Leigh McAlister, Raji Srinivasan, & MinChung Kim

Advertising, Research and
Development, and Systematic Risk
of the Firm
Marketing executives are being urged to speak in the language of finance to gain internal support for marketing
initiatives. Responding to this call, the authors examine the impact of a firm’s advertising and its research and
development (R&D) on the systematic risk of its stock, a key metric for publicly listed firms. They hypothesize that
a firm’s advertising and R&D expenditures create intangible assets that insulate it from stock market changes,
lowering its systematic risk. They test the hypotheses using a panel data on 644 publicly listed firms between 1979
and 2001, consisting of five-year moving windows. They scale the firm’s advertising and R&D expenditures by its
sales. After controlling for factors that accounting and finance researchers have shown to be associated with
systematic risk, the authors find that advertising/sales and R&D/sales lower a firm’s systematic risk. The article’s
findings extend prior research that has focused primarily on the effect of marketing initiatives on performance
metrics without consideration of systematic risk. For practice, the ability of advertising and R&D to reduce
systematic risk highlights the multifaceted implications of advertising and research programs. The article’s findings
may surprise senior management, some of whom are skeptical of the financial accountability of advertising

here is a growing consensus that senior management
and finance executives focus on maximizing shareholder value and do not value marketing performance
metrics (e.g., awareness, sales growth, loyalty, customer
satisfaction, repeat purchase), because they do not understand how or even whether these metrics are of interest to
the firm’s shareholders (Ambler 2003). Consequently, marketing executives are urged to “speak in the language of
finance” with their finance colleagues and senior management (Srivastava and Reibstein 2005) because financial
return is the dialogue required to access funds from the
financial purse strings that are crucial for the implementation of marketing programs. To address this gap, in this article, we examine whether a firm’s advertising and researchand-development (R&D) expenditures affect a metric of
interest to both finance executives and senior management:
the firm’s “systematic risk,” or β (we use these terms inter-

changeably throughout the article to denote the systematic
risk of the firm’s stock).1
Portfolio theory (Lintner 1965; Sharpe 1964), a key
development in finance, posits that investors can diversify
away a portion of the risk associated with a firm’s stock by
constructing a portfolio of stocks whose returns correlate
imperfectly with one another. In equilibrium, the risk that is
priced in the stock market is the stock’s systematic risk,
which is a function of the extent to which the stock’s return
changes when the overall market changes.2 This marketdriven variation in a firm’s stock returns, which cannot be
diversified away, is its systematic risk. By construction, the
stock market, as a whole, has a β of 1.0. A stock whose
return falls (or rises) more than the market’s return falls (or
rises) in response to a change in the market has a β greater
than 1.0. If a stock’s return falls (or rises) less than the market’s return falls (or rises) in response to a change in the
market, its β is less than 1.0. Thus, β, a measure of the
stock’s sensitivity to market changes, is an important metric
for publicly listed firms.
In this article, we examine the relationship between a
firm’s advertising and R&D and its systematic risk. Recent
developments in the market-based assets theory (Srivastava,
Shervani, and Fahey 1998) suggest that a firm’s advertising
creates intangible market-based assets (e.g., brand equity)
and that these assets strengthen performance—including
sales growth, market share, and profitability (Boulding and
Staelin 1995; Erickson and Jacobson 1992)—and shareholder value (Joshi and Hanssens 2004; Rao, Agarwal, and


Leigh McAlister is Professor and H.E. Hartfelder/The Southland Corporation Regents Chair for Effective Business Leadership (e-mail: Leigh.
McAlister@mccombs.utexas.edu), Raji Srinivasan is Assistant Professor
of Marketing (e-mail: raji.srinivasan@mccombs.utexas.edu), and
MinChung Kim is a doctoral candidate in Marketing (e-mail: MinChung.
Kim@phd.mccombs.utexas.edu), McCombs School of Business, University of Texas at Austin. The authors thank Jay Hartzell, Robert Parrino,
and Laura Starks for comments and perspectives that helped shape this
article and Robert Jacobson, Don Lehmann, and Natalie Mizik for comments on previous drafts of the article. The authors also thank Hans
Baumgartner, Christophe Van den Bulte, Joe Cote, Rajdeep Grewal,
Joonho Lee, and Carl Mela for their inputs. The authors acknowledge the
constructive suggestions of the three anonymous JM reviewers in the article’s development.

1As we discuss subsequently, to eliminate potentially confounding effects of firm size on systematic risk, in the empirical estimation, we scale the firm’s advertising and R&D by its sales.
2Thus, β is measured as (covariance [stock and market])/(variance [market]).

To read or contribute to reader and author dialogue on JM, visit

© 2007, American Marketing Association
ISSN: 0022-2429 (print), 1547-7185 (electronic)


Journal of Marketing
Vol. 71 (January 2007), 35–48

We develop hypotheses that relate a firm’s advertising and R&D to its systematic risk.Dahloff 2004). proposing that a firm’s advertising and R&D lower its systematic risk. 57) note that because of increased firm awareness due to advertising. We test the hypotheses using data on publicly listed firms that we obtained from the COMPUSTAT and Center for Research on Stock Prices (CRSP) databases for the period between 1979 and 2001. We control for the firm’s growth. Consistent with these theoretical developments. Frieder and Subrahmanyam (2005. and Nejadmalayeri (2005) report a significant. Faircloth. We organize the article as follows: In the next section. and dividend payout. In 1970. Fehle. Singh. Faircloth. the measures.” Grullon. We estimate the effect of a firm’s advertising/sales and R&D/ sales on its systematic risk using a fixed-effects model formulation that accounts for unobserved firm heterogeneity and serial correlation of errors. we examine the impact of a firm’s advertising and R&D on its systematic risk. Following precedent in the finance literature (Damodaran 2001). leverage. We anticipate that this broader ownership may insulate the stock’s return from market downturns. two recent studies (Madden.e. p. such as a more general sample that includes poorly performing firms. all else being equal. and the results. individualistic) risk associated with those stocks. and the results are not driven by multicollinearity. The model includes two additional control variables that may affect a firm’s systematic risk: firm age and competitive intensity in the industry. factors that finance and accounting scholars have shown to be associated with systematic risk. Fehle. The remaining variability. The key idea of portfolio theory is that investors can construct a portfolio of stocks with imperfectly correlated returns and thus eliminate nonsystematic (i. Singh. that is. we estimate the firm’s systematic risk using 60 months of stock returns in a fiveyear moving window and equal-weighted stock market returns.. in firms’ stocks (Lintner 1965. To eliminate the potentially confounding effects of firm size on systematic risk. and Fournier 2006. Beaver. Faircloth. Mossin 1966. “investors prefer holding stocks with high recognition and consequently. Madden. and Nejadmalayeri 2005) have explored the relationship between a firm’s advertising and its systematic risk. R&D creates intangible market-based assets that insulate the firm from the negative impact of stock market downturns. after we control for factors that prior research has shown to be associated with systematic risk. Our findings are novel and important and hold implications for both marketing theory and practice. are controlled for (but were not controlled for in Madden. thus lowering the firm’s systematic risk. we scale the firm’s advertising and R&D expenditures by its sales. market returns and relax the restriction that all 60 months of stock returns must be present to estimate β) and to alternative measures of advertising and R&D (we scale them by assets rather than by sales). Singh. January 2007 Accordingly. 36 / Journal of Marketing. There is also evidence in the literature that links a firm’s R&D expenditure to its financial performance (Boulding and Staelin 1995. we propose another way that advertising-created market-based assets might affect a firm’s systematic risk. Sharpe 1964). as opposed to equalweighted. earnings variability. issues that have not been examined in prior research. Kanatas. and Scholes (hereinafter. the firm’s systematic risk. On the basis of developments in the finance literature. or with measures of advertising scaled for firm size (to remove the confounding effect of firm size)? Will the negative relationship between advertising and systematic risk hold when other accounting characteristics that have been shown to link to a firm’s systematic risk. We then describe the proposed estimation approach. and opportunities for further research. Fehle. and a portfolio of all firms—and find that the portfolio of firms with strong brands had higher returns and lower systematic risk than the other two portfolios. thus lowering the firm’s systematic risk. Fehle. and Nejadmalayeri find that greater advertising expenditure (operationalized as advertising dollars) is associated with lower systematic risk. we provide a brief overview of systematic risk. and Weston (2004) find that a firm’s advertising results in broader ownership of the stock. which may vary across portfolios. Erickson and Jacobson 1992) and shareholder value (Jaffe 1986). and Fournier (2006) compare the performance of three stock portfolios—a portfolio of firms with strong brands (using Interbrand’s measure of brand strength). there is a research opportunity to incorporate unobserved firm heterogeneity (to rule out endogeneity caused by omitted variables) and to use lagged predictor variables (to rule out reverse causality). and Fournier’s (2006) studies raise intriguing research questions: Will the negative relationship between advertising (or brand strength) and systematic risk hold under other conditions. greater information precision. and Hoenig 1990. An Overview of Systematic Risk A central issue in portfolio theory in finance is the maximization of returns for people who invest in assets. its limitations. Singh. which resulted in the creation of a panel data set of 19 five-year moving windows with 3198 observations for 644 firms. Capon. the data. The results strongly support the hypotheses that higher advertising/sales and higher R&D/sales lower a firm’s systematic risk. Farley. and Fournier’s study)? In addition. Faircloth. negative relationship between a firm’s advertising and its systematic risk. reflects the extent to which its stock’s return responds to movement of the average return on all stocks in the market. These two effects are robust to alternative estimates of systematic risk (we estimate β using value-weighted. We hypothesize that as in the case of advertising. and Nejadmalayeri’s (2005) and Madden. We conclude with a discussion of the study’s contributions. We suggest that the consumer loyalty and the bargaining power over distribution channel partners inherent in these intangible market-based assets help insulate the firm from the impact of stock market downturns. Using a sample of “bestperforming firms” from the Stern–Stewart database for the period between 1998 and 2001. asset size. liquidity. BKS) related systematic risk to variables that describe the finan- . a portfolio of firms excluding firms with strong brands. Kettler.

