One common measure, "Jensen's Alpha" subtracts index return from stock return. For example, the Nifty is up 64 per cent in the past year, while Infosys is up 89 percent.
Hence, Infosys' alpha is +25. Some risk-related measure, usually standard deviation of return, which highlights volatility, is used as well.
An ideal portfolio has high returns with low standard deviation (SD). To beat the market, it needs positive alpha and low correlation to market (to ensure positive returns in bearish phases).
The standard methods of portfolio creation involve checks of correlation of each asset with market, cross-correlations (covariance) between assets, and "efficient frontier" (EF) analysis.
If return versus risk is mapped for a set of assets (or portfolios), the efficient frontier is the curve where the maximum return is available for a given risk. Or to put it another way, the EF indicates the minimum risk required to generate a desired return.
If the same basket of assets is combined in different weights, the return:risk profile will vary. EF analysis helps to work out the weights of a given basket of assets for a required return-risk profile.
It's convenient to use historical data for such calculations. But last year's high-alpha assets are not necessarily going to be the best picks of the next 12 months. Linear projections of historical CAGR and SD can cause large errors and false expectations.
Sophisticated quants make forward projections of risk, return and asset-correlation. They use historical data. But via Monte Carlo (MC) simulations, a randomness that is closer to actual volatility is introduced.
In each run, a MC maps random sets of returns. After many iterations, an MC will calculate many possible return:risk pairs along with associated probabilities of their occurrence. Even simple MC models can be more useful than linear CAGR projections.
Another de-risking strategy is a so-called "market neutral" portfolio. Most MN portfolios involve opposed long-short trades. A decent MN will yield positive returns most years.
But MN under-performs in big bull markets. Also, long-short plays are hard to manage, as well as being expensive. MN involves frequent trading and complex margining.
Can one build an MN portfolio from pure buy and hold assets? As mentioned above, a portfolio of assets that have positive alpha, low covariances with each other, and low correlation with the market index, may generate positive returns even in a bear-market.
Geoff Considine, who heads Quantext, a pioneer at creating forward-looking portfolio management solutions for individuals, claims it is very possible for individual investors to create MN portfolios consisting purely of long-term stock (or mutual fund) holdings. He offers some simple rules.
Pick profitable stocks with lower price-earnings than the index average PE. Each stock's monthly return should have low correlation with index return. There should be low covariance within the basket. Use MC to project correlations, covariances, SD, CAGR, etc. Find the efficient frontier in terms of weights for this basket of assets.
Considine's studies are US-specific. But there is no reason why this could not work in India. Some details such as the ideal number of stocks are a question of choice.
Other tricky details such as acceptable covariance levels, weights, etc., are optimised by data-crunching. The number crunching can be done using spreadsheets, or by writing simple programs. The data is widely available.
Examining Considine's rules, a value-investing fundamentalist's ideal portfolio is probably a close initial fit. The first criteria is common.
In turn, consistently profitable low PE stocks will include many with low market correlation. Many value-investors also instinctively pick businesses with low covariance & Keynes called this holding "opposed risks"
However, few fundamentalists actually rigorously check covariance and look for the EF. Conversely, few quants look at value-seeking fundamental criteria. The evidence suggests that combining the quant and fundamentalist approaches can be used to generate higher returns for lower risks.
At the least, it's worth running historical CAGR, covariance and SD numbers on any portfolio you may hold. Even an efficient frontier analysis based on the historical record may throw up insights about weights. An MC-based projection could do better.
If Considine is right, and his arguments are compelling, even small tweaks in weights could result in markedly improved portfolio performance along with insulation from market cycles. It is definitely an avenue worth exploring.