Carry trade performance

indi-cating that the conditions for the carry trade to work appear to be in place. It is important to note that the results are based on developed currencies, but in practice, investors are even able to invest in emerging markets countries, where interest rates are high, and thereby obtain higher interest rate differentials.

for all strategies in Table 5.2, which is in line with the findings of Nelly and Weller (2013). The nominal p-values presented in the table is obtained from the two-sided test described in Section 4.6.1. All the performance measures are statistically significant at a 5% significance level, except for Carry1M1C, Carry3M1C, Carry12M1C and Carry12M5C. In line with previous studies, the mean return is economically significant for all strategies. However, only 16 out of the 20 strategies are statistically significant.

Table (5.2) Performance metric of carry trade strategies over the full sample period

This table reports the annualised mean returns, Sharpe ratios and nominal p-values of a two-sided test, of the 20 carry trade strategies obtained in the sample period from 30-06-1999 to 29-05-2020. The results are based on ask-, mid- and bid

prices to account for the impact of transactions cost. The mean returns are marked by a 3-colour scale, using the 50th percentile, in which the highest mean returns hold the darkest colour and the lowest value hold the brightest colour. The

rest of the values are coloured proportionally. The abbreviations used to denote the strategies are described in Appendix A.1.

1M 3M 6M 12M

1C

Mean return 2.77% 2.46% 2.97% 2.79%

Sharpe ratio 0.407 0.386 0.445 0.398

Nominalp-value 0.063 0.078 0.043 0.069

2C

Mean return 2.57% 2.70% 2.82% 2.80%

Sharpe ratio 0.491 0.496 0.519 0.518

Nominalp-value 0.025 0.024 0.018 0.018

3C

Mean return 2.58% 2.66% 2.85% 2.84%

Sharpe ratio 0.528 0.534 0.564 0.557

Nominalp-value 0.016 0.015 0.010 0.011

4C

Mean return 2.70% 2.66% 2.79% 2.53%

Sharpe ratio 0.557 0.520 0.553 0.480

Nominalp-value 0.011 0.018 0.012 0.029

5C

Mean return 2.71% 2.46% 2.61% 1.94%

Sharpe ratio 0.551 0.470 0.500 0.353

Nominalp-value 0.012 0.032 0.023 0.742

Sub-sample periods

Even though most of the strategies are profitable over the full sample pe-riod, these are not necessarily profitable in every sub-sample period. This is clear from Figure 5.2-5.3 in Section , illustrating how the cumulative re-turn decreased substantially during the financial crisis. In relation to investor behaviour, it is likely that most investors would not tolerate a carry trade generating severe losses as those observed during the financial crisis. As a consequence, they would not stick to carry trade after longer periods of losses, although the strategy appears to be profitable in the long run.

Table (5.3) Performance metric of carry trade strategies over the three sample periods

This table reports the mean returns, Sharpe ratios and nominal p-values of the 20 carry trade strategies. The results are based on the carry returns obtained in the sub-sample period from 30-06-1999 to 30-06-2006, 30-06-2006 to 28-06-2013 and 28-06-2013 to 29-05-2020 in which transactions cost are taken into account. The

abbreviations used to denote the strategies are described in Appendix. A.1

Sub periods 1999-2006 2006-2013 2013-2020

1M 3M 6M 12M 1M 3M 6M 12M 1M 3M 6M 12M

1C

Mean return 5.40% 5.34% 6.41% 5.89% 1.41% 0.64% 0.79% 0.70% 0.91% 0.91% 1.30% 1.28%

Sharpe ratio 0.901 0.937 0.981 0.972 0.168 0.082 0.101 0.078 0.158 0.167 0.237 0.233 Nominalp-value 0.019 0.015 0.011 0.011 0.656 0.827 0.788 0.836 0.677 0.660 0.533 0.539 2C

Mean return 4.18% 4.98% 5.49% 5.17% 1.74% 1.38% 0.98% 1.35% 1.33% 1.27% 1.54% 1.43%

Sharpe ratio 0.926 1.176 1.148 1.136 0.260 0.199 0.145 0.195 0.321 0.268 0.348 0.328 Nominalp-value 0.016 0.002 0.003 0.003 0.491 0.597 0.701 0.604 0.399 0.481 0.360 0.388 3C

Mean return 5.14% 5.89% 6.45% 6.17% 0.90% 0.09% 0.26% 0.41% 1.29% 1.59% 1.41% 1.46%

Sharpe ratio 1.117 1.396 1.411 1.369 0.149 0.015 0.042 0.063 0.343 0.386 0.366 0.391 Nominalp-value 0.004 0.000 0.000 0.000 0.693 0.968 0.911 0.867 0.367 0.310 0.336 0.304 4C

Mean return 4.18% 6.64% 6.69% 6.91% 1.12% -0.15% -0.21% -1.12% 0.89% 1.09% 1.50% 1.40%

Sharpe ratio 1.210 1.438 1.414 1.402 0.194 - 0.023 - 0.034 - 0.170 0.233 0.287 0.413 0.385 Nominalp-value 0.002 0.000 0.000 0.000 0.607 0.951 0.928 0.652 0.539 0.450 0.277 0.311 5C

Mean return 5.40% 6.82% 7.54% 7.00% 0.78% -0.85% -1.22% -2.33% 0.93% 1.09% 1.17% 0.84%

Sharpe ratio 1.240 1.443 1.548 1.359 0.133 -0.129 -0.189 -0.339 0.249 0.293 0.323 0.226 Nominalp-value 0.001 0.000 0.000 0.001 0.723 0.733 0.615 0.370 0.512 0.440 0.395 0.552

In order to evaluate the performance of the strategy over time, the sample period is decomposed into three consecutive periods of seven years, namely:

1999-2006, 2006-2013 and 2013-2020. The performance measures of the 20 strategies in each of the three sub-periods are presented in Table 5.3.

