Effect of analysts’ earnings pressure on marketing spending and stock market performance

Abstract

Despite the clearly visible effects of analysts’ pressures on C-level executives in the popular press, there is limited evidence on their effects on marketing spending decisions. This study asks two questions. First, how do analysts’ pressures affect firms’ short-term marketing spending decisions? Based on a sample of 2706 firms during 1987–2009 compiled from Institutional Brokers Earning System, COMPUSTAT, and CRSP databases we find that firms cut marketing spending. Second, more importantly, we ask if firms which remained more committed in the past to marketing spending under analysts’ pressures have higher longer-term stock market performance. We find that the stock market performance of firms more committed to marketing spending under past periods of analysts’ pressures is higher. The findings are replicated for R&D spending and are robust across measures, controls, and methodologies. Consideration of two industry-based moderators, R&D spending and revenue growth, and one firm-based moderator, whether the firm is among the industry’s top four market share or other lower share firms, reveals that the findings are stronger for high R&D and growth industries and lower market share firms. One key implication is that top executives respond to analysts’ pressures by cutting marketing spending in the short term; however, if they can resist these pressures, longer-term stock market performance is higher.

“When all companies are quarterly earnings–obsessed, the market starts punishing companies that aren’t yielding an instant return. This not only creates a big incentive for bogus accounting, but also it inhibits the kind of investment that builds economic value.” Turner 2005

Introduction

Marketing departments struggle to justify obtaining resources from the firm, usually being the first to receive cuts when times get tough. This problem becomes harder on the marketing department as CEOs and CFOs experience pressures from analysts to meet earnings expectations. On the one hand, firms must continue to invest in order to meet or exceed customer expectations by providing quality products and services to established and new markets. On the other hand, analysts’ earnings expectations are a top priority for C-level executives responsible for mandating cuts in spending, partly because they have performance measures that are closely tied to meeting analysts’ expectations. The continuing need for investment in the firm in the presence of analysts’ pressures creates a debate on cutting versus maintaining marketing spending, and exacerbating this debate is the lack of knowledge on the extent and timing of market performance for each course of action.

Consequently, this study asks two key unique questions to inform the debate. First, do pressures from analysts’ earnings expectations affect firms’ marketing spending decisions in the short term? Analysts’ pressure is defined as the gap between the consensus of analysts’ earnings forecast and the company’s expected earnings. Expected earnings during a period are computed as the sum of earnings in the previous period and the estimated change in earnings during the period. Marketing spending is defined as the difference between the actual and expected marketing spending to sales ratio. Based on I/B/E/S, COMPUSTAT, and CRSP data on 2706 firms during 1987–2009, our first finding is that firms under analysts’ pressures cut marketing spending in the short term. The result is consistent with the accounting view that marketing is a discretionary expense readily cut under analysts’ pressures. The implication is that marketing spending is not just based on the evolving needs of markets, products, or services over time but on the evolving needs of top executives whose primary responsibility is to grow stock price and hence deal with analysts’ pressures. An analysis of potential moderating variables, why certain firms or industries are more prone to reduce marketing spending under analysts’ pressure, reveals that our first finding is driven by industries with higher R&D spending.

Second, we ask if firms with greater past commitment to marketing spending under analysts’ pressures have higher stock market performance in the longer term. This question is important because commitment to marketing spending could be overspending on the firm’s part; however, overspending on the firm’s part should not result in higher stock market performance. We employ a strategic variable: past commitment to marketing spending under pressure from analysts’ earnings expectations. Commitment to marketing spending under analysts’ pressures is defined based on the extent to which actual marketing spending is greater than expected during each past period when the firm was under analysts’ pressures. We focus on past periods of analysts’ pressures because these periods are the most difficult and tenuous periods in a firm’s history, with conflicting demands on spending versus cutting back. Such conflicting demands make it particularly difficult for firms to continue marketing spending to build assets such as brand equity, which can result in higher stock market performance. Our second finding is that firms more committed in the past to marketing spending under analysts’ pressures have higher stock market performance in the longer term, implying that marketing spending creates brand assets which improve stock market performance. An analysis of potential firm- and industry-based moderator variables reveals that this second finding is driven by industries with higher R&D and growth, and firms belonging to the industry’s lower market share firms.

The key managerial implication of our findings is straightforward: top executives indeed reduce short-term marketing spending in response to analysts’ pressures; however, if they can resist these pressures and stay the course, they enjoy higher longer-term stock performance. C-level executives need to be explicitly aware of the longer-term gains that accrue from commitment to marketing spending under analysts’ pressure. In many firms, top executives may not be aware of such gains. Our paper is aimed at the gap in awareness. In some firms, top executives may be aware of such losses but continue to tradeoff losses for a number of reasons such as short-term incentive-based compensation (Currim et al. 2012), private information (Stein 1989), Financial Accounting Standards Board’s (FASB) Generally Accepted Accounting Principles (GAAP),¹ or executives’ limited time horizons. How do such firms ensure that top executives under analysts’ pressures do not cut marketing spending in the short term, a decision which is counter-productive for the firm’s longer-term stock performance (Lodish and Mela 2007)? Why is this important? These questions are addressed under managerial implications in the final section of the paper.

Background

This section briefly reviews background literature in two steps. First, we (1) provide more information on analysts, who they are, and what they do and (2) review literature from accounting, finance, and strategy, because a few works in those literatures have directly explored how executives respond to analysts’ pressures. Second, we review the marketing literature because that literature has explored a few financial drivers (not analysts’ pressure) of marketing spending and the relationship between marketing spending and stock market return. Our main purpose is to clearly outline our contribution relative to other works in both literatures.

Literature in accounting, finance, and strategy

Financial analysts are specialists, employed by investment banks and brokerage firms, who issue research reports that reflect their understanding of the firm’s industry, strategy, and management quality, along with specific stock recommendations and earnings forecasts (Schipper 1991). There are two findings regarding why analysts’ forecasts result in pressure on managers. First, past literature shows an optimism bias in analysts’ forecasts and recommendations (Chopra 1998; Dreman and Berry 1995), which has been attributed to conflicts of interests, such as incentives to get or maintain underwriting businesses with the covered firm (Dechow et al. 2000; Dugar and Nathan 1995; Lin and McNichols 1998; Michaely and Womack 1999), or to generate sales and trading commission for the brokerage house (Cowen et al. 2006). Second, the optimism bias is exacerbated by a herding bias, due to career and reputation concerns. Inexperienced analysts who deviate from the consensus and miss the forecast are more likely to lose their jobs (Dechow et al. 2000). Consequently, analysts tend to release forecasts close to those previously announced by other analysts, even when this is not supported by their information or judgement (Trueman 1994). These optimism and herding biases increase the earnings pressure problem for managers (Zhang and Gimeno 2010).

Past research in accounting, finance, and strategy has explored four ways in which top executives respond to pressures from analysts’ earnings expectations: (1) by ignoring it (King 2004), (2) managing expectations (earnings guidance) by communicating more effectively with capital market agents (Bernhardt and Campello 2007), (3) engaging in “creative accounting,” such as capitalizing rather than expensing some costs (Burgstahler and Eames 2006; Degeorge et al. 1999) or (4) altering business decisions (e.g., cutting spending) to accommodate analysts’ pressures (Graham et al. 2005). Our study focuses on the fourth type of response to analysts’ earnings expectations.

