What Drives Stock Price Movements?

[Pages:36]Downloaded from at University of Notre Dame on February 24, 2013

RFS Advance Access published February 23, 2013

What Drives Stock Price Movements?

Long Chen Olin School of Business and CKGSB

Zhi Da Mendoza College of Business, University of Notre Dame

Xinlei Zhao Office of the Comptroller of the Currency and Kent State University

A central issue in finance is whether stock prices move because of revisions in expected cash flows or discount rates, and by how much of each. Using direct cash flow forecasts, we show that stock returns have a significant cash flow news component whose importance increases with the investment horizon. For horizons over two years, cash flow news is more important. These conclusions hold at both the firm and aggregate levels, and diversification plays a secondary role in affecting the relative importance of cash flow and discount rate news. Our findings highlight the importance of cash flows in asset pricing. (JEL G12, E44)

As investors, policymakers, and economists during the recent financial crisis were debating the likelihood of another great depression in 2008, accompanying the stock market plunge, the financial market revised downward its forecasts of five-years-ahead aggregate earnings. This is a consistent pattern over the period 1986?2010: The correlation between the revision of the five-year earnings forecasts and a recession dummy is -71% (Figure 1). It seems natural to conclude that a significant portion of stock price movement occurs because, when evaluating stocks, investors revise their expectations of future cash flows.

Yet this is not what one would conclude from the bulk of the asset pricing literature that examines the drivers of stock price movement. Conceptually, stock prices can move unexpectedly because investors update expectations

We thank Bo Becker, George Benston, Michael Boldin, Jeff Callen, John Campbell, John Cochrane, Ilan Cooper, Peter Easton, Pengjie Gao, Keejae Hong, Narasimhan Jegadeesh, Timothy Johnson, Raymond Kan, Andrew Karolyi, Soohun Kim, Jun Liu, George Pennacchi, Richard Priestley, Jesper Rangvid, Dan Segal, Jay Shanken, Jianfeng Shen, Rene Stulz, Michael Weisbach, Avi Wohl and Dexin Zhou; and Geert Bekaert (the editor) and three anonymous referees; and seminar participants at Beijing University, the Central Bank of Denmark, Copenhagen Business School, Hebrew University of Jerusalem, Hong Kong University of Science and Technology, McGill University, Norwegian School of Management (BI), Ohio State University, Tel Aviv University, Tsinghua University, University of Emory, University of Hong Kong, University of Illinois at UrbanaChampaign, University of Toronto, Villanova University, Washington University in St. Louis, and 2008 WFA annual meeting for their helpful comments. The usual disclaimer applies. The views expressed in this paper are those of the authors and do not necessarily reflect the views of the Office of the Comptroller of the Currency or the U.S. Department of the Treasury. Send correspondence to Zhi Da, Mendoza College of Business, University of Notre Dame, Notre Dame, IN 46556; telephone: (574) 631-0354. E-mail: zda@nd.edu.

? The Author 2013. Published by Oxford University Press on behalf of The Society for Financial Studies. All rights reserved. For Permissions, please e-mail: journals.permissions@. doi:10.1093/rfs/hht005

Downloaded from at University of Notre Dame on February 24, 2013

The Review of Financial Studies / v 0 n 0 2013

0.06 0.04 0.02

0 -0.02 -0.04

1986

1989

1992

1995

1998

2001

2004

2007

1 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0 2010

Figure 1 Analyst forecast revisions for five-year ahead earnings The figure plots the change of aggregate earnings forecasts scaled by last year's aggregate book equity. The data are aggregated from the firm level. The earnings forecast data are from IBES; the book equity data are from Computstat. The shaded bars represent NBER recessions. The data cover 1986?2010.

of future cash flows or discount rates. Since neither expected cash flows nor discount rates are observable, the traditional approach is to predict them, and calculate cash flow news and discount rate news as functions of the predictive variables. Because returns have been much easier to predict than dividends post-World War II, the prevailing view is that almost all aggregate stock return innovation is driven by discount rate news, and almost none by cash flow news.1