Eskew 1979. advertising. in the impact of advertising and R&D on β. For example. growth. In summary. sales force expenditure. the excessive earnings opportunities may erode when new firms enter the industry. they suggested that greater systematic risk is related to the following: •Higher growth because. leverage. Kroll. focuses on the prediction of the firm’s systematic risk in a future period (Elgers 1980. Singh. and Systematic Risk / 37 . •Smaller asset size because smaller firms have greater default risk. much less is known about the relationship between important indicators of marketing strategy (e. Rather. and Heiens 1999) explore the relationship between aspects of a firm’s marketing strategy and its risk. To eliminate the potentially confounding effects of firm size on systematic risk. developments in brand equity (Aaker 1996. Two diverse streams of empirical research have emerged from BKS’s (1970) study. which is more pertinent to the current study. earnings funds flow and cash flow (Ismail and Kim 1989). and Fournier (2006) find that higher levels of advertising expenditure are associated with lower systematic risk. Indeed. profit (Erickson and Jacobson 1992). Fama and French (1992) examine whether expected returns are better predicted by the firm’s previous βs than by other variables. The second stream of research. We first discuss the effects of the firm’s advertising on its systematic risk and then discuss the effects of R&D on systematic risk. Fehle. Note that there is a vigorous. systematic risk is an important metric for publicly listed firms that measure their stocks’ vulnerability to market downturns. although there is much work that relates a firm’s accounting characteristics (e. liquidity. a surrogate for total risk. As we noted previously. consequently. and strategic profiles (Veliyath and Ferris 1997). financial structure (Hill and Stone 1980). Considering two periods (1947–1956 and 1957–1965). and Nejadmalayeri (2005) and Madden.g. •Greater leverage because the earnings stream of common shareholders becomes more volatile as debt increases. Wright. relative price) are associated with lower variability in return on investment. we scale the firm’s advertising and R&D by its sales. which is not related to the current study’s objectives. increased 3Using historical data. •Higher levels of earnings variability because this results in a lower payout to stockholders.g. advertising. Although variability in return on investment. and firm value (Joshi and Hanssens 2004).. Ismail and Kim 1989). BKS (1970) (1) regressed the aforementioned firm characteristics on systematic risk in the first period and (2) examined whether a model of systematic risk from the first period predicted systematic risk in second period better than did systematic risk in first period. Fidelity. these results suggest a relationship between firms’ marketing activities and their systematic risk. a review of current investment practices indicates that leading investment firms (e. dividend payout. and Fournier (2006) provide two exceptions. operating leverage (Mandelker and Rhee 1984). not on β as a predictor of future firm value. two important manifestations of the firm’s marketing strategy. we examine the effects of a firm’s advertising and R&D. using the Profit Impact of Marketing Strategy database) decreases the surrogate measure of risk. each study adds some new variables to a subset of the variables in BKS’s study. two studies (Bharadwaj and Menon 1993. the covariance of firms’ cash flows relative to a market portfolio of equities. To address this research gap. promotional expenditure. our focus is on β as a measure of risk. they find a stronger empirical correlation of future firm value with a firm’s book-to-price and size but not with the measure of its historic βs. Faircloth. Advertising. To reiterate its central role in investment practice. earnings variability) to its systematic risk. and Heiens (1999) consider a surrogate for systematic risk. Theory We now develop hypotheses that relate a firm’s advertising and R&D to its systematic risk. confounds systematic risk with nonsystematic risk. Kroll. Specifically. Faircloth. Similarly. Keller 1998) suggest that firms’ advertising efforts create consumer and distributor brand equity. an intangible market-based asset with important strategic and performance implications. Reinforcing these performance rewards to advertising. Merrill Lynch. Wright.. Although there are several studies in this stream. whereas other aspects of marketing (e. Using realized average returns. Variables considered in prior research include dividend policy (Bildersee 1975). •Lower dividend payout because the need to offer steady dividends causes firms with greater volatility to pay out a lower percentage of earnings. and Value Line) use β extensively in the construction of investment portfolios. and •Higher earnings covariability with the market because this results in higher earnings volatility. Advertising A large body of work indicates that advertising has a direct effect on various firm performance metrics. Singh. our review indicates a lack of cumulative knowledge building in this area. on its systematic risk. Using service strategic business units from the Profit Impact of Marketing Strategy database.3 However. The first stream of research.g. again lowering the return on the stocks.. Bharadwaj and Menon (1993) find that some aspects of marketing (e. augments the predictor variables in BKS’s study with additional firm characteristics that may explain systematic risk. in the empirical estimation.cial position of a firm. international diversification (Goldberg and Heflin 1995).g. and they find that the superior product quality of strategic business units (again. as testimony of its importance. including sales (Leone 1995). ongoing debate about the usefulness of systematic risk for predicting future firm value in the finance literature (Fama and French 1992). shareholders and senior management of publicly traded firms are interested in β and.g. Thus. R&D. R&D expenditures) and systematic risk. customization) are associated with higher variability in return on investment... and Nejadmalayeri (2005) and Madden. Fehle. in a competitive economy. asset size. •Lower liquidity because liquid or current assets result in less volatile returns than do fixed assets.

