![]() To make it work for financial markets, one also needs to convert all the macro factor data (growth, inflation, earnings, credit spreads etc.) to the same units (“normalising” the data). The important point is that this approach is tried and tested, and allows one to deal effectively with the problem of “too many correlated factors”. Valiant efforts in there using red wine and teapots, among other things, as examples. I did find a site where several people try and explain PCA in a way that your grandmother would understand.( ). It has been around about 100 years, and it’s tricky to explain the intuition. One such approach is Principal Component Analysis. There is a way to convert the large set of correlated macro factors to a small set of uncorrelated variables (“components”) that still capture the majority of the information. The More Technical Bit – Dimensionality Reduction So your beta estimates will be biased (and the more correlated the factors are, the more biased the estimates will be). Unfortunately that won’t work because many of the factors are correlated. What about running a simple multiple regression with Apple on the left and the various factors on the right? Ok, that sounds plausible enough as a model. Specify the model for Apple stock: US and global GDP growth (use Nowcasting for daily GDP estimates), US and global inflation expectations, Apple 12 month forward consensus earnings expectations, high yield credit spreads, energy prices, EM market volatility, real US and global rates, global sovereign stress (CDS). And if you run a correlation between Apple vs China Growth and Apple vs USD they will indicate that neither of those two factors impact Apple stock, when in fact they are the only two macro factors that matter. If over recent months China growth has moved higher but the US Dollar has also strengthened, you will find that Apple stock moves roughly sideways. Let’s say you want to understand the macro exposure of Apple stock, and in this simple construct let’s say Apple stock cares about only 2 things: 1) Apple benefits from stronger China GDP growth and 2) Apple stock benefits from a weaker US Dollar. At a minimum, you should start with the hypothesis first and see if the data backs it up.Įven then, things are not so simple, because two-variable correlation can be highly misleading. (Some fun examples of spurious correlation here: ). But if you data mine in this way, and try enough variables, you will probably find a historical match that will have absolutely no forward looking value at all. Now, if you wanted to understand the driver of the S&P500, you could simply look at all sorts of correlation charts (PMIs, growth, inflation etc.). This is the quant approach that fits Global Macro, as it is based on the idea that macro information moves asset prices. Trading opportunity arises from a dislocation / large difference between the actual price of the asset and the model value (or “fair value”). specify the model), and then build a model that explains price movements. The next step is to define the right information set (i.e. Modelling the Price: This approach starts out with the idea that price is a function of information.Sometimes these are high frequency strategies. An adaptive and computationally intensive approach is often used (“stat arb”). When these two are out of line, trades are executed on both sides of the equation and profits ensue when relationships move back into line (sometimes with painful detours). Relative Value: This approach predicts the price movement of the asset relative to the price of other assets.This is perhaps the most popular systematic strategy. That is, prices do not exhibit a random walk, but rather the price tomorrow has some relationship to the price today (persistence). ![]() Price Based: This is based on the observation that prices often trend for relatively long periods of time.We closed that piece with a historical chart showing the S&P500 historical price overlayed with a macro model value. This can be frustrating if you want to base decisions on evidence.Ī key point from our last piece ( see “ Quant – The Future of Global Macro", April 12th 2019) was that in order to predict price, one must first understand what drives the price. Indeed, one of the features of Global Macro is that it is often based on stories and spinning tales. It is relatively easy to talk the general macro talk.
0 Comments
Leave a Reply. |
AuthorWrite something about yourself. No need to be fancy, just an overview. ArchivesCategories |