Investors are often tempted by rumors of calendar effects on financial markets. After all, the crashes in 1929 and 1987 both came in October. Maybe it is better to avoid the stock market in October. This exercise uses daily stock returns from 2010 through 2015 and divides the year by month. Use the data from the first three years 2010–2012 to estimate the needed model, as if you were an investor at the end of 2012. Reserve the data for 2013–2015 for part (g).

Motivation
(a) Explain why it would be surprising to find stock returns statistically significantly higher year after year in a specific month or on a specific day of the week.

(b) If a stock-market trader repeatedly tests for statistically significant differences in market returns by month or by day of the week every year, what problem is likely to happen? Method

(c) Describe how to test for a statistically significant difference in average daily returns on the stock market in 2010–2012 by month?

(d) What assumption of the MRM is likely to be violated in this context?

Mechanics

(e) Perform the analysis, using May as the baseline category. Do you find a statistically significant difference among months in 2010–2012? If so, for which months?

(f) Based on the estimated coefficients in your model, which two months have the largest difference in returns. It is important that May is the baseline category.

Message

(g) Explain how an investor would have been misled by t-statistics from the model estimated using data from 2010–2012 to identify a statistically significant difference. Use the data from 2013– 2015 to support your answer.