US equity risk factors and portfolios · Daily
Loading.
Eighteen daily long-short factors for US equities, built from individual stocks in the spirit of the NEFIN factors for Brazil. Prices come from Yahoo Finance and accounting data from SEC filings. Nothing here is taken from a published factor library.
The window buttons at the top apply to every exhibit and table on this page.
CAGR geometric. Sharpe and Sortino on excess returns over the T-bill, annualised; Sortino uses downside deviation. t is the t-statistic of the mean excess return. VaR and CVaR are historical, 95%, one day. Beta and information ratio are against the S&P 500. Factor rows are zero-investment spreads, so Sharpe uses them as is.
Daily returns, selected window, times 100. Bold where the absolute correlation is 50 or more.
Rp − Rf = α + β1·MKT + β2·SMB + β3·HML + … + ε
import pandas as pd
f = pd.read_csv("https://macroconvexity.github.io/arthurmota/risk/data/factors_daily.csv", index_col=0, parse_dates=True)
f <- read.csv("https://macroconvexity.github.io/arthurmota/risk/data/factors_daily.csv")
f$date <- as.Date(f$date)
Cite as Mota (2026), US Risk Model, macroconvexity.github.io/arthurmota/risk.
Survivorship bias, the main limit. The model holds companies that exist today, so companies that failed or were delisted are missing. This flatters long-only portfolios, small caps most of all, and it works against factors that short risky stocks (IVOL, MAX, QMJ, BAB), because the worst outcomes on the short side are absent. Use the sorted portfolios for exposures, not for the level of return. Long-short factors are less affected because both legs share the bias.
Check against the academic library. The table compares each factor with its Ken French counterpart as a diagnostic. The definitions differ, so the match is not expected to be exact.
Educational research. Not investment advice. Past returns are not predictive. Prices can be revised.