US Dollar Carry Trades in the Era of “Cheap Money”

In this paper, we employ a unique dataset of actual US dollar (USD) forward positions against a number of currencies taken by so-called commodity trading advisors (CTAs). We investigate the extent to which these positions exhibit a pattern of USD carry trading or other patterns of currency trading over the recent period of ultra-loose US monetary policy. Our analysis indeed shows that USD positions against emerging-market currencies are characterised by a pattern of carry trading. That is, the US dollar, as the lower-yielding currency, is associated with short positions. The payoff distributions of these positions, moreover, are found to have positive Sharpe ratios, negative skewness and high kurtosis. On the other hand, we find that USD positions against other developed-market currencies have a pattern completely opposite to carry trading, which is in line with the uncovered interest parity trading; i.e. the lower-yielding (higher-yielding) currency is associated with long (short) positions.


Introduction
In this paper, we employ and analyse a unique dataset of actual USD forward positions versus a number of emerging-and developed-market currencies.Our objective is to shed light on the characteristics of the currency trading styles implied by these positions during a recent sample period, with emphasis on USD carry trading. 1 The motivation behind this emphasis is the near-zero US interest rate over the vast majority of our sample period.
Currency carry trading implies that traders invest in higher-yielding currencies (investment or target currencies) using borrowings in lower-yielding currencies (funding currencies).So, by "USD carry trading" we mean the carry trades in which the US dollar is the funding currency.Under the assumption of the covered interest rate parity (CIP), this strategy can be implemented in the foreign exchange (FX) forward markets by taking long positions in currencies which are traded on the forward discount (high-interest-rate currencies) and short positions in currencies which are traded on the forward premium (low-interest-rate currencies).The motivation behind currency carry trading is the well-established finding of the downward bias in the unbiasedness hypothesis (UH) predictions, i.e. the forward premium bias puzzle (see, for example, Fama, 1984;Frankel and Chinn, 1993;Bansal and Dahlquist, 2000;and Frankel and Poonawala, 2010, among others).The standard expression of this hypothesis is through the Fama regression of:2    and   t k is a white noise process which implies that the currency excess return is expected to be zero.
In contrast, the well-documented finding of significantly less than unity and, more often, a negative slope coefficient implies that a positive currency excess return can be achieved by trading on currency interest differentials (on the carry).Carry trades have been found to be profitable on average with an attractive Sharpe ratio compared to stock and bond markets (see, for example, Hochradl and Wanger, 2010;Pojarliev, 2005;Burnside et al., 2006Burnside et al., , 2007Burnside et al., , 2008;;Gilmore and Hayashi, 2011;Menkhoff et al., 2012).The traditional common target currencies are found to include the Australian dollar, New Zealand dollar, Mexican peso, Brazilian real and Indian rupee, while funding currencies include mainly the Japanese yen and the Swiss franc (see Bilson, 2013;Galati and Melvin, 2004;Galati et al., 2007;McGuire and Upper, 2007;Gagnon and Chaboud, 2007;Curcuru et al., 2010).
In the wake of the 2007-2008 financial crisis, many countries, especially developed countries including the United States, adopted unconventional loose monetary policies with the purpose of stimulating their sluggish and unstable economies.This period is termed in the financial press as "the era of cheap money".On the other hand, other countries, especially emerging markets, maintained relatively high interest rates over the same period.Because of the potential impact of these effects on the trading decisions of FX traders, it is worthwhile to consider currency trading in general and USD carry trading in particular over the sample period of the paper.For example, Gilmore and Hayashi (2011) find a strong relation between currency excess return and the carry.Spronk et al. (2013) show that the more important the interest-rate differential, the more attractive the currency carry trading.It is also suggested that the relatively high-yielding emerging markets have been major recipients of carry trade flows in the wake of the crisis, and that this flow represents a "global search for yield", which is triggered by the unconventional expansionary monetary policy of developed economies (see, for example, Kim, 2015;Mishra et al., 2014).For a set of major currencies, Brière and Drut (2009) document the superiority of "fundamentals-based" trades over carry trades when uncertainty is high.
In light of this, the crux of the paper is to analyse our dataset of the USD forward positions to find out the extent to which they show characteristics of USD carry trading or another trading strategy over the recent period of record-low US interest rates.In other words, we investigate whether these positions exhibit a response to the very low US interest rates by having a pattern of USD carry trading or if other patterns of trading strategies can be identified across different currency markets.The distinctive feature of this study is that we have access to a dataset of dailyaggregated USD forward positions against a number of developed-and emergingmarket currencies.This dataset is collected from a Swedish investment specialist, Risk & Portfolio Management AB (RPM), which is a fund of hedge funds investing in managed futures strategies which are also known as commodity trading advisors (CTAs).CTAs engage in various strategies like trend-following, short-term trading and global macro that often employ carry trading as a sub-strategy.
By exploiting and analysing our private dataset, we find significant long-run equilibrium relationships which directly relate the USD forward positions to their forward premium.The relationships point to different trading strategies for emerging-and developed-market currencies.For emerging-market currencies, we find that these relationships are consistent with carry trading, i.e. the lower-yielding currency (USD) is associated with short positions and vice versa.This carry trading pattern of forward shorting a lower-yielding currency is induced by the expectations that the loweryielding currency will not actually appreciate as much on average as the forward rate implies, or it will even depreciate.This in turn implies a profit at maturity on average.On the other hand, we find that the reverse holds between the US dollar and developed-market currencies.In other words, we find a pattern of "fundamentalsbased" trading consistent with the uncovered interest parity condition.That is, the lower-yielding (higher-yielding) currency is associated with more long (short) positions.These anti-carry positions can reflect the unattractiveness of the USD carry trading against developed-market currencies due to the increased uncertainty and narrow interest differentials for these markets over the period following the recent crisis.
Given that our data set is collected from FX traders, which are mainly trend followers, the results of the different trading strategies for emerging-market and developed-market currencies shed some light on the trading behavior of this group of FX market participants.On the one hand, the characteristics of carry trades for emerging-market currencies, which involve a long high-interest currency against a lowinterest-rate currency, reflect a trend-following strategy based on the expectation that a high-interest-rate currency will appreciate-i.e. based on the appreciation trend of the high-interest-rate currency.On the other hand, the characteristics of "fundamentals-based" trades for developed-market currencies which involve a long lowinterest-rate currency against a high-interest-rate currency reflect a trend-following strategy which is based on the expectations that the low-interest-rate currency will appreciate-i.e. based on the appreciation trend of the low-interest-rate currency.This is in line with the heterogeneous agents model developed by Spronk et al. (2013), which demonstrates that depending on the dominant trend in the market, FX trend followers can be in the same line of either carry traders or fundamentalists.In this sense, our results provide some insights into these features of FX trendfollowing traders.
Our work is also mainly related to the studies of tracking and providing evidence on currency carry trading.The findings of the existing studies on tracking carry trading activities are, to a large extent, deemed implicit and indirect.This is because they only use publicly available datasets such as Bank for International Settlements (BIS) reports and statistics, FX turnovers and FX futures positions (see Galati et al., 2007;McGuire and Upper, 2007;Gagnon and Chaboud, 2007;Curcuru et al., 2010).In contrast, our dataset enables us to document a direct relationship between forward positions and the forward premium, thus providing explicit evidence on these activities.Despite speculation that "the unprecedented low interest rates of the US could have induced large-scale carry trades against high-yielding emerging-market currencies" (Aizenman et al., 2014), the existing literature, to our knowledge, lacks such direct evidence on this trend, especially in currency forward markets.Tracking currency carry trades is important because of such trades' vital implications.For example, identifying periods of increasing carry trades is relevant, as it has been suggested and found that carry trades increase the risk of currency crashes for investment currencies (see, for example, Brunnermeier et al., 2008;Breedon et al., 2016) and that a sudden and massive unwinding of carry positions can contribute substantially to the volatility shocks of the FX and other financial markets, especially for target countries (see, for example, Gagnon and Chaboud, 2007;Nishigaki, 2007;Galati et al., 2007;Eichengreen and Gupta, 2014;Mishra et al., 2014).Therefore, tracking carry trades can help enhance the understanding of markets' volatility dynamics.Currency carry trades can also play a role in the violation of the uncovered interest parity (UIP) by creating "self-enforcing" speculation opportunities which are intensified by the FX trend followers (see, e.g.Plantin and Shin, 2007;Gagnon and Chaboud, 2007;Spronk et al., 2013).
Moreover, our dataset allows us to analyse the performance of actual and not synthetic carry trading strategies.Earlier papers relying on hypothetical carry positions, assuming short positions in lower-yielding currencies and long positions in higher-yielding currencies, have found carry trades to be profitable.These studies have also documented high kurtosis and negative skewness for carry trade returns (see, for example, Burnside et al., 2008;Burnside et al., 2007;Burnside et al., 2006;Menkhoff et al., 2012).In contrast to these studies, we evaluate the performance of actual USD positions against emerging-market currencies where we have explicit evidence that carry trades were being executed.We investigate the extent to which the properties documented based on synthetic positions also apply to our actual positions.
The remainder of the paper is organized as follows: Section 2 describes the dataset and Section 3 explains the methodology.The results are presented in Section 4 and Section5 concludes the paper.

