The effect of foreign exchange exposure on stock price : an analysis of the Chinese market

(1)

The Effect of Foreign Exchange Exposure on Stock Price:

An Analysis of the Chinese Market

University of Amsterdam

Amsterdam Business School

MSc Finance

Master Specialisation Quantitative Finance

Author:

Yang Zhong

Student number:

10825835

Thesis supervisor: Dr. Jan Lemmen

Finish date:

June 2018

(2)

ii

STATEMENT OF ORIGINALITY

This document is written by Yang Zhong who declares to take full responsibility for the contents of this document.

I declare that the text and the work presented in this document are original and that no sources other than those mentioned in the text and its references have been used in creating it.

The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.

(3)

iii

ABSTRACT

This paper examines the effect of contemporaneous and lagged changes in exchange rate on the value of Chinese firms on an industry basis. Based on a sample of 1,347 firms from 2009 to 2016, this paper suggests that Chinese firms are exposed to both contemporaneous and lagged variations in exchange rate, with the lagged changes playing a more significant role. Asymmetry also exists in exchange rate variations. Under domestic currency appreciation, the stock market crash worsens the negative effect of exchange rate exposure when the coefficient is negative and hinders the positive effect when the coefficient is positive.

Keywords: Foreign Exchange, Financial Crises, International Financial Markets, Multinational Firms, Panel Data Models

(4)

iv

LIST OF TABLES

Table 1 Stock return based on industry page 12 Table 2 Change in foreign exchange rate page 12 Table 3 Industry-level exchange rate exposure page 14 Table 4 Lagged effect of exchange rate exposure page 15 Table 5 Asymmetric nature of exchange rate exposure page 19 Table 6 Lagged asymmetric nature of exchange rate exposure page 20 Table 7 Robustness check of model 1 for the full sample period page 22

LIST OF FIGURES

(6)

1

CHAPTER 1. Introduction

Due to globalization, an increasing number of firms are doing businesses with non-domestic companies. In the early 21st_{century, countries are more closely related to each other through trade in} goods and services that involves exports and imports (Krugman, Obstfeld, and Melitz, 2012). As can be seen from Figure 1 which depicts export and import quotes, all countries in the Organization for Economic Cooperation and Development (OECD) database rely heavily on international trade. This makes the exchange rate crucial because it allows the comparison between prices of goods and services produced in different countries. The exchange rate therefore poses uncertainty on multinationals when it is not fixed. This gives rise to the definition of exchange rate exposure: a firm exhibits exchange rate exposure as long as its value is affected by exchange rate (Adler and Dumas, 1984). Whether the exchange rate against a certain currency is fixed or floating depends on the exchange rate regime each country adopts.

Figure 1. Trade in goods and services: Exports/Imports, % of GDP, 2016

Source: Organization for Economic Cooperation and Development.

China was absolutely averse on trade before 1979 and was seen as an extreme version of import substitution (Wei, 1995). Since the introduction of the open-door policy in 1978, China has gradually built her existence in international trade. The volume of exports and imports increased dramatically over the years. As stated in the World Trade Statistical Review 20171_{published by the} World Trade Organization, China was the world’s biggest trader for merchandise exports and imports (in terms of value) in 2016. In the commercial services sector, China ranked fifth in export and second

(7)

2 in import. During the period from 1997 to 2005, the exchange rate of the Chinese currency Renminbi (RMB hereafter) was pegged to USD, protecting Chinese investors and producers from exchange rate risk. However, on July 21, 2005, the People’s Bank of China lifted the peg and announced a managed floating exchange rate regime with reference to a basket of currencies. The RMB exchange rate was expected to be more flexible, with its value determined more by the market demand and supply (Goldstein & Lardy, 2006). Xu, Mao and Tong (2016) reported a 39% appreciation of RMB exchange rate in real term against US dollar based on unit labour cost from 2000 to 2007. Chinese firms, especially those that rely heavily on exports and imports, have to deal with floating exchange rates since then. Exchange rates therefore started to affect the value of Chinese firms.

Existing literature that examined the effect of exchange rate exposure mainly focused on developed countries, but paid little attention to developing countries. Since the growth of the Chinese economy is unprecedented and the change in exchange rate regime is not similar to any developed counterparties, it is interesting to check the effect of RMB exchange rate exposure on the value of Chinese firms. Out of the few studies that focused on the Chinese market, either only one basic method to estimate the exchange rate exposure was used (Aggarwal et al., 2011) or firms from all industries were mixed together (Zhao, 2010). Besides the paucity of research on the effect of exchange rate exposure in China, the crisis period was hardly examined at all. In the summer of 2015, the Chinese stock market suffered from a tremendous crash. The Shanghai Stock Exchange (SSE) Composite Index dropped nearly 50 per cent from June 12, 2015 to January 28, 2016. Investors and global trading partners were surprised and largely affected by the turbulence. This paper therefore aims to fill in the gap of the crisis period. Hence the main research question is: What is the effect of RMB exchange rate exposure on stock price of Chinese firms during the 2015 stock market crash on an industry basis?

This paper investigates the exchange rate exposure at the industry level. In addition to the basic model introduced by Adler and Dumas (1984), a lagged change in exchange rate is added to test the lagged effect in addition to the contemporaneous effect. The asymmetric nature of exchange rate exposure is also evaluated. Data for 1,347 Chinese firms are collected and the period from January 2009 to December 2016 is divided into three sub periods so as to examine the effect of the stock market crash. The general conclusion is that there exist both contemporaneous and lagged effects of exchange rate exposure on stock return of Chinese firms, but the lagged effect is statistically significant for a higher percentage of industries. The asymmetric nature of exchange rate exposure cannot be ignored either. The stock market crash generally intensifies the effect of both the contemporaneous and the lagged effect.

The remainder of this paper will be structured as follows: Chapter 2 will perform a literature review on exchange rate exposure. The empirical methodology will be introduced in Chapter 3 followed by a description of relevant data. The next chapter will explain the empirical results. The conclusion and limitations of this paper will be presented in the last chapter.

(8)

3

CHAPTER 2. Literature Review

In this chapter, three types of exchange rate exposure will first be introduced followed by an explanation on why different industries should experience varying extent of exchange rate exposure. Sections 2.3 and 2.4 will discuss the effect of lagged variations in exchange rate and the asymmetric character of exchange rate exposure respectively. This chapter will end with a brief description of the Chinese stock market.

2.1 Different types of exchange rate exposure

Researchers have classified exchange rate exposure into three different types: transaction, translation and operating exposure (Shapiro, 1996 and Nydahl, 1999). Sercu and Uppal (1995) defined the combined effect of transaction and operating exposure as economic exposure.

Transaction exposure arises when a company has an account receivable or payable that is denominated in a foreign currency (Nydahl, 1999). The existence of a time lag between the date when a firm commits to a financial obligation and the actual transaction date gives rise to the possibility of a difference in purchase price between these two dates if the obligation is denominated in the counterparty’s currency (Arcelus, Gor, and Srinivasan, 2013). As mentioned in Bodnar et al.’s survey (1995), many companies consider transaction exposure as a problem. Nevertheless, this kind of exposure is generally well defined and short-term since only the period between the two dates matters. Therefore, it can be hedged by proper derivatives if the firm is willing to (Nydahl, 1999).

Translation exposure arises from the need to translate the financial accounting statements from local currencies of the countries where the foreign affiliates are located in to the currency of the parent firm (Hagelin, 2003). Consider a Chinese multinational firm that has subsidiaries in several different countries operating in local currencies. Subsidiaries do not suffer from exchange rate risk when financial statements are expressed in local currencies. However, shareholders of the parent firm are more interested in RMB, making themselves exposed to exchange rate fluctuations since the financial statements of foreign affiliates need to be translated to RMB-based. Jorion (1990) summarized translation exposure as the difference between short-term foreign monetary assets (fully exposed to exchange rate risk) and domestic monetary assets (not exposed to exchange rate risk), leaving the value of real assets affected by exchange rate changes.

Operating exposure arises when future operating cash flows fluctuate with respect to changes in real exchange rates, regardless of whether the cash flows are denominated in local or home currencies (Grant and Soenen, 2004). Most firms consider operating exposure as a more severe source of risk than transaction exposure since transaction exposure is short-term, whereas operating exposure affects a firm as long as it has foreign buyers, suppliers or competitors (Grant and Soenen, 2004). Operating exposure also exists when a firm’s domestic buyers or suppliers are connected with foreign buyers, suppliers or competitors. It is therefore considered to be dependent on the operations of the firm, for example, the competitive structure and locations of factories (Nydahl, 1999). This suggests

(9)

4 that the ability to relocate factories from countries with higher to countries with lower variations of exchange rates could mitigate operating exposure (Jorion, 1990).

