Not logged in.
Quick Search - Contribution
Contribution Details
| Type | Bachelor's Thesis |
| Scope | Discipline-based scholarship |
| Title | Is there an Alphabet Bias in Stock Returns? |
| Organization Unit | |
| Authors |
|
| Supervisors |
|
| Language |
|
| Institution | University of Zurich |
| Faculty | Faculty of Business, Economics and Informatics |
| Number of Pages | 18 |
| Date | 2016 |
| Zusammenfassung | The main topic of this thesis was the assessment of whether the alphabetic position of a stock is a determinant of its corresponding return, or not. Based on the empirically suggested argument that turnover is a predictor of stock returns, while turnover itself is dependent on the alphabetic position, I formulated the thesis that some of the effect of turnover on return might be due to the alphabetic position of the stock. For the purpose of the study, I created a sample of 441 US companies of which accounting and capital market data back to the year 1997 were available. I then decided to conduct a first test, in which I used a cross sectional regression in order to determine if the alphabetic position of those stocks is able to explain their average return. For adjusting the data and calculating the input parameters, I primarily used the program “MatLab” because of its ability to cope with large matrices. In contrast, the regressions themselves were performed with the program “R project”. The regressions of the first test were based on the annual averages of various variables between 1998 and 2014, since most accounting data was just available on an annual basis. The first two regressions I estimated were based on the Fama and French three, or respectively Fama and French five factor model. The company specific averages of “Book-to-Market”, “Size”, “Profitability”, and “Investment Behaviour” were supplemented with the average logarithmic turnover of the company. This led to two benchmark regressions in which the effect of the mentioned variables on average return could be observed. After that, six regressions based on different alphabetic variables as well as the benchmark variables were estimated. Those regressions were featuring alphabetic dummies which acted as an interacting variable with logarithmic turnover. It turned out, that neither the alphabetic dummies alone, nor their interaction terms with turnover were showing any significant influence on average return. However, the tendency of the estimates did suggest that stocks of high alphabetic order are resulting in lower average return. Furthermore, not even logarithmic turnover seemed to have a significant impact on average return. Since the adjusted R2 values of the regressions were below 10%, I decided that a more precise test was needed in order to draw a final conclusion. In order to conduct a more precise test, I decided to run two fixed effect panel regressions. Again, the first one should serve as a benchmark, whereas the second one should assess the direct and indirect effect of alphabetic order on return. This time, a monthly observation frequency was used, which led to a total of 83349 observations. The first panel regression controlled for time and entity specific fixed effects, for the reason that a dependency of capital market data and accounting data on the observed entity and time can be assumed. The benchmark regression provided a solid adjusted R2 value of 24% and allowed an assessment of its 441 entity specific intercepts. The intercepts suggested that an entity located at the upper 5% range of alphabetic order started out on average, with a monthly return that was 0.18% lower than the return of their 95% counterparts. Again, this was not enough to draw a final conclusion, since not all entity specific intercepts were significant in regards to their t values, and furthermore the measured effect could have been caused by another omitted entity specific factor different to the alphabetic position of the entities. Additionally, a second panel regression was conducted, only controlling for time specific fixed effects. This was necessary since the entity specific fixed effect and the alphabetic position of a stock suffered from perfect multicollinearity. It was therefore clear that the regression would suffer from an omitted variable bias (namely the entity specific fixed effect), yet it was perfectly capable of capturing nearly the same result for the variables that were used in the benchmark model. In addition, it again showed that the alphabetic position directly and indirectly lacked of a significant effect on monthly return. Since turnover itself proved to be significant in regards to its t value in both the benchmark model and the second panel regression, I concluded that turnover might have the proposed influence on return, but alphabetic position lacks the ability to improve the regression. This conclusion was supported by the slightly lower adjusted R2 of the second regression. I finally came to the conclusion that my results could be explained by observing them from three perspectives. As a first probable explanation, I argued that the failure to capture any significant effect of turnover or the alphabetic position on average return, might be caused by the unclear effect of turnover itself in regards to average return. Since high turnover can be observed in boom and bust stages of the market, on average a clear prediction of return with turnover seems rather unlikely. The same would therefore be true for alphabetic position which is a determinant of turnover. As a second probable explanation for the result, I referred to a theoretical argument, describing that returns are primarily a compensation of risk. It would therefore be rather unclear what kind of risk the alphabetic position would pose, and if it should be compensated at all. Last but not least, I additionally suggested that from an evolutionary finance point of view, a portfolio selection principle based on such a straight-forward approach, like choosing stocks located at the end of the alphabet, would be quickly discovered and eliminated. Since the “alphabet bias” is still present, I conclude that this bias might be of use for a portfolio when thinking about liquidation and marketability of certain assets, but poses irrelevant in regards to the return of those assets. |
| Export |
BibTeX
|