they are nonsystematic and can be diversified away (Lubatkin and O’Neill 1987). Finally. Grullon. although extant empirical research suggests that R&D can increase a firm’s nonsystematic risk. and Thompson 2002). Capon. and its systematic risk. including its advertising. which insulate the firm from market downturns. the brand equity of current flagship brands generates greater receptiveness of consumers and distribution channel partners to new product introductions (Kaufman. Kanatas. Jennings. and Hoenig (1990. and Fahey (1998) suggest. and Weston 2004). and 38 / Journal of Marketing. the number of a firm’s new product introductions lowers its systematic risk (Chaney. financial. technologies. and flexibility. Method Data The data for this study included all firms listed on the New York Stock Exchange (NYSE) during the period between 1979 and 2001. Laguerre. and Rose 2006) and enables the firm to migrate customers to more profitable products and/or to cross-sell products to existing customers (Kamakura et al.” As with advertising-created market-based assets. increased advertising and the resultant higher brand equity produce an asymmetric sales response to sales promotions (Blattberg. Erickson and Jacobson 1992). Kanatas. Although there is a debate about the sizes of the effects of R&D investments on different performance metrics (Boulding and Staelin 1995. Furthermore. an activity with uncertain returns. Measures The dependent variable. as Srivastava. and Leone 2002). a direct outcome of its increased advertising. and superior market value (Jaffe 1986). Shervani. Frieder and Subrahmanyam (2005) report that a firm’s increased brand perceptions (consistent with higher brand equity. acts as a barrier to competition. high stock returns (Chan. Briesch. Veliyath and Ferris (1997) report a relationship between the strategic profile of a firm. and Lehmann 2003) and lowers price sensitivities (Kaul and Wittink 1995. A firm’s systematic risk changes slowly over time. Pakes 1985). and marketing research that relates the intangible assets created by R&D to the firm’s financial performance. efficiency. Thus. R&D There is a large body of finance. increases ownership of the firm’s stock by individual investors (relative to institutional investors) because of individual investors’ preferences for stocks with higher-quality information (advertising plays an information role for a firm’s stockholders). Farley. 2003).and R&D-driven differentiation. and Lehmann 1997). as we discussed previously). In summary. Mizik and Jacobson 2003. To the extent that these kinds of volatility are specific to a firm or an industry. p. on its systematic risk. Thus: H1: The higher a firm’s advertising. brand equity also creates both consumer and distributor loyalty. thus lowering its systematic risk.advertising and the resultant brand equity increase the differentiation of a firm’s products (Kirmani and Zeithaml 1993) and make them less easily substitutable (Mela. This higher liquidity and increased breadth of ownership may help insulate the firm’s stock returns from market downturns. and provides bargaining power over distributors. We focus on the effects of a firm’s R&D. In a meta-analysis of 210 profitability studies. Neslin. this literature also suggests that R&D creates strategic differentiation. For example. R&D may increase total risk because R&D may decrease the predictability of a firm’s future income streams (Kothari. Analysts exhibit greater disagreement about year-ahead earnings for R&D-intensive firms than for other firms (Barth. systematic risk. advertising and the resultant brand equity also strengthen and stabilize the firm’s performance in new product markets. Kasznik. such that highly advertised brands are affected less (than less advertised brands) by competitors’ sales promotions. Devinney. Another study notes that postinvestment reported earnings are more highly variable for firms with higher R&D levels than for firms with lower R&D levels (Chambers. If the focus is on total risk (nonsystematic risk and systematic risk). enabling it to adapt to environmental changes. thus lowering its systematic risk. January 2007 Sougiannis 2001. is an inherently long-term construct that captures the extent to which a firm’s stock return covaries with market return (BKS 1970). “Dollars spent on R&D have an especially strong relationship to increased profitability. 1157) conclude. thus lowering its systematic risk. a firm’s brand equity may lower its systematic risk by serving as a capital market information channel to the firm’s stockholders (Frieder and Subrahmanyam 2005. This relationship between R&D and systematic risk occurs because a firm that invests in R&D exhibits greater dynamic efficiency and greater flexibility than its competitors (which invest less in R&D). brand equity can function as financial hedging contracts when firms enter new markets with new technologies. including in input prices. and customers (Miller and Bromiley 1990). these are all benefits that insulate a firm’s stock from market downturns and thus lower its systematic risk (Veliyath and Ferris 1997). R&Dcreated market-based assets may also insulate a firm’s stock from market downturns. and McNichols 2001). In addition to the benefits of advertising and brand equity in current product markets. and we obtained their stock prices for the computation of systematic risk from CRSP. Lakonishok. Grullon. We obtained accounting. it is well established that firms’ R&D investments generate persistent profits (Roberts 2001). Thus: H2: The higher a firm’s R&D. Sethuraman and Tellis 1991). and Fox 1995). and Winer 1991). In addition. Jayachandran. the lower is its systematic risk. Gupta. the lower is its systematic risk. Similarly. advertising. and Weston (2004) report that firms with higher advertising have higher liquidity and greater breadth of stock ownership. and R&D data on firms from COMPUSTAT. Increased brand equity also increases price premiums (Ailawadi. management. We followed the precedent in prior finance research (Damodaran . This efficiency and flexibility help insulate the firm from market downturns.

lagged advertising/sales. and Systematic Risk / 39 . DATA45 in COMPUSTAT includes the cost of advertising media (radio. which we measured by the number of years since its listing on the stock market. 2004). relative to the equalweighted return for the stock market for that period. we used the monthly stock returns for all firms on CRSP for 1979. Rmt is an index of the ex post return for all NYSE firms during month t (i. The number of firms in the sample by each moving window suggests that. asset size (log). we used the monthly stock returns for all firms for 1997. Given the theoretical processes discussed in the hypotheses development. Accounting variables. We are aware that some firms may report sales promotion expenditure as a part of their advertising expenditure. television. and Shanken 1995. In addition. to avoid problems associated with low-priced stocks. For the first moving window. as accounting researchers have done. and dividend payout (for the logic for the operationalization of these variables. Scaling the firm’s advertising and R&D expenditures by its sales rules out the alternative explanation that the negative effect of advertising and R&D on systematic risk may be due to larger firms having lower systematic risk. Second. we used the monthly stock returns for firms for 1980. a market price index of all firms on the NYSE at months t and t – 1. p. 1998. Thus. following the suggestion of an anonymous reviewer. we anticipate a lagged effect of a firm’s advertising and R&D on its systematic risk. we eliminated 52 observations in which the firm’s estimated systematic risk was negative. Systematic risk. For the second moving window. Accordingly. 1983.. and the control variables. We subsequently tested the robustness of the results to β estimates relative to the value-weighted returns and for β estimates when monthly stock returns were available for at least 50 of the 60 months of the moving window (which enables us to increase the number of firms in the data set). A total of 3198 observations (for 644 firms) had complete data on systematic risk. the market rate of return). we use monthly stock data to compute firm i’s systematic risk measure βi(hat). The inclusion of the firm’s asset size serves as a further control for the effects of the firm’s size on its systematic risk. Advertising and R&D. 1980. Hertzel et al. because firms do not indicate the split between advertising and sales promotions. We also included two additional control variables in the model. Finally. The fixed-effects formulation we employ to estimate the model precludes the inclusion of time-invariant industry dummies in the model. we accomplished this by using stock returns for the previous 60 months. which we measured with Herfindahl’s fourfirm concentration ratio as the proportion of market shares of the largest four firms to the industry’s sales at the twodigit Standard Industrial Classification level. because older firms may have lower systematic risk. We measured advertising by the mean of the firm’s advertising expenditure. we included a firm in the moving window only if it reported information on its advertising and R&D in COMPUSTAT for all years in the five-year moving window. and periodicals) and promotional expenses.g. Kothari. and Lehmann 1997) and the mean of the firm’s R&D expenditure. where Rit = Ln(Dit + Pit)/P′it – 1 and Rmt = Ln(Lt/Lt – 1). and Lt and Lt – 1 are the Fisher’s link relative. We obtained βi(hat) by estimating a separate regression using the monthly stock returns for each firm i for each five-year moving window. The slope of the regression equation βi(hat) is the empirical estimate of systematic risk βi of firm i. we included the industry’s competitive intensity. we included the firm’s age. Pit is the closing price of common stock i at the end of month t. Similar to BKS’s (1970) approach. resulting in up to 19 observations per firm. earnings variability.2001) and estimated the firm’s systematic risk using a fiveyear moving window. 1999. We subsequently test the model and report that estimation results are robust to scaling of the advertising and R&D by the firm’s assets (DATA6) instead of its sales. making our tests of hypotheses conservative. stock splits. to control for industry-specific effects. For the final moving window. we assume that most advertising expenditures reported in DATA45 in COMPUSTAT are related to communication of product benefits to customers. differences in the percentage of such misreported advertising expenditure that pertains to sales promotion add error to our estimates. for the five-year period from the annual data reported in COMPUSTAT. for a period by using a least squares regression of the form: (1) Rit = αi + βiRmt + εit. adjusted for dividends and all capital changes. First. where Rit is the ex post rate of return for stock i during period t. We are unable to include covariability of earnings. P′it – 1 is the closing price at the end of month t – 1 adjusted for capital changes (e. 2000. …. and 1983 to compute the firm’s systematic risk. 1981. and 2001. End. However. Given this study’s objective of exploring the effects of a firm’s advertising and R&D on its systematic risk. as reported in DATA46 and scaled by its sales reported in DATA12.. Gupta. and 1984. 1982. ex post.4 We measured R&D by 4Prior research has shown that sales promotion activities negatively affect brand loyalty (Mela. stock dividends). 1982. 1981. respectively. we do not hypothesize the directional effect of competitive intensity. which BKS include in their model. αi is the intercept of the fitted line of Rit using Rmt. have a nil effect on stock returns (Pauwels et al. as reported in DATA45 and scaled by its sales reported in DATA12. R&D. see BKS 1970. which preclude a potential reverse-causality explanation of the effects (Boulding and Staelin 1995). we excluded a stock from the five-year moving window if the average of its monthly closing stock prices was less than $2 (Ball. we included six accounting characteristics of the firm in the model: asset growth rate. 2002).e. because its calculation requires ten years of data. for the five-year period from the annual data reported in COMPUSTAT. newspapers. leverage. we use accounting variables that BKS (1970) use as control variables in our model. Additional control variables. lagged R&D/sales. 666). To ease interpretation of the results. t = Start. Thus. We provide the definitions of these measures and the data fields from COMPUSTAT used for BKS’s computation in the Appendix. liquidity. Advertising. Dit is the cash dividend payable on common stock i in month t. we used lagged measures of advertising/sales and R&D/sales. A priori.

