Perceived and actual market anomalies have been discussed for ages. They were initially identified as a flaw in the efficient market hypotheses and received much academic attention during the 80s and 90s. Most researchers sought to explain them as statistical quirks or proof of data torture. Examine the limited data set enough times and eventually something with high significance will show up. This article will commit every conceivable sin when it comes to data torture, but we only seek to illustrate an effect,not to explain in detail or verify it to the nth digit.
As an example of anomalies, Investopedia (a webpage) lists seven relatively well-known anomalies: the small-firm effect, the January effect, low book value, neglected stocks, reversal, day-of-the-week, and dogs-of-the-Dow. As you may imagine, the list is far from exhaustive. There are several other well-known anomalies, including the turn-of-the-month, low-beta, and holiday effects.
Sometimes anomalies are real, sometimes they are imaginary, and sometimes they disappear (or even worse, change sign) after discovery. The latter can be fairly problematic for portfolio managers seeking to exploit the anomalies. Some of them are profitable while others are non-persistent and fade away into the dark corners of history. In some cases, the disappearance of the anomaly is due to it having not been properly identified and in other cases because the market structure changed.
Anomalies appear to be sensitive to market structure issues, e.g., post-settlement trading on Fridays in the S&P 500 E-mini future that was introduced a few years ago, changed behaviour around option-expiries. When you, as a market researcher/trader/historian, want to understand the cause or potential profitability of a particular market irregularity you also need to know the circumstances under which an anomaly persisted.
This article will focus on one of the lesser-known market anomalies; the behaviour of a major equity market around the Federal Open Market Committee (FOMC) announcement.
Structural changes in the market
To illustrate an anomaly that was persistent for decades, but then went away, we will use the so-called day-of-the-week effect. The data-driven observation was that Mondays tended to have negative returns while Fridays tended to have positive returns.
The day-of-the-week effect was subject to academic debate in the 80s (French, 1980) . The standard explanation model predicted that Monday would be as profitable as Friday. If traders could see potential for profits before the weekend, they should discount that rationally on Thursday. Then if Thursdays were profitable, rational traders would front-run the effect on Wednesday and so forth. There should simply not be any room for a day-of-the-week effect if traders were rational. But the data gave a different answer: Fridays were as positive as Mondays were negative. The effect eventually disappeared and over the last two decades, both Friday and Mondays have a low positive realized return.
The persistence of the effect seems to have disappeared shortly after the “Black Monday” (October 17th) in 1987. Whether the market shock that the Black Monday represents (S&P 500 declined slightly more than 20% on that day) actually forced people to change risk management protocols or behave differently is unknown. After 1987, the day-of-the-week effect is simply not visible in the data anymore.
Similarly, there are also trading patterns that changed post-2008. Among other things, several equity indexes’ short-term mean reversion strategies have exhibited lower realized profitability. Whether or not that is related to 2008 is unclear but it seems to be slightly more than a pure coincidence.
When exploring data-driven anomalies, you need to try to understand the actual drivers of the behaviour. Is the behaviour persistent, semi-persistent or only an effect of data-mining?
Two NY Fed Researchers (Moench & Lucca, 2011)  wrote one of the first reports documenting the pre-FOMC drift. The paper stated that the FOMC announcement day, alone, is responsible for almost half of all equity market returns over the last decades. If that is the case, most types of other analytic work are essentially dominatedby the FOMC risk premium. As a side note, the FOMC is the committee that set the Federal Reserve Funding Rate and decides upon money supply among other things.
The market “glitch” is on par with other significant, well-disseminated, anomalies such as the turn-of-the-month or small cap against large cap. The anomaly, when viewed as monthly returns, achieves low correlation, practically uncorrelated, to equity markets, despite a (clear) long bias and fared reasonably well throughout the “Great Recession” and subsequent south European debacles. Despite this, there have been relatively few follow-up studies or published research papers on this particular announcement effect.
In this article, we will only cursorily discuss the potential reasons for the pre-announcement drift, but rather focus on the effect and how you, as a market participant, can exploit the effect. For reasons and discussions about the potential drivers of the effect, we would refer the reader to the original paper.
The FOMC announcements are one of the most important, regular events in financial markets. The excitement and pre-speculation can be experienced watching any news network the same day. Observing markets on a FOMC day is a waiting game. Akin to the gradual increase in excitement that is derived from waiting in a queue to a roller coaster. There is a long period of nothingness (waiting in the queue) and then everything is experienced in a very compressed time interval (the ride).