Several studies (e.g., Goyal and Welch 2008 and Chen and Zhao 2009) cast doubt on this prevailing view with the evidence that the traditional approach based on predictive regressions is sensitive to the choice of sample periods or predictive variables. A growing literature shows that, with different sample periods or cash flow measures, cash flow news can be more important than what is normally perceived (e.g., Ang and Bekaert 2007; Larrain and Yogo 2008; Chen 2009; Binsbergen and Koijen 2010; Chen, Da, and Priestley 2012).2

The relative importance of cash flow and discount rate news has an important bearing on the theoretical modeling of asset prices. For example, Campbell and Cochrane (1999) focus on time-varying discount rates with changing risk aversion, while the long-run risk literature highlights the role of cash flow risk (Bansal and Yaron 2004).

1 Cochrane (2011) summarizes the literature over the last 40 years by saying, "Previously, we thought returns were unpredictable, with variation in price-dividend ratios due to variation in expected cashflows. Now it seems all price-dividend variation corresponds to discount-rate variation (page 1047)."

2 Koijen and Van Nieuwerburgh (2011) in a survey of the recent literature on the predictability of returns and cash flows concluded that dividend growth is more predictable than commonly believed. Recently, Golez (2012) demonstrates that information from the derivatives market also helps to predict future dividend growth for the S&P 500 Index.

2

Downloaded from at University of Notre Dame on February 24, 2013

What Drives Stock Price Movements?

Given the importance of the subject and the sensitivity of the regressionbased approach, it is therefore useful to explore some alternative methods that do not rely on predictability. The alternative method we propose here uses direct expected cash flow measures. Given stock prices, we use the market prevailing forecasts for future cash flows (from IBES), for each firm and at each point of time, to back out the firm-specific implied cost of equity capital [see Pastor, Sinha, and Swaminathan (2008), among others]. Consequently, a price change can be decomposed into two pieces: (1) "CF news," defined as the price change holding the implied cost of capital (ICC) constant, and (2) "DR news," defined as the price change holding the cash flow forecasts constant. This decomposition holds by definition without resorting to predictability.

We emphasize that our CF and DR news are different from the traditional cash flow and discount rate news in the existing literature and our ICC is different from the more commonly studied one-period expected return. However, as demonstrated in Appendix A and consistent with the analysis in Pastor, Sinha, and Swaminathan (2008), ICC does convey information similar to the discount rate component of Campbell and Shiller's (1998) return decomposition. Consequently, our CF news and DR news also contain similar information as the cash flow news and discount rate news in the more traditional return decomposition framework.

A key assumption of the ICC approach is that analyst earnings forecasts are timely reflections of the marginal investors' belief regarding future cash flows. Any deviation from this assumption, such as stale or biased analyst forecasts could result in biased CF news estimates.3 We address these issues in a battery of empirical checks and find our conclusions from the ICC approach to be robust to biases in analyst forecasts and different assumptions governing the steady-state cash flows.

The ICC method provides new insights. During the sample period of 1985? 2010 when the earnings forecast data are available, we find that CF news contributes significantly to stock price variation. For example, at the one-year horizon, CF news accounts for 36% of the stock price variation at the aggregate level and 48% of the price variance at the firm level.

The extent of stock price variation explained by CF news increases with the investment horizon, and for horizons beyond two years, CF news outweighs DR news. At the aggregate level, the CF news portion is 53% at the two-year horizon, and 60% at the three-year horizon. At the firm level, the CF news portion is 63% at the two-year horizon, and 68% at the three-year horizon. CF news is more important at the firm level than at the aggregate level, suggesting that CF news is diversified away relatively more than DR news.

3 There is a literature documenting that stock prices respond to revisions of analyst forecasts. This literature includes, among others, Griffin (1976), Givoly and Lakonishok (1979), Elton, Gruber, and Gultekin (1981), Imhoff and Lobo (1984), Lys and Sohn (1990), Francis and Soffer (1997), and Park and Stice (2000).

3

Downloaded from at University of Notre Dame on February 24, 2013

The Review of Financial Studies / v 0 n 0 2013

This diversification effect is secondary, however, in the sense that, at longer horizons, CF news is only slightly more important at the firm level.

