If someone knows the best way for an Australian to buy US Put options, please let me know.

This post is somewhat unfinished, unedited, and much less detailed than I wanted it to be due me changing jobs.

Empirical prediction

My credence that the S&P 500 index will drop below 3029 at some point before this time next year, i.e. over a 30 percent decrease from current levels (at the time of writing), is much higher compared to both the market’s likely estimate (because rational investors would sell their stocks if they thought the index was overvalued at the current level, which at the time of writing is 4327) and compared to my base rate of such large crashes. If I had to estimate my current credence of a 30 percent devaluation happening within the next year (i.e. before July 16, 2022), I’d put it at around 65 percent.

Caveats

As you might see from the dates being behind the date at which this was posted, I’ve been writing this for a while. I’ve delayed posting this because I can’t fully justify the probabilities I’m assigning. I take methodology quite seriously (i.e. how do I know that my line of reasoning reliably leads to true conclusions), so I hate filling out my argument with weakly justified back-of-the-napkin calculations and gut feelings. But that’s all I’ve got here. So take everything I’ve written with a grain of salt.

The market actively fights any prediction. If an indicator of a future market movement is found, it will tend to become less effective over time, as other people learn to adjust their bets to accommodate that information. This means that all of my justifications based on historical data could be very wrong. I’ll be careful to, in each of these instances, give you reasons why a particular historical indicator might not indicate what it once did (note: ran out of time to do this).

To be clear, this is not financial advice. This is an empirical prediction with a justification that might have critical errors. I’m not a permabear, but I think the market conditions are heavily weighting the probability distribution of returns on the negative side of graph. To give you an idea of my confidence in my prediction, I sold all my stocks on 12 July 2021; the week after, I explained my position to some of my loved ones, and they sold their retirement stocks and held cash soon after; currently, I am deciding how to safely execute a large short position. So if I’ve made a mistake in my prediction, let me know.

The base rate of a now-to-trough drop of 30 percent

To calculate my base rate for a greater than 30 percent drop, I used the daily S&P 500 data since January 1986 (roughly when US inflation and US interest rates stabilized). (This data simply tracks the index without reinvested dividends, which is not ideal). On any random date since then, the probability of more than a 30 percent decrease at some point during the next year is 6.1 percent. This was way higher than I expected, so let me be extremely clear on what it means.

Put your left index finger on a random point on the S&P 500 graph between January 1986 and July 2020. Put your right index finger on the graph a year later. Put a thumb at the lowest point on the graph between your two index fingers. The probability that your thumb is more than 30 percent lower than your left index finger is 6.1 percent. This can be true even if your right index finger is higher than your left index finger. So why use this metric?

Why estimate now-to-trough?

I want to address a particular error I’ve seen a lot of people make. To quote a friend, “Stocks go up and down all the time, but if I leave it in there for 40 years, it’ll almost certainly be higher than it is now.” I actually agree with that statement, but it does not imply that you can’t get higher returns by temporarily selling your stocks (or shorting the market). If you knew with certainty that the market would drop 50 percent within the next month, you’d be a fool if you didn’t sell all your stocks now and then rebuy stocks closer to the bottom of the crash. The fact that we have to assign probabilities to crashes, rather than give guarantees, does not change anything of substance—it just adds extra steps in the maths. Suppose there’s only a 10 percent chance of the market crashing 30 percent tomorrow and a 90 percent chance of it going up 3 percent, then it would still be prudent to put your money into cash even though it’s a low probability event: Given those odds, it would be negative expected value to keep your wealth in stocks at that point, so there is no profitable Kelly bet.

The fact that the stocks will ultimately be higher does not mean that the people who sell now are making a mistake: There are mechanistic reasons why markets crash even if it’s guaranteed that the S&P 500 will be higher in 40 years than it is now (which I think it basically is). However, if we are going to try to “time the market”, we need to estimate a probability distribution for the highest peak and another for the lowest lows. If we sell too early, we lose. If we think the market will drop lower than it actually does, we’ll be waiting for a time that never comes. It is incredibly hard to correctly time the market, which is why investors who attempt it typically underperform the market in the long run. I’d generally recommend always having long positions, even when you think a crash is likely and that those stocks will fall in a market crash. This is to decrease your maximum drawdown if you’ve bet too early on a bubble. Then keep increasing the size of your short positions as the market rises (without an underlying reason for higher valueations, and decrease the short as the market drops.

