Cliff Asness was writing about how he can see when value is being invested in and when it is not. He looks at the spread between value and growth… if the price-to-factor on value and the price-to-factor on growth are narrowing, folks are betting on value - and vice versa. He says the spread gives a good longterm indicator of when the value style will outperform (larger the spread the better the future performance).
Since factor analysis here at p123 is followed by many quant funds, it would be good to see the “spread” on our factors. It would give us a sense of what factors the hedgies are piling into, or out of.
I would like to see average PriceToFactor over time for the ranks where all factors have a “price to” element. (To get price to book, create a rank with one factor, price to book; to get a complex price2 rank, use price to book, price to sales, and price to earnings; and so forth). I would like to see this spread for buckets 1 - 10 (assuming only 10 buckets; i.e., rank 0-9.99 = bucket 1, 10 - 19.99=2, … 90-100=10: something like that). We could get more granular, or even work directly with ranks/ranges of ranks, but this “bucket-based” spread would be enough for now.
So a graph would look like the rank performance line graph, but the lines would be the price to composite ranking (requires that elements in ranking all be Price2factors). I think we would see some divergence/convergence here, and adding some formulas to manipulate the spread, we could build timing systems. I suspect we would have seen a narrow spread on Price2Sales between the bottom rank and the 10th rank back in the second quarter of 2007 shortly before price to sales blew up.
So request is:
Spread on Price2Factor rankings for 10 buckets
Formulas to manipulate, e.g.,
SMA(#RankSpread(1,10),50) would be the 50 day SMA of the spread between bucket 1 and 10
View the feature request here.
This is an excellent request and I hope other p123 members vote for it, because I suspect it could provide early warning signals of imminent regime changes.
In the book “Quantitative Strategies for Achieving Alpha” you clearly see Value (or Quality) fall out of favor in 1999-2000 as Growth (or Momentum) ruled the markets. It was so extreme that many measures which are logical like P/S were turned upside down in their predictive ability.
Looks like we have a study that puts my notion into a nice, rigorous structure. It would be great if we could do this type of analysis to help select from our ports which to go with. Add to the HML netting feature the old request to have Ports as ETFs, so we could see if relative strength and/or mean reversion plays a role in port selection.
Alpha Architect has just posted another study on spread analysis of factor based portfolios. This time they feature a study on momentum that shows that when the gap in performance between your top ranked momentum portfolio (top 25% - top quartile) and your bottom ranked momentum portfolio (bottom quartile) performed poorly in the training period, then momentum will perform better ex-ante. So, momentum is a mean reverting factor. This study, along with previous studies listed in this thread, along with the Goldman Sachs “quantcentration” study I’ve posted elsewhere (use search to find) on this board, show that spread analysis of top & bottom factor performance is predictive of when a factor works. Sure would be nice to have this feature…
Good idea. You can now do a work-around to this, although you can’t chart it. You simply create 2 custom universes (one for top mo stocks and one for bottom mo stocks). Then create a custom series for each tracking their average momentum changes. Then, in your sim or screener, you take the difference in them.
However, you will be ‘working blind’ - so, if you we could call Custom Universes or use Frank inside of custom series - and then take moving averages of these - that would be much better.
So… a feature request like this - that is more generalizable, but still meets your specific needs, would probably be best.
We also need (ideally) the ability to call Custom Series in rankings. I’ve posted this before. Please vote for it.