funds to safe investments, often referred to as quality” or ”flight-to-liquidity”, ultimately leading to the poor performance of carry trades. The purpose of the dynamic strategies is therefore to avoid the effects of unwind-ing carry trades either by closunwind-ing the positions or takunwind-ing the reverse positions in an attempt to catch the momentum in which funds are flowing out of the high-interest and into safer currencies. As described in Section 4.4, the timing signals in this thesis are based on all available information up until the day of the signal. Using the contemporaneous value, this would require that traders know in advance what the implied volatility will be at the end of a trading day. For that reason, a regression analysis is performed for the carry trade returns on the change VIX the previous month one day lagged. The findings are reported in Table 5.5.
Table (5.5) Carry trade returns regressed on changes in VIX and VXY
This table reports the results of the regression of the next month-end carry trade returns with the change in VIXt−1 and VXYt−1. The results are based on data available in the sample period from 30-06-1999 to 29-05-2020. The reported figures
include the coefficient, standard error,z-score and associatedp-values as well as the 95% confidence interval.
Coefficient Standard error z p > |z| [0.025 0.975]
∆VIXt−1 -0.0011 0.000 -4.752 0.000 -0.002 -0.001
∆VXYt−1 -0.0065 0.000 -6.084 0.000 -0.009 -0.004
The findings of the table, show a significant negative relationship for predicting the next months carry returns. Taken together, VIXt−1 and VXYt−1 appear to be good indicators of volatile periods in which carry trades generate negative returns. In the following section these periods are identified using the threshold defined in Section 4.3.1.
5.2.2 Crisis periods
Figure 5.5 presents the events identified by the VIX and VXY signals, with a window of 253×1.5 and standard deviation of 1.5 over one day lagged sample periods between 29-06-1999 and 28-05-2020. During the period, the two indices signalled a crisis 14 times on the same date, which is approximately half of the identified crisis periods. The identified crisis periods are described in further detail in order to apprehend which periods that are defined as crisis periods in the exit- and reverse strategies. It is not the ambition to test whether the
signals are actually caused by a specific event, but rather identify and label the crisis periods on a granular level. Each period is therefore labelled by the known crisis that occurred over the sample period, even if the peak that led to a signal might have been caused by an event independent of the respective crisis.
Figure (5.5) Identified crisis periods
This figure presents the crisis periods identified by the VIX and VXY signal in the sample period from 30-06-1999 to 29-05-2020. The largest peaks in VIX are labelled by the known crisis period which occurred during the same period.
The first crisis periods identified by the VIX signal between 2001 and 2002, consist of three peaks reflecting the aftermath of the dot-com bubble crisis due to excessive speculation in the internet-related companies in the late 1990s.
Neither of these events was caught by the VXY signal. Instead, VXY identi-fied six other crisis periods between 2003 and 2006.
At the beginning of 2007, VIX was at its 10-year low level but started to in-crease steadily throughout the year. After the spike in the summer of 2007, the crisis started to escalate into a financial meltdown, culminating in the collapse of the Lehman Brothers in September 2008. In response to the extreme mar-ket disturbance after the collapse of the Lehman Brothers, the central banks, including the European Central Bank and Bank of Japan, started to provide direct lending to banks at a low or near-zero interest rate (Bauer and Neely,
2014). Both VXY and VIX identified fairly many crisis periods during the financial crisis.
Another crisis identified by the VIX and VXY signals is the fallout of the financial crisis of 2007-2008, the Greek government-debt crisis. The Greek crisis originated as early as late 2009. However, the period identified by the signals presumably reflects the poor performance of the Greek government in achieving conditions that were agreed upon for a bailout, consequently forcing them into a second bailout in 2011.
The next crisis period reflects the events of Brexit. Although the actual refer-endum took place in June 2016, both signals identified a crises period in 2015, possibly reflecting the events leading up the European Union Referendum Act (Walker, 2020).
Two years later, in 2018, the president, Donald Trump, imposed new tariffs on imported and exported products coming in and out of China. In the following months, this lead to the trade war between the USA and China, the events of which are reflected by the VIX and VXY signals. As many other countries rely on China being the worlds largest manufacturing hub, the ongoing trade war did affect a number of other currencies, besides the USD and CNY (Chinese Yuan Renminbi) (Swanson, 2020).
The most recent crisis period, identified by both signals, reflect the global recession, which is a consequence of the ongoing COVID-19 pandemic. The crisis period defined by the VIX signal, starts from February 2020, in line with the stock market crash that took place on 20 February 2020, where the first signs of recession started. However, the VXY signal did not identify the cri-sis period until the month after, in March 2020. The International Monetary Fund (IMF) projects the COVID-19 recession to become the most severe global economic downturn since the Great Depression during the 1930s (Gopinath, 2020). In general, it seems that both signals were able to identify the global economic crises, but there might have been particularly important periods, which were only caught one of the signals.