Brand Equity in Good and Bad Times:

What Distinguishes Winners from Losers in CPG Industries?

Koushyar Rajavi

Tarun Kushwaha

Jan-Benedict E.M. Steenkamp

Koushyar Rajavi (Email: koushyar.rajavi@scheller.gatech.edu; Voice: 404-894-7773) is an

Assistant Professor of Marketing at Scheller College of Business at Georgia Institute of
Technology. Tarun Kushwaha (Email: tkushwah@gmu.edu; Voice: 979-422-9710) is Professor of
Marketing at School of Business at George Mason University, and Jan-Benedict E.M. Steenkamp
(Email: jbs@unc.edu; Voice: 919-962-9579) is C. Knox Massey Distinguished Professor of
Marketing at the Kenan-Flagler Business School of the University of North Carolina.
Acknowledgments: The authors acknowledge help from AiMark and Kantar Worldpanel for
providing data for this study. They also acknowledge kind help of Hannes Datta, Maarten J.
Gijsenberg and Molly Borden. Authors acknowledge the constructive feedback received from the
Editor-in-Chief, the Associate Editor, and three anonymous reviewers.

1
 Brand Equity in Good and Bad Times:
What Distinguishes Winners from Losers in CPG Industries?

Abstract

We examine why some brands are able to ride the wave of macroeconomic expansions, while

other brands are better able to successfully weather contractions. Using a utility-based framework,

we develop hypotheses how the impact of these shocks on brand equity is moderated by six

strategic brand factors—price positioning, advertising spending, product line length, distribution

breadth, brand architecture, and market position. We utilize monthly data on 325 CPG national

brands in 35 categories across 17 years from the United Kingdom to obtain quarterly sales-based

brand equity estimates. The two pre-eminent brand factors are distribution and assortment.

Distribution is by far the most important factor in contractions. It is also the most important factor

in expansions. In short, in good times and bad times, extensively distributed brands win. In

expansions, a wide assortment is also a very strong contributor to brand equity, while it does not

destroy brand equity in contractions. We further find that advertising spending, premium price

positioning, umbrella branding structure, and market leadership matter in either expansions and/or

contractions, the magnitude of their effects on brand equity is relatively modest. We conclude with

managerial implications.

Keywords: Brand Management, Brand Equity, Sales-Based Brand Equity, Macroeconomic

Fluctuations, Brand Positioning, Contractions, Expansions

1
The ups and downs of macroeconomic cycles provide brand managers with opportunities to

grow their brand or insulate it from harm. In economic contractions, consumers have lower

disposable incomes and hence face tighter budgets. This makes them more price sensitive

(Gordon, Goldfarb, and Li 2013; van Heerde et al. 2013), less brand loyal (Pointer Media

Network 2009), and more inclined to shift their purchases to (cheaper) private labels (Lamey et

al. 2007, 2012; Scholdra et al. 2022). The opposite effects occur in good times. Consumers

change their cross-category consumption behavior across the business cycle (e.g., Deleersnyder

et al. 2004; Du and Kamakura 2008), but we know little about how do business cycles affect

different brands within a category? Dekimpe and Deleersnyder (2018, p. 54) raise this issue as an

important research question: “Are all brands equally affected?” We examine this issue for brand

equity, regarded by academics and practitioners alike as a key performance metric of a brand

(Aaker 1991; Datta, Ailawadi, and van Heerde 2017; Millward Brown 2017).

There is a rich literature on the effects of various marketing mix instruments on brand equity

(e.g., Ailawadi, Lehmann, and Neslin 2003; Sriram, Balachander, and Kalwani 2007; Yoo,

Donthu, and Lee 2000). This important work is short-term, tactical in scope. Academics

recommend that brand equity be built and maintained for the long run, using the various

elements of the marketing mix (Aaker 1991; Lodish and Mela 2007). In this paper we

complement previous work by adopting a strategic perspective. Our perspective is that of the

firm – and in particular brand management – that uses the marketing mix not only tactically, but

also strategically to position the brand vis-à-vis its competitors. We examine six strategic brand

factors: price positioning, advertising spending, line length, distribution breadth, brand

architecture, and market position.

2
 The purpose of this paper is to examine how brands with different positioning along these six

strategic brand factors are more or less able to weather economic shocks. Our research straddles

two important research streams—drivers of brand equity and the effects of macroeconomic

conditions. Our contribution is twofold. First, we adopt a strategic view on the effect of

managerial decisions on brand equity, by focusing on the role of strategic brand factors. Second,

we examine how and to what extent the effects of these strategic brand factors differs

systematically between expansions and contractions. The context in which we test our

hypotheses is consumer packaged goods (CPG) in the United Kingdom. We estimate the effect

of business cycles on brand equity of brands with different characteristics utilizing data on 325

CPG national brands in 35 categories across 17 years.

Background Literature

Macroeconomic Fluctuations

There is a rich and growing marketing literature on the effects of macroeconomic fluctuations on

marketing-related phenomena. Past research (see Web Appendix A for a summary) shows that

macroeconomic fluctuations influence consumers’ category preferences (Kamakura and Du

2012), budget allocation (Du and Kamakura 2008), purchase of durable goods (Deleersnyder et

al. 2004), shopping frequency, and purchase volume (Ma et al. 2011; Scholdra et al. 2022). They

also affect brands’ price elasticity (Gordon, Goldfarb, and Li 2013; van Heerde et al. 2013),

advertising effectiveness (e.g., Srinivasan, Lilien, and Sridhar 2011; Steenkamp and Fang 2011;

van Heerde et al. 2013), R&D effectiveness (Srinivasan, Lilien, and Sridhar 2011; Steenkamp

and Fang 2011), and marketing conduct over the macroeconomic fluctuations (Lamey et al.

2012). Consumers switch more frequently to private labels in downturns (Lamey et al. 2007,

2012; Scholdra et al. 2022). For the most part, past research did not focus on examining

3
customers’ heterogeneous behaviour across different types of brands and investigating why some

brands fare better than others during different economic conditions. We extend this body of

research by considering the link between strategic brand factors and brand equity during

expansions and contractions.

Brand Equity

A widely used definition of brand equity is the value added by the brand name to a product

(Farquhar 1989). The two basic approaches to operationalizing the value added to the products

by its brand name are consumer mindset metrics and market outcomes (Datta, Ailawadi, and van

Heerde 2017). The first approach is known as consumer-based brand equity (CBBE) and is

grounded in metrics such as awareness, attachment, and attitudes towards the brand. The second

approach, sales-based brand equity (SBBE), is based on market outcomes that can be attributed

to the brand, such as price, volume, or revenue premia (Ailawadi, Lehmann, and Neslin 2003;

Datta, Ailawadi, and van Heerde 2017; Sriram, Balachander, and Kalwani 2007). Extant research

has shown that SBBE and CBBE are positively related, but that the magnitude of the correlation

is modest, around .3 (Datta, Ailawadi, and van Heerde 2017), because what consumers think and

feel is far from perfectly aligned with what they actually do (Sheppard, Hartwick, and Warshaw

1988). Our interest is in SBBE, as it is the ability of managers to generate superior market

performance that provides the ultimate justification for spending money on branding activities.

Past research has examined the effect of marketing mix activities on SBBE. Ailawadi,

Lehmann, and Neslin (2003) and Sririam, Balachander, and Kalwani (2007) found that

advertising had a positive effect on brand equity while promotion had no effect. Sriram and

colleagues further found that innovation activity increased the equity of toothpaste brands, but

not for dish detergent brands. Previous research did not examine the heterogeneity in SBBE

4
across brand characteristics in different economic conditions. We extend research on SBBE by

examining how equity of different types of brands are affected during the business cycles.

