Exchange Rate Appreciations and the Distribution of Productivity: Is Importing Inputs Sustainable for Emerging Countries?

Abstract

A static model is used to investigate how the exchange rate movement affect the distribution of productivity within an industry. Quantile regression is then used to empirically test the effects of RMB exchange rate appreciation on the distribution of labor productivity within industries. Based on China’s manufacturing micro-enterprise survey data from 1998 to 2007, we characterize how exchange rate changes affect the distribution of productivity through three mechanisms. We find that the exchange rate appreciation increases the dispersion of the productivity distribution and decrease the efficiency of resource allocation and aggregate productivity. The distribution of the import intensity may be the main cause for the increase in the productivity dispersion and the deterioration of the industrial resource allocation efficiency, which implies that foreign inputs improve the mean productivity of firms but decrease the resource allocation efficiency. China should tradeoff the gain in productivity and loss in allocation efficiency when it aims to implement a more elastic RMB exchange rate regime.

1. Introduction

For an open economy, the exchange rate may be the most important price in international economic activities. Appreciations of the local currency increase the relative prices of domestic commodities, thereby affecting the exporting prices of domestic plants’ products and the importing costs of domestic firms. Therefore, exchange rate fluctuations would inevitably affect the allocation of resources within industries, which, in turn, affects the aggregate productivity of the country.

Industrial resource allocation efficiency improvement is an important source of aggregate productivity growth. Hsieh [1] found that if the resource allocation efficiency of China and India is same as that of America, the gain in manufacturing productivity would be 30–50% in China and 40–60% in India. An improvement in resource allocation efficiency in an industry usually means the exit of smaller less-productive firms, the entry of more productive enterprises, and the transfer of market shares and production factors to more productive organizations, and finally the growth of aggregate productivity [2]. There are many distortions and frictions in product and factor markets in China, which lead to resource allocation inefficiency within industries, and lower China’s aggregate productivity. Yin [3] found that distortions in the factor and product markets can both lead to resource misallocation, and the total factor productivity (TFP) loss due to product market distortions is on average 40.3%. Therefore, understanding how movements in the RMB exchange rates affect the dispersion of productivity is important for understanding the resource allocation efficiency, dynamics in aggregate productivity growth, and high-quality development strategies of China.

However, the related literature on industrial resource allocation efficiency that result from movements in the exchange rate is scarce Liu [4] investigated the effect of the RMB exchange rate movement on the resource allocation efficiency from the perspective of “pricing by market” of export enterprises, and found that an appreciation of the RMB exchange rate decreases the dispersion of markup and improved the efficiency in resource allocation. Liu [5] studied the effect of exchange rate movement on new entrants or exit firms (expansion margin) and continued firms (intensive margin) separately, and found that the allocation efficiency increase and markup dispersion decrease with RMB appreciation. Mao [6] focused on the mechanism by which exchange rate fluctuations affect markup dispersion, and found that changes in the RMB exchange rate had no significant impact on markup dispersion through export channels, while import competition and import intermediate channels would affect markup dispersion significantly. Although these studies analyzed the response of industry resource allocation efficiency to RMB exchange rate movement, they all use the markup dispersion to measure the resource allocation efficiency.

Productivity dispersion is also usually used to measure the efficiency in resource allocation in the literature, and which can afford more information on the exchange rate fluctuation effects. However, few studies have analyzed the effects of exchange rate changes on the productivity dispersion. Tomlin [7] used Canadian data to analyze how the exchange rate appreciation affects the productivity dispersion of the Canadian manufacturing industry, and found that exchange rate appreciation reduced the intra-industry productivity dispersion. However, this empirical study analyzes the impact of exchange rate changes in developed countries on productivity dispersion. The comparative advantage of China’s manufacturing industry is obviously different from that of developed countries such as Canada, and the Chinese position in the global supply chain is also very different from that of developed countries. Therefore, the empirical evidence that use developed countries’ data may not be relevant to China. We have not found related literatures on the effect of RMB exchange rate changes on the productivity dispersion of China’s manufacturing industry.

The current literature related to the effect of RMB exchange rate movements on productivity mainly focus on the mean productivity. According to Aghion [8], real exchange rate volatility could affect the productivity growth, and the effect depends on financial development. Yussif [9] used Ghana’s data to investigate the relationship between exchange rate volatility and international trade. The results showed that exchange rate volatility has a negative impact on export performances in the Ghanaian economy, but has no significant positive effect on imports. Liu [10] found that exchange rate appreciation is not positive to enterprise R&D investment, but the sample only include export enterprises. Zhang [11] and Wang [12] also investigated manufacturing export enterprises’ productivity and found that the appreciations of the RMB improved the quality of export products, but their samples include only firms in the manufacturing sector involved in trade. According to Xu [13], the net effect of RMB appreciation on the productivity of manufacturing enterprises is positive, and this effect is significantly heterogeneous with trade patterns, ownership, and technological levels. Yu [14] used the balanced panel data of manufacturing enterprises and found that the exchange rate appreciation made the manufacturing industry improve the productivity of enterprises through the transmission mechanism of market competition. Alfaro [15] analyzed that the promotion effect of the RMB exchange rate appreciations on the productivity of Chinese enterprises is related to export orientation. Cao [16] found the appreciations of RMB optimize the allocation of internal factors to avoid to be eliminated by the market competition and improve the TFP of firms. These studies all analyze the impact of RMB exchange rate changes on the productivity of enterprises in the manufacturing industry, focusing on the first moment of productivity. If the average productivity of firms within an industry increases, and the dispersion of industry productivity decreases, aggregate productivity may also decrease. Therefore, studying the impact of RMB exchange rate movements on the industry productivity dispersion has important theoretical and empirical meaning for a comprehensive assessment of the effect of exchange rates on gross productivity.

In this paper, we examine the relationship between movements in exchange rate and the distribution of productivity within industries. We explore three channels in a single framework: (1) export sales; (2) imported inputs; and (3) competition in the domestic market.

For export enterprises, an appreciation of the exchange rate leads to an increase in the price of export products, and the competitiveness of the export enterprises in foreign markets decreases, which leads to a decline in the production scale of the enterprises. Under the condition of increasing returns to scale, the productivity of the enterprises declines. Ekholm [17] found that the effect of exchange rate changes on the income of exporting enterprises is proportional to the share of the export value of the enterprises in total output. Bernard [18] found that the lower the marginal cost of the firms, the higher the share of export value. Therefore, for enterprises within the industry, the higher the productivity of the enterprise, the greater the negative impact of exchange rate appreciation on enterprise productivity through export sales channels. For non-exporting firms with lower productivity, the direct impact of exchange rate appreciation through the export channel is smaller. Therefore, exchange rate appreciation reduces industry productivity dispersion through the export sales channel.

