Which price ratio best identifies undervalued stocks?

It’s a fraught question, dependent on various factors including the time period tested, and the market capitalization and industries under consideration, but I believe a consensus is emerging.

The academic favorite remains book value-to-market capitalization (the inverse of price-to-book value). Fama and French maintain that it makes no difference which “price-to-a-fundamental” is employed, but if forced to choose favor book-to-market. In the Fama/French Forum on Dimensional Fund Advisor’s website they give it a tepid thumbs up despite the evidence that it’s not so great:

Data from Ken French’s website shows that sorting stocks on E/P or CF/P data produces a bigger spread than BtM over the last 55 years. Wouldn’t it make sense to use these other factors in addition to BtM to distinguish value from growth stocks? EFF/KRF: A stock’s price is just the present value of its expected future dividends, with the expected dividends discounted with the expected stock return (roughly speaking). A higher expected return implies a lower price.We always emphasize that different price ratios are just different ways to scale a stock’s price with a fundamental, to extract the information in the cross-section of stock prices about expected returns. One fundamental (book value, earnings, or cashflow) is pretty much as good as another for this job, and the average return spreads produced by different ratios are similar to and, in statistical terms, indistinguishable from one another. We like BtM because the book value in the numerator is more stable over time than earnings or cashflow, which is important for keeping turnover down in a value portfolio. Nevertheless, there are problems in all accounting variables and book value is no exception, so supplementing BtM with other ratios can in principal improve the information about expected returns. We periodically test this proposition, so far without much success.

research seemed to say (See, for example, Roger Ibbotson’s “Decile Portfolios of the New York Stock Exchange, 1967 – 1984,” Werner F.M. DeBondt and Richard H. Thaler’s “Further Evidence on Investor Overreaction and Stock Market Seasonality”). Josef Lakonishok, Andrei Shleifer, and Robert Vishny’s Contrarian Investment, Extrapolation and Risk, which was updated by The Brandes Institute as Value vs Glamour: A Global Phenomenon reopened the debate, suggesting that price-to-earnings and price-to-cash flow might add something to price-to-book.

A number of more recent papers have moved away from book-to-market, and towards the enterprise multiple ((equity value + debt + preferred stock – cash)/ (EBITDA)). As far as I am aware, Tim Loughran and Jay W. Wellman got in first with their 2009 paper “The Enterprise Multiple Factor and the Value Premium,” which was a great unpublished paper, but became in 2010 a slightly less great published paper, “New Evidence on the Relation Between the Enterprise Multiple and Average Stock Returns,” suitable only for academics and masochists (but I repeat myself). The abstract to the 2009 paper (missing from the 2010 paper) cuts right to the chase:

Following the work of Fama and French (1992, 1993), there has been wide-spread usage of book-to-market as a factor to explain stock return patterns. In this paper, we highlight serious flaws with the use of book-to-market and offer a replacement factor for it. The Enterprise Multiple, calculated as (equity value + debt value + preferred stock – cash)/ EBITDA, is better than book-to-market in cross-sectional monthly regressions over 1963-2008. In the top three size quintiles (accounting for about 94% of total market value), EM is a highly significant measure of relative value, whereas book-to-market is insignificant.

The abstract says everything you need to know: Book-to-market is widely used (by academics), but it has serious flaws. The enterprise multiple is more predictive over a long period (1963 to 2008), and it’s much more predictive in big market capitalization stocks where book-to-market is essentially useless.

What serious flaws?

The big problem with book-to-market is that so much of the return is attributable to nano-cap stocks and “the January effect”:

Loughran (1997) examines the data used by Fama and French (1992) and finds that the results are driven by a January seasonal and the returns on microcap growth stocks. For the largest size quintile, accounting for about three-quarters of total market cap, Loughran finds that BE/ME has no significant explanatory power over 1963-1995. Furthermore, for the top three size quintiles, accounting for about 94% of total market cap, size and BE/ME are insignificant once January returns are removed. Fama and French (2006) confirm Loughran’s result over the post- 1963 period. Thus, for nearly the entire market value of largest stock market (the US) over the most important time period (post-1963), the value premium does not exist.

