Post02 Output tariff

0.0005*** (3.26)

0.0020***

0.0002

0.0011***

0.0004

(5.34)

(0.87)

(2.98)

(1.24)

0.0002**

0.0003**

-0.0009***

-0.0009***

(2.08)

(2.16)

(-8.29)

(-8.25)

-0.0025***

-0.0026***

-0.0010***

(-7.33)

(-7.42)

(-3.13)

OutTat01×Post02

-0.0002***

× marketization index

(-2.85)

InTat01×Post02

0.0003* (1.95) -0.0019

-0.0015

-0.0020

-0.0015

rate Average profit rate

(-0.64)

(-0.49)

(-0.66)

(-0.49)

0.0004***

0.0005***

0.0004***

(5.46)

(4.49)

(5.49)

-0.0215***

-0.0220***

-0.0213***

(-14.18)

(-14.26)

(-13.99)

-0.0024

-0.0032**

-0.0024

intensity Herfindahl index

(-3.20) 0.0002** (2.07) 0.0001** (1.97)

-0.0265***

-0.0274***

-0.0265***

(-8.31)

(-7.76)

(-8.32)

(-7.76)

0.0005***

-0.0005**

-0.0006**

-0.0005**

-0.0006**

(4.51)

(-2.46)

(-2.22)

(-2.46)

(-2.22)

-0.0221***

-0.0298***

-0.0305***

-0.0296***

-0.0306***

(-14.34)

(-20.18)

(-20.17)

(-20.01)

(-20.16)

-0.0032**

-0.0086***

-0.0084***

-0.0086***

-0.0084***

MA

Average capital intensity

-0.0011***

-0.0273***

NU

Average sales growth

SC

× marketization index

Average size

-0.0003 (-1.52)

PT

Input tariff

-0.0001 (-1.02)

RI

OutTat01×Post02

(1)

(-1.59)

(-2.04)

(-1.61)

(-2.00)

(-5.84)

(-5.44)

(-5.85)

(-5.43)

-0.0202***

-0.0209***

-0.0201***

-0.0201***

0.0153**

0.0154**

0.0154**

0.0156**

(-3.35)

(-3.30)

(-3.22)

(2.55)

(2.49)

(2.56)

(2.52)

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Year effectsfixed effects Region-year fixed effects

Yes

Yes

Yes

Yes

Yes

Yes

Yes

0.30

Yes

Yes

Yes

Yes

Yes

53183

Yes

53165

Yes

53165

53183

53183

53165

53165

0.30

0.30

0.30

0.24

0.25

0.25

0.25

Note: Robust t-values are in parentheses are corrected for clustering at the industry-region level. ***, **, and * denote significant at the 1%, 5%, and 10% level respectively. The explanatory variables are lagged by one year to control for

CE

potential simultaneity and reverse causality.

AC

R-squared

Yes 53183

Yes

PT E

Observations

D

(-3.32) Industry-region fixed

50

ACCEPTED MANUSCRIPT Table 7. Further robustness checks (1)

(2)

(3)

(4)

(5)

(6)

Entry rate

Entry rate

Entry rate

Entry rate

Entry rate

Exit rate

0.0001*

0.0002**

0.0002**

0.0002**

0.0002**

(1.91)

(1.97)

(1.99)

(1.98)

(1.98)

-0.0005**

-0.0004**

-0.0004**

-0.0004**

(-2.33)

(-2.26)

(-2.23)

(-2.20)

-0.0045

-0.0045

-0.0045

(-1.41)

(-1.41)

(-1.40)

Average profit rate

0.0004***

0.0004***

0.0004***

(4.15)

(4.16)

(4.15)

(4.15)

(4.15)

(-2.47)

(-2.47)

(-2.48)

(-2.47)

(-2.47)

Average size

-0.0164***

-0.0165***

-0.0165***

-0.0165***

-0.0165***

-0.0276***

-0.0276***

-0.0276***

-0.0276***

-0.0276***

(-9.29)

(-9.30)

(-9.31)

(-9.31)

(-9.32)

(-15.45)

(-15.42)

(-15.44)

(-15.45)

(-15.45)

-0.0042**

-0.0042**

-0.0042**

-0.0043**

-0.0043**

-0.0070***

-0.0071***

-0.0071***

-0.0071***

-0.0071***

(-2.34)

(-2.34)

(-2.35)

(-2.37)

(-2.37)

(-3.88)

(-3.89)

(-3.90)

(-3.90)

(-3.90)