Research Framework and Hypotheses

Overview of Theoretical Framework

Figure 1 presents the research framework that guided our study. In our framework, we include

six strategic brand factors: price positioning (value vs. premium), advertising spending (low vs.

high), line length (short vs. long), distribution breadth (selective vs. extensive), brand

architecture (single-category vs. umbrella-category branding strategy), and market position

(follower vs. leader). Strategic brand factors are sticky but not fixed over time. For example, it is

possible to change the price positioning of the brand, if desired, but such a change should only be

executed gradually. You cannot change a value brand to a premium brand in the short run. The

same applies to the other strategic brand factors.

These six factors tap into the three components of brand image as identified by Keller (1993).

In Keller’s theory of customer-based brand equity, strong brands elicit strong, favorable

associations that are unique. Keller’s work has inspired various brand consultancies to propose

their own branding models. These models share broadly speaking the same components, albeit

they use different labels. In our work, we adopt Kantar’s BrandZ model because it is closest to

Keller’s original work. Kantar (2021) identifies three pillars of strong brands– differentiation

(akin to Keller’s uniqueness), meaningfulness (favorability), and salience (strength). According

to Kantar (2021, p. 13), differentiation refers to the brand being distinct from others. Price and

advertising are among the key strategic factors contributing to brand differentiation (Mela,

Gupta, and Jedidi 1998). Meaningful brands meet people’s heterogeneous needs and make

people feel emotionally connected to the brand. Line length (multiple SKUs to meet varying

5
consumer needs; Ataman, Mela, and van Heerde 2008) and advertising (to create emotional

connections; Aaker 1991) are strategic brand factors that contribute to meaningfulness. Finally,

salient brands are brands that come to mind quickly in purchase situations. Distribution and

market position contribute to brand salience, as does ubiquity across product categories

(umbrella brand) (Bronnenberg, Mahajan, and Vanhonacker 2000; Sharp 2010).

Figure 1: Research Framework

6
Brand Utility Framework

We examine whether and how the effect of macroeconomic conditions on brand equity plays out

differently depending on these six strategic brand factors through the lens of multiattribute

decision making under uncertainty and informational constraints (Meyer 1981; Pras and

Summers 1978). We draw upon Pras and Summers (1978), Erdem and Keane (1996), and

Erdem, Zhao, and Valenzuela (2004) and propose that the utility consumers derive from a brand

attribute l (Ul) depends on the brand’s perceived score on attribute l (Xl) and the importance of

attribute l to consumers (ωl), as well as the uncertainty about the ability of the brand to deliver

attribute l (σl), weighed by consumers’ tolerance for risk for that attribute l (rl): Ul = ωlXl – rlσl. 1

We assume that consumers on average are risk averse (Erdem, Zhao, and Valenzuela 2004; van

Ewijk, Gijsbrechts, and Steenkamp 2022), and thus rlσl captures the disutility from risk

associated with uncertainty in attribute delivery.

We distinguish between functional (tangible) and emotional/self-expressive (intangible)

attributes (Aaker 1996; Erdem, Zhao, and Valenzuela 2004; Myers and Shocker 1981).

Functional attributes refer are related to the tangible functions performed by the brand.

Emotional/self-expressive attributes refer to the intangible feelings the brand provides to

consumers and what the consumption of the brand tells others about the kind of person I am. We

aggregate across functional and emotional attributes and risks, and include the disutility of price.

Thus, the utility brand i provides can be expressed as a function of five elements:

𝑈𝑈𝑖𝑖 = −𝛼𝛼𝑃𝑃𝑖𝑖 + 𝝎𝝎𝑓𝑓 𝑿𝑿𝑓𝑓,𝑖𝑖 − 𝒓𝒓𝑓𝑓 𝝈𝝈𝑓𝑓,𝑖𝑖 + 𝝎𝝎𝑒𝑒 𝑿𝑿𝑒𝑒,𝑖𝑖 − 𝒓𝒓𝑒𝑒 𝝈𝝈𝑒𝑒,𝑖𝑖

Disutility Utility Disutility Utility Disutility

of price from from from from
functional functional emotional emotional
attributes risk attributes risk

1
Our development is for the aggregate consumer; hence we do not have a consumer subscript.

7
where 𝛼𝛼 is the price sensitivity, ωf and ωe are the importance attached to functional and

emotional attributes, rf and re denote the risk aversion for functional and emotional attributes, Xf,i

and Xe,i represent the vector of brand i’s perceived scores on the functional and emotional

attributes, respectively, and σf,i and σe,i indicate uncertainty about attribute delivery. We neither

claim to break new ground in utility theory nor will we estimate the different components

specified in the utility equation. Rather, we use this utility framework as a heuristic for

hypotheses development.

We propose that the relative importance of price, functional attributes and risks, and

emotional attributes and risks vary across the business cycle (i.e., change in magnitudes of 𝛼𝛼, 𝝎𝝎,

and 𝒓𝒓’s during the business cycle). In contractions, with tight budgets, consumers have lower

willingness to pay, hence αCON > αEXP, and thus the disutility for a given level of price will be

greater during contractions (Lamey et al. 2007; van Heerde et al. 2013). During contractions,

different motivational orientations are triggered than during expansions (Scholdra et al. 2022).

Contractions induce avoidance motivation and negative economic sentiments, while expansions

trigger approach motivation and positive economic sentiment (Millet, Lamey, and Van den

Bergh 2012). Hedonic attributes which trigger approach motivation are associated with

emotional attributes, while utilitarian attributes which trigger avoidance motivation are

associated with functional attributes (Higgins 2006; Tamir, Chiu, and Gross 2007). Conversely,

in expansions, incomes are on the rise and budgetary restrictions are less tight. Now, the

consumer has the opportunity to focus more on relevant emotional attributes (Lamey et al. 2012).

Relatedly, Kamakura and Du (2012) find that consumers’ share of expenditures on positional

goods (i.e., goods that people use to convey their relative standing within society) increases in

expansions. This is in line with “hierarchy of needs” (Maslow 1943); with more budgetary

8
restrictions in economic contractions, consumers are expected to prioritize their basic

physiological attributes over their social and self-actualization needs (Kamakura and Du 2012).

Thus, we expect the utility weights associated with functional attributes and risks to be greater

during contractions (ωf,CON > ωf,EXP and rf,CON > rf,EXP) and those associated with emotional

attributes and risks to be greater during expansions (ωe,EXP > ωe,CON and re,EXP > re,CON).

Predictions

We use these insights to develop hypotheses about the role of the six strategic brand factors in

moderating the effect of the business cycle on brand equity.

Price Positioning. Following van Heerde et al. (2013), we distinguish between value brands

and premium brands. Value brands are lower priced and utilitarian in scope (Steenkamp 2014).

They are positioned on tangible attributes, providing high value because they offer reasonable

quality for a low price. Premium brands cost more and offer better quality and excel on

emotional attributes (Aaker and Joachimsthaler 2000). Premium brands cost more (PPRM > PVAL),

but are also higher on functional and emotional attributes vis-à-vis value brands (Xf,PRM > Xf,VAL

and Xe,PRM > Xe,VAL; Steenkamp 2014). Premium brands also reduce consumers’ purchase risk.