Data
The empirical analysis in the paper draws on a private dataset. 3For the analysis, we employ a dataset of daily-aggregated short-term long and short USD forward positions against various developed-and emerging-market currencies.The complete dataset contains the positions' forward rate (F), spot rate (S), maturity date and the spot rate that transpired ex-post at the maturity date.The source of the dataset is RPM Risk & Portfolio Management AB (RPM), a specialist investment manager based in Stockholm, Sweden.RPM is a fund-of-funds specializing in managed futures strategies, i.e.CTAs and liquid global macro managers that trade in many futures markets such as currencies, bonds, equity indices and many other commodity futures.
It is well known that trend following is the strategy most widely used by CTAs.Galati and Melvin (2004) and Galati et al. (2007) point to the increasingly active role of CTAs in the FX market and their engagement in currency carry trades.Spronk et al. (2013) study the interactions of fundamental, trend-following and carry trade strategies in a theoretical model.They argue that carry traders have a directional role in driving the UH beta.When interest-rate differentials are persistent, carry traders introduce momentum effects in a currency that is picked up and extrapolated by trend followers.Furthermore, it is only due to the existence of trend followers that carry traders can have such a profound effect on FX markets.
We have daily-aggregated short-term USD forward positions against twelve developed-and emerging-market currencies.The developed-market currencies include EUR, JPY GBP, CHF, SEK and CAD, while emerging market currencies include INR, BRL, MYR, ZAR, CLP and MXN.Spot and forward exchange rates are expressed where the USD is the base currency (other currency units per 1 USD).From these positions, we construct a variable which we call "net position" NP.Against every other currency, the daily-aggregated NP in terms of the US dollar is calculated as the net of the long and short USD positions, i.e. the USD long positions minus short positions: where m is the number of position takers, and , ,  j Pos j L S is the position held in the currency, long and short, respectively.
Positive (negative) NP, therefore, implies a net long (short) position in USD against the other currency.NP is typically stationary so that in our cointegration analysis in this paper, we make use of the cumulative net position (CNP).CNP is the cumulative sum of the NP series: by construction, it is I(1) and its first difference is NP.Δ s is spot rate change in the log difference.fp is calculated as (LogF -LogS)*100.

Figure 1 Benchmark Policy Rates over the Sample Period
Note: End-of-month value of the benchmark policy interest rate for every currency obtained from DataStream.