Halpern [19] theoretically explained two mechanisms by which intermediate goods import have a positive effect on enterprise productivity: (1) the quality of imported intermediate goods is generally higher than that of domestic enterprises, and the use of imported intermediate goods can improve domestic enterprises product quality; and (2) there is incomplete substitution between the imported intermediate inputs and domestic ones. Imported intermediate input can ease the production constraints of domestic enterprises and improve firms’ production technology domestically. Brandt [20] empirically proved that importing intermediate inputs can improve the productivity of Chinese manufacturing enterprises. Due to the fixed cost of importing inputs, the higher the productivity of enterprises, the more varieties and quantities of intermediate inputs the firm import, the greater the promotion effect of the imported intermediate inputs on enterprise productivity. Therefore, when the exchange rate appreciates, the relative price of the imported input falls, and firms with a higher productivity can import a higher quantity and variety of higher-quality intermediate inputs, which increases the firms’ productivity. However, low-productivity enterprises are less competitive due to their inability to import, so their productivity declines. Generally, the proportion of FDI in higher productivity enterprises is higher, which strengthens the productivity promotion effect of imported intermediate inputs on high productivity enterprises [19]. Therefore, exchange rate appreciation increases the dispersion of industry productivity through the import of intermediate inputs.

When the exchange rate appreciates, the relative prices of the domestically imported products of foreign enterprises will decrease, the competitiveness of domestic enterprises will decline, low-productivity enterprises will withdraw from the market, and the market entry threshold will increase, which is the tail truncation effect [21]. Yu [14] and Tomlin [7] used data from China and Canada to prove that exchange rate appreciation makes low-efficiency firms exit, leading to an increase in average firm productivity. Therefore, exchange rate appreciation reduces the dispersion of industry productivity and increases the efficiency of resource allocation through the domestic market competition channel.

The impact of exchange rate changes on industry productivity dispersion depends on the relative magnitude of these three channel effects. China’s imports of intermediate products account for 74%, and that of the final products account for only 4%. If capital goods are included in intermediate products, then the proportion of intermediate imports in the total imports is as high as 93%. In fact, according to the analysis of the trade structure of the left and right tails of the industry productivity distribution, the ratio of the export dependence of the right tail (high productivity end) to the left tail (low productivity end) is 1.29 times, and the ratio of the import dependence of intermediate goods is 1.29 times. It is 2.8 times, and the ratio of FDI shares is 5.33 times. It can be seen from the above data that the difference in the degree of import dependence of intermediate goods between high-productivity enterprises and low-productivity enterprises is far greater than the difference in the degree of export dependence. Therefore, the main path through which RMB appreciation affects the productivity dispersion may be the import of intermediate goods.

Compared with the existing related research, the marginal contribution of our paper may be reflected in the following aspects: First, the degree to which the RMB exchange rate affects the dispersion of productivity at the industry level, the difference across industries, and the role of trade patterns are addressed in our empirical work, providing a new perspective for understanding the impact of exchange rate changes on the efficiency of resource allocation within an industry. Second, we provide empirical evidence for analyzing the impact of RMB exchange rate changes on the behavior and performance of firms in the tails of the intra-industry productivity distribution. The traditional least squares method pays more attention to the conditional mean of productivity, but less attention to the plants in the tail of the productivity distribution that are most vulnerable to exchange rate changes. Analysis of the behavior of firms in the tail of the productivity distribution contributes to efficiency and fairness during the related policy making. Third, in this paper we also analyze the heterogeneous impact of exchange rates on the efficiency of resource allocation in different industries. Using the standard deviation of productivity as the measure of industry resource allocation efficiency, the results obtained through regression can only analyze the average effect of exchange rate changes on industry resource allocation efficiency; thus, it cannot examine the resource allocation situation of a single industry, and it is not conducive to policy makers to implement according to industry characteristics policy. Analyzing the heterogeneous effect of exchange rate on productivity dispersion in different industries can help policymakers implement differentiated policies according to different industries.

2. Theoretical Model

To analyze the effects of exchange rate movement on firms in the same sector with different productivity, we set up a static model with two countries.

Production function. Firms in an industry are indexed by $i=1,\dots ,J$. The production technology of firm $i$ is given by the function

$$Y_i=\exp ({\omega }_i)K_i^{\alpha }{L}_i^{\beta }{\displaystyle \prod }_{s=1}^N{X}_{is}^{{\gamma }_s}$$

(1)

where $Y_i$ is the output of firm $i$. $K_i$ and $L_i$ denote the capital and labor used by $i$ in production process. $X_{is}$ is quantity of intermediate composite good $s$ used by firm $i$.${\omega }_i$ is firm’s total-factor productivity (TFP). ${\gamma }_s$ denotes the importance of intermediate input $s$, and the sum of them is $\gamma ={{\displaystyle \sum }}_s{\gamma }_s$.

Firm $i$ purchase each input variety supplied both from domestic and foreign markets:

$$X_{is}={[{(B_{is}^{*}{X}_{isF})}^{\frac{\theta -1}{\theta }}+{(X_{isH})}^{\frac{\theta -1}{\theta }}]}^{\frac{\theta }{\theta -1}}$$

(2)

where $X_{isF}$ and $X_{isH}$ denote the quantity of foreign and domestic inputs that the firm purchases, and $\theta$ is the elasticity of substitution between foreign inputs and domestic ones. $B_{is}^{*}$ denotes the quality advantage of the imported inputs over domestic inputs.

Solve the cost-minimization problems in (2), and then the price of composite inputs $X_{is}$ can be found:

$$P_{is}={[P_{sH}^{1-\theta }+{(\frac{P_{sF}^{*}}{e{B}_{is}^{*}})}^{1-\theta }]}^{\frac{1}{1-\theta }}=P_{sH}{[1+{(\frac{P_{sF}^{*}}{e{B}_{is}^{*}{P}_{sH}})}^{1-\theta }]}^{\frac{1}{1-\theta }}$$

(3)

where $P_{sH}$ and $P_{sF}^{*}$ are the prices of the domestic and foreign varieties in local currency, respectively, and $e$ is exchange rate, which denotes the quantity of foreign currency per unit of RMB. We used $A_{is}=\frac{B_{is}^{*}{P}_{iH}}{P_{sF}^{*}}$, which denotes the price-adjusted quality advantage of imported inputs, and we assume that $A_{is}=A$. If a firm only uses domestic inputs, then $P_{is}=P_{sH}$. Substitute A in (3), and take the log:

$$a\left(e\right)=\frac{ln[1+\left(eA{)}^{\theta -1}\right]}{\theta -1}$$

(4)

where $P_{is}=P_{sH}\exp [-a\left(e\right)]$, and $a\left(e\right)$ measures the per-product import gain. $\frac{\partial a\left(e\right)}{\partial e}>0$ implies that the per-product import gain increase with RMB appreciation. $\frac{\partial a\left(e\right)}{\partial A}>0$ and $\frac{\partial a\left(e\right)}{\partial \theta }<0$ imply that $a\left(e\right)$ is higher when the price-adjusted quality A is higher and imperfect substitution $\theta$ is lower.

