On the predictability of carry trade returns: The case of the Chinese Yuan

Research in International Business and Finance 39 (2017) 358–376

Contents lists available at ScienceDirect

Research in International Business and Finance j o u r n a l h o me p a g e : w w w . e l s e v i e r . c o m / l o c a t e / r i b a f

Full length article

On the predictability of carry trade returns: The case of the Chinese Yuan Calvin W.H. Cheong ∗ , Jothee Sinnakkannu, Sockalingam Ramasamy Department of Accounting, Banking & Finance, School of Business, Monash University Malaysia, Malaysia

a r t i c l e

i n f o

Article history: Received 5 January 2016 Received in revised form 13 August 2016 Accepted 14 September 2016 Available online 17 September 2016 JEL classification: G12 G15 F31 Keywords: Chinese Yuan Carry trade Uncovered interest parity FX volatility Fama-Macbeth

a b s t r a c t The Chinese Yuan or Renminbi has not been given due attention in the literature despite its importance in providing stability in global FX markets and attaining IMF reserve currency status. Unlike other currencies, the Yuan is subject to strict monetary controls by the People’s Bank of China. We explore factors that explain Yuan carry trade returns during the dollar-peg and managed float regimes by observing its response to the dollar risk factor, global FX volatility innovations and, liquidity. Using the traditional Fama and Macbeth (1973) two-pass ordinary least squares regression, we find that the effects of the dollar risk factor, global FX volatility innovations and liquidity on Yuan carry trade returns: (1) are unlike those on other currencies; (2) vary monotonically as the maturity of the carry trade position increases; and (3) are affected by the exchange rate regime in force. Our results suggest that short-term Yuan carry trade portfolios may serve as a hedge against market volatility. 1-year positions are far more resilient and deliver substantial returns especially during periods of low market volatility. Our results also exhibit the Yuan’s dependence on the dollar in its valuation besides the transfer of wealth between U.S. equity markets and dollar-denominated assets to the Yuan and Yuan-denominated assets. © 2016 Elsevier B.V. All rights reserved.

1. Introduction There is no shortage on studies that have examined the ability of forward exchange rates to be an unbiased predictor of future spot exchange rates. Often termed the “forward-premium puzzle”, foreign exchange (FX) market participants have long sought to exploit this ‘puzzle’ to earn excess returns with next-to-no risk at all; a strategy that has come to be known as carry trade. The carry trade strategy is designed to exploit deviations from uncovered interest parity (UIP). The theory posits that if UIP holds, the relative difference between the interest rates of two countries should be offset by a depreciation of equal magnitude in the currency with higher interest rates. This phenomenon of course, has been tested extensively across a wide range of currencies. Studies have shown (see Bilson, 1981; Fama, 1984; Hodrick, 1987; Engel, 1996) that generally, UIP does not hold, at least empirically. It has often been observed that currencies with higher interest rates appreciate instead of depreciate. Burnside et al. (2011) as a consequence, found carry trade to provide excess returns and Sharpe ratios that were double that of the US stock market. The literature in this regard, suggests that the sizeable carry trade returns do not come without a price; that the returns are the result of investors bearing commensurate risks. A number of studies have been conducted providing evidence to

∗ Correspondence author at: Monash University Malaysia, Jalan Lagoon Selatan, 47500 Bandar Sunway, Selangor D.E, Malaysia. E-mail addresses: [email protected], [email protected] (C.W.H. Cheong). http://dx.doi.org/10.1016/j.ribaf.2016.09.007 0275-5319/© 2016 Elsevier B.V. All rights reserved.

C.W.H. Cheong et al. / Research in International Business and Finance 39 (2017) 358–376

359

this effect. Burnside et al. (2011) argued that the sizeable carry trade returns are the result of what the authors termed, a peso problem i.e. a low probability of large negative payoffs. Brunnermeier et al. (2009) meanwhile believe that carry trade positions are exposed to crash risk, made worse by the sudden unwinding of positions whenever speculators are facing liquidity constraints. Burnside (2012) found that traditional risk factors used to price stock returns cannot explain the returns to carry trade while Christiansen et al. (2011) found that the level of FX volatility has an impact on carry trade return exposure towards stock and bond markets. Finally, Menkhoff et al. (2012) likewise found that carry trade investors are compensated for their exposure to global FX volatility risk through large carry trade payoffs. Motivated by the findings of the above mentioned studies, this paper examines the behaviour of carry trade returns in the specific context of the Chinese Yuan (CNY), or often synonymously, the Renminbi. The specific CNY focus is driven by a few reasons. First, instead of mere currency controls as has been implemented by various countries throughout history, China’s central bank – the People’s Bank of China (PBC) – have been accused by various parties but chiefly the U.S. of extensively manipulating the direction of the CNY on many occasions. Various authors have found that the CNY was undervalued by up to 30 percent against the USD; attributing this as the primary reason for China’s significant trade surplus (see Goldstein, 2004; Frankel, 2005; Funke and Rahn, 2005). Second, for the longest time, the CNY has only been convertible to the USD (Makin, 2011; Jaeger, 2010). It is only in the last decade that the PBC has allowed the CNY to be convertible to other safe-haven currencies such as the British Pound (GBP), the Euro (EUR) and the Japanese Yen (JPY). Third, China’s dominance in world trade resulted in the PBC’s rapid accumulation of substantial foreign currency reserves; reserves that were allegedly used to keep a tight hold on the CNY during the Global Financial Crisis between 2007 and 2009, effectively preventing the contagion from spreading across Asia (Sun and Zhang, 2009). Finally, the carry trade literature including those published in recent years, have left the CNY out of the equation despite its dominance in global FX markets. It would thus benefit academics and practitioners alike to have a better understanding of the CNY’s behaviour, especially since the International Monetary Fund’s (IMF) has said that it will elevate the CNY to reserve-currency status in October 2016 (Talley, 2015). In the empirical findings, we follow much of the recent literature by constructing portfolios of carry trade returns. The portfolios are constructed on the basis of either buying or selling the CNY forward against a unit of safe-haven currency (i.e. USD, GBP, EUR and JPY) on a 1-month, 3-month, 6-month and 1-year basis for a total of 4 portfolios. This is in contrast to Lustig and Verdelhan (2007), Lustig et al. (2011) and Menkhoff et al. (2012) whom sorted global currency portfolios according to their forward discount (i.e. interest rate). The reason behind the method of construction in this paper is so that the Yuan’s movement patterns throughout the sample period can be isolated and directly observed, rather than be confounded by the movements of a wide range of other currencies. This paper examines CNY behaviour through carry trade returns and its response to established FX risk factors mainly, the dollar risk factor (Lustig et al., 2011) and volatility innovations (Menkhoff et al., 2012). This paper also examines other possible predictors to the movements in CNY carry trade returns such as the excess returns to the value-weighted U.S. stock market, excess returns to the value weighted Chinese stock market, Carhart’s (1997) four factors, U.S. GDP growth, China GDP growth and U.S. industrial production growth. Brunnermeier et al., 2009 argue that liquidity is a key factor in a currency crash since currencies take massive hits whenever liquidity runs out. As Menkhoff et al. (2012) argue, liquidity may also be an important factor in understanding the cross-section of carry trade returns. Our analysis thus includes the Pástor and Stambaugh (2003) liquidity measure as another predictor for the movements of CNY carry trade returns. Using the traditional Fama and Macbeth (1973) (hereafter FMB) two-pass ordinary least squares (OLS) regression over a 15-year sample period from 2000 to 2014, our results show 1-month CNY carry trade returns to exhibit movement behaviour similar to those found by Lustig et al. (2011) and Menkhoff et al. (2012). That is to say, 1-month carry trade returns provided a form of hedge for investors during times of high market volatility. Having accounted for transaction costs, we also provide evidence that 1-year CNY carry trade positions were able to earn investors substantial excess returns regardless of market volatility conditions. This result corroborates the views of some commentators who believed that over the longer term, the PBC’s tight control over the CNY may have provided some form of stability in world asset markets following the recent financial crisis. We also find that liquidity can be attributed as one of the factors that contribute to the excess returns from CNY carry trade, irrespective of maturity. Finally, we find that the factors contributing to CNY carry trade returns behaviour is contrary to that of other currencies. We show that cross-sectional variation in CNY carry trade excess returns can be mostly explained by the dollar risk factor instead of global FX volatility innovations. In fact, our results show global FX volatility innovations ranks behind liquidity in explaining variations in CNY carry trade returns. From 1997, the CNY was pegged to the USD at a rate of 8.27 Yuan per USD until July 21, 2005 when the PBC lifted the peg causing a Yuan revaluation to about 8.11 per USD. Following the lifting of the peg, the PBC allowed the Yuan to appreciate against the dollar at a controlled pace before exercising tighter currency controls as the effects of the financial crisis started to be felt around the world (Robb, 2010). Although we excluded 2015 from our sample period, the market commentary on the CNY suggests that the PBC, despite its currency liberalization plans, still maintains a tight rein on the CNY as it held steady in spite of the huge Chinese stock market selloff in August. Consequently, this study also examines the impact of exchange rate regime changes on the CNY’s behaviour. Our results show that during the dollar-peg period of our sample (January 2000–July 2005), CNY carry trade returns were indeed compensation for global FX volatility, irrespective of maturity. But during the managed float period of our sample (August 2005–December 2014), we only find 1-month returns to exhibit such behaviour. In contrast to that which is commonly presumed under UIP, our results show that a stable currency does not in fact, make carry trade more lucrative.

360

C.W.H. Cheong et al. / Research in International Business and Finance 39 (2017) 358–376