The finding that the relative importance of cash flow/discount rate news changes with time horizon is intuitive. As long as discount rates are stationary, negative discount rate news in the current period (because discount rate goes up) will be offset by higher returns in the future. Therefore, the impact of discount rate news is temporary and attenuated with time. In the long-run limit, all stock return news must be cash flow news (e.g., Campbell and Vuolteenaho 2004; Hansen, Heaton, and Li 2008; Bansal, Dittmar, and Kiku 2009). This is a fundamental property that holds irrespective of economic models.

The finding that there is only a limited relative cash flow/discount rate diversification effect when moving from individual firms to the aggregate portfolio provides a stark contrast to the prevailing view (Vuolteenaho 2002) that, because of diversification, cash flow news dominates at the firm level but discount rate news dominates at the aggregate level.

We argue, however, that the cash flow diversification effect is likely overstated because the panel regressions in Vuolteenaho (2002) do not control for the firm-fixed effects. In panel regressions, there is a critical difference between cross-sectional and time-series predictability, an issue that has been largely overlooked in the current literature. The cross-sectional heterogeneity of cash flows is persistent and predictable (e.g., Lakonishok, Shleifer, and Vishny 1994; Fama and French 1995); it is thus easy to find that cash flow news dominates whenever panel data (without firm fixed-effects controls) are studied. In the time-series dimension, however, cash flows are less predictable than discount rates, and discount rate news is usually found to be more important in the pure time-series regressions that are common for aggregate portfolio analysis. The prevailing conclusion is thus the result of mixing the strong cross-sectional cash flow predictability with the weak time-series cash flow predictability. Such a conclusion is unreliable, as it compares apples with oranges. In an apples-to-apples comparison, we run time-series predictive regressions firm-by-firm. For the period of 1985?2010, cash flow news explains 48% of stock return at the firm level over the one-year horizon, a result very comparable to that using the ICC method.

To summarize, while the issue of what drives stock price movement is crucial for asset pricing because it reveals how investors evaluate securities, long overdue is a comprehensive investigation of the relative importance of cash flow versus discount rate news, at both the aggregate and firm levels, using different methods, and at different horizons.

Our paper attempts to fill in the void. Our main message is that, contrary to prevailing views, cash flow news is important in driving stock price movement at both the firm and aggregate levels. The previous conclusion that there is little cash flow news, albeit disconcerting, has provided an important empirical basis for theoretic modeling [e.g., Campbell and Cochrane (1999) versus Bansal and Yaron (2004)]. Our finding that there is significant cash flow news at reasonable

4

Downloaded from at University of Notre Dame on February 24, 2013

What Drives Stock Price Movements?

horizons suggests that cash flow news deserves a greater role in theoretical considerations.

We believe we are the first to use the ICC approach to study return decomposition. Our contribution is related to but distinct from the literature that uses the ICC approach to study asset valuation and risk-return tradeoff, including, among others, Kaplan and Ruback (1995), Liu and Thomas (2000), Claus and Thomas (2001), Gebhardt, Lee, and Swaminathan (2001), Jagannathan and Silva (2002), Easton and Monahan (2005), Pastor, Sinha, and Swaminathan (2008), and Da and Warachka (2009). Our approach is in the spirit of Graham and Harvey (2005), who use surveys of CFOs to measure the expected equity premium. Our results suggest that such an approach can shed fresh light on several fundamental issues in asset valuation.

Our findings complement the literature that studies relative return/cash flow predictability by the dividend yield (e.g., Campbell and Shiller 1988, 1998; Cochrane 1992, 2001, 2008; Goyal and Welch 2003; Lettau and Ludvigson 2005; Ang and Bekaert 2007; Larrain and Yogo 2008; Lettau and Van Nieuwerburgh 2008; Chen 2009; Binsbergen and Koijen 2010; Koijen and Van Nieuwerburgh 2011). This literature provides important evidence on predictability and on the information content of the dividend yield; we study price volatility without resorting to predictability.

