Zayed University

ZU Scholars

All Works

12-1-2018

Investor sentiment and advertising expenditure

G. Mujtaba Mian
Zayed University

Piyush Sharma
Curtin University

Ferdinand A. Gul
Deakin University

Follow this and additional works at: https://zuscholars.zu.ac.ae/works

Part of the Business Commons

Recommended Citation
Mian, G. Mujtaba; Sharma, Piyush; and Gul, Ferdinand A., "Investor sentiment and advertising expenditure"
(2018). All Works. 2142.
https://zuscholars.zu.ac.ae/works/2142

This Article is brought to you for free and open access by ZU Scholars. It has been accepted for inclusion in All
Works by an authorized administrator of ZU Scholars. For more information, please contact
[email protected], [email protected].
© 2018. This manuscript version is made available under the CC-BY-NC-ND 4.0 license http://creativecommons.org/
licenses/by-nc-nd/4.0/

Investor sentiment and advertising expenditure

Mian, M. and Sharma, P. and Gul, F.

Abstract

A strategic issue facing marketing managers is ‘how much and when’ to spend on

advertising. We argue that investor sentiment in the stock market may influence advertising

expenditure by affecting firms’ ability to raise new funds. We show that during periods of low

(high) investor sentiment, firms decrease (increase) their advertising expenditure, even

though the effectiveness of advertising is greater (lower) during such periods. We also find

that these results are stronger for financially constrained firms that rely more on external

financing. Our findings suggest that marketing managers can improve the efficiency of their

advertising expenditure by raising (reducing) it during periods of low (high) sentiment.

Keywords: Advertising expenditure; affordability method; marketing expenditure; investor

sentiment; stock market; advertising effectiveness.

1. Introduction

The extant marketing literature often recognizes the importance of ‘affordability’ in

determining the advertising budgets of firms, whereby managers spend more on advertising

when they have greater funds at their disposal and less when they are short of money 1 (e.g.,

Joseph & Richardson, 2002; Tellis, 1998). An important strand of this literature links the

notion of affordability to broad macroeconomic factors and examine how economic

recessions influence the levels of advertising expenditure (e.g., Deleersnyder et al. 2009;

Frankenberger & Graham, 2003; Srinivasan et al., 2005) as well as its effectiveness in terms

of firm-level outcomes such as profits and market share (Steenkamp & Fang 2011, Srinivasan

et al. 2011). However, there is hardly any research on how the conditions in the stock markets

may impact the availability of funds for the firms, and hence, their advertising expenditure.

We address this gap by studying the influence of investor sentiment in the stock market on

advertising expenditure.

By using investor sentiment, we look beyond the effects of macro-economic variables

such as state of the economy, because the concept of investor sentiment relates to conditions

in financial markets and depicts broad waves of investor optimism and pessimism in the stock

market, which do not always overlap with the economic cycles of recession, contraction, or

expansion. This also allows us to extend the aforementioned notion of ‘affordability’ by

emphasizing the availability of external funds to managers. In doing so we integrate and

extend the research on factors influencing advertising expenditure, including the state of the

economy and the availability of funds.

1
Advertising budgets are also derived using ‘percentage-of-sales’ and ‘competitive-parity’ methods (see, for
example, Blasko & Patti, 1984; Patti & Blasko, 1981), which are discussed later in the paper.

1
 The finance literature has long recognized that besides internally generated earnings, a

firm’s access to external funds (e.g., by issuing new shares in the stock market or taking a

new bank loan) is an important source of funds available to managers for investment

(Chirinko & Schaller, 2001; Polk & Sapienza, 2009; Shleifer, 2003). Access to external

funds, in turn, depends on conditions in the financial markets, which vary considerably over

time. In the midst of the Internet bubble of 1998-99, for example, many young technology

firms had little internal cash flows yet they were flush with resources as stock market

investors were keen to invest in such firms. Conversely, during the financial crisis of 2008-9,

many companies had trouble financing their urgent investment needs, even though they were

profitable, simply due to problems in the banking sector and stock market (e.g., Desperately

Seeking a Cash Cure, The Economist (November 20, 2008), and Cutting Costs, The

Economist (December 24, 2008). Conditions in the financial market therefore have an

important bearing on the availability of funds for firms.

We focus on the burgeoning literature in finance, which argues that financial markets

experience broad waves of investor optimism and pessimism, labelled as investor sentiment

(Baker & Wurgler, 2006, 2007). Such changes in sentiment cause significant variations over

time in investors’ perceptions about firms’ future prospects. During periods of high sentiment,

investors are excessively optimistic about the future and overestimate the expected cash flows

and/or underestimate the risk, resulting in an overvaluation of stocks. This overvaluation and

increased investor appetite for risk makes it easier and attractive for managers to use the

stock market to raise new funds for their firms’ investment needs. The converse happens

during periods of low sentiment, when investors are excessively pessimistic about the future.

They systematically underestimate the expected cash flows and/or overestimate risk, causing

an undervaluation of stocks. This undervaluation and investors’ heightened aversion to risk

make it difficult for managers to obtain new funds from investors (Luo, Jiang, & Cai, 2014).

2
This is especially true for financially constrained firms—that is, firms that have low or

negative internal cash flows and face substantial uncertainty about their future prospects—

whose access to external finances may dry up dramatically during low sentiment periods.

Indeed, managers’ ability to raise new funds from the stock market is so closely linked to

the prevailing sentiment in the stock market that the most common measure of investor

sentiment in the recent literature, the Baker and Wurgler sentiment index (or BW index), uses

the number of initial public offerings (IPO) and the first day return on these IPOs as

important inputs to compute the index (Baker & Wurgler, 2006, 2007). Hence, by default,

periods when the BW sentiment index is low (high) are those when managers find it difficult

(easy) to access new funds from the stock market. The affordability method of advertising

budgeting would, therefore, predict that the level of advertising expenditure is systematically

lower (higher) during periods of low (high) investor sentiment.

Advertising expenditure may be especially prone to funding constraints because unlike

spending on equipment or setting up of a factory, it is often viewed as discretionary, arguably

because its payoffs are relatively uncertain and accrue over longer periods (Erickson &

Jacobson, 1992; Mizik, 2010; Mizik & Jacobson, 2007). We therefore hypothesize that

managers would spend less on advertising (as a percentage of sales) during periods of low

sentiment than during periods of high sentiment. This would be especially true for financially

constrained firms because they depend to a greater extent on external financing and it

becomes especially difficult for them to raise funds during low sentiment periods, compared

to firms with lower financial constraints (Baker, Stein, & Wurgler, 2003).

To test these hypotheses, we employ Baker and Wurgler’s (2006, 2007) market-wide

investor sentiment index and examine how advertising levels vary with sentiment. We study

all firms in the COMPUSTAT dataset with listed stocks during 1971-2010. Our results show

3
that the firms in our sample spend significantly less on advertising, on average, during

periods of low investor sentiment than during periods of high sentiment. These results hold

after we explicitly control for the effect of macroeconomic variables, such as recessions and

consumer sentiment, as well as industry- and firm-level variables that depict firms’ growth

prospects. Using two popular measures of financial constraints, we also find that the positive

association between sentiment and advertising is stronger for financially constrained firms

that must rely more on external funding for their investment needs. We also provide

additional empirical evidence to justify our key premise that external financing is a

significant component for a typical firm in our sample. We also show that advertising

expenditure tends to be positively associated with external financing, especially for

financially constrained firms. Finally, firms tend to raise more external funds during high

sentiment periods, especially those that are financially constrained.

Allowing investor sentiment to influence advertising expenditure may lead to lower

advertising efficiency because changes in investor sentiment over time do not correlate

perfectly with product market considerations and consumer sentiment. For example, the

advertising budget may be sub-optimal when periods of low investor sentiment coincide with

periods of high consumer sentiment. During such periods, expectations of higher future

consumer demand and considerations of competitive parity may dictate increasing the

advertising; yet managers may be forced to cut down on advertising due to the limited

availability of funds. Moreover, because firms in aggregate would spend less (more) on

advertising during low (high) sentiment periods, advertising may be more (less) effective

during such periods due to lower (greater) clutter and cheaper (expensive) rates by

broadcasting channels. Therefore, it is likely that the effectiveness of advertising expenditure

is higher (lower) in low (high) sentiment periods. While it is easy to measure advertising

expenditure, there is little consensus on how to measure advertising effectiveness (Fischer &

4
Himme, 2016; Simon, 1984; Soberman, 2009). We measure advertising effectiveness by the

elasticity of profits to advertising (e.g., Kamber, 2002; Srinivasan et al., 2011). We find that

advertising is indeed more effective during low (vs. high) sentiment periods.

We contribute to the marketing literature in several ways. First, we provide insights for

managers on how collective advertising budgets vary systematically over time and how

perceptive managers who ‘buck the trend’ can benefit from such time variation. For example,

because most companies decrease their budgets during low sentiment periods, there is less

competitive clutter and better consumer recognition in the market, allowing some companies

to get more mileage from their advertising. Conversely, because most companies increase

their advertising budgets during high sentiment periods, some managers may find it useful to

channel some of the advertising budget to other marketing activities.

Second, our study extends the ongoing research endeavors that explore the links between

macroeconomic variables and time-variation in the level or effectiveness of advertising

expenditure (e.g., Srinivasan et al., 2011; Steenkamp & Fang, 2011) by examining the effect

of financial market conditions on firms' marketing decisions. Together, these studies provide

useful insights on how market-wide factors cause variations in firms' advertising expenditures

over time. Third, by linking investor sentiment in the stock market with firms’ advertising

expenditures, we help integrate the latest insights in the behavioral finance literature with

issues of interest to marketing professionals and academicians. Finally, our findings suggest

that the notion of ‘affordability of funds’ in the marketing literature need to be extended to

include access to ‘external’ funds available to a firm.

The rest of the paper proceeds as follows. We first review the related finance literature on

investor sentiment in the stock market and then develop our hypotheses. Next, we describe

our data, measures, and methodology, followed by our empirical findings. Finally, we

5
conclude the paper by discussing the conceptual and practical implications of our findings,

and their limitations.

2. Theoretical background and hypotheses

2.1. Investor sentiment, stock market valuations, and corporate decisions

The classical view in finance, often called ‘efficient markets hypothesis’, leaves little

room for investor psychology to influence markets (e.g., Fama, 1970). According to this

view, the market prices of financial assets, such as stocks, reflect the discounted present value

of the expected future cash flows. If prices ever move away from this correct valuation

benchmark, smart investors or arbitrageurs move in and trade aggressively, pushing prices

back to their correct valuation. In sharp contrast to this classical view, however, a burgeoning

literature in behavioral finance argues that investor psychology can and does play a

significant role in financial markets, in general, and the stock market in particular (Barberis &

Thaler, 2003; Shiller, 2003). This literature not only discusses a number of cognitive errors

that investors make when trading in the stock market, but also identifies the factors that limit

the ability of smart arbitrageurs to correct the effect of such errors on the market.