All three forms of exposure suggest an effect of currency risk on firm value. There has been an increasing number of studies on exchange rate exposure since a floating exchange rate system was introduced in the US in early 1970s (Kiymaz, 2002). Researchers have not yet reached an agreement on the effect of exchange rate exposure on firm value. Following the method suggested by Adler and Dumas (1984), Jorion (1990) collected a sample of 287 US firms for the period between 1971 and 1987 and reported statistically significant impact of exchange rate changes on stock returns for only 15 firms. This impact increases with the degree of foreign involvement. Similarly, Bortov and Bodnar (1994) found little effect of exchange rate exposure on stock prices for 208 US firms from 1978 to 1989. Their regression results show that contemporaneous changes in the US dollar hardly explain abnormal returns, but lagged changes are negatively correlated with abnormal returns. On the contrary, using a similar method, He and Ng (1998) investigated 171 Japanese multinationals from 1979 to 1993 and discovered that 25 per cent of the firms are statistically significantly exposed to exchange rate fluctuations. Lagged changes in exchange rate do not explain abnormal returns in Japan, unlike that in the US. Nydahl (1999) reported that about 26 per cent of a sample of 47 Swedish firms for the period from 1990 to 1997 has statistically significant exchange rate exposure and the lagged effect does not play a significant role. Since the results from existing literature do not agree with each other and that the Chinese stock market is developing fast, it would be interesting to examine the effect of exchange rate exposure on the Chinese market and the results could provide insight to investors on deciding their portfolio. As the value of a firm should be reflected in its stock price, the first null hypothesis of this paper is that foreign exchange exposure will not affect stock price.

2.2 Exchange rate exposure in different industries

Currency risk exposure does not affect firms in different industries in the same way. The extent of the influence should depend on what the industry does (Bodnar and Gentry, 1993). The degree of involvement in exports and imports, the market characteristics of upper stream industries to obtain inputs, and the amount of foreign investments determine the degree that an industry is connected to the international environment, which in turn affect its currency risk exposure. According to the activities they are involved in, Bodnar and Gentry (1993) divided the industries into six categories: non-traded good producer, exporter, importer, import competitor, user of internationally-priced inputs and foreign investor.

Traded goods industries and non-traded goods industries (goods whose international trade is hindered by high transportation costs) may be influenced differently by exchange rate exposure. As predicted by Dornbusch (1974) and Gavin (1988), an appreciation of the home currency generates a relocation of resources from traded to non-traded goods industries as long as capital is more sector specific than other inputs for production. This raises the market value of capital of non-traded goods

(10)

5 relative to that of traded goods in the short run, indicating a positive relationship between domestic currency appreciation and the value of non-traded goods industries (Bodnar and Gentry, 1993). An appreciation of the home currency reduces the amount of domestic currency needed to purchase one unit of foreign currency and increases the amount of foreign currency needed to purchase one unit of domestic currency. With the price of domestic products denoted in home currency unchanged, an appreciation of the home currency increases the amount of foreign currency required to purchase domestic products. Exporters are hampered due to the combination of a decrease in foreign demand and a lower price cost margin. Similarly, with the price of foreign products denoted in foreign currency unchanged, an appreciation of the home currency reduces the amount of home currency needed to purchase foreign products. Importers’ cash flows increase because of increasing demand and higher price cost margins, benefiting the import sector. Whereas under domestic currency appreciation, the less amount of home currency needed to purchase foreign products increases the price competitiveness of foreign imports and attracts more demand. The import competitors are negatively influenced by an appreciation of the domestic currency through loss in demand and reduction in price cost margins. When considering users of internationally-priced inputs which includes inputs imported from foreign countries as well as domestic inputs whose price is determined by the world market, an appreciation of home currency lowers the price of these inputs. The production cost decreases and the price cost margin improves, profiting the users of internationally-priced inputs. Foreign investors suffer from translating exposure since their current and future cash flows are denominated in foreign currency and are affected by exchange rate when translating back to home currency. The value of industries with foreign assets denominated in foreign currencies therefore decreases with an appreciation of the home currency. To sum up, an appreciation of the home currency decreases the value of exporters, import competitors, and foreign investors and increases the value of non-traded goods producers, importers and users of internationally-priced inputs.

Bodnar and Gentry (1993) added a trade-weighted exchange rate to the market model for a series of industries including traded and non-traded, manufacturing and services industries in Canada, Japan and the US. Taking ten-year data for Canada and the US and five-year data for Japan, they discovered that between 20 and 35 per cent of industries suffers from statistically significant exchange rate exposures for all three countries and variations in exchange rate affect industry returns at an economy-wide level. This effect is larger in Canada and Japan than in the US. They also proved that the extent of activities (as mentioned above) that an industry is involved in influences an industry’s exchange rate exposure as predicted. Williamson (2001) focused specifically on the automotive industry and confirmed the effect of foreign exchange exposure on firm value when analyzing a sample of automotive firms from the US and Japan. Furthermore, he reported that industry competition and firm operation are essential in determining the effect of exchange rate exposure. Akay and Cifter (2014) collected a sample of 173 Turkish firms from 2002 to 2010. Employing industry-weighted exchange rate indices, they concluded that the chemical, metal-machinery, non-metal and

(11)

6 textile industries are exposed to exchange rate variations since they are heavily involved in exports and imports. Among all variables examined, industry-based trade openness explains the exchange rate exposure to the largest extent. A consensus on the industry effect on exchange rate exposure is largely reached among existing literature, the second null hypothesis is therefore: the effect of exchange rate exposure on stock price is the same for all industries.

2.3 Lagged effect of exchange rate

Besides the contemporaneous effect, researchers also investigate the lagged effect of exchange rate exposure on stock prices. Bartov and Bodnar (1994) are considered to be the first to examine the lagged effect when they found little explanatory power in contemporaneous changes of exchange rate. A lagged change in exchange rate could possibly affect stock price because financial information is released to the public with a time lag and multinationals need time to adjust to a changing economic environment, in this case, a change in exchange rate. Their regression results show that lagged changes in exchange rate are negatively correlated with abnormal returns and that the effect is less substantial in the second half of the sample period from 1984 to 1989 than that in the first half of the sample period from 1978 to 1983. Examining a sample of 572 US firms from 1983 to 1994, Shin and Soenen (1999) supported Bartov and Bodnar (1994) by discovering a significant relationship between lagged exchange rates and stock prices for a one-month period after the fiscal year end, but the lagged effect fades after that. Using annual stock return from 1984 to 2009 of multinationals from eight developed countries and classifying them into nine major industries, Inci and Lee (2014) concluded that lagged exchange rate has a significant impact on stock returns for the majority of the countries and sectors. On the other hand, Doidge et al. (2006) employed firm-level data from 18 countries but found small economic magnitude and statistically insignificant effect of lagged exchange rate changes except for the US. Aggarwal and Harper (2010) also found that the effect of lagged changes in exchange rate is consistent with that of contemporaneous variations when looking at 1,265 US firms from 1990 to 2003. No consensus on the effect of lagged changes in exchange rate on stock prices could be drawn from existing literature. Therefore, this paper will also examine this perspective especially for the Chinese market. The third null hypothesis is as follows: lagged changes in exchange rate do not affect stock prices for Chinese firms.

2.4 Asymmetric exchange rate exposure

Most researchers, for example, Bodnar and Gentry (1993) and Jorion (1990), implicitly assume a symmetric exchange rate exposure, meaning that the effect of an exchange rate appreciation and depreciation on financial performance is symmetric. Under this assumption, net exporters are expected to experience a decrease in stock price under an appreciation of the domestic currency and an increase in stock price under a depreciation of the domestic currency. Similarly, net importers should undergo a negative change in stock price due to domestic currency depreciation and a positive change in stock

(12)

7 price due to domestic currency appreciation. However, this is not necessarily the case since corporate behaviors like pricing-to-market, hysteresis and hedging could result in asymmetric exposure on multinationals.