Earnings variability 9. and then declines to 162 in the last moving window (years = 1997–2001).003 –. The average value for systematic risk for all firms in a given window varies across the years.052 –.068 (.174 1. Age 11.164 –.068 –.112) . This drop in the number of firms in the moving windows over time occurred because of missing data for advertising and R&D.033 in Window 1. multicollinearity does not appear to be driving the results of the model estimation. Systematic risk 2.058 .000 –.065 –.005 . and window dummies on each of the predictor variables and systematic risk. in which we removed the effects of fixed effects. Ti. serial correlation.002 1. the model has the following structure: Yit = α + Xitβ + νi + εit = 1.154 .071 Notes: All correlations greater than . serial correlation.369 1.054 –. adding one predictor variable at a time to the model.000 .221) .152 (.104 . or window dummies in our model struc- ture.028 .11 are significant at p < .over time.000 .181 –.147 –. Lagged advertising/sales and lagged R&D/sales vary as follows: for average lagged advertising.153 (2.036 in Window 1. To explore this issue further.380 . and the sizes of the coefficients) across these stepwise regressions.079 –.009 . Growth 5.024 –. N. and Narendranathan 1982.083 . the number of firms in each moving window increases at first.076 –.951 million (Window 6). the model estimation results in Table 2 indicate that advertising/sales and R&D/sales have coefficients that are negative and significant at p < .005 1.388 (.000 . This “disconnect” occurs because the bivariate correlation matrix does not account for the fixed effects.853) 19. whereas average asset size (log) ranges from $6. Table 1 contains the descriptive statistics and correlation matrix of the measures.243 million (Window 14) to $4.000 .856 (Window 14) to 1. January 2007 .056 –. Liquidity 7.01. (2) 5We also performed stepwise regression analyses.000 .002 –.155 1. highest = . highest = .392 . Lagged R&D/sales 4. significance.128 .250 . serial correlation. and window dummies from the predictor variables by regressing fixed effects.040 (.000 2 3 4 5 6 7 8 9 10 –. Dividend payout 10. we estimated a fixed-effects.077 (.10. ….105 –.105 . Franzini.189 –.9% reported zero R&D expenditure.220 . Lagged advertising/sales 3.167 (2.036 –. However. reaches a maximum (N = 371 in Window 10 [years = 1989–1993]).000 .060 –.120) . Bhargava.527) . Competitive intensity M (SD) 1 1. Results Model Estimation Procedure Because the panel data set of moving windows consists of repeated observations of firms. showing that there are no harmful effects of multicollinearity.004 –. For example. The pattern of correlations in the adjusted correlation matrix corresponds closely to the regression results in Table 2. …. Leverage 6. and Figure 1 contains the frequency distribution of asset size (log) and systematic risk of observations in the study. t = 1.019 .989) 5. and 3.042 (.243 –.220 are significant at p < . ranging from .022 1.000 –. Woolridge 2002).05.237) 3.051 1.025 . Asset size 8.629 . cross-sectional.012 .10.000 –.399 (.016 .004 . we created an “adjusted” bivariate correlation matrix.112) 1.01. and all correlations greater than . We used the residuals from each of these regressions to create an adjusted correlation matrix (this and all other results that we do not report herein are available on request).410 (.182 –. and found consistent results for the effects of lagged advertising and lagged R&D (the directionality. and the correlation between advertising/sales and β is negative and significant only at p < .046 –.107 –. for average lagged R&D. 3% reported zero advertising expenditure.047 . time-series regression model with a correction for serial correlation of errors (Baltagi and Wu 1999. all correlations greater than . A perusal of Table 1 and Table 2 (which we discuss in detail subsequently) suggests that the pattern of bivariate correlation matrix in Table 1 is different from the pattern of regression results in Table 2.01. Specifically.081 (.112) .140 (Window 18). Of the observations.340 1.389 .009 –.055 in Window 15 to lowest = .121 1. in the bivariate correlation matrix in Table 1.228 .023 are significant at p < .088 in Window 15 to lowest = .053) . 40 / Journal of Marketing. the correlation between R&D/sales and β is positive and significant at p < .159) . TABLE 1 Descriptive Statistics and Bivariate Correlation Matrix for Variables in Proposed Model Variable 1.5 Thus.

and the R-square (within) for the proposed model is .i.01) are positively associated with a firm’s systematic risk. Given the changes in stock market regimes over time and the refinements in the estimation procedure we use herein (i. the differences in this study’s coefficients relative to those that BKS report are not surprising. The Xit include the accounting variables of growth.01).527).where εit = ρεit – 1 + ηit (|ρ| < 1. incorporating fixed effects and serial correlation in errors).167 (2. as Woolridge (2002) proposes. dividend payout. earnings variability.01) is negatively associated with systematic risk. and the Durbin–Watson statistic (Baltagi and Wu 1999) is not significant (t = 1. growth (b = . we can explore a causal relationship between advertising/sales and systematic risk and between R&D/sales and systematic risk (Boulding and Staelin 1995). asset size. we used lagged measures of advertising/sales and R&D/sales. significant coefficients indicated that growth (b = . p < .” 7We also estimated the model that BKS (1970) estimate for the 20 years in this study by constructing two data sets—one for the period 1980–1989 and one for the period 1990–1999—by averaging the various accounting variables for the 10 years following the procedure in BKS’s article. We checked for first-order serial correlation in errors. The failure to correct for serial correlation of errors. p < .161. The estimated autocorrelation coefficient. The proposed model is statistically significant (F(671. and Systematic Risk / 41 .6 We estimated the model using the “xtregar” procedure in STATA 9. R&D. liquidity.01) and leverage (b = .042 (. is large at . We report the results of the estimation of the proposed model in Column 1 of Table 2.01).d.359. whereas asset size (b = –. M (SD) = 1.200. lation of errors. p < . M (SD) = 5. or AR(1). We note that BKS report significant effects for dividend payout (b = –.58.091.s. 2527) = 17. We first compare the coefficients of the accounting variables in this study with BKS’s (1970) results. and earnings variability (b = 3. In addition to the predictor variables of lagged advertising/sales and lagged R&D/sales. The results of the ordinary least squares regression for the model with accounting variables were significant in both 10-year periods: Period 1: F(6. Higher-order processes involving several periods are both intractable and place a high burden on the researchers to justify more complex time-series processes. p < .668. p. we also included BKS’s (1970) control variables and the two additional control variables of age and competitive intensity. in support of the inclusion of an autoregressive (AR1) disturbance term. 257) notes. if it is present. and ηit is independent and identically distributed [i. we use a panel data model. The R-squares are Advertising. Because this structure precludes a potential reverse-causality explanation.e. 338) = 22. ρ.01).0.90. see Bhargava.).] with mean 0 and variance σ2h) and νi are assumed to be fixed parameters that may be correlated with the covariates Xit.05). p < . reconfirming the need to adjust for serial correlation in errors.01) reinforces the need for fixed-effects correction.501. n. p < . p < . Franzini. can result in inflated standard errors of parameter estimates and incorrect tests of hypotheses (for the detailed statistics of the fixed-effects panel data model with correction for first-order serial corre- Figure 1 Frequency Distribution A: Firm Size B: Systematic Risk (β) Notes: For Panel A. “the first-order autoregression has withstood the test of time and experimentation as a reasonable model for underlying processes that probably.7 6The most frequently analyzed process in the empirical econometrics literature is the first-order autoregression. process in which the errors across time t and t – 1 are correlated. We included dummies for each moving window to account for any differences across windows. leverage. and for Panel B. in truth. Thus.84. 231) = 17. are impenetrably complex.01. age. Period 2: F(6. competitive intensity.03.01). p < . and advertising and R&D. The rejection of the hypothesis of null fixed effects (F-value is significant at p < . As discussed previously.221). and Narendranathan 1982)..742. as Greene (2003. p < . In our results.515. This test rejected the hypothesis that there is no first-order serial correlation (p < .01.