That last sentence bears repeating: For nearly the entire market value of largest stock market (the US) over the most important time period (post-1963),the value premium does not exist, which means that book-to-market is not predictive in stocks other than the smallest 6 percent by market cap. What about book-to-market in the stocks in that smallest 6 percent? It might not work there either:

Keim (1983) shows that the January effect is primarily limited to the first trading days in January. These returns are heavily influenced by December tax-loss selling and bid-ask bounce in low-priced stocks. Since many fund managers are restricted in their ability to buy small stocks due to ownership concentration restrictions and are prohibited from buying low-prices stocks due to their speculative nature, it is unlikely that the value premium can be exploited.

More scalable

The enterprise multiple succeeds where book-to-market fails.

In the top three size quintiles, accounting for about 94% of total market value, EM is a highly significant measure of relative value, whereas BE/ME is insignificant and size is only weakly significant. EM is also highly significant after controlling for the January seasonal and removing low-priced (<$5) stocks. Robustness checks indicate that EM is also better to Tobin’s Q as a determinant of stock returns.

And maybe the best line in the  paper:

Our results are an improvement over the existing literature because, rather than being driven by obscure artifacts of the data, namely the stocks in the bottom 6% of market cap and the January effect, our results apply to virtually the entire universe of US stocks. In other words, our results may actually be relevant to both Wall Street and academics.

Why does the enterprise multiple work?

The enterprise multiple is a popular measure, and for other good reasons besides its performance. First, the enterprise multiple uses enterprise value. A stock’s enterprise value provides more information about its true cost than its market capitalization because it includes information about the stock’s balance sheet, including its debt, cash and preferred stock (and in some variations minorities and net payables-to-receivables). Such things are significant to acquirers of the business in its entirety, which, after all, is the way that value investors should think about each stock. Market capitalization can be misleading. Just because a stock is cheap on a book value basis does not mean that it’s cheap 0nce its debt load is factored into the valuation. Loughran and Wellman, quoting Damodaran (whose recent paper I covered here last week), write:

Damodaran shows in an unpublished study of 550 equity research reports that EM, along with Price/Earnings and Price/Sales, were the most common relative valuation multiples used. He states, “In the past two decades, this multiple (EM) has acquired a number of adherents among analysts for a number of reasons.” The reasons Damodaran cites for EM’s increasing popularity also point to the potential superiority of EM over book-to-market. One reason is that EM can be compared more easily across firms with differing leverage. We can see this when comparing the corresponding inputs of EM and BE/ME. The numerator of EM, Enterprise Value, can be compared to the market value of equity. EV can be viewed as a theoretical takeover price of a firm. After a takeover, the acquirer assumes the debt of the firm, but gains use of the firm’s cash and cash equivalents. Including debt is important here. To take an example, in 2005, General Motors had a market cap of $17 billion, but debt of $287 billion. Using market value of equity as a measure of size, General Motors is a mid-sized firm. Yet on the basis of Enterprise Value, GM is a huge company. Market value of equity by itself is unlikely to fully capture the effect GM’s debt has on its returns. More generally, it is reasonable to think that changing firm debt levels may affect returns in a way not fully captured by market value of equity. Bhojraj and Lee (2002) confirm this, finding that EV is superior to market value of common equity, particularly when firms are differentially levered.

The enterprise multiple’s ardor for cash and abhorrence for debt matches my own, hence why I like it so much. In practice, that tendency can be a double-edged sword. It digs up lots of little cash boxes with a legacy business attached like an appendix (think Daily Journal Corporation (NASDAQ:DJCO)or Rimage Corporation (NASDAQ:RIMG)). Such stocks tend to have limited upside. On the flip side, they also have happily virtually no downside. In this way they are vastly superior to the highly leveraged pigs favored by book-to-market, which tends to serve up heavily leveraged slivers of somewhat discounted equity, and leaves you to figure out whether it can bear the debt load. Get it wrong and you’ll be learning the intricacies of the bankruptcy process with nothing to show for it at the end. When it comes time to pull the trigger, I generally find it easier to do it with a cheap enterprise multiple than a cheap price-to-book value ratio.