There is usually lower than the normal trading activity before the event. Once the announcement has taken place, there is a flurry of activity as traders seek to digest the message and adjust positions. However, the post-announcement effect is smaller (virtually unchanged, under higher volatility) than the pre-announcement effect.
As the Russian proverb says, “Trust but verify,” we will start with a short data exploration exercise. Using S&P 500 equity index as the benchmark, we will explore the nature of the effect and then we will discuss the magnitude and risk of the effect.
Since the operation of the US Federal Reserve is not stationary throughout time, we should expect changing behaviour. Nor should we expect it to be stationary regarding the chief policy maker or mandate. One could for instance, suspect that the increasing visibility of monetary policy compared to the ever more polarized fiscal policy has shifted trader attention to an independent central bank.
We will start our data voyage, covering the current Fed Chair (Janet Yellen) as well as her two predecessors (Alan Greenspan and Ben Bernanke) – a period of approximately 28 years and close to 270 FOMC meetings. We only use pre-announced meetings as any pre-FOMC anomaly has to be based on prior known information. This is a frequent event, with the FOMC typically holding seven to 10 meetings per year.
There are alternative cut-off points, for instance including Volker or including the part of history from the change of the Federal Reserve Act in 1977 or any other arbitrary date. One of the issues when going back in history is that communication methods have changed. The FOMC gradually increased transparency over time, starting in the early 1990s.
One could argue that picking the starting point to include October 1987 (“Black Monday”) handicaps the S&P 500. So our exposé will start in January 1998. This would favour the S&P as we exclude one of the worst data points, reducing the size of the anomaly.
From Fig.1, we note that the total return of the S&P 500 is diminished by a factor of approximately half, if we remove the FOMC days. Additionally, the returns on FOMC days, exhibit a Sharpe ratio that is significantly higher than on non-FOMC days.
Making things even more interesting is that returns from the FOMC day are virtually uncorrelated to equity market returns on a monthly basis. This is despite the fact that the FOMC anomaly is a long-only effect.
As seen from the Table 1, the effect is rather indifferent to the general equity market direction. The strong performance during the equity market drawdown of 2008 speaks to the faith in the US Federal Reserve as the “buyer of last resort”.
To understand the actual behaviour of the FOMC announcement, usually taking place at around 2pm Eastern Standard Time, we will look at the intraday behaviour. We will be using data from the S&P 500 E-mini future, one of the most liquid equity indices.
The advantage of the S&P 500 E-mini is that it traded almost continuously and that we can study the drift of the non-active trading hours as well.
Using intraday data, with higher resolution than hourly, will consume a large amount of data resources and will most likely exceed the available data in your average spreadsheet program. Hence, we will only study the effect from 1999 and onwards.
We perform an event study around the FOMC day, denoted t. The study extends three days before the announcement and three days after.
Fig.2 Intraday behaviour three days before and three days after FOMC (here denoted as t). As we can note from Fig.2, the equity market exhibits a large move during the announcement day. Furthermore, there is generally a slight positive spike in performance directly after the announcement, which fades into the close of the FOMC days.
In general terms, the market is range-bound to slightly higher three days after the announcement. The three days leading into the announcement are usually driven by a slightly lower equity market. Again, the effect is mostly concentrated around the announcement itself and other effects are generally smaller. Stretching the time window, to the middle of the trading day, there is potentially a better entry point as stated in the original research article.
To complete the circle, we also plot the behaviour of the average non-FOMC compared the FOMC day in Fig.3. We can see that the event on the announcement day is large compared to the average day in equity markets. The average non-FOMC day, over this time-period, has not produced a large positive drift. The event itself, has produced a strong positive return pattern. We should remember that this is the average and there is substantial volatility around the announcement. The trade is not exactly riskless.
As with any other short-term market effect, there are substantial implementation costs. Each FOMC trade would require a buy and a sell trade. Liquidity is not the same as during other days, which may increase your transactions costs. That is, if you are foolish enough to try to implement this strategy. Implementation cost today, using for instance futures, is a couple of basis points and would comfortably be within the expected per trade return.
That said, liquidity is not constant throughout time and has increased substantially over time. So, returns from the 80s and 90s were costlier to implement. The advantage of using indices as the preferred instrument is that there are multiple venues of achieving target exposure.
Even if an organization does not have the trading technology to execute within the transaction cost, knowledge of the FOMC drift can be exploited to postpone sell-orders or to accelerate buy-orders.
Does the chairperson matter?