In contrast to our work, other examples of making business decisions to accommodate analysts’ pressure have been based on limited samples, in a particular industry, or are limited to a particular product category or shorter time period. For example, in the strategy literature, Zhang and Gimeno (2010) find that analysts’ pressures encourage dominant firms in the U.S. electricity generation industry to exercise market power by restricting output and increasing prices to increase short-term earnings even if it undermines longer-term competitiveness. In the accounting literature, Cohen et al. (2010) find that firms reduce advertising spending to avoid losses and decrease in earnings, based on proprietary data on advertising spending gathered during 2001–2006 by monitoring multiple media channels. They classify “firms that are suspected to have managed earnings” as those which reported barely positive earnings, earnings growth, or earnings that just met consensus analysts’ forecast, an ex-post measure, in that their measure employs reported earnings subsequent (ex-post) to any business decision to identify suspect firms. Chapman and Steenburgh (2011) find that soup manufacturers expected to manage earnings upwards at year end during 1985–1988 employ more frequent temporary price promotions in two cities and shift promotions away from smaller revenue brands to larger revenue brands, even if it means sacrificing long-term value. In their study, an ex-post measure based on whether earnings targets were met in prior periods is employed. In contrast, our measure of analysts’ pressure is ex-ante, based on analysts’ pressure prior to the business decision, because we are interested in the effect of analysts’ pressure on marketing spending decisions and stock market return without using marketing spending or stock market return to identify suspect firms.

Furthermore, there are a couple of articles in the accounting literature on the effect of pressures from analysts’ expectations on R&D spending and investment decisions. Bushee (1998) studies the impact of institutional investors (rather than analysts) on R&D spending (rather than marketing spending) to meet short-term (rather than longer-term) earnings goals. Roychowdhury (2006) and Cohen and Zarowin (2010) study earnings management of discretionary expenditures, including R&D, to meet earnings targets and operating performance (rather than stock return), respectively. Marketing and R&D are conceptually different variables. In this study the correlation between the two spending variables is −0.06, which suggests that marketing spending is a separate and important variable for the marketing literature, i.e., we need to know managers’ behavior and the stock market’s reaction to their behavior.

In summary, while there are studies in the accounting, finance, and strategy literatures on the effects of analysts’ pressure, there is little literature on the effects of analysts’ pressures on business spending decisions (relative to engaging in creative accounting or communicating with capital market agents), and to the best of our knowledge, no literature on the effects of analysts’ pressures on marketing spending decisions based on a large number of firms, industries, and time period. Most importantly, no study focuses on whether the stock market performance of a firm with greater past commitment to marketing spending under pressure from analysts’ earnings expectations is higher or lower.

Literature in marketing

Selected works from the marketing–finance interface literature are presented in Table 1.

Table 1 Selected works from the marketing–finance interface literature

There are two types of studies based on the choice of the dependent variables (DV): (1) when the DV is spending on R&D (value creation) or marketing (value appropriation), and (2) when the DV is stock market performance (stock return, idiosyncratic risk, IPO underpricing and trading, firm value, etc.). The first type of study explores the drivers of R&D or marketing spending, while the second type of study explores the relationship between R&D or marketing spending and stock market performance. For example, the first type of study investigates how the equity to bonus ratio of top five executives, past stock returns and volatility, experience of CEOs, and marketing and R&D intensity drive R&D and marketing spending.

To the best of our knowledge there is very little if any work that explicitly considers our two research questions. To the extent that our results are indicative of myopic management of marketing resources, Mizik (2010) and Mizik and Jacobson (2007) address how myopic management of marketing spending, e.g., at the time of a seasoned equity offering (SEO), lowers stock market performance in the long run. Our two important findings contribute to the theory of myopic management by (1) considering a new antecedent variable other than SEO, i.e., analysts’ pressure, that is computed before marketing spending is realized, i.e., ex-ante (not ex-post), (2) introducing a new strategic variable, commitment to marketing spending during periods of analysts’ pressure, which is found to drive stock market performance, and (3) including all firms in the empirical analysis, not just firms with negative earnings surprises (Mizik 2010), or firms embarking on seasoned equity offerings (SEOs) (Mizik and Jacobson 2007). These differences are important because (1) analysts’ pressure is much more pervasive or frequent than SEOs and (2) the ex-ante variable (relative to the ex-post measure) is not susceptible to marketing spending manipulation during the period. In the additional analyses section, we show that the effect of analysts’ pressure on marketing spending is statistically significant even after we control for (1) SEOs and (2) other ex-post earnings pressure measures.

Hypotheses

In this section, we develop two hypotheses on the effects of (1) analysts’ pressures on marketing spending and (2) commitment to marketing spending through past periods of analysts’ pressures on stock market returns.

Effect of analysts’ earnings expectations on marketing spending

The hypothesis on the effect of analysts’ pressures on marketing spending is based on the simple notion and finding from accounting/finance research that managers are motivated to meet earnings expectations to avoid stock price declines and they may even do this by cutting spending in the short term in a way that compromises long-term performance. One reason is that the stock-based compensation of top five executives is based more on current/near-term stock prices (Currim et al. 2012).

When analysts’ pressures put the stock prices of firms at risk, top executives can employ a variety of cost reduction (e.g., cutting marketing, R&D, overhead maintenance, employee training programs, travel budgets, or overproduction to lower cost of goods sold) or revenue generating (e.g., selling assets or price discounts) strategies (Dechow and Sloan 1991; Roychowdhury 2006), all being possible ways to help boost earnings. Cost reduction or revenue generation strategies are important because decreases (increases) in earnings are associated with declines (increases) in price earnings multiples (Barth et al. 1999). In addition, firms that meet or beat current analysts’ earnings expectations enjoy a higher return than firms with earnings forecast errors that fail to meet expectations since investors know when analysts’ expectations are met (Bartov et al. 2002). However, cost reduction strategies, relative to revenue generation, have more certain short-term impacts. Mizik (2010) refers to this as “inter-temporal borrowing” of earnings or “myopic management” because the manager is increasing earnings in the short term by cutting marketing investments shown to have long-term effects (Srinivasan and Hanssens 2009), without necessarily letting investors know the tradeoff made (Dichev et al. 2012).

• H1: An increase in pressure from analysts’ earnings expectations will be associated with a current period decrease in firms’ marketing spending.

Effect of commitment to marketing spending under analysts’ pressures on stock market return

The second hypothesis suggests that greater commitment to marketing spending through past periods of pressures from analysts’ earnings expectations will be useful to build marketing assets over time and higher longer-term stock market performance. The hypothesis relies on studies in the marketing literature which link marketing assets to stock market returns and provide a rationale to support the effects of greater commitment to marketing spending through past periods of analysts’ pressures on future stock market returns. These studies focus on marketing assets such as perceived product quality (Aaker and Jacobson 1994), brand equity (Barth et al. 1998), brand attitude (Aaker and Jacobson 2001), customer lifetime value (e.g., Gupta et al. 2004), and customer satisfaction (e.g., see Gupta and Zeithaml 2006 for a review).

The theory is that these marketing assets protect the firm from price competition created by lower quality brands (Blattberg et al. 1995), because they reduce product substitutability (Mela et al. 1997), lower price sensitivity (Kaul and Wittink 1995), and increase customers’ willingness to pay a price premium (Ailawadi et al. 2003). These assets also increase the receptiveness and loyalty of consumers and distributors (Kaufman et al. 2006), help the firm enter new markets (Srivastava et al. 1998), up-sell and cross-sell customers (Kamakura et al. 2003), and protect pioneering firms by enabling a higher level of marketing effectiveness (Bowman and Gatignon 1996).