Pricing-to-market refers to the adjustment of export prices based on the degree of competition in foreign markets (Koutmos & Martin, 2003). Knetter (1994) suggested two scenarios that could result in asymmetric markups on exports. In one case which exporters face capacity constraints due to trade restrictions, greater pricing-to-market behaviors are associated with an appreciation of domestic currency since the constraints exclude the possibility of increasing sales volume. In the other case which exporters aim for market share, they would maintain its export price in domestic currency and let the export price in foreign currency fall under domestic currency depreciation so as to increase sales volume and market share. When there is domestic currency appreciation, exporters maintain its export price in foreign currency instead of allowing it to increase so as to protect its market share. Therefore, there is less pass-through effect under domestic currency appreciation than depreciation. These two scenarios result in different extent of pricing-to-market behaviors, which lead to varying cash flows, and hence suggest an asymmetric exchange rate exposure.

Hysteresis could also result in asymmetric exposures. Hysteresis refers to the situation when the effect remains after the original causes fade. In terms of exchange rate, a depreciation of the domestic currency attracts new export competitors to enter the market. Hysteretic behaviors occur when these new firms remain in the market once domestic currency starts to appreciate (Koutmos & Martin, 2003). They have to maintain high entry costs and sunk cost investments under domestic currency appreciation, inducing an asymmetric competitive environment. When considering cash flows, the entrance of new export competitors during a currency depreciation could possibly reduce the cash flows of existing exporters compared to the case if they do not enter the market. Hysteretic behavior of new entrants not leaving the market when domestic currency starts to appreciate is likely to drive a decrease in cash flows for both incumbents and new entrants (Christophe, 1997). Hence, hysteresis could cause an asymmetric exchange rate exposure.

Another corporate behavior that could result in asymmetric exchange rate exposure is hedging. Forward and futures could protect a company from financial losses due to unfavorable exchange rate variations. However, they also eliminate the chance of financial gains when the exchange rate moves in the favorable direction. Currency options, on the other hand, provide not only downside protection but also unlimited upside potential. The asymmetric payoff from currency options poses a non-linear effect on a firm’s cash flow and firm value, exposing firms to asymmetric exchange rate variations.

These three forms of corporate behaviors provide theoretical evidence for asymmetric exchange rate behavior. This is proved empirically when Di Iorio and Faff (2000) analyzed both daily and monthly data of 24 Australian industries from 1988 to 1996 and discovered exchange rate exposure asymmetry. They discovered more evidence of asymmetric behavior in the lagged changes in exchange rate than contemporaneous changes. The effect is more significant when employing daily

(13)

8 data as compared to monthly data. Koutmos & Martin (2003) also discovered asymmetric exchange rate exposure based on a sample of weekly data for Germany, Japan, United Kingdom and the United States. They found that 40 per cent of all significant exposures are asymmetric and asymmetry is more likely to occur in the financial (due to hedging) and consumer non-cyclical sector (due to pricing-to-market and/or hysteresis). The existence of asymmetric exchange rate exposure both theoretically and empirically gives rise to the fourth null hypothesis of this paper: Chinese firms do not exhibit asymmetric exchange rate exposure.

2.5 The Chinese stock market

The economic reform, known as the open-door policy, was introduced in 1978 to strengthen China’s economy by transforming it from centrally-planned to market-oriented. After the initial success, the effect of the reform started to fade. To raise fund and improve the efficiency of state-owned enterprises, the government allowed the issues of bonds and stocks, which were later allowed to trade in the Shanghai (late 1990) and Shenzhen (early 1991) stock exchanges. The Chinese stock market was officially established since then and has been growing fast. Since the Chinese stock market emerged and has been operated as a combination of planned and market-oriented elements, it is essentially different from modern stock markets like the US stock market in terms of institutional set up (Ma, 2017). This difference makes the research on the Chinese stock market beneficial for investors who are interested in China.

The Chinese stock market has experienced roller coaster dynamics in the 2014-2015 period. The latest bubble started to build up since July 1, 2014 until it reached the peak on June 12, 2015, corresponding to an approximate 150 per cent growth of the Shanghai Stock Exchange (SSE) Composite Index in less than one year. The stock market then collapsed by 40 per cent on September 28, 2015 and further reached its trough on January 28, 2016 when the value loss of SSE Composite Index reached almost 50 per cent. In one and a half year, the Chinese stock market encountered peak to trough, accompanied by a loss of $5.6 trillion or more than half of China’s GNP. Contrasting to other developed markets, the Chinese stock market is mainly constituted of small and medium size Chinese investors. Their susceptibility to rumors and speculative behaviors are considered as the main drivers of the stock market crash (Sornette et al., 2015). Since this crash has tremendous effect on Chinese firms and investors, it is interesting to investigate whether the stock market crash plays a role in determining the effect of exchange rate exposure. Therefore, the last hypothesis is that the effect of exchange rate exposure does not change under stock market crash. Given that this crash resembles the Chinese share price bubble from 2005 to 2008, this paper will consider an 8-year period from 2009 to 2016.

(14)

9

CHAPTER 3. Empirical Methodology and Data

Section 3.1 will first discuss some conflicting factors used in previous literature and explain which factor will be used in this paper. This section also includes the main models that will be employed. Section 3.2 will show the data collection procedure and descriptive statistics.

3.1 Empirical methodology

3.1.1 Real or nominal exchange rate

Previous literature differs on the choice of exchange rate. On the one hand, Williamson (2001) stated that a nominal exchange rate should not affect the real value of a firm when it does not possess foreign assets or liabilities. He therefore employed real exchange rate in his model. Similarly, other researchers like Bredin and Hyde (2011) and Doukas et al. (2001) specified that real exchange rate is used in their research. On the other hand, Bodnar and Gentry (1993) argued that using real exchange rate assumes the possibility to observe inflation instantaneously for calculations. All variables in the regression also need to be adjusted for inflation when the real exchange rate is used for consistency purposes (Khoo, 1994). Despite the difference between nominal and real exchange rate, Griffin and Stulz (2001) argued that in low inflation countries, the two forms of exchange rate are strongly correlated so that the choice of which exchange rate to use does not make a difference. Bodnar and Gentry (1993) documented a correlation of 0.97, 0.95, and 0.98 between nominal and real trade-weighted exchange rates in the sample period for the US, Canada and Japan respectively. Choi and Prasad (1995) agreed on the consistency of results between the two forms of exchange rate by checking the impact of both real and nominal exchange rate changes on 409 US multinationals from 1978 to 1989 and observed very marginal differences. The World Bank showed 1.6%, 2.1%, 0.5% and 1.6% increase in consumer price index, a main proxy for inflation, for China, the US, Japan and Canada in 20172_{. Since China’s inflation rate is comparable to these countries and is considered low,} nominal exchange rate changes will be used in this paper.

3.1.2 Trade-weighted exchange rate

Instead of bilateral nominal exchange rates, this paper will employ a traded-weighted nominal exchange rate index since it can capture the economy-wide and aggregate variations in domestic currency (Aggarwal & Harper, 2010). This index reflects the actual currency environment that a firm would face. Merging several exchange rates together into a convenient measure, trade-weighted exchange rate saves researchers from having to include many bilateral exchange rates (Jorion, 1990). In addition, it avoids the multicollinearity problem between bilateral exchange rates since some exchange rates are fixed or almost fixed relative to each other (Jorion, 1990; Chow & Chen, 1998). Although Williamson (2001) noted that using trade-weighted exchange rate ignores the fact that some

(15)

10 firms may only be exposed to one or a few currencies in the basket or to currencies that are not in the basket, many researchers employ a trade-weighted exchange rate index as the exchange rate used to estimate the exposure (He & Ng, 1998; Chaieb & Mazzotta, 2013; Hutson & Laing, 2014, etc.). In this paper, currencies of China’s top five trading partners will be used. The trading partners are determined based on the sum of export and import volume and are updated annually. The weight of each currency in the index is directly proportional to the sum of exports and imports.

