5.3 Crisis Robust Carry Trade
5.3.1 VIX signal
In this section, the performance measure of the two dynamic strategies based on VIX is evaluated. Similar to the static carry trade strategies, the full sam-ple period and the three sub-periods are investigated to determine the overall performance of the strategies.
Summarising the findings of this section, the exit- and reverse strategies im-prove upon the 1-month static carry trades, shown by the economically higher performance measures and statistically significant mean returns. Contrary to this, the mean return of the 3-, 6- and 12-month rebalance strategies are all insignificant and economically lower. The impact of the reverse strategies is twice the exit strategies, as the opposite positions are taken instead of unwind-ing them durunwind-ing an identified crisis.
Similar to findings of the previous section, the returns of the dynamic strategies vary greatly across sub-periods, with the first sub-period, on average, leading to the highest mean returns. Employing the alternative strategies, have little or no effect on the strategies performance of this period. During 2006-2013, the mean returns of the 1-month rebalance strategies are statistically signifi-cant and improved compared to the static carry trade. This might be due to a higher potential for improvement as a result of the financial crisis, and in par-ticular, the Lehman collapse, which led to large carry trade losses. During the last period, 2013-2020, VIX reached its highest level historically. However, the performance measures of the exit- and reverse strategies remained insignificant and economically small.
Full sample period
Table 5.6 present the performance measures of the dynamic strategies based on VIX over the full sample period. Evaluating the performance of the exit strategy, the risk-return trade-off of the 1-month rebalance strategies improved compared to the static carry trade in Table 5.4. In particular, the mean return of Exit1M1C increased from 2.84% to 4.15%, and the Sharpe ratio increased from 0.417 to 0.727. The same pattern is observed for the reverse strategy, where Reverse1M1C had a mean return of 5.46% and a Sharpe ratio of 0.757.
As the reverse strategy involves taking the reversed position of the carry trade, it follows that the gain from taking this position will be twice the exit strat-egy and vice versa the loss. As can be seen in Table 5.4, only the 1-month rebalance strategies exhibit statistically significant mean returns and Sharpe ratios, at a 5% significance level.
The majority of the strategies rebalanced every 3-, 6- and 12 month produced lower mean returns and Sharpe ratios compared to those of the static carry trade. The static carry trade, Carry3M1C, generated a mean return of 2.49%
whereas Exit3M1C and Reverse3M1C generated a mean return of 1.43% and 0.38% respectively. The poor performance of the 3-, 6- and 12- month rebal-anced strategies, could be due to the assumption, that if the implied volatility signal is above the threshold on the rebalance day, the traders maintain the alternative positions until the next rebalance period. This means that even though the volatility increases or decreases between two rebalance periods, the position is kept until the next rebalance period. Consequently, traders may end up unwinding their carry trade position in the entire rebalance period, even if this period generally were not characterised as a crisis period. In real-ity, it is more likely that traders would have followed the market movements, especially when employing a dynamic strategy. It is therefore unrealistic to assume, that traders in practice, would have kept the 3-, 6- and 12 month rebalance strategies over the entire period. Since the 3-, 6- and 12 month rebalance periods under-performed compared to the static carry trade, this in-dicates that investors should change their position more frequently to account for changes in volatility.
Table (5.6) Performance metric of exit- and reverse carry trade strategies over the full sample period based on the VIX signal The table reports the mean return, Sharpe ratio and nominal p-value of the 20 exit- and reverse carry trade strategies using VXYt−1 as timing signal. The results
are based on end-of-month mid prices available in the period between 30-06-1999 to 29-05-2020. The mean returns are marked by a 3-colour scale, using the 50th
percentile. The abbreviations used to denote the strategies are described in Appendix. A.1
Exit Reverse
1M 3M 6M 12M 1M 3M 6M 12M
1C
Mean return 4.15% 1.43% 2.28% 2.08% 5.46% 0.38% 1.58% 1.37%
Sharpe ratio 0.727 0.245 0.366 0.303 0.817 0.060 0.234 0.194 Nominal p-value 0.001 0.262 0.095 0.166 0.000 0.785 0.284 0.374 2C
Mean return 3.87% 1.97% 2.18% 2.27% 5.13% 1.23% 1.53% 1.75%
Sharpe ratio 0.839 0.390 0.426 0.426 1.014 0.225 0.280 0.321 Nominal p-value 0.000 0.075 0.052 0.052 0.000 0.304 0.200 0.143 3C
Mean return 3.83% 2.22% 2.34% 2.44% 4.99% 1.74% 1.82% 2.03%
Sharpe ratio 0.893 0.479 0.486 0.484 1.051 0.346 0.357 0.397 Nominal p-value 0.000 0.029 0.027 0.027 0.000 0.114 0.103 0.070 4C
Mean return 3.98% 2.32% 2.32% 2.22% 5.14% 1.93% 1.83% 1.90%
Sharpe ratio 0.925 0.492 0.479 0.425 1.096 0.376 0.360 0.360 Nominal p-value 0.000 0.025 0.029 0.053 0.000 0.086 0.101 0.100 5C
Mean return 3.92% 2.27% 2.18% 1.68% 5.01% 2.04% 1.74% 1.39%
Sharpe ratio 0.881 0.475 0.438 0.308 1.047 0.389 0.331 0.253 Nominal p-value 0.000 0.031 0.046 0.160 0.000 0.076 0.130 0.248
Sub-sample periods
In order to further elucidate the performance of the dynamic strategies, these are employed on the same three sub-periods used in Section 4.2 for the static carry trade. The performance measures of the exit- and reverse strategies are reported in Table 5.7 (a) and (b).
