W CHICAGO JOURNALS Economic forces and the stock market Author(s): Nai-Fu Chen, Richard Roll and Stephen A. Ross R evlewea wor k(s) Source: The Tournal of Business, Vol. 59, No. 3(JuL, 1986), pp. 383-403 Published by: The University of Chicago Press StableUrl:http://www.jstor.org/stable/2352710 Accessed:04/12/201203:43 Your use of the JSTOR archive indicates your acceptance of the Terms Conditions of Use, available at http://www.jstor.org/page/info/about/policies/termsjsp JStOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about JSTOR, please contact support(@jstor. org The University of Chicago Press is collaborating with JSTOR to digitize, preserve and extend access to The Journal of Business. 的d http://www.jstororg his content downloaded by the authorized user from 192. 168.82.218 on Tue, 4 Dec 2012 03: 43: 28 AM All use subject to JSTOR Terms and Conditions
Economic Forces and the Stock Market Author(s): Nai-Fu Chen, Richard Roll and Stephen A. Ross Reviewed work(s): Source: The Journal of Business, Vol. 59, No. 3 (Jul., 1986), pp. 383-403 Published by: The University of Chicago Press Stable URL: http://www.jstor.org/stable/2352710 . Accessed: 04/12/2012 03:43 Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at . http://www.jstor.org/page/info/about/policies/terms.jsp . JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about JSTOR, please contact support@jstor.org. . The University of Chicago Press is collaborating with JSTOR to digitize, preserve and extend access to The Journal of Business. http://www.jstor.org This content downloaded by the authorized user from 192.168.82.218 on Tue, 4 Dec 2012 03:43:28 AM All use subject to JSTOR Terms and Conditions
Nai-Fu Chen University of Chicage Richard roll Los angeles Stephen A. Ross Yale University Economic Forces and the stock market I. Introduction This paper tests Asset prices are commonly believed to react sen- acroeconomic varI- sitively to economic news. Daily experience ables are risks that are seems to support the view that individual asset ded in the stock variety of market. Financial unanticipated events and that some events have the followggests that a more pervasive effect on asset prices than do economic anab.. others. Consistent with the ability of investors to should systematically diversify, modern financial theory has focused ffect stock market on pervasive, or"systematic, influences as the returns likely source of investment risk. The general conclusion of theory is that an additional compo interest rates, expected nd unexpected infla nent of long- run return is required and obtained tion, industrial produc henever a particular asset is influenced by sys tion and the between high-and low- can be earned by(needlessly) bearing diversifi- ade bonds. We find hat these sources of ble risk k are priced. Furthermore The authors are grateful to their respective universities to the Center for Research in Security Prices, to the National neither the market Science Foundation for research support, and to Ceajer Chan rtfolio nor aggregate onsumption are priced Cornell, Eugene Fama, Pierre Hillion, Richard Sweeney, and eparately. We also Arthur Warga were most helpful, as were the comments of find that oil price risk articipants in workshops at Claremont Graduate School is not separately re- Stanford University, the University of Tor the univer. warded in the stock sity of Ca market rst revision was written = For example, the APT (Ross 1976) and the models of Merton(1973)and Cox, Ingersoll, and Ross(1985)are consis- tent with this view (Journal of Business, 1986, vol. 59, no. 3) c 1986 by The University of Chicago. All rights reserved 0021-93988659030001s01.50 83 ontent downloaded by the authorized user from 192.168. 82.218 on Tue, 4 Dee 2012 03: 43 28 AM
Nai-Fu Chen University of Chicago Richard Roll University of California, Los Angeles Stephen A. Ross Yale University Economic Forces and the Stock Market* I. Introduction Asset prices are commonly believed to react sensitively to economic news. Daily experience seems to support the view that individual asset prices are influenced by a wide variety of unanticipated events and that some events have a more pervasive effect on asset prices than do others. Consistent with the ability of investors to diversify, modern financial theory has focused on pervasive, or "systematic," influences as the likely source of investment risk.' The general conclusion of theory is that an additional component of long-run return is required and obtained whenever a particular asset is influenced by systematic economic news and that no extra reward can be earned by (needlessly) bearing diversifiable risk. This paper tests whether innovations in macroeconomic variables are risks that are rewarded in the stock market. Financial theory suggests that the following macroeconomic variables should systematically affect stock market returns: the spread between long and short interest rates, expected and unexpected inflation, industrial production, and the spread between high- and lowgrade bonds. We find that these sources of rsk are significantly priced. Furthermore, neither the market portfolio nor aggregate consumption are priced separately. We also find that oil price risk is not separately rewarded in the stock market. * The authors are grateful to their respective universities, to the Center for Research in Security Prices, to the National Science Foundation for research support, and to Ceajer Chan for computational assistance. The comments of Bradford Cornell, Eugene Fama, Pierre Hillion, Richard Sweeney, and Arthur Warga were most helpful, as were the comments of participants in workshops at Claremont Graduate School, Stanford University, the University of Toronto, the University of California, Irvine, the University of Alberta, the University of Chicago, and unknown referees. The University of British Columbia provided a stimulating research environment where part of the first revision was written during August 1984. 1. For example, the APT (Ross 1976) and the models of Merton (1973) and Cox, Ingersoll, and Ross (1985) are consistent with this view. (Journal of Business, 1986, vol. 59, no. 3) ? 1986 by The University of Chicago. All rights reserved. 0021-9398/8615903-0001$01.50 383 This content downloaded by the authorized user from 192.168.82.218 on Tue, 4 Dec 2012 03:43:28 AM All use subject to JSTOR Terms and Conditions
