Seasonal patterns in agricultural commodities seem intuitively easier to understand than seasonal patterns in the foreign exchange market, where the turnover far and away exceeds the international trade. Often analysts and journalists claim a seasonal pattern in the dollar that does not meet a robust statistical standard. The sample size is frequently too small, or they are too quick to reject the null hypothesis that there is no pattern. Putting weight on an average move, as some analysis does, also seems misleading, given the variance. Alas, many might be fooled by randomness.
With Fordham global finance graduate student Nicholas Hatzis, we reviewed the major currencies and gold price action over the past twenty years, using monthly data. Because of the small sample size, we want to establish a high threshold before claiming a pattern and are suspicious if the direction was not sustained for 15 of the 20 years. In a review of our findings below, we also note patterns of 14 years as the next data point could meet our minimum threshold.
Claims of the yen's seasonal pattern may be the most popular. Japan's fiscal year-end (Mar. 31) is often linked to asset managers adjusting positions. First repatriation and then at the start of the new fiscal year, boosting foreign investments. We did not find such a pattern. However, August is the most notable. In 14 of the past 20 years, the dollar has fallen against the yen in August.
The euro has been unable to reach the threshold of moving in the same direction in a given month 15 times over the past 20 years. However, it came close a few times. For example, in June and December, the euro has risen against the dollar in fourteen of the past 20 years.
Of the major currencies, sterling has the most robust patterns. In the past twenty years, sterling has appreciated against the dollar in 18 of the past 20 Aprils, including last month. The pattern is just as strong in May but in the opposite direction. It has fallen in 17 of the past 20 Mays. August meets our threshold. Sterling has fallen in August 15 of the past 20 years. June and November are close, with sterling advancing 14 times and falling 14 times in the past 20 years, respectively.
The Australian dollar is interesting. Over the past 20 years, the Aussie has risen in 15 years in February, April (including this year), and June. Conversely, it has fallen 14 times in May and 15 times in August.
The US dollar typically falls against the Canadian dollar in April. It has moved lower in 16 of the past 20 Aprils, including last month when it fell by about 2.2%. Our threshold was approached in June, when the dollar has depreciated 14 of the past 20 years in June, and in November when the greenback rose 14 times of the past 20 years.
The strongest month for gold, which has largely been de-monetized, though still held in reserves by most central banks, is August. It has risen in 15 of the past 20 Augusts. Likewise, it has increased in 14 of the past 20 Januarys and Aprils while falling just as often in March.
Contrary to what is often suggested, charts and data do not often speak for themselves. There are often spurious correlations and more complicated relationships. As a starting point, begin probing possible reasons for the seasonal pattern, we looked at foreign buying of US Treasuries. There are indeed some strong seasonal patterns. Several factors could influence when Treasuries are bought, including new supply, reserve management decisions, the schedule of coupon payments (twice a year).
The UK's purchases of Treasuries appear to have the strongest seasonality. For the past 20 years, the UK has not failed to buy US Treasuries. It has been a buyer in 19 of the past 20 Februaries. The UK has been a net buyer 18 times in the past 20 Mays, 16 Marches and Augusts, and 15 Junes. The role of London as a financial center likely accounts for the strong buying.
No other country approaches the UK. Japan appears to be the closest. It has bought US Treasuries in 16 of the past 20 Februaries and 15 Julys. Canada has bought Treasuries 14 of the past 20 years in February, April, November, and December, and 16 times in March. China's pattern is less robust. May is the most consistent month of purchase of US Treasuries, 15 of the past 20 years. It has been a buyer for 14 of the past 20 years in July and December.
To be sure, in large series, what appears to be a pattern can seem to exist for extended periods but is no part of the randomness. Ideally, a robust pattern can be accompanied by an explanatory narrative. Without the latter, it is difficult to feel confident in most claims of a seasonality. Sterling's gains in April and subsequent decline in May appear to be the strongest pattern among the major currencies. May is also when the UK regularly buys US Treasuries (18 of 20 years). However, in July, when the UK has not failed to buy Treasuries in the past 20 years, sterling's behavior has no robust pattern. Caveat Emptor.