1999-2006
Evaluating the first period in Table 5.7, the performance measure of the exit-and reverse strategy are relatively unchanged for all 20 strategies, compared
to those of the static carry trade (see Appendix B.2). During this period, the signal identified three crisis periods, reflecting the aftermath of the dot-com bubble. Though it should be pointed out, this did not have any notable impact on the carry trade. As can be seen in Figure 5.4 the cumulative return of the static carry trade increased, even though the crisis periods were identified by VIX. As for the static carry trade, all the dynamic strategies performed substantially better during this period compared to the other sub-periods, due to the high-interest rate differentials. Even though the performance measures are generally high, the 1-month rebalanced strategies generated a slightly lower mean return compared to those of the static carry trade. While Carry1M1C generated a mean return of 5.54%, this decreased to 4.47% and 3.39% for the Reverse1M1C and Exit1M1C respectively. The carry trade strategies produced positive returns, in the crisis periods identified by VIX, leading to a negative impact on the mean returns after employing the alternative strategies.
During this period, the signal identified three crisis periods, reflecting the aftermath of the dot-com bubble. Though it should be pointed out, this did not have any notable impact on the carry trade as can be seen in 5.7.
2006-2013
The performance of the dynamic strategy improved for all 1-month rebalanc-ing strategies in the period 2006-2013, with the most profound improvement, being Reverse1M1C with a mean return of 8.92% and a Sharpe ratio of 1.111.
In Appendix B.8, the fraction of months with negative returns for each strat-egy is found. The implementation of the reverse stratstrat-egy, Reverse1M1C, led to a reduction of negative returns from 40.0% to 34.1% during this period. Con-trary to the static carry trade, all of the 1-month rebalance strategies, obtain performance measures which are statistically significant at a 5% significance level. This improvement may be explained by the financial crisis and in par-ticular, the Lehman collapse, which led to an extremely volatile period and at the same time a major change in the interest rate environment affecting the carry trade return.
Looking at the 1-month strategies, the more diversified strategies, such as Reverse1M5C, produced lower performance measures compared to the less di-versified strategies like Reverse1M1C. The Reverse1M1C strategy exploits the most extreme interest rate differential, whereas Reverse1M5C takes a position in all currencies. In Table 5.7 it can be seen that Reverse1M1C has a mean return of 8.92% and Reverse1M5C has a mean return of 6.05%. Since carry trade and reverse positions are held at the same point in time for all 1-month
Table (5.7) Sub sample test: VIX
This table reports the mean return, Sharpe ratio and nominal p-value of the 20 exit- and reverse carry trade strategies using the VIXt−1 crisis signal. The results are presented for three different sub-periods, 1999-2006, 2006-2013 and 2013-2020.
The mean returns are marked by a 3-colour scale, using the 50th percentile. The abbreviations used to denote the strategies are described in Appendix A.1.