119) .584)*** –. 3172) Equal weighted Changes in variablesa . ***p < .000 (.031)*** –.467 (.005 (.018 (.028) .021 (.785)*** –.006 (.42 / Journal of Marketing.175)*** .379 (.598 (.034)* .338)* . 3172) Equal weighted Changes in variables –.006)*** –. some of which are significant at p < .032) .459 (.180)*** . 2719) Equal weighted Levels of variables –.001 (.000) –.474 (.004)** .004) –. The models also include window dummies.022 (.421 (738.014 (.530 (. R&D.010) –.051 (671.037)*** –3. Minimum number of returns observations for estimating β Market return used in estimating β Model specification Intercept Lagged advertising/sales Lagged R&D/sales Growth Leverage Liquidity Asset size Earnings variability Dividend payout Age Competitive intensity Serial correlation (ρ) R2 Number of firms (observations) F (d.) Variable TABLE 2 Advertising.278 (.161 644 (3198) 17.05.190)*** .025 (.001 (.039)*** – 2.004)*** .004 (.359 (.542 (.202 (.004)*** .052 60 Column 6 .065 (.002 (.022 (.001 (.419 (.091 (.515 (.090)*** .031) –.023 (.093 (.000 (.011) –.021 (.010) –.006)*** –.000) –.705)*** –. January 2007 60 Value weighted Levels of variables –.073)*** .01.032) .000 (.004) — –.001 (. and Systematic Risk: Estimation Results 644 (3198) 6.059 (.102 50 Column 4 *p < .608 (. aWe used factor-scored predictor variables for this model.221 (.335)*** .754)*** –.339 (.010) –.845 (.668 .001 (.732)*** –1. Notes: Coefficient (standard errors) are in the columns.106 644 (3198) 11.183)*** .021 (.910 (25.008 (.014)* –.001 (. **p < .f.004) –.032)*** –.085)** –. 2719) Value weighted Levels of variables –.327)*** .032)*** –.000 (.004) –.598 (.989 (.007)*** –.038)*** –3.830 (.081)*** .065 (.004)*** –.007) –.034 (.346)* .115)*** .081 (.472 (738.654 .045) –2.000) — –.885 (.107 (.124)*** .018 (.282 (.187 (.084)*** .029) .403 (.643 .129) .052 60 Column 5 644 (3198) 6.041)*** –1.331) — .000) –.009 (.033)*** –. 2527) Column 2 60 Column 1 711 (3457) 18.443 (.901 (671.000) . 2527) Equal weighted Levels of variables –.026 (.656 .01.000 (.092 (.118)* .155 50 Column 3 711 (3457) 11.005) — .910 (25.089)** .501 (.036 (.007)** –.10.000 (.

We created the difference value for a variable as the difference between the variable at time t and at time t – 1.607. We next discuss the two hypothesized effects. and the baseline model = 3741. indicating that after we control for factors that accounting researchers have shown to affect the firm’s systematic risk. From these results. n. for Period 2. which includes lagged advertising/sales and lagged R&D/sales. we orthogonalized predictor variables using factor analysis. baseline model that includes only the accounting measures of growth. the remaining differences in our findings might be due to change in the marketplace. The effects of advertising/sales and R&D/sales on systematic risk using the larger data set of firms with as few as 50 months of stock returns are similar to those we obtained using the smaller data set.285. Following an anonymous reviewer’s suggestion.501. the firm’s age (b = –. leverage.01). n. Changes regression.01). earnings variability (+ and p < . those in BKS’s study.10). Columns 3 and 4 of Table 2 contain the results of model estimation using equal-weighted and value-weighted measures of market return in the β estimation for the expanded sample. p < . In the panel data set. earnings variability.01) are significantly associated with lower systematic risk. recall that BKS eliminated leverage and asset size in a stepwise regression of data averaged over 10 years. growth (+ and p < .05). and those of BKS’s model approach with the 1980–1999 data (see n. we explored the robustness of the results using a changes regression (Boulding and Staelin 1995).01) and higher lagged R&D/sales (b = –.000.001.315). higher lagged advertising/sales (b = –3.01). and Systematic Risk / 43 .018. dividend payout. but asset size and leverage were not associated with β.) are not related to systematic risk. we obtain consistent results for the effects of advertising/sales and R&D/sales on the firm’s systematic risk with valueweighted β estimates (see Column 2 of Table 2).s. The model with lagged advertising/sales and lagged R&D/sales outperforms the baseline model on the basis of the Schwarz Bayesian information criterion (a lower number indicates superior fit): The proposed model = 3705.34) and leverage (ρ = –. but we find no support for this interaction effect. We also reestimated the model using the interaction effect of lagged advertising and lagged R&D.).). The raw correlation structure in our panel data is consistent with this conjecture.01) and competitive intensity (b = –. The direction and significance levels of the coefficients of BKS’s (1970) variables in this baseline model without lagged advertising/ sales and lagged R&D/sales are unchanged. We conjecture that panel-structure-adjusted correlations may have shown that leverage and asset size were (as they are now) good predictors of systematic risk. driven by our up-to-date econometric estimation approach. p < . and dividend payout were significantly associated with β. and dividend payout (b = . with no significant effect for the squared terms. to explore nonlinear effects. we reestimated the proposed model. leverage (+ and p < . With BKS’s data and methodology.022.01). Thus. or they might be because BKS’s use of stepwise regression artificially eliminated leverage and asset size that our proposed model (with its panel structure) identifies as important predictors of β. In summary. Varimax rotation ensured that each factor had only one predictor variable that loaded heavily on that factor (variables’ loadings on their respective factors exceeded . which includes firms for which all 60 months of stock returns were available to estimate β. The accounting variables of liquidity (b = . correlations of liquidity with both asset size (ρ = . and asset size (– and p < . In terms of why BKS did not identify asset size and leverage as predictors of β.885. especially between “change in advertising scaled by sales” and “change in R&D scaled to sales. earnings variability (+ and p < .187. We compared the performance of the proposed model.01) lower the firm’s systematic risk. R&D. and earnings variability and the two additional control variables of age and competitive intensity (we do not report the results here). The results indicate that as H1 and H2 predicted. we can observe the impact of methodology. in large part.01). After correcting for methodology.98). and asset size (– and p < .s. 6) arise because of differences (1) in stock market regimes across the years that result from changes in the range of predictor variables and (2) in the econometric estimation procedure between our study and BKS’s study—we incorporate information on the panel data structure in the estimation. including the square of lagged advertising/sales and the square of lagged R&D/sales. To examine the robustness of the proposed model’s results to alternative specifications for systematic risk. we suggest that the difference between our results and BKS’s results are. growth (+ and p < . which we estimated with 60 months of equal-weighted returns using both raw predictor variables and factor-score predictor variables (to adjust for potential multicollinearity in the raw predictor Advertising. and Period 2 = . increases in both advertising/sales and R&D/sales lower a firm’s systematic risk. p < . We discuss this finding in detail in the “Discussion” section. The following accounting variables were significant: for Period 1. with a as follows: Period 1 = . the estimation results strongly support H1 and H2. which includes observations with 50 or more months of stock returns for the 60-month moving window.” which perhaps is not surprising because increases in R&D are likely to be associated with increases in advertising.05). Additional Analyses Explanatory power of proposed model. liquidity. Following an anonymous reviewer’s suggestion.We infer that differences among our findings. The correlation matrix of the differenced variables suggested potential multicollinearity problems. Alternative estimates of systematic risk. respectively. we expanded the data to include firms for which only 50 of the 60 months of stock returns in a moving window were available for β estimation. The proposed model we report in Column 1 of Table 2 uses equalweighted market returns to estimate the firm’s systematic risk. The results were similar to those we report in Column 1 of Table 2.37) are high. p < . growth.873. whereas BKS do not. n. earnings variability (b = –. and in general. leverage (+ and p < . again confirming that there is no evidence of multicollinearity of lagged advertising/sales and lagged R&D/sales with the control variables. we also estimated β for the five-year moving window using the value-weighted market return. We report the results of the changes regression models using difference scores for β.s. To examine the robustness of the negative effects of advertising/sales and R&D/sales on β. asset size. To reduce the impact of this multicollinearity. With respect to the other control variables.