Consequently, greater commitment to marketing spending (advertising, customer service, etc.) through past periods of analysts’ pressures can create and reinforce marketing assets, which can improve financial performance (Barth et al. 1998). In contrast, lower commitment to marketing spending through past periods of analysts’ pressures can lead to price discounting (Neslin 2002) and brand erosion, which lowers reference price (Kalyanaram and Winer 1995) and future financial performance (Barth et al. 1998). This is a key reason why managers may want to resist cutting marketing spending under analysts’ pressures. What makes greater commitment to marketing spending under analysts’ pressures special is that it allows continuous reinforcement and maintenance of marketing assets even through the most challenging times during which there is pressure at the highest levels in a firm, i.e., the CEO and CFO, to reduce spending and allow assets to deteriorate, in order to avoid a lower short-term stock market return. Consequently:

• H2: Firms’ greater commitment to marketing spending through past periods of analysts’ pressures will be positively associated with current period stock market returns.

If H2 is found to be true, in order to achieve the higher stock market performance associated with commitment to marketing spending and asset creation (H2), managers will need to resist the short-term motivation to cut marketing spending when under analysts’ pressures (H1). In our context, the difference between analysts’ forecasts and the firm’s expected earnings is observed at the beginning of the year, which influences marketing spending during the year, and consequently affects firm performance at the end of the year (Fig. 1). There may be other performance based drivers of spending (H1), such as Market-to-Book, ROA, and liquidity, which will be employed as controls described in the model section to test H1.

Fig. 1

An example of the timeline of events

Basically, if H1 and H2 are supported, the story can be summarized as follows: Firms respond to analysts’ pressures by cutting short-term marketing spending (H1); however, if they can resist these pressures and stay the course with their marketing investments, they will enjoy higher longer-term stock performance (H2).

Moderation effects

As indicated earlier, we explore firm- and industry-based moderator variables to determine whether and why certain firms or industries (1) are more prone to reduce marketing spending under analysts’ pressure and (2) have higher stock market performance associated with commitment to marketing spending. For example, we explore industries with high versus low (1) R&D spending and (2) growth. We chose industry R&D spending and growth because industries with high R&D spending (and growth) are innovation oriented, e.g., focused on value creation through new products over value appropriation through marketing spending. Consequently, when under analysts’ earnings pressure, firms in these industries are more likely to decrease short-term marketing spending. However, it is also possible that high R&D spending and growth industries need to inform and persuade markets more frequently about their more frequent product introductions, which would suggest a greater importance of commitment to marketing spending in the longer term or association with stock market performance. We also explore firms that belong to an industry’s top four market share firms versus other lower market share firms. We chose high versus low market share firms because when under analysts’ pressure, lower market share firms, because of their follower status, may be more likely to cut short-term marketing spending than leading firms. However, it is also possible that lower market share firms may benefit more in the longer term by remaining committed to marketing spending, i.e., experience higher stock market performance, because the commitment to marketing spending builds brand assets in the longer term, relative to higher market share firms with established brand assets.

Model

We present the model in two steps: (1) the major variables involved in hypothesis testing, i.e., marketing spending, analysts’ pressure, commitment to marketing spending under past periods of analysts’ pressure, and stock market returns, followed by (2) the models for hypothesis testing.

Marketing spending

Marketing spending is defined as the difference between firm i’s actual marketing spending (M_it) and the normal (expected or predicted) marketing spending (${\widehat{\mathrm{M}}}_{\mathrm{it}}$) to sales ratio at time t, in order to consider the unexpected marketing spending during the period. This definition allows marketing spending to be benchmarked by what is predicted or expected during the time period based on observations of past spending and returns on assets. Following Anderson and Hsiao (1982), Mizik and Jacobson (2007), and Mizik (2010), the expected marketing spending is estimated using a time-series panel data model with two period lags:

$$ \begin{array}{c}{\mathrm{M}}_{\mathrm{it}}-{\overline{\mathrm{M}}}_{\mathrm{t}}={\updelta}_{0\mathrm{i}}+{\updelta}_1\left({\mathrm{M}}_{\mathrm{it}-1}-{\overline{\mathrm{M}}}_{\mathrm{t}-1}\right)+{\updelta}_2\left({\mathrm{M}}_{\mathrm{it}-2}-{\overline{\mathrm{M}}}_{\mathrm{t}-2}\right)+{\updelta}_3\left({\mathrm{ROA}}_{\mathrm{it}-1}-{\overline{\mathrm{ROA}}}_{\mathrm{t}-1}\right)\\ {}+{\updelta}_4\left({\mathrm{ROA}}_{\mathrm{it}-2}-{\overline{\mathrm{ROA}}}_{\mathrm{t}-2}\right)+{\mu}_{\mathrm{it}}\end{array} $$

(1)

where M_it = Marketing spending to sales ratio of firm i during year t where marketing spending is defined as SG&A excluding R&D, ${\overline{\mathrm{M}}}_{\mathrm{t}}=$ Mean for the M_itseries during year t to adjust for the time effect of marketing spending, ROA_it = Return on assets of firm i during year t, and ${\overline{\mathrm{ROA}}}_{\mathrm{t}}=$ Mean for the ROA_it series during year t. We then predict ${\overset{\frown }{\mathrm{M}}}_{\mathrm{it}}$ using the estimated coefficients in the model above.

$$ \begin{array}{c}{\overset{\frown }{\mathrm{M}}}_{\mathrm{it}}-{\overline{\mathrm{M}}}_{\mathrm{t}}={\widehat{\updelta}}_{0\mathrm{i}}+{\widehat{\updelta}}_1\left({\mathrm{M}}_{\mathrm{it}-1}-{\overline{\mathrm{M}}}_{\mathrm{t}-1}\right)+{\widehat{\updelta}}_2\left({\mathrm{M}}_{\mathrm{it}-2}-{\overline{\mathrm{M}}}_{\mathrm{t}-2}\right)+{\widehat{\updelta}}_3\left({\mathrm{ROA}}_{\mathrm{it}-1}-{\overline{\mathrm{ROA}}}_{\mathrm{t}-1}\right)\\ {}+{\widehat{\updelta}}_4\left({\mathrm{ROA}}_{\mathrm{it}-2}-{\overline{\mathrm{ROA}}}_{\mathrm{t}-2}\right)\ \end{array} $$

(2)

Finally, we compute ${\mathrm{M}}_{\mathrm{it}}-{\overset{\frown }{\mathrm{M}}}_{\mathrm{it}}$ to determine the unexpected marketing spending to sales ratio or whether the firm increased or decreased marketing spending relative to the expected marketing spending to sales ratio. Although SG&A spending has limitations, two primary advantages over advertising spending are that SG&A spending (1) is reported more frequently than advertising spending and (2) includes other promotion or commercialization effects, e.g., direct sales, distribution, market research, trade promotions, and related activities, which are important in industries where commercialization is primarily accomplished through means other than advertising (Brower and Mahajan 2013). Consequently, while the SG&A–R&D measure contains some non-marketing spending, it contains more types of marketing spending than advertising does. Table 1 shows there is precedence in the literature for the use of SG&A–R&D based marketing spending measures.

