Background
In finance, there exists certain pricing anomalies that disappear as soon as they become known. For instance, stock price returns on Monday are expected to be lower than the rest of the week. The exact cause of these anomalies are difficult to find out, but likely have to do with investor behavior rather than theories of asset pricing. To read more about pricing, check out the Wikipedia page.
These anomalies can be valuable for investors because they allow for a market edge. In fact, if one identifies a pricing anomaly and keeps it to oneself, that person has generated alpha for their portfolio.
Living in Japan, I was amazed at the amount of work holidays there were. In 2019 alone, there were 25 public holidays (although 2019 was an anomaly itself, since an extremely rare 10 day holiday was held for Japan’s Emperor abdication). On these days, the banks are closed. The bank’s frequent holiday closure begs the question: is there an anomaly in the USD/Yen exchange rates on these days?
Hypothesis
My hypothesis is that there is an abnormality about exchange rates on Japanese public holiday dates. The reasoning behind this guess is that most days, Western markets react to news in Japan that happened while they were asleep. However, on holidays, the Japanese markets are the ones to react. To put in more concrete terms, I predict that the mean daily USD/Yen exchange rate on Japanese holidays is statistically different than the mean daily USD/Yen exchange rate on days before and after a holiday.
Methods
To go about testing my prediction, I first downloaded 30 years of historical USD/Yen data from FRED. Then, I used the holidays library in Python to download the past 30 years of Japanese holiday dates. Then, in a Jupyter notebook, I used Pandas and Matplotlab to manipulate and visualize the data.
To test my hypothesis, I conducted a two-sample t-test. My null hypotheses I wanted to test were: 
Results
| Null Hypothesis | T-Test Statistic | P-Value |
|---|---|---|
|
| -0.571 | 0.568 |
|
| -0.134 | 0.893 |
The first test yielded a t-test statistic of -0.57 and a p-value of 0.56. Therefore, the averages are not different at any significance level.
The second test yielded a t-test statistic of -0.134 and a p-value of 0.893. Similar to the previous test, the averages are not different at any significance level.
Conclusion
Taken together, these results suggest that there are no exchange rate anomalies during Japanese holidays. Indeed, the daily percent change of the USD/Yen exchange rate is no higher on a holiday than the day preceding or following a holiday.
Graph 1 illuminates this point well. Even at a cursory glance, it is apparent that the daily percent change in exchange rate is scattered randomly, instead of mostly falling below the 30 day moving average, or mostly above the 30 day moving average.
Pricing anomalies do exist, but this analysis shows that there is none on a Japanese holiday. When a holiday rolls around, traders are better off relaxing than trying to exploit a (nonexistent) exchange rate anomaly.
Links
Github:
Dominant Strategy 