-0.0063

-0.0064

-0.0064

-0.0066

-0.0065

0.0175**

0.0175**

0.0174**

0.0173**

0.0173**

(-0.89)

(-0.89)

(-0.90)

(-0.92)

(-0.91)

(2.42)

(2.42)

(2.41)

(2.40)

(2.39)

0.0004

0.0001

0.0001

0.0001

0.0001

0.0003

0.0004

0.0003

0.0003

0.0003

(1.23)

(0.30)

(0.34)

(0.38)

(0.36)

(0.91)

(1.15)

(0.88)

(0.85)

(0.85)

-0.0003

-0.0001

-0.0005

-0.0008

-0.0008

0.0002

0.0005

0.0004

0.0003

0.0003

(0.19)

Herfindahl index

Output tariff × pgdp

Input tariff × pgdp

(-0.36) Output tariff × open

Input tariff × open

Year fixed effects Industry-year fixed effects Observations R-squared

-0.0002**

-0.0002**

-0.0002**

-0.0003**

-0.0003**

(-2.24)

(-2.09)

(-2.15)

(-2.11)

(-2.36)

(-2.35)

-0.0045

-0.0044

-0.0251***

-0.0250***

-0.0250***

-0.0250***

-0.0250***

(-1.40)

(-1.39)

(-7.29)

(-7.28)

(-7.28)

(-7.28)

(-7.28)

0.0004***

0.0004***

-0.0006**

-0.0006**

-0.0006**

-0.0006**

-0.0006**

PT

-0.0004**

(-0.76)

(-0.77)

0.0035**

0.0035**

Exit rate

(0.57)

(0.42)

(0.30)

(0.30)

-0.0006*

-0.0006*

-0.0005*

-0.0005*

(3.11)

(3.12)

(2.39)

(2.37)

(-1.71)

(-1.70)

(-1.69)

(-1.67)

0.0027

0.0031

0.0034

0.0032

-0.0025

-0.0024

-0.0002

-0.0002

(0.94)

(1.09)

(1.13)

(1.11)

(-0.94)

(-0.91)

(-0.04)

(-0.05)

-0.0000

-0.0000

-0.0001

0.0001

0.0001

0.0001

(-0.22)

(-0.28)

(-0.32)

(0.71)

(0.63)

(0.59)

0.0006

0.0007

0.0007

0.0001

0.0003

0.0003

(1.37)

(1.38)

(0.32)

(0.56)

(0.60)

-0.0002

-0.0002

0.0007**

0.0006*

D

Region-year fixed effects

(-1.79)

(-0.51)

PT E CE AC

Input tariff × competition

-0.0001*

(-1.94)

0.0034***

(1.36)

Output tariff × competition

-0.0001*

(-2.21)

(-0.09)

Input tariff × fdi

Input tariff × processshare

-0.0002**

(-2.26)

0.0034***

Output tariff × fdi

Output tariff × processshare

-0.0002**

(-2.40)

RI

Average capital intensity

Exit rate

(10)

-0.0002**

SC

Average sales growth rate

(9)

Exit rate

NU

Input tariff × marketization index

(8)

Exit rate

MA

Output tariff × marketization index

(7)

(-0.13)

(-0.14)

(1.97)

(1.94)

0.0070*

0.0072*

0.0029**

0.0029**

(1.95)

(2.37)

(1.91)

(2.36)

0.0004

-0.0004*

(0.40)

(-1.78)

0.0039

0.0005

(1.16)

(0.13)

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

53165

53165

53165

53165

53165

53165

53165

53165

53165

53165

0.28

0.28

0.28

0.28

0.28

0.24

0.24

0.24

0.24

0.24

Note: Robust t-values are in parentheses are corrected for clustering at the industry-region level. ***, **, and * denote significant at the 1%, 5%, and 10% level respectively. The explanatory variables are lagged by one year to control for potential simultaneity and reverse causality.