The size of the effect could very well be due to a specific influence of the chairperson. Was the Greenspan Put worth more than the Bernanke Put?
Unfortunately the sample set only gives us two full-length periods and one that has barely started. The testing is fairly straightforward. We will not assume that there is a gradual transition period, but rather a hard and fast change.
From Fig.4 we note that the FOMC days seem to have been as profitable during Ben Bernanke’s reign as during Alan Greenspan’s – despite the fact that Ben Bernanke ran the Fed for a much shorter period. Janet Yellen’s impact on the equity market is a bit too early to tell, but so far, it has been positive.
The fact that Ben Bernanke comes out on top is partially related to the fact that he was the chairman of the Federal Reserve during the financial crisis of 2008.
As noted in the prior segment, the macro environment and decision by the Fed to infuse liquidity into the system had an impact on the FOMC day returns. We will perform a cursory investigation into two macro variables, the interest rate environment and the trend in equity markets.
For equity markets, we will be using a 12-month average as the indicator of general direction. For the interest rate environment, we look at the last decision to infer the monetary bias by the US Federal Reserve.
Equity market direction
We set up a quick momentum study, examining the year-on-year equity market direction prior to the meeting (using monthly data). While not a sophisticated momentum indicator, it is fairly robust and gives an overall indication if the market is moving up or down.
In general, we find that the percentage return on each of the FOMC days is approximately the same. Risk on the FOMC day, is generally higher during equity market declines reducing the risk adjusted returns of the anomaly.
The FOMC day drift does not seem to be related to the overall equity market. If anything, the pre-FOMC drift seems a bit friendlier to upward trending equity markets. It is difficult to disentangle the two effects. That is, we do not de-trend the data. The sample of negative year-on-year equity market periods is somewhat limited making the observation merely a reflection.
Like in the prior section, we study the impact of the most recent interest rate decision. We are interested to find out if the effect is weaker or stronger when the FOMC is deciding to increase, decrease or hold the policy rate. The policy rate, is only one of several variables in the US Fed’s toolbox.
The scope of this article does not allow us to perform a detailed analysis of the verbal message that is also communicated that may drive expectations about the rate. Pre-emptive communication messages may get the market “in-line” with the Fed’s view before the action. There are few observations in each bucket, so any data driven conclusion should be interpreted with great care.
First, let us point to the fact that number of state-changes are relatively few. Once the FOMC starts raising or lower rates, it tends to continue in the same direction for some time. There is not a large difference in the returns depending on the environment for the pre-FOMC drift. The returns from the pre-FOMC are slightly larger when the committee is reducing rates, but the difference is small.
We cannot determine any larger negative impact on the FOMC day when the rate cycle is changing. Potentially this is explained by the decision being widely communicated and the number of surprise decisions being low.
The FOMC drift has shown itself to be a strong driver of equity market performance. As we have shown in this article, there is a substantial drift around this announcement. Potentially, this is one of the most important return drivers for equity markets over the last three decades.
Furthermore, the effect is largely not driven by the actual announcement itself, but rather by the fact that the market is producing a significant drift, approximately 24 hours before the announcement of the decision.
The argument should perhaps not be stretched too far. In a hypothetical world, without a FOMC, we would most likely end up with a result close to the long-term equity market returns. Potentially driven by some other type of policy event.
The results could also be the effect of other events. However, the fact remains; the drift largely occurs before the announcement. The anticipation and market belief that the FOMC will deliver something that is positive for equities seems to be strong.
There are arguments that could point in the direction of the effect relating to the so-called “Fed Put”. The FOMC will reduce rates and improved liquidity in response to falling/panicking equity markets.
As long as the perception of the US Federal Reserve as financial stability friendly and an effective backstop for any larger equity market sell-offs persists, the market will probably continue to discount the action as a positive driver of equity market performance.
Linus Nilsson, CFA is an independent investment consultant. Most recently he held the position as fund manager/CIO for John Locke, a French Hedge Fund. He has worked for IMQubator, Man Investments, Norges Bank and RPM AB. He holds an MSc in Electrical Engineering from Chalmers University of Technology in Gothenburg and a MSc in Financial Economic from the University of Gothenburg. In addition, he also holds a number of industry designations, among them the CFA.
 French, K. R. (1980). Stock Returns and The Weekend Effect. Journal of Financial Economics , 55-69.
 Moench, E., & Lucca, D. O. (2011). The Pre-FOMC Announcement Drift. New York, USA: FRBNY Staff Reports no 512.