The theoretical contributions of this paper are as follows. Generally, the Chinese Yuan is remarkably resilient against global FX volatility innovations and is unaffected by Chinese as well as global macroeconomic conditions. Instead, the CNY is sensitive towards USD volatility and U.S. stock market liquidity, a clear indication that despite the PBC’s currency liberalization policies, CNY movements are highly dependent on the USD as well as U.S. financial market conditions, contributing to a wealth transfer from U.S. markets and dollar assets towards CNY and Yuan-denominated assets in the form of hot money. We also extend conventional practice by observing not just 1-month but 3, 6 and 12-month periods. Our results show that the relationship between carry trade returns and FX factors begin to deteriorate as the time to maturity of the carry trade position increases, before moving in the opposite direction at the 1-year mark. This paper is also useful for practitioners. Our method of portfolio construction reduces the mathematical effort required of those seen in prior studies. A common method used in the literature is to build portfolios with a large number of currencies that are rebalanced monthly and sorted on the basis of their forward discounts. Our portfolios are much simpler; consisting of just 4 assets (i.e. USD, EUR, GBP, and JPY against the CNY). Despite its simplicity, our 1-month portfolio has similar hedging capabilities to the 40+ assets monthly-rebalanced portfolio. Also, our rolling 1-year CNY carry trade portfolio is able to deliver substantial excess returns to investors even during periods of supposedly high market volatility, even after taking transaction costs into consideration. In contrast, the portfolio construction method commonly used in the literature delivers positive excess returns only during periods of low to moderate market volatility but turns negative when volatility is high. However, it is not our objective to prove the superiority of our method. Rather, our methods and findings show that investors are able to replicate the hedging effects and/or generate substantial excess carry trade returns with reduced effort and transaction costs simply by including the CNY in the portfolio. The remainder of this paper is organized as follows. In Section 2, we provide a review of the recent literature on the predictability of carry trade returns; and a summary of China’s exchange rate regime and the Chinese Yuan’s behaviour throughout history. Section 3 presents the data and methods used in this study, and the descriptive statistics while Section 4 shows the main results on volatility risk for the full, pre- and post-depegging sample, and other potential explanatory factors. In Section 5 we discuss some implications of our study. Section 6 concludes. 2. Literature review 2.1. Explaining carry trade returns Uncovered Interest Parity (UIP) posits that currencies with higher interest rates should depreciate against those with lower interest rates in proportion to the difference in interest rates. However, the empirical evidence on the UIP relationship throughout history suggests otherwise. Hansen and Hodrick (1980) and Fama (1984) show that instead of depreciating as theorized by UIP, higher interest rates led to a further appreciation of currency. The authors also found that investors could earn significant excess returns simply by holding bonds in currencies with interest rates that were higher than usual. More recently, studies have also shown that investors are able to earn significant excess returns through holding bonds of currencies with interest rates that are higher than those of other currencies (Lustig and Verdelhan, 2005, 2007), implying that the UIP fails empirically in the cross-section. Since then, rather than test the validity of the UIP hypothesis on various currencies, studies have shifted their focus to explaining or predicting the returns earned from exploiting deviations from UIP i.e. carry trade. The fundamental premise of carry trade is to borrow a low-interest currency and invest in a high-interest currency. Carry trade may also be executed by selling the foreign currency forward when it is at a premium, and buying the foreign currency forward when it is at a discount. The returns generated from these two carry trade processes should theoretically speaking, be the same (Burnside et al., 2011). Studies have thus attempted to explain the cause of the sizeable excess returns generated through carry trade, with varying levels of success. The findings of Bansal and Dahlquist (2000) show that despite being pivotal to the UIP condition, interest rates are just one of the many possible factors that determine carry trade returns while Ranaldo and Soderlind (2010) argue that the safe-haven reputation of some currencies possessed resulted in lower risk premiums as compared to currencies that were riskier. Burnside et al. (2011) similarly argued that the high average payoff from carry trade was a form of compensation for the risk investors had to bear. But instead of the traditional risk factors such as consumption growth, stock market returns and the Fama and French (1989) factors, the authors found that the high carry trade payoff was the result of a ‘peso problem’ i.e. low-probability events that do not occur in the sample. Assuming a foreign currency is at a forward premium, the investor would sell the foreign currency forward. But given that interest parity does not hold, there is a remote chance that the foreign currency might appreciate substantially. Facing the risk of substantial losses to his position, the investor is thus compensated in the form of high carry trade payoffs. Lustig et al. (2011) argue that heterogeneity in exposure to common risk explains the carry trade returns since without exposure to common risk, the risk premium on carry trade should be zero. Simply shorting currencies with low interest rates and going long in currencies with high interest rates does not expose investors to any country- or currency-specific risks. The findings of Cenedese et al. (2014) likewise show that FX market variance provides useful information about whether subsequent losses will occur following a significant carry trade loss. The authors also find that carry trade strategies that are conditional on market variance outperform conventional strategies. Menkhoff et al. (2012) too find that the sizeable excess returns from carry trade are a form of compensation for the risk borne by investors. Since unexpectedly high volatility

C.W.H. Cheong et al. / Research in International Business and Finance 39 (2017) 358–376

361

worsens an investor’s risk-return tradeoff, it follows that there must be a negative volatility risk premium (Menkhoff et al., 2012). And since high unexpected volatility results in lower returns, assets that are positively covaried with market volatility should act as a hedge. In this regard, studies such as Ang et al. (2006a), Ang et al. (2006b), Adrian and Rosenberg (2008) and Da and Schaumburg (2009) have shown that aggregate volatility in stock markets has a significant impact on stock returns. Christiansen et al.(2011) likewise found aggregate volatility to have a significant impact on the correlation between excess returns of stock markets and currencies. In a related study, Dobrynskaya (2014) found the global downside market risk of currencies to provide fair compensation for carry trade returns. Liquidity has also been presented as an explanatory factor for carry trade returns. Bandi et al. (2008) find that besides volatility, liquidity too is an important pricing factor. Brunnermeier et al. (2009) argue that carry trade positions are highly exposed to currency crash risk since exchange rate movements between high- and low-interest currencies are negatively skewed. The negative skew is a consequence of sudden unwinding of carry trades; itself a result of falling risk appetites and funding liquidity drying up.

2.2. China’s monetary regime and the renminbi’s behaviour Central planning is prevalent in China’s financial sector despite opening its borders to international trade and investment in the late 1970s (DeRosa, 2005). Prior to 1994, the People’s Bank of China (PBC) established a dual exchange rate policy where an official and a market exchange rate coexisted. To create stability and promote international trade, the market exchange rate was set at 2.8 Yuan per USD. Meanwhile, the official exchange rate was used only for non-trade transactions. The PBC abolished the dual exchange rate policy in 1994, replacing it with a carefully managed exchange rate system that allowed the CNY to float at a marginal 0.1% against the USD i.e. between 8.27–8.28 Yuan per USD; a system that can only be described as a fixed exchange rate system (Guo, 2009). The system remained in place until July 21, 2005 when the PBC announced that the dollar peg would be abolished and currency reforms introduced where the value of the CNY would be linked to a basket of currencies that included the USD, Japanese Yen, the Euro and the South Korean Won, Australian Dollar, Canadian Dollar, Pound Sterling, Malaysian Ringgit, Russian Rouble, Singapore Dollar and the Thai Baht, at undisclosed weights. Following the announcement, the CNY was revalued to 8.11 per USD. The reforms announced in July 2005 was supposed to allow the CNY to appreciate monotonically in accordance with market forces but as Goldstein and Lardy (2006) rightly pointed out, little had changed to that effect as the exchange rate was only 8.07 as of December 2005. It was only in Q1 of 2008 did the Yuan breach the 7.00 mark, before the PBC held it steady at 6.82 all throughout the crisis period from April 2008 to June 2010. The Yuan gradually appreciated against the USD following the crisis, reaching a high of 6.08 in May 2015 before the PBC decided to devalue it in August 2015 amidst slowing economic growth and domestic stock market turmoil. The PBC has been widely criticized by various countries for giving China an unfair advantage in international trade through their currency manipulation (Zhang, 2010; Bhide and Phelps, 2007; Nair and Sinnakkannu, 2011; Makin, 2011). At one point, it was alleged that the PBC had spent close to USD 20 billion a month for currency intervention. But despite these criticisms, the PBC and its strict currency policies has been credited for preventing the effects of financial crises from spreading further on more than one occasion, beginning with the Asian Financial Crisis of 1997. As the currencies of Malaysia, Indonesia, Thailand and South Korea fell drastically amidst strong FX speculation, it was feared that China may choose to follow suit (Sinnakkannu and Nassir, 2008; Das, 2006). However, the PBC held the CNY steady, providing the much needed stability in Asia’s FX markets besides playing a key role in bolstering international trade in the region. Also, through its alleged currency manipulation and burgeoning trade balance, China has been able to amass substantial foreign currency reserves. In fact, it can rightly be said that China had a key role in stabilizing the U.S. economy during the most recent financial crisis. The U.S. Treasury Department reported that China had in 2008 overtook Japan as the largest foreign holder of U.S. securities, accounting for more than a third of total new foreign purchases of these securities (Morrison, 2009) thereby injecting the liquidity the U.S. markets desperately needed. In light of the financial crisis, the U.S. Federal Reserve slashed interest rates to prevent a liquidity shortage and to avoid a meltdown in the financial markets. China meanwhile initiated a de-facto peg to stabilize international trade and industrial production whilst maintaining interest rates that were relatively higher than the U.S. The gradual appreciation of the CNY since its depegging in 2005, and the historically low U.S. interest rates generated a one-way bet in the foreign exchange markets, attracting a substantial amount of hot money inflows into China (McKinnon and Schnabl, 2009). The economic conditions between China and the U.S. created a double incentive for investors to borrow at almost zero cost in the U.S. and investing those funds in China to earn a hypothetical 2 percent return, assuming no transaction costs with the possibility of earning much more assuming the CNY continued on its path of appreciation; leading to a resurgence of the carry trade phenomenon (Schnabl, 2010).

3. Methodology 3.1. Carry trade The literature highlights a variety of strategies that exploits the failure of the uncovered interest parity (UIP). Here, we focus on carry trade, one of the most widely used strategies by practitioners (Galati and Melvin, 2004). Fundamentally, carry

362

C.W.H. Cheong et al. / Research in International Business and Finance 39 (2017) 358–376

trade consists of borrowing a low-interest-rate currency and lending a high-interest-rate currency. Assuming no transaction costs, the payoff to the carry trade, denominated in dollars, is



yt (1 + ri )

St+1 − (1 + rt ) St



(1)

where St is the spot exchange rate expressed as US dollars per Chinese Yuan. The variables rt and ri represent the U.S. and Chinese interest rate, respectively. For ease of computation, the amount of dollars bet on this strategy is normalized to one. The amount of dollars borrowed, yt is determined as:



yt =

+1ifrt < ri ,

(2)

−1ifri ≤ ri .

Consider the case in which St is a martingale: E t St+1 = S t

(3)

where Et is a time-t conditional expectations operator. This martingale is property is not an implication of market efficiency but is a reasonable description of the data (Burnside et al., 2011). Eq. (3) implies that the expected payoff for the carry trade is positive and equal to the difference between the higher and the lower interest rate: yt (ri −rt ) > 0. Alternatively, carry trade may also be executed by selling the Chinese Yuan forward when it is at a forward premium (Ft ≥ St ) and buying the Chinese Yuan forward when it is at a forward discount (Ft < St ). The value of wt , the number of Chinese Yuan sold forward is given by



wt =

+1/Ft ifFt ≥ St ,

(4)

−1/Ft ifFt < St .

The value of wt is equivalent to buying or selling one dollar forward. The dollar-denominated payoff to this strategy at t + 1, denoted xt+1 , is xt+1 = wt (F t −S t+1 ).

(5)

Covered interest parity implies that: (1 + rt ) =

Ft (1 + ri ) . St

(6)

When Eq. (6) holds, the payoff of Eqs. (4) and (2) should be equal (Burnside et al., 2011). In this study, we focus on investments in the forward and spot currency markets as it offers two distinct advantages (Lustig et al., 2011). First, carry trade is easily implemented in these markets; their data complete and readily available as opposed to foreign fixed income markets. Second, forward contracts have minimal default and counter-party risk. As such, we adopt the strategy defined in Eq. (4). 3.2. Transaction costs We follow Lustig et al.’s (2011) method of incorporating transaction costs by using bid-ask quotes for the spot and forward exchange rates. The net currency excess returns for an investor who goes long in Chinese Yuan is l a xt+1 = ftb − st+1

(7)

that is to say, the investor buys the Chinese Yuan forward or equivalently sells the dollar forward at the bid price (f b ) in a ) in the spot market in period t + 1. period t, and sells the Chinese Yuan or equivalently buys the dollar at the ask price (st+1 Similarly, for an investor who is long in the dollar i.e. short the Chinese Yuan, the net currency excess return is given by s b xt+1 = −fta + st+1

. (8)

3.3. Global FX volatility & dollar risk factor Following Menkhoff et al. (2012), global FX volatility is the absolute daily log return |rdk | (= |sd |) for each currency k on each day d in our sample. We then average over all currencies in the sample on any given day and average the daily values up to the monthly frequency. The measure of global FX volatility in month t is thus defined as:

⎡

tFX

1  ⎣ = Tt d ∈ Tt

k ∈ Kd



|rdk | Kd



⎤ ⎦

(9)

C.W.H. Cheong et al. / Research in International Business and Finance 39 (2017) 358–376

363

where Kd represents the number of currencies in the sample on day d while Tt is the total number of trading days in month t. Despite its similarities to realized volatility, we use absolute returns and not squared returns to minimize the impact of outlier returns.1 In the empirical analysis, we observe volatility innovations (tFX ) as a nontraded risk factor. Ang et al. (2006a,b), takes the first difference of the volatility series but, as is commonly found in FX-based data, autocorrelation may exist leading to spurious results. For computational ease, we measure volatility innovations as a simple change in the volatility level.2 Following its application in Lustig et al. (2011) and Menkhoff et al. (2012), we define the dollar risk factor (DOL) as the average excess returns of all currencies in our global sample, given as: rxt+1 =

