2. Literature Reviews
2.2 Carry trade
國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
16
against pound and receives buy order, he will charge higher quote price for the buy order. So the forward premium is high when the US dollar appreciates. This phenomenon also calls averse-selection. Averse-selection can give one explanation of the existence of forward premium puzzle.
Clarida, Davis and Pedersen (2009) propose that the well-known results of negative slope coefficient in Fama regression may not hold under different volatility regime. The authors divide the sample periods into two volatility regime. One is high-volatility regime and the other is low-volatility regime. They use 9 currencies to test whether the slope coefficient of Fama regression is positive or negative. The results show that the slope coefficient is negative in low-volatility regime. However, the slope coefficient is positive in high-volatility regime although the slope coefficient is negative for the entire sample period. Furthermore, the slope coefficient under high-volatility regime is larger than 1. That is, the low-interest-rate currency is prone to appreciate more than implied by interest rate difference. So, the authors conclude that the volatility-regime may give one explanation of the forward premium puzzle.
2.2 Carry trade
Darrvas (2009) proposes that previous papers confirm that leverage will increase the exchange rate risk which is the major reason to drag down the return of carry trade.
The author considers the leverage effect in the carry trade positions which is different from previous literatures that assume non-leverage carry trade positions. The author use not only common base currency such as us dollar and British pound and also another 9 major currencies to act as base currency. The empirical results suggest that the leverage negatively affects skewness and exhibits U-shaped relationship with the return. That is, when leverage increases to certain level, return continues to fall. The
‧
leverage, excess returns tend to present. However, when positions consider leverage, the excess returns vanish. So the author argues that leverage may act as one explanation that UIP theory fails.Dunis and Miao (2007) compare three trading strategies (Benchmark MACD model, Carry model and Combined & MACD model) by using annual return, information ratio and maximum drawdown to determine which model performs best in both in-sample period and out-of-sample period. The empirical results show that carry model has the highest annual return and Sharpe ratio among three models. Then, the authors add the volatility filters (“no trade” filter and reverse filter) to three models. The results suggest that the combined carry/ MACD model has the lowest maximum drawdown among three models. They also find that strategy equipped with volatility filter has lower maximum drawdown than strategy without volatility filter.
So they conclude that adding volatility filter to carry trade strategy can reduce volatility of carry trade’s return significantly.
Becker and Clifton (2007) extend previous literatures to find the relationship between hedge fund activities and carry trade strategy. However, due to limited data of hedge fund returns and related activities, former papers usually use indirect data to draw statistical inference. The authors use the banking data from BIS to proxy for hedge fund data. They use Japanese yen and Swiss Franc to act as funding currency and use the 10-year government bond yields of Australia and New Zealand to calculate the profits hedge fund may earn. Due to large amounts of money to implement carry trade, capitals of hedge fund flow into high-yield currencies usually cause exchange rate to appreciate. However, when the exchange rate volatility increases, hedge funds tend to unwind their positions which may cause turmoil in foreign exchange markets.
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
18
Clarida, Davis and Pedersen (2009) use the G10 currencies to examine the relationship between return and risk of the carry trade positions. They calculate the realized volatility by using exponentially weighted moving averages (EWMA). They define the low-volatility regime if the realized volatility is less than 25th percentile for the entire sample period and the high-volatility regime if the realized volatility exceeds 75th percentile for the entire sample period. The results show that returns are higher under low-volatility regime than those under high-volatility regime. The authors also use the Kernel regressions, a non-parametric method, to do the robustness check. The findings are consistent.
Brunnermeier, Nagel and Pedersen (2009) argue that high-interest rate currencies are subject to currency crash. That is, high-interest rate currencies tend to appreciate in short notice. Although failure of uncovered interest rate parity let carry trade strategy earn excess return on average, high-interest rate currencies depreciate abruptly could have negative impact on carry trade positions. The authors propose that high-interest rate currencies such as AUD and NZD tend to have conditional negative skewness. This paper knows that under the statistical properties, asset returns with negative skewness are prone to have extremely negative returns. So high-interest tare currencies with negative skewness are consistent with our observation in FX market. Then, the authors use the simple regressions to examine the relationship between quarterly carry trade return, interest rate difference mince change in log return on FX, and the interest rate differential. They find that the slope coefficient of regressions is positive. This implies that positive returns could be predicted by high interest rate differences. The authors also run the OLS regressions to test the relationship between futures positions held by speculators and interest rate differential.
The result shows that speculators such as hedge funds tend to hold large long
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
19
positions in high-interest rate currencies. This observation also could act as another explanation that UIP theory doesn’t hold.
Fong (2013) uses the data from Commodity Futures Trading Commissions (CFTC) to examine whether hedge funds have significant impact on FX markets or not. The preliminary analysis of this paper suggests that hedge funds have net positive positions in high-interest-rate currencies such as AUD and NZD and have net negative positions in low-interest-rate currencies such as EUR and USD. This observation is consistent with what this paper observed in the FX market. That is, the hedge funds usually try to exploit UIP failure to make profits. Then, the author use both OLS regression and structural VAR model to examine whether hedge funds follow momentum strategy. To state differently, this paper wants to examine whether hedge funds exhibit trend chasing behavior or not. The empirical results are consistent with previous papers that hedge funds indeed follow momentum strategy. Finally, the main contribution of this paper is to examine whether hedge funds activities will impact FX market or not. The author uses Quantile regression to divide the data into low quantiles (represent normal market) and high quantiles (represent turbulent market) and test the relationship between future FX returns and hedge fund net positions.
Under turbulent market, the coefficients of the hedge funds net positions are negative among all six currencies. In particular, the coefficients of AUD and NZD in terms of absolute value are greater than those of other currencies. Under normal market, the coefficients of the hedge funds net positions are positive and mostly insignificant among all six currencies. These two results imply that hedge fund activities have a reversal effect on FX market under turbulent market and virtually no impact under normal market.
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
20