The rest of the paper proceeds as follows. In Section 1, we describe the method we use to construct the ICC, the CF news, and the DR news. In Sections 2 and 3, we report the evidence at the aggregate and firm levels using the ICC method and the more traditional predictive regression, respectively. A brief conclusion is provided in Section 4. Appendix A compares our ICC-based decomposition to the more standard Campbell and Shiller (1988) loglinear return decomposition; examines the discount rate news over the long run; and studies the impact of a persistent cash flow forecast error. Appendix B provides details on computing the ICC.

1. The Implied Cost of Equity Capital Model

We back out the ICC for each firm quarter following Pastor, Sinha, and Swaminathan (2008). Appendix B provides the details on the sample selection and the calculation of ICC.

The equity value is the present value of future dividends and a terminal value:

Pt

=

T k=1

FEt+k (1-bt (1+ qt )k

+k

)

+

FEt+T +1 qt (1+ qt )T

(1)

= f ct ,qt ,

(2)

where Pt is the stock price, FEt+k is the earnings forecast k years ahead, bt+k is the plowback rate (i.e., 1-bt+k is the payout ratio), and qt is the ICC. T is set to 15 years. In short, the stock price Pt is a function of the vector of cash

5

Downloaded from at University of Notre Dame on February 24, 2013

The Review of Financial Studies / v 0 n 0 2013

flow forecasts ct (available at time t) and the discount rate qt . As detailed in Appendix B, FEt+k are calculated from at least three analyst forecasts from IBES: the earnings forecasts for the current fiscal year (FEt+1), the next fiscal year (FEt+2), and the long-term growth forecast (gt+3). While this calculation also depends on the assumptions of steady-state earnings growth rates and plowback rates, our decomposition results should not be materially affected by these assumptions for two reasons. First, since our decomposition focuses on price changes over time, as long as the assumed steady-state growth rates and plowback rates are slow-moving, they should not have a large impact on our decomposition results. Second, long-term cash flows are more heavily discounted relative to those in the nearer term (determined directly by IBES forecasts) in Equation (1).

1.1 CF and DR news The proportional price difference or capital gain return (Retx) between t +j and t is then (subscript changed from t to t +j ):

Retxj

=

Pt

+j - Pt

Pt

f =

ct+j ,qt+j

-f (ct ,qt )

Pt

= CFj + DRj ,

(3)

where:

CFj =

f (ct+j ,qt+j )-f (ct ,qt+j ) + f (ct+j ,qt )-f (ct ,qt )

Pt

Pt

/2

(4)

DRj =

f (ct ,qt+j )-f (ct ,qt ) + f (ct+j ,qt+j )-f (ct+j ,qt )

Pt

Pt

/2.

(5)

Take CFj as an example. It is labeled as CF news because the numerator is calculated by holding the discount rate constant, and CFj captures the price change driven primarily by the changing CF expectations from t to t +j . Note that for both CFj and DRj , the differences are computed at t and t +j . This is a balanced approach as CF and DR changes are defined symmetrically, neither given a higher weight in the definition.

It is important to note that our decomposition in (3) is different from the more standard loglinear return decomposition in Campbell and Shiller (1988) in at least three respects. First, we work with capital gain returns, which do not include dividends. However, exclusion of dividends should not have a material impact on our conclusions as dividends play a minor role in the total return volatility. For example, during 1926?2010, the average quarterly total return for the CRSP value-weighted portfolio is 2.91% (std. dev. = 11.32%); the average quarterly return excluding dividends is 1.94% (std. dev. = 11.25%).

6

Downloaded from at University of Notre Dame on February 24, 2013

What Drives Stock Price Movements?

During 1985?2010, the average total return is 2.96% (std. dev. = 8.86%); the average return excluding dividends is 2.36% (std. dev. = 8.80%). Therefore, dividend payout affects only the level of returns; its impact on return volatility is negligible. Second, while Campbell and Shiller (1988) interpret return variations through a log-linearization approximation of the present value formula, our approach does not linearize the formula, and nonlinearity is implicit in the ICC method.

