Portfolio Risk : Variance-Covariance Matrix and Correlation Analysis

Risk (Part 3) – Variance and Covariance Matrix

The previous chapters established how variance measures the risk of an individual stock and how covariance captures the relationship between two stocks. This chapter brings those tools together to analyse risk at the portfolio level, which is where the practical application of these concepts becomes most valuable for any investor engaged in equity investment.

The path through this chapter follows a clear sequence. First, the Variance-Covariance Matrix is constructed for a portfolio of multiple stocks. From there, a Correlation Matrix is derived to make the figures more interpretable. Finally, these results feed into the calculation of overall portfolio variance, which is the ultimate objective of this entire analytical thread. Portfolio variance tells an investor precisely how much risk they are carrying by holding a particular combination of stocks in the stock market.

This chapter will also lay the groundwork for two related topics that follow in subsequent chapters: asset allocation and its effect on portfolio risk and returns, and an introduction to Value at Risk.

What Exactly Is the Variance-Covariance Matrix?

Before proceeding, it is worth addressing a question that arises naturally at this point. Is the Variance-Covariance Matrix a combination of two separate matrices, one for variance and one for covariance, or is it a single unified matrix?

It is one matrix, but it contains both types of information. For a portfolio of five stocks, the matrix presents the variance of each individual stock alongside the covariance between every possible pair of stocks. The diagonal of the matrix, running from the top left to the bottom right, contains the variance figures for each stock. Every other position in the matrix contains the covariance between the two stocks corresponding to that row and column.

For those seeking a visual and intuitive explanation of matrix multiplication before proceeding, the Khan Academy resource at https://www.youtube.com/watch?v=kT4Mp9EdVqs is a reliable starting point.

Building the Matrix: A Five-Stock Portfolio

To keep the exercise manageable without sacrificing clarity, the example here uses a portfolio of five stocks rather than the 10 to 15 holdings that a typical retail investor might carry. The logic and methodology scale directly to larger portfolios, but beginning with five stocks keeps the matrix at a 5 by 5 dimension, which is far easier to follow without becoming overwhelming.

The five stocks selected for this illustration span genuinely different sectors of the Indian economy.

A pharmaceutical company. A telecommunications firm. An entertainment and leisure operator. A multiplex cinema chain. A specialty chemicals and pharmaceuticals manufacturer.

This cross-sector selection is deliberate. Stocks from different industries tend to respond differently to the same economic conditions, which means their covariance figures are likely to vary considerably. This variation is precisely what makes the matrix informative.

The size of the Variance-Covariance Matrix for any portfolio is determined by the number of stocks it contains. For a portfolio of k stocks, the matrix dimensions are k by k. A five-stock portfolio therefore produces a 5 by 5 matrix containing 25 individual cells. Of those 25 cells, five along the diagonal contain variance figures and the remaining 20 contain covariance figures between pairs of stocks.

It is worth noting that the matrix is symmetrical. The covariance between stock one and stock three is identical to the covariance between stock three and stock one. This means that whilst 20 cells contain covariance figures, only 10 of those represent unique pairings. The symmetry of the matrix is a property that becomes useful when performing matrix multiplication in subsequent steps.

From Variance-Covariance to Correlation

The Variance-Covariance Matrix, once constructed, presents a challenge in interpretation. The raw figures are difficult to compare directly because they are expressed in squared units and their magnitude depends heavily on the scale of the underlying return figures. A covariance of 0.00005 between two stocks does not immediately convey whether the relationship between them is strong or weak.

This is where the Correlation Matrix becomes essential. By standardising the covariance figures using the standard deviations of the individual stocks, the correlation matrix converts each covariance into a correlation coefficient. Correlation coefficients always fall between minus one and plus one, making them directly comparable regardless of the stocks involved.

A correlation of plus one indicates that two stocks move in perfect synchrony. A correlation of minus one indicates that they move in perfect opposition. A correlation close to zero indicates that their movements are largely independent of each other. For portfolio construction purposes, holdings with low or negative correlations provide the most meaningful diversification benefit.

Why Portfolio Variance Is the Ultimate Objective

Individual stock variance and pairwise covariance figures are intermediate steps. The destination is portfolio variance, which measures the total risk of the entire collection of holdings taken together. This figure accounts not only for how risky each stock is individually but also for how the stocks interact with one another.

A portfolio of individually high-variance stocks can exhibit surprisingly low portfolio variance if those stocks carry negative or near-zero covariances with each other. Conversely, a portfolio of individually low-variance stocks can carry substantial portfolio risk if all the holdings move together in the same direction under adverse conditions.

This is the central insight that the Variance-Covariance Matrix makes quantifiable. It is also the reason why stock selection alone is insufficient for sound risk management. The relationships between holdings matter as much as the individual characteristics of each stock.