3.1.3 Empirical models 3.1.3.1 The basic model

Adler and Dumas (1984) suggested a simple regression model to estimate the exchange rate exposure on stock price in which the stock return is the dependent variable and the exchange rate changes is the independent variable. The stock price is used to represent the firm’s value. The following equation can be used to obtain the exchange rate exposure:

𝑅

_𝑛,𝑡

= 𝛼

₀

+ 𝛽

₁

Δ𝑆

_𝑡

+ 𝛼

₁

𝑅

_𝑚𝑡

+ 𝑒

_𝑛,𝑡 (1)

𝑅

𝑛𝑡 is the rate of return on the nth firm’s stock in period t. Δ𝑆𝑡 is the percentage change in nominal

trade-weighted exchange rate index in period t. In this paper, the exchange rate is defined as the amount of foreign currency per unit of RMB such that a positive Δ𝑆𝑡 indicates an appreciation of the

home currency. 𝑅_𝑚𝑡 is the rate of return on market portfolio, which is used to control for market movements. Therefore 𝛽₁ measures the firm’s exposure to exchange rate changes, after taking into account the overall market’s exposure to the exchange rate variations. As discussed earlier, an appreciation of domestic currency makes exported goods more expensive in terms of foreign currencies, leading to a decrease in foreign demand and foreign sales revenue. Exporting firms’ value will be adversely affected by a home currency appreciation. Whereas importing firms benefit from this appreciation since imported goods are cheaper in terms of domestic currency, reducing their cost. Therefore the 𝛽₁

coefficient should be positive for net importers and negative for net exporters.

3.1.3.2 The lagged effect model

To investigate the effect of lagged changes in exchange rate on stock price, a lagged variable Δ𝑆_𝑡−1 in addition to the dependent variables in the basic equation is added. This term represents the percentage change in nominal exchange rate in the previous period. The original equation becomes:

𝑅

𝑛,𝑡

= 𝛼

0

+ 𝛽

1

Δ𝑆

𝑡

+ 𝛽

2

Δ𝑆

𝑡−1

+ 𝛼

1

𝑅

𝑚𝑡

+ 𝑒

𝑛,𝑡 (2)

3.1.3.3 The asymmetric model

To test for asymmetry in exchange rate exposure, this paper will follow Di Iorio and Faff (2000) by defining the following three dummy variables according to the sign of exchange rate movement:

𝐷𝑛𝑒𝑔 is a dummy variable that takes a value of one in period t if the exchange rate

(16)

11 𝐷𝑝𝑜𝑠 is a dummy variable that takes a value of one in period t if the exchange rate

has appreciated for more than 0.5 per cent and zero otherwise;

𝐷𝑛𝑒𝑢𝑡 is a dummy variable that takes a value of one in period t if both 𝐷𝑛𝑒𝑔 and

𝐷𝑝𝑜𝑠 are zero and takes a value of zero otherwise.

Equation (1) and (2) are restructured to accommodate the dummy variables:

𝑅𝑛,𝑡= 𝛼0+ 𝛼1𝑅𝑚𝑡+ 𝛽1𝐷𝑛𝑒𝑔Δ𝑆𝑡+ 𝛽2𝐷𝑝𝑜𝑠Δ𝑆𝑡+ 𝛽3𝐷𝑛𝑒𝑢𝑡Δ𝑆𝑡+ 𝑒𝑛,𝑡 (3)

𝑅𝑛,𝑡= 𝛼0+ 𝛼1𝑅𝑚𝑡+ 𝛽1𝐷𝑛𝑒𝑔Δ𝑆𝑡+ 𝛽2𝐷𝑝𝑜𝑠Δ𝑆𝑡+ 𝛽3𝐷𝑛𝑒𝑢𝑡Δ𝑆𝑡

+𝛽4𝐷𝑛𝑒𝑔,t−1Δ𝑆𝑡−1+ 𝛽5𝐷𝑝𝑜𝑠,𝑡−1Δ𝑆𝑡−1+ 𝛽6𝐷𝑛𝑒𝑢𝑡,𝑡−1Δ𝑆𝑡−1+ 𝑒𝑛,𝑡

(4)

3.2 Data

Monthly stock prices for companies with active A shares listed on Shanghai and Shenzhen Stock Exchange can be obtained from Datastream for an 8-year period from January 2009 to December 2016. A total of 1,347 firms are selected and are classified into 14 industries according to the China Securities Regulatory Commission (CSRC). The nominal exchange rate can be retrieved from the International Financial Statistics database by IMF. The trade-weighted exchange rate index is calculated based on the trading volumes of China’s top five trading partners. This information is determined according to the import and export volume, which can be obtained from the World Integrated Trade Solution organized by the World Bank. The Shanghai Stock Exchange Composite Index is used as a proxy for the change of the overall Chinese stock market. This data is also retrieved from Datastream. The entire sample period is divided into three sub periods: before the stock market crash (January 1, 2009 to June 30, 2014), during the stock market crash (July 1, 2014 to January 31, 2016), and after the stock market crash (February 1, 2016 to December 31, 2016).

Table 1 reports the descriptive statistics of the industry returns. As can be seen from Column 1, the Chinese stock market is mainly constituted by firms from the manufacturing, construction, real estate and mining industry. The average monthly stock return, reported in Column 2, varies between 1.67% and 3.14%. The information technology and scientific research industry have the highest stock returns whereas the transportation and utilities industry have the lowest. The mean stock return for each industry does not differ much, with the average for the entire sample being 2.41%. When considering stock return volatility which is shown in Column 3, the real estate and finance industry are associated with the highest volatility whereas the health and social work and transportation industry are with the lowest. The standard deviation varies between 12.77% and 21.50% and the overall standard deviation for the entire sample is 16.48%. Table 2 reports the variations in trade-weighted exchange rate index. For the whole sample period, the index increased by 0.06% per month. The change in the exchange rate index is positive before and during the stock market crash but is negative after the crash. During the stock market crash, the exchange rate index experiences the greatest increase as well as the highest volatility.

(17)

12 Table 1. Stock return based on industry

Obs Mean Std dev.

Agriculture 23 0.0233 0.1406

Mining 117 0.0213 0.1692

Manufacturing 522 0.0248 0.1622

Utilities 75 0.0201 0.1620

Construction 170 0.0236 0.1486

Wholesale and retail 74 0.0221 0.1367

Transportation 48 0.0167 0.1358 Information technology 28 0.0314 0.1632 Finance 26 0.0216 0.1785 Real estate 128 0.0271 0.2150 Scientific research 45 0.0289 0.1542 Resident service 22 0.0244 0.1390

Health and social work 4 0.0265 0.1277

Entertainment 65 0.0252 0.1745

Total 1,347 0.0241 0.1648

Table 2. Change in foreign exchange rate

Full sample Pre-crash Crash Post-crash

Mean 0.0006 0.0010 0.0027 -0.0054

(18)

13

CHAPTER 4. Results and Analysis

This chapter presents the empirical results and discussions on the four models mentioned above. Section 4.1 investigates the contemporaneous effect of exchange rate exposure whereas Section 4.2 adds the effect of lagged exchange rate variations. Section 4.3 moves on to assess the asymmetric nature of exchange rate exposure based on both the basic and the lagged effect model. Section 4.4 performs robustness checks on the results obtained.

4.1 The exchange rate exposure of Chinese industries

Table 3 shows the industry-level exchange rate exposure, 𝛽₁ as defined in equation (1), for the full sample period and the three sub periods. The robust standard errors are presented in parentheses under the coefficients. Positive 𝛽1 suggests that an appreciation of RMB against other currencies benefits the

stock return of Chinese firms in this industry.

During the full sample period, 6 out of the 14 industries classified according to the China Securities Regulatory Commission are associated with negative exchange rate exposure. The coefficients are statistically significant at standard levels for 3 out of 6 industries that have negative exposure and 4 out of 8 industries that have positive exposure. Negative coefficients indicate that firms in these industries suffer from a loss in stock return when there is domestic currency appreciation. To discover the effect of stock market crash on exchange rate exposure, the full sample period is divided into three subsample periods. Looking at the results for these three periods, the coefficients in the period before the stock market crash are generally consistent in sign and magnitude with that of the full sample period. Nevertheless, the stock market crash can be seen as an important factor that can influence the effect of the exchange rate exposure. 64% of the industries (9 out of 14) obtains a coefficient that is smaller than that for the period before the stock market crash. This means that when domestic currency appreciates, firms enjoy a less increase in stock return under a stock market crash than the time before the crash when the coefficient is positive and bear more loss in stock price when the coefficient is negative. The stock market crash diminishes the positive effect and intensifies the negative effect of domestic currency appreciation. During the period after the crash, the coefficients get generally smaller, with some of them changing from positive to negative. This suggests that the hindering effect of the stock market crash on the currency risk exposure does not fade after the crash. In some industries, firms even start to suffer from a loss. This could be explained by hysteresis: the stock market crash could still exert an effect even after it is over.

