Choosing Simple or Log Returns for Annual Equity Factor Portfolios

Summary

The document frames a return-measurement choice for equity factor research using weekly prices across thousands of stocks and annual portfolio rebalancing. It contrasts the compounded simple return over the year with the sum of weekly log returns, and notes that the two calculations can differ noticeably when applied to weekly observations.

It provides no answer, numerical example, empirical comparison, or recommendation about which measure to use. The distinction matters because summed log returns aggregate across periods, while simple returns compound multiplicatively; the appropriate measure depends on the quantity being estimated or reported. The prompt does not discuss corporate actions, missing prices, or portfolio weighting, so it leaves those implementation details open.

Key ideas

• Annual simple returns are calculated by compounding period returns across the year.
• Log returns aggregate by summing the log price relatives for each period.
• The question concerns weekly data, where the two annual return measures may diverge materially.
• The document does not provide a recommendation or evidence favoring either measure for factor construction.

Tags

# Log returns vs normal returns with weekly prices


# Log returns vs normal returns with weekly prices












I am constructing equity factors and I am given weekly prices for several thousand stocks. Every year the portfolio should be rebalanced, so I am always calculating the returns for a single year. Now I am unsure, whether to use normal returns: $$\prod (\frac{Price_{t}}{Price_{t-1}}) -1$$ or log returns for that matter: $$\sum \ln \frac{Price_{t}}{Price_{t-1}}$$

As far as I know, log returns are normally used for daily returns, since the deviations are not as high, since when I use it on weekly returns, the discrepancy between both methods is quite significant. I appreciate any feedback.

Kind regards, semicolon

Shown in full with attribution under the source's licence. Licence: CC BY-SA 4.0 (Stack Exchange)

This summary was written by Stratmill's research agent from the original; it is not a copy of the source.