What Moves the Stock and Bond Markets? A Variance Decomposition for Long-Term Asset Returns John Y, Campbell; John Ammer The Journal of Finance, Volume 48, Issue 1 (Mar., 1993), 3-37. Stable URL: bhtp:flinks,jstor-org/sici?sici~0022-1082%28199303%2948%3A %3C3%3AWMISAB3E2.0.CO%3B2-T ‘Your use of the ISTOR archive indicates your acceptance of ISTOR’s Terms and Conditions of Use, available at up: srw jstor org/aboutterms.html. ISTOR's Terms and Conditions of Use provides, in part, at unless you have obtained prior permission, you may aot download an entite issue of a journal or multiple copies of articles, and ‘you may use content in the ISTOR archive only for your personal, non-commercial use Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the sereen or pfinted page of such transmission. ‘The Journal of Finance 1s published by American Finance Association. Please contact the publisher for forther permissions regarding the use ofthis work. Publisher contact information may be obaines at Lupo tor org/journalsfaina hel. The Journal of Finance ©1993 American Finance Association ISTOR and the JSTOR logo are trademarks of JSTOR, and are Registered in the U.S, Patent and Trademark Otic. For more information on ISTOR contact jstor-nfo@umnich edu, (©2003 JSTOR hup:thevwwjstor.orgy Tue Aug 26 02:10:11 2003{18 SOUMIYAL.OF FINANCE # VOL RLV, NO. + ABH 1980 What Moves the Stock and Bond Markets? A Variance Decomposition for Long-Term Asset Returns JOHN Y. CAMPBELL and JOHN AMMER’ ABSTRACT ‘This paper uses a vector autoregressive model to decompane excess stack and 10-year bond returns into changes in expectations of fare stack divdenc, infla- tion, short tarm real interest ratee, and excess stock and bond retsnus. ti maith postwar U.S. data, tock and bond vetuens are driven langely by news about fate excess stock retuons and inflation, respectively Neal interest rates have litle Impact on returns, although they do affect the akart-erm macnn interest rate ard the slope ofthe term structure. Theae findings help to explain the low corvelation between exeess stade and hand returns MAINSTREAM RESEARCH LN EMPIRICAL asset priciig has traditionally treated the variances and covariances of asset returns as being exogenous, The questions asked concern the optimal response of utility-maximizing agents to these variances and covariances, and the resulting equilibrium pattern of mean returns on securities. Recently, however, a number of authors have challenged the finance profes- sion to bring the second moments of asset returns within the set. of phenom- ana to be explained. One of the first researchers to pose this challenge was Robert Shiller, who argued in the early 1980s that it ia hard to account for the variance of stock retutns using a model with constant discount rates.‘ Richard Rall has issued a similar challenge in a recent Presidential Address to the American Finance Association (1988), saying that “The immaturity of our science is illustrated by the conspicuous Inek of predictive content about some of its most intensely interesting phenomena, particularly changes in asset prices.” One straightforward way to meet this challenge is to regress asset price ‘changes on contemporaneous news events. Roll (1988) does this for individual, stocks and finds that less than 40% of the variance of price changes is * Woodrow Wilson Scheol, Princeton University and International Finance Divisten, Board of Governors of the Feders| Reserve System, This psper was pretentes at the annus] meeting of the American Fleance Asvoristion, Washingtan DC, Decerier 1990, We are grateful to Ben Bernanke, Robert Stambaugh, and tom anonymous referees for helpful comments, and ta Rick Mishkin for assistance with the dats Campbell acknowledges financial support from the National Science Foundation and the Sloan Foundation. Tha apimions exprected in thia paper are ‘those of the authors and do not necessarily reflect the views af the Board of Govemors or ather ‘exployess of the Federal Reserve System: “This research is reprinted and susimarized in Shiller 990) a4 ‘The Journal of Finance typically explained by the regressions. Eugene Fama (1990a) has applied a similar methodology to the aggregate stock market. He finds that almost two-thirds of the variance of aggregate stock price movements cant be ac- counted for by variables proxying for corporate cash flows and investors’ discount rates. Fama uses leads of some variables as well as contemporane- ‘ous values, 2s an informal way to allow for extra information that market participants may have about fature macroeconomic developments. Other recent papers using this approach include Collins, Kothari, Shanken, and Sloan (1992), Cutler, Poterba, and Summers (1989), Kothari and Shanken (1992), and Stambaugh (1990). ‘The use of contemporaneous regressions to explain asset price variability is appealing because it is simple, and because it is an extension of the well- established event study methodology in finance. However, this approach has litte to say about the channels through which macroeconomic news variables affect asset prices. Suppose that innovations to a particular variable, say industrial production, are associated with stock market movements. This could reflect an association of industrial production with changing expecta- tions of future cash flows, or an association with changing discount. rates (perhaps because both industrial praduetion and stock prices are responding to interest rate movements). The contemporaneous regression approach can- not distinguish these possibilities, or tell us about. their relative importance? In this paper we use an alternative approach developed in Campbell and Shiller (19882, 1988b) and Campbell (1991). Those papers study the stock market in isolation, whereas here we try to account for the variance of stack returns jointly with the variance of long-term nominal bond returns and the covarianea between stock and bond returns, We first express the innovation to a long-term asset return as the sum of revisions in expectations of future real cash payments ta investors, and revisions in expectations of future real returna on the asset. (Expected future returns are further broken down into expected future real interest rates and expected future excess returns on the Tong-term asset.) In general this asset-pricing framework holds as a logelinear approximation (Campbell and Shiller (1988a)}. However, we provide an exact log-linear relation for the ease in, which the long-term asset is a zero-coupon bond. We combine the asset-pricing framework with a vector autoregression (VAR) in fong-term asset returns, interest rates, inflation, and other informa- tion that helps to forecast these variables. We assume that the VAR ade- quately captures the information used by investors. From the VAR we ean calculate revisions in multiperiod forecasts of real returns and eash flows, and thus we can break asset returns into several components. In the case of stocks, the components are changing expectations of future real dividends, fature real interest rates, and future excess returns on stocks, In the case of long-term nominal bonds, the components are changing expectations of future 2 this statement dacs not apply tothe eross-sectional anslysis of Kothart and Shacken (19922, whieh ineludas a correction for discount rate taverents