Journal of Business The theory has been silent, however, about which events are likely to influence all assets. a rather embarrassing gap exists between the of systen our complete ignorance of their identity. The comovements of asset prices suggest the presence of underlying exogenous influences, but we have not yet determined which economic variables, if any, are respon Our paper is an exploration of this identification terrain. In Section Il, we employ a simple theoretical guide to help choose likely candi- dates for pervasive state variables In Section Ill we introduce the data and explain the techniques used to measure unanticipated movements in the proposed state variables. Section IV investigates whether expo- sure to systematic state variables explains expected returns. As specific alternatives to the pricing influence of the state variables identified by our simple theoretical model, Section IV considers the value-and the equally weighted market indices, an index of real con sumption, and an index of oil prices. Each of these is found to be unimportant for pricing when compared with the identified economic state variables. Section V briefly summarizes our findings and suggests some directions for future research Il. Theory No satisfactory theory would argue that the relation between financial markets and the macroeconomy is entirely in one direction. However stock prices are usually considered as responding to external forces (even though they may have a feedback on the other variables) It is apparent that all economic variables are endogenous in some ultimate sense. Only natural forces, such as supernovas, earthquakes, and the ike, are truly exogenous to the world economy, but to base an asset- pricing model on these systematic physical factors is well beyond our current abilities. Our present goal is merely to model equity returns as functions of macro variables and nonequity asset returns. Hence this paper will take the stock market as endogenous, relative to other mar- By the diversification argument that is implicit in capital market theory, only general economic state variables will influence the pricing of large stock market aggregates. Any systematic variables that affect the economys pricing operator or that influence dividends would also nfluence stock market returns. Additionally, any variables that are necessary to complete the description of the state of nature will also be part of the description of the systematic risk factors. An example of such a variable would be one that has no direct influence on current cash fows but that does describe the changing investment opportunity ontent downloaded by the authorized user from 192.168. 82.218 on Tue, 4 Dee 2012 03: 43 28 AM
384 Journal of Business The theory has been silent, however, about which events are likely to influence all assets. A rather embarrassing gap exists between the theoretically exclusive importance of systematic "state variables" and our complete ignorance of their identity. The comovements of asset prices suggest the presence of underlying exogenous influences, but we have not yet determined which economic variables, if any, are responsible. Our paper is an exploration of this identification terrain. In Section II, we employ a simple theoretical guide to help choose likely candidates for pervasive state variables. In Section III we introduce the data and explain the techniques used to measure unanticipated movements in the proposed state variables. Section IV investigates whether exposure to systematic state variables explains expected returns. As specific alternatives to the pricing influence of the state variables identified by our simple theoretical model, Section IV considers the value- and the equally weighted market indices, an index of real consumption, and an index of oil prices. Each of these is found to be unimportant for pricing when compared with the identified economic state variables. Section V briefly summarizes our findings and suggests some directions for future research. II. Theory No satisfactory theory would argue that the relation between financial markets and the macroeconomy is entirely in one direction. However, stock prices are usually considered as responding to external forces (even though they may have a feedback on the other variables). It is apparent that all economic variables are endogenous in some ultimate sense. Only natural forces, such as supernovas, earthquakes, and the like, are truly exogenous to the world economy, but to base an assetpricing model on these systematic physical factors is well beyond our current abilities. Our present goal is merely to model equity returns as functions of macro variables and nonequity asset returns. Hence this paper will take the stock market as endogenous, relative to other markets. By the diversification argument that is implicit in capital market theory, only general economic state variables will influence the pricing of large stock market aggregates. Any systematic variables that affect the economy's pricing operator or that influence dividends would also influence stock market returns. Additionally, any variables that are necessary to complete the description of the state of nature will also be part of the description of the systematic risk factors. An example of such a variable would be one that has no direct influence on current cash flows but that does describe the changing investment opportunity set. This content downloaded by the authorized user from 192.168.82.218 on Tue, 4 Dec 2012 03:43:28 AM All use subject to JSTOR Terms and Conditions