Are we in the 6.1 percent?

Assuming these odds of now-to-trough drops of over 30 percent haven’t changed too much, the question then is “How much more likely is it that we are in that 6.1 percent today?”

The S&P’s CAPE ratio and PS ratio

The cyclically-adjusted price-earnings (CAPE) ratio of the S&P 500 is over double its historical average, and it’s a metric designed to evaluate when the stock market is overvalued. The only time the CAPE ratio has ever been higher is the lead up to the Dot Com crash. Right now, the CAPE is higher than it was at the market peak before the 1929 crash that led to the Great Depression. If the market does fall, and if this metric is reliable, the market has a long way to go down.

If you run a regression on the CAPE ratio to either the average now-to-peak or now-to-trough, you see a negative relationship at all timescales. I.e., as the CAPE increases, your expected upside is lower and your expected downside is greater, whatever your investment time horizon is. In my rough model, when the CAPE is below 33, peaks exceed troughs at all timescales. At 34, troughs are bigger only on a timescale of 1 to 5 months, but at all other times, the peaks are bigger. At a CAPE of 36.6, the 12-month trough finally exceeds the 12-month peak, with 6 months being the biggest gap. In July, the actual CAPE was around 38.

The price-to-sales ratio (PS ratio) of the S&P500 tells a similar story, being at more than double its median value since Jan 2001. Looking at data that goes back to 1993, this is the highest the PS ratio has ever been. The ratio is 38 percent higher than what it was before the Dot Com Crash.

Note that there could be reasons why the CAPE and PS ratios might not return to their historical average, e.g. increased savings chasing investments, which are not perfectly elastic. There may be risk of persistent inflation, which stocks hedge against. But I don’t think anything that’s happened justifies a CAPE of 38 when its average is 17, except perhaps perpetually low interest rates, and there’s some bad news on that front.

US inflation is high and interest rates are very low

At the time of writing, annualized month-to-month US inflation has been over double the long-run target which, if it persists, will force the Fed to take actions that increase interest rates earlier than their 2023 forecast. Higher than normal inflation is something that the Fed predicted, but inflation went beyond the Fed’s predictions, being higher and more persistent. The Fed is arguing that this inflation is transitory. What they do not have, however, is good justification for why they think all the extra inflation is transitory, which was pointed out by former U.S. Treasury Secretary Larry Summers.

Even if inflation “should” be transitory, people’s mere expectation of higher inflation can cause higher inflation: A shop that expects the money it receives to be worth less will raise its prices, and if all shops do so, then the money is indeed worth less. Because inflation is heavily influenced by expectation, it’s hard to predict with precision. The labour market is also going to create inflationary pressure. Wage growth is at a recent high, which can cause inflation: More dollars are competing for the same goods.

In order to reduce inflation, interest rates usually have to increase.

As I’ve been writing this, the Fed has brought forward its reduction in asset purchases, while also stating that it will not increase the interest rate earlier than intended. However, when the Fed reduces its purchases of bonds, they move the demand curve for all bonds to the left, which decreases the price of the bonds, which raises effective interest rates for everyone except the banks.

Interest rates significantly affect asset prices. Unexpected increases in the interest rate increase the discount rate that investors use in their (explicit or implicit) net present value calculations, which decreases the true value of stocks. Rational investors will sell until the trading price reaches that new true value.

The federal debt-to-GDP ratio is very high

When the Fed buys assets, it is adding to the demand side, meaning higher prices and lower interst rates. The Fed is now reducing asset purchases. When they raise interest rates on the US’s enormous debt, the US is more likely to default, which would likely cause a huge decrease in the S&P 500 index. A mere increase in the likelihood of default should cause a decrease in the index.

Margin debt to GDP ratio is at an all-time high

The total money that is in margin loans is at both an absolute all-time high, and a relative all-time high compared to GDP. Right before the Dot Com crash, the margin-debt-to-GDP ratio peaked at 3 percent. Right before the 2008 crash, the ratio peaked just below 3 percent. Today, the ratio is at 3.9 percent. This is the highest it has ever been. Before each of these three major crashes, the ratio rose rapidly over the course of a few months, rising 52% before 2000 and 60% in 2007. From the bottom of the Covid crash, the rise has been 77%.

