# To understand FOMC events and its impact on the market

Last month when FOMC meeting decision went out that fed would start to exit QE3, immediately we saw a deleveraging effect: SPY went down, GLD went down, and LQD (bond) went down, but US dollars went up.

Most of all, I would like to know what's everyone's view on what was driving these market movements after the announcement.

Also, an interesting idea is that the covariance structure at these kind of special moment would degenerate to $ee^{t}$ form, which means a perfect correlation.

I wonder if there is any reference on quantitative research on deleveraging's measurement, detection and prediction?

Another way to understand the event is by applying signal analysis, for example the impulse response view. I wonder if anyone has a comment on how the trend/reversal formation for event like this?

By the way, I got a vote down for my previous post because of "common knowledge to those who study finance". As someone who worked for a hedge fund prototyping a high dimensional trading strategy with sharpe ratio >1.8, I really don't know this knowledge would be common. The reason is obvious, because the market is of a high dimension monster, so there would never be one or two factors like in "classic" economy view that you see on TV or in textbooks, as every one who really practice with serious investment would know. So why not let us be a little bit more open so people from different background can participate, so that everyone can learn ?

-

My answer will be very non-quantitative but the resulting models are actually quite mathematical but I like to stick to a general overview because of the proprietary nature of those models. Here couple thoughts though:

• You can't just try to explain market moves by a few indicators or a single Fed speech (by the way, the market hugely misread those statements, 2 days ago in my book was a game changer and put and end to all QE-end speculation, it will continue at current rates without question). Currencies, equities, ... move as a function of a whole host of variables and the relationships are entirely dynamic and ever changing. I recommend to learn to look at the big picture and take everything into perspective when modeling a new strategy rather than trying to get bogged down by forcing a relationship between Employment and SPX returns, for example.

• The market reacted to the Fed speech a few months ago entirely differently than it reacted to the Q&A 2 days ago in a for many unpredictable pattern (though I repeat I found it a clear game changer where the dollar will be sold into the next couple weeks). This fact points to the possibility (and for me empirical fact) that market participants on average may react at most times in predictable patterns but such patterns are strongly anchored to, what I call, stage-disequilibriums. What I mean with that is that investors, traders, central banks react as a function of what their current utility is and that utility is dynamic. As a result, estimating and modeling the correct utility and "calibrating" your model as a function of the current "market utility" is a much more fruitful endeavor than looking to squeeze static relationships out of equity markets and economic indicators.

• To elaborate what I mean with "market utility": There where market stages when injecting money by a central bank was seen as a huge negative to equity markets because market participants thought that the central bank had knowledge of less optimistic economic developments than what the broad public had. The point leads to the need to find some sort of utility (I use this term very loosely here) that reflect the way how investors react to a complete set of market driving variables. So, instead of modeling each variable and trying to estimate its impact on future market returns, the exercise would be to find one utility that optimizes how the whole set of variables and their changes impact market behavior. When such utility can be shown to be consistent and persistent over time then one can speak of having identified a specific "market stage" or "utility stage". Key is to appreciate the fact that a) dependent variables my drop out or be added from one stage to the next and b) that the relationship between the dependent variables remains static within the same stage but may change entirely between stages.

• Utility or utility stages can and are very different from one asset class to another for logical reasons: A perfect example is the different utility stages between the U.S. dollar and the Australian dollar: A big function of the Australian currency utility is its dependency on the Chinese economy and changes in market perception of future Chinese growth rates, labor conditions, imports and exports. The U.S. dollar on the other hand hardly displays any dependency on such variables and is currently much more driven by when investors will dissect the intended rhetoric of Ben Bernanke. For example, right now most market participants still believe that Fed action will very much be a function of economic results this year, 2013, while I built into the current utility that no matter whether unemployment drops below 6.5% or not Fed will not reduce its accommodative state in 2013 anymore. This points to two important facts: a) my model predicts dollar moves right now in very different way from the majority of market participants, my fair value of the dollar lies a lot lower than what the market believes, and b) my USD utility model is much less sensitive to economic indicators at this moment but much more sensitive to market perception of expected Fed behavior vs my model Fed behavior.

• I like to stress, as it is a very important point that this approach is not predictive in nature but reactive. The model takes into account how the market just reacted to certain economic releases or a speech. Thus, if the model is in disagreement with market reaction then the model would suggest to implement positions against short-term momentum, if the model agrees with market reaction but established a more elaborate market participation then the model would suggest a position in alignment with short-term momentum, to give just two examples.

The key take-away here is that I witnessed that identifying different market stages quantitatively will improve the quality of the model (reduction in tracking error, variability between model output and future realized market behavior) by orders of magnitude in terms of interpreting market reaction to key variables correctly. The big difference between this approach and other models is that an attempt is not made to fit individual dependent variables to a market that has proven to strongly dislike static correlation assumptions and it also differs from multi factor models that look for a linear or non linear relationship between different dependent variables to predict market returns. My approach is not predictive in nature but reactive and this approach has shown to be fairly accurate in identifying and reflecting changes in the overall utility.

** One sentence summary**: A model can be identified that reflects changing perception (utility) by market participants to what market-drivers identify as influential variables, thus such model does not attempt to establish a static relationship (which may be re-calibrated but is nonetheless static) between market returns and a combination of dependent variables, but it identifies different market stages where in each stage market perception to a model-dependent variable combination has been shown to be extremely stable.

-