The earnings variable: EBITDA

There’s a second good reason to like the enterprise multiple: the earnings variable. EBITDA contains more information than straight earnings, and so should give a more full view of where the accounting profits flow:

The denominator of EM is operating income before depreciation while net income (less dividends) flows into BE. The use of EBITDA provides several advantages that BE lacks. Damodaran notes that differences in depreciation methods across companies will affect net income and hence BE, but not EBITDA. Also, the McKinsey valuation text notes that operating income is not affected by nonoperating gains or losses. As a result, operating income before depreciation can be viewed as a more accurate and less manipulable measure of profitability, allowing it to be used to compare firms within as well as across industries. Critics of EBITDA point out that it is not a substitute for cash flow; however, EV in the numerator does account for cash.

The enterprise multiple includes debt as well as equity, contains a clearer measure of operating profit and captures changes in cash from period to period. The enterprise multiple is a more complete measure of relative value than book-to-market. It also performs better:

Performance of the enterprise multiple versus book-to-market

From CXOAdvisory:

• EM generates an annual value premium of 5.8% per year over the entire sample period (compared to 4.8% for B/M during 1926-2004).
• EM captures more premium than B/M for all five quintiles of firm size and is much less dependent on small stocks for its overall premium (see chart below).
• In the top three quintiles of firm size (accounting for about 94% of total market capitalization), EM is a highly significant measure of relative value, while B/M is not.
• EM remains highly significant after controlling for the January effect and after removing low-priced (<$5) stocks.
• EM outperforms Tobin’s q as a predictor of stock returns.
• Evidence from the UK and Japan confirms that EM is a highly significant measure of relative value.

The “value premium” is the difference in returns to a portfolio of glamour stocks (i.e., the most expensive decile) when compared to a portfolio of value stocks (i.e., the cheapest decile) ranked on a given price ratio (in this case, the enterprise multiple and book-to-market). The bigger the value premium, the better a given price ratio sorts stocks into winners and losers. It’s a more robust test than simply measuring the performance of the cheapest stocks. Not only do we want to limit our sins of commission (i.e., buying losers), we want to limit our sins of omission (i.e., not buying winners). 

Here are the value premia by market capitalization (from CXOAdvisory again):Ring the bell. The enterprise multiple kicks book-to-market’s ass up and down in every weight class, but most convincingly in the biggest stocks.

Strategies using the enterprise multiple

The enterprise multiple forms the basis for several strategies. It is the price ratio limb of Joel Greenblatt’s Magic Formula (the other limb is of course return on invested capital, which I like about as much as Hunter S. Thompson liked Richard Nixon, about whom he said in his obituary:

[The] record will show that I kicked him repeatedly long before he went down. I beat him like a mad dog with mange every time I got a chance, and I am proud of it. He was scum.

But I digress.) It also forms the basis for the Darwin’s Darlings strategy that I love (see Hunting Endangered Species). The Darwin’s Darlings strategy sought to front-run the LBO firms in the early 2000s, hence the enterprise multiple was the logical tool, and highly effective.

Conclusion

This post was motivated by the series last week on Aswath Damodaran’s paper ”Value Investing: Investing for Grown Ups?” in which he asks, “If value investing works, why do value investors underperform?” Loughran and Wellman also asked why, if Fama and French (2006) find a value premium (measured by book-to-market) of 4.8% per year over 1926-2004, mutual fund managers couldn’t capture it:

Fund managers perennially underperform growth indices like the Standard and Poor’s 500 Index and value fund managers do not outperform growth fund managers. Either the value premium does not actually exist, or it does not exist in a way that can be exploited by fund managers and other investors.

Loughran and Wellman find that for nearly the entire market value of largest stock market (the US) over the most important time period (post-1963), the value premium does not exist, which means that book-to-market is not predictive in stocks other than the smallest 6 percent by market cap (and even there the returns are suspect). The enterprise multiple succeeds where book-to-market fails. In the top three size quintiles, accounting for about 94% of total market value, the enterprise multiple is a highly predictive measure, while book-to-market is insignificant. The enterprise multiple also works after controlling for the January seasonal effect and after removing low priced (<$5) stocks. The enterprise multiple is king. Long live the enterprise multiple.