The result that 7 out of 14 of the Chinese industries are statistically significantly exposed to exchange rate changes differs from previous evidence in developed countries. Bodnar and Gentry (1993) reported that only 11 of 39 US industries, 4 of 19 Canadian industries and 7 of 20 Japanese industries have significant exchange rate exposures. On the other hand, Akay and Cijfter (2014) found 4 of 7 Turkish industries express exchange rate exposure that is statistically significant, this is close to

(19)

14 the results from this paper. The bigger percentage of significant exposures that Chinese industries experience than developed countries is in line with her great existence in international trade. Being the top exporter and importer of merchandise and commercial services, it is understandable that more industries are affected by exchange rate variations than other developed countries. Another reason for this larger percentage of significant exposure could be due to the fact that there are more Chinese firms not able to hedge the exposure than firms in other countries. The Chinese market is not fully developed in hedging yet.

Table 3. Industry-level exchange rate exposure

Model: 𝑅𝑛,𝑡= 𝛼0+ 𝛽1Δ𝑆𝑡+ 𝛼1𝑅𝑚𝑡+ 𝑒𝑛,𝑡

Industry

𝛽1

Full sample period Before During After

Agriculture -0.4573** -0.4730** -0.3834*** -0.6432 (0.2258) (0.2568) (0.1455) (0.6622) Mining -0.0264 -0.0227 -0.0258 -0.0240* (0.1246) (0.1268) (0.0172) (0.0125) Manufacturing -0.1396** -0.2010** -0.2055*** -0.5611*** (0.0615) (0.0809) (0.0429) (0.1353) Utilities 0.3903 0.2226 0.1683*** -0.3592 (0.2749) (0.4316) (0.0954) (0.2677) Construction 0.5232** 0.3305*** 0.3178*** -0.0828 (0.2602) (0.1061) (0.1078) (0.2402)

Wholesale and retail -0.1496 -0.5542*** -0.3901 -0.4025

(0.1356) (0.1578) (0.3542) (0.3345) Transportation 0.3710*** 0.4096*** 0.2564** -0.0453 (0.0885) (0.1453) (0.1112) (0.2843) Information technology 0.2395 0.3015 0.4823*** -0.4943*** (0.2592) (0.3181) (0.1601) (0.1656) Finance 0.6970*** 0.7249** 0.5059 -0.3511 (0.2565) (0.2639) (0.4321) (0.3148) Real estate -0.3888*** -0.7236*** -0.3952 -0.0717 (0.1458) (0.1814) (0.3168) (0.2731) Scientific research 0.0678 -0.0169 0.4260 -0.4059*** (0.1825) (0.2197) (0.4676) (0.1083) Resident service -0.2407 0.0264 -0.2750 -0.5348* (0.2252) (0.2880) (0.5190) (0.2883)

Health and social work 0.3763*** 0.5299* 0.5013** -0.5956

(0.1227) (0.3007) (0.2557) (0.4468)

Entertainment 0.2250 0.1021 0.1005 -0.6016***

(0.1510) (0.1806) (0.3696) (0.1854)

***, **, and * indicate statistical significance at 1%, 5% and 10% significance levels. Robust standard errors in parentheses.

(20)

15 Table 4. Lagged effect of exchange rate exposure

Model: 𝑅𝑛,𝑡= 𝛼0+ 𝛽1Δ𝑆𝑡+ 𝛽2Δ𝑆𝑡−1+ 𝛼1𝑅𝑚𝑡+ 𝑒𝑛,𝑡

Industry

𝛽1 𝛽2

Full Before During After Full Before During After

Agriculture -0.4477*** -0.3524 -0.3791*** -0.3093 -0.4680** -0.4354 -0.6639*** -0.4524 (0.1652) (0.2762) (0.1467) (0.7601) (0.2053) (0.2776) (0.1179) (0.5067) Mining -0.0705 -0.0396 -0.0839 -0.0537 -0.0675 -0.0891 -0.1192 -0.0545 (0.1308) (0.1512) (0.1444) (0.0394) (0.1538) (0.1505) (0.2292) (0.2959) Manufacturing -0.1510** -0.2084*** -0.2115*** -0.2447*** -0.2229*** -0.2290*** -0.2541*** -0.3866*** (0.0684) (0.0546) (0.0454) (0.0900) (0.0658) (0.0548) (0.0417) (0.0607) Utilities 0.3099 0.2943 0.3592** -0.0669 0.3560** 0.3210* 0.3675 0.2919 (0.2620) (0.4030) (0.1517) (0.3104) (0.1496) (0.1778) (0.2302) (0.3948) Construction 0.3755*** 0.4140*** 0.3673** 0.2103 0.1771 0.3773*** -0.1017*** -0.0794** (0.1042) (0.1222) (0.1622) (0.2541) (0.1234) (0.1162) (0.0377) (0.0326) Wholesale and retail -0.2765 -0.3383* -0.4089*** -0.3061*** -0.2635* -0.2587 -0.3317* -0.3568***

(0.1711) (0.1823) (0.1904) (0.1023) (0.1553) (0.1691) (0.1816) (0.1122) Transportation 0.2697*** 0.2727 0.2569* 0.1853*** 0.3239** 0.3633** 0.2557*** -0.0763 (0.0837) (0.1751) (0.1381) (0.0508) (0.1589) (0.1632) (0.0907) (0.0937) Information technology 0.2323 0.3240 0.2307 -0.3443* 0.2426 0.4079 0.3173** -0.0541 (0.3169) (0.3807) (0.1716) (0.1969) (0.3175) (0.3953) (0.1250) (0.5498) Finance 0.3165 0.6527*** -0.2835 -0.3640 0.5764*** 0.5783 0.5292** 0.0351 (0.3057) (0.2444) (0.4318) (0.3332) (0.2043) (0.4048) (0.2293) (0.4757) Real estate -0.2606 -0.5306*** -0.4343*** -0.0296 -0.2549* -0.2272 -0.2473** -0.2743 (0.1621) (0.1919) (0.1266) (0.3333) (0.1499) (0.1678) (0.1009) (0.4284) Scientific research 0.0322 -0.0222 0.0576 -0.4947*** -0.0507 0.0113 -0.0692 -0.1774 (0.2122) (0.2482) (0.1661) (0.1726) (0.2230) (0.2518) (0.1264) (0.6904) Resident service -0.1353 0.0847 0.0651 -0.3945 -0.1747*** -0.1238 -0.1615** -0.1609 (0.2653) (0.3218) (0.5899) (0.6588) (0.0586) (0.3435) (0.0630) (0.3149) Health and social work 0.3684** 0.3478** 0.5921 0.2890 0.3492** 0.4041*** 0.3846** -0.1994

(0.1592) (0.1746) (0.4073) (0.6635) (0.1699) (0.1055) (0.1603) (0.3986)

Entertainment 0.3781** 0.1274 0.4716*** -0.2981*** -0.3044* 0.1586 -0.4130*** -0.0289

(0.1767) (0.2010) (0.1034) (0.1110) (0.1665) (0.1862) (0.1298) (0.5827) ***, **, and * indicate statistical significance at 1%, 5% and 10% significance levels.

(21)

16

4.2 The effect of lagged changes in exchange rate

To test the lagged effect of exchange rate variations, a lagged variable is added to the basic model and the results are shown in Table 4. 𝛽₁ and 𝛽2 are the coefficients of the contemporaneous and lagged

changes in exchange rate and are shown in Columns 1 to 4 and Columns 5 to 8 respectively for different time periods.