If the firm imports foreign intermediate inputs, the optimal expenditure share of the foreign inputs in the total spending for each variety,

$$S\left(e\right)=\frac{P_{sF}^{*}{X}_{isF}}{e{P}_{is}{X}_{is}}=\frac{{(eA)}^{\theta -1}}{1+{(eA)}^{\theta -1}}$$

(5)

satisfies $0<S<1$. Since $\partial S/\partial e>0$, the foreign expenditure share increases with RMB appreciation.

The relative importance for production of the intermediate inputs the firm chooses to import can be expressed as

$$G\left(n_i\right)=\frac{{{\displaystyle \sum }}_{s=1}^{n_i}{\gamma }_s}{{{\displaystyle \sum }}_{s=1}^N{\gamma }_s}=\frac{{{\displaystyle \sum }}_{s=1}^{n_i}{\gamma }_s}{\gamma }$$

(6)

$G(•)$ satisfy $\frac{\partial G(•)}{\partial n}>0$ and $\frac{{\partial }^{2}G(•)}{{\partial }^{2}n}<0$. Conditional on $n_i$, expenditure on all imported inputs $M_i^F={{\displaystyle \sum }}_{s=1}^N{P}_{sF}^{*}{X}_{is}/e$, and expenditure on all inputs. The spending share on imports equals

$$\frac{M_i^F}{M_i}=S\left(e\right)\frac{{{\displaystyle \sum }}_{s=1}^{n_i}{\gamma }_s}{\gamma }=S\left(e\right)G\left(n_i\right)$$

(7)

Conditional on the Cobb–Douglas production function, expenditure on all inputs equals

$$M_i={\displaystyle \prod }_{s=1}^N{({\gamma }_s/\gamma )}^{-{\gamma }_s/\gamma }{\displaystyle \prod }_{s=1}^N{P}_{is}^{{\gamma }_s/\gamma }{\displaystyle \prod }_{s=1}^N{X}_{is}^{{\gamma }_s/\gamma }$$

(8)

where $P_{is}=P_{sH}\exp [-a\left(e\right)]$. Let $\mathsf{\Gamma }$$={\displaystyle \prod _{S=1}^N{({\gamma }_S/\gamma )}^{-{\gamma }_S/\gamma }}$. Substitute them in (8) and we get

$$\begin{array}{ll}{M}_i& =\mathsf{\Gamma }{\displaystyle \prod }_{s=1}^N{P}_{iH}^{{\gamma }_s/\gamma }{\displaystyle \prod }_{s=1}^N{X}_{is}^{{\gamma }_s/\gamma }\exp [-a\left(e\right){\displaystyle \sum }_{s=1}^N{\gamma }_s/\gamma ]\\ & =\mathsf{\Gamma }{\displaystyle \prod }_{s=1}^N{P}_{iH}^{{\gamma }_s/\gamma }{\displaystyle \prod }_{s=1}^N{X}_{is}^{{\gamma }_s/\gamma }\exp [-a\left(e\right)G\left(n_i\right)]\end{array}$$

(9)

Let $\rho =\mathsf{\Gamma }$${\displaystyle \prod _{S=1}^N{P}_{sH}^{{\gamma }_S/\gamma }}$, and then the quantity of the composite input follows that

$${\displaystyle \prod }_{s=1}^N{X}_{is}^{{\gamma }_s}=M_i^{\gamma }{\rho }^{-\gamma }\exp [a\left(e\right)G\left(n_i\right)]$$

(10)

We then use (10) to express firm output and take the log. Then the production Function (1) implies

$$y_i=\alpha k_i+\beta l_i+\gamma \left(m_i-\rho \right)+\gamma a\left(e\right)G\left(n_i\right)+{\omega }_i$$

(11)

where the lowercase variables $y_i,k_i,l_i$, and $m_i$ denote logs.

Consumers have the following preference over manufacturing varieties i,

$$Q_T=[{\displaystyle \underset{i\in {\Omega }_T}{\int }{Q}_i^{(\eta -1)/\eta }di}+{\displaystyle \underset{i\in {\Omega }_T^{*}}{\int }{Q}_i^{(\eta -1)/\eta }di{]}^{\eta /(\eta -1)}}$$

(12)

$Q_T$ is the gross demand for the industry. ${\Omega }_T$ and ${\Omega }_T^{*}$ denote the sets of domestically produced and imported varieties, respectively, which are given, and $Q_i$ is the consumption of individual i. Given preference (12), demand faced by firm i is

$$Q_i={(p_i/P_T)}^{-\eta }{Q}_T$$

(13)

$$Q_i^{*}={(p_{i}e/P_T^{*})}^{-\eta }{Q}_T^{*}$$

(14)

Here $Q_i$ and $Q_i{}^{*}$ are domestic and foreign demand faced by firm i; $p_i$ is the price charged by firm i; $P_T$ is the price index of the industry; $Q_T$ is demand for the CES aggregate by consumers. Both are taken as given by firms. Then we can get the inverse demand function $p_i=Q_i^{-\frac{1}{\eta }}{Q}_T^{\frac{1}{\eta }}{P}_T$ and $p_i=Q^{*}{}_i^{-\frac{1}{\eta }}{Q}^{*}{}_T^{\frac{1}{\eta }}{P}_T{}^{*}/e$. Using optimal pricing, it is easy to show that the fraction of domestic sales is given by ${\nu }_i\left(e\right)\equiv \frac{Q_i}{Q_i+Q_i^{*}}$. Since $Q_i=v_i\left(e\right)Y_i$, we have that $Q_i{}^{\frac{\eta -1}{\eta }}=v_i{(e)}^{\frac{\eta -1}{\eta }}{Y}_i{}^{\frac{\eta -1}{\eta }}$. For an exporting firm, we can then write total revenue $R_{it}$ as

$$\begin{array}{l}{R}_i=p_i{Q}_i+p_i{Q}_i^{*}=Q_i^{\frac{\eta -1}{\eta }}{Q}_T^{\frac{1}{\eta }}{P}_T+Q^{*}{}_i^{\frac{\eta -1}{\eta }}{Q}^{*}{}_T^{\frac{1}{\eta }}{P}_T{}^{*}/e\\ =Y_i^{\frac{\eta -1}{\eta }}[{\nu }_i{(e)}^{\frac{\eta -1}{\eta }}{Q}_T^{\frac{1}{\eta }}{P}_T+(1-{\nu }_i\left(e\right){)}^{\frac{\eta -1}{\eta }}{Q}^{*}{}_T^{\frac{1}{\eta }}{P}_T{}^{*}/e]\\ =Y_i^{\frac{\eta -1}{\eta }}{H}_i\left(Q_T,Q_T^{*},e\right)\end{array}$$

