Exploiting Anomalies in the Risk-Return Relationship

...

2.2 Volatility Effect

Another interesting effect that we will discuss is the volatility effect. Blitz and Van Vliet (2007) show that stocks with low historical volatility exhibit superior risk adjusted returns, both in terms of Sharpe ratios and CAPM alphas. They do so by creating decile portfolios that are based on a ranking of stocks on their historical return volatility. This ranking process is similar to the ranking process mentioned for momentum. The ranking based on historical volatility can be compared to a ranking of stocks on their CAPM betas. This is because beta is equal to it’s correlation with the market portfolio multiplied by it’s historical volatility divided by the volatility of the market portfolio.

Ang et al. (Ang, 2006) found that stocks that are highly sensitive to aggregate volatility have low average returns, and the same holds for stocks with high idiosyncratic volatility. This made Blitz and Van Vliet research this effect more thoroughly. Normally, we would expect high-risk stocks to provide higher returns, but the persistence of the volatility-effect shows that that it is positive to generate a positive alpha by following this strategy. Later on, we will discuss our methodology for testing this effect. As the finding that low-volatility stocks exhibit superior risk adjusted returns clearly questions the efficient market hypothesis, testing this result on real-life data will provide us insight in if this effect still holds.

But what exactly causes this intuitively anomalous behaviour of low-volatility stocks? Some explanations for this volatility-anomaly have been discussed by Blitz and Van Vliet (2007). To take fully advantage of the returns of low-risk stocks, they state that it’s necessary to apply leverage. In practice, many investors are not allowed or unwilling to apply leverage, especially on the scale needed to take advantage of this effect. This causes the volatility-effect to remain.

Another explanation lies in the fact that asset managers have an incentive to load on high-volatility stocks, as this is a relatively simple way to generate above-average returns. This leads high-risk stocks to become overpriced, and low-risk stocks to become underpriced. Generating significant positive alpha with the volatility-effect in this way becomes possible. Apart from this, Blitz and Van Vliet (2007) conclude that the volatility-effect could be caused by behavioural biases among private investors. Divergence from risk-averse behaviour may cause high-risk stocks to be overpriced and low risk stocks to be underpriced.

Blitz and Van Vliet (2007) also find that low-risk portfolios underperform the market during boom months, but outperform the market during bear months. The underperformance during up months is hereby less than the outperformance during down months, which overall leads to a higher average return. The high-risk portfolios exhibit the opposite behaviour. In this research, we will test if there’s a significant positive alpha to be found when applying a low-volatility strategy.

2.3 January Effect

In 1976, Rozeff and Kinney found that returns in January are significantly higher than in other months of the year. In the meantime a lot of research has been done on this phenomenon, but a simple explanation has not been found.

Keim (1989) and Roll (1983) argued that the bid-ask spread was the cause of the January effect. End of December usually shows a lot of selling activity, therefore prices are mainly bid quotes in that month. Contrarily, in January buying behaviour can be observed that causes stock prices to be ask quotes.

Keim (1983) also found that returns are the highest for small firms in January, especially in the first week of the year. Ritter and Chopra (1989) showed that small companies exhibit much higher returns than the market in January, both in an increasing and a declining market. This is in line with the findings of Haugen and Lakonishok (1988) regarding window dressing. This implies that at the end of the year painful losses are being sold so they do not show on the annual statement. Small caps are mostly blamed since bad luck on large caps can happen to anyone. In January more risky positions are taken again, which implies the buying of small stocks which drives their price up.

Another explanation is tax-loss selling, done primarily by individual investors. Since losses are tax deductable, individual investors sell their worst performing stocks at the end of the year, often small caps since they are more volatile (Roll, 1983). According to Ritter (1988) investors then ‘park the proceeds’ of the tax-loss selling to January and then prefer small caps again. The last possible explanation for the January effect is the fact of more cash inflows in the stock market in January due to cash payments at the end of the year (Ogden, 1990).

The January effect has been around for a long time and has been extensive researched ever since. However much research has been done on it, a clear answer as to why this effect has persisted for so long has not yet been given. It could be because of the bid-ask spread, a size effect, window dressing, portfolio rebalancing or tax-loss selling, or a combination of everything mentioned above. As Haugen and Jorion (1996) state, the January effect will most likely fade away and will slide into the previous year until it completely disappears.

---------------------------------------------------------------

3 Data description

The analysis is based on all stocks listed on FTSE Eurotop 100. This index provides a broad sample of 106 most highly capitalized blue chip companies in Europe. The sample period was chosen to be the past ten years, starting in January of 2005 and ending in January of 2015. Six stocks did not show returns for the full period so they were excluded, leaving 100 observations. The monthly return index data was gathered by Datastream provided by the Erasmus Data Service Centre. With the monthly return indices, logarithm stock returns are calculated by Rt,n= ln(RIt, n/RIt, n-1), where RIt, n represents the adjusted closing return index of stock t on the last trading day in month n. Instead of price index, the total return index is used to adjust for dividends and stock splits. Over the last decade, the monthly returns of all constituents have an average of 0.715% (or 8.9% annually).

Table 1. Descriptive statistics of monthly log returns

Mean

Minimum

Maximum