Correlation Conundrums

ETFs are huge in the United States, and are likely to become big in Australia.

The size of the ETF market in the US is mind boggling. The granddaddy of all ETFs, the SPDR S&P 500 (SPY) has approximately $78 billion in holdings. In total, ETFs hold approximately $1 trillion in assets in the US.

The phenomenal growth of ETFs may, while having many positive outcomes for investors, have an unintended impact on how we invest. This process of stock picking has been part and parcel of equity markets ever since the first equities were developed and traded. Stock picking is of course full of risks. The alternative to stock picking is to generally invest in a broad set of securities to achieve diversification.

The risks associated with investing and stock picking brought to the fore the science of portfolio management, beginning with Harry Markowitz’s portfolio theory of mean-variance analysis.

The important contribution of Markowitz was that the key to reducing risk in a portfolio is not so much the risk of the individual security, rather it is the return correlation between pairs of securities that counts.

Since this seminal financial economic work first appeared, there has been an explosion in products and funds that are largely based on the basic precepts of the principles behind portfolio theory. In the contemporary financial markets, index funds, and now broad market-based ETFs are the most prominent manifestations of these theories.

The very products that seek to implement these core principles of portfolio theory and enhance the efficiency of the markets, may be causing substantial distortionary outcomes in both the returns of securities and, significantly, on correlation structures. Evidence is mounting that, with equities at least; increased correlations are the “new normal.”

The implications were clearly articulated by Barclays Capital’s equity research team when stating:

“Equity correlation is close or higher than its high levels scalded during the credit crisis. This has profound implications even for non-derivatives investors in that it indicates that stock-picking skills are less useful in the current environment.”

Research in the US by Birinyi Associates shows that the correlation of stocks in the S&P500 is at approximately at 81%, or in other words, the average S&P500 stock in the US does the exact same thing as the market about 81 per cent of the time.

What is pointing to this sustained, structural shift in the correlation structure?

Many point to the increased volatility of the markets as a reason for the increases in correlation levels. The chart below on the NYSE’s S&P500 suggests that though the standard deviation of the past three or so years has been above the long term mean, it is by no means unusual.

Most now believe that the shift to ever higher correlation levels is largely being driven by massive changes in the structure of equity markets induced by the widespread adoption of ETFs.

It is a logical argument to make, with the rise of index funds and broad market-based ETFs, the equity fund flows associated with these products would, logically, lead to a secular increase in equity correlation.

What is the impact of increased correlations? If “stock picking” starts to become irrelevant then the traditional role of the broker and similar advisers will be forever changed. Rather than being experts on stocks or mutual funds, they will have to become expert asset allocators on a tactical basis. To a large extent this is already occurring, however one could have reasonably argued that the equity market alone provided enough asset diversification in the past.

Are we witnessing the commoditisation of the equity market, where shares are traded in lockstep like wheat or beef? Differences between individual companies probably mean less now than they did to the performance of their shares.