In 1952, Harry Markowitz published a 14-page paper that won a Nobel Prize. The core insight was elegant: what matters for a portfolio is not the risk of each individual asset, but how the assets move in relation to each other. Diversification is not about owning many things | it's about owning things that don't all fall at the same time.
Modern Portfolio Theory (MPT) is the mathematical formalisation of this insight. It's taught in every finance course, referenced in every investment mandate, and embedded in every portfolio optimisation tool. Understanding it | including its real limitations | is essential foundation for Path 4.
The core insight: correlation and diversification
Consider two stocks. Stock A returns 15% with 20% annual volatility. Stock B also returns 15% with 20% volatility. If you combine them 50/50, what do you get?
- If A and B are perfectly correlated (ρ = +1): portfolio volatility = 20%. No benefit from combining them.
- If A and B are uncorrelated (ρ = 0): portfolio volatility ≈ 14.1%. You get the same return with meaningfully less risk.
- If A and B are negatively correlated (ρ = -1): portfolio volatility = 0%. Perfect hedge | returns cancel out the risk entirely.
This is the mathematical heart of diversification. The lower the correlation between assets, the more risk-reduction benefit you get from combining them. The expected return of the combination is always just the weighted average of individual returns | but the risk is always less than the weighted average, as long as correlation is below +1.
The efficient frontier
Markowitz showed that for any given set of assets, there's a curve of portfolios that offer the maximum return for each level of risk. This is the efficient frontier. Portfolios on the frontier are "efficient" | you can't get more return without taking more risk, or reduce risk without giving up some return.
The practical implication: a well-diversified portfolio should sit on the efficient frontier. If your portfolio is below the frontier, you're earning less than you should for the risk you're taking | you need better diversification.
For systematic factor investors, the efficient frontier concept translates directly to factor diversification | the reason combining Momentum and Value works. They have low correlation, so combining them moves your factor portfolio closer to the efficient frontier than either factor alone.
What MPT gets right
What MPT gets wrong | the practical limitations
The practical lesson: Use MPT's conceptual framework | think in terms of correlation and diversification | but don't use mean-variance optimisation with historical inputs to set portfolio weights. The inputs are too noisy. Instead, use simpler, more robust weighting schemes (equal weight, inverse volatility weight) that don't depend on precise correlation estimates.
What actually matters for systematic portfolio construction
From 17 years of building systematic strategies, here are the MPT-derived principles that genuinely hold up in practice:
- Diversification across factors, not just stocks | owning 30 momentum stocks gives less diversification than owning 15 momentum + 15 quality stocks, because the factors have lower correlation
- Correlation matters more than individual stock selection | two moderately good factors with low correlation typically outperform one great factor on a risk-adjusted basis
- Don't overweight concentrated sector bets | even a well-run momentum strategy can become 50%+ concentrated in one sector. Sector caps (max 30% in any sector) add robustness without significant alpha cost
- Volatility-based position sizing works better than equal weight in volatile markets | giving lower weight to high-volatility stocks produces smoother returns than equal weight, because it prevents a few extreme movers from dominating the portfolio
RupeeCase strategies use equal weighting as the default (simplest, most robust) with an option for inverse-volatility weighting (gives lower allocation to high-volatility stocks). Sector concentration is monitored and flagged when a strategy's portfolio exceeds 35% in a single sector. The full correlation matrix of factor returns is available in the strategy analytics, so you can see how diversified your current portfolio actually is. Available at invest.rupeecase.com.
Why MPT under-delivers on Indian data specifically
Markowitz built MPT on US large-cap data with relatively stable correlations and deep liquidity. Indian equity markets break two of those quiet assumptions, which is why naive mean-variance optimisation often produces portfolios that look elegant in the spreadsheet and perform poorly in practice.
Correlation regime breaks. Indian equity correlations are not stationary. In normal markets, large-cap and mid-cap correlation runs around 0.7. In a sell-off triggered by FII outflows, that correlation jumps to 0.92 or higher. The diversification benefit you optimised for evaporates exactly when you need it. Mean-variance solutions that treat the long-run correlation as the working number are over-optimistic about diversification in stress. The fix in practice is to use stress-period correlations (or a blended estimate) when deciding allocation, not the full-history average.
Liquidity and impact-cost asymmetry. The MPT efficient frontier ignores the cost of getting in and out. On Nifty 50 names, the round-trip impact at sensible size is 5 to 15 basis points. On Nifty Smallcap 250 names, the same trade can cost 100 basis points or more. A mean-variance optimiser handed the entire Nifty 500 universe will happily over-allocate to small caps with high backtest returns. Live, the trades to maintain those weights eat the edge. The fix is a liquidity floor. Most disciplined Indian systematic books cap the impact cost per name at a hard ceiling (say 50 bps round trip) and exclude anything that fails the screen.
FII concentration risk. India's free-float is dominated by foreign holders in many large caps. A risk-off move in the Federal Reserve cycle, an INR weakening event or a macro shock can pull capital out across the entire FII-heavy basket together. Stocks that look unrelated by sector show up as correlated when the trigger is a single rates move. A pure sector-diversification frame misses this; an FII-holding-aware overlay catches it. The shareholding pattern files filed quarterly with NSE and BSE are the data source.
Mean-reversion in factor returns. Indian factor returns mean-revert harder than US data suggests. A momentum factor that underperformed for 18 months has historically been close to its inflection point, not its decline. Mean-variance optimisation done on a trailing window over-weights factors that just performed well and under-weights those mid-drawdown. The honest MPT solution in Indian markets is to pair the optimiser with a regime overlay, or to fall back on equal-weight blending for the practical robustness it delivers despite the theoretical inefficiency.
The payoff is not that MPT is wrong. The core insight stands. The payoff is that the inputs and constraints have to match the market. A Nifty 500-aware optimiser with stress-period correlations, a liquidity floor and a regime overlay produces portfolios that survive the moments the textbook version blows up in.
Glossary
Sources & further reading
- → Markowitz (1952) — Portfolio Selection, Journal of Finance
- → Markowitz, H. (1959). Portfolio Selection: Efficient Diversification of Investments. Wiley.
- → DeMiguel, V. et al. (2009). Optimal Versus Naive Diversification. Review of Financial Studies. (Shows equal weight often beats optimised portfolios)
- → NSE India Research — Portfolio Analytics Resources
Quick check, Module 4.1
Efficient Frontier Point
Compute the expected return and volatility of any two-asset weight mix. Sweep weight to trace the frontier.