Fama French and Momentum Factors: Data Library for Indian Market

Changes in data sources

Compared to the legacy releases, we have made the following changes in our data sources:

1. We are now running all the computations ourselves while for the legacy releases, the computations were carried out by CMIE using the methodology provided by us.
2. We are now using CMIE’s Prowess DX which provides three data releases every year (March, September and December). Accordingly, we also plan to have three releases every year.
3. We now cover both BSE and NSE listed companies while the legacy releases used data only from BSE and therefore excluded the companies which are exclusively listed on NSE. Also for stocks listed on both exchanges, we pick up prices from the exchange with greater aggregate liquidity.
4. While legacy releases computed total returns by adjusting the price returns for the dividends paid by the company, we now rely on the total returns field available directly in Prowess DX.

We also took advantage of the break in the series to make some improvements and refinements in the methodology.

1. Based on feedback from market participants and researchers, we have eliminated micro-cap stocks and penny stocks that are not part of the investable universe for most investors. Specifically, we apply two filters:
   • The market cap filter excludes all stocks with a market cap during the preceding year below 10% of the median market cap.
   • The price filter excludes stocks whose median price during the preceding year is less than INR 1.00.
   For example, using the data for 2019-20 (October to September) to set the filters that will be used for computing the factor portfolios and returns for the year 2020-21 (October to September):
   • The micro cap cutoff came to INR 109.5 million and excluded the bottom 25.8% of the universe of stocks.
   • The penny stock filter eliminated 5.3% of stocks from the universe.
   • As the stocks that fall under the two criteria overlap significantly, the two filters together excluded 26.8% of the stocks.
2. Some double-sort portfolios (especially the Large-Cap Value portfolio) have very few stocks in some months. Since factor portfolios need at least a minimal degree of diversification, we do not allow any double-sort portfolio to have less than five stocks in any month. If the Large-Cap Value portfolio is absent after applying this rule, the value factor is computed as the difference between Small Growth and Small Value.
3. We found that our survivorship bias algorithm has not identified any “vanishing companies” after the mid 1990s. This is not surprising in the light of the regulatory reforms that took place after the mid 1990s “vanishing companies” episode. We have therefore discontinued the ongoing running of this algorithm, but we do provide survivorship bias free files that correct for the mid-1990s “vanishing companies”.

The data in the new series does not therefore exactly match the data in the Legacy Releases and in our Working Paper, but the correlations between the two series are very high and the discrepancies are attributable to the more comprehensive data sources and the improvements in the methodology as explained above.