When I looked at forecasting stock market returns, I began by finding a base expectation: stock prices rise over a month-long period about 59% of the time. To continue the analogy to weather forecasting, we know that weather depends upon the season. Stock returns also seem to depend on the time of year, which is referred to as calendar effects.

Some well-known calendar effects include the Santa Claus rally, the January effect (especially for small caps), and “sell in May and go away”. And, of course, the fact that the worst market crashes have all happened in October. Keep in mind that because these effects are well-known, they are not very pronounced. At the same time, there may be an element of self-fulfilling prophecies.

Looking at the TSX since 1979, these are the probabilities of the stock market rising in each month.

January: 59% (typical)

February: 59% (typical)

March: 58% (typical)

April: 58% (typical)

May: 62% (slightly better)

June: 46% (worse)

July: 62% (slightly better)

August: 59% (typical)

September: 43% (much worse)

October: 57% (surprisingly typical)

November: 65% (better)

December: 81% (much better)

The real standouts are June, September, November and December. I was really surprised that October is essentially typical in how often the month will produce a positive or negative return. It’s just that when it’s negative, it’s a crash. I had never thought of June being as risky as September, or of December as being pretty close to a sure thing.

Over the period of 1979 to 2016, the average monthly return of the TSX is 0.60%. This translates to a yearly return of 7.18%. Looking at the frequency of positive returns gives an idea of the probability of a rising stock market. Looking at the average return for the month gives an idea of how far the stock market rose. Most are pretty close to the average, with the following exceptions.

October has a negative average return (-0.15%), even though it is positive more often than negative. That means that when it’s negative, it’s really bad. September is the worst, with an average monthly return of -1.62%. The only other month with a negative average return is June, at -0.43%, which is worse than October (more often, but to a lesser degree).

My forecast for the current month, June 2016, based on the calendar effect would be 46% probability of a positive return. Add to that the fact that May 2016 was positive (1.44%), and I’ll adjust upward to 51%. I’m willing to lay it on the line: there is a 51% chance that the TSX will end the month higher than 14,063.54.

I fully recognize that’s not really helpful. I’ll continue looking for variables that can help refine my model. Next is the business cycle and sector rotation, but I’m also open to suggestions.

Forecasting refinement – calendar effects

One thought on “Forecasting refinement – calendar effects

Leave a Reply

Your email address will not be published. Required fields are marked *