Third, while the cash flow and discount rate news in Campbell and Shiller (1988) sum up to the unexpected return (realized return minus expected return), our CF and DR news add up to the realized price change. However, the variation in expected returns is typically small relative to the variation in realized returns for stocks.

To understand the differences more directly, Appendix A compares these two decomposition methods empirically using the annual stock market data for the S&P 500 Index during the sample period of 1871?2009. We find that unexpected returns and realized price changes are highly correlated. Further, the cash flow (or discount rate) news components computed using the two decompositions are also highly correlated. The high correlations suggest that, as long as similar forecasts for future cash flows are used, our ICC-based decomposition and the more standard Campbell and Shiller (1988) decomposition, though implemented differently, produce very similar inferences.

We can then study the variance of the capital gain return through CF news and DR news:

VAR(Retxt ) = COV(CFt ,Retxt )+COV(DRt ,Retxt )

(6)

1 = COV(CFt ,Retxt ) + COV(DRt ,Retxt ) ,

(7)

VAR(Retxt )

VAR(Retxt )

where

VAR

and

COV

are

variance

and

covariance

operators.

COV(CFt ,Retxt ) VAR(Retxt )

is

the

slope

coefficient

of

regressing

CFt

on

Retxt

;

COV(DRt ,Retxt VAR(Retxt )

)

is

the

slope

coefficient

of regressing DRt on Retxt . In other words, to understand the portion of capital

gain return variance that is driven by CF news and DR news, one only needs to

regress CF and DR news on the capital gain returns, and draw inferences based

on the slope coefficients.

1.2 Expected return, discount rate component, and implied cost of equity capital

The discount rate in our model is the ICC. Hughes, Liu, and Liu (2009) show that ICC, the single discount rate that applies to all horizons, might deviate from the expected next-period return. This is not a concern for us because our goal is not to estimate the expected return for the next period but to capture price variations due to the changes in expected returns for all future horizons, or changes in the discount rate component of the Campbell and Shiller (1988)

7

Downloaded from at University of Notre Dame on February 24, 2013

The Review of Financial Studies / v 0 n 0 2013

decomposition (DRCS):

DRCt S =

j

-1Et

(rt +j

)

k 1-

-

(pt

-

dt

)

+

j-1Et (

dt+j ),

(8)

j =1

j =1

where and are the loglinear constants; p, d, and d are the log price, log dividend, and log dividend growth, respectively.

Similarly, the ICC solves:

Pt

=

j =1

Et (Dt+j ) (1+ ICCt )j

(9)

where Dt+j is the dividend in time t +j .

Following Jagannathan, McGrattan, and Scherbina (2000), a linearized

version of Equation (9) yields:

ICCt

y-g 1+g

2+g -(y

-g)

Pt Dt

+

j =1

j

Et

Dt+j

(10)

Dt +j

=

Dt +j Dt+j -1

-

1

1+ g j-1

j = 1+y

,

where g is the mean expected dividend growth rates, and y is the mean ICC. Appendix A provides detailed derivations of Equation (10) and a direct comparison between DRCS and ICC.

Comparing the ICC in (10) to the DRCS in (8), we find them to contain similar information. Subject to the difference in linear and loglinear approximations, both are linear combinations of the current price-to-dividend ratio and the weighted average of future expected dividend growth rates. Appendix A, using the S&P annual stock market data of 1871?2009, shows a correlation exceeding 0.98. For this reason, Pastor, Sinha, and Swaminathan (2008) in their Equation (3) go so far as to define ICC as a scaled version of DRCS. As a result, return decompositions using ICC, like those in the Campbell and Shiller (1988), should also shed light on the fundamental question of whether cash flow or discount rate news is the main drivers of stock price movement.

1.3 ICC vs. predictive regression Return innovations can be decomposed into cash flow news and discount rate news. Since usually neither expected cash flows nor discount rates are observable, the common practice in the current literature is to predict cash flows and returns, and cash flow and discount rate news are then computed as functions of the predictive variables.

8

 ................
................

In order to avoid copyright disputes, this page is only a partial summary.

Google Online Preview   Download