1 i rxt+1 N

(10)

i

where N is the number of currencies in our global sample. 3.4. Coskewness We include the measure of coskewness introduced by Harvey and Siddique (1999, 2000) and Ang et al. (2006a,b), and applied in Menkhoff et al. (2012) and Dobrynskaya (2014) in order to generate additional inferences towards the relationship between CNY carry trade returns

and market volatility. Coskewness is calculated by: coskew =

E (rk −k )(rm −m )2 (rk ) 2 (rm )

, (11)where rk and rm are respectively, the returns of portfolio k and the market benchmark

while  and  are respectively the mean and standard deviation. Portfolios with high coskewness values are normally assumed to deliver high returns during periods of high market volatility and vice versa. Implicitly, investors of high coskewness portfolios are compensated for the additional risk they bear, which in effect, means such portfolios act as a hedge against volatility and as such, should earn lower returns. 3.5. Data Our dataset on spot and forward exchange rates obtained from Bloomberg covers the major currencies i.e. U.S. Dollar (USD), Euro (EUR), British Pound (GBP) and Japanese Yen (JPY); quoted as units of CNY per one unit of major currency. The currencies of 37 countries included to calculate the global FX volatility innovations (VI) and dollar risk factor (DOL) are: Australia, Austria, Belgium, Brazil, Bulgaria, Canada, Croatia, Czech Republic, Denmark, Egypt, Finland, Hong Kong, Hungary, Iceland, Kuwait, Malaysia, Mexico, Netherlands, New Zealand, India, Ireland, Israel, Russia, Slovenia, Switzerland, Saudi Arabia, South Africa, Taiwan, Singapore, South Korea, Thailand, Slovakia, Sweden, Ukraine, Norway, the Philippines, and Poland; quoted as units of foreign currency per one unit of USD.3 The data consists of daily observations for bid and ask spot exchange rates and, one-month, three-month, six-month, and one-year forward exchange rates. The data spans the period January 2000–December 2014. 3.6. Portfolio construction We construct the carry trade portfolios from the perspective of an investor who simultaneously buys or sells the Chinese Yuan forward against the four major currencies. Four portfolios are constructed in total, representing the maturity of each forward CNY transaction studied i.e. 1 month, 3 months, 6 months, and 1 year. Each portfolio comprises four long/short CNY forward positions. Portfolio 1 consists of 1-month long/short CNY forward positions against the USD, EUR, GBP and JPY. Portfolios 2–4 represent the 3- and 6-month and 1-year long/short CNY forward positions. We assume equal weights of the four currencies in each portfolio i.e. the USD, EUR, GBP and JPY trades in all portfolios each represent 25 percent of j the portfolio. We compute the excess return xt+1 for portfolio j by taking the average of the currency excess returns in each portfolio j. As opposed to similar studies on carry trade, we do not rebalance the portfolios in order to observe how a rolling CNY carry trade strategy behaves over the sample period. 3.7. Descriptive statistics for portfolios Descriptive statistics for the four carry trade portfolios are presented in Table 1. The top panel displays results for the full sample period, January 2000–December 2014. The middle panel shows the descriptive statistics for the four portfolios

1

See Andersen et al. (2001) and Menkhoff et al. (2012) for a more detailed discussion on this matter. Menkhoff et al. (2012) used the residuals from a simple AR(1) estimation of volatility levels as the proxy for volatility innovations. Their results remained robust when a simple change in volatility levels was used as the proxy for volatility innovations. 3 The Austrian Schilling, Belgian Franc, Finnish Markka and Dutch Guilder were in circulation until 2002; the Slovenian Tolar was in circulation until 2007; while the Slovak Koruna was in circulation until 2009; before being replaced by the Euro. We dropped these currencies and recalculated Equation (9) and (10) when the countries officially adopted the Euro. The removal of these currencies did not alter the global FX volatility and DOL values significantly. 2

364

C.W.H. Cheong et al. / Research in International Business and Finance 39 (2017) 358–376

Table 1 Descriptive Statistics. 2000–2014: Full Sample Portfolio

1

2

3

4

Mean Median Standard Deviation Kurtosis Skewness Sharpe Ratio AC(1) Coskew (DOL) Coskew (USMKT) Coskew (CMKT)

0.0175 0.0025 0.0791 4.0471 0.8624 0.2218 0.0344 0.6968 0.5858 0.0871

0.0280 0.0078 0.1494 2.8911 0.2280 0.1876 0.0521 0.6449 0.5839 0.0314

0.0293 0.0154 0.1569 6.5232 0.4385 0.1871 0.0876 0.1581 −0.0416 −0.1028

0.0657 0.0326 0.3377 2.9049 0.6506 0.1944 0.0381 −0.0821 −0.0390 0.0975

Jan 2000–July 2005: Dollar Peg Period Mean Median Standard Deviation Kurtosis Skewness Sharpe Ratio AC(1) Coskew (DOL) Coskew (USMKT) Coskew (CMKT)

0.0075 0.0013 0.0488 7.3950 1.7377 0.1541 0.1213 −0.0477 −0.1225 −0.0877

0.0197 0.0057 0.0778 10.5150 1.7246 0.2535 0.1489 −0.0178 −0.0650 0.0110

0.0186 0.0146 0.0175 1.2678 1.2516 1.0624 0.1937 −0.0177 −0.0396 −0.0591

0.0668 0.0351 0.1794 2.6589 0.8502 0.3725 0.1445 0.0371 −0.0028 −0.0239

0.0014 0.0011 0.0035 0.8085 0.0082 0.1843 0.0885 1.0452 0.9719 0.0449

0.0356 0.0184 0.1968 3.0691 0.2573 0.1809 0.0935 0.2654 −0.0427 −0.1292

0.0653 0.0003 0.4027 1.5193 0.5677 0.1615 0.0968 −0.1600 −0.0630 0.1742

August 2005–December 2014: Managed Float Period 0.0237 Mean 0.0193 Median Standard Deviation 0.0917 Kurtosis 2.5909 Skewness 0.5885 0.2546 Sharpe Ratio AC(1) 0.0733 Coskew (DOL) 1.1465 1.0084 Coskew (USMKT) Coskew (CMKT) 0.1857

The table reports mean and median returns, annualized standard deviations, kurtosis and skewness of the four portfolios constructed on the basis of CNY forward maturities. We also report annualized Sharpe ratios, and AC(1), the first-order autocorrelation coefficient. Coskew(.) is the Harvey and Siddique (2000) measure of coskewness with respect to the dollar risk factor (DOL), the excess returns of the U.S. stock market (USMKT), or the Chinese stock market (CMKT). Portfolio 1–4 consists of 1-month, 3-month, 6-month and 1-year long/short CNY forward positions respectively against the four major currencies i.e. USD, EUR, GBP and JPY. Positions in each portfolio are equally weighted. All returns have been adjusted for transaction costs as explained in Section 2.2..

during the dollar peg period between January 2000 and July 2005 while the bottom panel presents the descriptive statistics of the portfolios during the post-depegging or managed float period between August 2005 and December 2014. Average returns seem to rise monotonically from Portfolio 1–4 while skewness remains relatively unchanged; a pattern that can be observed in all sample periods. Kurtosis likewise does not display any discernible pattern. However, standard deviation seems to rise in a monotonic pattern from Portfolio 1–4. Also, there does not seem to be any evidence of return autocorrelation across all portfolios. However, there is a hint of return autocorrelation during the dollar peg period; unsurprising seeing that CNY exchange rates then remained relatively constant. Finally, we look at the coskewness of each carry trade portfolio against the dollar risk factor (DOL), the U.S. stock market returns (USMKT) and China stock market returns (CMKT).4 Generally, we can see that Portfolio 1 and 2 displays relatively high coskewness throughout the sample period. That is to say, these two portfolios provide high returns during times of high market volatility; which implies that the 1-month and 3-month carry trade portfolios may serve as a hedge against volatility. In contrast, we can see that Portfolio 3 and 4 displays relatively low and sometimes, negative coskewness throughout the sample period i.e. returns are higher when the markets are less volatile. Fig. 1 shows cumulative excess returns for each of the carry trade portfolios. As can be seen, 1-year CNY carry trades are the most profitable. Although carry trade returns of other maturities have generated positive returns, their performance can be described as erratic at best. We can also see that 1-year carry trade returns throughout the sample period are higher when volatility innovations are low (i.e. small change to market volatility), consistent with our earlier inference of its coskewness. Similarly, Fig. 1 also suggests that short-term carry trade positions (1-month, 3-month) deliver lower returns during periods of high volatility, supporting the notion that portfolios with high coskewness serve as a hedge against market volatility. Meanwhile, 6-month carry trade returns rose monotonically until 2009 before moving tandem with the rise and

4

We use the New York Stock Exchange and Shanghai Composite Index’s daily returns data as a proxy of U.S. and China stock market returns.

C.W.H. Cheong et al. / Research in International Business and Finance 39 (2017) 358–376

365

Cumulave Carry Trade Returns

Cumulave excess returns (in % p.a.)

60 50 40 Porolio 4

30 20 Porolio 2

10

Porolio 1

Porolio 3

0 2000

2001

2003

2004

2006

2007

2009

2010

2012

2013

-10

Risk Factors 0.014

2 1.5 Dollar Risk Factor

Volality Innovaons

0.01

1

0.008 0.5 0.006 0

0.004 0.002 2000 0

Volaly Innovaons

Dollar Risk Factor

0.012

-0.5 2001

2003

2004

2006

2007

2009

2010

2012

2013 -1

Fig. 1. Returns to carry trade portfolios. The upper panel of this figure shows cumulative excess returns of each of the four carry trade portfolios, labeled accordingly. The lower panel shows a time series plot of volatility innovations (red line) and the dollar risk factor (light blue line). The sample period is January 2000–December 2014. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

fall of market volatility. Generally, we can see that CNY carry trade returns are not sensitive towards global macroeconomic conditions. In fact, the carry trade returns across all portfolios do not seem to display any significant reactions towards the removal of the dollar peg in mid-2005. Likewise, Portfolios 1–3 do not seem to show any sensitivity towards the most recent financial crisis. Portfolio 4 returns in contrast seems to have picked up as a result of the financial crisis, which may suggest that the stability of the CNY and the Chinese financial markets relative to the USD and the U.S. may have made the CNY and Yuan-denominated assets more attractive to investors, especially those from the U.S. 4. Empirical results 4.1. Graphical representation We provide a rudimentary visual analysis of the relationship between CNY carry trade returns of differing maturities against global FX volatility innovations in Fig. 2. We divide our sample of carry trade returns and volatility innovations into quartiles corresponding to the magnitude of the volatility innovations. The first subsample comprises of the 25% periods that have the lowest volatility innovation (“low”) while the last subsample comprises of the 25% periods that have the highest volatility innovation (“high”). Following that, we calculate average excess returns for Portfolios 1–4 for each of the four volatility innovation subsamples. The top left panel of Fig. 2 shows the results for the sample period of 2000–2014, the top right panel is for the dollar peg period (January 2000–July 2005) and the lower panel is for the managed float period (August 2005–December 2014). The bars represent the mean annualized returns of Portfolios 1–4. From Fig. 2, it is apparent that carry trade returns rise in tandem with rising volatility innovations, irrespective of maturity. However, when volatility innovations are at its peak (“high”), carry trade returns for all maturities fall. 1-year carry trade returns are hit hardest, with returns falling by close to 5%. Returns of the other portfolios are more resilient, falling by only about 2%. Fig. 2 provides some preliminary evidence that CNY carry trade performs poorly during periods of high volatility. Fig. 2 also suggests that CNY carry trade positions with maturities of less than a year significantly underperform the 1-year position, but deliver returns that are more resilient during periods of high volatility. The following section tests this relationship empirically.