Price premium is associated with reduction in uncertainty and greater trust (Ba and Pavlou 2002)

and higher incentives to provide consistent quality (Klein and Leffler 1981). Thus, premium

brands will have lower functional risk than value brands (σf,PRM < σf,VAL). In expansions,

emotional considerations gain importance (ωe,EXP > ωe,CON) (Millet, Lamey, and Van den Bergh

2012), which benefits premium brands. Therefore:

H1EXP: In expansions, premium brands perform better on brand equity than value
brands.

In contractions, both price (αCON > αEXP; Lamey et al. 2007; van Heerde et al. 2013) and

functional utility (ωf,CON > ωf,EXP) attain greater importance, while functional risk aversion

9
increases as well (rf,CON > rf,EXP) (Millet, Lamey, and Van den Bergh 2012). These forces are

contradictory. Value brands benefit from lower disutility of price but are hurt by lower functional

utility and higher functional risk. Because of the opposing forces, we refrain from proposing a

formal hypothesis for price positioning’s role in contractions. 2

Advertising Spending. We distinguish between low and high advertising spender brands.

Economists (e.g., Klein and Leffler 1981; Kihlstrom and Riordan 1984) derived analytically that

advertising expenditure is positively related to product quality. This confirms the old dictum that

it does not make sense to advertise a bad product. Kirmani and Wright (1989) showed

empirically that high advertising expenditure is perceived by consumers as an indicator of

marketing effort, which is a clue to the marketer’s confidence in product quality. Consequently,

high advertising spender brands should be perceived by consumers as being higher on functional

utility (Xf,HI-AD > Xf,LO-AD), which suggests that in contractions, they fare better on brand equity

than low advertising spender brands, given that functional attributes weigh more heavily in bad

times (ωf,CON > ωf,EXP). Advertising is a major marketing instrument to imbue a brand with

emotions and to communicate the emotional attributes to consumers (Aaker 1996). Thus, we

expect that the high advertising spender brands deliver more emotional utility (Xe,HI-AD > Xe,LO-

AD) and that consumers have a clearer idea about the emotional attributes delivered by high

advertising spender brands (σe,HI-AD < σe,LO-AD). This suggests that in economic expansions, when

the emotional attributes (ωe,EXP > ωe,CON) and disutility from emotional risks are higher (re,EXP >

re,CON), high ad spender brands do better on brand equity than low advertising spender brands.

H2EXP: In expansions, high advertising spender brands perform better on brand equity
than low advertising spender brands.
H2CON: In contractions, high advertising spender brands perform better on brand equity
than low advertising spender brands.

2
Web Appendix B reports the impact of each affected component on the utility function.

10
 Line Length. Line length refers to the number of SKUs offered by a brand in a category.

The more SKUs a brand carries, the more difficult it is for consumers to accurately gauge

their respective qualities. Consumers may be exposed to varieties (e.g., taste) about which

they have little idea. In a recent study, van Ewijk, Gijsbrechts, and Steenkamp (2022)

document that adding new SKUs has a ‘dark side’ as it increases consumer uncertainty about

quality of the brand. This suggests that longer line length is associated with higher functional

risk: σf,LNG > σf,SHR. This means that in contractions, when risk aversion is higher (rf,CON >

rf,EXP), the higher disutility from functional risk disadvantages longer line length brands

versus shorter line length brands.

Brands that carry a wider assortment are able to more closely meet the heterogeneous

needs of consumers (Nevo 2001) and allow consumers to choose the product that aligns best

with their psycho-social values. This is likely to lead increase consumer perceptions of

emotional attributes (Xe,LNG > Xe,SHR), which is more highly valued in expansions (ωe,EXP >

ωe,CON), leading to higher emotional utility for brands with a longer line length. Thus, we

propose:

H3EXP: In expansions, brands with longer line length perform better on brand equity than
brands with shorter line length.
H3CON: In contractions, brands with shorter line length perform better on brand equity
than brands with longer line length.

Distribution. Wider distribution is a key factor to market success of CPG brands (Ataman,

van Heerde, and Mela 2010; Srinivasan, Vanhuele, and Pauwels 2010). Although Klein and

Leffler (1981) focus on advertising as brand-specific marketing program investment, their

analytical conclusions apply to any kind of observable brand-name expenditures (Milgrom and

Roberts 1986, pp. 799-800), including distribution (Rao and Mahi 2003). Consumers interpret a

brand’s ubiquitous presence as a sign of its consistent performance across different markets.

11
Extensive distribution costs, associated with high expenditures on slotting allowances, in-store

promotion material, and other expensive retail investments would be lost if the brand does not

deliver on its promises (Rao and Mahi 2003). Thus, extensively distributed brands score higher

on functional attributes than selectively distributed brands (Xf,EXT > Xf,SEL). Additionally, there

are more stores where the brand can be bought, which offers opportunities to buy the brand for a

lower price, which suggests that the disutility of price is lower for extensively distributed brands

(PEXT < PSEL). This suggests that extensively distributed brands should perform better in

contractions than selectively distributed brands. Extensive distribution further contributes to

brand trust (Rajavi, Kushwaha, and Steenkamp 2019), which has been shown to correlate with

brand affect (Chaudhuri and Holbrook 2001, p. 89). Brand affect is brand’s potential to elicit

positive emotional response. This suggests that extensively distributed brands perform better on

emotional attributes in the minds of consumers than selectively distributed brands (Xe,EXT >

Xe,SEL), and as such, are valued more during expansions. Thus:

H4EXP: In expansions, brands with extensive distribution breadth perform better on brand
equity than brands with selective distribution breadth.
H4CON: In contractions, brands with extensive distribution breadth perform better on
brand equity than brands with selective distribution breadth.

Brand Architecture. We distinguish between umbrella brands and single-category brands

(Erdem 1998; Erdem and Sun 2002). Umbrella branding helps consumers in cross-category

learning which helps the umbrella brand in transferring favorable brand associations from one

category to another (Erdem and Chang 2012). Firms that adopt umbrella branding have more

incentives (vis-à-vis single category brands) to maintain and improve quality of their offerings as

they face greater risk of poor-quality attribution (Montgomery and Wernerfelt 1992; Erdem

1998; Miklós-Thal 2012): Xf,UMB > Xf,SIN and σf,UMB < σf,SIN. As functional considerations weigh

heavily in contractions, we expect umbrella brands to perform better in bad times than single-

12
category brands. However, umbrella branding strategy also has risks associated with it. Umbrella

brands may be forced to adopt a uniform brand positioning strategy across many categories,

while relevant emotional associations may differ across categories. This suggests that emotional

risks are higher for umbrella brands: σe,UMB > σe,SIN. As emotional aspects matter more in

expansions (Millet, Lamey, and Van den Bergh 2012), the negative impact of emotional risk will

reduce utility for umbrella brands more than single-category brands:

H5EXP: In economic expansions, single-category brands perform better on brand equity than
umbrella brands.
H5CON: In economic contractions, umbrella brands perform better on brand equity than
single-category brands.

Market Position. Here we distinguish between whether the brand is a leader versus a follower

in the category. Aaker (2007, p. 17) maintains that “the most influential exemplars [of leader

brands] will be those that are perceived to be superior in terms of quality, performance, and

reliability.” Market leader brands have greater incentives to maintain higher quality and meet the

brand’s promise as financial consequences of failure are much larger for them (Milgrom and

Roberts 1986): Xf,LEA > Xf,FOL , which benefits leader brands especially in contractions.

