Anatomy of a Market Crash (Part One)

December 12, 2019
Domestic Equities Valuation

Anatomy


A 1,000-Foot View

It is well understood that the macroeconomic environment and business cycle heavily influence financial markets. For instance, collective wisdom suggests that recessions are correlated with underperformance in risky assets. However, the evidence does not confirm this assertation, at least not on the surface. For example, in the financial crisis, the U.S. economy was still contracting through the end of 2009, with a -0.3% year over year growth of real GDP (Gross Domestic Product) in the fourth quarter, while the equity markets were up nearly 30%, respectively. So why is there a disconnect?

Like my theories on macroeconomics, financial markets generally follow a rhythm. While the rhythm is not always the same, there is a pattern that is associated with the overall performance of different financial markets. This often correlates with the business cycle, and for that reason, we spend a lot of resources in researching the business cycle. But some key relationships make the equity market different, inspiring us to write a piece on the anatomy of a market crash.

Structurally, it is our opinion that market crashes often lead to recessions and are correlated with these recessions. But causality of this correlation is not directly linked, instead is indirectly linked. For us, it does not make sense why a typical slowdown of 2% to 3% contraction in real GDP would destroy wealth in the financial markets of 40% to 50%, respectively.

We attempt to answer this disconnect with what we call the anatomy of a market crash or the causality of a market crash to identify important factors that are often associated with sizeable meaningful market events. Examples would include the tech bubble in 2000 and the financial crisis in 2008. We will approach this series as we do with our macroeconomic analysis starting with long-run drivers and slowly working down to short-run factors. Throughout this series, we will be analyzing three distinct time frames: long-run structure, intermediate (business cycle), and the short-run outlook. Each time frame and associated factors interreact with equal importance.

Part one will focus on the long-run structure of the domestic equity market, with drilling down to other time frames in subsequent articles. By the end of this series, we hope to provide clarity to the market that is infamous for incomprehension. The world is one significant math problem, and we aim to write the formulas to make sense of it.

Part One: Structural or Long-run Risk and Valuation Anchors

The safest place to start in our research is the academic axiom about what moves equity prices. “Equity prices depend primarily on the underlying earnings of companies. More specifically, they depend on how the market currently values the expected future earnings stream of a company.” (McGee 20151) Its future earnings expectation values the stock market, or in other words, stocks move relative to their valuation levels.

From a long-run perspective, this boils down to a valuation anchor. Various methods valuation methods exist, such as dividend yield, price-to-value ratios, and price-to-earnings ratios, etc. We prefer “Buffet Indicator” for long-term valuation, which takes market capitalization and presents it as a percentage of nominal GDP. The theory is that financial markets should grow in line with the output of the economy. Any variation above or below the long-run averages identifies over or undervaluation periods.

Back in 2001, Warren Buffet explained to Forbes that his indicator is “probably the best single measure of where valuations stand at any moment.” (Patton 20152) The fact that you have the “Oracle of Omaha” making such a statement should validate the method, but we will double-check the math anyway. We confirm his theory by testing our version of the Buffet Indicator over all the available sample data that we have.

The original Buffet Indicator uses corporate equities market capitalization. However, this data is reported quarterly and is hard to use in the real world with that amount of reporting lag. To account for this time delay, we substitute corporate market cap with the Wilshire 5000 Full Cap Price Index, the broadest domestic price index weighted by market capitalization. This modification allows us to use the theory of the Buffet Indicator in real-time. We call this the MBI (Modified Buffet Indicator). We have seen other sources use this same method in addition to us, so we will not name it the “Soderstrom Indicator,” which has a nice ring to it.

Table 1: Valuation Rubrics
Ratio Valuation
Ratio < 50% Significantly Undervalued
50% < Ratio < 75% Modestly Undervalued
75% < Ratio < 90% Fair Valued
90% < Ratio < 115% Modestly Overvalued
Ratio > 115% Significantly Overvalued

With the current Wilshire 5000 Index over 140% of nominal GDP, we consider the market valuation significantly high.

