Another interesting topic is that there seems to be a relationship between volatility and correlations.
In asset markets, correlations seem to rise with volatility (e.g. when the market crashes, all stocks crash together). There might be various reasons for this observation, I'll provide one example: when participants hold leveraged portfolios, then, when some stocks in the portfolio drop significantly, traders need to sell parts of their portfolio in order to avoid margin calls.
For the Forex market, the volatility/correlation relationship is more ambiguous. One could hypothesize that during bad times, investors flight for safety, and we see a more pronounced negative correlation between safe-haven and "risky" currencies.
There are various methods to examine the relationships described above, including a multivariate GARCH, if I'm not mistaking. However, such econometric models tend to get rather sophisticated.
Since there are at least a handful of people interested in market facts, I hope someone has the interest to conduct the following exploratory research. (If no one shows interest, I might do it myself and post the results).
Compute a moving-window volatility of a large (preferably international) stock index. This volatility is a proxy for global market risk. Set dummies based on this volatility. For example, classify volatilities below the 25th percentile as "low risk", and volatilities above the 75th percentile as "high risk".
Next, compute three cross-correlations of a number of currency pairs: uncondtional correlations; "low risk" correlations and "high risk" correlations. We might expect to find that negative correlations of safe-haven and risky currencies are more pronounced during risky periods, as capital flights for safety.
Naturally, there are better techniques to approach this problem. However, I suggest this method because I think that for some basic exploratory research, this would be sufficient.
In asset markets, correlations seem to rise with volatility (e.g. when the market crashes, all stocks crash together). There might be various reasons for this observation, I'll provide one example: when participants hold leveraged portfolios, then, when some stocks in the portfolio drop significantly, traders need to sell parts of their portfolio in order to avoid margin calls.
For the Forex market, the volatility/correlation relationship is more ambiguous. One could hypothesize that during bad times, investors flight for safety, and we see a more pronounced negative correlation between safe-haven and "risky" currencies.
There are various methods to examine the relationships described above, including a multivariate GARCH, if I'm not mistaking. However, such econometric models tend to get rather sophisticated.
Since there are at least a handful of people interested in market facts, I hope someone has the interest to conduct the following exploratory research. (If no one shows interest, I might do it myself and post the results).
Compute a moving-window volatility of a large (preferably international) stock index. This volatility is a proxy for global market risk. Set dummies based on this volatility. For example, classify volatilities below the 25th percentile as "low risk", and volatilities above the 75th percentile as "high risk".
Next, compute three cross-correlations of a number of currency pairs: uncondtional correlations; "low risk" correlations and "high risk" correlations. We might expect to find that negative correlations of safe-haven and risky currencies are more pronounced during risky periods, as capital flights for safety.
Naturally, there are better techniques to approach this problem. However, I suggest this method because I think that for some basic exploratory research, this would be sufficient.
The nail that sticks out gets hammered back in