Stock prices can be written as expected discounted dividends: E(c) (1) where c is the dividend stream and k is the discount rate. This implies that actual returns in any period are given by dp [(c) dk It follows(trivially) that the systematic forces that infuence returns are those that change discount factors, k, and expected cash flows, E(c) The discount rate is an average of rates over time, and it changes with both the level of rates and the term-structure spreads across dif- ferent maturities. Unanticipated changes in the riskless interest rate will therefore influence pricing, and, through their infuence on the time value of future cash flows, they will infuence returns. The discount rate also depends on the risk premium; hence, unanticipated changes in the premium will influence returns. On the demand side, changes in the indirect marginal utility of real wealth, perhaps as measured by real consumption changes, will influence pricing, and such effects should also show up as unanticipated changes in risk premia Expected cash flows change because of both real and nominal forces. Changes in the expected rate of inflation would influence nomi nal expected cash flows as well as the nominal rate of interest. To the xtent that pricing is done in real terms, unanticipated price-level changes will have a systematic effect, and to the extent that relative prices change along with general inflation, there can also be a change in asset valuation associated with changes in the average inflation rate Finally, changes in the expected level of real production would affect the current real value of cash flows. Insofar as the risk-premium mea- sure does not capture industrial production uncertainty, innovations in the rate of productive activity should have an infiuence on stock re turns through their impact on cash flows Ill. Constructing the Economic Factors Having proposed a set of relevant variables, we must now specify thei measurement and obtain time series of unanticipated movements. We ould proceed by identifying and estimating a vector autoregressive model in an attempt to use its residuals as the unanticipated innova- 2. Since we are only concerned with intuition, we are ignoring the second-order terms from the stochastic calculus in deriving eq. (2). Also notice that the expectation is taken with respect to the martingale pricing measure(see Cox et al. 1985)and not with respect to the ordinary probability distribution ontent downloaded by the authorized user from 192.168. 82.218 on Tue, 4 Dee 2012 03: 43 28 AM
Economic Forces and the Stock Market 385 Stock prices can be written as expected discounted dividends: - E(c) where c is the dividend stream and k is the discount rate. This implies that actual returns in any period are given by dp + c d[E(c)] A + c (2) p p E(c) k P It follows (trivially) that the systematic forces that influence returns are those that change discount factors, k, and expected cash flows, E(c).2 The discount rate is an average of rates over time, and it changes with both the level of rates and the term-structure spreads across different maturities. Unanticipated changes in the riskless interest rate will therefore influence pricing, and, through their influence on the time value of future cash flows, they will influence returns. The discount rate also depends on the risk premium; hence, unanticipated changes in the premium will influence returns. On the demand side, changes in the indirect marginal utility of real wealth, perhaps as measured by real consumption changes, will influence pricing, and such effects should also show up as unanticipated changes in risk premia. Expected cash flows change because of both real and nominal forces. Changes in the expected rate of inflation would influence nominal expected cash flows as well as the nominal rate of interest. To the extent that pricing is done in real terms, unanticipated price-level changes will have a systematic effect, and to the extent that relative prices change along with general inflation, there can also be a change in asset valuation associated with changes in the average inflation rate. Finally, changes in the expected level of real production would affect the current real value of cash flows. Insofar as the risk-premium measure does not capture industrial production uncertainty, innovations in the rate of productive activity should have an influence on stock returns through their impact on cash flows. III. Constructing the Economic Factors Having proposed a set of relevant variables, we must now specify their measurement and obtain time series of unanticipated movements. We could proceed by identifying and estimating a vector autoregressive model in an attempt to use its residuals as the unanticipated innova- 2. Since we are only concerned with intuition, we are ignoring the second-order terms from the stochastic calculus in deriving eq. (2). Also notice that the expectation is taken with respect to the martingale pricing measure (see Cox et al. 1985) and not with respect to the ordinary probability distribution. This content downloaded by the authorized user from 192.168.82.218 on Tue, 4 Dec 2012 03:43:28 AM All use subject to JSTOR Terms and Conditions