Analysts’ pressure

We define analysts’ pressure, AP_it, following the strategy literature, as the gap between consensus of analysts’ earnings forecast and the company’s expected earnings (Zhang and Gimeno 2010). The firm’s expected earnings are not a simple extrapolation of past earnings but based on earnings in the past period as well as recent stock returns, which includes all new market information such as changes in the external environment (e.g., demand) and internal operations (e.g., new products). Specifically, following Zhang and Gimeno (2010), we estimate the analysts’ pressure that firm i faces at the beginning of year t as the difference between a consensus of analysts’ earnings per share (EPS hereafter) forecasts for year t, F_it, and the estimate of its expected earnings, E[EPS_it], standardized by its stock price at the end of year t-1:

$$ {\mathrm{AP}}_{\mathrm{it}}=\left({\mathrm{F}}_{\mathrm{it}}-\mathrm{E}\left[{\mathrm{EPS}}_{\mathrm{it}}\right]\right)/{\mathrm{Stock}\ \mathrm{Price}}_{\mathrm{it}-1} $$

(3)

We adjust earnings pressure by the firm’s stock price at the beginning of the year, following Skinner and Sloan (2002). The main reason for this adjustment is that, although earnings are comparable across firms, EPS are not comparable because firms vary in the number of shares outstanding. In other words, the number of shares outstanding is decided by each firm because companies differ on their tendency to split stocks to make shares more affordable, hence EPS varies across firms. For example, a $1 EPS gap (between the consensus of analysts’ EPS forecast and the company’s expected earnings) for Facebook is large (latest EPS = $3.47, outstanding shares = 2.35B), but will be very small for Berkshire Hathaway (latest EPS = $14,155, outstanding shares = 784,669). Dividing by stock price (per share) provides an adjustment to earnings pressure as a percentage of stock price, which enables comparisons across firms. The reason for employing a lagged value of stock price (t-1) instead of a concurrent stock price (t) is as follows. Analysts’ pressure at time t is based on expected EPS at time t which in turn is dependent on realized EPS at t-1 and the expected change in EPS from t-1 to t. Therefore, we scale the numerator by Stock Price at t-1 in the denominator. Expected earnings E[EPS_it] is calculated as the sum of the firm’s earnings in year t-1 and the estimated change of its earnings in year t:

$$ \mathrm{E}\left[{\mathrm{EPS}}_{\mathrm{it}}\right]={\mathrm{EPS}}_{\mathrm{it}-1}+\mathrm{E}\left[\Delta {\mathrm{EPS}}_{\mathrm{it}}\right] $$

(4)

Our operationalization is identical to Zhang and Gimeno’s (2010) method, adopted from Matsumoto (2002) in the accounting literature, to estimate the change of a firm’s earnings in year t. In particular, we use the following regression to estimate the changes of EPS using all other firms in the same industry in the prior year:

$$ \Delta {\mathrm{EPS}}_{\mathrm{it}}/{\mathrm{P}}_{\mathrm{it}-1}={\upalpha}_{0\mathrm{jt}}+{\upalpha}_{1\mathrm{jt}}\left(\Delta {\mathrm{EPS}}_{\mathrm{it}-1}/{\mathrm{P}}_{\mathrm{it}-2}\right)+{\upalpha}_{2\mathrm{jt}}{\mathrm{CRET}}_{\mathrm{it}-1}+{\uptheta}_{\mathrm{it}} $$

(5)

where ∆EPS_it = EPS for firm i in year t minus EPS for the same firm in year t-1; P_it = price per share for firm i at the end of year t; and CRET_it = cumulative daily excess return for firm i in year t obtained from CRSP. Returns are cumulated from 3 days after the last yearly earnings announcement to 5 days prior to the last round of analysts’ earnings forecasts made before year t starts. The specification of the equation above adjusts differential earnings gaps by stock price (high versus low) to allow comparisons of differential earnings gaps across firms (growth versus value). The rationale is that growth firms have higher earnings expectations but also higher stock prices relative to their EPS bases. Subsequently, we use the parameter estimates from the prior firm-year to calculate the expected changes in EPS at the beginning of year t by:

$$ \mathrm{E}\left[\Delta {\mathrm{EPS}}_{\mathrm{it}}\right]=\left[{\widehat{\upalpha}}_{0\mathrm{t}-1}+{\widehat{\upalpha}}_{1\mathrm{t}-1}\left(\Delta {\mathrm{EPS}}_{\mathrm{it}-1}/{\mathrm{P}}_{\mathrm{it}-2}\right)+{\widehat{\upalpha}}_{2\mathrm{t}-1}{\mathrm{CRET}}_{\mathrm{it}-1}\right]\times {\mathrm{P}}_{\mathrm{it}-1} $$

(6)

This measure of analysts’ pressure is different from the one used by Payne and Robb (2000). They use the gap between earnings forecast consensus and pre-managed earnings, where pre-managed earnings are calculated as reported earnings minus discretionary accruals (Jones 1991), to measure pressure to meet earnings forecast consensus. Although the Payne and Robb (2000) measure is appropriate to identify accounting manipulations to meet earnings expectations, it is less applicable in our context. Our focus is to determine whether and how analysts’ pressure at the beginning of the period (i.e., ex-ante) subsequently impacts marketing spending during the period. Their measure can only be calculated after earnings are announced (i.e., ex-post), and only capture analysts’ pressure that could not be met through changes in marketing spending or other business decisions. Our measure of analysts’ pressure is also different from other ex-post measures employed by Mizik (2010) and Chapman and Steenburgh (2011) to motivate earnings manipulation, i.e., (1) whether and how much EPS_t is lower than analysts’ consensus forecast in period t, (2) whether and how much EPS_t-1 is lower than EPS_t-2, and (3) whether and how much EPS_t-1 is lower than analysts’ consensus forecast in period t-1. The correlation between our and their two measures is about 0.2. In our analysis and additional analysis sections we show that our results hold when we do not, as well as when we do control for these other pressure measures, respectively.

Commitment to marketing spending under past periods of analysts’ pressure

MCmt_it is past commitment to marketing spending under analysts’ pressure and defined as follows:

$$ \begin{array}{l}{\mathrm{M}\mathrm{Cmt}}_{\mathrm{it}}=\sum_{\mathrm{p}=0}^{\mathrm{P}}{\mathrm{d}}_{\mathrm{it}-\mathrm{p}}{\mathrm{AP}}_{\mathrm{it}-\mathrm{p}}\left({\mathrm{M}}_{\mathrm{it}-\mathrm{p}}-{\widehat{\mathrm{M}}}_{\mathrm{it}-\mathrm{p}}\right){\mathrm{e}}^{-\mathrm{p}}\kern0.5em \hfill \\ {}\mathrm{where}\ {\mathrm{d}}_{\mathrm{it}-\mathrm{p}}=\left\{\begin{array}{c}\hfill 1\kern0.5em \mathrm{if}\ \left({\mathrm{M}}_{\mathrm{it}-\mathrm{p}}-{\widehat{\mathrm{M}}}_{\mathrm{it}-\mathrm{p}}\right)\ge 0\ \mathrm{and}\ {\mathrm{AP}}_{\mathrm{it}-\mathrm{p}}>0\hfill \\ {}\hfill 0\kern0.5em \mathrm{otherwise}\kern9.75em \hfill \end{array}\right.\kern0.5em \hfill \end{array} $$