Figure 2 Average Interest-Rate Differentials
Note: The interest-rate differential is calculated as the other currency interest rate minus the USD interest rate.
Table 1 reports the sample period for every currency pair along with descriptive statistics of the variables of interest.We note that the US dollar typically has net short (long) positions whenever it is at a forward premium (discount), except for in the case of the SEK/USD pair.
Figure 1 depicts the benchmark policy interest rate for every country in the sample.The main feature to be noticed is the difference between developed-and emerging-market interest rates.For developed countries including the US, we can see a pattern of very low interest rates during the period after late 2008.For emerging countries, though many have cut their rates, they still have maintained relatively high interest rates.This further motivates the study of carry trading during the sample period.
To make this clearer, Figure 2 depicts the monthly average of interest-rate differentials, calculated as the other currency interest rate minus the USD interest rate.We obviously note very low interest-rate differentials between the developedmarket currencies and US dollar compared to those of emerging-market currencies.

Methodology
The existence of a cointegrated relationship between variables implies that there is a long-run equilibrium relationship binding them together, which also reveals important insights about their dynamic behavior.Thus, multivariate Johansen cointegration analysis techniques are employed to test for the existence and the number of cointegrated relationships between the USD forward positions ( ) CNP , spot exchange rates ( ) LogS and forward exchange rates ( ) LogF in order to investigate how they are related over the long run.We employ cointegration analysis, as we first intend to document whether or not these variables are cointegrated, which implies that there are common forces driving the variables over the long run, and then this is used to cast light on what the resulting cointegrated relationships imply for currency trading styles.The cointegration setup has the advantage that it provides an overview of how the variables are related over the long run with a more dynamic framework apart from the possible short-run deviations on the observation-by-observation basis.
Given that there is a significant long-run equilibrium relationship binding our variables together, a vector error correction mechanism (VECM) can be estimated according to Engel and Granger (1987).Such a VECM as specified below, along with an unrestricted VAR model, will enable us to evaluate the relationships between the variables and investigate the direction of causality among them.
where  are constants,  are adjustment coefficients, 1 ˆ t e is the error correction term and  are error terms.
In the framework of an unrestricted VAR model, specified similarly as in the equations above but with the exclusion of the error correction term 1 ˆ t e , i.e. setting  = 0, a Granger Causality/Block Exogeneity Wald test is performed to determine whether the lags of other variables Granger cause the respective dependant variable.In other words, the test shows whether the lags of other variables can be excluded from the respective dependent variable equation without losing relevant information.
For example, in the Δ t CNP equation we can test the joint null hypothesis for the coefficients of ΔLogS lags as 1 1 ... 0 and similarly for ΔLogF lags as 1 1 ... 0 , and we can test for the coefficients of the two variables' lags together as 1 . This test will provide insights on the Granger causality among the variables, as well as the exogeneity/endogeneity of the variables.
For the cases in which we have evidence on carry trading, we evaluate the performance of this trading strategy and investigate the properties of its payoff distribution.For every trading day, we calculate the payoff of all positions taken.The dailyaggregated payoff    t in terms of the quote currency is then computed as follows: where Pos stands for long and short positions where long positions take a positive sign and short positions take a negative sign,  t k S is the spot rate at the position's maturity and F is the position's forward rate.

