Do government bond returns worldwide exhibit seasonal effects analogous to those of stock market returns? In their August 2016 draft paper entitled “Seasonality in Government Bond Returns and Factor Premia”, Adam Zaremba and Tomasz Schabek investigate seasonal patterns in government bond returns across countries, focusing on regression tests of January and sell-in-May (May-October versus November-April) effects. They also examine whether four bond risk premiums (volatility, credit risk, value and momentum), each specified in multiple ways and measured via long-short portfolios formed from monthly sorts, exhibit these two seasonal effects. Using monthly total return bond indexes hedged against the U.S. dollar spanning 25 countries and allocated to five term ranges (1-3 years, 3-5 years, 5-7 years, 7-10 years and 10+ years) during January 1992 through June 2016, they find that:
- Over the sample period and across all countries:
- Average monthly total returns for term ranges 1-3 years, 3-5 years, 5-7 years, 7-10 years and 10+ years are 0.32%, 0.43%, 0.50%, 0.55% and 0.71%, respectively.
- Average volatilities of monthly returns for term ranges 1-3 years, 3-5 years, 5-7 years, 7-10 years and 10+ years are 0.46%, 1.02%%, 1.36%, 1.79% and 2.85%, respectively.
- In aggregate over the sample period, there are clear volatility, value and momentum premiums in the expected directions, but the credit premium (based on sovereign bond ratings) is inconsistent and mostly insignificant.
- Regarding any January effect:
- Only one of 25 countries exhibits a consistently significant January effect across maturity ranges (a positive effect in Italy).
- Significantly high returns in January occur for some maturities in only five other countries (Belgium, Ireland, New Zealand, Norway and Spain).
- For combined data, while returns are somewhat elevated during January across maturities, results are not significant.
- Regarding any sell-in-May effect.
- There are significantly low returns during November through April across maturity ranges only in Canada, the U.S. and (for four of five maturity ranges) Australia.
- For combined data, while returns are somewhat depressed during November through April across maturities, results are not significant.
- The four factor premiums do not exhibit any significant seasonality.
In summary, evidence indicates that any January and sell-in-May effects in government bonds worldwide are minor.
Cautions regarding findings include:
- Monthly returns used in the study come from indexes, which do not account for the costs of maintaining a liquid fund. These returns likely overstate returns realized by investors.
- Reported returns for seasonal effects are gross, not net. Trading to exploit any seasonal effects would incur costs that reduce these returns.
- In citing countries exhibiting seasonal effects, the authors apparently do not correct statistics for the snooping bias (false findings) associated with trying the same seasonal specifications factors across 125 country-term subsamples that embed randomness.
- Since returns are likely correlated to some degree across countries, effective country sample size is less than 25 countries.
- The sample period is one of secular decline in interest rates (tail wind for bond returns). Findings may differ for other interest rate environments.
See also “Mirror Image Seasonality for Stocks and Treasuries?” and “Bonds During the Off Season?”.