509. indicate a reasonable model fit (F is significant at p < . we obtain similar results for the models that use β computed with value-weighted market return.05) and R&D/ assets (b = –1. our fixed-effects formulation rules out endogeneity that might be caused by omitted variables.e. an important metric of risk for publicly listed firms. 44 / Journal of Marketing.. particularly. In the results discussed thus far. there are few insights that relate a firm’s marketing initiatives to its systematic risk. We examine the impact of a firm’s advertising and R&D. The results (we do not report these here) using a model identical to the model reported in Column 1 of Table 2. However. we also attempted to rule out potential endogeneity explanations of the firm’s advertising/sales and R&D/sales. Both models included fixed effects.05) lower the firm’s systematic risk. including the regression of differenced variables. it is possible that either because of inertia (i. following the work of Boulding and Staelin (1995). the model we estimated considered the impact of the firm’s advertising/sales in Period t – 1 and R&D/sales in Period t – 1 on systematic risk in Period t to rule out reverse causation.01) and R&D/sales (b = –. We also empirically rule out reverse causality (i. two important manifestations of a firm’s marketing strategy. and earnings variability or cash flow variability. and window dummies.8 Reverse causality. Second. Not surprisingly. In addition. Following the instrumental variable estimation. To rule out reverse causality. we regressed advertising/sales in Period t as a function of all predictor variables in the proposed model and β in Period t – 1. p < .01). The results of this estimation indicate a reasonable model fit and suggest that increases in a firm’s lagged advertising/sales and lagged R&D/sales reduce the firm’s systematic risk.05. but with advertising/sales and R&D/sales replaced by the advertising/assets and R&D/assets.variables) in Columns 5 and 6 of Table 2.737. First. respectively. pp. an autocorrelation error term.05) on systematic risk. Discussion The accountability of marketing initiatives. including equalweighted or value-weighted market returns and 60 months or 50 months of returns. advertising/assets (b = –. systematic risk lowers advertising and R&D) and endogeneity of both advertising and R&D expenditures. We used the firm’s advertising/sales and R&D/sales in Period t – 2 as the instrument for the firm’s advertising/sales and R&D/sales in Period t – 1 and reestimated the model that relates a firm’s two-period-lagged advertising/sales and R&D/sales to its systematic risk. the results (we do not report these here) indicate a reasonable model fit (F is significant at p < ..e. on its systematic risk. The directionality of the parameter estimates of the predictor variables are consistent with those we report in Column 1 of Table 2. Specifically. the additional analyses strengthen our confidence in our key findings that both lagged advertising/sales and lagged R&D/sales lower a firm’s systematic risk. we performed two regressions to rule out potential reverse-causality explanations. Thus. we reestimated the model using cash flow variability as a control variable rather than earnings variability. we checked for endogeneity of lagged advertising/sales and lagged R&D/sales using an instrumental variable estimation procedure. Following an anonymous reviewer’s suggestion. alternative measures of systematic risk. In addition. we scaled both the firm’s advertising and R&D expenditures by its sales. we performed Granger-causality Wald tests for each time series in the data set using a bivariate approach (Leeflang and Wittink 1992) between (1) the firm’s systematic risk and its advertising/sales and (2) the firm’s systematic risk and its R&D/sales. advertising budgets and R&D budgets are set as a percentage of sales) or because managers are forward looking. especially as measured by metrics of interest to a firm’s shareholders. again following an anonymous reviewer’s suggestion. we also performed the Davidson–MacKinnon test (Woolridge 2002. January 2007 1 of Table 2. we reestimated the model with measures of the firm’s advertising and R&D scaled by its assets.024. Following the explanation that a firm’s managers may be forward looking. the results are consistent for the effects of both advertising/ sales (b = –3. The results of the regressions (we do not report these here) indicate that consistent with the lack of reverse causality we established previously. The results of the Wald tests indicate that a firm’s systematic risk does not “Granger cause” either advertising or R&D.443. Second. we regressed R&D/sales in Period t as a function of all predictor variables in the proposed model and β in Period t – 1. Granger 1969). Potential endogeneity of advertising and R&D. is under increasing scrutiny from senior management and finance executives who control marketing budgets. Again. we performed the Grangercausality Wald tests (Dekimpe and Hanssens 2000. First. As we noted. p < . as BKS (1970) use. the results are robust to alternative model specifications. . In summary. p < . marketing scholars have turned their attention to relationships between various aspects of a firm’s marketing strategy and shareholder value. empirically ruling out the reverse-causality explanation. p < . systematic risk. producing a wealth of insights that indicate an important role for marketing in shareholder wealth creation. Alternative measures of advertising and R&D. β in Period t – 1 does not affect either advertising/sales or R&D/sales in Period t. 118–22) of endogeneity for lagged advertising and lagged R&D and found no support for endogeneity of either advertising or R&D. though the significance level changes to p < .01). The results of this instrumental variable estimation procedure (we do not report these here) are consistent with those we obtained with the one-period-lagged predictor variables of advertising/sales and R&D/sales reported for the proposed model in Column 8In general. in support of H1 and H2. Alternative measure of earnings variability. advertising and R&D budgets in Period t – 1 may be related to Period t’s firm characteristics. alternative measures of advertising and R&D scaled by the firm’s sales or assets. However.

As finance and accounting researchers have done previously.283. sales.g. Faircloth. and Weston 2004) and the finding in the marketing literature that advertising has a direct impact on stock price beyond its indirect effect through increased sales (Joshi and Hanssens 2004). not only reducing its financial performance. this is the first empirical study to cover a broad. and/or time in the relationship between advertising/sales and β. Managerial Implications The study’s findings also generate useful implications for managerial practice. Although the capital asset pricing model assumes that firms with greater systematic risk can expect higher future returns. who do not control for accounting and finance factors and firmsspecific effects) holds up for all firms across a period that extends from 1979 to 2001. we address the several calls for marketing scholars and practitioners to speak in the language of finance (Rust et al. Shervani. We offer two possible explanations for this phenomenon. industries.01). and marketing variables (in this case. and Systematic Risk / 45 . and Nejadmalayeri [2005] in a limited empirical context and by Madden. which lends confidence to the study’s findings. This study’s finding that a firm’s advertising/sales and R&D/sales lowers its systematic risk. Consistent with the notion that advertising and R&D increase a firm’s stock returns.013. market shares) and financial (e. If firms that invest in advertising and R&D are able to raise their returns while lowering their risk. in the interest of generality. and stock returns but also increasing its systematic risk. industries.. we include accounting. cost of capital. our findings that a firm’s higher advertising/sales and R&D/sales lower its systematic risk are novel and useful.Theoretical Implications To our knowledge. Such an effect would be consistent with the finding in the finance literature that investors prefer highly advertised firms (Grullon. perhaps there are diminishing returns to increased advertising/sales (and R&D/sales) for some range of advertising/sales (and R&D/sales) values.. Empirical research that explores this issue further would be valuable and would contribute to both the marketing and the finance literature. which also includes financial variables. cash flow) performance outcomes. Tobin’s q] = . We believe that the study’s findings may surprise senior management and finance executives. Specifically. A reduction in a firm’s advertising or R&D can have a double negative effect. a post hoc analysis indicated that lagged advertising/sales and lagged R&D/sales were highly correlated with the firm’s intangible value (ρ[advertising/sales. p < . some of whom may view their firms’ advertising and R&D programs as discretionary activities. The finding of a nonsignificant effect of the quadratic term for a firm’s advertising/sales and R&D/sales on its systematic risk is notable and merits discussion. Indeed. industry. and discount rate. Tobin’s q] = . it might be expected that there would be no empirical evidence of the high risk–high return link. we conjecture that the risk-lowering and return-enhancing effects of advertising and R&D may contribute to the anomaly that Fama and French (1992) identify. Kanatas. the linear effect of advertising/sales may occur because a firm’s senior management and finance executives lack the tools to evaluate the potential impact and thus set advertising/sales below levels that yield the lowest systematic risk. Fehle. stock returns) would be valuable to senior marketing executives. Marketing executives can use these findings to stress the multifaceted role of strong advertising and R&D programs. Further research that explores firm. leads to an interesting conjecture. By focusing on the firm’s systematic risk. return on assets. Further research that examines other interchangeable effects of other aspects of firms’ marketing and financial strategies on metrics of interest to capital markets (e.. combined with results from other studies (Madden. and Fournier [2006]. we estimate our model with the broadest possible cross-section of firms and industries and for an extended period. Such a situation may occur if firms are wise enough to set advertising/sales (and R&D/sales) for optimal financial returns and if the optimal financial return level is lower than that required to deliver the lowest β. The negative relationship between a firm’s advertising expenditure and its systematic risk (shown by Singh.g. who are often under pressure from senior management teams to justify investments in advertising and R&D programs. Given the increasing calls for the accountability of marketing initiatives. Mizik and Jacobson 2003) that show that a firm’s advertising increases its stock return. The negative impact of advertising/sales and R&D/sales on systematic risk in our model. R&D. an important metric of considerable interest to senior executives of publicly listed firms. and perhaps advertising/sales (and R&D/sales) values have been constrained well below that optimum level. Fehle. ρ[(R&D/ sales. increases in advertising/sales and R&D/sales lower a firm’s systematic risk.01. Given the dual benefits of advertising and R&D for firm value through effects on both stock returns (Mizik and Jacobson 2003) and systematic risk. Alternatively. firms must be cautious in cutting back on their advertising and R&D programs. cost of capital. and Fahey 1998). and Fournier 2006. intangible value. First. we believe that marketing executives can raise potentially provocative questions about Advertising. suggests the potential interchangeability between a firm’s marketing and financial choice variables in managing systematic risk. Fama and French find that there is no relationship between a firm’s risk in one period and its return in the future. or periods for which such diminishing returns may be identifiable. A second possible explanation for our inability to detect a quadratic effect when there are diminishing returns to increases in advertising/sales is heterogeneity across firms. multi-industry sample of firms over a 22-year period to demonstrate that after accounting and finance factors related to systematic risk are controlled for. such that the relationship is linear. p < . and/or time-specific effects that moderate the effects of advertising on a firm’s systematic risk may uncover firms.e. financial. used only accounting and financial measures). beyond their effects on market (e. attendant cash flows. In a departure from most prior studies of systematic risk that have used a silo-based approach (i. 2004. advertising and R&D) in our model of systematic risk. Srivastava..g.