USD Forward Positions and Forward Premium/Discount
Recall that in the FX forward market, the carry trading condition implies taking short positions in (i.e.selling forward) the currency that is at a forward premium and taking long positions in (i.e.buying forward) the currency that is  at a forward discount.Therefore, the conjecture is that, based on our cointegration analysis for every currency pair, finding (or not) a long-run equilibrium relationship meeting this condition would provide direct evidence on the position-takers' behavior with respect to USD carry trading over the sample period of the paper.
Tables 2 and 3 report the results of the two specifications of the ADF and PP stationarity tests for emerging-and developed-market currencies, respectively.The results reveal that the variables in level exhibit a unit root, but not in their first differences, i.e. they are integrated of order one.
These results enable us to proceed to test for the existence and the number of cointegrated relationships among variables using Johansen cointegration tests.LogF , except for the GBP/USD case, where both statistics indicate two cointegrated vectors.
In Panel A of Table 5 we report the variables' cointegration coefficients normalized on CNP4 and in Panel B we report the results of a significance test on the coefficients.For the emerging-market currencies, the three variables have significant cointegration coefficients at the 5% significance level.For the developed-market currencies, and LogS LogF have significant cointegration coefficients at the 1% sig- nificance level, but results vary with respect to CNP .In the JPY/USD, CHF/USD and SEK/USD equations, CNP's coefficient is significant at the 1% level, whereas it is significant only at the 10% level in the EUR/USD case and insignificant in the GBP/USD and CAD/USD cases.A cointegration relationship between the spot and forward rates is normally expected.So, the point to be noticed here is the significant cointegration coefficient of the CNP for most of the cases.
In order to illustrate what the resulting cointegration equations imply for USD carry trading, we take the INR/USD case as an example.where Log(S/F) is the forward premium/discount on the base currency.Apart from the scale differences, the positive sign of the forward premium/discount coefficient is particularly interesting.The cointegration equation above implies that an increase in Log(S/F)-which means that the base currency is at a discount in the forward market-is associated with an increase in CNP (i.e. more long positions in the base currency) and vice versa.Table 6 reports the cointegration equation for every currency pair.Note that the same relationship, as described above, holds for all emerging-market currencies, but it is the opposite for developed-market currencies.For emerging-market currencies, the relationships between the USD forward positions and forward premium are in line with the carry trading condition.As the US dollar is the lower-yielding is estimated for the levels of the three variables and then the optimal number of lags is evaluated based on information criteria including the sequential modified LR test statistic, final prediction error, Akaike information criterion, Schwarz information criterion and the Hannan-Quinn information criterion.Finally, the optimal lag length is chosen where the majority of the information criteria indicate the same number of lags.This also holds for the cointegration results in other tables.currency against these emerging-market currencies over our sample period, i.e. it is a forward premium-quoted currency, these long-run equilibrium relationships imply that the record-low US interest rate was being exploited against the higher rates of these emerging markets through USD carry In order to demonstrate this point and further illustrate the direction of carry trading in these positions, Figure 3 depicts the CNP for every emerging-market currency along with the corresponding spot exchange rate.Over our sample period, it is clear that USD is almost consistently on the short side of the position.Most importantly, USD carry trading in these positions appears obvious upon observing the behavior of the CNP, where it is typically on a downward trend.This is consistent with the US dollar being the funding currency.We also note that the downward trend of the CNP is associated with periods over which the US dollar exhibits a depreciation trend.On the other hand, a change in the CNP's behavior can be noticed for some currencies approximately after mid-2011, the period in which we can note an appreciation trend in the US dollar.This change in the CNP trend can imply a change in the trading strategy and thus reflect the actions of position takers to reduce carry trade losses resulting from the appreciation of the funding currency (USD) even though these actions could contribute to further funding currency appreciation.
For developed-market currencies, on the other hand, the cointegration equations imply patterns which are opposite to the carry trading condition, i.e. the forward discount-quoted currency is associated with more short positions and vice versa.This trading style is indeed consistent with the uncovered interest rate parity condition which implies that a low-interest-rate currency should appreciate against a highinterest-rate currency.Although the equations for the GBP/USD and CAD/USD cases appear to satisfy the carry trading condition, the CNP variable in these two cases has insignificant cointegration coefficients as shown in Panel B of Table 5.In other words, the resulting cointegration equations for these two cases come only from the LogS and LogF variables.
In accordance with the literature on the currency carry trading determinants, we argue that the reasons behind the anti-carry USD/developed-market currency positions are the increased uncertainty which followed the recent financial crisis in these markets and the narrow interest-rate differentials during the vast majority of the sample period, especially from late-2008 onwards.The reasoning behind our argument comes from the dependence of currency carry trading payoffs on exchangerate volatility as well as interest-rate differentials (see, for example, Menkhoff et al., 2012;Clarida et al., 2009;Coudert and Mignon, 2013;Spronk et al., 2013;Hoffmann, 2012;Gilmore and Hayashi, 2011;Brière and Drut, 2009).Specifically, with a low interest-rate differential, a slight adverse exchange-rate movement would wipe out any gains from the interest-rate differential.But with a relatively high interest-rate differential, the exchange rate needs a large adverse movement to cancel out the interest-rate differential, creating space for profitable currency carry trades even with minor adverse movements in the exchange rate.In this sense, it is not enough for a currency to be at a forward premium or discount in order for position takers to decide to engage in currency carry trades, as the magnitude of the forward premium/discount is also important.Notes: The table reports the OLS estimation results for the regression of net positions, NP, (in millions of US dollars) against the lagged carry, Carry-1, (lagged forward premium), lagged spot return, ∆s-1, lagged net position, NP-1, second lagged carry, Carry-2.t-statistic in parentheses.Menkhoff et al. (2016) study the relationship between different FX end-user (spot) order flows and several lagged explanatory variables with the aim of identifying the trading styles implied by these flows.They find different trading strategies among their customer groups.For example, asset managers are found to be trading against the interest-rate differential and are best characterized as trend followers.Surprisingly, the authors find that carry trading is not the dominant trading style for hedge funds and that corporate customers trade with the interest-rate differential.Their final customer group of private clients are found to be best described as contrarians.
