Skip to content
All library documents

Using Annual Carbon Data in Monthly Fama–MacBeth Regressions

Article Quant Q&A · Author: Jane
Read the original

Summary

The document frames a panel regression question about estimating a carbon risk premium with the Fama–MacBeth method. The proposed monthly cross-sectional setup predicts next-period excess returns from size, book-to-market, and carbon intensity measured in the prior period. The central data-frequency issue is that returns and conventional firm characteristics are monthly, while carbon intensity, defined as emissions divided by revenue, is annual.

The author asks whether carrying the same annual carbon intensity observation through each month of the year is sensible. No answer, empirical results, or timing convention is provided. A valid implementation would need to ensure the annual emissions and revenue figures were publicly available at each portfolio formation date, and consider reporting lags and stale measurements; otherwise information from the future could enter the regression. The text does not discuss those safeguards or evaluate alternative update schemes.

Key ideas

  • The proposed Fama–MacBeth regression uses monthly next-period excess returns and firm characteristics from the prior period.
  • Carbon intensity is available annually while returns, size, and book-to-market are monthly.
  • The document asks whether annual carbon intensity can be held constant across the intervening months.
  • Correct timing depends on when annual emissions and revenue data became available, but the document gives no answer or empirical test.

Tags

EquitiesFactor investingStatisticsBacktesting
Full text
# Estimating risk premium with cross sectional regression


# Estimating risk premium with cross sectional regression












I am trying to estimate a carbon risk premium according to the Fama & MacBeth methodology using a cross-sectional regression approach. Therefore, I regress the excess return in period t+1 on the size, B/M, and carbon intensity in period t.

Now, I have the problem that my carbon intensity(=carbon emissions/revenue) is data with yearly frequency whereas my returns and size and B/M are monthly data. Does the approach make sense using the same value of carbon intensity throughout each month of the year?

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.