A Dynamic Model of Characteristic‐Based Return Predictability

• Author: SHERIDAN TITMAN,AYDOĞAN ALTI
• Published date: 01 December 2019
• DOI: http://doi.org/10.1111/jofi.12839
• Date: 01 December 2019

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

A Dynamic Model of Characteristic-Based

Return Predictability

AYD O ˘

GAN ALTI and SHERIDAN TITMAN∗

ABSTRACT

We present a dynamic model that links characteristic-based return predictability to

systematic factors that determine the evolution of firm fundamentals. In the model,

an economy-wide disruption process reallocates profits from existing businesses to

new projects and thus generates a source of systematic risk for portfolios of firms

sorted on value, profitability, and asset growth. If investors are overconfident about

their ability to evaluate the disruption climate, these characteristic-sorted portfolios

exhibit persistent mispricing. The model generates predictions about the conditional

predictability of characteristic-sorted portfolio returns and illustrates how return

persistence increases the likelihood of observing characteristic-based anomalies.

SINCE THE LATE 1970S,FINANCIAL economists have identified a number of firm

characteristics like valuation ratios, profitability, and asset growth rates that

explain the cross-section of stock returns. Historically, market-neutral portfo-

lios that are sorted on these characteristics have exhibited very high Sharpe

ratios—at least as high as the Sharpe ratio of the market. Although the litera-

ture has mostly focused on these historical return patterns, there is relatively

little research on the conditional relationships between characteristics and

returns. This is somewhat surprising because the cross-section of firm char-

acteristics in the economy, as well as the relation between characteristics and

firm values, exhibits substantial time variation (e.g., the increased prevalence

and valuations of growth firms in the late 1990s).

∗Both authors are with the University of Texas at Austin. An earlier draft of the paper was cir-

culated under the title “Creative Destruction, Investor Beliefs and the Evolution of Stock Returns.”

We benefited from comments and suggestions by WeiXiong (the Editor); the Associate Editor; two

anonymous referees; Zhanhui Chen (discussant); Nicolae Gˆ

arleanu (discussant); David Hirshleifer;

Travis Johnson; Mao Lei (discussant); Kristian Rydqvist(discussant); David Solomon (discussant);

and participants in presentations of the paper at the 2016 Rising Stars Conference at Fordham

University,Workshop in Memory of Rick Green at Carnegie Mellon University, 2016 Asian Bureau

of Finance and Economic Research Conference in Singapore, Asset Pricing Conference in Honor of

John Wei at HKUST, 2016 UBC Summer Finance Conference, 2017 China International Confer-

ence in Finance, 2018 American Finance Association Meeting, Emory University, HEC Lausanne,

Office of Financial Research, SAIF, Universidad de Chile, University of Colorado Boulder and

Denver, University of Georgia, University of Southern California, University of Texas at Austin,

and Vienna University of Economics and Business. We thank WillShuo Liu for excellent research

assistance. Altı has no conflicts of interest to disclose. Titman has advised firms in the asset

management business. These businesses have not provided any financial support or in any way

influenced the content of this research.

DOI: 10.1111/jofi.12839

C2019 the American Finance Association

3187

3188 The Journal of Finance R

This paper explores the evolution of firm characteristics and their links to re-

turn predictability within the context of a dynamic behavioral model. The model

assumes that the profitability and growth rates of firms are affected by what we

refer to as the disruption climate, which is an economy-wide factor that creates

losers as well as winners. The behavioral elements of the model arise because

investors are overconfident about their ability to assess the disruption climate,

and because of this, firms with different exposures to disruption—for example,

value versus growth firms—exhibit predictable return differences. The model’s

testable implications relate these return differences to observable measures of

the conditional bias in investor beliefs. For instance, cross-sectional dispersions

of characteristics, which reflect the hard and soft indicators of disruption that

investors observe, predict subsequent returns of characteristic-sorted portfo-

lios. The model also illustrates how the persistence of characteristic-sorted

returns, which obtains due to slow-changing investor beliefs, can substantially

increase the likelihood of observing characteristic-based anomalies such as the

value premium in samples that are of comparable length to those in empirical

studies.

Firms in the model are characterized by differences in their current access to

new growth opportunities, as well as by their different histories. Growth firms

are endowed with new projects each period while value firms simply harvest

the profits from their existing projects. The emergence of new projects, as well

as the demise of existing ones, is determined by a systematic factor that we

refer to as the disruption climate. A favorable disruption climate increases the

arrival rate of new projects, which benefits young growth firms, but because

these new projects compete with existing businesses, a favorable disruption

climate harms the profits of assets in place, and is thus detrimental to mature

value firms. The model thus captures the Schumpeterian notion of creative

destruction, whereby innovation creates losers as well as winners.

To abstract from differences in risk premia, investors in our model are as-

sumed to be risk-neutral. These investors learn about the disruption climate

from two sources, the realized rate of disruptive innovations, and a soft in-

formation signal that represents, for example, news reports and expert opin-

ions. Since both sources are noisy indicators of the disruption climate, investor

expectations contain estimation errors, which imply that even fully rational

investors are sometimes too optimistic and sometimes too pessimistic about

the rate of future disruptive innovations. However, these estimation errors do

not generate predictable returns when investors are rational—some degree of

irrationality is needed to generate asset pricing anomalies.

We introduce the possibility of biased inferences by assuming that investors

are overconfident about the precision of their soft information, which implies

that their estimates of the disruption climate puts too much weight on the soft

information signal. This behavioral bias does not cause investors to systemat-

ically over- or underestimate the disruption climate, that is, the unconditional

or the long-run expected return associated with disruption rate surprises is

zero. However, because overconfident investors learn slowly, conditional ex-

pected returns differ from zero and change slowly over time. Put differently,