Such ongoing dialogue may be instructive to senior management and finance executives who control advertising and R&D budgets but are skeptical about the financial accountability of returns to these investments. including aspects of the programs’ effectiveness. In summary. sales. Disaggregated measures of a firm’s advertising and R&D programs for all publicly listed firms are not available. as a fixed percentage of sales) still apply. return on assets) but also for its systematic risk? We anticipate that the answers to these and related questions could guide the development of benchmarks to assess the returns to advertising and R&D programs.g. therefore. and marketing executives about the important “financial” role of their firms’ advertising and R&D expenditures (Rust et al. especially from the perspective of senior management and finance executives. of its marketing strategy. market share) and financial objectives (e.g.g.g. Theoretical research using complementary methods (e. two important indicators of the firm’s emphasis on differentiation and. Could marketing executives rightfully argue that some proportion of the firm’s advertising and R&D budgets be considered a financial expenditure aimed at lowering its interest burden? Although we are mindful about this study’s limited influence in changing finance managers’ mind-sets about the uncertain returns to their firms’ advertising and R&D investments. both independently and compared with those of other firms in the industry? What are the implications of advertising and R&D budgets going forward not only for the firm’s marketing objectives (e. where Pt and Et are the market value of common stocks and earnings at time t.g. entry timing). Further research that focuses on a few industry contexts and uses disaggregated measures of the various elements of advertising programs and new product development programs. which do not account for differences in implementation of advertising (e. 2004).. new product success rates. we focused on the relationship between a firm’s systematic risk and its advertising and R&D. finance executives. we hope that it serves as an impetus for an ongoing dialogue among senior management. what are the firm’s historical levels of advertising/sales. they represent consolidated input measures. and systematic risk.. For example. Indeed.g. efficiency of media planning) and new product development programs (e. R&D/sales. marketing channels) to systematic risk will be useful in setting a research agenda for further empirical research.. annual dollar amounts scaled by the firm’s sales. January 2007 . APPENDIX Proposed Model Predictor Variable Definitions and Measures Variable Definition Measure The five-year moving average of advertising/sales.. Furthermore. field studies) to develop a conceptual model and propositional inventories that relate other elements of marketing strategy (e. Limitations and Further Research Given data availability constraints for publicly listed firms. we view this study as an important first step in establishing that advertising and R&D lower systematic risk of a firm’s stock. respectively Advertising/sales R&D/sales Dividend payout Growth Variability of earnings 46 / Journal of Marketing. intellectual property rights.whether extant allocation norms for advertising and R&D (e.. we measured the firm’s advertising and R&D using aggregated. creativity of advertising campaigns.g. We hope that the study’s findings stimulate further work in this area. in-depth interviews. cash flows.. Although advertising and R&D expenditures are important. could provide a useful extension to generate actionable managerial implications regarding the effects of various elements of a firm’s advertising and new product development program on its systematic risk.. surveys. the study’s specific findings can guide marketing executives to initiate a dialogue with their finance counterparts. (1/5) × Σ Data45/Data12 for five years The five-year moving average of R&D/sales (1/5) × Σ Data46/Data12 for 5 years The five-year moving average of cash dividends/earnings (1/5) × Σ Cash dividend/Available income for five years The five-year moving average of terminal total assets/initial assets (1/5) × Ln(Total assets5/Total assets1) Leverage The five-year moving average of total senior securities (preferred stocks and bonds)/total assets (1/5) × Σ Total senior securities/Total assets for five years Liquidity The five-year moving average of current ratio (1/5) × Σ Current assets/Current liabilities for five years Asset size The five-year moving average of total assets (1/5) × Σ Ln(Total assets) for five years The five-year moving average of standard deviation of earnings:price ratio (Σ[Et/Pt – 1 – E{Et/Pt – 1}]2/4)1/2 for period t = 2 to t = 5.

and Dominique M. Wu (1999). 40 (1). 814–23. “Brand Perceptions and the Market for Common Stock. Ambler. (1979). “TimeSeries Models in Marketing: Past. “Excess Returns to R&D Intensive Firms. 17 (2–3). Advertising.” Journal of Marketing.” International Journal of Research in Marketing. Notes: We define the measures for the various predictor variables for each firm i for each of the 19 five-year moving windows. 107–118. Scholes (1970). “The Association Between the Level of International Diversification and Risk. Heflin (1995). Timothy M. 39 (February). Badi H. 57–85. 654–82. 67 (October). London: Financial Times/Prentice Hall. 19–40.” Journal of Business. “The Forecasting Ability of Accounting Risk Measures: Some Additional Evidence. 427–65.” Management Science. Capon. “The CrossSection of Expected Stock Returns. “Determinants of Success in Service-Based Industries: A PIMS-Based Empirical Investigation.F. Dennis. and M. 1–17. Corporate Finance: Theory and Practice. Ross Jennings. 38 (1). and W. R&D. Eugene F.” Marketing Science.” Management Science.” Journal of Financial & Quantitative Analysis. 133–58. 7 (July–September).” Review of Accounting Studies. Erickson. 14 (3). 50 (1). David (1996). Building Strong Brands. Tim (2003). Lehmann (2003). Ball. and Edward J. Josef Lakonishok. 56 (6). William and Richard Staelin (1995). Stephen R. Gary and Robert Jacobson (1992). 1143–59.. Kothari.” Journal of Accounting Research. Farley. and Scott Hoenig (1990). “The Association Between Market-Determined and Accounting-Determined Risk Measures. Dekimpe. G122–32. 183–93.. “Identifying Generalizable Effects of Strategic Actions on Prior Performance: The Case of Returns to Research and Development Spending. 45 (October). and Jay Shanken (1995). Aswath (2001). All the variables have a subscript i for each firm. and Frank L. and Anil Menon (1993). and Systematic Risk / 47 . 389–408.” Econometric Theory. Eskew. Richard Briesch. 533–49. Bildersee. Pieter T. 55 (3). L.. Boulding. “Problems in Measuring Portfolio Performance: An Application to Contrarian Investment Strategies. Hanssens (2000).” Accounting Review. “Analyst Coverage and Intangible Assets. Kettler. Louis K. G222–36. 573–610. Fama. 2431–56. Kusum L. Bharadwaj. Winer (1991). Frieder. Present and Future. Paul. COMPUSTAT Data Items Used in Constructing Variables Name of Variable Component Advertising/sales R&D/sales Total assets Income available for common stockholders Market value of common stock (Pt) Total senior securities (TotalSecurity) Current asset (CurrentAsset) Current liabilities (CurrentLiability) Cash dividends (CashDividend) Herfindahl’s concentration index COMPUSTAT Annual Data Items DATA45/DATA12 DATA46/DATA12 DATA6 DATA20 DATA24 × DATA25 × 1000 DATA5 + DATA9 + DATA10 DATA4 DATA5 DATA21 DATA12 REFERENCES Aaker.” Journal of Services Marketing. 6 (1). 64 (4).” Accounting Review.APPENDIX Continued Variable Definition Measure Firm age Number of years since the stock’s first listing on the stock market ⎯ Competitive intensity The four-firm Herfindahl’s concentration index Proportion of market share of the top four firms in the industry defined by two digits of the Standard Industrial Classification code. Laura and Avanidhar Subrahmanyam (2005). Robert K. “Revenue Premium as an Outcome Measure of Brand Equity. “Gaining Competitive Advantage Through Discretionary Expenditures: The Returns to R&D and Advertising. Mary E. and M. Neslin. Franzini. Baltagi. 79–107. John U. Marnik G. McNichols (2001). John S.S.. (1980).” Journal of Financial Economics. Ailawadi. Narendranathan (1982). Scott A. “Accounting-Based Risk Predictions: A Re-Examination. and Ping X.” Review of Economic Studies. S. “Unequally Spaced Panel Data Regressions with AR(1) Disturbances. (1975). and Russell S. 15 (6). Robert C. and Theodore Sougiannis (2001).. Fox (1995). Thompson (2002). A.. Elgers. New York: The Free Press. New York: John Wiley & Sons. Ray. 49 (1). P. 54 (1).” Journal of Finance. 7 (4). Barth. Blattberg. and Kenneth R. 81–98. “Serial Correlation and Fixed Effects Models. 47 (2). Noel. Chaney. and Donald R. Damodaran. and Robert P. Chambers.” Accounting Review. “The Impact of New Product Introductions on the Market Value of Firms. Devinney. 38 (9). Sundar G. “Determinants of Financial Performance: A Meta-Analysis.” The Accounting Review.” Marketing Science.” Journal of International Financial Management and Accounting. 1–25. Goldberg. 14 (3). 36 (10). “The Association Between a MarketDetermined Measure of Risk and Alternative Measures of Risk. Marketing and the Bottom Line. Chan.P. “How Promotions Work. Bhargava. Ron Kasznik. French (1992).” Journal of Finance. 1–34. Beaver. 1264–79. W. “The Stock Market Valuation of Research and Development Expenditures.

