Thoughts on market crashes

Nine years into a bull market following the financial crisis, I think many investors are wondering whether the next market crash is imminent. I for one have been fairly preoccupied about this since the end of last year and have been planning to write a post sharing my musings on market timing for quite some time. Can the timing of the next market crash be predicted with any precision at all? Is there anything that can be done to prepare for it, or is it just not worth worrying about?

Market crashes are rare

The natural starting point is to take a high level look at recent history. I’m only really interested in major market crashes where prices fall by significantly more than 20% (the traditional definition of a bear market).

Smaller ‘corrections’ may feel painful at the time but are really not worth worrying about, as quality stocks are often relatively unscathed and it’s very difficult to anticipate them well enough to make evasive action worthwhile anyway. Attempts to evade smaller market corrections by selling on signs of danger will tend to lose money, as rebounds are sudden and it is very hard to pick the bottom to reinvest again.

So what do recent crashes tell us? The main point that leaps out is that there isn’t really that much to go on. Going back to 1950, there have only been about six or so significant market crashes (depending on what exactly you call significant), occurring on average about every 10-11 years. Before 1950 they were much more frequent, but economic policy has changed significantly since then. Here’s a logarithmic chart of bear markets in the S&P500 since 1900 marked in red:


As I’m writing this I’ve noticed that there actually seems to be quite a distinctive pattern to the market crashes since 1950. There have been two long 15-20 year bull runs from 1949 to 1968 and from 1975 to 2001, each briefly interrupted by a short sharp crash then rebound about half-way through (the 1962 ‘Kennedy Flash Crash’ and the 1987 Black Monday October crash) and a few smaller corrections (e.g. in 1957 and 1980). Both of these bull runs were followed by 10 year periods of stagnation, each punctuated by two major market crashes (the oil shocks of the 70s and the bubble and financial crisis of the last decade). If this pattern were to repeat again we’d be due a significant flash crash around now, followed by another 10 years or so of bull market. This would fit more or less with my intuition of what’s likely to happen over the coming years. However, a sample of two has little predictive power and this pattern is more likely to be spurious than meaningful. My main point is that it’s not really possible to statistically predict future market crashes with any confidence from such a short history.

The second related point is that significant market crashes are less frequent than most investors seem to believe. The chances of a market crash in any given year are actually pretty low, even when valuations are at apparently high levels. If economic policy continues to improve and investors slowly learn from their previous mistakes, I’d expect them to occur even less frequently in the future. So a rational stance is to be optimistic in general about where the markets are likely to go. Given this, investors, including myself, seem to worry about market crashes disproportionately. Surveys of investors have found that they over predict the likelihood of severe one-day crashes by an order of magnitude above their actual historical frequency. I often hear of other private investors liquidating large proportions of their portfolios in anticipation of danger ahead. I’ve been guilty of this myself to some extent. Given what I understand about human behaviour, this fixation with crashes makes sense – we have evolved to be finely attuned to spotting possible signs of danger. Investors feel the pain of failing to predict a market crash much more acutely than incorrectly predicting crashes that don’t occur, even though the financial cost can be just as bad. That’s not to say it’s not worth trying to anticipate crashes at all. When they happen they can be severe. The last crash was an especially severe one that still remains fresh in the investing collective consciousness. However, it is important not to overreact with excessive caution.

What causes stock market crashes

From a narrow perspective, crashes are driven by investor sentiment. Some sort of catalyst, say the threat of a trade war, can cause some investors to sell and prices to fall. Normally lower prices should bring out more buyers and prices would stabilise, but under certain conditions a second more powerful effect can take hold. Instead of making investments look more attractive, for some investors falling prices can lead to fear and the expectation of further price falls. Even if other investors believe such fears to be irrational, they may still sell if they believe other investors are likely to do so. This can then generate further selling, in a vicious circle that can be hard to escape once the process has started. The psychological and self-fulling nature of this process make it unstable and consequently market crashes are difficult to predict. There is no single measurable factor or even set of factors that determines when a crash will happen. It happens when a sufficient proportion of investors are afraid that price falls will be met by further selling rather than buying.