A strand of the behavioral finance literature examines the effect of broad market-wide

waves of investor optimism and pessimism, or investor sentiment, on the stock market (e.g.,

Baker & Wurgler, 2007). It defines investor sentiment broadly as ‘beliefs about future cash

flows or discount rates that are not supported by the prevailing economic fundamentals’

(Baker & Wurgler, 2006, 2007), and constructs indexes of investor sentiment in the stock

market by combining information contained in several metrics of investor exuberance after

purging the effects of macroeconomic fundamentals (Baker & Wurgler, 2006; Lemmon &

Portniaguina, 2006). Studies in this line of research not only provide empirical evidence that

validates the proxies of investor sentiment as measures of investor optimism and pessimism,

6
but also show that investor sentiment causes over- and under-pricing of stocks in the stock

market. These studies also demonstrate that the effect of sentiment is distinct from that of

macroeconomic variables.

Sentiment-driven mispricing in the stock market also has significant effects on managerial

investment decisions (Baker & Wurgler, 2012; Dong, 2010; Gervais, 2010; Polk & Sapienza,

2009). When stock market valuations are high (low) in high (low) sentiment periods,

managers tend to over- (under-) invest (Chirinko & Schaller, 2001; Lamont & Stein, 2006;

McLean & Zhao, 2014). In bullish times when their stock is overpriced and investors are

ready to provide them with capital at very low cost, mangers invest more and finance the new

investments by issuing new shares at inflated prices (Lamont & Stein, 2006; Polk &

Sapienza, 2009; Shleifer & Vishny, 2003). Conversely, when sentiment is low and the stock is

underpriced, new funding is prohibitively expensive and managers tend to significantly

reduce their investments (Stein, 1996; Baker, Stein, & Wurgler, 2003).

2.2. Determinants of advertising expenditure

Early research on advertising expenditure identifies ‘affordability’, ‘percentage-of-sales’

and ‘competitive-parity’ as the most common methods by which firms decide their

advertising budgets (Blasko & Patti, 1984; Patti & Blasko, 1981). Subsequent studies show

that none of these methods may be optimal and as a result, most firms end up overspending

on advertising (Aaker & Carman, 1982; Lodish, et al., 1995). Among all these methods,

‘affordability’ seems quite popular because it allows firms to spend on advertising what they

can afford. Under this method, firms with greater resources may spend more on advertising

than what is necessary or desirable, and conversely, firms with little resources may

underspend on advertising (Joseph & Richardson, 2002). The main reason why marketing

managers may prefer this method to others is the inherent uncertainty about the exact effect

7
of advertising on sales. It is also suggested that allocating a part of the discretionary budget to

advertising is like purchasing insurance, and it makes sense to buy more insurance when the

firm has access to greater cash at a lower cost of capital (Joseph & Richardson, 2002).

In contrast, in ‘competitive-parity’ approach firms benchmark their advertising budgets

against the advertising expenditure of their competitors in the same industry (Dean, 1951;

Lee, 1994; Mitchell, 1993; Prasad & Sethi, 2004). For example, firms may decide to spend a

percentage of the total competitive advertising expenditure in an industry that is equal to its

share of the market, referred to as ‘share of voice’ (Jones, 1990). The advocates of this

approach argue that the total advertising spend and its allocation to all the competitors in an

industry represents their combined wisdom. However, it assumes that the competitors know

what they are doing and that they all share the same goals as each other, which is generally

not true and may lead to inefficient decisions (Lilien, Kotler, & Moorthy, 1992). Prior

research also identifies other drivers of advertising expenditure, such as changes in

advertising strategy and/or creative approach, and profit contribution goals or other financial

targets (Belch et al., 2008; Prendergast, West & Shi, 2006).

Overall, prior research shows that firms use a variety of heuristics to decide their

advertising budgets, even though many of them may not involve strictly ‘rational’ decisions.

There is also a growing realization that marketing managers do not always use complex

methods to arrive at their advertising budgets (Batra, 2009). An extensive review of

marketing literature over six decades suggests that advertising decisions are usually not

derived from rigorous empirical generalizations (Rasmussen, 1952; Rossiter, 2012).

Moreover, there is still limited research on the ‘external’ or ‘environmental’ factors that could

influence marketing managers' decisions about advertising expenditure (Albers, 2012;

Hanssens et al., 2016; Martín-Herrán et al., 2012).

8
2.3. Investor sentiment and advertising expenditure

As pointed out earlier, it is well accepted in the behavioral finance literature that firms’

access to new capital varies significantly with investor sentiment, wherein firms find it

relatively easier to raise new funds in high sentiment periods and face serious funding

constraints during low sentiment periods. In fact, Baker and Wurgler (2006) construct their

sentiment index using variables such as the number of initial public offerings (IPOs) and the

first day return on the IPOs, which reflect the ease with which firms can raise new funds for

their investments. The affordability argument would, therefore, suggest that periods of high

sentiment are associated with higher expenditure on advertising. The converse would be true

for low sentiment periods, when firms are short of funds and are forced to curtail their

advertising budgets (Joseph & Richardson, 2002; Luo et al., 2014; Tellis, 1998).

Marketing expenditure may be especially prone to sentiment-related changes in access to

external funds because advertising may be viewed as discretionary, and its contribution to

sales and earnings may not be readily visible (Srinivasan & Hanssens, 2009). Furthermore,

the marketing benefits may accrue over several quarters or even years (Steenkamp & Fang,

2011). The behavioral finance literature argues that the effect of managerial incentives is the

most pronounced in settings where the quality of managers’ actions is difficult to evaluate

and the feedback is noisy (Gervais, 2010). Marketing expenditure appears to fit this

description well. Based on the above, we state our first hypothesis as follows:

H1: Firms spend less on advertising (as a percentage of sales) during periods of

low investor sentiment than during periods of high sentiment.

It is instructive to reiterate and clarify that HI is underpinned by three premises: (i)

external finance, such as new stock or bond issues or new borrowing from banks, is an

important component of total funds available to a firm; (ii) a firm’s advertising budget is

9
positively associated with the availability of external funds; and (iii) firms can raise external

funds more easily during high sentiment periods than during low sentiment periods. In a later

section, we provide empirical support for each of these three premises.

2.4. Moderating effects of firm attributes

If time varying access to external funds indeed underpins the relation we hypothesize in

H1, we should find that it is more pronounced when firms are financially constrained. A firm

is perceived as financially constrained if it has low or negative internal cash flows and finds it

difficult to fund its investment needs at a given point in time. Such firms typically have high

uncertainty and volatility, which makes it difficult for them to raise external funds (Whited &

Wu, 2006, Hadlock & Pierce, 2010). Their investment budgets are more likely to depend on

the conditions prevailing in the financial markets for two reasons. First, due to limited

internal funds, these firms may rely more on external sources to fund their investment needs

(Stein, 1996, Baker, Stein, & Wurgler, 2003). Second, because they tend to be more volatile

and risky, investors’ willingness to invest in them may vary more with conditions in the

financial markets (Baker & Wurgler, 2012).

When investor sentiment in the stock market is high, investors’ risk aversion decreases.

They become willing to invest in firms with limited internal cash flows and that appear

volatile and risky. This happens not only because investors are willing to take the risks

associated with these companies, but also because they may perceive that these companies

have greater growth prospects. Such firms therefore find it easier to raise new capital during

periods of high sentiment (Baker & Wurgler, 2012). In contrast, when sentiment is low,

investors’ risk aversion increases. They have limited appetite for investing in firms with weak

internal cash flows and that appear risky and volatile. This makes it very difficult for such

firms to raise money for their operations (see, for example, Desperately Seeking a Cash Cure,

10
The Economist (November 20, 2008), and Cutting Costs, The Economist (December 24,

2008). In sum, because financially constrained firms may have to rely more on external funds

to finance their advertising budgets, and because their ability to raise external funds varies

more with sentiment, their advertising budgets would be more prone to changes in investor

sentiment than other firms. 2 We, therefore, predict a stronger positive association between

advertising and investor sentiment for financially constrained firms than for other firms. 3

H2: The effect of investor sentiment on advertising expenditure is stronger when

firms are financially more constrained.

3. Research methodology

3.1. Data

We obtain data on firms’ annual advertising expenditure from COMPUSTAT. The primary

source of the COMPUSTAT data is the publically disclosed accounting statements of firms.

The data coverage is sparse before 1971, so our sample period ranges from 1971 to 2010 and

covers a large number of companies across a broad set of industries. Prior to 1994,

accounting regulations in the US required listed commercial and industrial firms to make

public a ‘Supplementary Income Statement Information’ schedule that included information

about advertising. In December 1993, Accounting Standards Executive Committee (AcSEC)

issued a new rule SOP 93-7, which made it optional for companies to report their advertising

expenditure. This rule, which became effective from June 15, 1994, also curtailed firms’

2
In a later section, we provide empirical evidence for these two premises. Specifically, we show that (i) the
advertising budgets of financially constrained firms have a stronger positive association with external financing,
and (ii) the external financing of such firms is more strongly associated with sentiment.
3
As we discuss later, we employ two indices popular in the finance literature—the WW index and SA index—
to measure firms’ financial constraints.

11
latitude in terms of how they report their advertising expenditure and increased the

uniformity of the reported expenses across firms. Nevertheless, the dataset covers a large

number of companies across a broad set of industries.

We merge the advertising and accounting data from COMPUSTAT with the stock price

and returns information from the Centre for Research on Security Prices (CRSP). We retain

firms with common shares listed on US stock exchanges and remove specialized firms such

as closed-end mutual funds, American Depository Receipts, and real estate investment trusts

from our analysis. We retain only those firms that report a positive value for advertising

expenditure in our analysis. We also remove a tiny number of firms that report negative net

sales. We further remove observations for which we do not have complete data for all of the

control variables and measures of financial constraints. Our final sample includes 54,209

firm-year observations covering 1971-2010. The number of unique firms in our sample is

8,201. Figure 1 reports the number of firms in our sample by year, which is typically about

1,400 firms per year. The dip in the number of firms in 1994 is due to the change in

regulation noted above, whereby the reporting of advertising became voluntary.

< Insert Figure 1 about here >

Next, we explain some of our key measures and variables in detail. Appendix A provides

a complete list of all the variables used in our study.