(7)

where p is defined as all past periods in the firm’s history during which it demonstrates commitment to marketing spending (M_it- ${\widehat{\mathrm{M}}}_{\mathrm{it}}$) ≥ 0 under analysts’ pressure (AP_it > 0).² In such periods, d_it-p will take on a value of 1, and 0 otherwise. For each of the past periods during which the firm is under analysts’ pressure and actual marketing spending is higher than expected, we consider the magnitude of unexpected marketing spending, i.e., deviation of actual marketing spending from the expected marketing spending to sales ratio (Mizik 2010; Mizik and Jacobson 2007), and the magnitude of analysts’ pressure (Matsumoto 2002; Zhang and Gimeno 2010), and accumulate the resulting commitment over past periods in the firm’s history. We take the product of unexpected marketing spending and analysts’ pressures because a higher level of marketing spending relative to what is expected during the time period, in the presence of a higher level of analysts’ pressure represent a higher level of the firm’s commitment to marketing spending.³ We employ discounting because past research has indicated that the benefit of accumulated experience and investments to organizations may decay over time (Argote et al. 1990). We assumed exponential (1/exp.(p), where p = 1 to 3, 5, 10, or all past years), linear (1/p), and square discounting (1/p²), however, we note in the results section that the results of testing H2 are not sensitive to different ways and time horizons of discounting.⁴

Compounded unexpected stock return

CAR, the dependent variable, is the compounded unexpected stock return defined following Mizik (2010) as follows:

$$ {\mathrm{CAR}}_{\mathrm{it}}= \log {\prod}_{\mathrm{m}=1}^{12}\left[1+\left\{{\mathrm{Ret}}_{\mathrm{im}}-\mathrm{E}\left({\mathrm{Ret}}_{\mathrm{im}}\right)\right\}\right] $$

(8)

where $\mathrm{E}\left({\mathrm{Ret}}_{\mathrm{im}}\right)={\widehat{\varsigma}}_{1\mathrm{i}}\left({\mathrm{Ret}}_{\mathrm{m}\mathrm{arket},\mathrm{m}}-{\mathrm{Ret}}_{\mathrm{risk}\hbox{-} \mathrm{free},\mathrm{m}}\right)+{\widehat{\varsigma}}_{2\mathrm{i}}{\mathrm{SB}}_{\mathrm{m}}+{\widehat{\varsigma}}_{3\mathrm{i}}{\mathrm{HL}}_{\mathrm{m}}+{\widehat{\varsigma}}_{4\mathrm{i}}{\mathrm{MO}}_{\mathrm{m}}$. (Ret_market,m − Ret_{risk‐free,m}) is the risk-free market return; SB_m is the difference between the return on a value-weighted portfolio of small stocks and the return on a value-weighted portfolio of big stocks; HL_m is the difference between a value-weighted portfolio of high book-to-market stocks and the return on a value-weighted portfolio of low book-to-market stocks; MO_m is the momentum factor, the difference between the average return on the two (small and large size) high prior-return portfolios and the average return on the two (small and large size) low prior-return portfolios computed in month m; which are available from Kenneth French’s data library posted on his Web site. ς _1i ,  ς _2i ,  ς _3i ,  and ς _4i are generated from estimating the Fama and French three-factor model (Fama and French 1992, 1996) augmented with the momentum factor as in the Carhart model (Carhart 1997) for each firm i:

$$ {\mathrm{Ret}}_{\mathrm{im}}-{\mathrm{Ret}}_{\mathrm{risk}\hbox{-} \mathrm{free},\mathrm{m}}={\varsigma}_{0\mathrm{i}}+{\varsigma}_{1\mathrm{i}}\left({\mathrm{Ret}}_{\mathrm{m}\mathrm{arket},\mathrm{m}}-{\mathrm{Ret}}_{\mathrm{risk}\hbox{-} \mathrm{free},\mathrm{m}}\right)+{\varsigma}_{2\mathrm{i}}{\mathrm{SB}}_{\mathrm{m}}+{\varsigma}_{3\mathrm{i}}{\mathrm{HL}}_{\mathrm{m}}+{\varsigma}_{4\mathrm{i}}{\mathrm{MO}}_{\mathrm{m}}+{\upvarphi}_{\mathrm{im}} $$

CAR_it at t = 0 or CAR₀ is defined as current period abnormal returns, i.e., summed over months within a year. As a robustness check, we use (1) compounded or total stock returns in the current year (TSR₀) (Mizik 2010; Mizik and Jacobson 2007) and (2) unexpected stock returns suggested by Barber and Lyon (1997),⁵ CTSR₀ in the current year. This measure employs, for each sample firm, a control firm in the same industry with close market values and book-to-market ratios, enabling the computation of unexpected returns based on the difference between the stock returns for the sample firm versus the matched firm (Mizik and Jacobson 2007). Note that TSR are raw returns in the current year hence firm and time fixed effects are included when testing H2 while CTSR are returns relative to matched firms in the same year consequently time effects are not included. Firm effects are included to allow for imperfections in matching. If matching is perfect firm effects will be insignificant.

Model to test H1

The model to test H1 is:

$$ {\mathrm{M}}_{\mathrm{it}}-{\widehat{\mathrm{M}}}_{\mathrm{it}}={\upbeta}_0+{\upbeta}_1{\mathrm{AP}}_{\mathrm{it}-1}+{\upbeta}_2{\mathrm{M}\mathrm{TB}}_{\mathrm{it}-1}+{\upbeta}_3{\mathrm{SLACK}}_{\mathrm{it}-1}+{\upbeta}_4{\mathrm{ROA}}_{\mathrm{it}-1}+{\upbeta}_{5\mathrm{t}}{\mathrm{YEAR}}_{\mathrm{t}}+{\upvarepsilon}_{\mathrm{it}} $$

(9)

For H1 to be supported, β₁ < 0. Earlier we provided details on the definition and estimation of the main dependent and independent variables, marketing spending (${\mathrm{M}}_{\mathrm{it}}-{\widehat{\mathrm{M}}}_{\mathrm{it}}\kern-0.25em$) and analysts’ pressure (AP_it-1), respectively. Note that marketing spending is defined as actual marketing spending relative to predicted spending in a period. We looked at several finance/accounting and marketing papers to decide on the main controls. Consequently, while controls selected have precedence in the literatures, their inclusion is also supported by the following qualitative reasoning. When attempting to determine the effect of analysts’ pressure on unexpected or abnormal marketing spending, one must control for other factors that may drive unexpected marketing spending. One such factor is the firm’s resources measured either by previous period returns, e.g., ROA in the financial market (Hubbard 1998) or firm liquidity or cash on hand, e.g., SLACK (Hubbard 1998). Another factor which can drive unexpected marketing spending is the growth opportunity of the firm, e.g., Market to Book (MTB) (Kaplan and Zingales 1997). Finally we include time fixed effects (YEAR) which control for other variables related to time such as the economy, competition, etc. MTB is calculated as the firm’s market value divided by total book value of assets. SLACK is calculated as 1 minus the ratio of the firm’s total book value of long-term debt-to-equity. ROA is calculated as the firm’s net income divided by total book value of assets. These three control variables are lagged by one period because decisions in the current period are based on outcomes and opportunities observed in the prior period.

