Abstract
In this paper I examine the power of multifactor regression models in describing the returns to New Zealand stocks and portfolios from January 2006 to October 2010. The analysis focuses on making additions and adjustments to the market model, the Fama-French model and the Carhart model in order to determine if these are adequate for use with thinly traded equities. I show that a market model regression has negligible explanatory power for thinly traded equities when compared to thickly traded equities. The standard approach of using portfolio returns as the dependent variables in the multi-factor regressions is used in the first instance in order to determine if the factors are significant. Once the significance of these is shown according to the accepted methodology, the models are applied to the returns of individual thinly traded stocks. The study period is segmented into three equally-sized consecutive time periods with very different market conditions in order to determine if there is time or market variance in risk exposure and premia. Key findings are that beta estimation for thinly traded stocks is greatly improved by using the average of the bid and ask price rather than the close price to calculate returns and by using an equal-weighted benchmark to proxy the market. By using this method, estimations of beta are significantly higher and also are negatively correlated with size and liquidity, which is contrary to previous findings. I find size, book-to-market value of equity, liquidity and momentum all to be significant systematic or common risks, over and above market risk, in New Zealand. However, the significance of the risks and the premia associated with them vary with time and/or the prevailing market. GARCH effects are also significant across time periods and this significance is generally robust to changes in explanatory variables. Furthermore, I provide evidence in support of size risk being related to, but not a proxy for, liquidity risk. I show that liquidity measures in New Zealand are not equal, and that liquidity may be a multidimensional risk itself. However, I conclude the power of these regression models, even with numerous modifications, is generally poor for individual thinly traded equities.