Sign up for email updates!

Don't miss out on what matters. Sign up for email updates!

Stay informed! Sign up for e-mail updates:

Wednesday, March 18, 2020

Market Volatility: Echoes from Black Monday

The latest ups and downs on the stock market are only partly related to the current public-health issues. There are some big, underlying factors at work here.

---


You have probably not escaped the volatility on the stock market, which is being blamed almost entirely on the coronavirus epidemic. While that plays a role, it is far from the whole story. We have seen a relatively sudden drop in oil prices, which some investors consider to be a negative factor for the stock market. Oil companies do suffer, for sure, and oil-producing states could respond to declining tax revenue with rash fiscal measures. At the end of the day, though, lower oil prices work as a stimulus for the economy and should be viewed as such by the stock market. Consumers are left with more money at the pump - as it happens at the very right point in time - and down the road industries that depend on oil as an input will see a nice little cost alleviation.

If we look beyond the immediate news cycle, this market correction makes sense. That does not mean it is good or even desirable, but it is entirely explicable. It is related to macroeconomic trends that I have been discussing on this blog since last fall: 



In other words, last fall the U.S. economy shifted gear, from a long stretch of growth and improving household finances to a softer period. It was only a matter of time before the stock market would correct for this; a review of historic data furthermore suggests that the correction was not going to be minor. The big question was how smoothly the correction was going to happen.

It was not going to be as big as what we are seeing now, though. The current public-health hysteria is conspiring with an increase in computerized stock-market trading to amplify what is, again, really just a course correction (albeit not a minor one). To get an idea of what role these news-cycle events play, and how they balance against macroeconomic data, let us go back to the Black Monday event of 1987. Back then, the S&P 500 fell over 20 percent in one day (measuring day-to-day closing prices) and derivatives markets basically crashed. 

The worst we have seen so far is a 12-percent one-day decline in S&P 500.

Back in 1987, the economy was coming out of a boom, inspired in good part by the Reagan tax cuts earlier in the decade. Similarly, we are now at the end of a long expansion period, with the strongest growth of that period seen in the past couple of years. Back then, the substance of the growth period had already passed when the stock market started correcting: in terms of capital formation, a.k.a., business investments, the situation was noticeably worse than it is today. Figure 1 compares annual capital-formation growth rates, reported quarterly, for the periods 1982-1987 and 2014-2019:

Figure 1
Source: Bureau of Economic Analysis

Our current expansion period is less conspicuous in terms of growth rates, but it is also smoother in terms of endurance. That said, with the dip into negative territory with shrinking investments at the end of last year, we should not be surprised at all at what the stock market is doing right now.

The timing of the correction is also important. In the 1980s the macroeconomic slowdown had begun earlier and should reasonably have been accounted for by the time of the Black Monday event on October 19, 1987. However, there was also a strong expansion in the financial sector at the time, due in part to global financial deregulations and the fact that we had already seen a five-year long, strong growth period in global trade. We had also seen a sustained period of financial globalization; it was in the 1980s that we got used to the idea that you can instantly move a billion dollars from one continent to the next. 

In other words, when the macroeconomic fundamentals underlying the stock market began showing some signs of a soft landing, the stock market itself basically saw good reasons to continue to grow. This, however, also marked the beginning of a long trend where the stock market partially detaches itself from its macroeconomic fundamentals. 

Today, that detachment is much more pronounced, contributing to a somewhat more erratic behavior of the market in times of stress. Figure 2 gives us an idea of just how big that detachment has become, reporting the S&P 500 market value per $1 current-price GDP:

Figure 2
Sources of raw data: NASDAQ (S&P500); Bureau of Economic Analysis (GDP)

Plainly: compared to 1987 the stock market over-values our economic production by a factor of 30. This is, again, a reason in itself to expect more volatility when the stock market adjusts to changing macroeconomic fundamentals.

There is more, however. The increased use of computer algorithms in stock-market trade appears to have added to the very problem the algorithms were supposed to counter. The algorithms and automated trade orders were introduced partly in response to the Black Monday events for the purpose of stabilizing trade and avoid more volatility. In reality, the computerization of trade has actually added to the volatility of disruptive episodes on the market.

Without highly specific micro-trade data, we cannot isolate the specific contributions from these algorithms. However, institutionally speaking - at the aggregate level of the market - we can compare the volatility of the market in 1987 to where it is today. This gives us a bottom-line view of whether or not computerization of trade has made the market more or less resilient in volatile times.