Journal of Business tions in the economic factors. It is, however, more interesting and (perhaps) robust out of sample to employ theory to find single equa tions that can be estimated directly. In particular, since monthly rates of return are nearly serially uncorrelated they can be employed as innovations without alteration. The general impact of a failure ade quately to filter out the expected movement in an independent variable is to introduce an errors-in-variables problem. This has to be traded off against the error introduced by misspecification of the estimated equa- tion for determining the expected movement A somewhat subtler version of the same problem arises with proce- dures such as vector autoregression Any such statistically based time series approach will find lagged stock market returns having a signifi- cant predictive content for macroeconomic variables. In the analysis of pricing, then, we will indirectly be using lagged stock market variables to explain the expected returns on portfolios of stocks. Whatever econometric advantages such an approach might offer, it is antithetical to the spirit of this investigation, which is to explore the pricing in- fluence of exogenous macroeconomic variables. For this reason much as for any other, we have chosen to follow the simpler route in constructing the time series we use Throughout this paper we adopt the convention that time subscripts apply to the end of the time period. The standard period is 1 month Thus, E( t-1)denotes the expectation operator at the end of month t-1 conditional on the information set available at the end of month t 1, and X(o denotes the value of variable X in month t, or the growth that prevailed from the end of t- l to the end of I. A. Industrial Production The basic series is the growth rate in U.S. industrial production. It was obtained from the Survey of Current Business. If IP(o) denotes the rate of industrial production in month t, then the monthly growth rate is MP(t)= loge IP(r)-loge IP(t-1) and the yearly growth rate is YP(1)= loge IP(t)-loge IP(I-12) (see table 1 for a summary of variables Because IP(t)actually is the flow of industrial production during month t, MP(t)measures the change in industrial production lagged by at least a partial month. To make this variable contemporaneous with other series, subsequent statistical work will lead it by 1 month. Except for an annual seasonal, it is noisy enough to be treated as an in novation autocorrelations in their returns arising from the nontrading effect ontent downloaded by the authorized user from 192.168. 82.218 on Tue, 4 Dee 2012 03: 43 28 AM
386 Journal of Business tions in the economic factors. It is, however, more interesting and (perhaps) robust out of sample to employ theory to find single equations that can be estimated directly. In particular, since monthly rates of return are nearly serially uncorrelated, they can be employed as innovations without alteration. The general impact of a failure adequately to filter out the expected movement in an independent variable is to introduce an errors-in-variables problem. This has to be traded off against the error introduced by misspecification of the estimated equation for determining the expected movement. A somewhat subtler version of the same problem arises with procedures such as vector autoregression. Any such statistically based timeseries approach will find lagged stock market returns having a significant predictive content for macroeconomic variables. In the analysis of pricing, then, we will indirectly be using lagged stock market variables to explain the expected returns on portfolios of stocks. Whatever econometric advantages such an approach might offer, it is antithetical to the spirit of this investigation, which is to explore the pricing influence of exogenous macroeconomic variables. For this reason, as much as for any other, we have chosen to follow the simpler route in constructing the time series we use.3 Throughout this paper we adopt the convention that time subscripts apply to the end of the time period. The standard period is 1 month. Thus, E( It - 1) denotes the expectation operator at the end of month t - 1 conditional on the information set available at the end of month t - 1, and X(t) denotes the value of variable X in month t, or the growth that prevailed from the end of t - 1 to the end of t. A. Industrial Production The basic series is the growth rate in U.S. industrial production. It was obtained from the Survey of Current Business. If IP(t) denotes the rate of industrial production in month t, then the monthly growth rate is MPMt) = loge IPMt) - loge IP(t - 1), (3) and the yearly growth rate is YP(t) = loge IP(t) - loge IP(t - 12) (4) (see table 1 for a summary of variables). Because IP(t) actually is the flow of industrial production during month t, MP(t) measures the change in industrial production lagged by at least a partial month. To make this variable contemporaneous with other series, subsequent statistical work will lead it by 1 month. Except for an annual seasonal, it is noisy enough to be treated as an innovation. 3. In addition, the pricing tests reported below used portfolios that have induced autocorrelations in their returns arising from the nontrading effect. This content downloaded by the authorized user from 192.168.82.218 on Tue, 4 Dec 2012 03:43:28 AM All use subject to JSTOR Terms and Conditions