366

C.W.H. Cheong et al. / Research in International Business and Finance 39 (2017) 358–376

Fig. 2. Excess returns and volatility. The figure shows average excess returns for the four carry trade portfolios against the lowest to highest quartile of global FX volatility innovations (labeled “low” to “high”) on the x-axis of each panel. The bars represent the average excess returns for each carry trade position. The panels are, clockwise from top left: the full sample period (2000–2014), the dollar peg period (January 2000–July 2005) and the managed float period (August 2005–December 2014).

4.2. Fama-Macbeth regression – dollar risk factor & global FX volatility In the following tables, we report our empirical results from using the traditional Fama and Macbeth (1973) (hereafter FMB), two-pass ordinary least squares (OLS) method. Our two-pass procedure is the conventional one detailed in Chapter 12 of Cochrane (2005). The first step is estimating a time-series regression to obtain in-sample betas for portfolios 1–4. In the second step, these portfolio betas are used in the cross-sectional average excess returns on the estimated time-series betas. We do not include a constant in the second stage of the FMB regression as we obtain similar results when the DOL factor is replaced with a constant in the second stage of the FMB regression. As Menkhoff et al. (2012) suggests, the DOL factor serves the same purpose as a constant that allows for common mispricing. Additionally, because the traditional two-pass approach (such as in FMB) are prone to the “errors-in-variables” problem (Shanken, 1992) which may lead to beta and risk factor estimation errors, we compute Shanken (1992)-corrected standard errors as well as Heteroscedasticity and Autocorrelation Consistent (HAC) Newey and West (1987) standard errors computed with optimal lag selections (Andrews, 1991).5 Table 2 below presents the estimates of the FMB regression using the four CNY carry trade portfolios as the test assets. The factors included in the regression are the dollar risk factor (DOL) and global FX volatility innovations (VI). From Table 2, we can see that among the four CNY carry trade portfolios for the full sample period, only Portfolio 1 displays a negative factor price estimate for VI, implying that it delivers lower risk premia when its returns co-move positively with volatility innovations. Meanwhile, DOL returns a positive factor price estimate implying a CNY carry trade portfolio’s sensitivity towards the USD. The negative (positive) factor price estimate for VI (DOL) seems to hold even when the sample period is split into the dollar-peg and managed-float sample periods. The estimates suggest that short-term CNY carry trade positions (i.e. 1-month) react in a manner similar to carry trade positions using a portfolio of currencies. The estimates also suggest that the strict monetary control imposed by the PBC has had no effect on the behaviour of CNY carry trade positions, at least in the short-term. Our findings are consistent with those of Menkhoff et al. (2012) who, using monthly observations of carry trade portfolios, found VI (DOL) to have negative (positive) factor price estimate. However, when we extend the maturity of the carry trade position, the estimates begin to differ. The 3-month CNY carry trade portfolio (P2) has positive factor price estimates for both the DOL and VI factors, for the full and managed float sample periods. Both the 6-month (P3) and 1-year (P4) CNY carry trade portfolios meanwhile have negative (positive) factor price

5 Two common forms of residuals dependence exist: time-series (i.e. firm effect) (Wooldridge, 2007), and cross-sectional (time effect) (Petersen, 2009); both of which may occur when employing the FMB approach. Although the FMB approach is designed to overcome the time effect, FMB standard error estimates are biased downwards in the presence of firm (or at least in this study, currency) effects (Petersen, 2009). To address both effects, we included currency fixed effects and then estimated standard errors clustered by time. The standard error estimates were similar to our computed Shanken (1992)corrected standard errors as well the HAC Newey and West (1987) standard errors (see Tables A1 and A2 in Appendix).

C.W.H. Cheong et al. / Research in International Business and Finance 39 (2017) 358–376

367

Table 2 Cross-Sectional Asset Pricing: Factor Prices. Factor Prices FMB

DOL

VI

2SH

2NW

DOL

VI

2SH

2NW

DOL

VI

2SH

2NW

P1

␭ (SH) (NW)

8.66 (0.26) (0.68)

−0.01 (0.02) (0.01)

0.57 (0.49)

1.60 (0.21)

2.68 (0.19) (0.40)

−0.01 (0.01) (0.01)

0.57 (0.49)

0.44 (0.51)

10.26 (0.32) (0.08)

−0.01 (0.02) (0.01)

0.55 (0.48)

1.78 (0.18)

P2

␭ (SH) (NW)

7.40 (0.32) (0.19)

0.01 (0.02) (0.03)

0.40 (0.57)

0.14 (0.70)

1.97 (0.18) (0.10)

−0.01 (0.01) (0.02)

0.60 (0.38)

0.04 (0.85)

8.76 (0.39) (0.16)

0.02 (0.03) (0.03)

0.41 (0.56)

0.30 (0.58)

P3

␭ (SH) (NW)

−19.33 (0.45) (0.35)

0.04 (0.04) (0.05)

0.82 (0.29)

0.30 (0.58)

5.98 (0.11) (0.35)

−0.01 (0.01) (0.00)

0.80 (0.29)

2.80 (0.09)

−21.86 (0.48) (0.24)

0.05 (0.04) (0.04)

0.89 (0.25)

0.77 (0.38)

P4

␭ (SH) (NW)

−31.58 (0.52) (0.39)

0.01 (0.04) (0.06)

0.61 (0.38)

0.64 (0.42)

75.94 (0.63) (0.54)

−0.11 (0.20) (0.09)

0.57 (0.49)

1.91 (0.17)

−45.33 (0.15) (0.37)

0.02 (0.03) (0.08)

0.70 (0.33)

1.50 (0.22)

1.56

1.28

1.45

1.39

1.82

1.66

Bai-Perron

The table reports cross-sectional pricing results for the linear factor model based on the dollar risk factor (DOL) and volatility innovations (VI). The test assets are excess returns to CNY carry trade positions against the USD, EUR, GBP and JPY for 1-month (P1), 3-month (P2), 6-month (P3) and 1-year (P4) maturities for the full sample period (left panel), dollar peg period (center panel), and managed float period (right panel). The panels display coefficient estimates of factor risk prices ␭, obtained by Fama and Macbeth (1973) (FMB) cross-sectional regressions. We report standard errors and 2 test statistics (p-values in parentheses) that are Shanken (1992) adjusted (SH) or Newey and West (1987) approach with optimal lag selection (NW). Bai and Perron (2003) structural break test F-statistics are reported in the last row. Table 3 Cross-sectional Asset Pricing: Factor Betas. Factor Betas Portfolio

Full Sample Period

Dollar Peg Period

Managed Float Period

␣

DOL

VI

R

␣

DOL

VI

R

␣

DOL

VI

R2

1

0.001 (0.003)

0.003 (0.001)

1.443 (0.217)

0.795

−0.001 (0.002)

0.006 (0.004)

0.531 (0.672)

0.794

0.001 (0.001)

0.003 (0.001)

1.202 (0.001)

0.750

2

0.009 (0.013)

0.002 (0.001)

0.644 (0.316)

0.732

−0.001 (0.002)

0.001 (0.001)

−0.451 (0.084)

0.899

0.008 (0.010)

0.002 (0.001)

0.583 (0.152)

0.791

3

0.027 (0.028)

0.001 (0.001)

0.288 (0.764)

0.731

0.001 (0.001)

0.004 (0.001)

−1.014 (0.001)

0.830

0.023 (0.023)

0.001 (0.001)

0.112 (0.368)

0.662

4

0.029 (0.013)

−0.002 (0.001)

−1.377 (0.316)

0.831

−0.023 (0.044)

−0.206 (0.110)

−14.31 (0.356)

0.745

0.023 (0.022)

−0.001 (0.001)

−0.292 (0.497)

0.840

Bai-Perron

1.18

1.58

1.34

1.24

1.41

1.62

1.33

1.26

1.75

2

2

The table reports cross-sectional pricing results for the linear factor model based on the dollar risk factor (DOL) and volatility innovations (VI). The test assets are excess returns to CNY carry trade positions against the USD, EUR, GBP and JPY for 1-month (P1), 3-month (P2), 6-month (P3) and 1-year (P4) maturities for the full sample period (left panel), dollar peg period (center panel), and managed float period (right panel). The panels report results for time-series regressions of excess returns on a constant (␣), the dollar risk factor (DOL) and volatility innovations (VI). Newey-West HAC standard errors (with optimal lag selection) are reported in parentheses. The full sample period is January 2000–December 2014; the dollar peg period is January 2000–July 2005 while the managed float period is August 2005–December 2014 and we use weekly transaction cost adjusted returns. Bai and Perron (2003) structural break test F-statistics are reported in the last row.

estimates for the DOL (VI) factors for the full and managed float sample periods. But as can be seen in Table 2, P2, P3, and P4 have negative (positive) factor estimates for VI (DOL). P2 factor estimates may imply that 3-month CNY carry trade delivers returns that do not compensate the investor for neither global FX volatility nor dollar risk factors. 6-month and 1-year CNY carry trade returns on the other hand seem to be more sensitive towards USD volatility. The negative DOL factor price for P3 and P4 suggests that the returns serve as a hedge against USD-specific rather than global FX volatility, a relationship that persists even when the dollar peg was removed. During the dollar-peg period from January 2000–July 2005, we can see that the signs for the factor price estimates are consistent across the four portfolios; that their returns can be attributed to changes in global FX volatility. A final observation that can be made from Table 2 is that, regardless of maturity, CNY carry trade positions seem to be more sensitive towards USD-specific risk rather than global FX volatility judging from the magnitude of the estimated factor prices. Table 3 below shows time-series beta estimates for the four CNY carry trade portfolios for the full, dollar-peg, and managed float sample periods. For the full sample period, we can see that estimates for ˇDOL and ˇVI indicate that 1-month CNY carry trade returns co-move positively with the dollar risk factor and global FX volatility innovations. As the time to maturity of the carry trade position increases, we find ˇDOL and ˇVI estimates to decline in a stark monotonic manner; to the point where

368

C.W.H. Cheong et al. / Research in International Business and Finance 39 (2017) 358–376

Table 4 Cross-Sectional Asset Pricing: Liquidity. Panel A: Factor Prices FMB

Panel B: Factor Betas IL

2SH

2NW

Portfolio

␣

IL

R2

P1

␭ (SH) (NW)

0.08 (0.08) (0.08)

0.88 (0.29)

1.37 (0.12)

1

−0.002 (0.04)

0.702 (0.04)

0.866

P2

␭ (SH) (NW)

0.12 (0.22) (0.25)

0.63 (0.36)

0.64 (0.42)

2

0.001 (0.06)

0.293 (0.08)

0.361

P3

␭ (SH) (NW)

(0.05) (0.37) (0.15)

0.56 (0.49)

0.27 (0.89)

3

−0.027 (0.02)

−0.259 (0.06)

0.433

P4

␭ (SH) NW

0.57 (0.49) (0.27)

0.94 (0.22)

0.56 (0.51)

4

−0.002 (0.09)

−0.801 (0.04)

0.532

Bai-Perron

1.36

1.51

Bai-Perron

1.43

The table presents cross-sectional pricing results for the linear factor model based on the dollar risk factor and volatility innovations (not reported), and Pástor and Stambaugh (2003) liquidity measure i.e. innovations in aggregate liquidity (IL) obtained from Wharton Research Data Services. The test assets are excess returns to CNY carry trade positions against the USD, EUR, GBP and JPY for 1-month (P1), 3-month (P2), 6-month (P3) and 1-year (P4) maturities for the full sample period. Panel A displays coefficient estimates of factor risk prices ␭, obtained by Fama and Macbeth (1973) (FMB) cross-sectional regressions. We report standard errors and 2 test statistics (p-values in parentheses) that are Shanken (1992) adjusted (SH) or Newey-West approach with optimal lag selection (NW). Panel B presents results for time-series regressions of excess returns on a constant (␣), the dollar risk factor and volatility innovations (not reported), and Pástor and Stambaugh (2003) liquidity measure i.e. innovations in aggregate liquidity (IL). Newey-West HAC standard errors (with optimal lag selection) are reported in parentheses. The full sample period is January 2000–December 2014 and we use weekly transaction cost adjusted returns. Bai and Perron (2003) structural break test F-statistics are reported in the last row.