1999-2006

Table 5.3 indicates a clear difference in the performance of carry trade between the three sub-sample periods. Evaluating the sub-sample period 1999-2006, the performance of the carry trades were substantially better compared to the two subsequent periods. The mean return and Sharpe ratio are statistically significant for all 20 strategies at a 5% significance level. Although there is no clear pattern between performance and frequency of rebalancing, all 20 strategies generated economically high returns between 4.18% and 7.54%.

Furthermore, the Sharpe ratios are remarkably high even for carry trades, with the most profound being Carry6M5C with 1.548. Hence, the conditions for the carry trade to work appeared to be in place during this period, meaning low volatility and high interest rate differentials of the currencies.

2006-2013

Evaluating the performance of the sub-sample periods going from 2006 to 2013, all of the performance measures are statistically insignificant at a 5%

significance level. Looking more closely, the more diversified and less frequent rebalanced portfolios, such as Carry6M5C and Carry12M5C, exhibit negative mean returns during the sub-sample period. During this period, including the financial crisis, massive changes in the interest rate environment occurred. This led to substantial changes in the optimal high and low portfolios. For instance, the interest rate differential between GBP and USD went from being positive to negative and should optimally be reallocated from the high portfolio to the low portfolio around the Lehman collapse in September 2008. Consequently, investors who rebalance their portfolios with 3-, 6- and 12-month frequency, would not have been able to reallocate the currency pairs accordingly. In context to real-life portfolio management, most investors would presumably trade actively in such periods to prevent losses such as the ones observed.

2013-2020

In the sub-sample period going from 2013 to 2020, all of the mean returns are positive. However, these are not statically significant, similarly to the previous period. In Figure 3.1 Section 3, the interest rate differentials over the period 2013 to 2020 declined rapidly towards zero. Around 2014, central banks in developed countries, lowered their interest rates to support their economies, leading to an undesirable change in the interest environment. The lower re-turns, although positive, could therefore be due to the equivalent decrease in

the interest buffer as this normally protects against adverse currency move-ments.

Summarising Table 5.3, the carry trades performed excellently in the first sub-period and rather poorly in the two subsequent sub-periods. The poor performance of the two subsequent sub-periods gives the premise to implement the alterna-tive strategies and possibly to improve the return in periods where carry trade undoubtedly underperforms.

Carry trade with no transaction costs

Among previous literature, Swinkels and Egbers (2015), include transaction cost in order to estimate the actual realised excess return. However, Darvas (2009) found that rollover costs are typically small for the G10 currencies while Briere and Drut (2009) argues that transaction costs are so low, it is not necessary to take it into account. Hence, the actual difference in the performance measure of the actual realised excess return and the carry trade without transaction cost is investigated in this section. The pay-off of carry trade without transaction cost is based on the assumption that investors can buy and sell currencies at the average of the bid and ask price. In Table 5.4, the performance metric of the full sample period is shown for estimates based on mid prices. Excluding transaction cost, does not seem to have any particular impact on the mean returns. In fact, the mean returns are still around 2-3%, and the Shape ratios lie between 0.356 and 0.581. The Sharpe ratio seems to increase going from less diversified portfolios, such as Carry1M1C, to more diversified portfolios such as Carry1M5C. All performance measures are statistically significant, except for Carry1M1C, Carry3M1C, Carry12M1C and Carry12M5C as observed in Table 5.2. The performance of carry trade over the three sub-sample periods 1999-2006, 2006-2013 and 2013-2020 is found in Appendix B.2. The carry trade performance after excluding the transaction cost displays the same pattern in mean returns and is only slightly larger compared to the excess return, including transaction costs. Due to this small impact of including transactions cost, these will be excluded going forward.

Hence, the analysis and model implementation of the dynamic strategies are based solely on mid prices.

Table (5.4) Performance metric of carry trade strategies over the full sample period (no transaction costs)

The table reports the mean returns, Sharpe ratios and nominal p-values of the 20 carry trade strategies obtained in the sample period from 30-06-1999 to 29-05-2020.

The results are based on the mid prices, to exclude the impact of transactions cost.

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

1M 3M 6M 12M

1C

Mean return 2.84% 2.49% 2.98% 2.79%

Sharpe ratio 0.417 0.389 0.446 0.399

Nominalp-value 0.057 0.076 0.042 0.069

2C

Mean return 2.60% 2.71% 2.82% 2.80%

Sharpe ratio 0.498 0.499 0.520 0.518

Nominalp-value 0.023 0.023 0.018 0.018

3C

Mean return 2.67% 2.70% 2.86% 2.84%

Sharpe ratio 0.546 0.540 0.567 0.559

Nominalp-value 0.013 0.014 0.010 0.011

4C

Mean return 2.82% 2.70% 2.81% 2.54%

Sharpe ratio 0.581 0.529 0.557 0.482

Nominalp-value 0.008 0.016 0.011 0.028

5C

Mean return 2.84% 2.51% 2.63% 1.96%

Sharpe ratio 0.577 0.479 0.505 0.356

Nominalp-value 0.009 0.029 0.021 0.104

In document Can traders benefit from the currency carry trade during crisis periods? (Sider 43-49)