Where Is the Risk in Value? Evidence from a Market‐to‐Book Decomposition

• DOI: http://doi.org/10.1111/jofi.12836
• Author: ANDREY GOLUBOV,THEODOSIA KONSTANTINIDI
• Date: 01 December 2019
• Published date: 01 December 2019

THE JOURNAL OF FINANCE •VOL. LXXIV, NO. 6 •DECEMBER 2019

Where Is the Risk in Value? Evidence from a

Market-to-Book Decomposition

ANDREY GOLUBOV and THEODOSIA KONSTANTINIDI∗

ABSTRACT

Westudy the value premium using the multiples-based market-to-book decomposition

of Rhodes-Kropf, Robinson, and Viswanathan (2005). The market-to-value component

drives all of the value strategy return, while the value-to-book component exhibits

no return predictability in either portfolio sorts or firm-level regressions. Existing re-

sults linking market-to-book to operating leverage, duration, exposure to investment-

specific technology shocks, and analysts’ risk ratings derive from the unpriced value-

to-book component. In contrast, results on expectation errors, limits to arbitrage, and

certain types of cash flow risk and consumption risk exposure are due to the market-

to-value component. Overall, our evidence casts doubt on several value premium

theories.

THE POSITIVE RETURN DIFFERENTIAL between high book-to-market (value) and

low book-to-market (growth) stocks is one of the most pervasive phenomena

in the behavior of stock prices, having been documented in many markets

around the world (e.g., Fama and French (1998,2012), Asness, Moskowitz,

and Pedersen (2013)). Naturally, a substantial stream of asset pricing research

is concerned with the economic origins of the book-to-market effect. Multiple

theories attempt to reconcile the value premium with models of investor and

firm behavior, and many of these theories find empirical support. Considerable

debate remains, however, about the exact mechanism giving rise to the value

∗Andrey Golubov is at Rotman School of Management, University of Toronto. Theodosia Kon-

stantinidi is at Cass Business School, City, University of London. We thank two anonymous ref-

erees; the Editor (Stefan Nagel); Pat Akey; Malcolm Baker; Eli Bartov; Akash Chattopadhyay;

Alexandre Corhay; Ettore Croci; Zhi Da; Patricia Dechow; Olivier Dessaint; Huseyin Gulen;

Raymond Kan; Alexandros Kostakis; Yan Li; Juhani Linnainmaa; Mamdouh Medhat; Partha

Mohanram; Panos Patatoukas; Bradley Paye; Dimitris Petmezas; Christopher Polk; Matthew

Rhodes-Kropf; George Serafeim; Mikhail Simutin; Theodore Sougiannis; Chay Ornthanalai; Ali

Sharifkhani; Richard Sloan; Alfred Yawson; Huizhong Zhang; participants at the AAA 2016, AFA

2018, CFEA 2016, FMA 2016, and NFA 2016 annual conferences; as well as seminar partici-

pants at Aalto University, Bocconi University, City, University of London (Cass), London School

of Economics, Universidad Auton´

oma de Madrid, and University of Toronto (Rotman) for helpful

comments and suggestions. We also thank Petri Jylha and Joni Kokkonen for sharing arbitrage

capital data with us. Any errors are our own. The authors have read the Journal of Finance Conflict

of Interest Disclosure Policy and have nothing to disclose.

DOI: 10.1111/jofi.12836

C2019 the American Finance Association

3135

3136 The Journal of Finance R

premium and whether some of the proposed theories are more consistent with

the data than others.1

In this paper, we show that a number of prominent theories related to the

value premium are actually at odds with the data, and the few stories that

withstand our tests face other challenges. We test the various theories using

the market-to-book decomposition introduced by Rhodes-Kropf, Robinson, and

Viswanathan (2005, RRV hereafter) in their study of merger waves. In par-

ticular, we decompose market-to-book into market-to-value and value-to-book

components, where value is an estimate of fundamental value based on in-

dustry valuations and a set of observable characteristics. The market-to-value

component represents the stock price deviation from the valuation implied by

long-run industry multiples, which we refer to as the total error. This compo-

nent is further decomposed into the stock price deviation from the contempora-

neous peer-implied valuation (firm-specific error) and the deviation of the latter

from the valuation implied by long-run industry multiples (sector error).2

Our baseline results show that the entire value premium concentrates in

the market-to-value component. Over the 1975 to 2013 period, a long-short

portfolio strategy based on the conventional market-to-book ratio produces an

average return of 0.75% per month in return-weighted (RW) portfolios and

0.59% in value-weighted (VW) portfolios. The same strategy based on market-

to-value produces an average RW (VW) return of 0.75% (0.43%), while the

return spread between low and high value-to-book portfolios is about 10 basis

points per month and statistically insignificant regardless of the weighting.

