Skip to content

Latest commit

 

History

History
43 lines (35 loc) · 2.33 KB

test-results.md

File metadata and controls

43 lines (35 loc) · 2.33 KB

ExpectedReturns Tests

0: Pre-req

1: Easy

  • Download/install went smoothly

2: Intermediate

  • Time-series-momentum.Rmd: Doesn't build because ff.dates does not match index of data.

3: Harder

  • Commodities-long-run.Rmd:
    • Analysis:
      • Disclaimer: I'm no finance expert (yet), but I did my best.
      • Macro Indicators
        • The regressions show statistical significance of the macroeconomic states in the short-term for an equally-weighted portfolio of commodities. This suggests that changes in macroeconomic factors have effects in the short-term, but fade over longer horizons. Additionally, since there wasn't statistical significance for the long-short portfolio, it's possible that the strategy intentionally adjusts for macroeconomic factors, which could explain the low statistical significance.
        • The time-series factor model showed alphas of 0.007% and 0.004% for equally-weighted and long-short portfolios. From what I can gather, this is a pretty small alpha, giving a tiny edge in the market. Lastly, the residual volatilities are low--about 0.05% for both portfolios. Thus, the model has captured most of the variation in asset returns.
      • Investment Styles
        • For the short term (horizon of 0 months), momentum was found to be very statistically significant. Momentum was also significant for an 11-month horizon using the long-short portfolio. For the other horizons of 11 and 59 months, however, no relationship was found.
        • Value did not show statistical significance for expected returns.
        • Carry showed statistical significance in the long term (horizon of 59 months) for the long-short portfolio, but not the equally-weighted portfolio.
    • Repetitive code:
      • lapply(unlist(macro.regs, recursive=FALSE), summary) was used multiple times to provide summary statistics for multivariate regressions.
      • This could be a function called multivariateRegression to provide summary stats given the dependent/independent variables, horizons, and data.