From characteristics to prices
In this capstone you estimate what the stock market pays for three simple traits, 12-month momentum, last month's return and 12-month volatility, and you decide how much to believe the answer. You turn daily closes into monthly returns of the single stocks, build for every stock-month the traits known at that date and the return that follows, and run the Fama–MacBeth procedure: one cross-sectional regression of next-month returns on the standardised traits for every month, then the mean of the monthly slopes with its standard errors. Statsmodels does the regressions and you write the rest.
Two details make the answer honest. The standard error of the mean slope must respect that monthly slopes can be autocorrelated, so you code the Newey–West t-statistic yourself, with Bartlett weights. And a premium is only interesting if it survives the known factors, so you form a long-short portfolio on volatility and regress its monthly return on the Fama–French market, size and value factors with a HAC covariance. The intercept of that regression, its alpha, is what remains unexplained.
- Two passes: monthly cross-sectional regressions first, then the mean and spread of the monthly slopes.
- Robust errors: a Newey–West t-statistic for the mean slope and a HAC covariance for the alpha regression.
- Survivorship bias: a list of today's winners flatters the premium, so quote it as an upper bound.
Lab · what's graded
- Monthly returns10 pts
- The stock-month panel15 pts
- One month's prices10 pts
- The first pass10 pts
- A Newey–West t-statistic15 pts
- The second pass10 pts
- A long-short portfolio10 pts
- Alpha against the factors10 pts
- The whole study10 pts