Model to test H2

The model to test H2 is:

$$ {\mathrm{CAR}}_{\mathrm{it}}={\upbeta}_0^{\prime }+{\upbeta}_1^{\prime }{\mathrm{M}\mathrm{Cmt}}_{\mathrm{it}}+{\upbeta}_2^{\prime }\ \sum_{\mathrm{p}=1}^{\mathrm{P}}{\mathrm{AP}}_{\mathrm{it}\hbox{-} \mathrm{p}}+{\upbeta}_3^{\prime }\ \sum_{\mathrm{p}=1}^{\mathrm{P}}\left({\mathrm{M}}_{\mathrm{it}\hbox{-} \mathrm{p}}-{\widehat{\mathrm{M}}}_{\mathrm{it}\hbox{-} \mathrm{p}}\right)+{\upbeta}_{4\mathrm{t}}^{\prime }{\mathrm{YEAR}}_{\mathrm{t}}+{\upvarepsilon}_{\mathrm{it}}^{\prime } $$

(10)

For H2 to be supported, ${\upbeta}_1^{\hbox{'}}$> 0 where ${\upbeta}_1^{\hbox{'}}$ is higher performance (if ${\upbeta}_1^{\hbox{'}}$> 0) or lower performance (if ${\upbeta}_1^{\hbox{'}}$< 0) for commitment to marketing spending. Note that CAR defined in Eq. 8 is relative to the firm’s predicted return in the period (not observed return in the previous period) which accounts for firm but not time fixed effects. Therefore, time fixed effects (YEAR) were included in Eq. 10. MCmt defined in Eq. 7 involves an interaction effect of abnormal marketing spending and analyst pressure accumulated over time. Consequently, the corresponding main effects of analyst pressure accumulated over time,$\sum_{\mathrm{p}=1}^{\mathrm{P}}{\mathrm{AP}}_{\mathrm{it}\hbox{-} \mathrm{p}}$, and marketing spending accumulated over time,$\sum_{\mathrm{p}=1}^{\mathrm{P}}\left({\mathrm{M}}_{\mathrm{it}\hbox{-} \mathrm{p}}-{\widehat{\mathrm{M}}}_{\mathrm{it}\hbox{-} \mathrm{p}}\right)$, are included in Eq. 10. Hence, market spending levels are included as differenced from predicted spending in the main and interaction explanatory variables in Eq. 10.

Empirical test

Sample

Our sample was compiled from Institutional Brokers Earning System (I/B/E/S), COMPUSTAT, and CRSP databases. The I/B/E/S database provides institutional analysts’ expectations of several financial performance variables for U.S. publicly traded companies and summary statistics (e.g., mean, median, standard deviation) of institutional analysts’ annual expectations for each company. Standard & Poor’s COMPUSTAT database comprises financial information for all U.S. publicly traded companies, including marketing spending. CRSP’s database maintains the stock price, return, and volume data for the NYSE, AMEX, and NASDAQ stock markets. We combined the I/B/E/S, COMPUSTAT, and CRSP databases by year employing the CUSIP/GVKEY match assigned to each firm. We combined the databases by year because firm budgetary decisions including marketing budgets are decided annually (Mizik 2010; Mizik and Jacobson 2007), although we tested the hypotheses employing quarterly data as well to check the robustness of the results under the additional analysis section.⁶ We built our sample from all companies that were in the COMPUSTAT database during the period of 1987–2009. We begin with 1987 because of the sparse observations in the I/B/E/S database before 1987. For each period or year in the annual data sets from COMPUSTAT we consider the analysts’ pressure closest to the beginning of the year.

Results

Descriptive statistics of our sample are presented in Table 2. In Panel A, 1987–2009, we observe that on average, 58% of firms are under analysts’ pressures, and 29% of firms cut marketing spending when under analysts’ pressures. These observations reveal that analysts’ pressure is a pervasive phenomenon, and that there is good variance in the sample on firm marketing spending behavior when firms are under analysts’ pressures.⁷ In Panel B, we provide the definitions, means, and standard deviations for our dependent and independent variables including the sources of the variables. Equation 1 on unexpected marketing spending is estimated in order to test H1. That estimation involved 17,969 observations from 2415 firms, with a mean number of 6 observations per firm. A frequency distribution of the number of firm-year observations employed for the prediction of unexpected marketing spending is provided in Online Appendix Table 1. A correlation matrix of all variables employed in the corresponding models is provided in Online Appendix Table 2. The correlation between AP and control variables in Eq. 9 to test H1 was found to be less than 0.2. The correlation between MCmt and other independent variables in Eq. 10 to test H2 was found to be less than 0.2. Consequently, collinearity concerns are mitigated.

Table 2 Descriptive statistics

Next, we describe results of models testing H1 (Eq. 9) and H2 (Eq. 10). When estimating each of the two models to test H1 and H2, we dropped missing values. In other words, we estimated models based on the maximum sample available to maximize the quality of the parameter estimates employed to test H1 and H2. First, we describe results of testing H1 (Table 3). As hypothesized in H1, an increase in pressure from analysts’ earnings expectations is found to be associated with a current year decrease in marketing spending relative to the expected marketing spending (p < .01). Consequently, H1 is supported.⁸ Regarding the effect size, Online Appendix Fig. 1 Panel A shows that when the average earnings pressure doubles from 0.03 to 0.06, unexpected marketing spending declines from −6.4% to −6.8% of total sales, or from −27.8% to −29.5% of total marketing spending. H1 is supported when we consider the marketing spending to asset ratio as well as the marketing spending to sales ratio. Regarding controls utilized, a firm’s MTB ratio, SLACK, and ROA are found to be positively associated with unexpected marketing spending (all p < .01), indicating that its growth potential as judged by the capital market, its ability to spend as measured by its cash position, and its financial returns, all result in a higher than expected marketing spending to sales ratio.

Table 3 Effect of pressure from analysts’ earnings expectations on marketing spending to sales ratio decisions

Second, to test H2, we report results in Table 4 for compounded unexpected stock return CAR_t (Eq. 10) for t = 0. The model is estimated using a panel regression approach. In addition, we report results for total stock return, TSR_t, t = 0, and unexpected stock return (Barber and Lyon 1997), CTSR_t, t = 0, to test whether the finding based on CAR_t is robust to different measures of stock market return. As hypothesized in H2, commitment to marketing spending through past periods of pressure from analysts’ earnings expectations is found to be positively associated with stock market returns in the current period. Consequently H2 is supported. Regarding the effect size, Online Appendix Fig. 1 Panel B shows that when average marketing commitment doubles from 0.03 to 0.06, compounded abnormal return increases from −4.68% to −4.65%. We used the Hausman-Wu endogeneity test (Baum et al. 2003) to test whether marketing commitment is independent from remaining contemporaneous errors. We implemented the test using instruments that are lagged one period beyond the error term. The F-statistic was not significant (F_1,13,113 = 1.76, p > 0.1). This indicates that marketing commitment is not correlated with remaining contemporaneous errors, and consequently we do not need instruments to control for endogeneity.

Table 4 Relationship between commitment to marketing spending to sales ratio under analysts’ earnings pressures and stock market return

Moderation effects

We explored the following industry based moderation effects that might strengthen or weaken the relationships we find in our analysis: (1) high versus low R&D industries (based on a median cut on R&D spending as a ratio of sales across industries), which reveals that our results for H1 (the short-term effect of analysts’ pressure on marketing spending) and H2 (the longer-term effect of commitment to marketing spending under analysts’ pressure on stock market performance) are found to be driven by high R&D industries (Tables 5 and 6 respectively),⁹ and (2) high versus low industry growth (based on a median split of 5 year compounded industry revenue growth), which reveals that while our results for H1 are observed for both high and low growth industries (Table 7), our results for H2 are driven by high growth industries (TSR and CTSR, Table 8). We also explored a firm-based moderator, whether the firm belongs to the industry’s top four market share (based on the four-digit industry SIC code) or other lower market share firms, and found that while H1 is supported for both types of firms (Table 9),¹⁰ the results on H2 are driven by lower market share firms (Table 10).