Looking at 𝛽₁, 6 out of 14 industries have a negative coefficient, which indicates a decrease in stock return under a domestic currency appreciation in the full sample period. Among the 6 industries, the negative coefficients for agriculture and manufacturing are statistically significant at standard levels. 8 industries obtain a positive coefficient, allowing them to benefit from domestic currency appreciation. The coefficients are statistically significantly different from zero for 4 of the 8 industries at standard levels. Overall, 43% of the industries have a statistically significant contemporaneous exchange rate exposure in the full sample period. Changes in 𝛽1

across different subsample periods is

similar to that in the basic model in terms of sign and magnitude. During the stock market crash, 57% of the industries obtain a statistically significant coefficient, greater than that before or after the stock market crash (43% for both). The lagged effect, which is inferred from 𝛽2, differs from the

contemporaneous effect to a small extent. As can be seen from Column 5, 10 out of 14 industries have coefficients that are statistically significantly different from zero at standard levels in the full sample period. This means that lagged changes in exchange rate could explain the change in stock price for firms in 71% of the industries. Half of the industries have a negative 𝛽2 whereas the other half are

associated with positive coefficients. When dividing the full sample period into three subsample periods, the coefficients for the period before the stock market crash are generally consistent with that during the full sample period in terms of sign and magnitude. Except for utilities, all other industries obtain a smaller coefficient during the stock market crash than that before the crash. This is consistent with the contemporaneous effect: the stock market crash worsens the adverse effect of domestic currency appreciation on stock price when the coefficient is negative and diminishes the positive effect when the coefficient is positive. The difference between the contemporaneous and lagged effect is that, 𝛽2 is statistically significant at standard levels for 11 industries, corresponding to 79% of all

industries. This percentage is greater than that for 𝛽1 (57%). There are more industries obtaining a

significant coefficient during the stock market crash when considering the lagged effect than the contemporaneous effect. The F-test that examines whether both 𝛽1 and 𝛽2 are zero obtains a p-value

(not shown in the table, available from the author) that is smaller than 0.05 for almost all industries in all periods suggest that both contemporaneous and lagged effect of exchange rate variations explain the change in stock returns.

The results regarding the lagged effect of exchange rate variations from this paper is in accordance to Bartov and Bodnar (1994) and Inci and Lee (2014). Both papers reported a significant

(22)

17 explanation power of lagged changes in exchange rate on stock return when analysing 208 US firms from 1978 to 1989 and multinationals from eight developed countries, classified into nine major industries from 1984 to 2009 respectively. This paper further agrees with Inci and Lee (2014) regarding the contemporaneous effect. Both the contemporaneous and lagged effect play essential roles in affecting stock price. An analysis that only includes the contemporaneous relation between exchange rate changes and stock return is limited and lacks the potential dynamic effect of exchange rate variations on stock prices over time. The results prove that multinationals need time to fully accommodate changes in exchange rates.

4.3 Asymmetric exchange rate exposure

Table 5 and Table 6 report the findings of the asymmetric nature of exchange rate exposure. The results for the basic model augmented by the dummy exchange rate variables are shown in Table 5 and the coefficients for the augmented lagged model are presented in Table 6, with Panel A to D showing the values for the full sample period, the period before, during, and after the stock market crash respectively.

Similar to the general findings of the basic model, there is evidence of statistically significant contemporaneous effect of exchange rate exposure when taking into consideration the potential asymmetric behaviour. During the full sample period, 6 industries have significant coefficients in the negative case (0.5% domestic currency appreciation or more) at standard levels, 4 industries in the positive case (0.5% domestic currency depreciation or more) and 5 in the neutral case (depreciate or appreciate for less than 0.5% of domestic currency). The periods before and after the stock market crash obtain results that are close to that for the whole period whereas the period during the stock market crash shows stronger statistical evidence of asymmetric effect. 5, 7, and 8 industries have significant coefficients in the negative, positive, and neutral cases respectively at standard levels during the stock market crash. Stronger evidence of asymmetric effects is shown when lagged changes in exchange rate are added to the regression. Specifically, as shown in Table 6 Panel A, for the full sample period, 8 coefficients are significant for the negative lagged case, 5 are significant for the positive lagged case, and 6 are significant for the neutral lagged case. The period before and after the stock market crash obtain comparable results as that for the full sample period. Panel C reports the findings for the period during the stock market crash. It is clear that more coefficients are statistically significant at standard levels: 9, 7, and 8 industries have statistically significant coefficients for the negative, positive, and neutral lagged cases.

There is indeed asymmetric effect in both the contemporaneous and lagged changes in exchange rate, but the lagged sensitivity to exchange rate variations is supported by stronger statistically significant evidence. The stock market crash intensifies the effect of both the contemporaneous and lagged terms. This conclusion regarding the asymmetric nature is consistent with Iorio and Faff’s (2000) results for the lagged term but contradicting for the contemporaneous

(23)

18 term. Iorio and Faff (2000) found little evidence of statistically significant contemporaneous effect of exchange rate changes but strong evidence for the lagged term when using both daily and monthly data. Since the coefficients in the basic model are statistically significant for around half of the industries, it is understandable that the contemporaneous effect is significant when adding the potential asymmetry.

4.4 Robustness checks

To determine the robustness of the results, the relations are re-examined by using alternative measures of the variables. Table 7 shows the results of robustness checks for the basic model during the full sample period. Column 1 reports the coefficients in Table 3 as a comparison. As suggested by Aggarwal and Harper (2010), an equally weighted exchange rate index of the top five trading partners for China is used as an alternative for the trade-weighted exchange rate index. The results for this measure are shown in Column 2. Moreover, the Shenzhen Stock Exchange A share price index is used to replace the Shanghai Stock Exchange Composite Index as the proxy for the change of the overall Chinese stock market and the results are tabulated in Column 3.

Similar to the results in Column 1, 8 out of 14 and 7 out of 14 industries obtain significant coefficients at standard levels when equally weighted exchange rate index and Shenzhen Stock Exchange A share price index are used respectively. The coefficients are close to each other for the three measures in terms of sign and magnitude. The results provide statistically significant evidence of contemporaneous effect of exchange rate variations on stock return. The same robustness checks are applied to the three subsample periods and the rest of the models as well to test the lagged effect and the asymmetric nature of exchange rate exposure. Due to space constraint, the coefficients are not listed specifically (The tables are available from the author). The results from the robustness checks when changing specifications largely coincide with that from the original measure. The coefficients from the robustness checks offer strong and consistent support to reject the null hypothesis that the contemporaneous and lagged changes in exchange rate does not affect stock return.

(24)

19 Table 5. Asymmetric nature of exchange rate exposure

Model: 𝑅𝑛,𝑡= 𝛼0+ 𝛼1𝑅𝑚𝑡+ 𝛽1𝐷𝑛𝑒𝑔Δ𝑆𝑡+ 𝛽2𝐷𝑝𝑜𝑠Δ𝑆𝑡+ 𝛽3𝐷𝑛𝑒𝑢𝑡Δ𝑆𝑡+ 𝑒𝑛,𝑡

𝛽1 𝛽2 𝛽3

Industry Full Before During After Full Before During After Full Before During After

