Replication of Fama & French (2015) - A five factor asset pricing model

I replicate the five Fama-French factors on the monthly frequency. The results of my replication in terms of correlation are as follows:

1. SMB: 97.19%
2. HML: 95.15%
3. RMW: 91.81%
4. CMA: 97.30%

The "Cumulative returns" folder contains a visual comparison of the replicated factors with the original by plotting their cumulative returns from July 1963 to December 2021.

FormatCRSPdata.py

I use the CRSPSift application for Windows to extract the following data from the CRSP monthly tape:

1. RET: total return
2. PRC: end-of-period price
3. SHROUT: number of shares outstanding
4. EXCHCD: exchange code
5. SHRCD: security share code
6. PERMNO: security identifier
7. PERMCO: company identifier

I extract everything for the period Jan 1926 to Dec 2021 and then I subset it.

CreateFirmCharacteristicsFF5Dataset.py

I extract all data necessary to construct Book-to-Market, Operating Profitability and Investability variable as described in the paper from Compustat. The only filter I use is that items must be reported in USD. Then I proceed to define the aforementioned variables and merge them with market equity data from CRSP.

FamaFrench2015FF5.py

I subset the data to include only common ordinary shares (SHRCD = 10, 11) that trade in NYSE, AMEX and NASDAQ (EXCHCD = 1, 2, 3) after June 1963. Then I proceed to construct the factors using the methods of PortSort. Check https://github.com/ioannisrpt/portsort for more details.