51

ACCEPTED MANUSCRIPT Table 8. The role of subcomponents of the marketization index

Input tariff × MI1

(2)

(3)

(4)

(5)

(6)

Entry rate

Entry rate

Entry rate

Entry rate

Entry rate

Exit rate

0.0003**

-0.0005**

(2.27)

(-2.51)

-0.0004*

-0.0002**

(-1.79)

(-2.07)

Output tariff × MI2

0.0002**

Input tariff × MI2

(7) Exit rate

(8) Exit rate

(9) Exit rate

Exit rate

-0.0003**

(2.01)

(-2.08)

-0.0002

-0.0002*

(-1.49)

(-1.93)

Output tariff × MI3

0.0001

Input tariff × MI3

0.0002

-0.0001*

RI

(1.26)

Output tariff × MI4

-0.0000 (-0.65) -0.0003 (-1.41)

Output tariff × MI5

(-0.20) -0.0001 (-1.56) 0.0001 (0.34)

NU

Input tariff × MI4

(-1.68)

-0.0000

SC

(1.11)

0.0001*

-0.0000

(1.77)

(-0.23)

-0.0009**

-0.0005****

MA

Input tariff × MI5

(-2.55)

Average sales growth rate

(10)

PT

Output tariff × MI1

(1)

(-2.69)

-0.0044

-0.0044

(-1.39)

(-1.39)

(-1.38)

0.0004***

0.0004***

0.0004***

(4.16)

(4.17)

(4.15)

(4.18)

Average size

-0.0165***

-0.0164***

-0.0164***

-0.0164***

(-9.32)

(-9.30)

(-9.31)

(-9.26)

(-9.24)

(-15.46)

(-15.46)

(-15.52)

(-15.48)

(-15.47)

Average capital intensity

-0.0043**

-0.0042**

-0.0043**

-0.0042**

-0.0042**

-0.0070***

-0.0070***

-0.0070***

-0.0069***

-0.0069***

Herfindahl index

PT E

Average profit rate

-0.0045

-0.0045

-0.0250***

-0.0250***

-0.0250***

-0.0250***

-0.0250***

(-1.41)

(-1.40)

(-7.26)

(-7.26)

(-7.26)

(-7.27)

(-7.26)

0.0004***

0.0004***

-0.0006**

-0.0006**

-0.0006**

-0.0006**

-0.0006**

(4.17)

(-2.47)

(-2.47)

(-2.47)

(-2.47)

(-2.47)

-0.0163***

-0.0276***

-0.0276***

-0.0277***

-0.0276***

-0.0276***

D

-0.0044

(-2.38)

(-2.36)

(-2.38)

(-2.35)

(-2.36)

(-3.84)

(-3.85)

(-3.84)

(-3.83)

(-3.82)

-0.0075

-0.0072

-0.0079

-0.0065

-0.0066

0.0173**

0.0175**

0.0174**

0.0179**

0.0180**

(-1.01)

(-1.12)

(-0.91)

(-0.94)

(2.40)

(2.43)

(2.42)

(2.48)

(2.50)

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Year fixed effects

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Observations R-squared

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

53165

Yes

53165

53165

53165

53165

53165

53165

53165

53165

53165

0.28

0.28

0.28

0.28

0.28

0.24

0.24

0.24

0.24

0.24

AC

Industry-year fixed effects

CE

(-1.06)

Region-year fixed effects

Note: Robust t-values are in parentheses are corrected for clustering at the industry-region level. ***, **, and * denote significant at the 1%, 5%, and 10% level respectively. The explanatory variables are lagged by one year to control for potential simultaneity and reverse causality.

52

ACCEPTED MANUSCRIPT Table 9. The differences of average productivity by types of firms. Exiting vs. surviving

Entering vs. exiting

(1)

(2)

(3)

-0.1420***

-0.3509***

0.2339***

(-6.73)

(-13.49)

(12.08)

-0.1862***

-0.4056***

0.3748***

(-8.97)

(-13.85)

(14.20)

-0.0343

-0.3231***

0.1592***

(-1.51)

(-12.91)

(8.81)

The regions with higher marketization level

-0.1340***

-0.3128***

(-6.93)

(-13.95)

(11.85)

The regions with lower marketization level

-0.1652***

-0.3819***

0.2570***

(-12.10)

(10.87)

The period before China’s WTO accession

The period after China’s WTO accession

RI

All firms

PT

Entering vs. surviving

(-6.62)

0.1974***

SC

Note: Robust t-values are in parentheses are corrected for clustering at the firm level, the numbers outside parentheses represent average productivity difference between different types. ***, **, and * denote significant at the 1%, 5%, and 10%

AC

CE

PT E

D

MA

NU

level respectively. Industry, region and year fixed effects are included in all regressions.

