ANOTHER LOOK INTO ANOMALIES IN CROSS-SECTIONAL STOCK RETURNS : THE CASE OF SINGAPORE

Abstract

Extensive research in the US has documented various anomalous effects in stock returns. For instance. firms with small market capitalisation. high Earnings-to-Price ratio (E/P) and Book-to-Market Value of Equity ratio (B/P) and so forth are found to exhibit returns that significantly exceed those of firms with large market capitalisation, low E/P and B/P. However, these results are not conclusive. Furthermore, they may be biased because their methodologies are based on unrealisable assumptions about the distributions of the sample data. This study tests the presence of four anomalies, namely, the size, E/P, B/P and Sales-to-Price (S/P) effects in the Stock Exchange of Singapore. We also include Total Assets-to-Price ratio (TA/P) and the return from the previous month to this study for the purposes of control and comparison. With the significance frequency of an anomaly variable from each cross-sectional regression across the sample period as the test instrument, we observe anomalies being present but they are far less pronounced than those documented by previous studies using other methodologies and cannot be used to form investment strategies even though their presence cannot be rejected statistically. Specifically, small size demonstrates noticeable explanatory power for cross-sectional stock returns. B/P and TA/P anomalies show lower significance than size but remain marginally accountable for stock returns. S/P anomaly has very weak impact on stock returns. There is hardly any relation between E/P and cross-section stock returns. The return from the previous month is related to stock return. Size is robust to the inclusion of other variables, so is S/P though the latter is not an important explanatory factor of stock return. There is a strong offsetting effect between B/P and TA/P. These same results hold even after adjustments for variable timing and heteroscedasticity. With the same data set, we also produce results using Fama & MacBeth (1973) (F&M) regressions - a commonly used statistical procedure in the test of anomalous returns. We find that the results are not consistent with those based on the significance frequency test. This is because F&M test is not robust to nonstationarity of regression coefficients. We then propose that the anomalies documented by previous literature may partly be statistical artifacts.