The current valuation suggests that in the long run, a market event that is large enough to affect earnings might cause a reversion of valuation levels towards their mean or below their average. The market tends never to stay fair valued for long. We classify the valuation level using a simple rubric (Table 1) based on detrended data and deviations from the mean.

The last time we had valuation levels, this high was the tech bubble in 2001, ending the secular bull market starting in the 1980s and started a period known as the lost decade. That was 10-years of no returns for passive investors, which seems to be in the cards again!

Check the Math


You might think that we are crazy to suggest that there is a chance of another lost decade, defined as an average return of near zero over the next 10-years. Please ask yourself the same question at the end of this article.

To test the ability of the Modified Buffet Indicator to forecast returns, we decided to put it through a series of regression analyses with time being the variable that we change. The regression is the levels of the MBI compared to the annualized returns of the S&P 500 Index for X years into the future. The annualized return is the total return for the X years into the future, starting at the test variable of MBI. If the MBI variable is the January 1, 1970 sample and X variable tested is 10-years, then the annualized return would be calculated from January 1, 1970, to January 1, 1980.

By changing X in the process, we were able to determine if there was any linear or non-linear relationship with the MBI and future returns of the stock market represented by the S&P 500 Index. We were searching for a high r-squared value, which explains the variation in the regression analysis. The higher the r-squared value, the better the model did at forecasting the future.

What we have found from this study was positive and negative. The good: the long-run Modified Buffet Indicator can explain almost 80% of the variability of 10 to 15-year return variation (Figure 2). The bad: it is not a crystal ball and did a very poor job forecasting 5-years or less.

Thus, making investment decisions on valuation alone is not recommended and should be avoided. The market tends to remain overvalued or undervalued for extended periods, and sometimes longer than one can stay solvent. However, valuation does have a good track record for forecasting returns over a long period.

The optimal forecast period (10-15 years) is around the same length as the average business cycle over the sample from 1970 to today. The shorter the time frame, the less valuation explains the variation. For this reason, the 10-year model for MBI fair value sets the structural long-run valuation anchor. The deviation around the anchor from one extreme to the other is followed by a reversion back towards the mean (i.e., a market crash). Our analysis shows the optimal forecast range at 13 years, however merely explaining the model, we set our secular valuation forecast at ten years.

This mean reversion often associated with valuation drifting too far from its anchor can warn investors if the financial markets are in a period of vulnerability. Meaning an overvalued equity market combined with some shock to the financial system will often correct itself in the form of a market crash. The depth of the drawdown can be magnified depending on how high the valuation is relative to its anchor. Put, the higher you are, the farther you fall type situation.

The data tells us that investing when the market is cheap (undervalued) has typically delivered strong subsequent returns. Investing when the market is expensive (overvalued), it tends to offer below-average returns over the same horizon. It might help to visualize this relationship with a regression plot (Figure 3).

Considering the current MBI is at 140% of GDP and is literally off the charts, a general assumption that the next market crash could be massive destruction of wealth. Regression analysis is tricky, and the inability to forecast returns in the short run using valuation makes it hard to use this information in a timely fashion. It is sobering to think that we are once again at an excessive valuation.

We have some theories to explain why we are at this point that includes growth by excessive debt and low-interest rates to spur reflation this past decade. What matters now is the math and if there is a way to confirm this analysis.

Re-checking the Math


After discovering how compelling the results were, we decided additional analysis was warranted. So, we dove deeper into the study to challenge or confirm the Modified Buffet Indicator with other long-term valuation models.

First, let’s address the drawbacks of using the Wilshire 5000. For a good reason, we want to use the broad market to replicate the Buffet Indicator in real-time, but it comes with one apparent issue: data. We have tested the theory since 1970, but we can explore further by exchanging the Wilshire 5000 Index with an index of a longer track record, the S&P 500 Index. While this change to the model will not have the same percentage as the original, it will allow us to test back to 1929, where the data starts, right before the Great Depression.