1-year carry trade returns co-move negatively with both the dollar risk factor and global FX volatility. We also observe a similar decline in DOL and VI beta estimates for the dollar-peg and managed float sample periods. Particularly striking is the large negative ˇVI for the 1-year carry trade returns during the dollar-peg period. Our estimates for Portfolio 1 are consistent with the findings of Menkhoff et al. (2012); that 1-month carry trade returns provided a form of hedge during turbulent times; an effect that can be observed in our estimates for Portfolio 1. In contrast, our beta estimates for Portfolio 4 suggests that 1-year CNY carry trade returns are counter-cyclical to USD and global FX volatility. Due to the 1-year carry trade’s ability to deliver excess returns resilience despite variations to volatility (see Figs. 1 and 2), investors are able to earn substantial returns by continuously maintaining a 1-year CNY carry trade position. 4.3. Fama-Macbeth regression – liquidity We also examine Yuan carry trade return behaviour against a range of other factors besides the dollar risk factor and global FX volatility innovations. The other factors include the excess returns to the value-weighted U.S. stock market (Burnside et al., 2011), excess returns to the value weighted Chinese stock market, Carhart’s (1997) four factors (excess return to the value-weighted U.S. stock market, the size premium (SMB), the value premium (HML) and momentum (UMD)), U.S. GDP growth, China GDP growth and U.S. industrial production growth. As we mentioned earlier, liquidity also has an important role in the understanding of FX risk premia (Brunnermeier et al., 2009; Menkhoff et al., 2012). We use Pástor and Stambaugh’s (2003) innovations in aggregate liquidity to explore this relationship in the context of CNY carry trade. Data for the liquidity measure were obtained from Wharton Research Data Services (WRDS). For brevity purposes, we only report results for the liquidity measure since among all the possible factors, only liquidity exhibited robust findings.6 To examine the relationship between liquidity and CNY carry trade returns, we run the same FMB two-pass OLS regression as above, substituting the dollar risk factor and global FX volatility innovations with the Pástor and Stambaugh (2003) innovation in aggregate liquidity. The fundamental understanding behind the Pástor and Stambaugh (2003) liquidity measure is that asset price movements caused by changes to liquidity have to be reversed eventually such that stronger price reversals reflect lower liquidity. The measure is constructed in a way where a larger value reflects higher liquidity. Following the literature, we would expect the factor loading and price of liquidity to be positive. Panel A of Table 4 reports the factor loadings and prices for the Pástor and Stambaugh (2003) liquidity measure. From Panel A, we find the factor loadings and prices for liquidity to be positive for all CNY carry trade portfolios (except Portfolio 3), consistent with the findings from the literature. Comparing our estimates for factor price and loadings in Panel A of Table 4 to those in Table 2, we can see that the magnitude of the liquidity factor is greater than that of global FX volatility

6 Apart from the Pástor and Stambaugh (2003) liquidity measure, the other factors we explored: (1) did not exhibit statistically significant relationships; (2) had high standard errors even after correction; (3) low R2 values or; (4) all of the above.

C.W.H. Cheong et al. / Research in International Business and Finance 39 (2017) 358–376

369

Table 5 Cross-sectional Asset Pricing: Central Bank Intervention. Panel A: Factor Prices FMB

Panel B: Factor Betas

RRR

 ln FX

REP

2SH

2NW

Portfolio

␣

RRR

 ln FX

REP

R2

P1

␭ (SH) (NW)

−0.06 (1.08) (1.06)

0.05 (1.57) (1.79)

−0.56 (1.11) (1.98)

9.88 (0.009)

7.17 (0.001)

1

−0.102 (1.04)

−0.102 (1.95)

0.559 (1.52)

−0.232 (1.72)

0.216

P2

␭ (SH) (NW)

−0.45 (1.12) (1.25)

0.18 (1.33) (1.37)

−0.23 (1.05) (1.11)

7.36 (0.003)

6.41 (0.004)

2

−0.011 (1.96)

−0.045 (1.01)

0.253 (1.31)

−0.879 (1.58)

0.161

P3

␭ (SH) (NW)

−0.16 (1.07) (1.15)

0.29 (1.92) (1.91)

−0.41 (1.34) (1.33)

5.61 (0.001)

7.37 (0.008)

3

−0.217 (1.12)

−0.217 (1.11)

0.334 (1.87)

−0.676 (1.41)

0.133

P4

␭ (SH) NW

−0.97 (1.39) (1.17)

0.45 (1.77) (1.82)

−0.33 (1.41) (1.18)

9.94 (0.002)

8.56 (0.005)

4

−0.582 (1.99)

−0.582 (1.67)

0.762 (1.33)

−0.214 (1.11)

0.102

1.22

1.11

1.35

Bai-Perron

1.21

1.24

1.38

1.67

Bai-Perron

The table presents cross-sectional pricing results for the linear factor model based on the dollar risk factor and volatility innovations (not reported), and measures of central bank intervention i.e. reserve requirement ratio (RRR); change in natural log of foreign exchange reserves ( ln FX); and the 7-day repo rate (REP). The test assets are excess returns to CNY carry trade positions against the USD, EUR, GBP and JPY for 1-month (P1), 3-month (P2), 6-month (P3) and 1-year (P4) maturities for the full sample period. Panel A displays coefficient estimates of factor risk prices ␭, obtained by Fama and Macbeth (1973) (FMB) cross-sectional regressions. We report standard errors and 2 test statistics (p-values in parentheses) that are Shanken (1992) adjusted (SH) or Newey-West approach with optimal lag selection (NW). Panel B presents results for time-series regressions of excess returns on a constant (␣), the dollar risk factor and volatility innovations (not reported), and measures of central bank intervention. Newey-West HAC standard errors (with optimal lag selection) are reported in parentheses. The full sample period is January 2000–December 2014 and we use weekly transaction cost adjusted returns. Bai and Perron (2003) structural break test F-statistics are reported in the last row.

innovations but still less than that of the dollar risk factor. Time-series regression estimates also show coefficient signs that are consistent with those we found in Table 3. R2 values however, fell substantially as compared to the model where the dollar risk factor and global FX volatility innovations were the regressors. 4.4. Robustness Throughout our sample period of 2000–2014, the People’s Bank of China (PBC) was often met with accusations of currency manipulation through central bank interventions; keeping CNY values low to benefit the Chinese export market. The Chinese authorities have vehemently denied these allegations; that the interventions were to tame forex speculation and ensure convergence towards their CNY target. Academics and practitioners, however, seem to think otherwise (e.g. Nair and Sinnakkannu, 2010; Goldstein, 2004; Funke and Rahn, 2005). In a joint conference by the PBC and the IMF, Sun Guofeng, Deputy Director General of the Monetary Policy Department revealed that the PBC’s common tools to manage monetary policy is the reserve requirement ratio, management of currency flows as well as central bank lending to commercial banks i.e. the repo rate (Sun, 2014). We include these additional variables: the reserve requirement ratio; changes to foreign exchange reserves; and the 7-day repo rate into our FMB and time-series estimations. Data for these variables were obtained from the Bloomberg and Bankscope databases. Panel A of Table 5 reports the factor loadings and prices for the three measures of PBC intervention. From Panel A, we find the factor loadings and prices for China’s reserve requirement ratio (RRR) and 7-day repo rate (REP) to be negative for all CNY carry trade portfolios. The negative sign is unsurprising since a lower RRR and REP suggests a greater amount of liquidity in the markets; consistent with our findings in Table 4. Factor loading and price for change in foreign exchange reserves ( ln FX) meanwhile is positive, suggesting that through accumulation of foreign exchange reserves (measured in USD), more CNY floods the market creating more liquidity. However, corrected standard errors as well as 2 statistics for all portfolios are large. Meanwhile, the time-series regression estimates in Panel B exhibit similar coefficient signs but standard errors are large while R2 values are low. We performed additional time-series regression estimation by introducing dummies representing the date of monetary policy report announcements by the PBC besides conducting a Chow breakpoint test for these dates. The results are presented in Tables 4 and 6. We also employed a simple event study methodology to test the statistical significance of the cumulative average carry trade returns (CAR −1, +1) surrounding the PBC quarterly monetary policy announcement dates. We only considered the 1-month and 3-month carry trade portfolio returns in the time-series regression and the event study. We did not consider 6-month and 1-year carry trade returns because at much longer investment horizons, directly attributing their returns to quarterly announcement dates would be spurious as there will be many other factors that could have had an impact on the returns. 1-month and 3-month carry traders on the other hand, are more able to adjust their expectations in response to PBC announcements due to their maturity periods that can be more closely matched to the timing of the announcements. These results are presented in Table 8, supplemented with Fig. 3.

370

Table 6 Cross-Sectional Asset Pricing at PBC Monetary Policy Announcement Dates, 1-month Portfolio. Panel A: Dollar Peg Period

Panel B: Managed Float Period ␣

RRR

2004

−0.782 −0.835 −0.655 −0.532 −0.335 −0.513

−0.869 −0.247 −0.235 −0.009 −0.266 −0.905

2005

Q1 Q2 Q3 Q4 Q1 Q2

 ln FX

REP

0.037 0.360 0.190 0.709 −0.217a −0.478c

−0.907 −0.806 −0.110 −0.067 −0.054 −0.288

R

0.114 0.105 0.116 0.188 0.214 0.377

Chow-F 0.7846 0.6284 0.4874 0.3012 0.3198 0.1832

Monetary Policy Announcement

␣

RRR

 ln FX

REP

R2

Chow-F

2005

−0.543 −0.326 −0.172 −0.249 −0.795 −0.694 −0.084 −0.437 −0.541 −0.766 −0.716 −0.300 −0.922 −0.223 −0.965 −0.079 −0.484 −0.404 −0.327 −0.304 −0.579 −0.612 −0.203 −0.214 −0.384 −0.917 −0.102 −0.411 −0.054 −0.563 −0.310 −0.434 −0.381 −0.658 −0.875 −0.361 −0.864 −0.424

−0.541 −0.689 −0.269 −0.376 −0.390 −0.715 −0.829 −0.655 −0.657 −0.398 −0.419 −0.636 −0.112 −0.932 −0.952 −0.412 −0.170 −0.734 −0.466 −0.361 −0.122 −0.684 −0.073 −0.566 −0.293 −0.609 −0.366 −0.087 −0.616 −0.846 −0.450 0.017 −0.520 0.934 0.938 0.045 −0.612 −0.239

−0.786 −0.677b 0.979 0.461 0.167 0.214 0.830 0.411 0.076 0.453 0.025 0.298 0.337 0.278 0.949 0.253 0.488 0.703 0.172 0.468 0.287 0.260 0.649 0.166 0.609 0.399 0.405 0.956 0.957 0.864 0.419 0.212 0.481 0.520 0.871 0.610 0.416 0.955

−0.706 −0.708 −0.032 −0.212 −0.118 −0.248 −0.852 −0.222 −0.272 −0.771 −0.683 −0.932 −0.764 0.469 0.718 −0.601 −0.234 −0.187 −0.775 −0.106 −0.769 −0.409 −0.568 −0.017 0.018 0.910 0.033 −0.057 −0.497 −0.940 0.591 −0.218 −0.349 −0.289 −0.085 −0.185 −0.791 −0.277