Economic forces and the TABLE 1 lossary and Definitions of Variables Symbo riable Definition or Source Inflation Log relative of U.S. Consumer Treasury-bill End-of-period return on I- month LGB (1958-78: Ibbotson and Baa Return EWNY Equally weighted equities Return on equally weighted port- VWNY Value-weighted equities Return on a value-weighted port- folio of NYSE-listed stocks (CRSP) Growth rate in re geton [1982]; Survey of Cur- rent Busines Oil prices og relative of Producer Price Index/Crude petroleum series Bureau of Labor Statistics Derived Series log [IP(rIP(I-1) YP(1) Annual growth, industrial pro- loge [P(r )IP(I- 12)1 E[() Expected infation Fama and Gibbons(1984 Unexpected inflation Real interest(ex post) DEI(t) Change in expected inflation E[(t+1)]-E[()t-l URP(n) Risk premium Baa(r)-LGB() UTS() Term structure The monthly series of yearly growth rates, YP(t), was examined because the equity market is related to changes in industrial activity in the long run. Since stock market prices involve the valuation of cash flows over long periods in the future, monthly stock returns may not be highly related to contemporaneous monthly changes in rates of indus- trial production, although such changes might capture the information pertinent for pricing. This month's change in stock prices probably reflects changes in industrial production anticipated many months into ontent downloaded by the authorized user from 192.168. 82.218 on Tue, 4 Dee 2012 03: 43 28 AM
Economic Forces and the Stock Market 387 TABLE 1 Glossary and Definitions of Variables Symbol Variable Definition or Source Basic Series I Inflation Log relative of U.S. Consumer Price Index TB Treasury-bill rate End-of-period return on 1-month bills LGB Long-term government bonds Return on long-term government bonds (1958-78: Ibbotson and Sinquefield [1982]; 1979-83: CRSP) IP Industrial production Industrial production during month (Survey of Current Business) Baa Low-grade bonds Return on bonds rated Baa and under (1953-77: Ibbotson [1979], constructed for 1978- 83) EWNY Equally weighted equities Return on equally weighted portfolio of NYSE-listed stocks (CRSP) VWNY Value-weighted equities Return on a value-weighted portfolio of NYSE-listed stocks (CRSP) CG Consumption Growth rate in real per capita consumption (Hansen and Singleton [1982]; Survey of Current Business) OG Oil prices Log relative of Producer Price Index/Crude Petroleum series (Bureau of Labor Statistics) Derived Series MP(t) Monthly growth, industrial loge[IP(t)/IP(t - 1)] production YP(t) Annual growth, industrial pro- loge[IP(t)/IP(t - 12)] duction E[I(t)] Expected inflation Fama and Gibbons (1984) UI(t) Unexpected inflation I(t) - E[I(t)lt - 1] RHO(t) Real interest (ex post) TB(t - 1) - I(t) DEI(t) Change in expected inflation E[I(t + 1)It] - E[I(t)It - 1] URP(t) Risk premium Baa(t) - LGB(t) UTS(t) Term structure LGB(t) - TB(t - 1) The monthly series of yearly growth rates, YP(t), was examined because the equity market is related to changes in industrial activity in the long run. Since stock market prices involve the valuation of cash flows over long periods in the future, monthly stock returns may not be highly related to contemporaneous monthly changes in rates of industrial production, although such changes might capture the information pertinent for pricing. This month's change in stock prices probably reflects changes in industrial production anticipated many months into This content downloaded by the authorized user from 192.168.82.218 on Tue, 4 Dec 2012 03:43:28 AM All use subject to JSTOR Terms and Conditions
388 ournal of Business he future. Therefore, subsequent statistical work will lead this vari tble by I year, similar to the variable used in Fama(1981) Because of the overlap in the series, YP(r)is highly autocorrelated A procedure was developed for forecasting expected YP(n)and a series of unanticipated changes in YP(r), and changes in the expectation itself were examined for their influence on pricing. The resulting series of- fered no discernible advantage over the raw production series, and, as a consequence, they have been dropped from the analysis. 4 B. Inflo Unanticipated infiation is defined as UI(t)=I()-E[(t)t-1 where I(n)is the realized monthly first difference in the logarithm of the Consumer Price Index for period t. The series of expected inflation E[I(nt-1] for the period 1953-78, is obtained from Fama and Gib bons(1984). If RHO(n)denotes the ex post real rate of interest applica ble in period t and TB(t-1)denotes the Treasury-bill rate known at he end of period t- 1 and applying to period t, then Fisher's equation asserts that TB(t-1)=E[RHO()t-1]+E[()t-1l Hence, TB(t-1)-I(r)measures the ex post real return on Treasury bills in the period From a time-series analysis of this variable, Fama and Gibbons(1984)constructed a time series for E[RHO(nr-1].Our expected inflation variable is defined by subtracting their time series for the expected real rate from the TB(t-1)series Another inflation variable that is unanticipated and that might have an influence separable from UI is DEI(1)=E[I(t+1)-EI()t-1]l the change in expected inflation. We subscript this variable with t since is (in principle) unknown at the end of month t-1. while, strictly speaking, DEl(t) need not have mean zero under the additional as sumption that expected inflation follows a martingale this variable may be treated as an innovation, and it may contain information not present in the ui variable. This would occur whenever inflation forecasts are influenced by economic factors other than past forecasting errors (Notice that the Ui series and the dei series will contain the inform tion in a series of innovations in the nominal interest rate, TB. )5 4. Results that include these series are available in an earlier draft of the paper, which is available from the authors on request pated inf tively correlated with the unanticipated change in the This follows from th bservation that the Fisher equation(6)holds for reali s as well as for expecta- tions. The UI(n) series also has a simple correlation of unanticipated inflation Fama(1981) ontent downloaded by the authorized user from 192.168. 82.218 on Tue, 4 Dee 2012 03: 43 28 AM