The results match better than expected when looking back for an additional 41 years to the model. R squared dropped from .80 to .70 but is still statistically significant. Furthermore, the valuation model is at 14% of GDP, suggesting the next decade will have approximately an average annualized return of zero, matching the forecast using the Wilshire 5000 Index.

Rechecking the Math, yet again


It will benefit the analysis if we change direction and test long-run valuation with an entirely different model. The one that stands out is the Shiller CAPE, which is a P/E ratio that is cyclically adjusted for the business cycle, giving a long-term view on P/E ratios. Thanks to the noble prize winner, Dr. Shiller, and his data-gathering ability, we have data going back to 1929.

TThe result shows an r-squared value of 0.41, explaining about 41% of variation over a 10-year horizon. It was exciting to find out that our version of a valuation anchor explaining 70% of the standard error was much better than Dr. Shiller’s 41% over the same variables. We do not usually like to toot our own horn, but because our method is statistically superior over a noble prize winner, “toot-toot”!

As far as the forecast is concerned, the CAPE ratio is currently at 27.21. Suggesting a 10-Year average annualized return of approximately zero, and once again confirms the outlook for the next ten years to be rocky.

Cleaning up the CAPE ratio model a little by rerunning the analysis in modern times and matching the time frame of the Wilshire 5000 Index (starting in 1970), shows us r-squared values back in the higher ranges (Figure 6). The r squared result increased back to the expected range to 0.69. Forecasting the next ten years to have an average 3% annualized return. Still lower than the long-term average of 8% and below periods with a lower valuation.

Valuation Anchors and Market Crashes


With our “anatomy of market crashes” process, the valuation anchor used to gauge the vulnerability and likelihood of a market crash over the next 10-years. Because of the inability of valuation anchors to forecast short-term market movements, valuation alone is not enough for decision making.

The valuation anchor is needed in the overall process to monitor and prepare for a market crash of considerable magnitude. Additional math can help explain how low the drawdown might be by following our rubrics.

For example, if the event causing the mean reversion is mild, one could expect a pullback of approximately 23%. On the other hand, if the event is dramatic, like a financial crisis, the mean reversion might drop below the anchor value (fair value) to significantly undervalued, suggesting a drawdown of an estimated 66%. Compression to the mean or anchor value would indicate a reduction in the amount of 49%. (Table 2)

Table 2: Estimated Drawdown Magnitude Based on Rubrics
Ratio Valuation Est..Drawdown.Magnitude
Ratio < 50% Significantly Undervalued -66.61% or less
50% < Ratio < 75% Modestly Undervalued -66.61% to -49.99%
75% < Ratio < 90% Fair Valued -49.99% to -39.99%
90% < Ratio < 115% Modestly Overvalued -39.99% to -23.22%
Ratio > 115% Significantly Overvalued -23.22% or more
Current Ratio: 149.78% Significantly Overvalued

So far, we have discussed multiple models for a thesis around developing a valuation anchor. The valuation model chosen was the Modified Buffet Indicator (MBI), using the Wilshire 5000 index as a proxy for the total market capitalization of the U.S. domestic equity market. Currently, the valuation anchor is at 149.78%, significantly overvalued.

This anchor suggests a macroeconomic event could cause another market crash bringing valuation levels back to the mean. In other words, the market is currently overvalued and is in a vulnerable period for a market shock in the years to come.

In part two of this article, we will shorten our horizon and look at what is needed to take place over this 10-year horizon for a market crash to occur. Part Three, we will look at the short-term horizon to help pinpoint areas of stress in the domestic market and help prepare asset allocation models for shocks.


References:

  1. McGee, Robert T. Applied Financial Macroeconomics and Investment Strategy (Global Financial Markets)

  2. Mike Patton Under the Hood: Stock Market Valuation, Pt. 1: Buffett’s Choice



Omega Icon
Aaron Soderstrom
Chief Investment Strategist
550 Reserve St Suite 190 Southlake, TX 76092
Office 817-500-0556
aaron@omega2capital.com
Omega2wealth.com
Facebook Icon LinkedIn Icon Twitter Icon