India Six-Factor Premia, Attribution & Regime Analysis
Overview
Indian equities are not a scaled copy of US markets. Large, mid, and small caps can diverge sharply around credit events; quality-oriented segments often hold up better in stress; momentum can deliver strong long-run premia but also violent drawdowns; and two funds marketed with similar labels can post very different paths once factor exposures are unpacked.
This project applies a Fama–French-style six-factor framework to India: market (MF), size (SMB5), value (HML), profitability (RMW), investment (CMA), and momentum (WML). We use monthly data over a multi-year window, align factors with Nifty style benchmarks, and run the same models on a small set of Indian equity mutual funds so you can see both index-level and fund-level results in one place.
The work is aimed at investors and researchers who want evidence, not anecdotes: Which factors earned positive long-run premia in India? How correlated are they? How much of a Nifty 200 Value or Smallcap 250 return is explained by market beta alone versus a full factor model? Do funds earn alpha after controlling for known risks? Does momentum interact with value or the market factor? And does the quality factor (RMW) still pay when the market factor is negative?
Interactive exhibits in the factor-model and what-you-will-see sections organize results by theme — market puzzles, factor premia, index and fund attribution, momentum risk, and a formal hypothesis test.
Why factor models for India?
Single-index benchmarks (e.g. Nifty 50) hide large dispersion across cap tiers and style tilts. A mid-cap value fund and a large-cap growth fund may both be “equity” but carry different factor bets. Multi-factor models separate market beta from size, value, profitability, investment, and momentum effects so performance can be compared on a level field.
India-specific research (including work on four- and five-factor models in Indian equities) shows that classical Fama–French structures are informative but not identical to US calibrations — premia, correlations, and crash episodes differ. This study documents those differences with transparent regressions rather than relying on US factor histories as a proxy.
Six-Factor Specification
For asset or index in month , excess return is the linear factor model
MF is the market excess return. SMB5 is a size spread. HML is value (high book-to-market minus low). RMW is profitability (robust minus weak). CMA is investment (conservative minus aggressive). WML is winners minus losers, the 12–1 momentum spread. CAPM is the same equation with every slope except constrained to zero. The four-factor variant frees SMB, HML, and WML and still constrains RMW and CMA. Nesting is the point of reporting all three: the increment in from CAPM to six factors is the fraction of variance those extra premia actually span.
Slopes and are ordinary least squares on the monthly sample. Alphas shown in annualised units are , a scaling of the monthly intercept, not a compounded rate. Rolling 36-month regressions re-estimate the same equation on so a change in is a change in exposure, not a change in the factor definition.
The quality stress test restricts the RMW premium to months in which and asks whether the conditional mean of RMW stays positive. That is a subsample mean, not an interaction term inside the index regression, and the charts label it as such.
What you will see in the results
Market puzzles — cumulative paths for large, mid, and small caps, quality vs benchmark, momentum, fund dispersion, and volatility through major macro events (demonetisation, IL&FS, COVID).
Factor premia — summary statistics, correlations, and long-run cumulative factor returns for the Indian FF6 panel.
Index regressions — how much of each Nifty style index’s return is explained by CAPM, four-factor, and six-factor models; rolling factor exposures on a highlight index.
Momentum — WML profile, drawdowns, interaction with value and market factors, and long-only vs long–short momentum behaviour.
Funds — FF6 attribution for sample Indian equity funds, model fit across specifications, rolling factor exposures, and comparison of total return vs factor-explained performance.
Hypothesis test — whether profitability (quality) earns positive returns when the market factor is negative, with robustness checks.
| Factor | Ann. mean | Ann. vol | Sharpe | Max DD |
|---|---|---|---|---|
| MF | 9.800% | 17.100% | 0.574 | -34.500% |
| SMB5 | 9.700% | 9.100% | 1.065 | -9.900% |
| HML | 7.800% | 14.200% | 0.550 | -26.000% |
| RMW | 3.700% | 7.800% | 0.470 | -22.600% |
| CMA | -6.600% | 10.200% | -0.644 | -52.700% |
| WML | 5.200% | 14.500% | 0.359 | -34.100% |
| MF | SMB5 | HML | RMW | CMA | WML | |
|---|---|---|---|---|---|---|
| MF | 1.000 | 0.152 | 0.390 | 0.031 | -0.533 | -0.052 |
| SMB5 | 0.152 | 1.000 | 0.260 | 0.081 | -0.067 | -0.040 |
| HML | 0.390 | 0.260 | 1.000 | 0.186 | -0.346 | -0.315 |
| RMW | 0.031 | 0.081 | 0.186 | 1.000 | -0.081 | -0.206 |
| CMA | -0.533 | -0.067 | -0.346 | -0.081 | 1.000 | 0.133 |
| WML | -0.052 | -0.040 | -0.315 | -0.206 | 0.133 | 1.000 |
| Fund | α (ann.) | R² | β MF | β SMB5 | β HML | β RMW | β CMA | β WML |
|---|---|---|---|---|---|---|---|---|
| ('Fund_A', 'NIFTYBEES.NS') | 5.500% | 0.947 | 0.965 | -0.043 | -0.020 | 0.115 | 0.203 | -0.032 |
| ('Fund_B', 'JUNIORBEES.NS') | 2.600% | 0.903 | 1.047 | 0.344 | -0.044 | -0.172 | 0.072 | 0.079 |
| ('Fund_C', 'MOM100.NS') | 0.300% | 0.885 | 0.972 | 0.709 | -0.042 | 0.070 | -0.102 | 0.023 |
Puzzle 1 — Does size matter?