388 Journal of Business the future. Therefore, subsequent statistical work will lead this variable by 1 year, similar to the variable used in Fama (1981). Because of the overlap in the series, YP(t) is highly autocorrelated. A procedure was developed for forecasting expected YP(t) and a series of unanticipated changes in YP(t), and changes in the expectation itself were examined for their influence on pricing. The resulting series offered no discernible advantage over the raw production series, and, as a consequence, they have been dropped from the analysis.4 B. Inflation Unanticipated inflation is defined as {JI(t) = I(t) -E[I(t)lt - 1], (5) where I(t) is the realized monthly first difference in the logarithm of the Consumer Price Index for period t. The series of expected inflation, E[I(t)lt - 1] for the period 1953-78, is obtained from Fama and Gibbons (1984). If RHO(t) denotes the ex post real rate of interest applicable in period t and TB(t - 1) denotes the Treasury-bill rate known at the end of period t - 1 and applying to period t, then Fisher's equation asserts that TB(t - 1) = E[RHO(t)lt - 1] + E[I(t)lt - 1]. (6) Hence, TB(t - 1) - I(t) measures the ex post real return on Treasury bills in the period. From a time-series analysis of this variable, Fama and Gibbons (1984) constructed a time series for E[RHO(t)lt - 1]. Our expected inflation variable is defined by subtracting their time series for the expected real rate from the TB(t - 1) series. Another inflation variable that is unanticipated and that might have an influence separable from UI is DEI(t) = E[I(t + 1)It] - E[I(t)lt - 1], (7) the change in expected inflation. We subscript this variable with t since it is (in principle) unknown at the end of month t - 1. While, strictly speaking, DEI(t) need not have mean zero, under the additional assumption that expected inflation follows a martingale this variable may be treated as an innovation, and it may contain information not present in the U1 variable. This would occur whenever inflation forecasts are influenced by economic factors other than past forecasting errors. (Notice that the UI series and the DEI series will contain the information in a series of innovations in the nominal interest rate, TB.)5 4. Results that include these series are available in an earlier draft of the paper, which is available from the authors on request. 5. As an aside, the resulting unanticipated inflation variable, UI(t), is perfectly negatively correlated with the unanticipated change in the real rate. This follows from the observation that the Fisher equation (6) holds for realized rates as well as for expectations. The UI(t) series also has a simple correlation of .98 with the unanticipated inflation series in Fama (1981). This content downloaded by the authorized user from 192.168.82.218 on Tue, 4 Dec 2012 03:43:28 AM All use subject to JSTOR Terms and Conditions
Economic Forces and the Stock Markel c. Risk premia To capture the effect on returns of unanticipated changes in risk pre- mia, we will employ another variable drawn from the money markets The variable, UPR. is defined as UPR(t)="Baa and under"bond portfolio return(t)-lGB(r),(8) where LGB(r)is the return on a portfolio of long-term government bonds obtained from Ibbotson and Sinquefield(1982)for the perio 1953-78. From 1979 through 1983, lGB(t) was obtained from the Cen- ter for Research in Securities Prices(CRSP)data file. Again, UPR is not formally an innovation, but, as the differences in two return series it is sufficiently uncorrelated that we can treat it as unanticipated, and we will use it as a member of the set of economic factors The low-grade bond return series is for nonconvertible corporate bonds, and it was obtained from R. G. Ibbotson and Company for the period prior to 1977. a detailed description of the sample is contained in Ibbotson( 1979). The low-grade series was extended through 1983 by choosing 10 bonds whose ratings on January 1966 were below Baa. By 1978 these bonds still were rated below Baa, but their maturity was than that of the long-term government bond series. These 10 were then combined with three that were left over from the Ibbotson series at the end of 1978 to create a low-grade bond portfolio of 13 bonds in all. The returns on this portfolio were then used to xtend the upr series beyond 1977 and through 1983. two further difficulties with the series are that the ratings have experienced consid erable inflation since the mid-1950s and that the low-grade series con- tains bonds that are unrated The upr variable would have mean zero in a risk-neutral world, and it is natural to think of it as a direct measure of the degree of risk aversion implicit in pricing(at least insofar as the rating agencies main tain constant standards for their classifications). We hoped that U would reflect much of the unanticipated movement in the degree of risk aversion and in the level of risk implicit in the market,'s pricing of D. The Term structure To capture the infuence of the shape of the term structure, we employ another interest rate variable UTS(t)= LGB()- TB(t-1). 6. It could be argued that UPR captures a leverage effect, with highly levered firms with lower ratings. Furthermore, UPR is also similar to a measure of ity returns since a substantial portion of the value of low-grade bonds comes from the me sort of call option(behind secured debt) as for ordinary stock ontent downloaded by the authorized user from 192.168. 82.218 on Tue, 4 Dee 2012 03: 43 28 AM