53

ACCEPTED MANUSCRIPT Table 10. Decomposition of productivity growth. Between

Entry

Exit

(1)

(2)

(3)

(4)

(5)

0.1337

0.0783

-0.0023

0.0082

0.0494

[58.58]

[-1.70]

[6.13]

[36.96]

0.0771

0.0028

-0.0023

0.0397

[65.72]

[2.39]

[-1.97]

[33.84]

0.0884

-0.0215

0.0090

0.0667

[61.97]

[-15.07]

[6.33]

[46.76]

0.0802

-0.0073

0.0093

0.0625

[55.45]

[-5.05]

[6.41]

[43.18]

0.0691

0.0060

-0.0082

0.0399

[5.62]

[-7.67]

[37.40]

0.1173

Post China’s WTO accession period

0.1427

High marketization region

0.1446

Low marketization region

0.1068

AC

CE

PT E

D

MA

NU

SC

[64.71]

PT

Before China’s WTO accession period

Within

RI

All firms

Total

54

ACCEPTED MANUSCRIPT Table 11. The effect of trade liberalization and marketization on aggregate productivity. Between

Entry

Exit

(1)

(2)

(3)

(4)

(5)

(6)

-0.0055*

-0.0001

-0.0003

0.0000

0.0004

-0.0002

(-1.88)

(-0.26)

(-0.87)

(0.62)

(1.36)

(-1.38)

-0.0230**

-0.0008

-0.0003*

-0.0000

0.0010

-0.0015*

(-2.40)

(-0.30)

(-1.71)

(-0.20)

(0.13)

(-1.72)

-0.0010**

-0.0005**

0.0003

-0.0002

-0.0006**

(-2.09)

(-2.17)

(1.23)

(-1.58)

(-2.38)

-0.0041***

-0.0018***

-0.0002

-0.0009*

-0.0012**

Output tariff × marketization index

Input tariff × marketization index

(-3.47)

-0.1407***

-0.0201

(-4.05)

(-4.12)

Industry fixed effect

Yes

Yes

Region-year fixed effects

Yes

Yes

Year fixed effects

Yes

Yes

16410

16410

0.09

0.10

Observations R-squared

(-1.17)

(-1.95)

(-2.46)

-0.0035**

0.0105

-0.1276***

(-0.99)

(-2.36)

(1.60)

(-3.65)

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

16410

16410

16410

16410

0.15

0.04

0.06

0.10

NU

(-2.80) -0.1364***

Herfindahl index

PT

Within

RI

Input tariff

Aggregate

SC

Output tariff

Aggregate

Note: Robust t-values are in parentheses are corrected for clustering at the industry-region level. ***, **, and * denote

MA

significant at the 1%, 5%, and 10% level respectively. The explanatory variables are lagged by one year to control for

AC

CE

PT E

D

potential simultaneity and reverse causality.

55

ACCEPTED MANUSCRIPT Appendix A. Measure of Total Factor Productivity (TFP) We use the augmented Olley and Pakes (1996)’s method (the OP method hereafter) to calculate total factor productivity. Following Amiti and Konings (2007) as well as Fan et al. (2015), we consider a Cobb-Douglas production function as estimation specification: ln Yijt  0   k ln Kijt  l ln Lijt  ijt

(A1)

PT

where Yijt , K ijt , and Lijt denote firm i’s value added, capital and labour in industry j in year t, respectively. The main characteristic of OP method is to use firm investment as proxy for unobserved

RI

firm-specific productivity shocks, therefore it is necessary to construct a real investment variable

SC

when using this method. Specially, we construct the real investment variable by adopting the perpetual inventory method and using the depreciation ratio suggested by Brandt et al. (2012).

NU

In this paper, we extend Olley and Pakes’s (1996) method in threefold. First, following Biesebroeck (2005), we add firm’s export decision into the investment function, that is to say, we take

MA

the export decision as one of the key variables for firm’s investment function. Second, since China’s accession to the WTO can bring a positive demand shock which will push firms to expand their scale of production, and then will in turn increase the simultaneous bias when calculating TFP. To address

D

such a concern, we take the event of China’s accession to WTO into account, that is, including a WTO

PT E

dummy in the investment function. Last but not least, as our firm data set is from 1998 to 2007, we thus include a RMB exchange rate reform dummy (i.e., take one for a year after 2005 and zero for

CE

before) in the investment function.

It is worth noting that we adopt several different price deflators to deflate the nominal terms. In particular, we use output deflators and input deflators provided by Brandt et al. (2012) to deflate

AC

firm’s output and input respectively, furthermore, following Li et al. (2015), we deflate capital stock using the regional fixed asset price index obtained from the National Bureau of Statistics of China.

