Full report
The Dukascopy Bank SA research department continues its work on statistical properties of currency pair returns. The previous part of the research marked out relationships amongst returns. It was therefore dedicated to a somewhat determinate aspect of their behaviour. In this issue we look into uncertainty of the movements and investigate the parameters that can help reduce it.
The most widely quoted measure of uncertainty in asset movements is volatility. Some sources position it merely as a measure of risk, while others go as far as proposing associated trading strategies. Therefore we set a goal to investigate properties of volatility and see what information it can actually carry.
We base the study on our five top traded currency pairs: EUR/USD, GBP/USD, USD/JPY, USD/CHF and EUR/JPY, - as they seem to be of the greatest interest to our clients. The data used are different frequency exchange rates (2012 ten minute and one hour rates, and 2000-2012 one day rates).
Methodology
Although we mostly use volatility to describe exchange rates per se, the calculations are not based on prices directly. Instead, we derive volatility from returns.
Returns reflect the changes of prices, and asset instability gets captured in their fluctuations. Namely, interchanging positive and negative returns of great magnitude define high instability, while smaller variations indicate calmer movements (see Figure 1).
Volatility Calculation
There are several commonly used formulas for calculating volatility, but they all reflect similar ideas. In fact, the most popular ones – squared returns, return absolute values, and return standard deviations, - are derived from the same notion.
Results
1. Each currency pair has a specific volatility range.
2. Volatility tends to linger on the achieved level.
3. Volatility of high frequency rates is periodical.
4. Volatility scaling often gives significantly inaccurate results.
Conclusion
In this research we have looked into instability of exchange rates and possibilities to reduce its effects. It appeared that under certain circumstances some prognoses are possible. Their accuracy, however, is limited to expecting one or the other level of volatility.
We have found that the magnitude of exchange rate fluctuations changes gradually, and that less stable times generally last longer than calm ones. In addition, high frequency rates seem to be more volatile in the middle of the day, while evenings and mornings offer a calmer picture. This might be useful for both risk-averse and risk tolerant investors while choosing a more preferable trading time.
Scaling appeared to be a rather inaccurate way to convert volatility. It tends to under or overestimate risks, depending on the magnitude of the scaling factor. However, sometimes it is impossible to manage without conversion. In such cases it is important to estimate not only the value, but also its possible error.
Finally, we have established that each currency pair has a typical volatility level, and that some pairs are more unstable than others. Thus, it is possible to balance risks and profits by taking into account this “instability ranking” of currency pairs. However, we have also seen that the levels can change over time. Therefore, to try to use volatility, one must monitor it closely to capture any changes in behaviour.