(15)

Taking logs and plugging in production Function (11):

$$r_i=\frac{\eta -1}{\eta }\left[{\alpha }_0+\alpha k_i+\beta l_i+\gamma \left(m_i-\rho \right)+\gamma a\left(e\right)G\left(n_i\right)+{\omega }_i\right]+h_i\left(Q_T,Q_T^{*},e\right)$$

(16)

Then we can construct an empirical measure of the level revenue-based productivity (TFPR) as

$$tfp{r}_i=r_i-p_T-\left[{\alpha }^{*}{k}_i+{\beta }^{*}{l}_i+{\gamma }^{*}\left(m_i-\rho \right)\right]={\tilde{\alpha }}_0+{\gamma }^{*}a\left(e\right)G\left(n_i\right)+h_i\left(q_T,q_T^{*},e\right)+{\omega }_i^{*}$$

(17)

Taking derivatives with respect to the exchange rate, we can get the elasticity of productivity on exchange rate:

$$\frac{\partial tfp{r}_i}{\partial e}={\gamma }^{*}\frac{\partial a\left(e\right)G\left(n_i\right)}{\partial e}+\frac{\partial g_i\left(q_T,q_T^{*},e\right)}{\partial e}={\gamma }^{*}\frac{\partial a\left(e\right)}{\partial e}G\left(n_i\right)+{\gamma }^{*}\frac{\partial G\left(n_i\right)}{\partial e}a\left(e\right)+\frac{\partial h_i\left(q_T,q_T^{*},e\right)}{\partial e}$$

(18)

The first two items in (18) represent the imported intermediate goods channel. This channel can be divided into two effects. The first term indicates that the cost reduction effects are induced through the quality difference and incomplete substitution between the imported inputs and domestic ones. This effect is related to the number of types of intermediate goods imported by enterprises. Since $\frac{\partial a\left(e\right)}{\partial e}>0$, an appreciation will improve the quality of the imported inputs after price adjustment. This improvement effect is related to the number of types of intermediate inputs imported by enterprises. The more types of imported inputs, the greater the effect of improving the quality of the imported inputs brought about by exchange rate appreciation. The second item is that exchange rate appreciation affects firm productivity through the changes in the number of types of intermediate imported inputs. This effect is related to the price-adjusted quality advantage of the imported inputs. The third item in (18) implies the effect of exchange rate movement on the productivity of the enterprise is through the export channel, which is mainly due to the decrease in the share of the export quantity and the total export value of the enterprise when the exchange rate appreciates, resulting in a decline in productivity. Since $\frac{\partial h_i\left(q_T,q_T^{*},e\right)}{\partial e}<0$, the exchange rate appreciation decreases the productivity through the export channel. For two firms with different productivity in the same industry (assuming ${\omega }_i>{\omega }_j$), the difference in the impact of exchange rate changes on the productivity of the two firms can be expressed as

$$\begin{array}{l}\frac{\partial tfp{r}_i}{\partial e}-\frac{\partial tfp{r}_j}{\partial e}={\gamma }^{*}\frac{\partial a\left(e\right)}{\partial e}\left[G\left(n_i\right)-G\left(n_j\right)\right]+{\gamma }^{*}a\left(e\right)[\frac{\partial G\left(n_i\right)}{\partial e}-\frac{\partial G\left(n_j\right)}{\partial e}]\\ +[\frac{\partial h_i\left(q_T,q_T^{*},e\right)}{\partial e}-\frac{\partial h_j\left(q_T,q_T^{*},e\right)}{\partial e}]\end{array}$$

(19)

Since ${\omega }_i>{\omega }_j$ and $\frac{\partial \pi }{\partial \omega }>0$, we can get $n_i>n_j$. Because $\frac{\partial G(•)}{\partial n}>0$, $G\left(n_i\right)-G\left(n_j\right)>0$. $\frac{{\partial }^{2}G(•)}{{\partial }^{2}n}<0$ implies that $\frac{\partial G\left(n_i\right)}{\partial e}-\frac{\partial G\left(n_j\right)}{\partial e}<0$ when ${\omega }_i$ << ${\omega }_j$. As ${\nu }_i\left(e\right)<{\nu }_j\left(e\right)$, $\frac{\partial h_i\left(q_T,q_T^{*},e\right)}{\partial e}<\frac{\partial h_j\left(q_T,q_T^{*},e\right)}{\partial e}$($\frac{\partial h}{\partial e}<0$). Therefore, the difference in the response of the productivity to exchange rate movement between i and j is determined by $\gamma$, $\eta$, A, $\theta$, and the aggregate demand of the industry at home and abroad.

Exchange rates movements affect the import price of the final products and affect the distribution of enterprises by affecting the industry prices. Exchange rate appreciation causes industry prices to drop; the relative prices of domestic products then rise, and domestic firms’ profits decline or they even exit the market.

In light of these arguments, we propose our hypothesis:

Hypothesis 1. 

Appreciations of the RMB exchange rate decrease the industrial resource allocation efficiency of the Chinese manufacturing sectors. The imported intermediate inputs channel is the main cause of the negative effects of RMB exchange rate on the resource allocation efficiency.

3. Data Description

The firms’ production-side information as the benchmark was mainly obtained from the Annual Surveys of Industrial Production (ASIP), conducted by the National Bureau of Statistics of China (NBC), over 1998–2007. Outliers were removed according to the following methods: excluding sample values with less than eight employees; total output equal to or less than 0; total assets less than fixed assets; negative fixed costs; and negative inputs. The total output value, value-added, and intermediate inputs were converted to their actual values. The firms’ labor productivity is the actual output and employees discounted using the deflator. The total factor productivity of the companies was calculated using the actual output in the database of industrial companies according to the method provided by Olley and Pakes [22]. In this paper, attention was only paid to the manufacturing industry. Therefore, only the Chinese industry categories (CIC) coded 13–42 were finally kept to obtain about 1.5 million samples from 1998 to 2007. As a robust test, we also introduce the samples from 2008 to 2013 to test the benchmark regression results.