1 -0.3335*** 0.0462 -0.0921 -0.3751*** 0.1522 0.1450 0.2152 0.1788 -0.1289 -0.2139 -0.3104* -0.1148 (0.1048) (0.1727) (0.1141) (0.1361) (0.4147) (0.4142) (0.4533) (0.1333) (0.0887) (0.3148) (0.1768) (0.1215) 2 0.2006 0.3579* 0.2956** -0.0854 -0.1924 -0.2357 -0.1961 -0.2551* -0.1472 -0.2067 -0.1981 0.0850 (0.2027) (0.2155) (0.1272) (0.1659) (0.1914) (0.2392) (0.2321) (0.1457) (0.1898) (0.2023) (0.1817) (0.1334) 3 0.2875*** 0.3745*** 0.2363 0.2533 -0.2882*** -0.3342** -0.2607** 0.2825 -0.3140** -0.4352 -0.2530** -0.0814 (0.1012) (0.1259) (0.2026) (0.2506) (0.1043) (0.1395) (0.1275) (0.1765) (0.1503) (0.3465) (0.1241) (0.1648) 4 0.3786 0.3563 0.5032 0.2040 -0.2082 -0.3128 -0.2051 -0.2274 -0.1928 -0.0089 -0.0568 -0.0250 (0.2607) (0.3584) (0.4889) (0.2171) (0.1941) (0.2118) (0.2975) (0.3362) (0.1731) (0.0636) (0.0902) (0.0430) 5 0.4036** 0.4023 0.5125*** 0.5213*** 0.3746** 0.2874 0.4176*** 0.4455** -0.5915* -0.5465** -0.6765*** -0.2755* (0.1616) (0.3818) (0.1803) (0.1782) (0.1584) (0.1957) (0.1402) (0.2057) (0.3156) (0.2778) (0.2538) (0.1415) 6 -0.1038 0.0108 -0.2835 -0.2703*** 0.0456 -0.0241 0.1340** 0.0350 0.3254** 0.2114 0.4792*** 0.2767 (0.2281) (0.2739) (0.5220) (0.0901) (0.2587) (0.2891) (0.0528) (0.0724) (0.1617) (0.1524) (0.1095) (0.4082) 7 0.2549 0.2633 0.0083 -0.0908 0.0997 0.1780 0.1066 0.0695 0.4008*** 0.2135** 0.3187* 0.2012 (0.2361) (0.2487) (0.2159) (0.1391) (0.3777) (0.2928) (0.3321) (0.3345) (0.1161) (0.1073) (0.1766) (0.2597) 8 -0.3812*** -0.2804 -0.1831** -0.2733* -0.2418 -0.3223*** -0.1408 -0.2255* -0.1934 -0.2038 -0.2789** -0.0103 (0.1215) (0.2657) (0.0886) (0.1633) (0.2239) (0.1015) (0.2810) (0.1252) (0.2550) (0.5111) (0.1320) (0.1423) 9 0.1039 0.0160 0.2204** -0.0087 0.2907*** 0.0107 0.1796 0.4127 0.1619 0.1937 0.0448 -0.1076 (0.0709) (0.1639) (0.1033) (0.0159) (0.1107) (0.0882) (0.1903) (0.3733) (0.1505) (0.2921) (0.1267) (0.1610) 10 0.4503** 0.3177 0.2348 0.1654 0.3226 0.5598** 0.5171** 0.3080** 0.1936 0.1958 0.1698*** 0.0769** (0.2164) (0.2641) (0.4161) (0.1500) (0.2533) (0.2295) (0.2171) (0.1541) (0.1729) (0.1650) (0.0486) (0.0381) 11 0.2408 0.2075 0.3194 0.3402** -0.1999 -0.1782 -0.2148** -0.1533** -0.0411 -0.0512 -0.0564 -0.0236 (0.3265) (0.3763) (0.2861) (0.1296) (0.3406) (0.4222) (0.1063) (0.0698) (0.2910) (0.2385) (0.2507) (0.2216) 12 0.2775*** 0.3096*** 0.2873** 0.1988** -0.0512 -0.1478** 0.0911 -0.0722 -0.2182 -0.3318 -0.3839 -0.1327 (0.1040) (0.1054) (0.1359) (0.0835) (0.4500) (0.0667) (0.1613) (0.0669) (0.1947) (0.2159) (0.2904) (0.2076) 13 0.1974 0.0134 0.2312 0.1901 -0.2289 -0.3805 -0.5949* -0.2205 -0.3401*** -0.4164** -0.5806** -0.2167 (0.1975) (0.1881) (0.1695) (0.2854) (0.1478) (0.3043) (0.3239) (0.2218) (0.1319) (0.2031) (0.2899) (0.1439) 14 0.1085 0.2518* 0.0398 -0.0669 0.1686** -0.1268** 0.1819** -0.0630 -0.0685 -0.0950 -0.1588 -0.0008 (0.2498) (0.1504) (0.0409) (0.1012) (0.0804) (0.0506) (0.0982) (0.1393) (0.1636) (0.0726) (0.2728) (0.0486) ***, **, and * indicate statistical significance at 1%, 5% and 10% significance levels.

(25)

20 Table 6. Lagged asymmetric nature of exchange rate exposure

Panel A: Full sample period

Model: 𝑅𝑛,𝑡= 𝛼0+ 𝛼1𝑅𝑚𝑡+ 𝛽1𝐷𝑛𝑒𝑔Δ𝑆𝑡+ 𝛽2𝐷𝑝𝑜𝑠Δ𝑆𝑡+ 𝛽3𝐷𝑛𝑒𝑢𝑡Δ𝑆𝑡 +𝛽4𝐷𝑛𝑒𝑔,t−1Δ𝑆𝑡−1+ 𝛽5𝐷𝑝𝑜𝑠,𝑡−1Δ𝑆𝑡−1+ 𝛽6𝐷𝑛𝑒𝑢𝑡,𝑡−1Δ𝑆𝑡−1+ 𝑒𝑛,𝑡 Industry 𝛽1 𝛽2 𝛽3 𝛽4 𝛽5 𝛽6 Agriculture 0.4062** -0.1563 -0.3542* -0.2892* 0.2762 -0.1908** (0.1932) (0.1089) (0.2091) (0.1510) (0.2039) (0.0914) Mining 0.1034 -0.1908 -0.2029 -0.1033 0.1871 -0.3541 (0.1570) (0.2109) (01592) (0.2103) (0.2520) (0.3034) Manufacturing 0.2054** -0.1977* -0.1033 -0.2984** 0.3211** -0.1981 (0.0910) (0.1032) (0.1290) (0.1322) (0.1431) (0.1732) Utilities 0.4462** 0.1230 -0.3095** -0.3599*** 0.3214 -0.2520 (0.2051) (0.2135) (0.1286) (0.1112) (0.2898) (0.2439) Construction 0.2025 0.1086 -0.3223 -0.2025** 0.2132 -0.3539** (0.2290) (0.2231) (0.5870) (0.0992) (0.1890) (0.1744) Wholesale and retail 0.3015 -0.1892** -0.3521*** 0.0911 0.5541** -0.4948**

(0.2589) (0.0921) (0.1192) (0.2526) (0.2391) (0.2129) Transportation 0.3039 0.1023 -0.4139** -0.1503** -0.3981 0.2840 (0.1990) (0.2086) (0.1766) (0.0722) (0.3340) (0.2921) Information technology 0.4474 -0.4870** 0.3090 -0.0511 0.3312* -0.1881** (0.3982) (0.2053) (0.2912) (0.1462) (0.1919) (0.0898) Finance 0.4980** 0.4592** -0.2908 0.1044 0.3242 0.5422 (0.2360) (0.2022) (0.3329) (0.2334) (0.3001) (0.4848) Real estate -0.1124 -0.2534 -0.4399 0.4129** 0.2521*** -0.4359* (0.2019) (0.1829) (0.3981) (0.1673) (0.0918) (0.2290) Scientific research 0.3587* 0.1749 -0.1056 0.2571 0.3303 -0.2588 (0.2099) (0.2454) (0.1920) (0.3490) (0.3913) (0.3321) Resident service 0.1093 -0.1872 -0.4091* 0.3029** -0.2218 -0.1888 (0.2019) (0.2095) (0.2325) (0.1540) (0.3521) (0.4549) Health and social work 0.3381*** -0.2193 0.2921 -0.2022** 0.4618** -0.2013**

(0.1092) (0.1628) (0.2248) (0.1023) (0.1910) (0.1011)

Entertainment 0.2510 -0.1033 -0.0239 -0.0911 0.3589 -0.3020

(0.2039) (0.1340) (0.0988) (0.2013) (0.2871) (0.2921)

Panel B: Before stock market crash

Agriculture 0.5317** -0.1780 -0.4448* -0.2851* 0.3824** -0.1848* (0.2001) (0.1262) (0.2511) (0.1488) (0.1669) (0.1014) Mining 0.0622 -0.2839 -0.1960 -0.0983 0.1476 -0.2267 (0.2570) (0.2741) (0.1198) (0.2793) (0.2699) (0.2237) Manufacturing 0.1516* -0.2161** -0.0744 -0.3250** 0.2971** -0.1617 (0.0806) (0.0856) (0.1533) (0.1594) (0.1325) (0.1116) Utilities 0.5020 -0.4304 -0.3118** -0.3348*** 0.3042 -0.3178 (0.4257) (0.2800) (0.1553) (0.1055) (0.4405) (0.2551) Construction 0.3625* -0.0453 -0.4058 -0.4180** 0.1908 -0.4322** (0.2161) (0.2125) (0.2916) (0.2122) (0.1955) (0.1909) Wholesale and retail 0.2461 -0.2016* -0.2486 -0.2051 0.5933** -0.5579**

(0.3319) (0.1071) (0.1892) (0.3146) (0.2689) (0.2749) Transportation 0.3696 -0.0246 -0.5143*** -0.2358** -0.4754 0.2480 (0.2965) (0.3118) (0.1946) (0.0946) (0.3074) (0.2686) Information technology 0.5147 -0.5801** 0.3803 -0.0506 0.2995** -0.1903 (0.5433) (0.2759) (0.3914) (0.1353) (0.1232) (0.1394) Finance 0.5931** 0.5866 -0.4923 0.1417 0.2971 0.6188 (0.2808) (0.5647) (0.5909) (0.5303) (0.2539) (0.5411)