In order to shed more light on the drivers of our net positions and the direction they take, we run the regression the net positions for every currency pair against the lagged forward premium (carry), lagged spot return and the lagged net position itself.We perform two regression specifications: in the first specification, the first lagged carry is the lone explanatory variable; in the second specification, the explanatory variables include the first and second lagged carry, lagged spot returns and lagged net positions.The estimation results are reported in Table 7.
For emerging-market currencies, the results of the first specification show that unlike MYR, which has insignificant positive coefficients on the lagged carry, and MXN, which has zero coefficient, BRL, ZAR, INR and CLP have negative coefficients.The coefficient for BRL and ZAR is significant at the 5% significance level, while for INR it is significant at about the 15% significance level.These negative coefficients mean that the net position is related to the lagged carry through a pattern of carry trading.Note that the strongest carry relation is for those currencies which have the highest interest rates.The results of the second specification show that the net position is significantly and positively related to the lagged spot return (for the six currency pairs) and to its own lagged net position (all currencies except MXN).The positive coefficients on the lagged spot return are consistent with trendfollowing trading.However, even after accounting for these additional variables, the coefficient on the first lagged carry remained significantly negative for BRL and ZAR, and for MYR the second lagged carry has a significantly negative coefficient.
For developed-market currencies, the lagged carry coefficient in the first specification is positive for all cases and it is significant for EUR, JPY, GBP and CHF.In the second specification, where the lagged spot return and lagged net position are accounted for, the lagged carry coefficient is significantly positive for the six currencies except SEK.These positive coefficients indicate an anti-carry trading pattern which is specifically consistent with trading on the uncovered interest parity condition.The estimation results of the second specification also show that the lagged spot return has a positive coefficient for all cases except GBP and it is significant for JPY, SEK and CAD.In addition, the lagged net position has significantly negative coefficients in the cases of EUR and CHF, but significantly positive coefficients in the cases of GBP, SEK and CAD.
Overall, these results show that even after accounting for the trend-following aspect of the positions, the carry remained influential and its effects are in different directions; while the carry effect tends to be in line with carry trading for emergingmarket currencies, the effect is in line with "fundamentals-based" trading, i.e.UIP, for developed-market currencies.These different trading strategies for emerging-and Notes: VECM is estimated as described by equations ( 3)-( 5) in the methodology section, t-statistics in brackets.
Boldface denotes significance minimally at the 10% level.
developed-market currencies give rise to the trend followers' trading behavior as introduced in the heterogeneous agents model of Spronk et al. (2013).The model consists of fundamentalists, who may form their expectations according to the UH; carry traders, who trade against the UH predictions; and chartists, who simply follow the dominant trend in the market.In our analysis for emerging-market currencieswhere USD carry trading is very attractive-carry traders dominate the market trend.Consequently, CTAs, which are mainly trend followers, behave just like carry traders.On the other hand, for the currencies of developed countries, carry trading is unattractive, so the market is dominated by the fundamentalists who trade in accordance with the UH predictions.Thus, CTAs' USD dollar positions against these currencies once again follow the dominant, but very different, trend.Given these findings, the analysis will henceforth focus on emerging-market currencies where we have clear evidence of currency carry trade activities.
In order to evaluate the long-run relationships and investigate the direction of causality between our variables, we estimate the VECM and unrestricted VAR model as specified in the methodology section-see equations ( 3)-( 5).
Table 8 reports the variables' adjustment coefficients.In the cases of INR/USD, BRL/USD, MYR/USD and MXN/USD, the three variables all have significant adjustment coefficients.This means that in the short run they respond significantly to the departure from the reported long-run equilibrium relationships.In the cases of ZAR/USD and CLP/USD, only the ΔLogS variable has significant coefficients.
Table 9 reports the results of the VAR Granger Causality/Block Exogeneity Wald test as described in the methodology section.This framework enables us to assess whether cumulative net forward positions could be driving the spot and forward rates in a way that could counter the strategy.For all currency pairs, the lags of the ΔCNP variable are jointly insignificant in the ΔLogF and ΔLogS equations, and  thus could be excluded from the equations without losing relevant information.Also, these results suggest that ΔCNP does not Granger cause ΔLogF or ΔLogS .On the other hand, for the equation of the ΔCNP variable we note that, at least, either the lags of the ΔLogF variable are jointly significant or the lags of the two variables taken together are jointly significant; "Both" cannot be excluded the ΔCNP equation.
One exception is the BRL/USD case, where neither the lags of the two variables taken separately nor taken together are jointly significant.The results of this test shed light on the endogeneity of the CNP variable, at least in the short run.In addition, the lack of any adverse casual effect from forward shorting to the forward rate implies that the strategy pays off on average.