What about emotional payoff ? On the one hand, it has been argued that brands with

dominant market position might generate more positive emotions because of the bandwagon

effect – the pleasure that consumers have from using a product when more people are using it

(Hellofs and Jacobson 1999; Edeling and Himme 2018), and the “fitting in” effect that enhances

consumers’ sense of belonging to a larger social group (van Herpen, Pieters, and Zeelenberg

2009). On the other hand, it has been argued that using popular and well-known brands might

decrease consumers’ emotional utility because of the loss of exclusivity effect: “consumers feel

worse about the product and perhaps even themselves (through loss of image) when the brand

they are using is popular” (Hellofs and Jacobson 1999, p. 18). Thus, leading brands may or may

13
not be more favorably perceived on emotional attributes than their follower counterparts. Given

the competing theoretical arguments, we refrain from hypothesizing for market position’s effect

during expansions:

H6CON: In contractions, market leader brands perform better on brand equity than
follower brands.

Method

Our empirical strategy consists of two general steps: 1) estimating brand equity using the sales-

based brand equity (SBBE) approach, and 2) explaining heterogeneity in the SBBE estimates

using strategic brand factors (SBFs) and their interactions with macroeconomic expansions and

contractions. Following Datta, Ailawadi, and van Heerde (2017) and Sriram, Balachander, and

Kalwani (2007), we operationalize SBBE using the brand intercept method, where, after

accounting for marketing mix investments and tangible product characteristics, what is left in the

brand intercept is a measure of the ability to leverage the brand to generate sales. In the first step,

we follow Datta, Ailawadi, and van Heerde (2017), and estimate quarterly brand intercepts using

a model with marketing activities and product attributes of the focal brands, and other control

variables as predictors, and brand volume market share as the dependent variable. In Step 2, we

use six SBFs (i.e., Price Positioning, Ad Spending, Distribution Breadth, Line Length, Brand

Architecture, and Market Position), as well as their interactions with the magnitude of

macroeconomic expansions and contractions to explain the variation in the quarterly brand

intercepts, i.e., brand equity estimates. 3

We investigate our hypotheses in the context of CPG categories in the UK. We acquired UK

household scanner panel data from Kantar Worldpanel for 35 CPG categories. The monthly

3
We acknowledge that the estimation can alternatively be done in one stage. However, the shared variance between
marketing mix instruments and strategic brand factors is likely to lead to severe collinearity issues.

14
brand-level data covers 17 years from January-1994 to November-2010 (203 months) and

includes information on marketing conduct and performance of national brands in each CPG

category. We retained all brands that satisfied the following two conditions: a) non-zero sales in

at least 95% of the months during the data window, and b) average monthly volume market share

exceeding 0.1%. Our resulting sample consists of 325 national brands. We complement our data

with monthly ad expenditures for brands in our sample which we get from Nielsen Media UK. 4

Step 1: Estimating SBBE

We follow Datta, Ailawadi, and van Heerde (2017) and use market share attraction model

(Cooper and Nakanishi 1988) at the monthly level to estimate SBBE at brand-quarter level. We

specify a model that allows for heterogeneous brand-specific coefficients (Gielens 2012; Datta,

Ailawadi, and van Heerde 2017). Market share of brand i in category j during month t is

expressed as the attraction of that brand (Aijt) relative to the aggregate attractions of the Ij brands

in category j during month t (Ij represents number of brands in product category j):
𝐴𝐴𝑖𝑖𝑖𝑖𝑖𝑖
(1) 𝑀𝑀𝑆𝑆𝑖𝑖𝑖𝑖𝑖𝑖 = 𝐼𝐼𝑗𝑗
∑𝑘𝑘=1 𝐴𝐴𝑘𝑘𝑘𝑘𝑘𝑘

where MSijt is the market share of brand i in category j during month t. We specify attraction of

each brand as a function of brand-quarter dummies (i.e., SBBE estimates), marketing mix

instruments (advertising stock, regular price, price promotion depth, product line length, and

distribution intensity), and product attributes. 5 To control for state dependence in market share,

we also include lagged market share as a regressor in the model (Gielens 2012). By including

4
Our dataset is similar to the data used by van Heerde et al. (2013). Two notable differences are: 1) whereas van
Heerde et al. (2013) examine leading national brands (average of 4.1 brands in a category), our analysis covers a
broader set of national brands, with an average of 9.3 brands in each category, and 2) we had to drop two product
categories (dry soup and peanut butter) because we could only identify two national brands that satisfied our
selection criteria and with only two brands it was not possible to estimate the market share attraction model.
5
We provide category-specific summary of market shares in Web Appendix C, marketing mix instruments in Web
Appendix D, and product attributes in Web Appendix E.

15
Gaussian copulas, we account for potential endogeneity of marketing mix variables that might

arise due to unobservables that are not accounted for in our model (Park and Gupta 2012; Datta,

Ailawadi, and van Heerde 2017; Datta et al. 2022; Papies, Ebbes, and van Heerde 2017): 6, 7

(2) 𝐴𝐴𝑖𝑖𝑖𝑖𝑖𝑖 = exp (∑𝑄𝑄

𝑞𝑞=1 𝛼𝛼𝑖𝑖𝑖𝑖𝑖𝑖 ∗ 𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝑡𝑡𝑡𝑡 + 𝛽𝛽𝑖𝑖𝑖𝑖1 𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑖𝑖𝑖𝑖𝑖𝑖 + 𝛽𝛽𝑖𝑖𝑖𝑖2 𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑖𝑖𝑖𝑖𝑖𝑖

+𝛽𝛽𝑖𝑖𝑖𝑖3 𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑖𝑖𝑖𝑖𝑖𝑖 + 𝛽𝛽𝑖𝑖𝑖𝑖4 𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑖𝑖𝑖𝑖𝑖𝑖 + 𝛽𝛽𝑖𝑖𝑖𝑖5 𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑖𝑖𝑖𝑖𝑖𝑖 + 𝛽𝛽𝑖𝑖𝑖𝑖6 𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑖𝑖𝑖𝑖𝑖𝑖−1

𝑛𝑛
𝑗𝑗
+ ∑𝑎𝑎=1 𝛾𝛾𝑎𝑎𝑎𝑎𝑎𝑎 𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎 + ∑5𝑐𝑐=1 𝛿𝛿𝑐𝑐𝑐𝑐𝑐𝑐 𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐 + 𝜀𝜀𝑖𝑖𝑖𝑖𝑖𝑖 )

where q denotes quarters and DUMQTRtq represents quarterly dummies and hence its coefficient

(αijq) holds brand- and quarter-specific intercepts. ADSTOCKijt, PRICEijt, PROMOijt, LLijt, and

DISTijt represent advertising stock, regular price, price promotion depth, product line length, and

distribution by brand i in category j during month t, respectively, and ATTRaijt (a=1…nj)

represents different product attributes, which are defined separately for each category (see Web

Appendix E). Operationalization of the variables used in the first stage are presented in Table 1. 8

Model Estimation. The attraction model for each product category j can be written as a

system of Ij equations that is estimated simultaneously using seemingly unrelated regression

(SUR). After substituting Equation (2) into Equation (1), the system of equations can be

linearized and normalized by first taking its logarithm, followed by using either of the two

approaches discussed by Cooper and Nakanishi (1988): 1) normalizing with respect to a base

brand (base-brand approach), or 2) normalizing by centering (log-centering approach). The two

approaches are equivalent (Cooper and Nakanishi 1988) and we use the latter. Finally, we

6
For example, our model does not account for feature/display activity of brands or their slotting allowances. In case
such variables that are not observed in our model are correlated with the predictors in the model, if we do not
account for endogeneity, our estimates might be biased.
7
A necessary identification requirement for the Gaussian copula approach is non-normality of the endogenous
regressors. Using Shapiro-Wilk tests, normality of all five log-transformed marketing mix instruments were strongly
rejected at .01 level, hence allowing us to specify Gaussian copulas.
8
We tested the stationarity of the variables included in our first-stage model using Levin-Lin-Chu and Fisher-type
panel unit root tests. Across both tests, the null of presence of unit root was strongly rejected (p<.001).