(26)

21 Real estate -0.2229 -0.3453** -0.5600 0.3991*** 0.2818** -0.4755* (0.3413) (0.1721) (0.4976) (0.1448) (0.1091) (0.2623) Scientific research 0.4402 -0.0799 -0.1313 0.2361 0.3087 -0.2581 (0.4253) (0.4494) (0.2074) (0.4235) (0.3740) (0.4139) Resident service 0.0796 -0.2063 -0.4383*** 0.2682* -0.3652 -0.2227 (0.5786) (0.3369) (0.1202) (0.1379) (0.5465) (0.5222) Health and social work 0.4274*** -0.8742** 0.2341 -0.1388 0.4436 -0.3007

(0.1455) (0.3648) (0.1818) (0.1164) (0.3003) (0.2557)

Entertainment 0.3335 -0.2304 -0.0603 -0.1302 0.3202 -0.2882

(0.3255) (0.3618) (0.0972) (0.3403) (0.2878) (0.3183) Panel C: During stock market crash

Agriculture 0.4879** -0.1589 -0.4042* -0.3029** 0.3051 -0.2008** (0.2128) (0.1232) (0.2128) (0.1431) (0.2659) (0.0854) Mining 0.1238 -0.2126 -0.2349 -0.1143 0.2314 -0.4089 (0.1612) (0.2981) (01897) (0.1562) (0.2891) (0.3551) Manufacturing 0.2879*** -0.2326*** -0.1348 -0.3055** 0.4511** -0.2213*** (0.1010) (0.0769) (0.1398) (0.1467) (0.2020) (0.0762) Utilities 0.4871*** 0.1457 -0.3621*** -0.3962*** 0.3619 -0.2760 (0.1688) (0.2349) (0.1357) (0.1012) (0.3001) (0.2567) Construction 0.2459 0.1386 -0.4023 -0.2319** 0.2501** -0.3981** (0.2321) (0.2059) (0.4811) (0.1102) (0.1217) (0.1651) Wholesale and retail 0.4319*** -0.2650** -0.3879*** 0.1537** 0.5231** -0.5009**

(0.1562) (0.1128) (0.1231) (0.0742) (0.2217) (0.2500) Transportation 0.3439 0.1542 -0.4432** -0.1862** -0.4055 0.3019 (0.2870) (0.2186) (0.1898) (0.0834) (0.4012) (0.2921) Information technology 0.4878 -0.5238** 0.3421 -0.0621 0.3651** -02081*** (04062) (0.2121) (0.2897) (0.1089) (0.1715) (0.0751) Finance 05231** 0.5089** -0.3198*** 0.1543 0.3320 0.5222 (0.1810) (0.2232) (0.1029) (0.1945) (0.3001) (0.4898) Real estate -0.1452 -0.2569 -0.3987 0.4561*** 0.2671*** -0.4825** (0.2015) (0.1935) (0.3927) (0.1543) (0.1018) (0.2219) Scientific research 0.3210 0.2016 -0.1264 0.2438 0.3387 -0.2989 (0.2234) (0.2458) (0.1794) (0.3012) (0.3914) (0.4012) Resident service 0.1099 -0.1957** -0.4521** 0.3879** -0.2248* -0.2008* (0.2028) (0.0868) (0.2029) (0.1666) (0.1331) (0.1101) Health and social work 0.4089*** -0.2493 0.2988 -0.2322** 0.5128** -0.2518**

(0.1238) (0.1893) (0.364) (0.1120) (0.2531) (0.1034)

Entertainment 0.2810 -0.1654 -0.0299 -0.1124 0.3599 -0.3129

(0.2114) (0.1440) (0.0760) (0.1977) (0.2798) (0.2980) Panel D: After stock market crash

Agriculture 0.2162** -0.2079 -0.4040* -0.2982* 0.2922** -0.1872** (0.1002) (0.1642) (0.2378) (0.1620) (01431) (0.0814) Mining 0.1120 -0.2128 -0.1798 -0.0923 0.1650 -0.3411 (0.0913) (0.2103) (01692) (0.2103) (0.2349) (0.2934) Manufacturing 0.1793** -0.2029* -0.0923 -03109** 0.3081** -0.1862 (0.0883) (0.1101) (0.1224) (0.1543) (0.1492) (0.1932) Utilities 0.4044 0.1120 -0.3278** -0.3279 0.3036 -0.2429 (0.3569) (0.1994) (0.1541) (0.2091) (0.2898) (0.2410) Construction 0.1329 0.1086 -0.3001 -0.2224** 0.2326 -0.3329** (0.1983) (0.1815) (0.4590) (0.1019) (0.2094) (0.1650) Wholesale and retail 0.2495 -0.1701** -0.3329*** 0.1214 0.5042** -0.3029**

(27)

22 Transportation 0.2989 0.1238 -0.3824** -0.1413** -0.2879 0.2541 (0.2339) (0.2003) (0.1819) (0.0637) (0.4211) (0.2134) Information technology 0.4089 -0.4012 0.2230 -0.0821 0.3037** -0.1762** (0.4231) (0.3551) (0.2511) (0.0912) (0.1528) (0.0768) Finance 0.5081** 0.3911** -0.2513 0.1086 0.3142 0.5204 (0.2242) (0.1913) (0.3322) (0.1872) (02789) (0.4828) Real estate -0.0819 -0.2512 -0.3917 0.4245** 0.2838*** -0.3920 (0.1889) (0.1956) (0.3980) (0.2154) (0.1020) (0.2806) Scientific research 0.3373 0.1523 -0.0930 0.2432 0.3024 -0.2188 (0.2512) (0.2329) (0.1120) (0.3941) (0.3750) (0.3823) Resident service 0.1125 -0.1762 -0.3612 0.2981 -0.2079 -0.2029 (0.1872) (0.1988) (0.3418) (0.2103) (0.3121) (03860) Health and social work 0.3022 -0.2022 0.2741 -0.1932** 0.4517** -0.1760**

(0.2671) (0.1549) (0.2981) (0.0823) (0.2010) (0.0872)

Entertainment 0.2341 -0.1019 -0.0220 -0.1011 0.3062 -0.2820

(0.2093) (0.1219) (0.0493) (0.1013) (0.2871) (0.2212) ***, **, and * indicate statistical significance at 1%, 5% and 10% significance levels.

Robust standard errors in parentheses.

Table 7. Robustness check of model 1 for the full sample period

Model: 𝑅𝑛,𝑡= 𝛼0+ 𝛽1Δ𝑆𝑡+ 𝛼1𝑅𝑚𝑡+ 𝑒𝑛,𝑡

Industry

𝛽1

Original measure Measure 1 Measure 2

Agriculture -0.4573** -0.4260** -0.3980** (0.2258) (0.2088) (0.1822) Mining -0.0264 -0.0307 -0.0282 (0.1246) (0.1562) (0.0205) Manufacturing -0.1396** -0.1826*** -0.1523*** (0.0615) (0.0706) (0.0579) Utilities 0.3903 0.3055 0.2847* (0.2749) (0.2574) (0.1654) Construction 0.5232** 0.4593** 0.3987** (0.2602) (0.2061) (0.1869)

Wholesale and retail -0.1496 -0.2045 -0.1093

(0.1356) (0.1635) (0.2047) Transportation 0.3710*** 0.4079*** 0.3487** (0.0885) (0.1561) (0.1702) Information technology 0.2395 0.3105** 0.2760** (0.2592) (0.1538) (0.1329) Finance 0.6970*** 0.7094** 0.5683 (0.2565) (0.3048) (0.3821) Real estate -0.3888*** -0.5293** -0.4025 (0.1458) (0.2078) (0.2822) Scientific research 0.0678 0.0396 0.1092 (0.1825) (0.2036) (0.4093) Resident service -0.2407 0.0135 -0.1875 (0.2252) (0.4021) (0.2980)

Health and social work 0.3763*** 0.2961** 0.4169**

(0.1227) (0.1490) (0.2022)

Entertainment 0.2250 0.1581 0.1836

(0.1510) (0.2079) (0.2124)

***, **, and * indicate statistical significance at 1%, 5% and 10% significance levels. Robust standard errors in parentheses.

The effect of foreign exchange exposure on stock price : an analysis of the Chinese market