Factor Model Information on the Web

Northfield Information Systems is one of the better providers of factor models and portfolio optimization systems, and their web site is a good source of information. The documentation for their "Everything Everywhere" risk model provides an especially comprehensive discussion of factor model design. It's a bit long (52 pages), so you might want to look at the web page that lists all of their product documentation and browse for yourself.

I've also added two lists of links, one to providers of factor model (a few of the many), and the other to discussions related to factor models (which are also called "risk models").

Parameter Stability - Part 1

My entry on March 27th showed a regression of 120 months of IBM returns against the Fama French market factor. I was working out the details of getting the data and using Mathematica to do the regression. Now I want to take a closer look at the results. If you are interested in building an economic factor model, you can work out a lot of the methodological details with just one stock. So, continuing to work with the IBM's exposure to Fama French Market Factor, let's look at the stability of the exposure over time. The image below shows a plot of 366 rolling 60-month factor betas, and the corresponding r-squares. Not very stable! However, notice the discontinuities in the r-square plot (enclosed by two red boxes). This strongly suggest that data outliers are causing problems. So, my next step will be to define procedures to identify and deal with outliers.

Look how little Mathematica code it took to do this work (Click on the image below to see the details). I like Mathematica more every day.

Parameter Stability - Part 2

Monthly Stock Returns

Since I want this blog to be useful to people that aren't investment professionals - i.e., don't have a Bloomberg or other validated data sources available to them - I'm going to include comments about data sources. You can download a history of monthly prices from Yahoo, but you have to download dividend history separately, and put prices and dividends together on a spreadsheet to calculate total returns. I wondered if there was an easier way to get monthly returns. I googled "ibm monthly return history," and found the chart below on AOL.


However, AOL's returns for IBM don't match Bloomberg's returns for IBM. Bloomberg shows -0.92 for Jan 08, -1.43 for Oct 07, and -1.26 for Jun 07. I didn't go back any further. June and October 07 are off only a basis point, but Jan 08 is significant. The months with discrepancies aren't months with dividends, so it's not that. For months which do have dividends, only the price return is shown. Although AOL provides a nice table, these returns won't work for our purposes.

Fama French Data Source

The Fama French data is fundamental to QEPM (the book), so I want my readers to know where to get that data. Just follow the link.

QEPM (Start Here)

This is going to be a website based on the content of Quantitative Equity Portfolio Management by Ludwig Chincarini and Daehwan Kim. Their book discusses the conceptual and methodological issues related to creating equity portfolio factor models. I think it's a terrific book, and I plan to use it as the basis for a lot of research in factor models and related topics. I will use this website to show the results of my efforts and to elicit discussion about issues that come up for me as I work on factor models, portfolio optimization, and other topics in the book. I'm going to tell the authors about my project, and hopefully, they will approve, and maybe even post from time to time.