0.285 0.332 0.172 0.188 0.143 0.167 0.136 0.160 0.167 0.105 0.168 0.141 0.134 0.209 0.102 0.192 0.177 0.182 0.151 0.108 0.123 0.138 0.153 0.107 0.147 0.108 0.138 0.159 0.124 0.199 0.102 0.139 0.155 0.162 0.145 0.163 0.187 0.141

0.0392 0.3847 0.3051 0.1963 0.8280 0.3651 0.1467 0.8454 0.4143 0.3036 0.5797 0.3715 0.6230 0.6538 0.7522 0.5986 0.1544 0.6923 0.8807 0.3274 0.2649 0.0793 0.5237 0.6475 0.7004 0.8759 0.9225 0.8647 0.7755 0.7219 0.8986 0.2392 0.9070 0.0538 0.8284 0.8437 0.2999 0.8001

2006

2007

2008

2009

2010

2011

2012

2013

2014

Q3 Q4 Q1 Q2 Q3 Q4 Q1 Q2 Q3 Q4 Q1 Q2 Q3 Q4 Q1 Q2 Q3 Q4 Q1 Q2 Q3 Q4 Q1 Q2 Q3 Q4 Q1 Q2 Q3 Q4 Q1 Q2 Q3 Q4 Q1 Q2 Q3 Q4

b

The table presents cross-sectional pricing results for the linear factor model based on the dollar risk factor and volatility innovations (not reported), and measures of central bank intervention i.e. reserve requirement ratio (RRR); change in natural log of foreign exchange reserves ( ln FX); and the 7-day repo rate (REP) at every PBC monetary policy announcement date for the 1-month carry trade portfolio. Policy announcement dates (available only from 2004) were sourced from the English version of the PBC’s official website. Panel A provides the estimates of the central bank interventions during the dollar-peg period (2004 to the first half of 2005). Panel B provides the estimates of the central bank interventions during the managed float period (second half of 2005–2014). Chow breakpoint test F-statistics are also presented. a Indicates significance at 10% level. b Indicates significance at 5% level. c Indicates significance at 1% level.

C.W.H. Cheong et al. / Research in International Business and Finance 39 (2017) 358–376

Monetary Policy Announcement

2

C.W.H. Cheong et al. / Research in International Business and Finance 39 (2017) 358–376

371

Cumulave Average Carry Trade Return (-1,+1) 8 6

2 0 -2

1M (-1,+1) 3M (-1,+1) 2004Q1 2004Q2 2004Q3 2004Q4 2005Q1 2005Q2 2005Q3 2005Q4 2006Q1 2006Q2 2006Q3 2006Q4 2007Q1 2007Q2 2007Q3 2007Q4 2008Q1 2008Q2 2008Q3 2008Q4 2009Q1 2009Q2 2009Q3 2009Q4 2010Q1 2010Q2 2010Q3 2010Q4 2011Q1 2011Q2 2011Q3 2011Q4 2012Q1 2012Q2 2012Q3 2012Q4 2013Q1 2013Q2 2013Q3 2013Q4 2014Q1 2014Q2 2014Q3

Return (%)

4

-4 -6

Fig. 3. Returns to carry trade portfolios. The figure above shows the cumulative average return of the 1-month (1 M) and 3-month (3 M) carry trade portfolios for (−1,+1) periods from the People’s Bank of China’s quarterly monetary policy announcement date The sample period is January 2004–December 2014.

The time-series regression estimates in Tables 6 and 7 show that the only time that the PBC’s monetary policy announcement had an impact on carry trade returns was when the dollar-peg on the Yuan was abolished in July 2005. In Table 8, we see that the policy announcement in Q2 2005 had a significant impact on carry trade returns. We can also see from Fig. 3 that as a result of the peg’s removal in Q2 2005, 1- and 3-month carry trade portfolios plummeted tremendously before rising again. Thereafter, 1- and 3-month returns seem to remain relatively constant over time, which validates our earlier claim from Fig. 2 where carry trade returns remained positive across varying degrees of exchange rate volatility. Similarly, the first half of 2007 seems to have a significant impact on carry trade returns (see Table 8). But far from being the result of monetary policy announcements, we strongly believe that other more significant global events that took place at that time i.e. the onset of the Global Financial Crisis was the cause. The estimates in Table 6 and 7 corroborates this inference as we found that monetary policy interventions had no statistically significant impact on carry trade returns during the same policy announcement dates. Rather, the stability of the Yuan and Yuan-denominated assets provided the much needed reprieve from the volatility of U.S. and European markets during this period; contributing to ‘hot money’ flows into China in the form of carry trade. 5. Implications In this study, we attempted to provide a deeper understanding of the Chinese Yuan’s (CNY) behaviour over the past 15 years; driven by the numerous allegations of unfair trade practices and currency manipulation directed at the People’s Bank of China. To observe the CNY’s behaviour directly, we constructed four portfolios of CNY carry trade positions with maturities ranging from 1 month to 1 year. This is in contrast to the methodology adopted in prior studies that constructed carry trade portfolios on the basis of interest rates. We examine how the CNY and its carry trade returns respond towards three well-established factors that drive exchange rate and carry trade returns movement i.e. the dollar risk factor (Lustig et al., 2011), global FX volatility innovations (Menkhoff et al., 2012) and liquidity (Brunnermeier et al., 2009). We found 1-month CNY carry trade returns to exhibit a behaviour that is similar to those found by Lustig et al. (2011) and Menkhoff et al. (2012). That is to say, 1-month carry trade returns provided a form of hedge for investors during times of high market volatility. In fact, we found carry trade returns of all other maturities to co-move positively with our measure of volatility. Having factored in transaction costs, our findings provide evidence that 1-year CNY carry trade positions were able to earn investors substantial excess returns regardless of market volatility conditions (see Figs. 1 and 2). We also examine the effects China’s change in exchange rate regime had on CNY behaviour. Our results shows that during the dollar-peg period of our sample (January 2000–July 2005), CNY carry trade returns were indeed compensation for global FX volatility, irrespective of maturity but after during the managed float period of our sample (August 2005–December 2014), we only find 1-month returns to exhibit such behaviour. Following the study by Brunnermeier et al. (2009), we also examined the effects of liquidity on CNY carry trade returns using the Pástor and Stambaugh (2003) measure of liquidity. Consistent with the findings of prior studies, we find that liquidity can be attributed as one of the factors that contribute to the excess returns from CNY carry trade, irrespective of maturity. Finally, although our findings show that the CNY does on average, exhibit similar behaviour to most currencies, the factors that contribute to its behaviour are different from other currencies. Lustig et al. (2011) found that the variable they coined, the carry trade risk factor (HML) explains most of the cross-sectional variation in foreign currency excess returns while Menkhoff et al. (2012) found that global FX volatility innovations accounted for most of the variation in returns. In contrast, our results show that cross-sectional variation in CNY carry trade excess returns can be mostly explained by the dollar risk factor instead of global FX volatility innovations. In fact, as we discussed earlier, global FX volatility innovations ranks behind liquidity in explaining variations in CNY carry trade returns. A few implications arise from our study. First, our findings show that the Chinese Yuan is remarkably resilient against the movements of global currencies. Our measure of global FX volatility included about 40 currencies. Yet, despite the

372

Table 7 Cross-Sectional Asset Pricing at PBC Monetary Policy Announcement Dates, 3-month Portfolio. Panel A: Dollar Peg Period

Panel B: Managed Float Period

Monetary Policy Announcement 2004

2005

␣

RRR

 ln FX

REP

R2

Chow-F

Monetary Policy Announcement

−0.320 −0.574 −0.854 −0.350 −0.651 −0.239

−0.960 −0.842 −0.002 −0.488 −0.275 −0.869

0.907 0.502 0.386 0.992 −0.373a −0.618c

−0.769 −0.816 −0.960 −0.810 −0.588 −0.729

0.162 0.139 0.146 0.108 0.136 0.235

0.541 0.018 0.407 0.286 0.474 0.283

2005 2006

2007

2008

2009

2010

2011

2012

2013

2014

Q3 Q4 Q1 Q2 Q3 Q4 Q1 Q2 Q3 Q4 Q1 Q2 Q3 Q4 Q1 Q2 Q3 Q4 Q1 Q2 Q3 Q4 Q1 Q2 Q3 Q4 Q1 Q2 Q3 Q4 Q1 Q2 Q3 Q4 Q1 Q2 Q3 Q4

␣

RRR

 ln FX

REP

R2

Chow-F

−0.819 −0.310 −0.442 −0.620 −0.123 −0.191 −0.536 −0.708 −0.438 −0.241 −0.332 −0.064 −0.469 −0.356 0.816 −0.068 −0.053 0.685 −0.920 −0.550 −0.918 −0.560 0.449 0.916 0.186 −0.083 −0.980 0.817 0.172 −0.003 −0.094 0.055 −0.408 0.899 −0.316 −0.934 −0.787 −0.622

−0.920 −0.951 −0.028 −0.534 −0.745 −0.653 −0.568 −0.866 −0.136 −0.718 −0.070 −0.604 −0.997 −0.017 −0.942 −0.400 −0.475 −0.401 −0.634 −0.935 −0.336 −0.444 −0.078 −0.368 −0.417 −0.365 −0.162 −0.188 −0.883 −0.029 −0.598 −0.373 −0.176 −0.148 −0.830 −0.883 −0.662 −0.427

−0.097a −0.671a 0.337 0.404 0.908 0.762 0.010 0.066 0.001 0.666 0.538 0.258 0.531 0.464 0.789 0.847 0.544 0.817 0.735 0.348 0.223 0.293 0.507 0.115 0.216 0.688 0.749 0.611 0.479 0.089 0.176 0.742 0.303 0.530 0.641 0.915 0.007 0.670

−0.066 −0.488 −0.079 −0.438 −0.896 −0.610 −0.011 −0.087 −0.255 −0.871 −0.125 −0.768 −0.529 −0.414 −0.456 −0.872 −0.931 −0.779 −0.787 −0.077 −0.387 −0.856 −0.942 −0.495 −0.418 −0.561 −0.917 −0.805 −0.824 −0.748 −0.201 −0.189 −0.897 −0.304 −0.653 −0.358 −0.981 −0.550

0.170 0.182 0.140 0.197 0.183 0.125 0.158 0.189 0.110 0.192 0.091 0.070 0.134 0.070 0.111 0.135 0.084 0.122 0.033 0.169 0.054 0.119 0.125 0.106 0.135 0.167 0.133 0.153 0.100 0.109 0.144 0.165 0.090 0.059 0.127 0.078 0.157 0.085

0.209 0.313 0.276 0.506 0.128 0.058 0.017 0.361 0.967 0.785 0.187 0.765 0.555 0.984 0.390 0.132 0.810 0.437 0.782 0.507 0.968 0.787 0.346 0.237 0.320 0.467 0.826 0.004 0.009 0.478 0.276 0.176 0.099 0.399 0.387 0.742 0.017 0.939

The table presents cross-sectional pricing results for the linear factor model based on the dollar risk factor and volatility innovations (not reported), and measures of central bank intervention i.e. reserve requirement ratio (RRR); change in natural log of foreign exchange reserves ( ln FX); and the 7-day repo rate (REP) at every PBC monetary policy announcement date for the 3-month carry trade portfolio. Policy announcement dates (available only from 2004) were sourced from the English version of the PBC’s official website. Panel A provides the estimates of the central bank interventions during the dollar-peg period (2004 to the first half of 2005). Panel B provides the estimates of the central bank interventions during the managed float period (second half of 2005–2014). Chow breakpoint test F-statistics are also presented. b Indicates significance at 5% level. a Indicates significance at 10% level. c Indicates significance at 1% level.

