Jim, Interesting. I downloaded the paper. I will look at the book as well. Thanks.
Here’s a different way of looking at it:
To create a portfolio, I want to use a variety of assets to decrease the impact of unsystematic risk that could cause an individual stock-company to damage the expected impact of the whole.
To create model factor, I want to use a variety of inputs to decrease the impact of unsystematic risk that could cause an individual data item to damage the expected impact of the whole.
So in fact, the thought process of behind the creation of a portfolio and a model factor are one and the same. In each case, we try to diversify away the risk that an aberration will unduly mess up the whole thing. The processes and the reasonings are identical.
The disconnect you express is coming in terms of defining the problem. Pardon, now if I’m a bit imprecise since I now have to bring my mind back to high school science, but I was taught that the basic component of life is the atom and that it has proton, neutrons and electrons. However, I’ve also come to understand that this is not as complete a statement as my high school teacher led us to believe. Seems that protons, neutrons and electrons have their own sets of components. When you set up a diversified portfolio, you’re building an atom by combining protons, neutrons and electrons. When you work with factors and formulas, you’re combining things to build your protons (your value factors), your neutrons (your quality factors) and electrons (your growth factors).
In case I really screwed up the atom analogy, bear in mind that the only way I was able to pass high school chemistry was by working for the teacher (extracurricular thing) and I graded my own tests.
Marc,
Thanks for the comments. You are righting about defing the problem.
Another way I thought about it was the use of the Dupont model to express ROE.
Maybe I am still looking at this wrong, but…the three uncorrelated (at least when looked at separately) ratios of profit, asset turnover and debt leverage, when used together, can express a higher ROE. But each by itself looks quite different from the others in terms of the impact on the business.
So these three seemingly uncorrelated factors work together to solve a problem.
I have had enough finance courses now that I see an underlying idea is one of interlocking relationships. They are expressed mathematically but they have basis in realistically how a business is run. Not just throwing factors against a wall and seeing what sticks.
Jim’s reference to this book and all of these comments have helped me gel some concepts about better model construction.
One wants ‘diversity’ in models - as long as they collectvely are all pulling in the right direction. Not to get philosophical about it…
Not just that – diversity within each idea expressed in the model. exampe: using pe to define value is like a one-stock portfolio. Defining value as the constellation of pe, ps, pb, pfcf, etc. is analogous to a diversified portfolio. You may, for example, diversifythe impact of a ps that looks to high because ttm sales includes a divested business, or pe that looks very low becasue last year’s eps was inflated by an asset-sale gain, etc. etc. etc.
James O’Shaughnessy seems to have evolved to the position of using multiple value factors.
I believe he had just one value ratio for Tiny Titans (price to sales).
I have lost track recently but he was up to at least 5 value ratios last time I looked.
-Jim