Having explored the history of the price-to-book ratio, we can now turn to its usefulness as a stock selection criterion. The data suggests a few important points about the price-to-book ratio:
- It has worked quite nicely in small-cap
- It has not worked as well in large-cap stocks
- Price-to-book delivers the best returns when it is used to compare each stock against all others, but requires taking large sector bets
- Price-to-tangible book may be a slight improvement over regular price-to-book, but not by much
- The very cheapest stocks (those with the lowest price-to-book ratios) have performed poorly
- Price-to-book should not be used in isolation
Here are decile returns (rebalanced on an annual basis) for price to book in two universes: large stocks (anything will a larger than average market cap) and all stocks (large stocks + any smaller stock down to $200MM in market cap) since 1963. Clearly, the factor has worked more effectively when smaller stocks are a part of the opportunity set (all stocks).
One key thing to note about price-to-book: it has tended to push investors towards financials, utilities, materials and industrials, and push them away from health care, technology, and consumer staples. Here is a historical sector weight distribution for the cheapest decile of stocks by price-to-book.
If you adjust the price-to-book calculation for economic sector or industry group (so you compare the ratio only to stocks in the same sector to determine relative cheapness), the factor works (not quite as well) but significantly smoothest sector exposures:
In my first post on the history of the price-to-book ratio, I noted that book value has become far less “tangible” over time: a large chunk of today’s total market book value comes from non-psychical assets like goodwill and intellectual property of various types. For example, here is the percentage of total assets represented by just goodwill across sectors.
And here are a few top names today, listed by goodwill as a percentage of total asset value.
A second way to calculate price-to-book adjusts for (by removing) the value of the non-physical assets. The result is price-to-tangible-book. Here is a comparison of the two ratios since 1988 (when data for goodwill and intangibles becomes widely available). For these results, I’ve REMOVED any stock with negative tangible book value (which would all fall into the worst decile, but perform fine, roughly in line with the overall market).
You can see that these ratios provide very similar results. For those uncomfortable with valuing intangible assets, price-to-tangible book may be the better ratio to use.
One key final point: price-to-book has NOT been useful for building concentrated portfolios, and should therefore only be considered as a factor tilt (think smart beta) rather than a tool to be used in isolation for individual stock selection.
To see why, consider this final chart (I’ve included a methodology behind this chart at the bottom, it’s a little tricky). The general idea is that we ran simulations which randomly lop off half the stocks eligible for selection and half the dates historical to test how the price-to-book ratio works on different subsets of the universe (you’d like to see the factor work in any random group of stocks). The results tell us that even if you get really unlucky with the batch of stocks you have to pick from, the ratio still has delivered some excess return (albeit very little). This test is done at different levels of concentration from 1-stock per month to 300-stocks per month.
Two messages stand out: 1) price-to-book has delivered some small excess return even when the universe is “unlucky,” but 2) you should not build concentrated portfolios based only on price-to-book as they’ve tended to lose to the market by wide margins.
This has been a somewhat high-level fly by on the price-to-book factor. What other details would you like to know? Let me know below. I started with this ratio as I believe it to be the least useful valuation ratio. It has worked, but not nearly as well as several others that we will explore. Next up: the price-to-sales ratio.
Methodology: To arrive at the final chart, we ran 1,000 simulations which did the following: 1) for every monthly date since 1962, randomly remove half of the stocks in the all stocks universe (so say go from 3,000 down to 1,500). Then, build a portfolio which buys x number of stocks per month with an annual holding period (listed on the x-axis). Then 2) calculate the forward monthly returns of that portfolio (i.e. for each level of concentration, be it a 5-stock or 300-stock portfolio based on price-to-book). Then 3) randomly lop off half of the monthly returns. Finally, calculate the excess returns above or below all other stocks and dates that made the cut. This does two things: it gives the investor confidence that a factor isn’t reliant on any one set of stocks to work. The middle line is the average observation, where we see a similar pattern as the decile results from earlier (price-to-book peaking at around 3% annualized excess return). The top and bottom lines are the 100th and 900th observations out of the 1000 that we ran, which act as confidence intervals.
Another Note: These results are all for U.S. domiciled stocks only (I’ve sometimes included ADRs in the past but for this factor series will focus on U.S. stocks)