16
 Table 1: Variables in Step 1
Construct Var. Name Operationalization
Volume Market MSijt Monthly brand volume market share for brand i in category j at
Share month t.
Brand Attraction Aijt Attraction of brand i in category j in month t
Sales-based Brand αijq Brand- and quarter- specific intercepts for brand i in category j
Equity (SBBE) at quarter q.
Quarter Dummies DUMQTRtq Quarterly time indicator which gets a value of 1 if month t is in
quarter q and 0 otherwise.
Advertising Stock ADSTOCKijt Advertising stock of brand i in category j in month t, where
ADSTOCKijt = λjADSTOCKijt-1 + (1-λj)ADijt and ADijt is monthly
advertising expenditures, adjusted by yearly consumer price
index in the UK, by brand i in category j in month t. The
smoothing parameter (λj) is determined separately for each
product category based on a grid search on the interval of [0,.9]
in increments of .1 (we report smoothing parameters [λj] of
different categories in Web Appendix F).
Regular Price PRICEijt Regular price of brand i in category j at month t, adjusted by
yearly consumer price index in the UK. Regular price
operationalized based on average price of a brand over a six-
month moving window (Gielens 2012).
Price Promotion PROMOijt 1 – (average paid price by consumers for brand i in category j in
month t / regular price of brand i in category j in month t);
higher values indicate deeper price discounts offered by the
brand. 9
Product Line Length LLijt the number of stock-keeping units (SKUs) offered by brand i in
category j at month t.
Distribution DISTijt Percentage of UK retailers that sold brand i's SKUs during
month t, weighted by retailer’s volume market share in the
category j in month t.
Product Attributes ATTRaijt Fraction of SKUs of brand i in category j that have a certain
product attribute at month t of year y. Quantity and nature of
(a=1…nj)
product attributes vary across the 35 product categories. nj
represents the number of attributes in category j; at most 9
attributes are defined for a category. Attributes for different
categories are listed in Web Appendix E.
Gaussian Copula COPULAcijt Five control functions based on the method proposed by Park
Control Functions (c=1…5) and Gupta (2012) for the five potentially endogenous marketing
mix instruments.

9
In our data, we only observe paid price. Using Gielens’ (2012, p. 412) approach, we decomposed paid price into
regular price (average price level of a brand defined over a six-month moving window) and price promotion depth
(the same approach has also been used in Geyskens, Gielens, and Gijsbrechts 2010). We thank an anonymous
reviewer for this suggestion.

17
estimate this system of seemingly unrelated equations using Feasible Generalized Least Square:

𝑀𝑀𝑆𝑆
(3) 𝑖𝑖𝑖𝑖𝑖𝑖
𝑙𝑙𝑙𝑙 � 𝑀𝑀𝑆𝑆
�
� = ∑𝑄𝑄 ′ ������������������
𝑞𝑞=1�𝛼𝛼𝑖𝑖𝑖𝑖𝑖𝑖 � ∗ 𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝑡𝑡𝑡𝑡 + 𝛽𝛽𝑖𝑖𝑖𝑖1 �𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑖𝑖𝑖𝑖𝑖𝑖 − 𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝐾𝐾𝚥𝚥𝚥𝚥 �
𝚥𝚥𝚥𝚥

�������������
+𝛽𝛽𝑖𝑖𝑖𝑖2 �𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑖𝑖𝑖𝑖𝑖𝑖 − 𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙 ���������������
𝚥𝚥𝚥𝚥 � + 𝛽𝛽𝑖𝑖𝑖𝑖3 �𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑖𝑖𝑖𝑖𝑖𝑖 − 𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝚥𝚥𝚥𝚥 �

��������
+𝛽𝛽𝑖𝑖𝑖𝑖4 �𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑖𝑖𝑖𝑖𝑖𝑖 − 𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙 �����������
𝚥𝚥𝚥𝚥 � + 𝛽𝛽𝑖𝑖𝑖𝑖5 �𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑖𝑖𝑖𝑖𝑖𝑖 − 𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝚥𝚥𝚥𝚥 �
𝑛𝑛𝑗𝑗
������������
+𝛽𝛽𝑖𝑖𝑖𝑖6 (𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑖𝑖𝑖𝑖𝑖𝑖−1 − 𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙 �����������
𝚥𝚥𝚥𝚥−1 ) + ∑𝑎𝑎=1 𝛾𝛾𝑎𝑎𝑎𝑎𝑎𝑎 (𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎 − 𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝑎𝑎𝑎𝑎𝑎𝑎 )

��������������� ′ 10
+ ∑5𝑐𝑐=1 𝛿𝛿𝑐𝑐𝑐𝑐𝑐𝑐 (𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐 − 𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝑐𝑐𝑐𝑐𝑐𝑐 ) + 𝜀𝜀𝑖𝑖𝑖𝑖𝑖𝑖

′
In the above equation, 𝛼𝛼𝑖𝑖𝑖𝑖𝑖𝑖 is our brand- and quarter-specific equity estimates that we will use in

′
the second stage (hereinafter, we refer to 𝛼𝛼𝑖𝑖𝑖𝑖𝑖𝑖 as SBBEijq). 11

Estimation Results. Table 2 reports the weighted mean marketing mix elasticities across all

325 brands (for category-specific results see Web Appendix F). All elasticities have the expected

sign and their meta-analytic Z-statistics (Rosenthal 1991) are significant. We find a small but

significant mean advertising elasticity (.0149), close to .0021 in van Heerde et al. (2013). Our

mean price elasticity (-.8895) is smaller in magnitude than the -1.4266 reported by van Heerde et

al. (2013). However, van Heerde et al.’s (2013) elasticities are based on absolute sales rather

than market share. 12 Bijmolt, van Heerde, and Pieters (2005, Table 2) report that price elasticities

based on market share are on average .32 smaller in absolute magnitude than price elasticities

based on sales. The weighted average price promotion elasticity of .1966 is in line with .146

reported by Srinivasan, Vanhuele, and Pauwels (2010). Mean distribution elasticity of .3392 is

consistent with .40 of Datta, Ailawadi, and van Heerde (2017) and .368 of Datta et al. (2022).

Finally, our mean elasticity for line length (.6396) is in the range of values (from .348 to 1.511)

reported by Jindal et al. (2020) and comparable with .459 reported by Datta et al. (2022).

10
Our exposition follows Cooper and Nakanishi’s (1988) Equation (2.13). 𝑀𝑀𝑆𝑆 � 𝚥𝚥𝚥𝚥 is the geometric mean of MSjt. The
bar operator (𝑋𝑋�) represents arithmetic mean.
11 ′
𝛼𝛼𝑖𝑖𝑖𝑖𝑖𝑖 is technically 𝛼𝛼𝑖𝑖𝑖𝑖𝑖𝑖 − ����.
𝛼𝛼𝚥𝚥𝚥𝚥 Thus, our brand- and quarter-specific SBBE estimates are relative to the category’s
′
average SBBE. Similarly, 𝜀𝜀𝑖𝑖𝑖𝑖𝑖𝑖 is 𝜀𝜀𝑖𝑖𝑖𝑖𝑖𝑖 − ����.
𝜀𝜀𝚥𝚥𝚥𝚥
12
Our analysis also covers more brands in each product category.