Figure 3 reports two 22-day episodes, both with daily closing prices for the S&P 500. The first period is the month of October 1987 (blue); the second covers the most recent 22 days in 2020 (grey):

Figure 3
Source of raw data: Wall Street Journal

Both periods start with eight days of relative calm. After that, the 1987 episode goes into a brief period of violent price swings, of a magnitude we have yet to see today. After that, though, it reverts back to relative calm. The 2020 episode, on the other hand, actually exhibits increasing volatility.

Without any detailed knowledge of how the algorithms contribute to trade, this pattern could very well be the work of automatically executed orders. An algorithm does not make judgments on its own: its trade strategy is dependent on the strategic thinking that goes into writing the algorithm. If the person who employs the algorithm is risk averse and wants to cut losses when the market is either plummeting or generally volatile, then that is what he is going to tell his algorithm to do. Since selling stocks in a downbound market means adding to the price plunge, automated sales for the purposes of risk mitigation will in fact increase the trend that triggered the sale in the first place.

This is really no more complicated than how many investors think about their portfolios. The difference is that a human being can change his strategy entirely on a whim; an algorithm is what it is until it is rewritten.

The stock market is really a microcosm of economic behavior in general. In terms of price stability, it captures well the tension between two key concepts in all economic theory: risk and uncertainty. Risk is the less-than-perfectly predictable future that we are trying to assess in order to make informed decisions today; whenever we can quantitatively estimate the possible outcomes of alternative actions - such as buying or or selling stocks - we can make a risk-based forecast of the near future.

When, on the other hand, we cannot attach reasonably confident numbers to possible scenarios, we are faced with a situation of uncertainty. Whenever we cannot make quantitative estimates of, say, the outcome of a stock-market investment, we tend to respond differently than under risk-based forecasts. Uncertainty, simply, makes people react less predictably and with stronger emphasis on aversion.

At the same time, by the very nature of their circumstances, decisions under uncertainty are much more qualitative and much less quantitative than decisions under risk. Therefore, it is quite a bit harder - not to say de facto impossible - to rely on algorithms for decisions under uncertainty. This is why we have reasons to be concerned about the latest volatility episode. 

It is, again, important to distinguish between a price trend and price volatility. A trend is predictable by extrapolation; volatility is not. Therefore, volatility injects uncertainty into decision making in a way that trends do not. Plainly, in general theoretical terms: it is better to live a predictable, miserable life with financial losses than to live with fundamental uncertainty as to whether or not you will be able to pay your bills tomorrow.

Applied to the stock market, this means that the pattern of price swings for 2020 exhibited in Figure 3 is a classic example of uncertainty. When the price amplitude is high, it becomes predictable after a while; when the amplitude is expanding, predictability is hampered. Every expansion of the amplitude grows the vector of possible future prices, without adding any information that allows us to assign probability numbers to any of those prices:

Figure 4a

The problem with computerized trading by algorithms written to reduce losses is that they actually exacerbate the problem. This can happen in either of two ways, the first being an expansion of the amplitude itself. For example, suppose an algorithm is designed to sell when prices plunge by a certain percentage within a given time period. This exacerbates the price plunge - adding to the excess supply that drives the price drop in the first place - thus driving more sell-offs of the same kind. Suppose, then, that another algorithm identifies a bargain opportunity and starts buying the stock. 

The theory behind these two algorithms is that they will balance one another out and stabilize the market. However, since the amplitude in prices is expanding in the first place, and these algorithms execute more trades the bigger the price swings, they actually amplify what they are supposed to mitigate. The price vector is, so to speak, stretched out:

Figure 4b

The other way to add more uncertainty to the situation is to load a given price vector with more information, i.e., more possible future prices. This happens when price swings become more frequent; you face a shorter time frame for making decisions on a given price vector:

Figure 4c

I have yet to see one computer program, in any setting, that can do the exact opposite of this, namely reverse the amplitudinal expansion of the price vector. Maybe they exist, but if they do they are apparently not being used on the stock market. 

There is hope, though. Here are the daily closing prices for the last four months of 2008, when the macroeconomic expansion period of the Bush Sr. presidency abruptly came to an end:

Figure 5
Source of raw data: Wall Street Journal

There was another spat of volatility not long after this one, but it was smaller and followed a similar pattern. Hopefully, level-headed decision makers on Wall Street will prevail and turn the current volatility episode into the temporary aberration it deserves to be. The underlying macroeconomic variables that precipitated a stock-market correction are nowhere near what they need to be to merit further volatility. 

Over time, we need a more sensible balance between the stock market and GDP, but that is a story for another day. 

No comments:

Post a Comment

Favorites!