3.6 Appdendix to Chapter 3
4.2.2 Seasonality in Bond Returns
While the seasonality in risky equity markets has received extensive coverage in existing literature as cited above, much less focus has been paid to seasonality in risk-free equity returns. However, arguably, investors who expect seasonality in risky asset returns might shift their investment portfolio to risk-free assets to avoid potential loss. Therefore, we expect reverse seasonal patterns in bond returns compared to seasonality in stock returns. .
Several studies have shown the days of the week effect also exists in risk-free asset market. Gibbons and Hess (1981) found a pattern in Treasury bill returns similar to those in risky equity. They investigated Treasury Bills for the period of December 1962 to December 1968 and observed Treasury bill returns on Mondays are lower than average. Flannery and Protopapadakis (1988) analysed seven Treasury Bills maturities ranging from one month to thirty years and overnight repurchase agreements. Their study showed that Monday Treasury returns and underlying maturity is negatively correlated. Johnston et al. (1991), Singleton and Wingender (1994),and Griffiths and Winters (1995)) further confirmed the existence of the days of the week effect in a types of debts, including federal funds and mortgage-backed securities. The Monday effect is consistent in both stock and bond
markets, our findings of Chapter 3 suggested that trading activities on Mondays are lower than the other days of the week, as Monday tends to be a day of strategic planning for institutional investors. Thus, it is expected that returns of both stock and bond market are lower on Mondays.
The May principle refers to negative or below average returns in stock markets from May to October. When a growing number of investors are expecting the May principle in stock markets, then it is expected that the returns of government bonds will also be affected during this period, as investors who sell their stocks in May may turn to the bond market. Athanassakos (2008) considered stock and government bond data from 1957 to 2003 and employed time series dummy GMM regressions to test for seasonality in the Canadian financial market. He documented significant seasonal patterns in both risky and risk-free markets in Canada. In addition, he also revealed the seasonality effect on stock market was in the opposite direction to that of bond markets, which is as expected. In detail, the average bond returns from November to April is higher than that from May to October, stock returns from November to April is lower than that from May to October.
The January effect has also been proven in the literature. Chang and Huang (1990) demon- strated the January effect in US long-term corporate bonds by studying the pricing of equally weighted long-term corporate bond portfolios in six different ratings. Their sam- ple consists of Moody’s classifications from Aaa-rated to B-rated bonds. Wilson and Jones (1990) also presented the January effect on corporate bonds and commercial paper returns . They examined a 131-year period of data for both series, and applied a pro- cedure that provides consistent estimates of the variance-covariance matrix. Their result suggested that the January effect persists for both corporate bond and commercial paper during the entire period. Clayton et al. (1989) documented that the long-term government
bonds had significantly lower returns in January than in the remaining year and argued the tax-loss selling in the equity markets together with investors investing their December sales proceeds in January provide a potential explanation for the January effect. Smith (2002) extended the study of the January effect to international government bond markets and provided evidence of the January effect in US, France, Germany, UK and Canada. He stated that the results provided considerable diversification opportunities for investors. However, the January effect in US treasury markets were challenged by the studies of Schneeweis and Woolridge (1979), Smirlock (1985) and Chang and Pinegar (1986), who suggested that the January effect is not consistent over time. Clayton et al. (1989) ques- tioned the findings of these three papers and argued that the datasets and methodologies they employed were not favourable for discovering the January effect in bond markets.
Literature is also available regarding the SAD effect in risk-free asset returns. Kamstra et al. (2014) examined 4 medium-to-long term US treasury returns and documented a seasonal pattern in US treasury returns. They performed seasonality tests by estimating Hansen (1982) generalized method of moments tests with Newey and West (1987, 1994) heteroskedasticity and autocorrelation consistent stand error to control for heteroskedas- ticity and autocorrelation in treasury returns. The result indicated the seasonal variation result an average 80 basis points swing in treasury returns from October to April. They included the SAD effect in their study on bond returns and provided evidence of a fall- winter and a OR bond return patterns. In the fall-winter pattern, bond returns are higher during September to November and lower in February to April. The OR variable, intro- duced by Kamstra et al. (2014), represents changes in the proportion of SAD symptoms sufferers, it is negative in winter and spring and positive in summer and fall. Including OR variable could constitute an improved version that links the bond returns to the clinical
evidence of SAD symptoms as SAD variables are only related to the length of daytime. A positive relationship between OR and US treasury returns discerned, this means a higher proportion of SAD sufferers in the population leads to higher treasury returns, which is opposite to the SAD effect on stock returns. They considered 11 alternative models with macroeconomic and risk factors are controlled, and showed that seasonality in these macroeconomic and risk factors does not respond to the seasonal Treasury return pattern. It is thus suggested that SAD, causing seasonally varying investor risk perception, is the most important factor behind seasonal variations in financial markets. Investors suffer from SAD symptoms and become more risk averse, they tend to avoid risk and turn to bond markets, leading higher bond returns in the fall and winter; when investors recover from SAD symptoms, they rebalance their portfolios and increase investment in risky assets.