3.1.1. Measures of investor sentiment and economic conditions

We employ Baker and Wurgler’s (2006, 2007) investor sentiment index in our analysis. 4

Baker and Wurgler (2006, 2007) describe how the index is constructed and validated. The

BW index is a composite measure that combines six individual proxies of investor sentiment,

4
We thank Jeffrey Wurgler for making the index available on his web site http://people.stern.nyu.edu/jwurgler/.
Every few years, the index is updated to include a more recent period.

12
namely closed-end fund discount, NYSE share turnover, number of IPOs, first day returns on

IPOs, share of equity issues in total debt and equity issues, and dividend premium (the log

difference of the average market-to-book ratio of payers and non-payers). To isolate investor

sentiment from economic fundamentals, each of these six proxies is first regressed on five

macroeconomic variables: the growth in industrial production index (Federal Reserve

Statistical Release G.17); growth in consumer durables, nondurables, and services (all from

BEA National Income Accounts Table 2.10); and a dummy variable for NBER recessions.

The first principal component of the residuals of these six regressions constitutes the investor

sentiment index, standardized to yield a mean of zero and a standard deviation of one.

Positive values signify periods of high (i.e., bullish) sentiment and negative values periods of

low (i.e., bearish) sentiment. Baker and Wurgler (2006, 2007) show that their sentiment index

coincides well with anecdotal accounts of investor exuberance and panic. The index has since

been widely used to examine the implications of broad waves of market-wide sentiment for

financial markets (Hribar & Mcinnis, 2012; Mian & Sankaraguruswamy, 2012; Stambaugh,

Yu, & Yuan, 2012, 2015; Yu & Yuan, 2011).

Before proceeding with our main analyses, we first examine the overlap between the

investor sentiment index and the measure of macroeconomic performance used in Steenkamp

and Fang (2011). We closely follow the procedure in Steenkamp and Fang (2011) to construct

their measure of economic recessions, Contr, which is based on the filter and the cyclical

component of GDP (Hodrick & Prescott, 1997). Specifically, Contr takes the value of zero

when the economy is growing at or above its long-term trend, but equals the magnitude of the

decline in the cyclical component of GDP when the economy is in recession. We find that

investor sentiment has a negligible correlation of 0.002 with the Steenkamp and Fang

measure. This is reassuring because it suggests that we are indeed investigating a

phenomenon that is empirically distinct and perhaps more novel than their measure.

13
 It is also useful to compare and contrast the BW measure of investor sentiment with

measures of consumer sentiment. To highlight the conceptual and empirical differences

between investor sentiment and consumer sentiment, we us the University of Michigan index

of consumer sentiment (ICS), which is one of the most widely used measures of consumer

sentiment in the US. While the ICS focusses on near-term consumer attitudes on the business

climate, personal finance, and spending, the BW investor sentiment index depicts the

optimism or pessimism, or risk aversion, of investors in the stock market. In contrast to

consumer sentiment, which is based on telephone interviews of US households, investor

sentiment is constructed based on data generated within the stock market. Empirically, we

confirm the distinct nature of consumer sentiment by collecting its data over our sample

period of 1971-2010. ICS is now available on a monthly frequency, but it was available on a

quarterly basis before 1978. We therefore use its quarterly observations over our entire

sample period, and take the average of the four calendar quarters to compute the average

value of the index for a year. We then compute the correlation between the annual

observations of ICS and investor sentiment. The correlation is 0.22, as reported in Table 1. To

confirm that the effect of investor sentiment on advertising that we document is distinct from

any effect of an economic recession and consumer sentiment on advertising, we include both

Contr and ICS in our regressions as controls.

< Insert Table 1 about here >

3.1.2. Measure of advertising intensity

To investigate how advertising expenditure varies with investor sentiment in the stock

market, it would be inappropriate to look at just the dollar amount of advertising expenditure

because sentiment could relate to firms’ revenues, which, in turn, could influence the dollar

14
amount firms spend on advertising. We obtain our measure of a firm’s advertising intensity by

dividing the firm’s advertising expenditure with its (average) sales revenue, as follows.

Advertising t
Advertising Intensity t = (1)
 Salest -1 + Salest 
 
 2 

3.1.3. Measures of financial constraints

We employ two popular composite measures of firm-level financial constraints from the

recent literature: the WW index and SA index. Whited and Wu (2006) developed the WW

index of constraints using an Euler equation approach from a structural model of investment.

Their index relies on six different factors to estimate the financial constraints a firm faces:

cash flows, dividends on preferred and common stock, book value of assets, the firm’s own

sales growth, the ratio of long term debt to total assets, and the growth of sales in the

industry. The index loads negatively on the first four factors and positively on the last two. A

higher value of the index for a firm signifies that the firm has higher financial constraints, and

must rely more on external funds for new investments. Specifically, the index is computed for

each firm in each year using the inputs and their relative weights specified in Whited and Wu

(2006, Equation 13) as follows:

WWIndext = −0.091CFt − 0.062 DIVPOS t + 0.021LTDt − 0.044TAt + 0.102 ISGt − 0.035SGt

(2)

The second measure—SA index—is developed more recently by Hadlock and Pierce

(2010). They first gather detailed qualitative information on financial constraints from

statements made by managers in financial filings, and use this to categorize firms in terms of

financial constraints. Using these derived financial constrained categories, they then estimate

ordered logit models to extract the quantitative firm variables that help identify the

constraints. They find that a combination of three simple factors (assets, assets squared, and
15
the number of years since a firm first appears in the COMPUSTAT dataset) works reasonably

well, and better than many more elaborate indices, in identifying firms with financial

constraints. We follow Hadlock and Pierce (2010), and compute the SA index for each firm in

each year based on the inputs and the relative weights as in Column (2) of Table 6 (p. 1928):

𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑡𝑡 = −0.737𝐴𝐴𝐴𝐴𝐴𝐴𝑡𝑡 + 0.043𝐴𝐴𝐴𝐴𝐴𝐴2𝑡𝑡 − 0.040𝐴𝐴𝐴𝐴𝐴𝐴𝑡𝑡 (3)

The explanatory variables include ATA, the log of total assets adjusted for inflation, and

Age, which is the number of years since a firm first appeared in COMPUSTAT dataset. The

SA index loads negatively on assets and age, and positively on the square of assets. It is

worth noting that Whited and Wu (2006) and Hadlock and Pierce (2010) do not find industry

association to be marginally informative about financial constraints of firms once other

variables are taken into account. 5 Moreover, the correlation of the aggregated yearly values of

WW index and SA index with investor sentiment is 0.074 (p-value = 0.65) and 0.058 (p-value

= 0.72), respectively. These low correlations are not surprising because the financial

constraint indices are based on fundamental accounting variables whereas investor sentiment

is constructed after purging the effect of the fundamentals.

3.1.4. Summary statistics

Table 2 reports the summary statistics of our variables. Panel A reports the statistics for

the sentiment and macroeconomic variables that are available at yearly frequency. The

statistics are consistent with those reported in prior studies. The investor sentiment index by

construction has a mean close to zero, and a standard deviation close to one. Contr has a

median of zero, consistent with the evidence in Steenkamp and Fang (2011), suggesting that

5
No prior evidence exists on whether the weights of the inputs in WW and SA indices are generalizable across
countries. This is, however, not a concern for us as we use the indices for the same country for which they were
developed, that is, for the US market.

16
fewer than one year in two experience economic contraction in the US. ICS, which is one of

our important control variables, has a mean of 85, consistent with the numbers reported in

Lemmon and Portniaguina (2006).

< Insert Table 2 about here >

We use winsorized variables in all our analyses. Panel B reports the statistics for the firm

and industry level variables based on pooled firm-year observations. All variables in Panel B

are winsorized at 1% and 99% by replacing all values above (below) the 99th (1st) percentile

of the respective pooled firm-year distribution with the value of the 99th (1st) percentile to

mitigate the effect of extreme and/or erroneous observations on the regression estimates. The

mean and standard deviation of the annual advertising expenditure (as a percentage of

average sales) is 3.5% and 4.8%, respectively, for the firms in our sample. The firms included

in our sample on average spend 7.3% of their sales on capital expenditure and 4.3% on R&D.

3.2. Model specification

To investigate the association between investor sentiment and advertising intensity, we

estimate a model similar to Chakravarty and Grewal (2011). Specifically, we regress a firm's

advertising expenditure (scaled by average sales) on the prevailing investor sentiment in the

stock market, as follows:

ln(𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴 𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼)𝑖𝑖,𝑡𝑡
= 𝑎𝑎0 + 𝑎𝑎1 ln(𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴 𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼)𝑖𝑖,𝑡𝑡−1 +𝑏𝑏1 𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑡𝑡 + 𝑏𝑏2 𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑡𝑡 𝑥𝑥𝐹𝐹𝐹𝐹𝑖𝑖,𝑡𝑡
+ 𝑏𝑏3 𝐹𝐹𝐹𝐹𝑖𝑖,𝑡𝑡 + 𝑏𝑏4 𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝑡𝑡 + 𝑏𝑏5 𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝑡𝑡 𝑥𝑥𝐹𝐹𝐹𝐹𝑖𝑖,𝑡𝑡 + 𝑏𝑏6 𝐼𝐼𝐼𝐼𝐼𝐼𝑡𝑡 + 𝑏𝑏7 𝐼𝐼𝐼𝐼𝐼𝐼𝑡𝑡 𝑥𝑥𝐹𝐹𝐹𝐹𝑖𝑖,𝑡𝑡
𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶 𝑅𝑅&𝐷𝐷
+ 𝑐𝑐1 𝐻𝐻𝑒𝑒𝑟𝑟𝑟𝑟𝑖𝑖,𝑡𝑡 + 𝑐𝑐2 𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼ℎ𝑖𝑖,𝑡𝑡 + 𝑐𝑐3 � � + 𝑐𝑐4 ln �1 + �
𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆 𝑖𝑖,𝑡𝑡 𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆 𝑖𝑖,𝑡𝑡
+ 𝐹𝐹𝐹𝐹𝐹𝐹𝐹𝐹 𝐹𝐹𝐹𝐹𝐹𝐹𝐹𝐹𝐹𝐹 𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸 +∈𝑖𝑖,𝑡𝑡

(4)

17
 The subscripts i and t in the model signify firm and year, respectively. We use the

(natural) log-transformed values of advertising intensity as the dependent variable because

the distribution of advertising intensity is positively skewed. 6 We include the lagged value of

the dependent variable as an explanatory variable to account for the inertia in the advertising

expenditure of firms. To test Hypothesis 1, the key coefficient of interest is b1, which captures

the effect of contemporaneous investor sentiment, Sent, on advertising budgets. The

hypothesis predicts this coefficient to be positive and significant. To test Hypothesis 2, we

include FC as a measure of a firm’s financial constraints. It alternately takes the values of

WW index and SA index, the two composite measures of financial constraints that we discuss

earlier. According to Hypothesis 2, the coefficient b2 on the interaction term Sent x FC should

be positive, indicating that the positive association between sentiment and advertising

intensity is more pronounced for financially constrained firms.