Currency Carry Trading Payoffs
Several studies investigate the performance and properties of currency carry trade returns.These studies are mainly based on synthetic carry positions.Generally, they create carry trade portfolios by sorting currencies periodically according to the forward premium/discount and then they assume short positions in lower-yielding currencies and long positions in higher-yielding currencies.The resulting payoffs from these positions are found to be, on average, profitable with relatively high Sharpe ratios and they have high kurtosis and negative skewness (see, for example, Burnside et al., 2006Burnside et al., , 2008Burnside et al., , 2011;;Menkhoff et al., 2012).In this section, we evaluate and explore the characteristics of payoffs generated from the carry positions against the emerging-market currencies.
We calculate the daily-aggregated payoff of our actual USD positions against every emerging-market currency as in equation ( 6).The properties of the calculated payoffs are then compared to the results of the earlier studies that rely on synthetic trading.We then have six currency-specific carry trades.In Table 10 we report the main descriptive statistics of the payoffs for each currency pair.For comparison purposes, we also report the same statistics for excess returns of a value-weighted portfolio of US stocks obtained from Kenneth French's website.
For all currency pairs, except for the case of ZAR/USD, the Sharpe ratios are positive.The INR/USD and MYR/USD pairs beat the US stock market based on the Sharpe ratio.Carry trading payoffs exhibit positive kurtosis and negative skewness.Moreover, in all cases, except for MYR/USD, the Sortino ratios are lower than that of the US stock market, indicating that they are more subject to large losses.Jarque-Bera p-values indicate that distributions are far from being normal.
The negative skewness of currency carry trading payoffs reflects a higher likelihood of large negative outcomes.This results from the tendency of target currencies to occasionally depreciate against the funding currency, which in turn results in large occasional negative profits.This payoff behaviour implies that currency carry trades are subject to so-called downside risk.Gyntelberg andRemolona (2007) andBrunnermeier et al. (2008) also find that currency carry trades are exposed to crash risk.
Our results of positive Sharpe ratios, high kurtosis and negative skewness are consistent with many studies which investigate the properties of currency carry trading payoffs, and confirm the common description of currency carry trading payoffs as being "picking up pennies in front of a truck".

