Performance measurement: Market timing and security selection

Mads B√łgelund

Student thesis: Master thesis


The Master thesis deals with models for performance measurement of market timing and security selection, which has been a hot topic since 1966. In that year Treynor & Mazuy proposed a groundbreaking model to test the timing of portfolios. Since then, a number of other models has been developed, all of which is analyzed in relation to validity and accuracy. The focus is on the kind of timing that involves investments in the risk-free asset. If a model is accurate, it should not find any selection in portfolios of so-called perfect or perverse timing. These artificial portfolios are constructed using returns from the risk-free asset and the value-weighted market to achieve the best and the worst possible timing of the market. No risky assets are preferred to others, so there is no selection present. It is argued that the overall performance should be measured by the differential return. This is the return above a relevant benchmark, of which the definition is very important. The proposed benchmark provides the opportunity for the development of a new model that contains only two estimates. The existing models are based either on total or excess return, but the analysis of differential return allows for an absolute interpretation of the models' estimates. The sum of the contributions from timing and selection is exactly equal to the average differential return, which measures the overall performance. The benchmark takes account of any restrictions on investment in the risk-free asset, which is not necessarily prohibited. Traditionally, any combination of the risk-free asset and the value-weighted market is regarded as a passive investment, of which the owner can be responsible. This is not correct however, as it is in direct conflict with strategies for timing. As a perspective on the problems, the solution is extended to allow for both types of timing, since there will also be an estimate for the investment in risky assets with a desired systematic risk. This differs from selection, where assets are chosen if their prices are too low. The extension of the model thus provides a breakdown of performance into selection and two kinds of timing, which together constitute the return above the benchmark. The subject does not include risk adjustments, which is consistent with existing models regarding timing and selection.

EducationsMSc in Finance and Accounting, (Graduate Programme) Final Thesis
Publication date2009
Number of pages80