There are two sources of trade data used in this paper. The trade data for calculating the real effective exchange rate and the data for counting the imports and exports of other countries with China were obtained from the CEPII database. The trade data of companies with different productivity levels within industries were obtained from the Chinese customs database, mainly from 2000 to 2006. The paper draws on the approach of Li [23]; the data from the customs database were processed to remove outliers in the following manner: excluding data with missing trade values and quantities; excluding samples of products with annual growth in unit export values and quantities in the top 5% and bottom 5% of the industry. Then the data in the customs database were matched with the industrial enterprise database, based on the enterprise name and year, which aims to exclude any samples of companies whose exports are more extensive than their outputs and whose imported intermediate inputs are more significant than the total intermediate inputs used. The trade statistics values of each industry from 2000 to 2006 were calculated based on the trade data of the matched companies, and the annual average values were taken as the proxy variables for studying the trade characteristics statistics of each industry. The real GDP and inflation indices, nominal exchange rate data, and Chinese TFP used in this paper were taken from the Penn World Table (PWT9.0). Data on China’s tariffs were obtained from the WITS database. Data on China’s FDI-restricted entry policies were obtained from Brandt [20]. The FDI forbidden data were from the Catalogue for the Guidance of Industries for Foreign Investment published by the Chinese government. Product market openness was from the Wind database (website: https://www.wind.com.cn/, accessed on 20 June 2022).

4. Regression Analysis

We adopt the regression quantiles model proposed by Koenker [24] to study the relationship between exchange rate and productivity. The regression quantile model can estimate the response of different quantiles of the dependent variable to changes in the explanatory variables, which can effectively reduce the bias of the estimated coefficients caused by the outliers, and can also analyze the entire productivity distribution, especially the change in the tails of the productivity distribution. Referring to Tomlin [7], the benchmark regression model used was as follows:

$$p{r}_{ikt}={\beta }_{k}re{r}_{kt}+{\alpha }_k{}^{CHN}{\tau }_{kt}^{CHN}+{\alpha }_k^{FOR}{\tau }_{kt}^{FOR}+\gamma \ln TF{P}_t+{\lambda }_k{x}_{it}+{\eta }_k{y}_{ikt}+{\epsilon }_{ikt}$$

(20)

where $p{r}_{ikt}$ is the logarithm of labor productivity for plant $i$, in industry $k$, at time $t$; $re{r}_{kt}$ is the logarithm of the trade-weighted real exchange rate; and ${\tau }_{kt}^{CHN}$ is the import tariff rates on products in industry $kk$ and ${\tau }_{kt}^{FOR}$ is the export-weighted export tariff on products in industry $k$; $\ln TF{P}_t$ is the logarithm aggregate total-factor productivity of China; $x_{kt}$ is the vector of industry-level controls (including the logarithm of real GDP of main trade-partners with China and FDI forbidden on industry $k$; and $y_{ikt}$ is the vector of plant-level controls (including age, scale, and ownership of the plants). We estimated the model at each decile of the conditional productivity distribution (i.e., the 10th, 20th, …, 90th percentiles) in (1) separately for each of the 29 industries in the data set, which provide 261 estimates of $\beta$.

We obtained the average estimates of $\beta$ across all industries in two ways. The average unweighted values of $\beta$ for each quantile $\theta$ is calculated as ${\tilde{\beta }}_{\theta }={{\displaystyle \sum }}_k\sigma {\widehat{\beta }}_{\theta ,k}$, where ${\widehat{\beta }}_{\theta ,k}$ is the estimates of $\beta$ for quantile $\theta$ of industry $k$, and $\sigma =\frac{1}{29}$. The average weighted estimate of $\beta$ is calculated as ${\tilde{\beta }}_{\theta }={{\displaystyle \sum }}_k{\sigma }_k{\widehat{\beta }}_{\theta ,k}$, where ${\sigma }_k$ is industry k’s share in the total output of the 29 industries. Figure 1 presents the two average estimates of $\beta$ across all industries, separately. In panel (a), the average unweighted values are reported on the y-axis. Panel (b) presents the average weighted estimates of $\beta$. The x-axis reports the quantile at which regression the model was estimated. We approximated the 95% confidence intervals by using $C{I}_{\theta }={\tilde{\beta }}_{\theta }\pm 1.95se\left({\tilde{\beta }}_{\theta }\right)$, where $se\left({\tilde{\beta }}_{\theta ,k}\right)$ is the average standard errors calculated using the following equation:

$$se\left({\tilde{\beta }}_{\theta }\right)={\left(\mathrm{var}\left[{\displaystyle \sum }_k\left({\sigma }_k\cdot {\widehat{\beta }}_{\theta ,k}\right)\right]\right)}^{1/2}={\left({\displaystyle \sum }_k{[{\sigma }_k\cdot se\left({\widehat{\beta }}_{\theta ,k}\right)]}^2\right)}^{1/2}$$

(21)

In panel (a), we set ${\sigma }_k=\frac{1}{29}$. In panel (b), ${\sigma }_k$ is industry k’s share in the total output of the 29 industries.

As in Figure 1, the results in both the unweighted and weighted cases are very similar. The upward sloping quantile regression curves show that the value of the estimated coefficient on the exchange rate varies over the conditional productivity distribution. It is clear that the average quantile estimate of $\beta$ tells a different story with the OLS estimates. For the unweighted average, a 1% appreciation of the exchange rate is associated with an increase of 0.36% in mean firm productivity and 0.2% in median firm productivity. However, at the lower end of the conditional productivity distribution, the average estimate is not significant at the 10th percentile, while a 1% increase in the exchange rate is associated with a 0.6% increase at the 80th percentile and 0.56% increase at the 90th percentile. The weighted results, which better reflect the aggregate results, are quite similar. The averaged coefficient at the 10th percentile is not significant. However, a 1% increase in the exchange rate leads to a 0.96% increase in productivity at the 90th percentile, and a 1.3% increase in 80th percentile. For both panels, the 95% confidence intervals suggest that the result of an upward sloping quantile curve is statistically meaningful.

5. Robust Tests

The robust tests include replacing the control variables and using TFP to replace labor productivity as dependent variable.

5.1. Different Control Variables

The benchmark regression uses the weighted averaged real GDP of the major trading countries to capture foreign demand. In order to more accurately measure the foreign demand, we refer to Tomlin and Fung (2015) and use the total import volume of China’s major exporting countries in this industry minus the volume imported from China as a proxy variable for this country’s demand, and then weighted by China’s share of the industry’s exports to the country as a foreign demand shock. To capture the reduction in the foreign production costs, we use the volume this industry exported by China’s main import source countries to other countries other than China as a proxy. The regression results are shown in Figure 2. Figure 2a,b represent the industry simple average and weighted average of the regression estimation coefficients for each quantile, respectively. The conclusion in Figure 2 is consistent with the conclusions in the benchmark regression. As the RMB exchange rate appreciates, the conditional mean of firm productivity increases, and the dispersion of industry productivity increases. After the foreign demand shock is controlled at the industry level, for firms with lower productivity, the productivity regression coefficient is negative, and the appreciation of the exchange rate makes the productivity dispersion increase more sharply.