We estimate the model in Equation (4) using pooled firm-year data with firm fixed

effects. The inclusion of firm fixed effects controls for any time-invariant firm characteristics

that may be relevant to advertising intensity, but that we do not include in our specification.

The t-statistics we report are robust to the presence of heteroscedasticity. They are also based

on standard errors clustered by year, to control for the time-specific correlations in

advertising budgets across firms (Petersen, 2009).7

6
We also experiment with two alternative specifications of Equation (4) to check the robustness of our
inferences. First, to mitigate the concern about simultaneity bias, we use one-period lagged values of the
explanatory variables on the right-hand side. Second, to examine the effect of scaling by sales of our firm-level
variables, we estimate the model using unscaled variables. In this formulation, we use natural log of the dollar
amount of advertising as the dependent variable, and include firm sales as an additional control. We find that
our inferences are robust to using these alternative specifications.
7
Simple OLS estimations compute standard errors under the assumption that all observations are independent.
However, this assumption is frequently violated in panel data estimations and the process of clustering adjusts
the standard errors and the t-statistics, for violation of this assumption. In our context, clustering by year is
meant to correct the t-statistics for correlation in advertising expenditures across firms within a given year.

18
 We use Contr and ICS as control variables to account for economic recessions and

consumer sentiment respectively in view of their importance as discussed earlier in the paper.

Because the effect of economic recession and consumer sentiment on advertising may also

differ for financially constrained firms, we also interact these variables with FC and include

these as additional controls in the model. Herf is the Herfindahl-Hirschman index, a common

measure of market concentration and competition in the firm’s industry. We compute it using

sales of firms within a two-digit Standard Industry Classification (SIC) code. We include it

because the degree of competition can influence firms’ advertising spending.

To control for the possibility that firms’ growth prospects could vary over time with

sentiment, we further include a few industry- and firm-level variables. We include growth

rate of sales from year t-1 to year t for all the firms in the industry the firm belongs to, where

we define industry by the two-digit SIC code. We also include capital expenditure scaled by

average sales, Capex, as another proxy for a firm’s growth prospects because firms with

greater growth opportunities would invest more in their fixed assets, and also because new

investments in fixed assets may necessitate additional advertising. Finally, we include

research and development (R&D) expenditure as a control variable because new product

developments can affect the advertising budgets of firms.

4. Main results and robustness checks

4.1. Main results

We begin our analysis by examining the association between advertising intensity and

sentiment in a simple bivariate setting. We do so by first computing the average value for

advertising intensity for each year based on the data available for all firms during that year.

This yields a time series of advertising intensity with 40 observations—one for each year in

our sample period of 1971-2010. The simple correlation between the annual time series of
19
advertising intensity and sentiment reported in Table 1 is 0.35, which seems substantial. We

also report the correlations between advertising and the two macro variables in Table 1.

Consistent with the findings of Steenkamp and Fang (2011) that advertising expenditure

reduces during recession, the correlation between advertising and their (inverse) measure of

economic recession is 0.34. Advertising is also positively correlated with consumer sentiment

indicating that firms spend more on advertising when they expect their customers to increase

their purchases in the near future.

We also divide our sample periods into periods of high and low sentiment and examine

the average advertising intensity across these periods. Because the BW sentiment index has a

mean of zero by construction, it is common in prior studies to use the zero cutoff to divide the

sample years into periods of high or low sentiment (Stambaugh, Yu, & Yuan, 2012, 2015).

Values of the index greater than zero signify periods of high sentiment, whereas values below

zero represent periods of low sentiment. In Table 3, we report the average advertising

intensity based on pooled firm-year observations across high and low sentiment periods. The

average advertising intensity is 3.2% during low sentiment periods and increases to 3.8%

during periods of high sentiment. This represents an increase of about 19% [= (3.8% - 3.2%) /

3.2%)] which again appears substantial.

< Insert Table 3 about here >

When we examine the relationship between investor sentiment and advertising

expenditure in a multiple regression framework using Equation (4), we obtain consistent

results in Table 4. When we estimate the model without financial constraints and its

interaction terms in Column 1, the coefficient on the key variable, Sent, is 0.017 (t-statistic =

2.72), confirming the robust positive relation between investor sentiment and firms’

20
advertising intensity. 8 Table 4 therefore shows that the results continue to support Hypothesis

1 when we control for other factors that affect the level of firms' advertising intensity.

< Insert Table 4 about here >

It is also instructive to note the coefficients of some of the control variables in Column (1)

of Table 4. The coefficient on ICS, which measures consumer sentiment, is positive and

statistically significant, suggesting that firms increase their advertising in anticipation of

consumers increasing their purchases. The significant positive coefficients on industry growth

(IndGrowth) indicates that firms increase advertising when their industry is growing. The

positive and highly significant coefficients on Capex and R&D indicate that firms spend more

on advertising when they expand the scale of their operations and when they engage in

greater product innovation, respectively. All the coefficients on the control variables are

consistent with prior literature and intuition.

To test H2, we estimate the complete model of Equation (4), in which we interact

indicator variable for financial constraints, FC, with investor sentiment and other macro

variables. The last two columns in Table 4 report these results. The effect of financial

constraints on the relationship between investor sentiment and advertising intensity is

depicted by the coefficients on the interaction variable Sent x FC in Columns (2) and (3). In

Column (2), we measure the FC variable through the WW index and in Column (3) by the SA

index. According to H2, the coefficients should be positive. The results provide strong

support for H2, as the coefficients are indeed positive and statistically significant. The

coefficient on the stand alone variable for financial constraints, FC, is positive in both

8
In a robustness check, we explore whether lagged sentiment has any marginal effect on advertising in the
presence of its contemporaneous value. The results indicate that it is not because the lagged value of sentiment
shows up as non-significant.

21
Columns (2) and (3) and is marginally significant at 10% level in the latter column. A

positive coefficient suggests that managers spend more on advertising when their resources

decline in the hope of reviving them. To summarize, Table 4 indicates strong support for both

H1 and H2. Firms’ advertising intensity seems to be positively associated with the investor

sentiment prevailing in the stock market. This association is stronger for firms that are

financially constrained, consistent with the idea that these firms depend more on the

availability of external funds to finance their advertising budgets.

4.2. Endogeneity concerns and robustness checks

A possible concern about the empirical relationship between advertising and investor

sentiment relates to endogeneity, especially the problem of omitted variables. Specifically,

despite the inclusion of a host of macro, industry and firm-level variables as controls in our

model, it is possible that the observed relationship between sentiment and advertising simply

arises due to some other variable—which we do not include in our main model in Equation

(4)—driving both sentiment and advertising. We address this concern in several ways.

First, we specify our model in Equation (4) in a way that helps us mitigate this issue. In

arriving at the specification of our model, we conduct a formal Hausman test to examine the

presence of fixed versus random effects in our model. The Chi-Square statistic turns out to be

very large (p < 0.000), which clearly rejects the null of random effects, and suggests the need

to include firm fixed effects. We therefore, include firm fixed effects in the model. Inclusion

of firm fixed effects accounts for any permanent differences in the advertising policies of

firms, and allows us to examine how over-time changes in sentiment relate to over-time

changes in advertising. In addition, we include lagged values of the dependent variable in our

model. This further controls for any omitted variables that affect the advertising expenditure

equally in both years, t and t-1.

22
 Second, to rule out the remaining possibility that some time-varying factor may be

causing endogeneity, we conduct the standard Hausman-Wu test of endogeneity (Davidson

and MacKinnon 1993; Baum et al. 2003). Specifically, we test whether sentiment (Sent) in

Equation (4) is independent from the contemporaneous random events, i.e., the error term εi,t.

We implement the test using the specification in Column (1) of Table 4, and use instruments

that are lagged one period beyond the error term. The results indicate that F-statistic is not

significant at conventional level (F1,41562) = 3.09, p-value > 0.05). We therefore conclude that

the null hypothesis—that sentiment is exogenous—cannot be rejected, and hence no explicit

correction for endogeneity is required in our model.

Finally, we note that the measure of advertising intensity, specified in Equation (1), is

obtained by dividing advertising expenditure by the average sales of the contemporaneous

and lagged years. It is possible that managers set advertising budgets based on planned sales

and not based on current or lagged sales, and expected sales are higher during period of high

sentiment. If so, it is possible that the positive association between advertising and sentiment

reported in Table 4 arises spuriously due to a positive association between future sales and

sentiment. To rule out this possibility, we conduct additional analysis. Specifically, we

compute a revised measure of advertising intensity by dividing the advertising spend with

future sales. That is, we divide the advertising expenditure in year t with sales in year t+1

(and in further robustness check, with sales in year t+2). We then examine the association

between this revised measure of advertising intensity and sentiment.

If the previously documented relationship between sentiment and advertising simply

arises because managers set advertising budgets based on expectations of higher future sales

following periods of high sentiment, then the new measure of advertising intensity should not

vary over time with investor sentiment. This is because the new measure scales, or deflates,

advertising using future sales. In untabulated analysis, we re-run our main regression model
23
in Table 4 by replacing its dependent variable with the revised measure of advertising

intensity. The results mirror those reported in Table 4, with a positive and highly statistically

significant coefficient for Sent, so the revised measure of advertising intensity exhibits almost

the same level of strong positive association with sentiment as our primary measure. This

suggests that the advertising-sentiment relation that we report does not spuriously emerge

because we ignore the relation between current advertising and future sales. To summarize,

our analyses in this section suggests that endogeneity may not be a significant concern about

the observed relationship between sentiment and advertising expenditure.

5. Additional analyses

5.1. Empirical validity of key premises underlying our hypotheses

While our main results in Table 4 provide support for our two hypotheses, it would also

be instructive to separately examine the empirical validity of the three premises that underpin

HI. As noted earlier, these premises include the following: (i) external financing is a key

source of funds for most companies, (ii) a firm’s advertising budget in a year is positively

associated with the external financing it receives during the year, and (iii) more external

financing becomes available during periods of high sentiment than those with low sentiment.

The empirical support for the first premise appears in Table 2, where we report the

summary statistics of the external financing that firms receive each year. We compute the

amount of external financing based on information from financial statements available

through COMPUSTAT. Specifically, we determine the external financing of a firm in year t

by taking the change in total assets from year t-1 to t, and subtracting from it the change in

retained earnings over the same period. This definition encompasses all types of external

financing, including net new stock issue, net new bond issues, short- and long-term

borrowing from banks, and so on. We scale external financing by average sales in year t-1
24
and t to make it comparable to the other firm-level variables we report, such as cash flows

and advertising. We report the summary statistics of firms’ external financing in our sample

for 1971 to 2010. The mean is around 24%, which is greater than the mean of the annual

internal cash flows that firms generate, which is close to zero. We compute internal cash

flows as income before extraordinary items plus depreciation and amortization, scaled by

average sales. The amount of external financing in a year also appears significant compared

to the annual capital expenditure of 7.3% and annual advertising expenditure of 3.5%.