Table 5 Moderation effect of high versus low industry R&D spending to sales ratio on H1 effect of analysts’ pressure on marketing spending to sales ratio decisions

Table 6 Moderation effect of high versus low industry R&D spending to sales ratio on H2 relationship between commitment to marketing spending to sales ratio under analysts’ pressures and stock market return

Table 7 Moderation effect of high versus low industry growth on H1 effect of analysts’ pressure on marketing spending to sales ratio decisions

Table 8 Moderation effect of high versus low industry growth on H2 relationship between commitment to marketing spending to sales ratio under analysts’ pressures and stock market return

Table 9 Moderation effect of top 4 market share vs. lower market share firms in the industry on H1 effect of pressure from analysts’ earnings expectations on marketing spending to sales ratio decisions

Table 10 Moderation effect of top 4 market share vs. lower market share firms in an industry on H2 relationship between commitment to marketing spending to sales ratio under analysts’ earnings pressures and stock market return

Additional analyses

We conducted a number of additional analyses aimed at testing the robustness of the results described in Tables 3 and 4 earlier. First, a commonly used approach in long-run event studies is the calendar time portfolio (CTP) approach developed by Jaffe (1974) and Mandelker (1974), and recommended by Fama (1998). For each calendar year we computed the return of an equally weighted portfolio of companies that have positive accumulated marketing commitment in the last year. The return of this “commitment portfolio” is denoted Ret_um, where u indicates that the portfolio consists of companies with positive accumulated marketing commitment and m denotes the calendar month. The Fama-French-Carhart four factor model (Carhart 1997; Fama and French 1992, 1996) is used to compute the unexpected return of this portfolio:

$$ {\mathrm{Ret}}_{\mathrm{um}}-{\mathrm{Ret}}_{\mathrm{risk}\hbox{-} \mathrm{free},\mathrm{m}}={\varsigma}_{0\mathrm{u}}^{\prime }+{\varsigma}_{1\mathrm{u}}^{\prime}\left({\mathrm{Ret}}_{\mathrm{m}\mathrm{arket},\mathrm{m}}-{\mathrm{Ret}}_{\mathrm{risk}\hbox{-} \mathrm{free},\mathrm{m}}\right)+{\varsigma}_{2\mathrm{u}}^{\prime }{\mathrm{SB}}_{\mathrm{m}}+{\varsigma}_{3\mathrm{u}}^{\prime }{\mathrm{HL}}_{\mathrm{m}}+{\varsigma}_{4\mathrm{u}}^{\prime }{\mathrm{MO}}_{\mathrm{m}}+{\upvarphi}_{\mathrm{um}}^{\prime } $$

where definitions of all independent variables are identical to those in the earlier section on Compounded Unexpected Stock Return. Assuming that the broad-market return and the Fama-French-Carhart four factors adequately describe average returns, the parameter of interest, ${\varsigma}_{0\mathrm{u}}^{\hbox{'}}$, can be interpreted as the average unexpected return associated with holding this simulated portfolio. The results show that for the following 12 months, the average monthly unexpected return, ${\varsigma}_{0\mathrm{u}}^{\hbox{'}}$, turns out as 0.2% (p < 0.001). This suggests that the positive effects of the marketing commitment variable found in Table 4 are also robust with the alternative approach of calendar time portfolio returns.

Second, while the main results for H1 and H2 are for the marketing spending to sales ratio, we conducted similar tests for H1 and H2 employing the marketing spending to assets ratio (Mizik 2010; Mizik and Jacobson 2007). H1 is supported (p < .05); H2 is supported for CAR, TSR, and CTSR (each p < .01).

Third, we tested H1 and H2 for the R&D spending to sales ratio. R&D spending is another important decision, a commitment to which can lead to assets such as innovative products, brand building and stock market return. H1 is supported (p < .01); H2 is supported for CAR and TSR (p < .01).

Fourth, in addition to commitment to marketing spending under analysts’ pressures there are three other strategic options: (1) commitment when not under pressure, (2) lack of commitment under pressure, and (3) lack of commitment when not under pressure. We constructed one variable for each of these three additional options in the model to test for the robustness of the H2 result and assess the effect of each additional option on stock return. The coefficients of the four commitment options indicate that greater commitment under pressure generates the largest positive returns for CAR and TSR (p < .01). Greater commitment under no pressure generates lower CAR and TSR (p < .01). Finally, a lack of commitment when not under pressure results in negative CAR (p < .01).

Fifth, we repeated the model estimations for testing H2 (Table 4) using additional controls for whether earnings were met or not. This is accomplished by adding three control variables: (1) earnings met with marketing spending cuts, (2) earnings met without marketing spending cuts, and (3) earnings not met with marketing spending cuts, all relative to the base level of earnings not met without a marketing spending cut. The results of testing H2 including the additional three controls are consistent with those reported in Table 4 without the three controls. Consequently, employing additional controls for whether earnings are met or not is found to reinforce the results described in the previous section.

Sixth, we computed the commitment measure employing an initialization period comprising 20% of the data. This initialization period is intended to overcome a potential weakness in measuring the commitment variable in earlier periods when there are not enough observations to compute the variable. The corresponding results when compared to Table 4 in the paper are remarkably similar, i.e., the coefficients of marketing commitment for CAR, TSR, and CTSR are positive (p < .05).

Seventh, we re-estimated models for quarterly data and found that H1 is also supported for quarterly data, i.e., the coefficient of AP_it is negative (p < .01). The results for H2 are similar for CAR (p < .05). Basically, there is overlap between the results of annual and quarterly data.

Eight, since R&D spending has been shown to be influenced by analysts’ pressures, we re-estimated the models controlling for R&D spending and found that the results of H2 are similar to those reported in Table 4 for CAR and TSR (both p < .05). Consequently, controlling for R&D spending reinforces the results described earlier which did not control for R&D spending.

Ninth, to investigate whether the results on H1 and H2 would be affected by error in the models on (1) marketing spending and (2) analysts’ pressures, we re-estimated the models to classify marketing spending (consensus of analysts’ EPS expectation) as being increased over predicted spending (the estimate of expected earnings) when marketing spending (consensus earnings expectation) was not just greater than predicted spending (estimate of expected earnings) but one standard deviation greater than predicted spending (estimate of expected earnings). The results of H1 are similar to those reported in Table 3, i.e., the coefficient of AP is negative (p < .01). The results for H2 are similar to those reported in Table 4, i.e., the coefficient of MCmt for CAR is positive (p < .05). These results indicate that small differences between actual and predicted marketing spending or between consensus of analysts’ earnings estimates and estimates of expected earnings are not driving the results reported earlier.