5.2. TFP as the Dependent Variable

The change in the dispersion of labor productivity may be caused by the per capita capital of the enterprise, not the change in the production efficiency of the enterprise. In order to further confirm the impact of the RMB exchange rate appreciation on industry resource allocation, we replaced the dependent variable labor productivity with firms’ TFP and performed quantile regression according to Formula (20). The simple mean and industry output share weighted mean of the estimated coefficients of each quantile of the TFP distribution on the exchange rate regression are shown in Figure 3a,b, respectively. Figure 3 shows that with the appreciation of the RMB exchange rate, all quantiles of the TFP distribution increase, and the productivity growth rate in the right tail of the distribution is greater than that in the left tail, and the productivity dispersion increases. This result is consistent with the benchmark results. A slight difference from the benchmark regression is that the estimated coefficient of productivity in the left tail of the TFP distribution is significantly positive; the expansion of TFP dispersion is smaller than that of labor productivity.

6. Further Investigation

6.1. Extent Samples

In order to verify the above regression results, we again used the data from 2008 to 2013 for further analysis. Since the global financial crisis in 2008 had a great impact on Chinese companies, the entry and exit of a large number of companies had a greater impact on the continuity of corporate data before and after the financial crisis, and the degree of matching was low. In 2011, the China Industrial Enterprise Database re-established the enterprise statistical standards. The previous statistical scale was adjusted from sales of 5 million and above to 20 million and above. The sample range has undergone great changes, and it is no longer appropriate to use enterprise-level data. One must perform quantile regression. Therefore, for the samples from 2008 to 2013, we used the standard deviation and interquartile range of China’s manufacturing industry productivity to measure the industrial production dispersion. Considering the influence of the sample size on the regression accuracy, we adopted the four-digit code level of the manufacturing industry. The calculated industry productivity dispersion was 1%. The regression model was as follows:

$$\begin{array}{l}dispersio{n}_{kt}=\beta re{r}_{kt}+{\alpha }^{CHN}tariff\_im{p}_{kt}+{\alpha }_{}^{FOR}tariff\_e{x}_{kt}+{\gamma }_{1}tf{p}_t+{\gamma }_{2}chn\_gd{p}_t\\ +{\gamma }_{3}for\_gd{p}_t+\lambda any\_fd{i}_{kt}+{\epsilon }_{kt}\end{array}$$

(22)

where $dispersio{n}_{kt}$ is the industrial productivity dispersion; ${\alpha }_k$ is industry fixed effects; and ${\epsilon }_{kt}$ the random error term. Others are the same as the variables in (19).

We first used the sample from 1998 to 2007 to further verify the benchmark regression results, but used (22) for the regression. The regression results are shown in

Table 1. The regression results using the sample from 1998 to 2007.

	(1)	(2)	(3)	(4)	(5)	(6)
Dependent Variables	Standard Deviation	Standard Deviation	Standard Deviation	Interquartile	Interquartile	Interquartile
rer	0.808 ***	0.608 ***	0.197 ***	1.079 ***	0.933 ***	0.388 ***
	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)
tariff_exp		0.005 *	0.004 ***		0.014 **	0.014 ***
		(0.097)	(0.001)		(0.011)	(0.000)
for_gdp		−0.177 ***	0.066 **		−0.117 *	0.204 ***
		(0.000)	(0.013)		(0.075)	(0.001)
tariff_imp		−0.012	−0.014		−0.017	−0.019
		(0.358)	(0.275)		(0.302)	(0.198)
tfp			3.724 ***			5.874 ***
			(0.000)			(0.000)
china_gdp			−0.798 ***			−1.215 ***
			(0.000)			(0.000)
_cons	−2.741 ***	0.972	1.698 ***	−3.700 ***	−1.215	0.232
	(0.000)	(0.143)	(0.001)	(0.000)	(0.362)	(0.839)
N	395	395	395	395	395	395
adj. R2	0.139	0.166	0.270	0.079	0.088	0.153

Notes: ***, **, and * indicate statistical significance at the 1%, 5%, and 10% level, respectively.

. Since the Hausman test cannot reject the fixed-effects model, the regression controls for the industry fixed effects. Columns (1)–(3) in Table 1 are the results of the regression using the standard deviation of labor productivity as a measure of the efficiency of resource allocation in the industry, and columns (4)–(6) indicate that the industry is measured by the quartiles of industry labor productivity distribution resource allocation efficiency. The exchange rate coefficient in Table 1 is positive and highly significant, indicating that the appreciation of the RMB exchange rate increases the dispersion of industry productivity and reduces the efficiency of industry resource allocation. If the real effective exchange rate of RMB appreciates by 1%, the 75% quantile of the labor productivity distribution will be 47% higher than the 25% quantile ($e^{0.388}-1=0.47$). This is consistent with the conclusions obtained from the benchmark regression.

Using the sample from 2008 to 2013, the regression results of the real effective exchange rate of RMB are shown in

Table 2. The regression results using the sample from 2008 and 2013.

	(1)	(2)	(3)	(4)	(5)	(6)
Dependent Variables	Standard Deviation	Standard Deviation	Standard Deviation	Interquartile	Interquartile	Interquartile
rer	0.351 ***	0.190 ***	0.177	0.530 ***	0.384 ***	0.607 *
	(0.000)	(0.000)	(0.408)	(0.000)	(0.000)	(0.052)
tariff_exp		0.013	0.013		0.019	0.019
		(0.263)	(0.266)		(0.344)	(0.339)
tariff_imp		−0.013	−0.014		−0.007	−0.010
		(0.107)	(0.102)		(0.391)	(0.232)
for_gdp		0.000 ***	0.000		0.000 ***	0.000
		(0.000)	(0.236)		(0.000)	(0.525)
chn_gdp			0.000			−0.000
			(0.817)			(0.877)
tfp			−1.211			−2.571
			(0.653)			(0.533)
_cons	−0.726 ***	−1.354 ***	−0.520	−1.291 ***	−2.027 ***	−1.400
	(0.000)	(0.000)	(0.840)	(0.000)	(0.000)	(0.721)
N	1023	1023	1023	1022	1022	1022
adj. R2	0.318	0.407	0.407	0.285	0.346	0.351

Notes: *** and * indicate statistical significance at the 1% and 10% level, respectively.

, which shows that no matter calculating with or without the other control variables, the regression coefficient of the standard deviation of labor productivity on the appreciation of the RMB exchange rate is positive and significant. In column (3), the coefficient on exchange rate is not significant, which may be caused by the contribution character of China’s GDP and aggregate TFP. When only one of these two variables is added, the coefficient is significant at the 1% level. This means that the appreciation of the RMB exchange rate increases the standard deviation of industry productivity, increases the dispersion of productivity, and reduces the efficiency of industry resource allocation.