18
 Table 2: Marketing Mix Elasticities Estimates
Marketing Instrument Elasticities Meana
Brand Advertising (Ad Stock) .0149**
Brand Regular Price -.8895***
Brand Price Promotion Depth .1966***
Brand Line Length .6396***
Brand Distribution .3392***
*p <.10; ** p <.05; *** p <.01. a Weighted means across 325 brands in 35 categories, with weights being the
inverse of the estimated standard errors. Significance tests are based on meta-analytic Z-values.

We illustrate quarterly SBBE estimates for some brands across four product categories

(Figure 2). As it can be seen, Mr Muscle, Lavazza, and Wilkinson are consistently valuable

brands in the UK. Some brands (e.g., Sensodyne) have experienced considerable growth over

years, while other brands have declined over time (e.g., Ajax, Mentadent), and others remained

fairly stable (e.g., Douwe Egberts, Cif). We report category-specific statistics on SBBE scores in

Web Appendix G.

Step 2: Explaining the Dynamics of Brand Equity

Operationalizing Business Cycles. We use quarterly data on inflation-adjusted gross domestic

product per capita (GDPPC) from UK’s Office for National Statistics to extract macroeconomic

fluctuations. We follow past research (e.g., Lamey et al. 2007, 2012) and adopt time-series

filtering to extract the cyclical component of (log-transformed) macroeconomic fluctuations

(lnGDPPCqcyc; see Web Appendix H for details). Following van Heerde et al. (2013), we use

lnGDPPCqcyc and define the magnitude of expansions (contractions) as the difference between

the actual level of the cyclical component of the macroeconomic fluctuations at quarter q and the

prior trough (peak):

𝑐𝑐𝑐𝑐𝑐𝑐 𝑐𝑐𝑐𝑐𝑐𝑐 𝑐𝑐𝑐𝑐𝑐𝑐
𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝐶𝐶𝑞𝑞 − �𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝 𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡ℎ 𝑖𝑖𝑖𝑖 𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝐶𝐶𝑞𝑞 � ; 𝑖𝑖𝑖𝑖 ∆𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝐶𝐶𝑞𝑞 >0
(4) EXPq = � 𝑐𝑐𝑐𝑐𝑐𝑐
0 ; 𝑖𝑖𝑖𝑖 ∆𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝐶𝐶𝑞𝑞 ≤0
𝑐𝑐𝑐𝑐𝑐𝑐
0 ; 𝑖𝑖𝑖𝑖 ∆𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝐶𝐶𝑞𝑞 >0
(5) CONq = � 𝑐𝑐𝑐𝑐𝑐𝑐 𝑐𝑐𝑐𝑐𝑐𝑐 𝑐𝑐𝑐𝑐𝑐𝑐
�𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝 𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝 𝑖𝑖𝑖𝑖 𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝐶𝐶𝑞𝑞 � − 𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝐶𝐶𝑞𝑞 ; 𝑖𝑖𝑖𝑖 ∆𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝐶𝐶𝑞𝑞 ≤0

19
 Figure 2: Sales-Based Brand Equity Estimates in Four Product Categories
2.5 Ground/Bean Coffee 2.5 Household Cleaners
Mr Muscle
1.5 Lavazza 1.5
Cif
0.5

Estimated SBBE
Douwe Egbert 0.5
Estimated SBBE

-0.5
Flash
-0.5
-1.5

-1.5 Ajax
-2.5 Rombouts

-3.5 -2.5

1994

1995

1996

1997

1998

1999

2000

2001

2002

2003

2004

2005

2006

2007

2008

2009

2010
1994

1995

1996

1997

1998

1999

2000

2001

2002

2003

2004

2005

2006

2007

2008

2009

2010
2 Razor Blades Toothpaste Sensodyne
2

1 Wilkinson Sword Colgate

1
Gillette

Estimated SBBE
Estimated SBBE

Bic 0
0
Oral B
-1

-1
-2
Personna Mentadent

-2 -3
1994

1995

1996

1997

1998

1999

2000

2001

2002

2003

2004

2005

2006

2007

2008

2009

2010
1994

1995

1996

1997

1998

1999

2000

2001

2002

2003

2004

2005

2006

2007

2008

2009

2010

* To avoid overcrowding the plots, we focus on a sample of 3-4 brands in each category.
20
EXPq (CONq) takes positive values during economic upturns (downturns) and 0 during

downturns (upturns). This operationalization allows us to capture the magnitude of expansions

and slowdowns, with the value of EXPq (CONq) capturing the percentage improvement (decline)

in the economy during expansions (contractions).

Model Specification. To examine how different strategic brand factors help (or hurt) brands

during expansions and contractions, we use the following model:

𝑚𝑚=7 𝑘𝑘 𝑚𝑚=14 𝑘𝑘
(6) 𝑆𝑆𝑆𝑆𝑆𝑆𝐸𝐸𝑖𝑖𝑖𝑖𝑖𝑖 = 𝛼𝛼0 + 𝛼𝛼1 𝑆𝑆𝑆𝑆𝑆𝑆𝐸𝐸𝑖𝑖𝑖𝑖𝑖𝑖−1 + ∑𝑚𝑚=2 𝛼𝛼𝑚𝑚 𝑆𝑆𝑆𝑆𝑆𝑆𝑖𝑖𝑖𝑖𝑖𝑖 + 𝛼𝛼8 𝐸𝐸𝐸𝐸𝐸𝐸𝑞𝑞 + ∑𝑚𝑚=9 𝛼𝛼𝑚𝑚 𝐸𝐸𝐸𝐸𝐸𝐸𝑞𝑞 ∗ 𝑆𝑆𝑆𝑆𝑆𝑆𝑖𝑖𝑖𝑖𝑖𝑖

𝑚𝑚=21 𝑘𝑘 𝑚𝑚=26 𝑙𝑙
+𝛼𝛼15 𝐶𝐶𝐶𝐶𝐶𝐶𝑞𝑞 + ∑𝑚𝑚=16 𝛼𝛼𝑚𝑚 𝐶𝐶𝐶𝐶𝐶𝐶𝑞𝑞 ∗ 𝑆𝑆𝑆𝑆𝑆𝑆𝑖𝑖𝑖𝑖𝑖𝑖 + ∑𝑚𝑚=22 𝛼𝛼𝑚𝑚 𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝑖𝑖𝑖𝑖𝑖𝑖

+ ∑𝐵𝐵1 𝜏𝜏𝑏𝑏 𝐵𝐵𝐵𝐵𝐵𝐵𝐵𝐵𝐵𝐵𝑏𝑏 + ∑𝑄𝑄 𝑌𝑌

1 𝛿𝛿𝑞𝑞 𝑄𝑄𝑄𝑄𝑄𝑄𝑄𝑄𝑄𝑄𝑄𝑄𝑄𝑄𝑞𝑞 + ∑1 𝛾𝛾𝑦𝑦 𝑌𝑌𝑌𝑌𝑌𝑌𝑌𝑌𝑦𝑦 + 𝜀𝜀𝑖𝑖𝑖𝑖𝑖𝑖
13

where i represents brands, j represents categories, and q represents quarters. We include lagged

brand equity (𝑆𝑆𝑆𝑆𝑆𝑆𝐸𝐸𝑖𝑖𝑖𝑖𝑖𝑖−1 ) as an independent variable to allow for inertia in brand equity (Sriram,

Balachander, and Kalwani 2007). 𝑆𝑆𝑆𝑆𝑆𝑆𝑖𝑖𝑖𝑖𝑖𝑖

𝑘𝑘
(k=1…6) represents the six strategic brand factors:

Price Positioning (value vs. premium; PRICEijq), Ad Spending (low vs. high; ADijq), Line Length

(short vs. long; LLijq), Distribution Breadth (selective vs. extensive; DISTijq), Brand Architecture

(single- vs. umbrella-category branding; ARCHij), and Market Position (follower vs. leader;

POSijq). The operationalization for the six SBFs, five control (𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝑖𝑖𝑖𝑖𝑖𝑖
𝑙𝑙
, l= 1…5), as well as

other variables used in Step 2 are presented in Table 3. In operationalizing the first four SBF

variables we use marketing mix activities of brands in the four quarters preceding the current

time period. Using a four-quarter rolling window increases the stability of our measures across

time, which is consistent with the nature of strategic factors, as they are unlikely to be transient

in the near term. The temporal separation also reduces endogeneity concerns as brand managers

13
We tested the stationarity of the dependent variable using different panel unit root tests. Across all of the tests, the
null of presence of unit root was strongly rejected (p<.001).