56

ACCEPTED MANUSCRIPT Table A1. Estimated coefficients of the production function. Labour

Capital

Processing of foods (13)

0.494

0.081

Manufacture of foods (14)

0.577

0.061

Beverages (15)

0.529

0.058

Textiles (17)

0.417

0.033

Apparel (18)

0.593

0.110

Leather (19)

0.544

0.052

Timber (20)

0.487

0.107

PT

Industry (Code)

Furniture (21)

0.630

0.034

0.438

0.164

0.445

0.187

0.544

0.105

0.205

0.061

0.307

0.089

0.468

0.186

0.390

0.030

0.330

0.116

0.416

0.105

0.333

0.047

0.454

0.130

0.381

0.167

0.459

0.127

0.378

0.097

0.396

0.104

0.507

0.101

0.438

0.064

0.504

0.157

0.409

0.052

Manufacture of artwork (42)

0.508

0.081

Waste resources recycling processing (43)

0.426

0.129

Paper (22)

RI

Printing (23) Articles for culture and sports (24)

SC

Petroleum (25) Raw chemicals (26) Medicines (27)

NU

Chemical fibres (28) Rubber (29) Plastics (30)

MA

Non-metallicminerals (31) Smelting of ferrous metals (32) Smelting of non-ferrous metals (33) Metal (34)

D

General machinery (35)

Transport equipment (37) Electrical machinery (39) Communication equipment (40)

AC

CE

Measuring instruments (41)

PT E

Special machinery (36)

57

ACCEPTED MANUSCRIPT Table A2. Robustness checks: 2-digit CIC industry.

Input tariff Average sales growth rate Average profit rate

(2)

Entry rate

Exit rate

0.0006** (2.08)

-0.0008*** (-3.71)

-0.0039***

-0.0040***

(-3.86)

(-3.90)

0.0065*

-0.0201*

(1.67)

(-1.92)

0.0031

-0.0082**

(1.48) Average size

-0.0154***

RI

(-3.17)

PT

Output tariff

(1)

Herfindahl index Industry fixed effects

-0.0131**

(-1.72)

(-2.29)

-0.0384**

0.0181

(-2.52)

(1.15)

Yes

Yes

Yes

Yes

NU

Region-year fixed effects Year fixed effects Observations

MA

R-squared

(-6.19)

-0.0028*

SC

Average capital intensity

(-2.30) -0.0267***

Yes

Yes

5366

5366

0.32

0.27

Note: Robust t-values are in parentheses are corrected for clustering at the industry-region level. ***, **, and * denote significant at the 1%, 5%, and 10% level respectively. The explanatory variables are lagged by one year to control for

AC

CE

PT E

D

potential simultaneity and reverse causality.

58

ACCEPTED MANUSCRIPT Table A3. Robustness checks: Controlling for non-tariff barriers. (1)

(2)

Entry rate

Exit rate

Output tariff

0.0004*** (2.95)

0.0010*** (8.41)

Input tariff

-0.0027***

-0.0018***

(-7.38)

(-5.18)

-0.0018

-0.0266***

Average sales growth rate

(-7.75)

Average profit rate

0.0005***

-0.0006**

Average size

-0.0213***

Average capital intensity

-0.0033**

-0.0088***

(-2.02)

(-5.63)

(4.43)

SC

RI

(-13.49)

Herfindahl index

Industry fixed effects Region-year fixed effects

MA

Year fixed effects Observations R-squared

(-2.23) -0.0296*** (-19.11)

-0.0178***

0.0170***

(-2.85)

(2.70)

0.0004

-0.0007

(0.17)

(-0.58)

Yes

Yes

Yes

Yes

Yes

Yes

53183

53183

0.25

0.21

NU

NTBsDum

PT

(-0.58)

Note: Robust t-values are in parentheses are corrected for clustering at the industry-region level. ***, **, and * denote

AC

CE

PT E

potential simultaneity and reverse causality.

D

significant at the 1%, 5%, and 10% level respectively. The explanatory variables are lagged by one year to control for

59

ACCEPTED MANUSCRIPT Table A4. LPM and logit estimate with firm fixed effect.