We test our second premise about the positive association between advertising

expenditure of a firm and the external financing it obtains using a specification similar to that

in Equation (4), as follows:

ln(𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴 𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼)𝑖𝑖,𝑡𝑡
= 𝑎𝑎0 + 𝑎𝑎1 ln(𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴 𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼)𝑖𝑖,𝑡𝑡−1 +𝑏𝑏1 𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝑡𝑡 + 𝑏𝑏2 𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝑡𝑡 𝑥𝑥𝐹𝐹𝐹𝐹𝑖𝑖,𝑡𝑡
𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶
+ 𝑏𝑏3 𝐹𝐹𝐹𝐹𝑖𝑖,𝑡𝑡 + 𝑐𝑐1 𝐻𝐻𝐻𝐻𝐻𝐻𝐻𝐻𝑖𝑖,𝑡𝑡 + 𝑐𝑐2 𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼ℎ𝑖𝑖,𝑡𝑡 + 𝑐𝑐3 � �
𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆 𝑖𝑖,𝑡𝑡
𝑅𝑅&𝐷𝐷
+ 𝑐𝑐4 ln �1 + � + 𝐹𝐹𝐹𝐹𝐹𝐹𝐹𝐹 𝐹𝐹𝐹𝐹𝐹𝐹𝐹𝐹𝐹𝐹 𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸 +∈𝑖𝑖,𝑡𝑡
𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆 𝑖𝑖,𝑡𝑡
(5)

In this specification, we replace the market-wide variables—investor sentiment, economic

recession, and consumer sentiment—with a firm-specific variable, ExFin, which measures

the external financing the firm raises in year t. We continue to include other firm-level

controls and firm fixed effects in the model. 9 We report the estimates from Equation (5) in

Table 5. The estimated coefficient on external financing is positive and highly statistically

significant at 0.072 (t-statistic = 7.63). This indicates that a firm’s advertising expense is

9
We do not include macro factors (i.e., Sent, Contr and ICS) in Equation 5 is because external financing is itself
a function of these external factors, as depicted in our later analysis in Equation (6). Putting these macro factors
in Equation 5 would lead to multicollinearity problem and would not allow us to obtain an unbiased estimate of
the coefficients b1 and b2.

25
positively associated with the external financing it raises in the year. Interestingly, the

additional results in Columns (2) and (3) suggest that this positive association is stronger for

firms that face financial constraints (as the significant positive coefficients on the interaction

variables ExFin x WWIndex in Column (2) and ExFin x SAIndex in Column (3) show). This is

consistent with the idea that firms rely more on external financing for their advertising

expenditure when they face more constrains in terms of internal financial resources.

< Insert Table 5 about here >

Finally, we examine the association between external financing and investor sentiment by

running the following regression.

𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝑖𝑖,𝑡𝑡 = 𝑎𝑎0 + 𝑏𝑏1 𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑡𝑡 + 𝑏𝑏2 𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑡𝑡 𝑥𝑥𝐹𝐹𝐹𝐹𝑖𝑖,𝑡𝑡 + 𝑏𝑏3 𝐹𝐹𝐹𝐹𝑖𝑖,𝑡𝑡 + 𝑏𝑏4 𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝑡𝑡 + 𝑏𝑏5 𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝑡𝑡 𝑥𝑥𝐹𝐹𝐹𝐹𝑖𝑖,𝑡𝑡
𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶
+ 𝑏𝑏6 𝐼𝐼𝐼𝐼𝐼𝐼𝑡𝑡 + 𝑏𝑏7 𝐼𝐼𝐼𝐼𝐼𝐼𝑡𝑡 𝑥𝑥𝐹𝐹𝐹𝐹𝑖𝑖,𝑡𝑡 + 𝑐𝑐1 𝐻𝐻𝐻𝐻𝐻𝐻𝐻𝐻𝑖𝑖 ,𝑡𝑡 + 𝑐𝑐2 𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼ℎ𝑖𝑖,𝑡𝑡 + 𝑐𝑐3 � �
𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆 𝑖𝑖,𝑡𝑡
𝑅𝑅&𝐷𝐷
+ 𝑐𝑐4 ln �1 + � + 𝐹𝐹𝐹𝐹𝐹𝐹𝐹𝐹 𝐹𝐹𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖 𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸 +∈𝑖𝑖,𝑡𝑡
𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆 𝑖𝑖,𝑡𝑡
(6)

The coefficient on Sent in this equation captures the association between external

financing and sentiment. We report the results of estimating Equation (6) in Table 6. The

coefficient on stand-alone Sent is positive in all columns, though it is statistically significant

only in the last column. A positive coefficient indicates that firms raise more external fund

during periods of high sentiment. Interestingly though, the results in Columns (2) and (3)

suggest that this positive association is stronger for firms that face financial constraints (as

the significant positive coefficients on the interaction variables ExFin x WWIndex in Column

(2) and ExFin x SAIndex in Column (3) show). Thus, financially constrained firms raise

especially more external financing during periods of high sentiment. Taken together, these

three additional pieces of evidence support the building blocks that underpin our hypotheses.

26
 < Insert Table 6 about here >

5.2. Association between sentiment and advertising effectiveness

If managers increase or decrease advertising budgets with sentiment due to changes in the

availability of funds, then the effectiveness of adverting may also vary with sentiment. For

example, Steenkamp and Fang (2011) show that changes in advertising linked to recession-

related funding constraints increases the effectiveness of advertising during recessions.

During periods of low sentiment, when firms have difficulty raising external funds, managers

may be forced to pick only those marketing investments that yield the highest payoffs and

leave out the other more marginal investments. During high sentiment periods, in contrast,

raising money would not only be easy, but it would be available at favorable terms, relieving

managers of the need to prioritize their advertising needs.

The effectiveness of advertising for individual firms could also vary across periods of

high and low sentiment due to the aggregate behavior of the firms. Because most firms spend

less (more) on advertising during low (high) sentiment periods, this would lead to advertising

by individual firms to be more (less) effective during such periods due to lower (greater)

clutter and cheaper (expensive) rates by broadcasting channels. Therefore, the effectiveness of

advertising expenditure is likely to be greater during periods of low sentiment than during

periods of high sentiment. To explore this possibility, we examine the association between

investor sentiment and the effectiveness of advertising for firms in our sample, by measuring

how the effectiveness of advertising varies with sentiment using the following model.

ln(𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂 𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼)𝑖𝑖,𝑡𝑡+1
= 𝑎𝑎0 + 𝑏𝑏1 ln(𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴)𝑖𝑖,𝑡𝑡 +𝑏𝑏2 𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑡𝑡 + 𝑏𝑏3 ln(𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴)𝑖𝑖,𝑡𝑡 𝑥𝑥 𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑡𝑡
+ 𝑏𝑏4 𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝑡𝑡 + 𝑏𝑏5 ln(𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴)𝑖𝑖,𝑡𝑡 𝑥𝑥 𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝑡𝑡 + 𝑏𝑏6 𝐼𝐼𝐼𝐼𝐼𝐼𝑡𝑡
+ 𝑏𝑏7 ln(𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴)𝑖𝑖,𝑡𝑡 𝑥𝑥 𝐼𝐼𝐼𝐼𝐼𝐼𝑡𝑡 + 𝑐𝑐1 𝐻𝐻𝐻𝐻𝐻𝐻𝐻𝐻𝑖𝑖,𝑡𝑡 + 𝑐𝑐2 𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼ℎ𝑖𝑖,𝑡𝑡
+ 𝑐𝑐3 ln(𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶 )𝑖𝑖,𝑡𝑡 + 𝑐𝑐4 ln(1 + 𝑅𝑅&𝐷𝐷)𝑖𝑖,𝑡𝑡 + 𝐹𝐹𝐹𝐹𝐹𝐹𝐹𝐹 𝐹𝐹𝐹𝐹𝐹𝐹𝐹𝐹𝐹𝐹 𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸 +∈𝑖𝑖,𝑡𝑡
27
 (7)
The dependent variable is the natural log of operating profits in year t+1. The coefficient

on ln(Advertising), b1, captures the effectiveness of advertising. Of greater interest for us is

the coefficient on the interaction variable, ln(Advertising) x Sent, which indicates how the

effectiveness of advertising varies with sentiment. A negative value for the coefficient, b3,

would indicate that that the average advertising dollars are more (less) effective in increasing

profits during low (high) sentiment periods. We continue to add other control variables that

may affect profits such as capital expenditure and R&D, as well as firm fixed effects to

account for unobservable differences across firms.

In Table 7, we report the estimates of Equation (7). As expected, the coefficient on

ln(Advertising) is positive indicating that greater spending on advertising is associated with

greater profits. More importantly, the coefficient of the interaction term, ln(Advertising) x

Sent, is negative and statistically significant, hence, the effectiveness of advertising indeed

varies over time with sentiment and is greater (lower) during low (high) sentiment periods.

< Insert Table 7 about here >

6. Contribution and implications

Our research makes a significant conceptual contribution to the literature that explores the

interplay between firms' marketing decisions and conditions in the stock market. While much

of this literature examines the impact of marketing decisions on stock prices (Aksoy et al.,

2008; Anderson, Fornell, & Lehmann, 1994; Anderson, Fornell, & Mazvancheryl, 2004), we

extend this research by showing that stock market events can also have a significant influence

on marketing decisions (Chakravarty & Grewal, 2011; Mizik, 2010). Specifically, we show

that investor sentiment in the stock market induces marketing managers to change the level of

their advertising expenditures. We show that such changes in marketing expenditures are not

28
associated with substantial gains in profits. Marketing managers can gain by recognizing

these heuristics when making decisions about advertising expenditures.

We also show that investor sentiment has a unique influence on marketing decision-

making different from that of overall economic conditions, thus extending research by

Steenkamp and Fang (2011) and Srinivasan et al. (2011). While these prior studies focus on

macroeconomic conditions such as contractions and recessions, this study shows that

sentiment in financial markets also influences managers’ decisions. Marketing managers

should pay attention to stock market investor sentiment and not just macroeconomic

conditions when assessing the extent and effectiveness of advertising expenditures. In fact,

marketing managers may also consider investor sentiment when making other marketing

expenditure decisions, such as for R&D, hiring sales forces where appropriate, and so on.