Further decomposition of market-to-value shows that return predictability is

driven by firm-specific error, whereas sector error exhibits no significant associ-

ation with future stock returns. Firm-level stock return regressions controlling

for numerous other firm-level characteristics produce consistent results: the

market-to-value component, and in particular firm-specific error, subsumes all

of the value premium.

Conceptually, deviations of market value from our estimates of fundamen-

tal value can arise for two reasons. First, industry-year multiples may fail to

fully capture cross-sectional differences in value-relevant attributes, leading

to biased estimates of fundamental value. If these differences represent priced

sources of risk, subsequent returns represent compensation for unmodeled

risk factors.3Second, deviations may be due to relative over/undervaluation,

1Another possibility is that return predictability in general is an artifact of data snooping (e.g.,

Lo and MacKinlay (1990), Fama (1991,1998), Conrad, Cooper,and Kaul (2003)). This is an unlikely

explanation for the value premium, however, as it has been documented in several time periods,

asset classes, and markets (see, for example, Barber and Lyon (1997), Fama and French (1998),

Davis, Fama, and French (2000), Asness, Moskowitz, and Pedersen (2013)). Further, Harvey, Liu,

and Zhu (2016) show that the t-statistic of the HML factor is comfortably above the critical t-value

adjusted for publication bias.

2We recognize that any estimate of value likely deviates from “true” fundamental value. We

therefore use the term “error” loosely.

3Toillustrate this point, assume that we attempt to value a firm that is riskier than its industry-

year peers. In this case, we would be using valuation multiples that are too high (discount rates

Where Is the Risk in Value? 3137

suggesting that subsequent returns represent corrections of prices toward fun-

damental value. The latter would require mechanisms through which stock

prices become and remain dislocated for a prolonged period of time (De Long

et al. (1990), Shleifer and Vishny (1997)).

We examine whether existing results in the value premium literature con-

tinue to hold for the component of market-to-book that is actually priced. Re-

cent evidence suggests that market-to-book captures exposure to investment-

specific technology (IST) shocks. Kogan and Papanikolaou (2014) find that

growth stocks are more sensitive to changes in prices of investment goods com-

pared to value stocks, and that this exposure earns a negative risk premium.

Technological shocks tend to lower the cost of investment goods, and value

stocks miss out on those benefits. We find that the value strategy does capture

exposure to IST shocks, but this is due largely to the value-to-book component.

Therefore, exposure to IST shocks is an unlikely explanation for the value

premium.

We further explore operating leverage—a focal feature of production-based

models that potentially gives rise to the value premium (Carlson, Fisher, and

Giammarino (2004), Zhang (2005), Novy-Marx (2011)). Operating leverage, in

the form of fixed costs of production, makes assets-in-place riskier than growth

options, and market-to-book is commonly believed to capture variation in the

mix of assets-in-place versus growth options. Using various proxies, we show

that differences in the mix of assets-in-place versus growth options across

market-to-book portfolios are due to value-to-book. There are no differences in

assets-in-place intensity across market-to-value portfolios. Therefore, even if

operating leverage is a priced source of risk, it is unlikely to be the mechanism

behind the value premium.

Cash flow duration is another firm characteristic that has been linked to

market-to-book (Lettau and Wachter (2007)). Several studies show empirically

that value stocks have shorter cash flow durations than growth stocks (Dechow,

Sloan, and Soliman (2004), Da (2009), Chen (2017)). We again show that differ-

ences in cash flow duration are due to the unpriced value-to-book component.

That is, duration cannot explain the value premium.

In the accounting literature, Lui, Markov, and Tamayo (2007) show that

equity analysts perceive value stocks to be riskier than growth stocks. While

we confirm a negative association between analysts’ risk ratings and market-

to-book, this correlation is once again driven by the unpriced value-to-book.

We emphasize that we do not take a stance on whether operating leverage,

duration, exposure to IST shocks, or analysts’ risk ratings represent priced

sources of risk—we only examine their relationship with the value premium.4

that are too low), resulting in an inflated estimate of fundamental value. This, in turn, would lead

to a lower estimate of market-to-value. Consequently, the higher returns earned by firms with

lower market-to-value would be consistent with risk-based pricing.

4The value premium has also been linked to distress risk (e.g., Fama and French (1992)).

However, this theory has found little empirical support, and thus, we do not reexamine it. See

Griffin and Lemmon (2002), Campbell, Hilscher, and Szilagyi (2008), and Da and Gao (2010).