(a)
Exit
Sub periods 1999-2006 2006-2013 2013-2020
1M 3M 6M 12M 1M 3M 6M 12M 1M 3M 6M 12M
1C
Mean return 4.47% 4.72% 5.88% 6.00% 5.16% -1.02% -0.65% -1.42% 2.19% 0.23% 1.32% 1.28%
Sharpe ratio 0.762 0.858 0.952 0.994 0.781 -0.146 -0.096 -0.166 0.477 0.047 0.240 0.233 Nominalp-value 0.046 0.025 0.013 0.010 0.041 0.699 0.799 0.660 0.210 0.900 0.528 0.539 2C
Mean return 4.33% 4.68% 5.07% 5.23% 4.74% 0.27% -0.46% -0.23% 2.04% 0.58% 1.55% 1.43%
Sharpe ratio 0.996 1.108 1.081 1.153 0.834 0.043 -0.076 -0.034 0.563 0.134 0.350 0.328 Nominalp-value 0.010 0.004 0.005 0.003 0.029 0.909 0.840 0.929 0.140 0.723 0.357 0.388 3C
Mean return 5.49% 5.72% 5.89% 6.23% 3.86% -0.27% -0.66% -0.79% 1.68% 0.84% 1.43% 1.47%
Sharpe ratio 1.261 1.372 1.290 1.384 0.767 -0.047 -0.115 -0.124 0.504 0.224 0.371 0.394 Nominalp-value 0.001 0.000 0.001 0.000 0.044 0.900 0.760 0.742 0.186 0.556 0.329 0.300 4C
Mean return 6.09% 6.65% 6.13% 6.95% 3.92% -0.38% -1.05% -2.05% 1.50% 0.31% 1.53% 1.42%
Sharpe ratio 1.344 1.483 1.307 1.410 0.802 -0.066 -0.183 -0.319 0.454 0.089 0.419 0.389 Nominalp-value 0.001 0.000 0.001 0.000 0.036 0.860 0.628 0.399 0.233 0.814 0.270 0.306 5C
Mean return 6.44% 6.91% 6.95% 6.99% 3.47% -0.85% -1.95% -3.15% 1.45% 0.41% 1.20% 0.85%
Sharpe ratio 1.362 1.528 1.455 1.357 0.691 -0.146 -0.326 -0.471 0.427 0.118 0.330 0.229 Nominalp-value 0.000 0.000 0.000 0.001 0.070 0.699 0.387 0.213 0.262 0.756 0.386 0.546
(b)
Reverse
Sub periods 1999-2006 2006-2013 2013-2020
1M 3M 6M 12M 1M 3M 6M 12M 1M 3M 6M 12M
1C
Mean return 3.39% 4.04% 5.32% 6.11% 8.92% -2.65% -2.08% -3.53% 3.40% -0.49% 1.32% 1.28%
Sharpe ratio 0.554 0.698 0.803 1.012 1.111 -0.344 -0.269 -0.399 0.602 -0.089 0.240 0.233 Nominalp-value 0.144 0.067 0.035 0.009 0.004 0.362 0.476 0.292 0.115 0.814 0.528 0.539 2C
Mean return 4.40% 4.35% 4.64% 5.29% 7.79% -0.82% -1.88% -1.79% 2.68% -0.15% 1.55% 1.43%
Sharpe ratio 0.976 1.012 0.956 1.167 1.237 -0.118 -0.279 -0.259 0.655 -0.030 0.350 0.328 Nominalp-value 0.011 0.009 0.013 0.003 0.001 0.755 0.460 0.493 0.087 0.936 0.357 0.388 3C
Mean return 5.75% 5.52% 5.31% 6.28% 6.76% -0.65% -1.59% -1.99% 1.94% 0.04% 1.43% 1.47%
Sharpe ratio 1.266 1.293 1.132 1.396 1.184 -0.104 -0.253 -0.307 0.519 0.009 0.371 0.394 Nominalp-value 0.001 0.001 0.003 0.000 0.002 0.783 0.502 0.417 0.174 0.981 0.329 0.300 4C
Mean return 6.36% 6.63% 5.56% 6.97% 6.62% -0.64% -1.91% -2.99% 1.97% -0.52% 1.53% 1.42%
Sharpe ratio 1.367 1.432 1.145 1.416 1.215 -0.100 -0.307 -0.458 0.522 -0.137 0.419 0.389 Nominalp-value 0.000 0.000 0.003 0.000 0.002 0.790 0.417 0.227 0.171 0.718 0.270 0.306 5C
Mean return 6.70% 6.95% 6.34% 6.97% 6.05% -0.89% -2.69% -3.98% 1.83% -0.32% 1.20% 0.85%
Sharpe ratio 1.386 1.473 1.265 1.351 1.084 -0.135 -0.421 -0.586 0.494 -0.086 0.330 0.229 Nominalp-value 0.000 0.000 0.001 0.001 0.005 0.721 0.266 0.123 0.195 0.820 0.386 0.546
rebalance strategies, the more extreme interest rate differentials seem to be more profitable. The poor performance of the 3-, 6- and 12 month rebalance strategies, indicate that investors need to actively manage their position by rebalancing more frequently during volatile periods. For the 3-, 6- and 12 month rebalancing strategies, all performance measures decreased, and for the most part, this resulted in negative mean returns. As for the carry trade, the poor results seems to be related to the unrealistic assumption that the carry traders hold on to same positions until the next rebalance period. In reality, they would more likely have unwound their position or taken the reverse carry trade position in order to reduce losses.
2013-2020
Looking at the last sub-period, all performance measures are insignificant at a 5% significance level, similar to those of the static carry trade. As discussed previously, Figure 3.1 showing the evolution of interest differential over the full sample period, displayed lower interest rate differentials in the most recent years, ultimately affecting the profitability of carry trade. Several crises oc-curred during this period, including Brexit, the US-China trade war and finally COVID-19, which led to the highest historical level of VIX. Yet, neither of the identified crisis periods took place on the 6- and 12 month rebalanced day, essentially leading to the same performance measures as the static carry trade.
With the carry returns being generally small during 2013-2020, the negative and positive impact of the mean returns were also inherently small.