We also make an important conceptual contribution to the literature that focusses on the

affordability method of advertising budgeting (e.g., Joseph & Richardson, 2002; Tellis,

1998). Much of this literature has viewed and operationalized the notion of affordability by

looking at the availability of internally generated funds, such as profits and cash flows. The

findings in our study suggest this may only provide an incomplete picture. Researchers need

to also consider the conditions in the financial markets and firms’ access to external funds to

properly assess the resources available to managers.

Besides these important conceptual contributions, our research can also help marketing

managers and investors gain a better understanding of the factors influencing managerial

decision-making on advertising expenditure, which accounts for a major portion of marketing

spends for most companies, especially in consumer businesses. It can also help perceptive

marketing managers improve the efficiency of their advertising expenditure. For example, our

finding that firms in general cut their advertising budgets during low sentiment periods offers

29
an opportunity to managers who are ready to go against the crowd. Low advertising by others

means less competitive clutter and better recall and recognition by consumers for companies

that maintain or increase their advertising during such times. Similarly, a general decline in

advertising could give managers the opportunity to extract better deals with advertising

agencies during low sentiment periods.

Our results caution marketing managers to prepare for the possibility that external

funding conditions might force them to cut their advertising budgets precisely when these

may be most effective in generating additional sales and enhancing the firms’ profits. Any

untimely cuts in marketing expenditure due to a sudden worsening of sentiment in the

financial markets may expose a firm to delays in the launch of new products or advertising

campaigns, which, in turn, may have an adverse impact on brand image and customer loyalty

in the long run, with potentially disastrous consequences for the firm's bottom line. In fact,

the adverse impact of such cuts in advertising spends may be even worse if a period of

negative investor sentiment coincides with or is followed by a period of positive consumer

sentiment, because having cut their advertising expenditure in the previous period, the firms

may not be in a position to capitalize on the buoyancy in the demand for their products and

services in the subsequent period that may be triggered by the rise in consumer sentiment. 10

10
There are a number of years in our sample where investor sentiment is below its time series mean whereas
consumer sentiment is above its time series mean. Some of the years that stand out in this regard are 1972, 1977,
2002 and 2003. These are the years in which investor sentiment was less than -0.5 (i.e., more than 0.5 standard
deviation below its mean of zero), but consumer sentiment was significantly above its mean of 85. To illustrate
our point, we pick the year 2002 and examine the behavior of firms in the industry ‘Food and Kindred Products’
(SIC code=20). The industry has 21 firms in 2002 in our sample. We note that the five firms who decided to cut
their advertising the most (including Sara Lee Corporation and Heinz), by an average of 62%, experience a
decline in their average operating profits of 6% over two-year period of 2002 and 2003. In contrast, the five
firms who decided to increase their advertising the most (including Dean Foods and Del Monte), by an average
of 121%, experience an increase in average operating profits of 34% over the same period. Thus, the importance
of increasing advertising during periods when investor sentiment is low but consumer sentiment is high.

30
 Conversely, our finding that firms in general increase their advertising budgets during

high sentiment periods may also offer an opportunity to some marketing managers. The

easier access to funding for their advertising budgets does not always imply that managers

should spend more on advertising. Higher advertising by all the players in the market would

only result in more competitive clutter with lower recall and recognition by consumers.

Similarly, a general increase in advertising could force managers to enter less favorable deals

with advertising agencies during high sentiment periods. Managers could therefore consider

whether it may be better to focus on their real needs, such as understanding changes in their

customers' needs through more consumer research and investing in developing new products

that satisfy these changing needs, which would give these firms a stronger competitive

advantage rather than simply spending more money on advertising. This would be especially

true if a period of positive investor sentiment coincides with negative consumer sentiment,

which may also lead to sub-optimal results because the negative consumer sentiment would

not allow the firms to generate additional consumer demand to justify their high advertising

expenditure in response to positive investor sentiment. Hence, marketing managers will be

well advised to keep an eye on both investor and consumer sentiments to ensure that their

advertising budgets are not out of sync with either of these.

Our findings also show that firms are more prone to fluctuations in their advertising

spends in reaction to investor sentiment in the stock market when they are more financially

constrained. This is an important finding for marketing managers whose firms face financial

constraints because it warns them of their greater susceptibility. Finally, by showing another

channel through which investor sentiment in the financial market distorts real managerial

decisions, our findings also imply that financial market regulators should play a more active

role in curbing extreme swings in sentiment, not only due to its direct impact on stock

markets, but also its indirect effect on the real economy and firm performance.

31
7. Limitations and future research

Our research is subject to a few limitations that future research can consider. First, before

AcSEC issued the SOP 93-7 rule in December 1993, firms had considerable flexibility in

deciding how much of their advertising expenditure to show as expenses and how much to

capitalize and show as an asset. Because our study uses longitudinal data to explore the

impact of sentiment on advertising expenditure over time, differences in the timing of when

firms report their advertising expenditure after it is incurred can introduce noise into our

empirical analysis. However, we expect only a minimal impact from any such noise because

we use a large dataset (42,192 firm-year observations) over a long period (1971-2010).

Second, some researchers argue that relative market share is a better indicator of

marketing performance because it shows how much better a firm is doing compared to its

competitors, unlike absolute market share, which only represents how well a firm is doing

(Mitchell & Singh, 1993). However, in this study, we focus on investor sentiment, which

affects the marketing decisions and performance of all firms. We thus operationalize

marketing effectiveness by looking at the impact of advertising on profits at an aggregate

level, in effect looking at how the whole profit pie increases due to more advertising by an

average firm. Future researchers should take notice of this distinction and include the

appropriate role of market share in their studies.

Third, we study the impact of investor sentiment on the advertising expenditure, but

marketing covers a broader range of activities, such as expenditure on the sales force,

promotional expenses, and new product development. Future research could therefore extend

our work by exploring the impact of investor sentiment on managerial decision-making in

these other domains. Fourth, our sample period does not cover recent years after 2010.

Finally, we consider financial constraints as a mediator in the relationship between investor

32
sentiment in the stock market and advertising decisions. We acknowledge that several

potential moderators exist, such as industry and corporate governance differences that could

affect the relationship. These and other issues provide fruitful avenues for future research.

References

Aaker, D. A., & Carman, J. M. (1982). Are You Overadvertising? Journal of Advertising
Research, 22(4), 57-70.
Aksoy, L., Cooil, B., Groening, C., Keiningham, T. L., & Yalçın, A. (2008). The long-term
stock market valuation of customer satisfaction. Journal of Marketing, 72(4), 105-122.

Albers, S. (2012). Optimizable and implementable aggregate response modeling for

marketing decision support. International Journal of Research in Marketing, 29(2),
111-122.

Anderson, E. W., Fornell, C., & Lehmann, D. R. (1994). Customer satisfaction, market share,
and profitability: Findings from Sweden. Journal of Marketing, 58(July), 53-66.

Anderson, E. W., Fornell, C., & Mazvancheryl, S. K. (2004). Customer satisfaction and
shareholder value. Journal of Marketing, 68(October), 172-185.

Baker, M., Stein, J., & Wurgler, J. (2003). When does the market matter? Stock prices and
the investment of equity-dependent firms. Quarterly Journal of Economics, 118(3),
969-1005.

Baker, M., & Wurgler, J. (2006). Investor sentiment and the cross-section of stock returns.
Journal of Finance, 61(4), 1645-1680.

Baker, M., & Wurgler, J. (2007). Investor sentiment in the stock market. Journal of
Economic Perspectives, 21(2), 129-152.

Baker, M., & Wurgler, J. (2012). Behavioral corporate finance: An updated survey. In G. M.
Constantinides, M. Harris & R. M. Stulz (Eds.), Handbook of the Economics of
Finance (Vol. 2): Elsevier Press.

Barberis, N., & Thaler, R. (2003). A survey of behavioral finance. Handbook of the
Economics and Finance, 1(B), 1053-1128.

Batra, R. (2009). Advertising Management, 5/e: Pearson Education India.

Baum, C., Schaffer, M., & Stillman, S. (2003). Instrumental variables and GMM: Estimations
and testing. The Stata Journal, 3(1), 1-31.

Belch, G. E., Belch, M. A., Kerr, G. F., & Powell, I. (2008). Advertising and promotion: An
integrated marketing communications perspective. Sydney: McGraw-Hill.

33
Blasko, V. J., & Patti, C. H. (1984). The Advertising Budgeting Practices of Industrial
Marketers. Journal of Marketing, 48(4), 104-110.

Chakravarty, A., & Grewal, R. (2011). The Stock market in the driver's seat! Implications for
R&D and marketing. Management Science, 57(9), 1594-1609.

Chirinko, R. S., & Schaller, H. (2001). Business fixed Investment and ‘bubbles’: The
Japanese case. American Economic Review, 91(3), 663-680.

Davidson, R., & MacKinnon, J. (1993). Estimation and inference in econometrics, Oxford
University Press.

Deleersnyder, B., Dekimpe, M. G., Steenkamp, J. B. E., & Leeflang, P. S. (2009). The role of
national culture in advertising's sensitivity to business cycles: An investigation across
continents. Journal of Marketing Research, 46(5), 623-636.

Dean, J. (1951). How much to spend on advertising. Harvard Business Review, 29(1), 65-74.
Dong, M. (2010). Mergers and Acquisitions. In H. K. Baker & J. Nofsinger (Eds.),
Behavioral Finance. Hoboken, NJ: Wiley.

Erickson, G., & Jacobson, R. (1992). Gaining comparative advantage through discretionary
expenditures: The returns to R&D and advertising. Management Science, 38(9), 1264-
1279.

Fama, E. F. (1970). Efficient capital markets: A review of theory and empirical work. The
Journal of Finance, 25(2), 383-417.

Fischer, M., & Himme, A. (2016). The financial brand value chain: How brand investments
contribute to the financial health of firms. International Journal of Research in
Marketing, 34(1), 137-153.

Frankenberger, K. D., & Graham, R. C. (2003). Should firms increase advertising

expenditures during recessions? Marketing Science Institute Report, (03-003), 65-85.

Gervais, S. (2010). Capital Budgeting and Other Investment Decisions. In H. K. Baker & J.
Nofsinger (Eds.), Behavioral Finance. Hoboken, NJ: Wiley.

Hadlock, C. J., & Pierce, J. R. (2010). New evidence on measuring financial constraints:
Moving beyond the KZ index. Review of Financial Studies, 23(5), 1909-1940.

Hanssens, D. M., Wang, F., & Zhang, X.-P. (2016). Performance growth and opportunistic
marketing spending. International Journal of Research in Marketing, 33(4), 711-724.

Hodrick, R. J., & Prescott, E. C. (1997). Postwar US business cycles: an empirical

investigation. Journal of Money, credit, and Banking, 29(1), 1-16.