Conclusion
Currency trading strategies in general and carry trades in particular have been of interest to many researchers.In this paper, we make use of a unique dataset consisting of daily-aggregated forward positions in the US dollar against several emerging-and developed-market currencies.The dataset is collected from RPM, a specialist investment manager based in Stockholm, Sweden.Our aim is to investigate whether these positions exhibit a carry trading behavior in which the US dollar represents the funding currency in response to the low US interest rates or if other trading styles can be identified.
By applying Johansen multivariate cointegration analysis, we find long-run equilibrium relationships between the USD/emerging-market currency forward positions and the forward premium/discount meeting the carry trading condition, which involves going short in currencies that are at a forward premium and going long in currencies that are at a forward discount.Furthermore, the USD forward positions against those emerging market currencies exhibit short position trends, implying a carry trade direction in which the USD represents the funding currency.
On the other hand, the cointegrated relationships for the developed market currencies exhibit a completely different pattern.Contrary to the carry trading condition, the relationships for these market currencies imply that a higher-yielding currency is associated with more short positions and vice versa.This pattern is indeed a "fundamentals-based" trading style which is line with the condition of the uncovered interest parity.We argue that the simultaneous low interest-rate differentials between the developed-market currencies and US dollar along with the increased uncertainty over the period after the recent financial crisis for these markets are possible reasons for this anti-carry pattern.
In sum, these findings suggest that over the recent period of ultra-loose US monetary policy, FX traders could have the tendency to engage in USD carry trading against emerging-market currencies but not against developed-market currencies, where they are found to follow a completely different trading style.The findings also provide more direct evidence on carry trades by explicitly relating forward positions to the carry.One of the most important implications of currency carry trades is their effect on the stability of the FX market, especially in times of unwinding carry positions.So, tracking carry trades is important because it can provide us with a better understanding of the FX market volatility dynamics.Moreover, tracking carry trades is relevant to potential target countries that usually tend to take actions against the undesirable influences of such speculative activities.These actions may include capital flow restrictions, taxation of foreign investments and limitations on foreign holdings.In this sense, evaluating and examining the alternative measures which can be taken along with their effectiveness can be the focus of future research.
Finally, we evaluate the realized payoffs of the actual USD positions against every emerging-market currency.We find that five out of six currency pairs yield positive Sharpe ratios, with two cases-INR/USD and MYR/USD-producing Sharpe ratios larger than the Sharpe ratio of a value-weighted portfolio of US stocks.In addition, the payoff distributions for all cases exhibit high kurtosis and negative skewness, similarly to previous studies that employed synthetic currency carry trade positions.