768–83. Strategic Brand Management. Peter W. (1969). and Philip Bromiley (1990). Rhee (1984). Weston (2004).” Managerial and Decision Economics. 39–57. Breadth of Ownership and Liquidity.” The American Economic Review. (2001). eds. 76 (5).P. Peter S. Sethuraman. “The Impact of the Degrees of Operating and Financial Leverage on Systematic Risk of Common Stock. Wittink (1992). 595–637. Lintner. 9 (1). 2–18. Hanssens (2004). 20 (1).” Accounting Review. “Merger Strategies and Capital Market Risk. and James P. Kent D. Kim (1989). ———. Hill. Adam B. 57 (6). Marketing Studies Center. “Diagnosing Competitive Reactions Using (Aggregated) Scanner Data. Tim Ambler. Michael and Hugh M. 160–74. Woolridge. 15 (3). Jeffrey M. “The Long-Term Impact of Promotion and Advertising on Consumer Brand Choice. 19 (3). Biel. “On the Association of Cash Flow Variables with Market Risk: Further Evidence. Greene. 239–50. (1964).” Journal of Marketing Research. NJ: Lawrence Erlbaum Associates. Tasadduq A. Michael Lemmon.” Journal of Marketing Research. 665–84. 13–37. George Kanatas. “Capital Asset Prices: A Theory of Market Equilibrium Under Conditions of Risk. Shervani. and Moon K. 20 (4). “Accounting Betas. Jaffe. T. Los Angeles.. 424–38.” Econometrica. Leone. Mizik. 68 (October).” Journal of Financial and Quantitative Analysis. 93 (2). and David Reibstein (2005).” The Review of Financial Studies. . Rose (2006). Wagner. Gershon N. Kumar. and Donald R. Srivastava. “Advertising Spending and Market Capitalization. (1986). Lehmann (1997). Perceived Quality.. 47 (1). NJ: Prentice Hall. 219–27. John (1965). Marketing Science Institute. “Capital Market Impact of Product Marketing Strategy: Evidence from the Relationship Between Advertising Expenses and Cost of Capital. and Denise Dahlhoff (2004). Robert P. Pakes. (2003). “Generalizing What Is Known About Temporal Aggregation and Advertising Carryover. Mossin. Kamakura.” Marketing Science.. “Long-Run Performance Following Private Placements of Equity. Roland T.” Academy of Management Journal. and Rajendra K. Sales Promotions. Jan (1966). Sunil Gupta. 34 (May). 19 (1). Rajendra K. 126–41. Econometric Analysis. “The Contribution of Product Quality to Competitive Advantage: Impacts on Systematic Variance and Unexplained Variance in Returns.” International Journal of Research in Marketing.” Journal of Marketing. and Susan Fournier (2006). Kevin Lane (1998). 125–36. “An Analysis of the Tradeoff Between Advertising and Price Discounting. and R.” Journal of Financial and Quantitative Analysis. Grullon. Natalie and Robert Jacobson (2003). and Dominique M.” Journal of the Academy of Marketing Science. Michel Wedel. Sheri Faircloth. “New Products. Kothari. Roberts. S. and Liam Fahey (1998).. Ferris (1997). University of California. G141–50. and Financial Leverage: A RiskComposition Approach to the Determinants of Systematic Risk. 48 / Journal of Marketing. Jose Antonio Rosa. 375–84. (2002). 355–82.” Journal of the Academy of Marketing Science.. Amna and Valarie A. “Innovation and Firm-Level Persistent Profitability: A Schumpeterian Framework. Hillsdale. 580–87.H. 432–44. 67 (January). and Firm Value: The Case of the Automobile Industry. January 2007 Madden. 63–76. C.” Journal of Marketing. 14 (3). “Brands Matter: An Empirical Demonstration of the Creation of Shareholder Value Through Branding. Gregory S. Heiens (1999). V. “MarketBased Assets and Shareholder Value: A Framework for Analysis. Sharpe. and Ghon S. “The Valuation of Risk Assets and the Selection of Risky Investments in Stock Portfolios and Capital Budgets. P. Badr E. Kaul. 34 (2).” Journal of Finance. and the Stock Market Rate of Return. 248–61.” Marketing Science. Vithala R. Miller. 30 (4). Rao. Frank Fehle. Agarwal. and A. “Advertising. Pauwels. Veliyath. MA: MIT Press. Upper Saddle River.” International Journal of Research in Marketing. and Lynn Rees (2002). O’Neill (1987). G151–60. 984–1001. Profits.” Journal of Business Research. 7 (4).” Journal of Political Economy. “How Is Manifest Branding Strategy Related to the Intangible Value of a Corporation?” Journal of Marketing. 45–57. Ariel (1985). Leone (2002). Ismail. Lubatkin. 64 (1). 2595–2617.” Academy of Management Journal. Rajaram and Stephen P. “Technological Opportunity and Spillovers of R&D: Evidence from Firms’ Patents. “Trading Off Between Value Creation and Value Appropriation: The Financial Implications of Shifts in Strategic Emphasis. Carl F. Systematic Operating Risk. 68 (October). Gustavo. M.” Journal of Marketing. 45–65. 142–56. 388. 33 (4). 143–82. 43 (November). Amit and Dominique M. Stone (1980). “Strategic Risk and Corporate Performance: An Analysis of Alternative Risk Measures. “Equilibrium in a Capital Asset Market. 756–79.J. 68 (October). “The Role of Relational Embeddedness in Retail Buyers’ Selection of New Products. William H. Keller.” Journal of Marketing Research. Anil and Dick R. Linck. 28 (May). 37 (3). Anderson School of Management. 224–35. Econometric Analysis of Cross Sections and Panel Data. R&D.” Working Paper No.” Review of Accounting Studies. Srivastava (2004). 439–61.Granger. and Bernell K. “Capitalization Versus Expensing: Evidence on the Uncertainty of Future Earnings from Capital Expenditures Versus R&D Outlays. Satish Jayachandran. “Advertising. Mandelker. and Jos Afonso Mazzon (2003). “Measuring Marketing Productivity: Current Knowledge and Future Directions.. Rust. William F. Manohar. Ned C. Raj and Gerard J. 34 (4). Kroll.” Econometrica: Journal of the Econometric Society.” in Brand Equity and Advertising. Wright. “On Patents.” Review of Economics and Statistics. “Metrics for Linking Marketing to Financial Performance. Hanssens (2004). 39 (3). and Randall L. 14 (3). 22 (4–5). 17 (2). David A. 62 (January). Peter. Carpenter. James S. Joshi. Kaufman. 390–409. Jorge Silva-Russo.” Journal of Marketing. Leeflang. (1995).” Strategic Management Journal. Mela. and Brand Image. and Dick R. Shuba Srinivasan. Wittink (1995).” Report No. Cambridge. “Agency Influences on Risk Reduction and Operating Performance: An Empirical Investigation Among Strategic Groups. Michael. Zeithaml (1993). Manoj K.W. New York: McGraw-Hill. 33 (4).” Journal of Finance. and Ali Nejadmalayeri (2005). “Investigating Causal Relations by Econometric Models and Cross-Spectral Methods. 05–200e. Laguerre. Aaker and Alexander L. “Cross-Selling Through Database Marketing: A Mixed Factor Analyzer for Data Augmentation and Prediction. Tellis (1991). Singh. and Market Value. Koen. Kirmani. Thomas J. Hertzel. 76–89. 425–42. “Empirical Generalizations About the Impact of Advertising on Price Sensitivity and Price.