Hribar, P., & Mcinnis, J. (2012). Investor sentiment and analysts' earnings forecast errors.
Management Science, 58(2), 293-307.

Jones, J. P. (1990). Ad spending: maintaining market share. Harvard Business Review, 68(1),
38-42.

34
Joseph, K., & Richardson, V. J. (2002). Free cash flow, agency costs, and the affordability
method of advertising budgeting. Journal of Marketing, 66(1), 94-107.

Kamber, T. (2002). The brand manager's dilemma: Understanding how advertising

expenditures affect sales growth during a recession. Journal of Brand Management,
10(2), 106-120.

Lamont, O., & Stein, J. (2006). Investor sentiment and corporate finance: Micro and macro.
American Economic Review, 96(2), 147-151.

Lee, D. Y. (1994). The impact of firms' risk-taking attitudes on advertising budgets. Journal
of Business Research, 31(2-3), 247-256.
Lemmon, M., & Portniaguina, E. (2006). Consumer confidence and asset prices: Some
empirical evidence. Review of Financial Studies, 19(4), 1499-1529.

Lilien, G. L., Kotler, P., & Moorthy, K. S. (1992). Marketing Models. . NJ, USA: Prentice
Hall.
Lodish, L. M., Abraham, M., Kalmenson, S., Livelsberger, J., Lubetkin, B., Richardson, B.,
& Stevens, M. E. (1995). How T.V. Advertising Works: A Meta-Analysis of 389 Real
World Split Cable T.V. Advertising Experiments. Journal of Marketing Research,
32(2), 125-139.
Luo, M., Jiang, D. X., & Cai, J. (2014). Investor sentiment, product features, and advertising
investment sensitivities. Asia-Pacific Journal of Financial Studies, 43(6), 798-837.

Martín-Herrán, G., McQuitty, S., & Sigué, S. P. (2012). Offensive versus defensive
marketing: What is the optimal spending allocation? International Journal of
Research in Marketing, 29(2), 210-219.

McLean, R. D., & Zhao, M. (2014). The business cycle, investor sentiment, and costly
external finance. The Journal of Finance, 69(3), 1377-1409.

Mian, G. M., & Sankaraguruswamy, S. (2012). Investor sentiment and stock market response
to earnings news. Accounting Review, 87(4), 1357-1384.

Mitchell, L. A. (1993). An examination of methods of setting advertising budgets: practice

and the literature. European Journal of Marketing, 27(5), 5-21.

Mitchell, W., & Singh, K. (1993). Death of the lethargic: Effects of expansion into new
technical subfields on performance in a firm's base business. Organization Science,
4(2), 152-180.

Mizik, N. (2010). The theory and practice of myopic management. Journal of Marketing
Research, 47(4), 594-611.

Mizik, N., & Jacobson, R. (2007). Myopic marketing management: Evidence of the
phenomenon and its long-term performance consequences in the SEO context.
Marketing Science, 26(3), 361-379.

Patti, C. H., & Blasko, V. J. (1981). Budgeting Practices of Big Advertisers. Journal of
Advertising Research, 21(6), 23-29.
35
Petersen, M. A. (2009). Estimating standard errors in finance panel data sets: Comparing
approaches. Review of Financial Studies, 22(1), 435-480.

Polk, C., & Sapienza, P. (2009). The stock market and corporate investment: A test of
catering theory. Review of Financial Studies, 22(1), 187-217.

Prendergast, G., West, D., & Shi, Y.-Z. (2006). Advertising budgeting methods and processes
in China. Journal of Advertising, 35(3), 165-176.

Rasmussen, A. (1952). The determination of advertising expenditure. Journal of Marketing,

16(4), 439-446.

Rossiter, J. R. (2012). Advertising management principles are derived mostly from logic and
very little from empirical generalizations. Marketing Theory, 12(2), 103-116.

Shiller, R. (2003). From efficient markets theory to behavioral finance. Journal of Economic
Perspectives, 17(1), 83-104.

Shleifer, A. (2003). Inefficient markets: An introduction to behavioral finance. New York:

Oxford University Press.

Shleifer, A., & Vishny, R. (2003). Stock market driven acquisitions. Journal of Financial
Economics, 70(3), 295-311.

Simon, H. (1984). Challenges and new research avenues in marketing science. International
Journal of Research in Marketing, 1(4), 249-261.

Soberman, D. A. (2009). Marketing agencies, media experts and sales agents: Helping
competitive firms improve the effectiveness of marketing. International Journal of
Research in Marketing, 26(1), 21-33.

Srinivasan, R., Lilien, G. L., & Sridhar, S. (2011). Should firms spend more on research and
development and advertising during recessions? Journal of Marketing, 75(May), 49-
65.

Srinivasan, R., Rangaswamy, A., & Lilien, G. L. (2005). Turning adversity into advantage:
Does proactive marketing during a recession pay off? International Journal of
Research in Marketing, 22(2), 109-125.

Srinivasan, S., & Hanssens, D. M. (2009). Marketing and firm value: Metrics, methods,
findings, and future directions. Journal of Marketing Research, 46(3), 293-312.

Stambaugh, R., Yu, J., & Yuan, Y. (2012). The short of it: Investor sentiment and anomalies.
Journal of Financial Economics, 104(2), 288-302.

Stambaugh, R., Yu, J., & Yuan, Y. (2015). Arbitrage asymmetry and the idiosyncratic
volatility puzzle. Journal of Finance, 70(5), 1903-1948.

Steenkamp, J.-B. E. M., & Fang, E. E. (2011). The impact of economic contractions on the
effectiveness of R&D and advertising: Evidence from U.S. companies spanning three
decades. Marketing Science, 30(4), 628-645.

36
Stein, J. (1996). Rational capital budgeting in an irrational world. Journal of Business, 69(4),
429-455.

Tellis, G. J. (1998). Advertising and Sales Promotion Strategy. Reading, MA: Addison-
Wesley.

Whited, T., & Wu, G. (2006). Financial constraints risk. Review of Financial Studies, 19(2),
531-559.

Yu, J., & Yuan, Y. (2011). Investor sentiment and the mean-variance relation. Journal of
Financial Economics, 100(2), 367-381.

37
 Figure 1
Number of Firms Included in our Sample by Year
This figure reports the number of firms included in our sample for each year. The period spans 1971-2010. We
include all firms that report annual advertising expenditure as per COMPUSTAT dataset, subject to the availability
of all the controls variables.

2,000

1,800

1,600

1,400

1,200
# of Firms

1,000

800

600

400

200

Year

38
 Table 1
Correlations
This table reports the correlations. All variables are described in Appendix.

Advertising Sentiment Contr (x -1) ICS

Advertising 1
Sentiment 0.346 1
Contr (x -1) 0.336 0.002 1
ICS 0.625 0.222 0.612 1

39
 Table 2
Summary Statistics
This table reports the summary statistics for the variables used in this study. Panel A reports the statistics for the
annual time series of investor sentiment and other macroeconomics variables that are available at aggregate level.
Panel B reports the statistics for advertising intensity and other variables that are available at firm or industry
level. The sample spans 40 years over 1971-2010 and includes 54,209 firm-year observations for most variables.
All variables are described in Appendix.

Standard
Variables # of obs. Mean deviation Minimum Maximum
Panel A: Macro variables (yearly observations)
Sentiment Variables:
Sent 40 -0.05 0.92 -2.24 1.79
Control variables:
Contr 40 1.29 2.34 0 8.79
ICS 40 85.01 12.07 63.75 107.58

Panel B: Firm or industry level variables (firm-year observations)

Advertising variables:
Advertising (in $ million) 54,209 34 116 0.009 871
Advertising Intensity (i.e.,
advertising scaled by avg. sales) 54,209 0.035 0.048 0.000 0.299

Control and other variables:

Herf 54,209 0.093 0.090 0.025 0.546
IndGrowth 54,209 0.106 0.087 -0.143 0.351
Capex (in $ million) 54,209 62 209 0.010 1,566
Capex (scaled by avg.sales) 54,209 0.073 0.117 0.001 0.796
R&D (in $ million) 54,209 18 78 0.000 631
R&D (scaled by avg.sales) 54,209 0.043 0.107 0.000 0.749
WWIndex 54,209 -0.235 0.116 -0.510 0.025
SAIndex 54,209 -2.997 0.764 -4.687 -0.976
ExFin (scaled by avg.sales) 54,209 0.235 0.792 -0.756 5.543
CF (scaled by avg. sales) 54,209 -0.007 0.400 -2.896 0.386
Operating Income after
Depreciation (in $ million) 42,447 160 493 0.001 3,419

40
 Table 3
Average Advertising Intensity across Periods of High and Low Investor Sentiment
This table reports the average values of advertising intensity for firm-year observations sorted into periods of high
and low sentiment. The sample consists of 54,209 firm-year observations over the period 1971-2010. High
sentiment periods are those when sentiment is above the mean, whereas low sentiment periods are those when
sentiment is below the mean. The t-statistic reported in parentheses in the last row is based on firm- and year-
clustered standard errors (in order to correct for the firm- and year-related autocorrelations and heteroscedasticity).
***, ** and * represent the 1%, 5% and 10% significance levels, respectively.

Advertising (scaled by avg. sales)

Sentiment # of Firm-Year Obs. Avg

High (Sent > 0) 20,543 0.038

Low (Sent < 0) 21,649 0.032

High - Low 0.006***

(t-statistic) (14.76)

41
 Table 4
Investor Sentiment and Advertising Expenditure – Regression Analysis
This table reports the estimates of Equation (4). Variable definitions are provided in Appendix. The t-statistics are
reported in parentheses and are based on standard errors that are clustered by year. We do not report the intercepts
as the inclusion of firm fixed effects in the model make them difficult to interpret. ***, ** and * represent the 1%,
5% and 10% significance levels, respectively.

Dependent Variable: ln(Advertising Intensity)

FC = WWIndex FC = SAIndex
Explanatory Variables (1) (2) (3)
ln(Advertising Intensity) t-1 0.590*** 0.589*** 0.584***
(36.10) (36.13) (36.26)
Sent 0.017*** 0.036*** 0.068***
(2.72) (2.95) (3.16)
Sent x FC 0.072** 0.015**
(2.05) (2.38)
FC 0.205 0.107*
(0.68) (1.98)
Contr -0.004 -0.004 -0.007
(-0.93) (-0.58) (-0.73)
Contr x FC -0.001 -0.001
(-0.06) (-0.35)
ICS 0.002*** 0.002* 0.001
(4.09) (1.70) (0.38)
ICS x FC 0.000 -0.000
(0.01) (-0.62)
Herf 0.091 0.073 0.035
(1.27) (1.05) (0.52)
IndGrowth 0.214*** 0.186** 0.154**
(2.96) (2.45) (2.08)
Capex/Sales 0.783*** 0.785*** 0.763***
(12.05) (12.19) (11.94)
ln(1+R&D/Sales) 0.928*** 0.915*** 0.937***
(7.25) (7.17) (7.42)

Fixed effects Firm Firm Firm

Cluster by Year Year Year

# of firm-year obs. 49,387 49,387 49,387

Adjusted R-squared 0.871 0.871 0.871

42
 Table 5
External Financing and Advertising Expenditure – Regression Analysis
This table reports the estimates of Equation (5). Variable definitions are provided in Appendix. The t-statistics are
reported in parentheses and are based on standard errors that are clustered by year. We do not report the intercepts
as the inclusion of firm fixed effects in the model make them difficult to interpret. ***, ** and * represent the 1%,
5% and 10% significance levels, respectively.