Input tariff

Output tariff× relative productivity

Input tariff× relative productivity

LPM

logit

logit

(1)

(2)

(3)

(4)

-0.0010***

-0.0008**

-0.0039***

-0.0030**

(-2.62)

(-2.43)

(-2.73)

(-2.10)

-0.0035***

-0.0024***

-0.0131**

-0.0104**

(-3.86)

(-2.63)

(-2.12)

(-2.01)

0.0003**

0.0001*

0.0005*

-0.0002

(1.99)

(1.67)

(1.69)

(-1.26)

0.0010**

0.0000

(2.33)

× marketization index

× marketization index

Input tariff× marketization index

-0.0002***

MA

Relative productivity

Herfindahl index

PT E

AC

Firm fixed effects

CE

Firm capital intensity

Year fixed effects Observations R-squared

(1.39) 0.0005** (2.28)

0.0002**

0.0007**

(2.45)

(2.39)

-0.0001

-0.0006**

(-0.80)

(-2.03)

-0.0002*

-0.0004

(-1.68)

(-1.19)

-0.0001*

-0.0012**

-0.0005

(-5.62)

(-1.66)

(-2.15)

(-1.19)

-0.0222***

-0.0221***

-0.2171***

-0.2150***

(-16.31)

(-16.17)

(-20.92)

(-20.60)

-0.0003*

-0.0003*

-0.0061

-0.0063

(-1.74)

(-1.74)

(-1.59)

(-1.61)

-0.0595***

-0.0590***

-0.4043***

-0.4090***

(-25.78)

(-25.71)

(-19.21)

(-19.38)

-0.0121***

-0.0122***

-0.1104***

-0.1107***

(-9.33)

(-9.43)

(-9.20)

(-9.20)

0.0170

0.0077

0.0745*

0.0728*

(1.34)

(0.61)

(1.74)

(1.70)

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

655,180

655,146

655,180

655,146

0.590

0.590 -45380

-45290

D

Firm sales growth rate

0.0008

(2.62)

NU

Output tariff× marketization index

Firm profit rate

(2.18)

SC

Input tariff× relative productivity

Firm size

(0.86) 0.0001***

0.0014**

RI

Output tariff× relative productivity

PT

Output tariff

LPM

Log likelihood

Note: Robust t-values are in parentheses are corrected for clustering at the industry-region level. ***, **, and * denote significant at the 1%, 5%, and 10% level respectively. The explanatory variables are lagged by one year to control for potential simultaneity and reverse causality.

60

ACCEPTED MANUSCRIPT Table A5. Robustness to alternative productivity estimations. Aggregate TFP: OLS

Aggregate labor

method

productivety

Aggregate TFP: LP method

Input tariff

(2)

(3)

(4)

(5)

(6)

-0.0045*

0.0001

-0.0017*

0.0015

-0.0009

-0.0003

(-1.83)

(0.18)

(-1.67)

(1.21)

(-1.62)

(-1.06)

-0.0236***

-0.0098*

-0.0234***

-0.0130*

-0.0147**

-0.0084

(-2.83)

(-1.68)

(-3.27)

(-1.77)

(-2.57)

(-1.57)

Output tariff × marketization index

-0.0008*

-0.0006*

(-0.64)

-0.0024**

-0.0018*

-0.0011**

(-1.99)

(-1.83)

(-2.08)

-0.2444***

-0.1468***

-0.1466***

-0.0965***

-0.0966***

(-5.01)

(-4.94)

(-4.20)

(-4.17)

(-2.62)

(-2.61)

Industry fixed effects

Yes

Yes

Yes

Yes

Yes

Yes

Region-year fixed effects

Yes

Yes

Yes

Yes

Yes

Yes

Year fixed effects

Yes

Yes

Yes

Yes

Yes

Yes

16410

16410

16410

16410

16410

16410

0.05

0.05

0.06

0.06

0.06

0.06

Observations R-squared

SC

-0.2465***

NU

Herfindahl index

(-1.73)

RI

(-1.85) Input tariff × marketization index

-0.0001

PT

Output tariff

(1)

MA

Note: Robust t-values are in parentheses are corrected for clustering at the industry-region level. ***, **, and * denote significant at the 1%, 5%, and 10% level respectively. The explanatory variables are lagged by one year to control for

AC

CE

PT E

D

potential simultaneity and reverse causality.

61

ACCEPTED MANUSCRIPT Highlights 1. This paper investigates the impact of trade liberalization on firm dynamics and productivity in the context of dramatic tariff reductions after China’s accession to the WTO. 2. This paper examines whether the impact of trade liberalization on firm dynamics depends on institutional environment where the firms are located.

PT

3. This paper finds regional marketization strengthens the impact of trade liberalization on firm dynamics.

RI

4. This paper finds firm dynamics effect is an important channel through which trade liberalization

AC

CE

PT E

D

MA

NU

SC

fosters productivity growth, and domestic market reform strengthens such an impact.

62