Puzzle 4 — Same category, different journeys
Puzzle 5 — Risk and return
| Index | CAPM R² | β MF | α (ann.) |
|---|---|---|---|
| ^Cnx100 | 0.958 | 0.934 | 3.300% |
| ^Nsmidcp | 0.871 | 1.030 | 4.300% |
| Smallcap.Ns | 0.138 | 0.262 | 35.500% |
| ^Nsei | 0.931 | 0.913 | 3.200% |
| Factor | Style proxy | Correlation |
|---|---|---|
| SMB5 | size_spread | 0.025 |
Nifty style indices — cumulative returns
WML cumulative return
Drawdown profile
| Start | End | Max DD |
|---|---|---|
| 2020-04-30 | 2023-11-30 | -34.100% |
| 2024-08-31 | 2026-11-30 | -19.700% |
| 2019-02-28 | 2019-06-30 | -11.400% |
| 2019-10-31 | 2020-02-29 | -10.800% |
| 2016-11-30 | 2017-01-31 | -7.100% |
36m rolling corr: WML vs HML
36m rolling corr: WML vs MF
Momentum strategies typically require frequent rebalancing; higher month-to-month variation in WML implies greater turnover and implementation cost in live portfolios.
Fund cumulative returns
| Fund | Total ret. (ann.) | OLS α (ann.) | FF6 R² |
|---|---|---|---|
| ('Fund_A', 'NIFTYBEES.NS') | 12.600% | 5.500% | 0.947 |
| ('Fund_B', 'JUNIORBEES.NS') | 14.100% | 2.500% | 0.903 |
| ('Fund_C', 'MOM100.NS') | 16.400% | 0.300% | 0.885 |
Rolling 36m exposures — ('Fund_A', 'NIFTYBEES.NS')
Rolling 36m exposures — ('Fund_B', 'JUNIORBEES.NS')
Rolling 36m exposures — ('Fund_C', 'MOM100.NS')
| Test | Stress n | Stress mean (m) | t-stat | p-value |
|---|---|---|---|---|
| RMW factor (full FF6 panel) | 54 | 0.600% | 1.734 | 0.089 |
| RMW in bottom-quartile MF months (robustness) | 32 | 0.500% | 0.956 | 0.346 |
- Look-ahead bias if fundamentals are not lagged.
- Multiple testing across many factors inflates false positives.
- Short stress windows reduce statistical power.
- Long-only quality indices differ economically from long-short RMW.
What the Regressions Identify
A higher means the factors span more of that index’s variance. It does not mean the index earned a premium. The intercept is the average return left after the estimated exposures are subtracted. A fund with a large SMB loading and a zero intercept earned the size premium; it did not earn an additional return beyond that premium.
Compare CAPM with six-factor on the same index. If almost all of the fit arrives with MF alone, style labels on that benchmark are mostly market beta. If HML or WML moves the fit and the rolling beta, the style label is visible in the covariance, which is the claim a factor attribution is supposed to test.
Momentum crash panels plot the WML drawdown path next to the market. Clustering of WML losses in rebounds, rather than in the decline, is the Daniel–Moskowitz pattern. It can appear as a negative loading of WML on a “down-market rebound” indicator even when the unconditional WML mean is positive.
Limitations
Indian factor series constructed from public index and return data are not a Ken French India library. SMB, HML, RMW, CMA, and WML definitions follow the standard long-short logic but inherit whatever universe and rebalance the pipeline uses. A different breakpoint for “small” or a different skip month will move the premium.
Monthly OLS standard errors ignore overlapping macro shocks and the fact that several indices are being read at once. Treat -statistics as descriptive. Fund attribution uses a short list of schemes, not the population of Indian equity funds. Results are research documentation, not a product comparison or investment advice.