Not every problem is a nail

Readers of these pages know that I am an advocate of thinking about your assumptions when modeling. Just because you have a useful tool doesn’t mean that you should use it on everything.

A colleague sent me the work from the folks at Equity Clock, who are doing research on market timing based on seasonality:

Criteria for model evaluation
When I look at this research, several questions come to mind:

I understand turn-of-year effects, which has been well documented in academic literature, but what is the economic basis for seasonality on sectors and industries like Consumer Staples, Utilities and Biotech? What about small cap growth and value?

If we were to accept the premise that seasonal effect observed in the Energy sector, why does the seasonal effect for natural gas not coincide with Energy stocks? Are their fundamental drivers that different?

When I evaluate the effectiveness of a quantitative model, here are some of the considerations that I think about:

• What is the alpha, or average incremental return over the test period?
• What is the batting average?
• How consistent is it? For example, did it work as well early in the test period as later in the test period?
• What is the turnover? What are the implementation costs, i.e. can we trade this model and make money?

Don’t use a single tool indiscriminately
A number of years ago I did some research into seasonality effects on selected industries. We found that there was a weak seasonal effect for US retailers around Black Friday. In addition, turn-of-year effects have been cited by many researchers in many markets around the world. CXO Advisory recently did some work on gold and gold stock seasonality, their conclusion was that you get very different conclusions depending on what test period you chose.

Seasonality is a tool that can be used in certain circumstances, but it’s not a universal tool. Don’t fall into the trap that if you have a hammer, every problem looks like a nail.