Trading in Risk
Following the the 1987 stock market crash, Federal Reserve Chairman Ben Bernanke signaled to financial markets that the Federal Reserve would backstop further trading losses. This was the famous “Burnanke Put” that has been in effect continuously since 1987.
The result of the Bernanke Put has been volatility suppression that has offered financial traders narrow bands for trading, entirely decoupled from the economy. The “risk” trade–or “carry trade” involves borrowing money from low interest rate money markets (such as repo markets, etc.), then leveraging the low-interest money into higher return investments. As a result of the leverage, traders could book much higher profits from the risky investments–so long as money markets remained stable. But when money markets periodically broke (such as in 2008, 2019), risk trades dried up and equity markets crashed, resulting in very large losses.
Portfolios based on diversified multifactor strategies can offer strong performance, while minimizing the risk of large losses.
But it’s critical that any multifactor strategy be based on common sense, emphasizing logical and explainable factors. In our view, over-reliance on statistical analysis (Modern Portfolio Theory and its variants) or data mining analysis (traditional factor investing) fail the common sense test, for the following reasons:
Statistical analysis relies on the assumptions of a “normal” distribution of market data from a mean (average); and standard deviations from that mean. In a normal distribution, the vast majority of data are grouped closely around the mean and two or three deviations from the mean, with only a few data points distributed into “long tails” of the distribution curve. Because the few data points relegated to the long tail of the distribution curve are “statistically unlikely,” they are ignored in statistical models. And yet, all recent shocks to the financial system have been triggered by these very same long-tail data points, otherwise known as “black swans.” Statistical models have no way of predicting or modeling black swan events; yet it is these statistically improbable data points that drive value changes in volatile and turbulent markets.
data mining analytics
Data mining analytics gathers very large data sets related to multiple factors, then attempts to correlate data points that indicate results based on certain factors emphasized in the data. This is an impressive approach–in theory.
But in practice, it is almost impossible to know how to weigh certain factors. Of course, in back-testing, one can more easily identify factors that appear to drive a result. But back-testing is not forecasting. It is entirely likely that back-tested results may be explained by an unseen, or improperly-weighted factor, that receives minimal (or no) attention in the back-tested data.
So, factor analysis based on data mining cannot be relied on to predictably forecast results looking forward.
In our view, multifactor analysis offers great promise for forecasting results, but using statistical or data mining tools, alone, is not the way to achieve reliable results. The answer to avoiding the limitations of computer models is to use old fashioned reasoning techniques that are easy to explain, challenge and verify, as follows:
identify likely factors that inter-relate and create a cause-and-effect thesis that tends to explain the inter-relationship based on common sense, logic and observable data.
Basically, this is the way Warren Buffet would likely approach the multi-factor problem–by setting up an observable and testable hypothesis that can be explained using human logic and common sense.
Tidepool focuses on portfolio selection that implement our core philosophy of minimizing the risk of large losses, as required by ERISA, while obtaining performance in any market. We believe the best results are obtained through a properly-weighted and well-understood multifactor approach.
The following diagram shows the way we think about how to minimize liquidity risk in fixed income markets.
The internal variables in this diagram are: asset quality (always the most important predictive factor), duration, leverage, market depth and transparency. These are the factors that a plan can, and should, take into account in selecting from among the thousands of available fixed income investing options.
The external inputs in this diagram are: interest rates, inflation, legal / regulatory concerns, corporate actions and collateral requirements. Any or all of these external factors can affect the value of a fixed income investment, as a change in external inputs works its way through the liquidity risk cycle, shown below:
Acting through our investment advisors, we select among thousands of available funds and ETFs that best capture a proven multifactor investing strategy, accounting for the variables and inputs shown in the above diagram. Funds and ETFs that implement this strategy are identified and made available to plan participants through our investment lineups.