Dependent Variable: ln(Advertising Intensity)

FC = WWIndex FC = SAIndex
Explanatory Variables (1) (2) (3)

ln(Advertising Intensity) t-1 0.593*** 0.592*** 0.589***

(36.70) (36.83) (36.78)
ExFin 0.072*** 0.139*** 0.210***
(7.63) (8.18) (6.69)
ExFin x FC 0.249*** 0.045***
(5.64) (4.93)
FC 0.119 0.052**
(1.10) (2.70)
Herf 0.055 0.036 0.005
(0.86) (0.58) (0.08)
IndGrowth 0.325*** 0.313*** 0.301***
(3.16) (3.05) (3.11)
Capex/Sales 0.659*** 0.620*** 0.615***
(10.52) (10.05) (10.25)
ln(1+R&D/Sales) 0.745*** 0.615*** 0.654***
(6.30) (5.21) (5.56)

Fixed effects Firm Firm Firm

Cluster by Year Year Year

# of firm-year obs. 49,387 49,387 49,387

Adjusted R-squared 0.871 0.872 0.872

43
 Table 6
Investor Sentiment and External Financing – Regression Analysis
This table reports the estimates of Equation (6). Variable definitions are provided in Appendix. The t-statistics are
reported in parentheses and are based on standard errors that are clustered by year. We do not report the intercepts
as the inclusion of firm fixed effects in the model make them difficult to interpret. ***, ** and * represent the 1%,
5% and 10% significance levels, respectively.
Dependent Variable: ExFin
FC = WWIndex FC = SAIndex
Explanatory Variables (1) (2) (3)
Sent 0.005 0.015 0.047**
(0.81) (1.33) (2.13)
Sent x FC 0.077** 0.015**
(1.99) (2.50)
FC 1.078 0.140
(0.94) (1.31)
Contr 0.002 -0.013 -0.022
(0.44) (-1.37) (-1.52)
Contr x FC -0.065 -0.008
(-1.33) (-1.52)
ICS 0.003** -0.004 -0.004
(2.22) (-1.39) (-1.12)
ICS x FC -0.025* -0.002*
(-2.02) (-1.85)
Herf 0.058 0.131 0.096
(0.73) (1.51) (1.15)
IndGrowth 0.348*** 0.477*** 0.384***
(2.79) (3.61) (2.95)
Capex/Sales 2.093*** 2.072*** 2.104***
(13.61) (13.77) (13.60)
ln(1+R&D/Sales) 2.842*** 2.923*** 2.845***
(9.16) (9.44) (9.22)

Fixed effects Firm Firm Firm

Cluster by Year Year Year

# of firm-year obs. 54,209 54,209 42,191

Adjusted R-squared 0.526 0.530 0.460

44
 Table 7
Investor Sentiment and Advertising Effectiveness – Regression Analysis
This table reports the estimates of Equation (7). Variable definitions are provided in Appendix. The t-statistics are
reported in parentheses and are based on standard errors that are clustered by year. We do not report the intercepts
as the inclusion of firm fixed effects in the model make them difficult to interpret. ***, ** and * represent the 1%,
5% and 10% significance levels, respectively.

Explanatory Variables Dependent Variable:: ln(Operating Income) t+1

ln(Advertising) 0.220***
(5.80)
Sent -0.019
(-0.88)
ln(Advertising) x Sent -0.013***
(-2.85)

Contr 0.003
(0.28)
ln(Advertising) x Contr 0.003*
(1.85)

ICS 0.000
(0.01)
ln(Advertising) x ICS 0.001
(1.66)

Herf -0.262
(-1.25)
IndGrowth -0.184
(-1.06)

ln(Capex) 0.244***
(20.88)
ln(1+R&D) 0.165***
(9.75)

Fixed Effects Firm

Cluster by Year

# of firm-year obs. 42,266

Adjusted R-squared 0.834

45
 Appendix A
Description of the Variables
This appendix provides detailed description of all the variables used in the paper. The two sentiment variables—ICS and Sent—are computed at the market-level for each year.
The sources of these data are noted in the table. Herfindahl-Hirschman index, Industry growth and Industry sales growth are computed at the level of SIC two-digit industry.
All remaining variables are computed at firm level for each year. The data for the firm- and industry-level variables come from S&P COMPUSTAT Annual file.

Variable Name Symbol Description

A1: Variables employed in our study
Advertising Advertising Dollar amount of annual advertising expenditure (COMPUSTAT item ‘xad’)
Advertising intensity Advertising Annual advertising expenditure divided by the average sales in the contemporaneous and preceding years, as
Intensity depicted in Equation (1)
Capital expenditure Capex Annual capital expenditure (COMPUSTAT item ‘capx’)
Economic recession Contr A measure of economic recession constructed by Steenkamp and Fang (2011). We closely follow the procedure
in Steenkamp and Fang (2011) to construct this measure using the filter and the cyclical component of GDP
(Hodrick & Prescott, 1997). Specifically, Contr takes the value of zero when the economy is growing at or above
its long-term trend, but equals the magnitude of the decline in the cyclical component of GDP when the economy
is in recession.
External finance ExFin External financing of firm in year t, computed by taking the change in total assets (COMPUSTAT item ‘at’) from
year t-1 to t, and subtracting from it the change in retained earnings (COMPUSTAT item ‘re’) over the same
period. This definition encompasses all types of external financing, including net new stock issue, net new bond
issues, short- and long-term borrowing from banks, and so on. We scale external financing by average sales in
year t-1 and t to make it comparable to the other firm-level variables we report.
Financial constraints FC Financial constraints, which alternately is measured by WW Index and SA index. It is computed for each firm-for
each year.
Herfindahl- Herf Herfindahl-Hirschman index is a commonly used measure of industry concentration and competitiveness. It is
Hirschman index calculated by squaring the market share of each firm competing in an SIC 2-digit industry, and then summing the
resulting numbers across all firms in the industry. The higher the index, the more concentrated, and less
competitive, the industry. Theoretically, the value of the index can range from close to zero (extremely
competitive industry) to 10,000 (complete monopoly—only one firm operating in the industry).
Index of consumer ICS Index of consumer sentiment reported by University of Michigan. It is compiled based on telephone interviews of
sentiment US households, and reflects consumer attitudes towards economy and planned purchases. ICS is now available on
a monthly frequency, but it was available on a quarterly basis before 1978. We therefore use its quarterly
observations over our entire sample period, and take the average of the four calendar quarters to compute the
annual value of the index. The data can be downloaded from http://www.sca.isr.umich.edu/tables.html.

46
Variable Name Symbol Description
Industry growth IndGrowth Average growth rate of sales for an SIC 2-digit industry. We first compute the sales growth for each firm by
taking the log difference in sales (COMPUSTAT item ‘sale’) over successive years, and then take the mean
of the individual firms’ growth rates in the industry.
Investor sentiment index Sent Index of investor sentiment in the stock market developed by Baker and Wurgler (2006), and downloaded
from the web site of Jeffrey Wurgler (http://people.stern.nyu.edu/jwurgler/). The index combines six
individual proxies of investor sentiment, namely closed-end fund discount, NYSE share turnover, number of
IPOs, first day returns on IPOs, share of equity issues in total debt and equity issues, and dividend premium
(the log difference of the average market-to-book ratio of payers and non-payers). To isolate investor
sentiment from economic fundamentals, each of these six proxies is first regressed on several
macroeconomic variables. The first principal component of the residuals of these six regressions constitutes
the investor sentiment index, standardized to yield a mean of zero and a standard deviation of one.
Operating income Operating Operating income after depreciation (COMPUSTAT item ‘oiadp’)
Income
Research and development R&D Annual research and development expenditure (COMPUSTAT item ‘xrd’)
expenditure
Sales Sales Annual net sales (COMPUSTAT item ‘sale’)

A2: Variables related to WW Index:

Cash flows CF Cash flows computed as net income before extraordinary items plus depreciation and amortization expense
(COMPUSTAT item ‘ib’ + COMPUSTAT item ‘dp’)
Dummy for dividends paid DIVPOS DIVPOS is a dummy variable that takes the value of one if the firm pays any dividend on preferred or
common stocks, and zero otherwise. Specifically, DIVPOS=1 if COMPUSTAT item ‘dvc’ or ‘dvp’ > 0; and
otherwise DIVPOS=0)
Firms sales growth SG Growth in a firm’s sales, computed by subtracting one from sales in year t divided by sales in year t-1.
Industry sales growth ISG Average sales growth of all the firms in the industry, based on SIC 3-digit industry classification. Sales
growth is first computed for each firm by subtracting one from sales in year t divided by sales in year t-1.
The mean across all firms in the industry is then used to represent the industry growth.
Long term debt to total LTD Ratio of long-term debt (COMPUSTAT item ‘dltt’) to book value of total assets (COMPUSTAT item ‘at’).
assets
Total assets TA Natural log of book value of total assets (COMPUSTAT item ‘at’)
Whited and Wu (2006) WWIndex Index of financial constraints constructed by Whited and Wu (2006). It is based on six inputs, namely cash
Index flows, dividends, book value of assets, firm’s sales growth, ratio of long-term debt to assets, and industry
sales growth. The computation is based on Equation (2). The weights for the inputs are taken from Whited
and Wu (2006, Equation 13 & Table 1).

47
Variable Name Symbol Description

A3: Variables related to SA Index:

Firm age Age Number of years since the firm first appears in COMPUSTAT dataset.
Hadlock and Pierce (2010) SAIndex Index of financial constraints constructed by Hadlock and Pierce (2010). It is based on two inputs, namely,
Index inflation-adjusted total assets and firm age. The computation is based on Equation (3). The weights for the
inputs are taken from Hadlock and Pierce (2006, Table 6, Column 2).
Total inflation-adjusted ATA Natural log of inflation-adjusted total assets (i.e., natural log of COMPUSTAT item ‘at’ divided by CPI
assets inflation deflator), as in Hadlock and Pierce (2010).

48