After watching a YouTube price analysis seshion by Tone Vays, one of my favorite analysts, talking about the FOMC announcement after the markets had closed on May 4th, and how he would trade it based on historical price data (and presumably make lots of money with high certainty, if he would have the time to do so), I decided this would be an interesting topic to write about. Since I had watched it later, while watching the markets going south, I had to smile, because it totally contradicted the profitability of the implied trading-proposition that he was talking about.

Just for clarification, I am a huge fan of Tone Vays. I really admire his ability to logically analyse topics and I like his unique style to explaining things. Especially, he has done some great work around clarifying the value proposition of Bitcoin and lining out, why other cryptocurrencies are either technically flawed, or are not able to provide any real long-term value as (crypto)currencies and should rather be seen as unregistered securities. He (like me) believes that most of them are basically scams, with the ambition of founders to get rich by basically creating and printing their own money. He has a “CryptoScam” podcast in which he used to provide analysis on some of the prominent projects (here is my favorite episode about Crypto Hedge Funds). Unfortunately, in recent years his focus has been mainly on providing technical analysis — primarily for crypto-traders. I understand it, since this is the topic that most of his followers are interested in — and he can monetize it. But I think he provides way more value in his other activities.

Paper Summary

Anyway, the paper in question was “The Pre-FOMC Announcement Drift”, a report by the Federal Reserve Bank of New York, first published in 2011 and then revised in 2013. (Who is interested in reading it can do so here). It is quite a popular paper, which was cited more than 500 times.

The gist of their findings is, that the stock market gains made during the 24-hours before the FOMC announcement, accounts for about 80% of the annualized returns since 1994.

Here is the Chart depicting their main findings:

Some important points about the study:

• The FOMC has 8 meetings a year and the research of the main outcome includes the data of 131 scheduled FOMC meetings from September 1994 to March 2011. The reason for this time frame is that before 1994, there were different changes in the amount of meetings, announcement time and other factors that would have added inconsistency to the examined data points.
• The data shows a mean of about a 0.49% pre-announcement gain.
• Accumulated, this results in a 3.89% per year, while the accumulated gain on all of the remaining trading days has only been 0.88%. Hence, 80% of the gains are made during the Pre-FOMC Announcement Drift.
• They also find, that the pre-FOMC drift shows no significant differences during the time of monetary policy easing cycles and tightening cycles.
• However, “it appears that pre-FOMC returns are somewhat higher around recessions and in periods of financial turbulence such as the 1987 stock market crash or the 1998 LTCM crisis”. They measured this based on the VIX level at the time and it came out that a one-standard deviation increase in the VIX increased the returns by 31 basis points.
• Importantly, the return from the announcement until the market closes is “essentially zero”. And their 95% confidence level for this is about 50 basis points. Keep that in mind.
• The paper also states that other major stock indexes also show similar excessive returns on their central banks policy announcement days. Interestingly, Japan is the only country where it doesn’t apply. Japan’s monetary policy is very weird and unique in many ways.
• Most interesting for me is that their findings suggest that these pre-announcement drift effects are purely on equities, not on fixed income securities. This is quite surprising, because it is the meeting where the Fed announces the federal funds rate for the forthcoming months. If it has at all a market effect, this announcement should primarily affect the Treasuries and secondarily on all other fixed income securities.

It is interesting that this trade suggests putting money into the stock market before the FOMC announcement. This announcement has important implications for the future of the market and therefore bears a lot of risk. Hence, it is essentially the opposite to what a risk-averse investor might do.

I would further assume that it is actually more likely that some investors pull out their money before the announcement and based on the announcement reinvest it otherwise. This should rather lead to some pullback before the announcement and higher volatility afterwards.

Given the data results, there is just a very short period of time right after the announcement with high volatility, usually resulting in a short dip, followed by a strong upward burst and a slight retreat until the market closes.

One possible explanation for this phenomenon (also mentioned in the paper) would be that due to a downward trending federal funds rate since the 1980s, the monetary policy news have arguably had a predominantly positive feedback-loop.

One interesting side note: This study completely runs contrary to any possible EMH (Efficient Market Hypothesis) assumptions.

The FOMC Announcement

Let’s look at how the S&P behaved upon the latest FOMC statement on May 4th.

The most fascinating observation is the fact that the 5-minute candle at the announcement time is exactly straddling the 0.49% gain. And it is true for both starting points explored in the paper, a) the opening of the market on the previous day, as well as the exact 24-hour trading window (you can see the candle at 14:15 on May 3rd going through the zero-line).

Next, we see the market swings, right after the announcement, seemingly exactly follow the pattern predicted by the empirical results at first glance. But looking at it more closely, it becomes obvious that while the directions of the candles follow the suggested pattern, the magnitude of those swings is way out of whack.

Especially, the gigantic surge until the closing bell is magnitudes outside of the confidence Intervall. I am actually not even sure how they are creating the 95% confidence interval in their figure (as indicated by the grey area around the mean line). It should be 2 standard deviations away from the mean and it appears to be at around 0.25% on both sides at the announcement time. However, in their regression model their result is a standard deviation of 1.215%.

Anyway, while the mean of the model gets precisely hit at the announcement time, everything that happens afterwards, first runs directionally in line with the predicted outcome pattern, but with a way stronger magnitude (way more than 3 standard deviations, or in other words, it is way exceeding even a 99% confidence interval!).

And lastly, on the next trading day, it goes completely counter to the predicted range of outcomes: A huge downward swing with enormous magnitude, the S&P 500 loosing more than 3,5%, which based on the underlying data-set should be statistically almost impossible, thereby contradicting all data-points collected over the 17 years.

Conclusion

In a way their main finding got precisely confirmed during the latest FOMC announcement. However, all of the trading after the announcement, contradicted the reliability of the statistical data supporting their findings.

What do we learn from this analysis?

• Many traders invest based on technical analysis of markets. Therefore some of these trading strategies can lead to self-reinforcing outcomes.
• However, traders & investors shouldn’t have too much trust in any market phenomenons, which are based on empirical analysis and historic data.
• Investors should keep in mind that all market behavior is based on human action, which in turn happens based on subjective valuations of people at any point in time and therefore cannot be predicted by historic data.