Stay informed with free updates
Simply sign up to the Currencies myFT Digest — delivered directly to your inbox.
Barclays’ currencies team has an interesting note out this week on something a lot of people believe in, even though logic says it can’t exist: forex seasonality.
Efficient, frictionless markets just shouldn’t be affected by the time of year, even when there are demand-led reasons to argue otherwise. Here’s a paper that correlates Venetian galley fleet launch dates to seasonality in the medieval FX market between 1383 and 1411, for example, and (perhaps more relevant to now) ones that link underperformance to a country’s tax year and the aggregate of its recent dividend payments.
What’s lacking is evidence that any seasonal trend is robust enough to survive in sunlight. The presumption is always that a known and profitable trading strategy will be arbitraged away long before it’s peer-reviewed. Claims about “a Santa rally” and “sell in May” are therefore considered useful only for identifying trading-platform shills.
Barclays’ Lefteris Farmakis and team test the theory by checking for performance anomalies in FX pairs versus the average. We’ve put their data (which uses so called dummy variable regressions, statistics nerds) into a couple of interactive charts:
Per the above, dollar underperformance against the safe currencies is only occasional, and it’s on the fringe of statistical significance only in December. The trend also appears to have weakened in recent years. A “sell in May” effect (where riskier currencies should underperform) is unconvincing in the long run, even if it is more apparent recently.
Here’s the same exercise for emerging markets currencies. The pattern of risk-off and risk-on in December and May are similar though seasonal distortions are a bit more frequent, probably owing to thinner liquidity:
Stage two of Barclays’ analysis is to test whether seasonal effects are diluted over time. This involves a lot of hard maths that’s beyond the scope of a blog post so we’ll skip to the conclusion, which is that among G10 currencies seasonal patterns have been persistent only for the New Zealand dollar and the Swedish krona. Moreover, while May effects have faded over time for both currencies, the December effect appears to have stuck around.
Why? Well, demand mostly, though it’s complicated.
New Zealand’s dairy exports drop sharply in August then recover strongly in December, which broadly matches the NZD trend. For Sweden, there’s a big flow of dividends from foreign investments in the second quarter that might contribute to the May effect, but that doesn’t explain why the krona tends to outperform in December. It’s particularly odd because the annual pension disbursement from the Swedish Pension Agency in that month should be krona negative, says Barclays.
Seasonality in emerging markets is more tenacious, with Barclays spotting repeating patterns in the Indian rupee, Indonesian rupiah, Philippine peso, Taiwan dollar, Brazilian real, Polish zloty, South African rand and the pesos of Chile and Mexico. The possible explanations include equity flows into India, EU disbursements to Poland, copper exports from Chile and Philippines overseas worker remittances.
But back to the main question: has Barclays proved the existence of Santa by isolating his rally?
[W]e find little systematic evidence for the two most famous market seasonals, and particularly for the alleged ‘sell in May’ pattern of broad-based USD strength. Even within the context of the ‘dummy variable’ approach this pattern flips from significant to insignificant depending on sample length, while there is no corresponding sign of seasonal effect in equity markets (proxied by the S&P 500 index). The same comments apply to the so-called ‘Santa rally’ in December, with the only caveat that more instances of statistically significant December seasonality survive our tests, particularly in EM FX.