So what conditions are required for this vicious circle to take hold? There are two main fundamental drivers. The first is the economic business cycle and more specifically investor expectations about where in the cycle the economy is at. Investors are naturally wary that an impending recession would hurt corporate profits. The stock market is a leading indicator of the business cycle i.e. it will fall in advance of recession rather than at the same time or afterwards. Market crashes are more likely when a large proportion of investors think a recession could be near.

The second driver is investor perceptions about equity valuations. Market crashes are more likely when investors perceive stocks to be too expensive. However, there are not clear benchmarks as to what defines expensive, as valuations depend heavily on how optimistic investors are about the future and how much they discount it, neither of which has a clearly defined answer.

Based on this, at the extreme you could think of two types of crashes: those caused by economic shocks and those caused by the unwinding of bubbles. The last two major market crashes we’ve experienced seem to fit neatly one into each category. More generally, crashes can be precipitated by both high valuations and in anticipation of an economic downturn, as the two tend to be highly correlated.

Indicators of a market crash

I’ve written about why market crashes are hard to predict and why investors perhaps shouldn’t worry about them as much as they do. But so far I’ve skirted around the main question I set out to answer: what indicators are actually useful for predicting market crashes?


The most commonly referred to indicators of potential market crashes, at least by the media, are valuation measures. The most common of these are the price earnings ratio of the market index, or the CAPE (cyclically-adjusted price earnings ratio of the market index). The simple and rather intuitive logic is that when these ratios are higher, indicating higher valuations, a crash is more likely. The following chart shows the CAPE over the long term:


The CAPE suggests that US equity valuations have never been higher other than in the boom. This seems like it could be cause for alarm, but in reality the CAPE (and other valuation measures) has very little predictive power for the timing of market crashes, as pointed out in this article.

When originally conceived, the CAPE was shown to be inversely correlated to long term future returns: high valuations would tend to lead to lower average returns in the future and vice versa. However, this does not necessarily tell us that high valuations cause market crashes. It could just reflect the inverse of that relationship i.e. that bear markets cause valuations to fall, while bull markets cause valuations to rise. All the inverse correlation between the CAPE and long term future returns actually tells us is that bear markets follow bull markets, i.e. there is a stock market cycle, not what causes it. Intuitively, it seems likely that a high valuation is a factor that makes a market crash more likely, but it’s not really that clear how important it is.

The bottom line is that there is no clear benchmark valuation level at which stocks are ‘expensive’. It is tempting to compare valuations to historical averages, but to me even this seems likely to be a fallacious comparison. As I pointed out above, the history of the modern stock market is very short. It appears that equity valuations have been rising over this period. My personal view is that equity valuations will continue to rise over the long term, as equities still seem undervalued relative to other asset classes. Equities have been perennially undervalued as an asset class, apparently because investors have priced in an excessive risk premium. This premium implies a probability of an economic meltdown from which corporate profits do not recover (or take a very long time) that I think is much greater than the underlying reality. Of course this is just my speculation – there is no way to know for sure!

Economic indicators

As mentioned above, the stock market is a leading indicator of the business cycle. This means that to predict market crashes you need to be able predict recessions a long way in advance. Needless to say this is pretty hard.

However, one factor that may be a fairly useful leading predictor is the interest rate. The relationship between short and long term interest rates, expectations of inflation, the business cycle and the stock market is complex.  My understanding is not as good as I’d like it to be and a cursory bit of research online suggests that a lot of popular understanding is contradictory and confused. A look at interest rates would warrant its own post in any event – some homework for my to-do list perhaps…

For now, I think the simplest and most useful intuition is that central banks set interest rates in anticipation of the business cycle and future inflation i.e. they raise interest rates when the economy is improving and the prospect of inflation is looming. This implies that stock market crashes are very often preceded by higher, rising interest rates.

Similar to valuation levels, there is no obvious interest rate benchmark for when a market crash will occur. However, one simple model that has been shown to be pretty good predictor of crashes is the bond stock earnings yield differential model (‘BSEYD’ – a special case of which is sometimes known as the ‘Fed model’). This compares the earnings yield from stocks to the interest rate (yield) from bonds. When bond yields rise significantly above equity earnings yields, either because of rising equity valuations or rising interest rates, a market crash pretty much always follows fairly soon after (it has commonly been a sufficient but not necessary condition for a crash to occur). It’s not clear exactly why this seemingly fairly arbitrary indicator works. The popular intuition behind it is that bonds and equities are competing assets, and bond yields rising above equity yields makes equities no longer look attractive. However, this intuition doesn’t necessarily make a whole lot of sense as it doesn’t compare like with like: true equity yields vary with future growth and inflation, while bond yields do not (as pointed out in more detail here). The reason the BYSED works well as a rule of thumb for predicting crashes is probably more simply because it focuses, in a straightforward way, on two of the most relevant predictors, equity valuations and interest rates. With interest rates where they are at the moment, this indicator isn’t showing imminent signs of danger.