21
 Table 3: Variables Used in Step 2
Construct Var. Name Operationalization
Sales-Based SBBEijq Estimated portion of quarterly brand volume market share that is not explained
Brand Equity by its marketing activities, product attributes, and other control variables in the
first stage.
Expansion EXPq Magnitude of expansion as the difference between cyclical GDP per capita and
the prior trough.
Contraction CONq Magnitude of contraction as the difference between cyclical GDP per capita and
the prior peak.
Strategic Brand 𝑘𝑘
𝑆𝑆𝑆𝑆𝑆𝑆𝑖𝑖𝑖𝑖𝑖𝑖 1
• Price Positioning (value vs. premium; 𝑆𝑆𝑆𝑆𝑆𝑆𝑖𝑖𝑖𝑖𝑖𝑖 = PRICEijq): Whether brand i's
Factors (k=1…6) average paid price in the four quarters before current time period is above
average of other brands in category j (=.5; premium) or not (=-.5; value).
2
• Ad Spending (low vs high; 𝑆𝑆𝑆𝑆𝑆𝑆𝑖𝑖𝑖𝑖𝑖𝑖 = ADijq): Whether brand i's average ad
expenditure in the four quarters before current period is above average of other
brands in category j (=.5) or not (=-.5).
3
• Line Length (short vs. long; 𝑆𝑆𝑆𝑆𝑆𝑆𝑖𝑖𝑖𝑖𝑖𝑖 = LLijq): Whether brand i's average line
length in the four quarters before current time period is above average of other
brands in category j (=.5) or not (=-.5).
4
• Distribution Breadth (selective vs. extensive; 𝑆𝑆𝑆𝑆𝑆𝑆𝑖𝑖𝑖𝑖𝑖𝑖 = DISTijq): Whether
brand i's average distribution intensity in the four quarters before current time
period is above average of other brands in category j (=.5; extensive) or not
(=-.5; selective).
• Brand Architecture (single-category vs. umbrella branding; 𝑆𝑆𝑆𝑆𝑆𝑆𝑖𝑖𝑖𝑖5 = ARCHij):
Whether brand i is offered in multiple categories (=.5; umbrella brand) or in
one category (=-.5; single-category brand).
6
• Market Position (follower vs. leader; 𝑆𝑆𝑆𝑆𝑆𝑆𝑖𝑖𝑖𝑖𝑖𝑖 = POSijq): Whether the brand is
among the top quartile of its category in terms of average market share in the
four quarters before current time (=.5; leader) or not (=-.5; follower).

Marketing 𝑙𝑙
𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝑖𝑖𝑖𝑖𝑖𝑖 1
Other brands’ quarterly paid price (𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝑖𝑖𝑖𝑖𝑖𝑖 =OTHERSPPijq): (Log-
Activity of (l=1…4) transformed) average brand paid price in category j, excluding focal brand i, in
Other National quarter q.
Brands 2
Other brands’ quarterly advertising (𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝑖𝑖𝑖𝑖𝑖𝑖 =OTHERSADijq): (Log-
transformed) average brand ad expenditures in category j, excluding focal brand
i, in quarter q.
3
Other brands’ quarterly line length (𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝑖𝑖𝑖𝑖𝑖𝑖 =OTHERSLLijq): (Log-
transformed) average brand line length in category j, excluding focal brand i, in
quarter q.
4
Other brands’ quarterly distribution intensity (𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝑖𝑖𝑖𝑖𝑖𝑖 =OTHERSDISTijq):
(Log-transformed) average brand distribution in category j, excluding focal
brand i, in quarter q.
Private Label 5
𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝑖𝑖𝑖𝑖𝑖𝑖 (=PLMSjq) category’s total private label volume market share in category j,
Market Share averaged across months in quarter q.

22
are unlikely to be able to accurately forecast the state of the economy several quarters in advance

and hence adjust their SBF-affecting actions in anticipation of the macroeconomic shock.

In Equation (6), α2 - α7 capture the main effect of SBF variables on SBBE; i.e., general

differences in SBBE due to the SBFs, irrespective of economic conditions. It should be noted

that main effect of ARCH is not identified since it is a time-invariant characteristic and hence the

effect is subsumed within brand fixed effects. α8 (and α15) hold the main effect of

macroeconomic expansions (contractions) on SBBE. α9 - α14 (α16 - α21) capture how equity of

brands with different SBFs are affected differentially during expansions (contractions). Thus, our

modeling approach distinguishes between general effect of SBFs, as well as how these effects

change during expansions and contractions, which is in line with van Heerde et al. (2013).

Control Variables and Fixed Effects. We include several control variables in the model

(𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝑖𝑖𝑖𝑖𝑖𝑖
𝑙𝑙
, l= 1…5). We account for marketing activities of other brands in the category by

averaging paid price, advertising, line length, and distribution of all other brands in the category.

We account for the presence and strength of private labels in a category by controlling for

category’s total private label market share (PLMSjq).

We also include several sets of fixed effects in our model. First, we include 324 brand

dummies (∑BRANDb) to account for unobserved time-invariant brand-specific factors that might

influence SBBE (e.g., country of origin, heritage). To control for seasonal fluctuations in SBBE

estimates in some categories (see Figure 2), we include three quarterly dummies (∑QUARTERq).

To account for general year-specific shocks to SBBE, we include yearly dummies (∑YEARy).

Multicollinearity. Having a large number of interaction terms might lead to multicollinearity.

In our empirical setting, all the variance inflation factor (VIF) values are well below 10 (average

VIF=2.80), thereby alleviating multicollinearity concerns. Further, as shown in Web Appendix I,

23
all correlations between our focal independent variables (and their interactions) are below .7.

Estimation. Since the dependent variable in Equation (6) is an estimated variable, we use

weighted least squares (WLS), with the inverse of SBBE’s standard errors from Equation (3) as

weights in our estimation (Bezawada and Pauwels 2013; Datta, Ailawadi, and van Heerde 2017).

We estimate standard errors using two-way cluster-adjusted robust standard errors (at brand and

quarter levels) that accounts for within-panel and within-time dependencies across observations

(Seiler, Tuchman, and Yao 2021).