However, increasing levels of automation in the financial markets causes me to expect higher margin-debt-to-GDP ratios in correctly valued markets. One reason for this is that when a computer program finds arbitrage, it can, theoretically, leverage infinitely with zero risk. Fewer companies were doing this in the early 2000s. Another reason may be that more investors are better maximizing long-run growth. To misuse a term (but get the idea across), they’re “Kelly betting”, which would require increasing leverage when the S&P 500 is undervalued, and decreasing leverage when the S&P 500 is overvalued. We can actually see that the troughs of the margin-debt-to-GDP ratio has gotten higher each time (though, this is not strong evidence, as there are only four datapoints). The ratio hasn’t been below 2% since late 2012. And will you look at that, I just had a quick look, and from 1960 to 1995, the ratio never went above 1%.

China’s housing bubble

China’s housing bubble is truly enormous. The China narrative I’ve heard for most of my life is that it will overtake the US and become the world’s dominant country. I strongly disagree with this claim. Government systems and culture are hard to quantitatively evaluate, but I also think they’re probably the most important drivers of long-run economic growth. Good systems, in any domain (not just government), ultimately have some robust logical justification. In the case of public policy and market rules, each good system provably satisfies some optimality criteria (free markets, for instance, satisfy Pareto efficiency). For the CCP to declare themselves as “Communist” means they are likely to disregard much granularity that good policy must have (the same can be said of policy makers who cheerlead Capitalism, who have privatised things in such a way that the incentives of the new system produce incredibly poor results, such as the US prison system).

When you have a government that identifies as Communist, that’s a threat that needs to be taken seriously. The evidence I’ve looked at this threat to evaluate this threat is anecdotal, which is often wrongly criticised. Certain anecdotes have much stronger weight than other anecdotes, which is often not considered by people who don’t understand Bayesian reasoning. Anecdotes are data, and data should update beliefs by an amount that’s subject to the strength in your belief of your priors, and the likelihood of the data occurring under the hypothesis you’re updating relative to the likelihood of the data occurring under all other hypotheses. Some anecdotes are incredbly strong evidence.

In this case, my anecdotal evidence is from Winston Sterzel and Matthew Tye, two Westerners who both lived in China for over a decade, both married Chinese women, and both had to escape China with their families under the threat of arrest under false charges of spying. (One of their Canadian friends was arrested, and subsequently released hours after Canada released an executive of Huawei who was legitimately arrested.) They have both experienced China’s extreme housing bubble. They’ve talked about

the CCP’s concerns about a housing bubble and how they’ve implemented loan restrictions
how married couples are getting on-paper divorces so that they can increase the loans they can legally attain
how people take out additional, illegal loans under the names of their family members and friends
the price of housing is something like 40x the annual salary
that if a house is previously lived-in, the buyer of a property will pay less because of a common superstition that you inherent the previous tenant’s bad luck, so a lot of these houses aren’t producing rent, meaning that house prices go well above the net present value of rental payments
the poor quality of these buildings and poor maintence, leading to cracks in the walls and apartments being demolishes before the 70 year lease is up (land in China is not privately owned, only leased by the CCP)

In early July, I told a friend that buying housing in China was like investing in a banana that you’re not going to eat. The investing, cultural, and political environments that these two people describe makes me very worried for the people of China, and we’ve seen before that large housing bubbles can take down the world economy. (I was not aware of any of particular housing businesses, such as Evergrande, when I wrote this.)

Conclusion

There are a lot of triggers for a potential sharp short-term drop in the S&P 500 index. Every major leading indicator is flashing red, which tells me the chance of a drop over 30 percent is much more likely than it normally is. Either forced sell offs (due to margin calls) or sell offs due to fear could cause a massive and rapid selling cascade. At the time of posting, the S&P 500 is 1.4 percent higher than when I made my prediction. I still claim there is a 65 percent chance that drop to 3029 will occur sometime before July 16, 2022.

The S&P 500 Will Drop Below 3029 Before July 16 (65 percent confidence)