Do you guys use Barra factors & models to make trading strategies?

Yes. I know many guys, who were ex-stat arb, and had used barra models for their risk purposes - and they happened to back-testing the barra factors as a return factor. Then they said "viola! why don't I dump my stat-arb model and switch to this easy stuff" and then launched a hedge fund with nothing more than a bloomberg terminal and subscription to barra factor. A few of them even ended up getting some money & its really amazing that there too many such cases of ex-stat arb traders who follow this route, not knowing that there are many others before them who have tried the same.It is better to actually backtest some of the elements of barra factors instead. recall their factors are more from a risk-variance explaination perspective. i had some huge amount of factor backtest of the barra factors and compared them to ones i calculated on my own from databases like compustat, worldscope - and you will be better off that way.(a) moreover, there are major practical issues in using a factor which is both used for the alpha and for the risk, if the objective is to often to have a risk-adjusted factor ... intuitively, its like dividing by the same thing etc ... (b) also, more often than not, many institutions, who ask for transparencies, will end up looking at the barra-factor exposures of the manager and conclude that the manager has only alternative beta via barra factor risk and not really doing any genuine stock picking and then start to pull out their money(c) finally, in quant equity strategies, the transaction cost often eats up 50%-70% of the back-tested returns ... but very few people factor this in their back-test exclusively. I met too many guys who just add a fixed number - and the most common is a negative 3% - whereas in equity market neutral (area of my expertise), the loss due to transacion costs is often 7%-12%. So, in case you are planning to do so, make sure you know all the smart ways to combine factors to get a higher information ratio/coefficients.

By saynig about risk and alpha factors you mean, that we should first regress on Barra factors, get residuals (excess return) and after it regress those residuals on our alpha factors?

yes ... thats one way - atleast if you start with the residuals after the barra factors, then no one will come at you and say that "hey, you are just doing factor-picking rather than stock-picking" - of course, I am asumming its stock-picking that what you want to do, right?

Right.The problem I see here is that Barra factors take into account some <potentially> valuable fund data and technical data.So, when we try to regress residuals we should not use those factors (or use them in some discrete way) - because otherwise it is smth like "double division" you mentioned before.Right?

correct. that's the idea. the actual implementation is another ball-game - just my 2-piece. regards

But this leaves us on the "backward-looking" side only with a limited amount of benchmarks - too many are in Barra. That's what bothers me...Thanks a lot for the explanations!

What's the difference between alpha and risk factor ?

QuoteOriginally posted by: alexoWhat's the difference between alpha and risk factor ?Oh please!

Would appreciate clear answer :let's say I calculate z-scores based on financial ratios, momentum, skews, whatnot.All of them have (hypothetically) some predictive power (significant in statisticalsense) for stock returns.So are they risk-factors or alpha factors ?Having calculated bunch of them and selected the ones statisticallysignificant, how would I determine which factors to put into my risk model and which should I use for alphas ?

QuoteOriginally posted by: alexoWould appreciate clear answer :let's say I calculate z-scores based on financial ratios, momentum, skews, whatnot.All of them have (hypothetically) some predictive power (significant in statisticalsense) for stock returns.So are they risk-factors or alpha factors ?Having calculated bunch of them and selected the ones statisticallysignificant, how would I determine which factors to put into my risk model and which should I use for alphas ?It can be a pretty amusing question depending upon whom you ask this one ... I can bet you that if you ask the same thing to my head of risk, he will say that everything is a risk factor. of course, there can be zillions of point-of-views on this one ... one def could be that all "return-factors" (i.e. those that generate significantly positive/negative returns over a long time) are also "risk-factors" (i.e. those factors that can reduce the unexplained variance in a cross-sectional regression), but not vice-versa.

Read several interesting posts by pb273, which prompt me joining the forum. Thank you. I've got a question here. Suppose I've got a portfolio of stocks, 50 say, long for now. What should I do if I need to hedge the Barra factor risks? If I calculate the Barra factor exposure of my portfolio, (R-Squared of portfolio returns against a Barra factor?) how to hedge this exposure? Thank you very much.

You mean - how to find hedging ptfs exposed to only one factor? 1m$ question

QuoteOriginally posted by: teddypiggyRead several interesting posts by pb273, which prompt me joining the forum. Thank you. I've got a question here. Suppose I've got a portfolio of stocks, 50 say, long for now. What should I do if I need to hedge the Barra factor risks? If I calculate the Barra factor exposure of my portfolio, (R-Squared of portfolio returns against a Barra factor?) how to hedge this exposure? Thank you very much.hey man, thanks for the comments ... i am sure its not true, but anyway ...btw, when I meant the barra factor exposure, I meant that start with the weights of the stocks in the portfolio and multiply them with the corresponding most recent exposure and sum it up to get the net exposures of the portfolio. the factor exposures are meant (may be?) to be linear ... in a pure long portfolio, you are severely constrained - i guess the best you can do is to use some optimizer to reduce the exposures. first - you should also know why you may want to hedge (or let's say reduce) the barra factor exposure. i would start with the barra vol factors - i have often found many varieties of portfolios, that have twisted towards lower barra vol exposures, to result in similar levels of returns with moderately reduced risks etc.

Can anyone recommend a book with a good relavent intro to Barra Factors/ any papers?Cheers