Column (6) shows that the real effective exchange rate of the RMB appreciation is 1%, and the 75% quantile of industry productivity distribution is 76% higher than the 25% quantile ($e^{0.607}-1=0.83$). The regression results in Table 1 and Table 2 are nearly consistent with the conclusions obtained from the benchmark regression.

6.2. Entry and Exit

In order to explore the impact of exchange rate appreciation on firms’ exit and entry in the market, we used LPM, probit, and logit models to estimate the effect of exchange rate appreciation on firms’ exit from the market and their entry probabilities. We controlled for a range of observable variables, such as firm lag-productivity, wages, and age; market variables such as industry tariffs, real GDP, and weighted average foreign real GDP; and firm time-invariant variables such as initial size and ownership, and so on.

Table 3. Impact of RMB appreciation on firm exit within industries.

Regression Model	LPM	Probit	Logit
rer	0.11 ***	0.71 ***	1.23 ***
20th∙rer	−0.02	−0.24	−0.4
30th∙rer	−0.04	−0.38 **	−0.61 *
40th∙rer	−0.04	−0.3 *	−0.39
50th∙rer	−0.08 ***	−0.53 ***	−0.84 **
60th∙rer	−0.11 ***	−0.73 ***	−1.25 ***
70th∙rer	−0.09 ***	−0.56 ***	−0.9 ***
80th∙rer	−0.08 ***	−0.4 **	−0.59 *
90th∙rer	−0.14 ***	−0.77 ***	−1.27 ***
100th∙rer	−0.13 ***	−0.67 ***	−1.1 ***

Notes: ***, **, and * indicate statistical significance at the 1%, 5%, and 10% level, respectively.

and

Table 4. Impact of RMB appreciation on firm entry within industries.

Regression Model	LPM	Probit	Logit
rer	−0.39 ***	−1.16 ***	−1.87 ***
20th∙rer	0.14 ***	0.37 ***	0.58 ***
30th∙rer	0.28 ***	0.8 ***	1.29 ***
40th∙rer	0.32 ***	0.93 ***	1.51 ***
50th∙rer	0.41 ***	1.25 ***	2.01 ***
60th∙rer	0.5 ***	1.54 ***	2.51 ***
70th∙rer	0.51 ***	1.6 ***	2.58 ***
80th∙rer	0.57 ***	1.81 ***	2.92 ***
90th∙rer	0.62 ***	1.94 ***	3.16 ***
100th∙rer	0.63 ***	2.00 ***	3.27 ***

Notes: *** indicate statistical significance at the 1% level, respectively.

show the regression results of firm exit and entry on the exchange rate, respectively. Table 3 shows that when the RMB exchange rate appreciates, firms with a lower productivity are more likely to exit. As productivity rises, the probability of exit becomes smaller and smaller. Table 4 shows that exchange rate appreciation increases the probability of entry of firms with a higher productivity, appreciation promotes the entry of high-productivity firms, and the higher the productivity, the more pronounced the effect is. The results obtained by the three models are consistent. The cost of importing intermediate goods and final goods will decrease with RMB appreciation, which lead to high-productivity enterprises entry and low-efficiency ones exiting. New entrants are easier to adjust, whether it is a fall in the prices of intermediate goods or intensified competition in the domestic product market, and thus faster productivity growth [20]. Especially for the import of intermediate inputs, new entrants can obtain more varieties of foreign inputs with higher quality, which can more conveniently adjust the production method and encourage enterprises to upgrade product quality.

7. Heterogeneity Analysis

We calculated the standard deviation of labor productivity as a measure of the productivity dispersion of subdivided industries in a four-digit code classification, and then regressed the exchange rate. The industry classification range is narrower, and the product substitution is stronger. We mainly selected the degree of export dependence, the degree of import dependence, the degree of product market openness, and FDI control as the interaction items of exchange rate for investigation.

7.1. Export Intensity

According to the theoretical model, the larger the export intensity is, the greater the negative effect of exchange rate appreciation on the firm’s productivity. In an industry, the higher the firm productivity and the greater the export intensity, the stronger the negative effect of exchange rate appreciation. Therefore, the greater the export intensity of the industry, the smaller the increase in productivity dispersion caused by exchange rate appreciation. The export intensity of the industry in this paper was obtained from the average annual enterprise export intensity of the industry. The industry export intensity was lagged to avoid endogeneity. The regression results are as shown in

Table 5. Impact of RMB appreciation on productivity dispersion within industries with different export intensities.

Export Intensity	(1)	(2)	(3)
rer	0.622 ***	0.622 ***	0.554 ***
	(0.000)	(0.000)	(0.000)
exp_int	2.895	3.048	3.205 *
	(0.138)	(0.120)	(0.100)
rer × exp_int	−0.641	−0.675	−0.708 *
	(0.135)	(0.118)	(0.099)
tariff_exp		−0.000	0.004
		(0.947)	(0.440)
for_gdp		−0.018	0.062 *
		(0.580)	(0.096)
tariff_imp		−0.011	−0.014
		(0.288)	(0.189)
tfp			−0.904
			(0.165)
china_gdp			0.113
			(0.347)
_cons	−1.915 ***	−1.608 **	−2.843 ***
	(0.000)	(0.030)	(0.000)
N	395	395	395
adj. R2	0.111	0.112	0.121

Notes: ***, **, and * indicate statistical significance at the 1%, 5%, and 10% level, respectively.

. Table 5 shows that the exchange rate coefficient is positive and significant, and the interaction term between the exchange rate and export intensity is negative and significant, which means that in industries with greater export intensity, exchange rate appreciation makes the decline in industry productivity dispersion smaller.

7.2. Import Intensity

The import intensity of the industry also adopted the average value of the import intensity of firms within the industry, and used the lag term to form an interaction term with the exchange rate for regression analysis (the results are shown in

Table 6. Impact of RMB appreciations on industry productivity dispersion with different import intensities.

Import Intensity	(1)	(2)	(3)
rer	0.401 ***	0.379 ***	0.279 ***
	(0.000)	(0.000)	(0.001)
imp_int	−2.833	−2.941 *	−3.178 *
	(0.104)	(0.094)	(0.072)
rer × imp_int	0.621	0.645 *	0.693 *
	(0.105)	(0.096)	(0.074)
tariff_exp		0.001	0.006
		(0.919)	(0.267)
for_gdp		−0.030	0.067
		(0.415)	(0.102)
tariff_imp		−0.012	−0.015
		(0.237)	(0.126)
tfp			−1.140 *
			(0.074)
china_gdp			0.145
			(0.225)
_cons	−0.912 **	−0.328	−1.762 **
	(0.012)	(0.692)	(0.034)
N	395	395	395
adj. R2	0.120	0.121	0.135

Notes: ***, **, and * indicate statistical significance at the 1%, 5%, and 10% level, respectively.

