Sorry if this has been covered, but I can't find an explicit answer. If I regress stock price A against stock price B and decide that the residuals are stationary, conventional pairs trading dictates that you go long on one stock and short on the other with the trade size ratio dictated by the 'slope' of the regression. My question is, how is the trade setup affected by the sign of the hedge ratio (regression slope), if the hedge ratio is negative does this mean I go long on both stocks (or not)?
i think you are asking something about pairs of stocks though your choice of words used to describe a trade of such is throwing me (my mind is somewhat antiqued and easily confused). i am sure you know this, but just in case you don't, the trade size of each stock in the pair is generally determined by the price of each stock so that you will be market neutral regarding capital. i am not sure where the slope fits in as i use a chart that has the two stocks combined as a single unit with whatever ratio is pertinent to the 2 financial instruments. i can not imagine a situation where you would buy both stocks unless it is with something like SDS and SPY or DIA. i mean, the whole idea of pairtrading is to be relatively market neutral. If you want to elaborate on "hedge ratio", i may be able to help answer your question or perhaps it is beyond me, i dunno.
Yeah, my post was not particularity clear I admit, pondering this yesterday and drawing a few graphs I realize now that it will always be a long/short setup. The "Hedge Ratio" refers the relative sizes of the trades in the pair. As I understand it this is the 'slope' of the regression line. Does this help?
i am still trying to figure out how there would be a negative hedge ratio with a pair, i mean is it not generally just the % of long $ to short $ or vice versa? if so, it would be equal $ each way so a ratio of 1 regarding $. if you are referring to options with deltas and all that jazz, i don't trade options and haven't a clue. i have a rudimentary understanding of regression lines and their possible negativity so perhpas you are asking if the pair's differential is negative, then is it time to buy or sell and what? if so, then very basically, no fundamentals, correlation, cointegration or long term technicals considered, the thing to do would be to buy the front stock used in the spread and short the second one. like if the spread of MSFT and ORCL (i.e.MSFT minus ORCL) has a negative slope, then buy MSFT and short ORCL. if this is even close to answering or you feel likke elucidating further, i am happy to throw out an answer
Correct, equal dollar value. Neither have I. Perhaps a better question would be how do you identify potential pairs?
First off, look at stocks in the same industry and determine if they are correlated. some good sites for this are http://www.market-topology.com/, http://www.csidata.com/index.html and one that a lot of people have good luck with http://pairtradefinder.com/, check out the pair trading strategy in the journal forum. next, try to figure out their cointegration to see if their spreads move in and out of their range as they move up and dowm. the more spread volatility, the more opportunity to get in and out, either intraday or a swing timeframe. this can be determined by either looking at an intraday chart that combines the 2 stocks or comparing the pair's differential ATR by the points the spread has moved in a certain time period. the higher the number, the better. from there, look to something like PTF (site above)to indicate which pairs have diverged a certain # of standard deviations outside of their sma and are thus, most likely to come back in. check the news to make sure there is nothing crazy going on with either of the stocks, be it earnings, upgrades/downgrades, etc. if you have heaps of cash in your account (or leverage), then you may also scale into and out of the pairs, which requires a consideration of the fundamentals as you may be holding them a while. If you determine a bias long or short then trade it accordingly. As you may be in the trades for months, adding to them as they go against you, it is imperative that you believe in your choice (whether technical or fundamentally based) so you do not get easily shaken out of your positions. Also, with this technique, IMO, it is best if you have at least 10 pairs on at all times spread over a variety of sectors(the more, the better so it is difficult for one pair to kill your day). when to get n on these is determined by where the spread is in the range of the pair's spread over some past time period (i use 2 years, as a general rule but more or less depending on the pair). For instance, if the differential of AMTD-SCHW has had a range of -6 to 2 (with a ratio of 1:1)over the past 2 years (which it has except for the period of wack volatility last fall) and i am fundamentally long (biased towards AMTD), I would start buying the pair (buy AMTD, short SCHW) when it got to about -3 on a day when it is at least down 70% of the 20 day ATR for the pair. this is just one of many ways to trade pairs and even the method of adding layers can be done in many ways. let me know if you have any more specific questions or maybe we will get some input from someone else.
you were right the first time. I see not too many people understand how it works. Simulate long short positions on the neg 'slope' pair and watch what happens (hint: it ain't neutral).
Thanks for all the answers, dtrader, I'm not sure which post you are referring to? Is it possible to define 'neutral' do you mean $ neutral? While I've got every-bodies attention, I'd like to ask another stupid question, why do people talk about doing regression on the log of the price? why is regressing the raw price a bad thing to do?
<ding, light goes on in head> I see! a negative ratio cannot be neutral, simple when you think it though...finally.