Economic Forces and the Stock Market 389 C. Risk Premia To capture the effect on returns of unanticipated changes in risk premia, we will employ another variable drawn from the money markets. The variable, UPR, is defined as UPR(t) = "Baa and under" bond portfolio return (t) - LGB(t), (8) where LGB(t) is the return on a portfolio of long-term government bonds obtained from Ibbotson and Sinquefield (1982) for the period 1953-78. From 1979 through 1983, LGB(t) was obtained from the Center for Research in Securities Prices (CRSP) data file. Again, UPR is not formally an innovation, but, as the differences in two return series, it is sufficiently uncorrelated that we can treat it as unanticipated, and we will use it as a member of the set of economic factors. The low-grade bond return series is for nonconvertible corporate bonds, and it was obtained from R. G. Ibbotson and Company for the period prior to 1977. A detailed description of the sample is contained in Ibbotson (1979). The low-grade series was extended through 1983 by choosing 10 bonds whose ratings on January 1966 were below Baa. By 1978 these bonds still were rated below Baa, but their maturity was shorter than that of the long-term government bond series. These 10 bonds were then combined with three that were left over from the Ibbotson series at the end of 1978 to create a low-grade bond portfolio of 13 bonds in all. The returns on this portfolio were then used to extend the UPR series beyond 1977 and through 1983. Two further difficulties with the series are that the ratings have experienced considerable inflation since the mid-1950s and that the low-grade series contains bonds that are unrated. The UPR variable would have mean zero in a risk-neutral world, and it is natural to think of it as a direct measure of the degree of risk aversion implicit in pricing (at least insofar as the rating agencies maintain constant standards for their classifications). We hoped that UPR would reflect much of the unanticipated movement in the degree of risk aversion and in the level of risk implicit in the market's pricing of stocks.6 D. The Term Structure To capture the influence of the shape of the term structure, we employ another interest rate variable, UTS(t) = LGB(t) - TB(t - 1). (9) 6. It could be argued that UPR captures a leverage effect, with highly levered firms being associated with lower ratings. Furthermore, UPR is also similar to a measure of equity returns since a substantial portion of the value of low-grade bonds comes from the same sort of call option (behind secured debt) as for ordinary stock. This content downloaded by the authorized user from 192.168.82.218 on Tue, 4 Dec 2012 03:43:28 AM All use subject to JSTOR Terms and Conditions
f busines Again, under the appropriate form of risk neutrality and this variable can be thought of as measuring the unanticipated return on long bonds. The assumption of risk neutrality is used only to isolate the pure term-structure effects; the variable UPR is used to capture the effect of changes in risk aversion E. Market Indice The major thrust of our effort is to examine the relation between non- quity economic variables and stock returns. However because of the smoothing and averaging characteristics of most macroeconomic time series, in short holding periods, such as a single month these series cannot be expected to capture all the information available to the mar- ket in the same period. Stock prices, on the other hand, respond very quickly to public information. The effect of this is to guarantee that market returns will be, at best, weakly related and very noisy relative to innovations in macroeconomic fact This should bias our results in favor of finding a stronger linkage between the time-series returns on market indices and other portfolios of stock returns than between these portfolio returns and innovations in the macro variables. To examine the relative pricing influence of the traditional market indices we used the following variables EWNY()= return on the equally weighted NYSE index VWNY(r)= return on the value-weighted NYSE index These variables should reflect both the real information in the indus- rial production series and the nominal influence of the inflation vari- F. Consumption In addition to the macro variables discussed above we also examined a time series of percentage changes in real consumption, CG. The series is in real per capita terms and includes service flows. It was con tructed by dividing the CitiBaSE series of seasonally adjusted real population estimates. The CG series extends from January 1959 to December 1983, and it is an extension of a series obtained from lars Hansen for the period through 1979. a detailed description of its con- struction can be found in Hansen and Singleton (1983) G. Oil Prices It is often argued that oil prices must be included in any list of the systematic factors that influence stock market returns and pricing. To test this proposition and to examine another alternative to the macro variables discussed above we formed the og series of realized ontent downloaded by the authorized user from 192.168. 82.218 on Tue, 4 Dee 2012 03: 43 28 AM
