Wednesday, February 7, 2018

Manias, Panics, Crashes, Market Structures, and a Different Model of Financial Markets

This post will demarcate the first in a series of posts about the general structure and behavior of financial markets. In the first post of the series, I'll discuss the general behavior of markets and their tendency to move in between extremes. This post will include the general use of leverage while emphasizing the role of both policymakers and the psychology of market actors, but will mainly be focused on market dynamics and how to construct a model of this world. This post will have 3 parts:
1. EMH and Basic Market Dynamics
2. Flip Between Bulls and Bears
3. Different Model of Financial Markets

1. EMH and Basic Market Dynamics
Before I begin with a basic idea of how financial markets operate, I'll first discuss an idea called the Efficient Markets Hypothesis (EMH). The general idea of EMH is that markets are correct ("efficient") in the way they operate. The idea also presumes market moves are uncorrelated, independent, and "truly random" (meaning it's not possible to predict direction or size of moves). EMH also assumes that market actors do not behave based on the behavior of other market actors, either directly or indirectly. In reality, none of these assumptions actually hold.

In reality, the behavior of market actors does correlate with the behavior of other market actors. Market moves also have some form of predictability in both predictability and size. In other words, market movements are not truly random and markets are not efficient. On the contrary, markets are often driven by what were called "animal spirits" by the great John Maynard Keynes. The underlying idea is that market fluctuations are driven largely by emotions, which overlook rationality.

In periods of exuberance and optimism, market actors can become oblivious to reality as expectations shift. Rising markets create the illusion of expectation of further gains in price, which result in more market actors buying. Over time, this becomes a positive feedback loop where more people buying creates rising prices which increase expectations resulting in more buyers and so forth. That structure is the underlying feedback loop in a bull market.

Also note that in bull markets, leverage often comes into play (this is more true for housing booms than for equity/stock bubbles, but it can be true for both). In the case of leverage, it will increase gains in asset booms and may even allow asset booms to last longer than they otherwise would. However, all of the gains that came in the boom will soon be wiped out in the subsequent bust while someone will run the risk of not getting paid back. That tipping point is called a financial crisis. In that period, we will usually see a panic and then a crash.

In the case of a bear market, all the same elements pushing markets up reverse. Prices fall and debts have to be paid. So panic selling appears. That places further pressure in falling prices. Leverage unwinds as firms fail. Liquidations begin to occur. Usually, a lender of last resort will come in to prevent total chaos for no reason, but everything else unwinds and you get a positive feedback loop as the economy goes into a downward spiral.

2. Flip Between Bulls to Bears
In the dynamics I laid out above, we can think of financial markets as having two different states: a low-volatility state ("bull market") and a high-volatility state ("bear market"). The vast majority of any cycle will be in the low-volatility state when things are going just fine and a positive feedback loop in asset prices is forming. In this state, there's not much volatility or fluctuation while asset markets tend to rise.

In low-volatility states, not only do asset prices rise, but leverage does too. As more and more market actors forget about the last crisis and more new participants enter who never witnessed the last crisis, the stable growth begets more growth. Expectations begin to slowly shift while policymakers actively shift policy to encourage growth after the crash. In this process, it becomes financially beneficial to lever up to buy assets because of expectations of future price gains.

However, leverage and psychology introduce financial factors that induce balance sheet fragility. Market participants, on the whole, end up seeing rising leverage, rising assets, and rising income from either assets being cashed out or from cash flow or for a "wealth effect" (or for many other reasons I'm not going to get into in this post). Eventually, the boom reaches the phase where many market participants rely on further gains in asset prices in order to sustain their ability to borrow and make payments on their liabilities rather than being able to make debt service payments from existing cash flows. Hence, asset valuations become more stretched as to dissuade long-term investors from participating or taking place.

As the low-volatility state progresses, long-term investors begin to slowly drop out of markets or behave like short-term speculators that act based on short-term indicators (like liquidity). If long-term investors drop out of a market, then the only way to entice them back into the market is for valuations to drop to a level wherein it makes sense for those who dropped out of the market to come back in. In other words, the only way to "stabilize" the underlying market becomes via a market correction.

So what happens when a crash is triggered: markets suddenly switch to a high-volatility state where fluctuations are large, where there's no large buyer/fundamental investor to come in when prices begin to fall precipitously, and wherein the psychology of market actors is focused on a very short-term time frame.

The flip between a low-volatility state to a high volatility state happens suddenly (often overnight) and the high-volatility state usually doesn't last very long. However, those high-volatility states contain the highest concentration of both upward moves and downward moves in a single day. They contain wild, rapid swings that don't get resolved until some event comes along that signifies either that long-term information is viable allowing long-term investors to re-enter the market or wherein a large enough correction in asset prices prompts investors to seek out profit opportunities that give them compensation for the risks they take.

3. New Model of Financial Markets (this is mathy, skip if you want):
The traditional model of financial markets assume the movement of securities prices are like that of Brownian motion (a "random walk") wherein every price movement occurs built on a distribution of a constant mean and constant variance. This model is a direct reflection of the EMH assumptions listed above. All of those assumptions necessarily imply that volatility should be a static variable.

In reality, however, volatility is not a static variable. It's a dynamic variable that's constantly changing. In mathematical terms, volatility is a stochastic variable, not a deterministic one. If volatility is a stochastic variable, then we need a model wherein it changes. Of course, I've built the assumptions of volatility as something that switches from two states (high volatility to low volatility).

The return will depend on the asset class, but I'll set this model up for stock prices. A play on a country's corporate business over a long-term time horizon is effectively a levered convex bet on NGDP growth. So the question is what to use as that leverage. I find ~1.5-2.5 as an appropriate long-term multiple growth factor. That's an overall number, but it may need to go up (~2-4) depending on the correction mechanism you use. You can even make that number stochastic if you want to, though it adds complexity that I'm not sure will improve model performance in practice.

Then, I'd use a valuation driven model that uses at least 80-90 yrs of data to determine the potential for correction dynamically. For my stock models, I use >100 year datasets and adjust them to account for more uncertainty, especially on the downside. Then, I take the long-term return, use a normal volatility average, add in some uncertainty, and add in a correction mechanism from the valuation based model. For the valuation based model addition, I prefer using bootstrap methods.

Note: I'm purposefully not going into detail in this model and am not being 100% rigorous cuz I'm here to give a vague, general description of something to follow. Many ways to adjust this model to your liking. 

2 comments:

  1. Hello, this is unrelated to your post, but what would you recommend for someone with a decent background in maths (i.e. someone who has at least completed Axler's Linear Algebra Done Right, Baby Rudin, Herstein's Topics in Algebra, and Munkre's Topology) who is looking to learn about economics (with very little present knowledge)?

    I would like a text (or texts) that have suitable rigor, but also help teach intuition (so including supply/demand graphs). I do not have a syllabus in mind, but something that will cover essential points microeconomics thoroughly, as well as matters such as trade, subsidies, and other matters of government intervention in markets.

    Thank you sir.

    ReplyDelete
  2. Anything by Michael Pettis. I also highly recommend a deep reading of financial history. My personal favorite historian is Charles Kindleberger. There's many great American financial historians ranging from H.W. Brands to Kindleberger to Roger Lowenstein to Calomoris and Haber.

    ReplyDelete