Tenth, we tested the robustness of our result for H1, controlling for (1) SEOs (Mizik and Jacobson 2007) and (2) other measures of earnings pressures in the literature, such as the measures employed by Mizik (2010) and Chapman and Steenburgh (2011): (a) whether and how much EPS_t is lower than analysts’ consensus forecast in period t, (b) whether and how much EPS_t-1 is lower than EPS_t-2 and (c) whether and how much EPS_t-1 is lower than analysts’ consensus forecast in period t-1. The results on H1 reported earlier in Table 3 hold when we control for SEOs and these other measures of earnings pressures.¹¹

Finally, we checked the results of H1 and H2 reported in Tables 3 and 4 respectively using (1) clustered standard errors by firm, and (2) (a) estimating Newey–West standard errors with first-order autocorrelation error structure for coefficients estimated by OLS (for CAR0) and (b) cross-sectional time-series regression models with first-order autoregressive disturbance terms (for TSR0 and CTSR0). We found that the results reported in Tables 3 and 4 are robust. The coefficient of analyst pressure is negative and significant (p < .01), which supports H1 for both tests. The coefficient of marketing commitment for CAR, TSR, and CTSR is positive and significant at p < .1, p < .01, and p < .05, respectively, which supports H2 for TSR and CTSR for clustered errors and p < .1 for all measures employing the autocorrelation correction.

Discussion

This study makes two theoretical contributions that have significant implications for top executives. First, we consider a new antecedent variable, pressure from analysts’ earnings expectations, which potentially drives firms’ marketing spending. Not a day goes by without stories in the popular financial press about differences between firms’ realized earnings and analysts’ expectations, with CEOs or CFOs commenting on the differences. While there is (1) accounting and financial theory which is equivocal about whether or not analysts’ earnings expectations drive general management decisions (Brown and Caylor 2005; Jensen 2005; Porter 1992; Stein 1989), (2) substantial anecdotal and some experimental- or survey-based evidence that analysts’ pressures impact top executive decisions (Graham et al. 2005), (3) some empirical literature in accounting, finance, and strategy on the effects of analysts’ pressures on general management and R&D decisions based on data from few firms in an industry over a limited time period (reviewed in the background section), and (4) limited empirical evidence about the effect of earnings pressures (e.g., from SEOs) on marketing spending (also reviewed in the background section), our main goal is to investigate and establish the short-term effect of analysts’ pressures on firms’ marketing spending decisions based on data on a large number of firms, industries, and time period. While earnings pressures from SEOs are important, analysts’ pressures resulting from earnings expectations are felt more frequently by larger number of firms across industries. We find that pressures from analysts’ earnings expectations are pervasive across the sample we study and, based on empirical models with extensive controls, are able to establish that marketing spending in the current year is negatively impacted by analysts’ pressures (H1). This effect is consistent with the view that marketing is viewed by top executives as a discretionary expense which may create intangible assets, but readily cut in times of earnings pressure (Currim et al. 2012). An analysis of firm and industry based moderator variables reveals that this result is driven by high R&D industries.

Second, we investigate whether firms’ past commitment to marketing spending under analysts’ pressures, a strategic variable, is associated with longer-term stock market performance. Based on an empirical model and multiple measures of stock market return, we are able to establish that the stock market performance of firms with greater past commitment to marketing spending under analysts’ pressures is higher (H2). In contrast to the first result, this second result demonstrates the longer-term financial value of commitment to marketing spending under analysts’ pressure, an important step in establishing marketing’s accountability, in particular for CEOs and CFOs accustomed to thinking about marketing spending as discretionary and cutting marketing spending under analysts’ pressure. An analysis of firm- and industry-based moderator variables reveals that this result is driven by lower market share firms and high R&D and growth industries.

Managerial implications

There has been considerable concern about marketing’s decreasing influence in the firm (Feng et al. 2015; Homburg et al. 2015; Reibstein et al. 2009; Rust et al. 2004), in the boardroom (Webster et al. 2005), and at the corporate strategy level (McGovern et al. 2004). Marketing is increasingly viewed as a cost and not as an investment (Morgan and Rego 2009). Strategically important aspects of marketing have moved to other functions in the organization (Sheth and Sisodia 2005) and the roles of financial managers have become more important than marketing managers (Nath and Mahajan 2008). One main reason identified for this decline in marketing’s influence is its lack of accountability (Lehmann 2004; Verhoef and Leeflang 2009). Further, global competition, recessions, and stock market pressures have only increased the demands for marketing accountability (Lehmann and Reibstein 2006).

Consequently, one main question is how the accountability of marketing spending can be established in a firm. Scholars have suggested that there is a gap in communication at the highest management levels and have suggested staffing solutions, i.e., appointing a Chief Marketing Officer (CMO) to the top management team, who can communicate marketing’s accountability to CEOs and CFOs (Germann et al. 2015). However, Moorman (2013) reports that demonstrating marketing’s value remains a challenge, even for CMOs. As a result, a key question is how CMOs can demonstrate the value of marketing spending to CEOs and CFOs, who consider marketing spending as discretionary, i.e., readily cut under analysts’ pressure.

The answer, in our view, is that CMOs employ a clear two-step empirical demonstration, (1) showing that firms like theirs, and/or firms which operate in the industry in which their firm operates, do cut marketing spending under analysts’ pressure (as we have established in H1) and (2) showing the value of commitment to marketing spending under analysts’ pressure, for the purpose of achieving better financial or stock market performance (as we have established in H2). Our analysis of firm- and industry-based moderator variables reveals that the commitment result is driven by lower market share firms and high R&D and growth industries. One potential explanation is that high R&D and growth industries are associated with frequent innovations which require customers be informed frequently, resulting in the importance of commitment to marketing spending. Another potential explanation is that lower market share firms have built less brand assets relative to top four market share firms, hence commitment to marketing spending under analysts’ pressure is more important for the positive financial performance of lower market share firms. Another staffing solution is a CEO with a background in marketing. However, the number of CEOs with marketing backgrounds has decreased (Verhoef and Leeflang 2009). Consequently, the two-step empirical demonstration remains a key priority for the CMO or the top marketing executive in the firm. Finally, what if CEOs and CFOs continue to cut marketing spending under analysts’ pressure despite demonstration of the value of commitment to marketing spending under analysts’ pressure? In such firms, the boards of directors will need to provide the correct advice for dealing with analysts’ pressures, and compensation committees will need to develop longer-term incentives for top executives (e.g., increasing the equity to bonus ratio), so that top executives under analysts’ pressures do not cut marketing spending in the short term, a decision which is counter-productive for firms’ longer-term financial performance.

Future research

It would be interesting for future research to focus on the following issues. First, our original analysis of H1, i.e., the effect of analysts’ pressure on marketing spending, considers the entire continuum of analysts’ pressure, i.e., when it is positive and negative. Future research can conduct separate analyses for positive and negative analysts’ pressures. Second, it would be useful to conduct a qualitative and quantitative study to investigate the mechanism through which firms withstand the pressure and improve long-term performance. Third, while we employed the Hausman-Wu endogeneity test to determine whether marketing commitment is independent from remaining contemporaneous errors, future research can investigate whether there are potential endogeneity concerns resulting from the CEO’s background, the presence of a CMO, or compensation. Fourth, there are studies showing that some analysts (e.g., star analysts) may have a more salient impact on firm stock performance (e.g., Loh and Stulz 2009). However, to the best of our knowledge, I/B/E/S has stopped providing the identities of individual analyst forecast data, most likely because revelation of identity could cause legal risks for analysts. Consequently, while differences in analysts and their impact on firm decisions is an interesting question to consider, it is not feasible, because of data reasons, to examine the influence of specific individual analysts on firm decisions. Perhaps future research can investigate the differences between star and other analysts, and the herding and optimism biases noted in the introduction of this article based on a survey. We hope others will investigate these issues in order to build on our work in academic and commercial settings.