390 Journal of Business Again, under the appropriate form of risk neutrality, E[UTS(t)lt - 1] = 0, (10) and this variable can be thought of as measuring the unanticipated return on long bonds. The assumption of risk neutrality is used only to isolate the pure term-structure effects; the variable UPR is used to capture the effect of changes in risk aversion. E. Market Indices The major thrust of our effort is to examine the relation between nonequity economic variables and stock returns. However, because of the smoothing and averaging characteristics of most macroeconomic time series, in short holding periods, such as a single month, these series cannot be expected to capture all the information available to the market in the same period. Stock prices, on the other hand, respond very quickly to public information. The effect of this is to guarantee that market returns will be, at best, weakly related and very noisy relative to innovations in macroeconomic factors. This should bias our results in favor of finding a stronger linkage between the time-series returns on market indices and other portfolios of stock returns than between these portfolio returns and innovations in the macro variables. To examine the relative pricing influence of the traditional market indices we used the following variables: EWNY(t) = return on the equally weighted NYSE index; VWNY(t) = return on the value-weighted NYSE index. These variables should reflect both the real information in the industrial production series and the nominal influence of the inflation variables. F. Consumption In addition to the macro variables discussed above, we also examined a time series of percentage changes in real consumption, CG. The series is in real per capita terms and includes service flows. It was constructed by dividing the CITIBASE series of seasonally adjusted real consumption (excluding durables) by the Bureau of Census's monthly population estimates. The CG series extends from January 1959 to December 1983, and it is an extension of a series obtained from Lars Hansen for the period through 1979. A detailed description of its construction can be found in Hansen and Singleton (1983). G. Oil Prices It is often argued that oil prices must be included in any list of the systematic factors that influence stock market returns and pricing. To test this proposition and to examine another alternative to the macro variables discussed above, we formed the OG series of realized This content downloaded by the authorized user from 192.168.82.218 on Tue, 4 Dec 2012 03:43:28 AM All use subject to JSTOR Terms and Conditions
Forces and the Stock Markel monthly first differences in the logarithm of the Producer Price Index/ Crude Petroleum series(obtained from the Bureau of Labor Statistics U.S. Department of Labor, DRI series no 3884). The glossary in table I summarizes the variables H. Statistical Characteristics of the Macro variables Table 2 displays the correlation matrix for the state variables. The orrelation matrices of table 2 are computed for several different pe- TABLE 2 Correlation Matrices for Economic variables EWNY VWNY UPR UTS A. January 1953-November 1983 VWNY DEI .163 UTS 159-394 -,752 270 139 .005 13 B. January 1953-December 1972 VWNY .147 DEI 059 C. January 1973-December 1977 VWNY UTS 554 335 361 174 D. January 1978-November 1983 VWNY UTS NoTE -VWNY return on the value-weighted NYSE index; EWNY (Baa and under return -long-term government bond return); UTS unanticipated change in the term structure (long-term government bond return Treasury-bill rate); and YP yearly gro ontent downloaded by the authorized user from 192.168. 82.218 on Tue, 4 Dee 2012 03: 43 28 AM
Economic Forces and the Stock Market 391 monthly first differences in the logarithm of the Producer Price Index/ Crude Petroleum series (obtained from the Bureau of Labor Statistics, U.S. Department of Labor, DRI series no. 3884). The glossary in table 1 summarizes the variables. H. Statistical Characteristics of the Macro Variables Table 2 displays the correlation matrix for the state variables. The correlation matrices of table 2 are computed for several different peTABLE 2 Correlation Matrices for Economic Variables Symbol EWNY VWNY MP DEI UT UPR UTS A. January 1953-November 1983 VWNY .916 MP .103 .020 DEI -.163 -.119 .063 UT -.163 -.112 -.067 .378 UPR .105 .042 .216 .266 .018 UTS .227 .248 - .159 -.394 - .103 - .752 YP .270 .270 .139 - .003 - .005 .113 .099 B. January 1953-December 1972 VWNY .930 MP .147 .081 DEI -.130 -.122 .020 UT -.081 -.021 -.203 .388 UPR .265 .214 .213 .068 - .072 UTS .110 .108 -.059 -.210 -.041 -.688 YP .260 .238 .128 - .013 -.032 .128 .063 C. January 1973-December 1977 VWNY .883 MP .022 -.118 DEI -.314 -.263 .004 UT -.377 -.352 -.004 .505 UPR .341 .231 .227 .032 -.289 UTS .217 .313 - .350 -.280 .026 -.554 YP .335 .361 .107 - .124 - .334 .221 .174 D. January 1978-November 1983 VWNY .937 MP .092 -.010 DEI - .143 -.073 .169 UT -.055 -.024 .168 .375 UPR - .275 - .319 .248 .458 .259 UTS .424 .431 - .277 - .512 - .239 - .890 YP .269 .261 .193 .053 .247 .018 .115 NOTE.-VWNY = return on the value-weighted NYSE index; EWNY = return on the equally weighted NYSE index; MP = monthly growth rate in- industrial production; DEI = change in expected inflation; UI = unanticipated inflation; UPR = unanticipated change in the risk premium (Baa and under return - long-term government bond return); UTS = unanticipated change in the term structure (long-term government bond return - Treasury-bill rate); and YP = yearly growth rate in industrial production. This content downloaded by the authorized user from 192.168.82.218 on Tue, 4 Dec 2012 03:43:28 AM All use subject to JSTOR Terms and Conditions