Figure 3
Figure 3 Spot Exchange Rate and Corresponding CNP

Table 2
Stationarity Tests for Emerging Markets This table reports the p-values of Augmented Dickey-Fuller, ADF and Phillips-Perron, and PP tests.
Con is constant.The p-values are based on MacKinnon (1996).

Table 3
Stationarity Test for Developed Markets This table reports the p-values of Augmented Dickey-Fuller, ADF and Phillips-Perron, and PP tests.
Con is constant.The p-values are based on MacKinnon (1996).
Table 4 reports the results of trace statistic and maximum eigenvalue tests.Both test statistics indicate the existence of only one cointegrated equation between CNP, and LogS

Table 4
Johansen Cointegration Test Results MacKinnon-Haug-Michelis (1999)s the null hypothesis that there is r cointegrating vectors against the alternative that there is generally more than r cointegrating vectors.While in the maximum eigenvalue test the null hypothesis is that there is r against the alternative r + 1 cointegrating vectors.CE is cointegrating equation.***denotessignificanceatthe 1% level.p-valuesarebasedonMacKinnon-Haug-Michelis(1999).We select the number of lags as follows:a Vector Autoregressive model(VAR)

Table 5
Co-integration Coefficients Normalized on CNP Notes: Panel A: cointegrating coefficients are normalized on CNP, Cons is constant, t-statistics in brackets.Panel B: ***,** and * denote significance at the 1%, 5% and 10% significance levels, respectively.The US dollar is the base currency.

Table 7
Drivers of Net Positions

Table 10
Payoffs are in millions of quote currency (the US dollar is the base currency).US Stock market is the market premium (Mkt -RF) on value weighted portfolio of US stock obtained from French's data library.