Sentiment indicators

Indicators of investor sentiment are commonly reported on but have little predictive power in practice. A common intuition is that indicators of excessive investor optimism or ‘euphoria’, such as overly consistently positive media commentary, low professional investor cash balances, extremely high levels of investor margin debt, extreme optimism in investor surveys and taxi drivers giving out stock tips, may indicate that a market top is near. However, these are pretty inconsistent predictors. Various sentiment indicators can suggest when conditions seem euphoric, but they are volatile and won’t predict when a market crash is coming. They will also miss the majority of market crashes, which occur when conditions are not actually that bubble-like. Sentiment indicators can be a bit better at indicating short term bounces when they show extreme bearish readings, but that’s not of much use for this post.

Market indicators

A different tack is to try to observe the behaviour of the overall market as a crash is happening and react quickly, rather than attempting to predict the crash in advance. This sort of approach is favoured by momentum investors like Mark Minervini and William O’Neill. I think it may have some mileage if you can learn what the relevant patterns are. Hopefully, with experience you can get better. Broadly, what you’re trying to do is identify signs that the vicious circle of negative sentiment and selling that I mentioned above is starting.

One obvious thing to look out for is the trend in the overall market index. A simple strategy of selling out when the indices fall below their long term moving average could have saved you from a lot of the damage in the last two market crashes, but to really benefit you would have also needed the balls to reinvest when the market fell further.

O’Neill suggests something slightly more sophisticated, which is to look for and count ‘distribution days’ in the market index. Distribution days are defined as days where the price falls on higher volume than the previous day. O’Neill suggests these may indicate selling by institutions, which tends to occur in waves as they become more anxious over time. A high number of distribution days (e.g. more than five or six) within six weeks or so may indicate that the market is becoming bearish. I’ve been following this a bit and have found it useful as a sort of warning light, in addition to observing the overall market trend. However, it seems to get triggered fairly often and so may over-identify crashes.

Strategies for dealing with crashes

Given all the tools available and their limitations, what are the best strategies for dealing with crashes?

I’m quite firmly of the view now that sitting put and remaining fully invested has a lot to be said for it. This is the most straightforward strategy to implement and is almost guaranteed to beat the average investor, who is prone to panic and sell at the wrong time (which given the rarity of market crashes is probably something like nine out of the ten times they might be tempted to sell). Instead of worrying about your market timing, why not just sit back and let compounding do its work? Sure you will get temporarily hit every now and again, but markets tend to bounce back pretty quickly even from severe crashes and if you allocate your portfolio to high quality defensive businesses you needn’t worry much. The ‘do nothing’ strategy sets a pretty high benchmark for other strategies to beat. There needs to be a clear rationale to deviate from it.

One possibility is to sell out of positions and raise cash only in very rare exceptional circumstances when you think a big market crash is likely (but only when you are very confident about this rather than just as a precautionary measure). This is tricky but may not be quite as difficult as it sounds. High valuations and high and rising interest rates (the BSEYD) and a change in market trend are a good starting point. The last few major crashes could have been predicted / reacted to quickly this way.

Another option is to raise a bit of cash as a precautionary measure when the markets are looking a bit wobbly by holding off reinvesting the proceeds of any sales. This gives you the option to either reinvest at lower prices if the market does crash further, or if not at least it allows you to rotate into higher conviction positions. It can also keep you on the front foot psychologically, as you can focus on hunting for bargains with the cash balance you’ve raised rather than passively and anxiously watching your portfolio decline in value. However, while it may be psychologically rewarding, I’m not convinced that it is a particularly profitable technique in general. Much more often than not the market will just bounce back and you end up a bit less well off compared to just holding. For me the jury’s out on whether this is actually a good idea or not.






Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google+ photo

You are commenting using your Google+ account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )


Connecting to %s