C.W.H. Cheong et al. / Research in International Business and Finance 39 (2017) 358–376

Q1 Q2 Q3 Q4 Q1 Q2

C.W.H. Cheong et al. / Research in International Business and Finance 39 (2017) 358–376

373

Table 8 Test Results for Carry Trade Returns. Monetary Policy Announcement Date

1-Month CAR (-1,+1)

p-value

3-Month CAR (-1, +1)

p-value

2004Q1 2004Q2 2004Q3 2004Q4 2005Q1 2005Q2 2005Q3 2005Q4 2006Q1 2006Q2 2006Q3 2006Q4 2007Q1 2007Q2 2007Q3 2007Q4 2008Q1 2008Q2 2008Q3 2008Q4 2009Q1 2009Q2 2009Q3 2009Q4 2010Q1 2010Q2 2010Q3 2010Q4 2011Q1 2011Q2 2011Q3 2011Q4 2012Q1 2012Q2 2012Q3 2012Q4 2013Q1 2013Q2 2013Q3 2013Q4 2014Q1 2014Q2 2014Q3 2014Q4

1.515 0.444 0.943 0.267 2.334 −4.427c 1.317b 2.432 2.624 1.593 1.742 1.978 2.199a 0.376a 1.186a 2.113 2.525 3.752 2.015 4.322 1.663 1.780 2.020 2.365 1.422 2.400 2.172 1.051 2.199 0.853 1.364 1.092 1.914 1.221 1.464 0.584 2.232 1.571 1.147 1.294 2.247 1.512 1.286 1.212

0.691 0.161 0.334 0.540 0.223 0.009 0.034 0.179 0.191 0.369 0.218 0.314 0.073 0.095 0.069 0.114 0.190 0.399 0.210 0.440 0.240 0.194 0.305 0.299 0.138 0.152 0.230 0.236 0.251 0.286 0.381 0.292 0.235 0.117 0.429 0.210 0.116 0.195 0.127 0.134 0.433 0.114 0.288 0.112

6.592 2.556 4.481 4.033 5.842 −1.699a 3.393b 5.930 6.846 3.101 −0.269 2.007 3.556b 2.181b 0.293a −0.209 1.437 3.536 4.509 4.635 4.569 1.528 1.269 2.051 3.732 1.691 2.018 2.308 2.831 2.033 2.096 2.501 2.336 3.833 3.288 2.837 2.087 2.384 2.714 2.827 2.002 3.549 3.439 3.340

0.223 0.199 0.225 0.232 0.280 0.065 0.036 0.379 0.258 0.282 0.107 0.205 0.042 0.039 0.072 0.159 0.158 0.203 0.257 0.334 0.463 0.391 0.285 0.176 0.247 0.130 0.232 0.396 0.233 0.177 0.261 0.168 0.261 0.374 0.303 0.163 0.182 0.154 0.207 0.330 0.238 0.130 0.152 0.178

The table presents the cumulative average carry trade returns (CAR) for the 1-month and 3-month Yuan carry trade portfolios at t − 1 and t + 1, where t = 0 is the date of the PBC’s monetary policy announcement. p-values for H0 : CAR = 0 are also presented. a Indicates significance at 10% level. The sample period is January 2004–December 2014. b Indicates significance at 5% level. c Indicates significance at 1% level.

inclusivity we find that the value of CNY is relatively unperturbed by the movements of other currencies around the world. We also find the CNY to be unaffected by Chinese and global macroeconomic conditions. Rather, the CNY is highly sensitive towards the dollar risk factor, a clear indication that despite the PBC’s currency liberalization policies, the CNY remains substantially exposed to the volatility of the dollar. The CNY is also sensitive towards liquidity risks from U.S. stock markets. Using Pástor and Stambaugh’s (2003) innovations in aggregate liquidity, we find that changes in the U.S. stock market’s liquidity have a veritable effect on CNY carry trade returns. This substantiates the claims by Morrison (2009) and McKinnon and Schnabl (2009) that existing macroeconomic conditions (at least during our sample period) have provided a liquidity boost to the markets and as a result, contributed to the hot money inflows into China in the form of purchases of the CNY or Yuan-denominated assets. CNY carry trade susceptibility to U.S. market liquidity also provides evidence of a wealth transfer between equity markets and FX markets. We deviated from conventional practice by not only observing 1-month but 3, 6 and 12-month periods as well. Our findings show that current understanding of interest parity conditions; carry trade and FX sensitivities hold, but only up to a month as the relationship between carry trade returns and FX factors begin to deteriorate as the time to maturity of the carry trade position increases, before moving in the opposite direction at the 1-year mark. Our study also has some practical implications for investors. First, prior studies on carry trade constructed portfolios that consisted of 40 or more currencies sorted on the basis of their forward discounts (i.e. interest rates) relative to a base

374

C.W.H. Cheong et al. / Research in International Business and Finance 39 (2017) 358–376

currency. Although their findings are robust, the mathematical effort can be daunting. We instead, choose to construct carry trade portfolios simply based on the value of the CNY against just 4 currencies i.e. the U.S. dollar, the Euro, the British Pound and the Japanese Yen for varying maturities. Despite the simplicity of our method, our findings are consistent with those of studies that used portfolios with a larger asset base. Implicitly, we can say that our rolling 1-month CNY carry trade portfolio that only consists of 4 assets has the ability to hedge against global FX volatility that is similar to a monthly-rebalanced carry trade portfolio of 40 or more assets. That is to say, due to the PBC’s existing currency policy, investors are able to achieve a level of FX diversification that is equivalent to a portfolio that has a large asset base simply by trading the CNY 1-month forward against the 4 major currencies. With regards to profitability, we provide evidence that a rolling 1-year CNY carry trade portfolio that consists of only 4 assets is still able to deliver substantial excess returns to investors even during periods of supposedly high market volatility. A common method of carry trade portfolio construction is to short the portfolio of currencies with the lowest interest rates and long the portfolio with the highest interest rates. As evidenced in prior studies, this portfolio is able to deliver returns in excess of 10 percent during periods of low to moderate market volatility, yet delivers substantial negative returns when volatility is high (see Menkhoff et al., 2012; Fig. 2). In contrast, our 1-year portfolio consistently delivers positive excess returns regardless of market volatility, even after accounting for transaction costs. Nevertheless, we do not seek to prove which method is superior. Rather, our study shows that the PBC’s currency policies and practices have created an ideal opportunity for investors to exploit. Understandably, our study raises the question of whether CNY carry trade contributes to the (mis)alignment of the CNY vis-à-vis its equilibrium value. That is to say, do CNY carry trade returns hedge against FX volatility, or is FX volatility a result of carry trade7 ? In its simplest form, carry trade involves borrowing from a low-interest funding currency, exchanging the borrowed funds for the target currency in the spot FX market and then investing in a high-yield account before exchanging the investment proceeds back to the funding currency in the spot FX market. In theory, exchanging funding currency for target currency and then back again creates a stir in the exchange rate between the two currencies, ceteris paribus, thereby contributing to FX volatility. However, a successful carry trade is heavily reliant on the condition that there is little to no movement in the exchange rate between the funding and target currencies. That is to say, rather than contribute to FX volatility, the biggest risk to carry trade is in fact, FX volatility. There has been as far as the literature suggests no evidence to show otherwise. Rather, historical evidence has shown that exchange rate movements follow a random walk (Curcuru et al., 2010) instead of uncovered interest parity – the driving theory behind carry trade. Meanwhile, academics and practitioners alike have attributed FX volatility to central bank policy interventions (Humpage, 2003) involving output and inflation volatility (Morana, 2009); the openness of an economy or significant news events in the country (Stancik, 2006); moral hazards as a result of the exchange rate regime in place (Eichengreen and Hausmann, 1999); market optimism and pessimism (De Grauwe and Kaltwasser, 2007); trade interdependence and financial market development (Devereux and Lane, 2003); or microstructural factors (Canales-Kriljenko and Habermeier, 2004; Lyons, 2001); instead of carry trade, especially since carry trade positions only represent a small portion of the total estimated FX trades that take place daily (Galati et al., 2007). As a consequence, it is unlikely that Yuan carry trade has an impact on the (mis)alignment of the Yuan exchange rate.

6. Conclusion China has often come under fire over the last decade due to their alleged currency manipulation policies, keeping the Chinese Yuan artificially undervalued thereby giving China a significant boost in their trade balance. However, China’s strict monetary policy has also been credited for providing the much needed stability in the FX markets during the 1997 Asian Financial Crisis and the more recent Global Financial Crisis of 2007. Despite the Yuan’s rising importance in global FX markets and its status as a future reserve currency, seminal papers on foreign exchange and carry trade dynamics have neglected the Yuan in their analysis for reasons unbeknownst to us. We fill that gap by examining the Yuan carry trade return behaviour using methods that have been well-established in the literature. We find short-term Yuan carry trade portfolios may serve as a hedge against market volatility. As the time to maturity of the carry trade position increases, the returns exhibit opposing behaviours providing investors with an opportunity to earn substantial returns during period of low market volatility. This study also shows that dollar-specific risk and U.S. market liquidity have a significant influence on Yuan carry trade returns as compared to global FX volatility. This provides evidence of firstly, the Yuan’s dependence on the dollar in its valuation and secondly, the transfer of wealth between U.S. equity markets and dollar-denominated assets to the Yuan and Yuandenominated assets. Finally, by observing Yuan carry trade returns during the dollar-peg and managed float period, we can surmise that contrary to the UIP condition, a stable exchange rate does not necessitate positive carry trade returns. Our findings are of course limited in generalizability due to our country-specific focus besides the imbalance in observations for the dollar-peg and managed float periods, which is due to a lack of data on Yuan forward rates before the year 2000. We encourage future research in this regard to build on our findings by exploring the nature of carry trade returns at maturities longer than a month using a wider range of currencies. In addition, we also encourage future researchers to examine the

7

We would like to thank the anonymous reviewer for raising this matter with us; making our discussion a more robust one.

C.W.H. Cheong et al. / Research in International Business and Finance 39 (2017) 358–376

375

possibility of a currency investment strategy that capitalizes the hedging capabilities of 1-month carry trade portfolios whilst exploiting the returns resilience of 1-year portfolios. Appendix. See Tables A1 and A2. Table A1 Cross-Sectional Asset Pricing: Factor Prices. Factor Prices FMB

DOL

VI

USD

EUR

GBP

JPY

2CE

2TE

2.35 (0.14)

1.68 (0.34)

P1

␭ (TSE)

7.68 (0.21)

−0.05 (0.05)

1.97 (0.32)

1.68 (0.11)

1.54 (0.41)

−0.89 (0.33)

P2

␭ (TSE)

6.51 (0.24)

0.03 (0.04)

1.58 (0.55)

1.88 (0.63)

1.69 (0.52)

−0.58 (0.12)

P3

␭ (TSE)

−15.34 (0.64)

0.06 (0.08)

1.44 (0.34)

1.31 (0.68)

1.78 (0.42)

−1.02 (0.19)

P4

␭ (TSE)

−26.55 (0.61)

0.03 (0.03)

1.64 (0.12)

1.78 (0.25)

1.12 (0.31)

−1.56 (0.52)

1.42

1.12

Bai-Perron

The table reports cross-sectional pricing results for the linear factor model based on the dollar risk factor (DOL) and volatility innovations (VI). The test assets are excess returns to CNY carry trade positions against the USD, EUR, GBP and JPY for 1-month (P1), 3-month (P2), 6-month (P3) and 1-year (P4) maturities for the full sample period (2000–2014). The panels display coefficient estimates of factor risk prices ␭, obtained by Fama and Macbeth (1973) (FMB) cross-sectional regressions. We report currency fixed effects (USD, EUR, GBP, and JPY) and time-clustered standard errors (in parantheses), as well as 2 test statistics (p-values in parentheses) for the currency effect (CE) and time-effect (TE). (2003) structural break test F-statistics are reported in the last row.