). Table 6 shows that the coefficient of the interaction term between the exchange rate and import intensity is positive and significant at the 10% confidence level, which means that the greater the import intensity of the industry, the greater the increase in the productivity dispersion of the industry due to the appreciation of the RMB exchange rate, and the lower the efficiency of resource allocation. When the industry relies more on imported intermediate inputs, the productivity increase of higher productivity enterprises from RMB appreciation is larger than that of lower productivity enterprises, which is consistent with the theoretical model.

7.3. Product Market Openness

The data on the openness of the industry product market in this paper comes from the product market development score in the Fan Gang marketization index in the wind database. Since the product market development score is the provincial level data, the product market openness of an industry is the industry average of the product market development score of the location of the enterprises in the industry.

Table 7. Impact of RMB appreciations on productivity dispersion within industries with different product market openness.

Product Market Openness	(1)	(2)	(3)
Rer	1.911 ***	1.777 ***	1.269 **
	(0.000)	(0.000)	(0.025)
market_openness	0.939 ***	0.849 ***	0.531 *
	(0.000)	(0.000)	(0.084)
rer × market_openness	−0.215 ***	−0.196 ***	−0.129 *
	(0.000)	(0.000)	(0.058)
tariff_exp		0.004 ***	0.004 ***
		(0.006)	(0.002)
for_gdp		0.082 ***	0.058 **
		(0.001)	(0.027)
tariff_imp		−0.014	−0.015
		(0.221)	(0.191)
Tfp			0.778
			(0.268)
chn_gdp			−0.051
			(0.723)
_cons	−7.464 ***	−8.115 ***	−5.325 *
	(0.000)	(0.000)	(0.060)
N	395	395	395
adj. R2	0.289	0.294	0.299

Notes: ***, **, and * indicate statistical significance at the 1%, 5%, and 10% level, respectively.

shows the regression results. As the results show, the coefficient of interaction between the exchange rate and the market openness index is negative and significant, which means that the more open the product market is in the industry, the appreciation of the RMB exchange rate makes the increase in the productivity dispersion smaller.

7.4. FDI Forbidden

The regression results shown in

Table 8. Impact of RMB appreciations on productivity dispersion within industries with different FDI forbidden levels.

FDI Forbidden	(1)	(2)	(3)
rer	0.819 ***	0.278 ***	0.228 ***
	(0.000)	(0.000)	(0.000)
any_fdi	1.298 ***	1.053 ***	1.027 **
	(0.008)	(0.025)	(0.024)
any_fdi × rer	−0.275 **	−0.225 ***	−0.223 **
	(0.010)	(0.028)	(0.025)
tariff_exp		0.004 **	0.005 ***
		(0.025)	(0.009)
for_gdp		0.028 ***	0.067 ***
		(0.232)	(0.003)
lntariff_imp		−0.013 *	−0.014 **
		(0.053)	(0.040)
Tfp			3.695 ***
			(0.000)
china_gdp			−0.791 ***
_cons	−2.797 ***	−0.475	1.518 ***
	(0.000)	(0.112)	(0.001)
N	395	395	395
adj. R2	0.037	0.065	0.178

Notes: ***, **, and * indicate statistical significance at the 1%, 5%, and 10% level, respectively.

are the results of adding the interaction term of the FDI forbidden and RMB exchange rate. The results in Table 8 show that the coefficient of the interaction term between the exchange rate and FDI control is negative and significant, which means that the industries with FDI entry forbidden have a large productivity dispersion and low resource allocation efficiency. Exchange rate appreciation makes the increase in productivity dispersion smaller and the decrease in resource allocation efficiency smaller if FDI forbidden exists.

8. Conclusions

Based on the data of manufacturing enterprises from 1998 to 2007 in the industrial enterprise database of the National Bureau of Statistics of China, we constructed a static model with a quality difference and imperfection of substitution between domestic intermediate inputs and foreign ones, and used quantile regression to analyze the relationship between resource allocation efficiency and RMB exchange rate appreciation. The empirical results show that an appreciation of the RMB exchange rate leads to an increase in the average productivity of enterprises, but also to an increase in the dispersion of productivity in the industry and a decrease in the resource allocation efficiency. The depreciation of the exchange rate reduces the dispersion of productivity and improves the industrial allocation efficiency. An exchange rate appreciation makes the productivity increase of the higher productivity enterprises in the industry larger than those with a lower productivity. The upward sloping quantile regression curves suggests that RMB appreciation is associated with an expansion of productivity distribution, and depreciations are associated with decreased productivity dispersion.

We find that the RMB exchange rate appreciations are not significantly associated with the change in productivity at the 10th percentile. While at the upper end of the conditional distribution, a 1% increase in the RMB exchange rate leads to a 0.6% increase in productivity at the 80th percentile, and a 0.56% increase at the 90th percentile. The weighted results, which better reflect the different sizes of the industries included in the sample, and which can be considered the aggregated results, are almost identical. Both the unweighted and weighted results are in the line with the predictions of the theoretical model outlined in Section 2. Moreover, both the 90% confidence intervals suggest that the results of the upward-sloping quantile curve is statistically meaningful, which implies that the resource allocation efficiency decreases with RMB exchange rate appreciations.

The increase in the dispersion of industry productivity distribution and the decline in resource allocation efficiency were mainly associated with differences in the intermediate goods that were imported by enterprises with different productivities in the industry and the distribution of the proportion of foreign-funded firms. Larger, more productive firms import more intermediate inputs, which account for a higher share of the production cost, and may be more for foreign, directly invested enterprises. Due to the fixed cost of importing intermediate goods, smaller and less productive firms import less intermediate goods. When the RMB exchange rate appreciates, the cost of more productive firms can decline more as the have a higher imported intermediate inputs share, and they can also import more high-quality inputs, which could improve their products’ quality and their production efficiency. Appreciations of the RMB exchange rate expose domestic plants to more competition as export opportunities shrink and import competition intensifies and increase the productivity of more productive enterprises, forcing smaller less productive firms from the market, which truncates the lower end of the productivity distribution. The entry of enterprises with a higher productivity and the improvement in production methods and production efficiency of the retained enterprises increase at the higher end of the productivity distribution. The productivity growth of the right tail of distribution is faster than that of the left tail, leading to an increase in the dispersion of industry productivity. This implies that importing input from foreign markets to improve productivity may not be sustainable.

Our results imply that when the Chinese government evaluates the impact of the RMB appreciations or monetary policies, they should pay more attention to the resource allocation efficiency. Appreciations have a positive effect on the productivity of firms on average, but have negative effects on industrial allocation efficiency, especially those that import more intermediate inputs. As the trade conflict worsens, Chinese trading patterns may also change. According to our theoretical framework, more effective monetary policy should help firms and industries that are affected by trade conflicts according to their trade status.