Table A2 Cross-Sectional Asset Pricing: Factor Betas. Panel B: Factor Betas Portfolio

␣

DOL

VI

USD

EUR

GBP

JPY

R2

1

0.001 (0.002)

0.004 (0.003)

1.311 (0.312)

1.64 (0.111)

1.87 (0.232)

1.98 (0.153)

−1.55 (0.341)

0.784

2

0.011 (0.021)

0.005 (0.002)

0.522 (0.351)

1.38 (0.251)

1.57 (0.221)

1.62 (0.283)

−2.01 (0.357)

0.754

3

0.031 (0.011)

0.001 (0.004)

0.312 (0.273)

2.12 (0.143)

2.34 (0.254)

2.41 (0.312)

−2.23 (0.223)

0.795

4

0.035 (0.015)

−0.004 (0.001)

−1.541 (0.387)

2.41 (0.331)

2.81 (0.382)

2.62 (0.422)

−2.31 (0.416)

0.865

Bai-Perron

1.22

1.54

2.01

2.11

1.64

1.58

1.91

The table reports cross-sectional pricing results for the linear factor model based on the dollar risk factor (DOL) and volatility innovations (VI). The test assets are excess returns to CNY carry trade positions against the USD, EUR, GBP and JPY for 1-month (P1), 3-month (P2), 6-month (P3) and 1-year (P4) maturities for the full sample period (2000–2004). We report results for time-series regressions of excess returns on a constant (␣), the dollar risk factor (DOL), volatility innovations (VI) and currency fixed effects (USD, EUR, GBP, JPY). Time-clustered standard errors are reported in parentheses. We use weekly transaction cost adjusted returns. Bai and Perron (2003) structural break test F-statistics are reported in the last row.

References Adrian, T., Rosenberg, J., 2008. Stock returns and volatility: pricing the short-run and long-run components of market risk. J. Finance 63, 2997–3030. Andersen, T.G., Bollerslev, T., Diebold, F.X., Labys, P., 2001. The distribution of realized exchange rate volatility. J. Am. Stat. Assoc. 96, 42–55. Andrews, D.W.K., 1991. Heteroskedasticity and autocorrelation consistent covariance matrix estimation. Econometrica 59, 817–858. Ang, A., Chen, J., Xing, Y., 2006a. Downside risk. Rev. Financ. Stud. 19, 1191–1239. Ang, A., Hodrick, R., Xing, Y., Zhang, X., 2006b. The cross-section of volatility and expected returns. J. Finance 61, 259–299. Bai, J., Perron, P., 2003. Computation and analysis of multiple structural change models. J. Appl. Econ. 18, 1–22. Bandi, F.M., Moise, C.E., Russell, J.R., 2008. The joint pricing of volatility and liquidity. In: Working Paper. University of Chicago. Bansal, R., Dahlquist, M., 2000. The forward premium puzzle: different tales from developed and emerging economies. J. Int. Econ. 51, 115–144. Bhide, A., Phelps, E., 2007. A dynamic theory of China-U.S. trade: making sense of the imbalances. World Econ. 8, 7–25. Bilson, J.F.O., 1981. The ‘Speculative efficiency’ hypothesis. J. Bus. 54, 435–451. Brunnermeier, M., Nagel, S., Pedersen, L.H. 2009. Carry Trades and Currency Crashes, NBER Macroeconomics Annual 2008, 23, 313–347. Burnside, C., Eichenbaum, M., Kleshchelski, I., Rebelo, S., 2011. Do peso problems explain the returns to the carry trade. Rev. Financ. Stud. 24, 853–891. Burnside, C., 2012. Carry trades and risk. In: James, J., Sarno, L., Marsh, I.W. (Eds.), Handbook of Exchange Rates. Wiley, Hoboken NY.

376

C.W.H. Cheong et al. / Research in International Business and Finance 39 (2017) 358–376

Canales-Kriljenko, J., Habermeier, K. 2004. Structural factors affecting exchange rate volatility: a cross-section study, IMF Working Paper, No. 04/147. Carhart, M.M., 1997. On persistence in mutual fund performance. J. Finance 52, 57–82. Cenedese, G., Sarno, L., Tsiakas, I., 2014. Foreign exchange risk and the predictability of carry trade returns. J. Bank. Finance 42, 302–313. Christiansen, C., Ranaldo, A., Soderlind, P., 2011. The Time-varying Systematic Risk of Carry Trade Strategies. J. Financ. Quant. Anal. 46, 1107–1125. Cochrane, J.H., 2005. Asset Pricing. Princeton University Press, NJ. Curcuru, S., Vega, C., Hoek, J. 2010. Measuring carry trade activity, Board of Governors of the Federal Reserve System, Working Paper. Da, Z., Schaumburg, E. 2009. The Pricing of Volatility Risk across Asset Classes and the Fama-French Factors, Northwestern University Working Paper. Das, D.K., 2006. The evolution of renminbi yuan and the protracted debate on its undervaluation: an integrated review. J. Asian Econ. 20, 570–579. De Grauwe, P.,Kaltwasser, P. R. 2007 Modeling optimism and pessimism in the foreign exchange market, CESifo Working Paper Series, No. 1962. DeRosa, D.F., 2005. China’s defense of the peg perpetuates central planning. Cato J. 25, 49–54. Devereux, M., Lane, P., 2003. Understanding bilateral exchange rate volatility. J. Int. Econ. 60, 109–132. Dobrynskaya, V., 2014. Downside risk of carry trades. Rev. Finance 18, 1885–1913. Eichengreen, B., Hausmann, R. 1999. Exchange rates and financial fragility, NBER Working Paper Series, Working Paper 7418. Engel, C., 1996. The forward discount anomaly and the risk premium: a survey of recent evidence. J. Empir. Finance 3, 123–192. Fama, E.F., French, K.R., 1989. Business conditions and expected returns on stocks and bonds. J. Financ. Econ. 25, 23–49. Fama, E.F., Macbeth, J., 1973. Risk, return and equilibrium: empirical tests. J. Polit. Econ. 81, 607–636. Fama, E.F., 1984. Forward and spot exchange rates. J. Monet. Econ. 14, 319–338. Frankel, J. 2005. On the Renminbi: the choice between adjustment under a fixed exchange rate and adjustment under a flexible rate, NBER Working Paper, No. 11274. Funke, M., Rahn, J., 2005. Just how undervalued is the chinese renminbi? World Econ. 8, 465–489. Galati, G., Melvin, M., 2004. Why has FX trading surged? Explaining the 2004 triennial survey. BIS Quart. Rev., 67–74. Galati, G., Heath, A., McGuire, P., 2007. Evidence of carry trade activity. BIS Quart. Rev., 27–41. Goldstein, M., Lardy, N., 2006. China’s exchange rate policy dilemma. Asian Curr. Matters 92, 422–426. Goldstein, M., 2004. China and the renminbi exchange rate. In: Bergsten, C., Williamson, J. (Eds.), Dollar Adjustment: How Far? Against What? Institute for International Economics, Washington. Guo, J., 2009. Examining the exchange rate regime for China. Int. Res. J. Finance Econ. 1, 64–77. Hansen, L.P., Hodrick, R.J., 1980. Forward exchange rates as optimal predictors of future spot rates: an econometric analysis. J. Polit. Econ. 88, 829–853. Harvey, C.R., Siddique, A., 1999. Autoregressive conditional skewness. J. Financ. Quant. Anal. 34, 465–477. Harvey, C.R., Siddique, A., 2000. Conditional skewness in asset pricing tests. J. Finance 55, 1263–1295. Hodrick, R.J., 1987. The Empirical Evidence on the Efficiency of Forward and Futures Foreign Exchange Markets. Harwood Academic. Humpage, O.F. 2003. Government interventions in the foreign exchange market, Federal Reserve Bank of Cleveland Working Paper, 03–15. Jaeger, M., 2010. Yuan as a Reserve Currency: Likely Prospects & Possible Implications, Research Briefing. Deutsche Bank Research. Lustig, H., Verdelhan, A. 2005. The Cross-section of Foreign Currency Risk Premia and Consumption Growth Risk, NBER Working Paper, No. 11104. Lustig, H., Verdelhan, A., 2007. The cross section of foreign currency risk premia and consumption growth risk. Am. Econ. Rev. 97, 89–117. Lustig, H., Roussanov, N., Verdelhan, A., 2011. Common risk factors in currency markets. Rev. Financ. Stud. 24, 3731–3777. Lyons, R., 2001. The Microstructure Approach to Exchange Rates. MIT Press, Cambridge. Makin, J.H., 2011. Can China’s currency go global? Economic outlook. American Institute of Public Policy Research. McKinnon, R., Schnabl, G., 2009. The case of stabilizing China’s exchange rate: setting the stage for fiscal expansion. China World Econ. 17, 1–32. Menkhoff, L., Sarno, L., Schmeling, M., Schrimpf, A., 2012. Carry trades and global foreign exchange volatility. J. Finance 68, 681–718. Morana, C., 2009. On the macroeconomic causes of exchange rate volatility. Int. J. Forecasting 25, 328–350. Morrison, W.M. 2009. China and the Global Financial Crisis: Implications for the United States, CRS Report for Congress, Congressional Research Service. Nair, M.K.V., Sinnakkannu, J., 2011. De-pegging of the chinese renminbi against the U.S. dollar: analysis of its effects on China’s international trade competitiveness. J. Asia Pacific Econ. 15, 470–489. Newey, W.K., West, K.D., 1987. A simple, positive semi-definite, heteroskedasticity and autocorrelation consistent covariance matrix. Econometrica 55, 703–708. Pástor, L., Stambaugh, R., 2003. Liquidity risk and expected stock returns. J. Polit. Econ. 111, 642–685. Petersen, M.A., 2009. Estimating standard errors in finance panel data sets: comparing approaches. Rev. Financ. Stud. 22, 435–480. Ranaldo, A., Soderlind, P., 2010. Safe haven currencies. Rev. Finance 10, 385–407. Robb G. June 10, 2010. ‘Critically Important’ that China Move on Currency: Geithner. Retrieved from http://www.marketwatch.com/story/geithner-takes-strong-stand-on-china-currency-2010-06-10?pagenumber=1, on December 01.10.15. Shanken, J., 1992. On the estimation of beta pricing models. Rev. Financ. Stud. 5, 1–34. Sinnakkannu, J., Nassir, A.M., 2008. Revisiting the causes of the east asian financial crisis: was it predictable? Int. J. Econ. Perspect. 2, 78–93. Schnabl, G. (2010). The role of the Chinese dollar peg for macroeconomic stability in China and the world economy. Working Papers on Global Financial Markets, 13(1), 1-32. Stancik, J. 2006. Determinants of exchange rate volatility: The case of the new EU members, CERGE-EI Discussion Paper Series, No. 2006-158. Sun, G. (2014). Chinese monetary policy tools. Paper presented at PBC and IMF Joint Conference – New issues on monetary policy: International experience and relevance for China, March 27, 2014, Beijing, China. Sun, T., Zhang, X. 2009. Spillovers of the U.S. Subprime Financial Turmoil to Mainland China and Hong Kong SAR: Evidence from Stock Markets, IMF Working Paper, WP/09/166. Talley, I. November 30, 2015. China Joins World’s Elite Currency Club. Retrieved from http://www.wsj.com/articles/imf-lifts-chinese-yuan-to-elite-lending-reserve-currency-status-1448903067?mod=e2fb, on 01.12.15. Wooldridge, J., 2007. Econometric Analysis of Cross Section and Panel Data. MIT Press, Cambridge. Zhang, J.X., 2010. Will RMB appreciation reduce trade deficit in the U.S.? J. Asia Pac. Econ. 17, 17.