Results

We present model-free evidence in Web Appendix J. We build our final model by successively

adding blocks of predictors to arrive at our full model (see Web Appendix K). Table 4 provides

the parameter estimates for equation (6). We find that long line length (α4=.0341, p<.01),

extensively distributed (α5=.0418, p<.01), and market leader brands (α7=.0246, p<.05) on

average have higher SBBE than selectively distributed, short line length, and market follower

brands. We do not find significant difference in SBBE of value vs. premium (α2=-.0012, p>.10)

and low vs. high ad spender (α3=.0083, p>.10) brands. The main effects of expansions (α8=-

.0949, p>.10) and contractions (α15=-.1011, p>.10) on SBBE are non-significant, suggesting that

SBBE of an ‘average brand’ does not change during expansions and contractions. 14

Expansions and Strategic Brand Factors

Although the main effect of expansions on SBBE is not significant, we find that brands with

different strategic characteristics are differentially affected by expansions. In line with H1EXP,

SBBE of premium brands is higher than SBBE of value brands during expansions (α9=.6165,

p<.05). This suggests that in good economic times when consumers have fewer budgetary

14
An ‘average brand’ is a brand that hypothetically scores zero on all six SBF variables. In an additional analysis we
removed all 12 interaction effects and both EXPq and CONq were again non-significant.

24
 Table 4: Main Results
Expected Std.
Predictors Effect Symbol Estimate Error
Intercept α0 .3095*** .0388
Past Level of Brand Equity (SBBEijq-1) α1 .8626*** .0120
Value vs. premium price positioning (PRICEijq) α2 -.0012 .0063
Low vs. high ad spenders (ADijq) α3 .0083 .0069
Strategic
Brand Short vs. long line length (LLijq) α4 .0341 ***
.0087
Factors Selective vs. extensive distribution (DISTijq) α5 .0418*** .0103
(SBFs)
Single- vs. umbrella-category branding (ARCHij) α6 NA†
Follower vs. leader market position (POSijq) α7 .0246** .0112
Magnitude of Expansion (EXPq) α8 -.0949 .1807
EXPq * PRICEijq H1EXP: + α9 .6165** .3690
Differential
Effect of EXPq * ADijq H2EXP: + α10 .1655 .6269
Expansions EXPq * LLijq H3EXP: + α11 .9600** .4982
for Different EXPq * DISTijq H4EXP: + α12 1.6825** .7785
Brands
EXPq * ARCHij H5EXP: - α13 .3898 .4422
EXPq * POSijq α14 .5672* .3279
Magnitude of Contraction (CONq) α15 -.1011 .2304
CONq * PRICEijq α16 .2116 .1423
Differential
Effect of CONq * ADijq H2CON: + α17 .3602* .2399
Contractions CONq * LLijq H3CON: - α18 -.5414*** .1752
for Different CONq * DISTijq H4CON: + α19 .6786** .3697
Brands
CONq * ARCHij H5CON: + α20 .4897*** .1931
CONq * POSijq H6CON: + α21 .2307 .4507
Other Brands’ Paid Price (OTHERSPPijq) α22 .0197 .0559
Other Brands’ Ad Expenditures (OTHERSADijq) α23 -.0001 .0009
Other Brands’ Line Length (OTHERSLLijq) α24 -.0053 .0068
Control Other Brands’ Distribution (OTHERSDISTijq) α25 -.2866*** .0377
Variables Private Label Market Share (PLMSjq) α26 -.0589 .0437
Brand Fixed Effects (324 Dummies) Included
Year Fixed Effects (15 Dummies) Included
Quarter Fixed Effects (3 Dummies) Included
*** p<.01; ** p<.05; * p<.10 (one-sided p-values for hypothesized effects and two-sided for others).
Significance assessed using 2-way cluster-adjusted SEs (at brand and quarter levels).
N=20,800 (due to the nature of operationalization of most SBF variables that utilize past four quarters of data, we
do not use first year of data [1994] in our second-stage analysis).
† Main effect of ARCH is not identified since it is a time-invariant characteristic and hence the effect is subsumed
within brand fixed effects.

25
restrictions, premium brands are preferred by them. We do not find significant evidence for

differences in SBBE of low vs. high ad spender brands during expansions (α10=.1655, p>.10),

hence H2EXP is not supported. During expansions, brands with long line length outperform brands

with short line length (α11=.9600, p<.05). Thus, H3EXP is supported.

In line with H4EXP, in expansions, extensively distributed brands do better equity-wise

compared to selectively distributed brands (α12=1.6825, p<.05). We do not find any difference in

brand equity of single-category vs. umbrella-category brands during expansions (α13=.3898,

p>.10), thus H5EXP is not supported. SBBE of market leader brands outperform that of follower

brands during expansionary periods (α14=.5672, p<.10), which provides support for the

bandwagon effect (i.e., the pleasure that consumers gain from using a product when more people

are using it). 15

Contractions and Strategic Brand Factors

While non-significance of CONq suggests that contractions do not generally affect SBBE of

brands, there is significant heterogeneity with respect to strategic brand factors. SBBE of value

and premium brands do not significantly differ during contractions (α16=.2116, p>.10). We

conjecture that higher product quality associated with premium brands (and hence higher

functional utility) provides a countervailing force to the higher price associated with them. In

line with H2CON, high ad spenders, compared to low ad spenders, have higher SBBE during

contractions (α17=.3602, p<.10).

Consistent with H3CON, we find that brands with short line length do better in contractions

compared to those with long line length (α18=-.5414, p<.01). Brands with extensive distribution

are estimated to have higher SBBE during contractions vis-à-vis brands with selective

To better understand the magnitude of the interaction effects, it should be noted that EXPq ranges from 0 to .032
15

while CONq ranges from 0 to .059.

26
distribution (α19=.6786, p<.05). Hence, H4CON is supported. 16 In line with H5CON, we find that in

contractions, umbrella-category brands have higher SBBE compared to single-category brands

(α20=.4897, p<.01). Finally, we do not find significant difference in SBBE of market leaders and

followers in contractions (α21=.2307, p>.10). Therefore, H6CON is not supported. It is possible that

during contractions when consumers lose trust in the economic system, they react more

negatively towards leading brands since these brands “may be seen to benefit most from this

unfair system” (Dekimpe and Deleersnyder 2018, p. 54). This alternative mechanism might have

weighted out the higher functional utility associated with market leaders. 17

Long-term Effects

Our main findings in Table 4 present the short-term differences in equity of different types of

brands during the business cycle. Such differences carry over into subsequent quarters because of

the inertia of brand equity, which is around .86 (Table 4), implying that 86% of brand equity

carries over into the next quarter. This renders brand equity stickier than revenues, which have a

quarterly carryover coefficient of .6 (Clarke 1976). The greater stickiness of brand equity makes

it even more worthwhile to invest in brands because the long-term differences across different

types of brands are substantially larger than their short-term differences. Figure 3 shows the

long-term implication of our main findings, using αLT = αST/(1-αSBBE(t-1)).

Figure 3 shows that entering expansions or contractions with different SBFs has considerable

long-term SBBE implications. In assessing the magnitude of differences observed in Figure 3, it

is worth noting that the average (median) brand-specific standard deviation in SBBE is .54 (.46).

16
Considering our finding regarding attributes of extensive distribution during expansions (α12=1.6825, p<.05), it
appears that the relation between distribution’s effect on SBBE and state of economy follows a V-shape: in regular
times, distribution’s effect on SBBE is smaller (yet significant) but in recessions or expansions, extensive
distribution is linked with higher SBBE.
17
We thank the AE for suggesting this explanation.

27
 Figure 3: Long-term Effects of the Business Cycle on Different Types of Brands

a. Value vs. Premium Pricing b. Low vs. High Ad Spender c. Short vs Long Line Length
.25 .25
.17 .17 .33
.20 .20 .35
Predicted SBBE

Predicted SBBE

Predicted SBBE
.15 .15
.20
.09 .10
.10 .10
.10 .10
.02 .02
.05
.05 .05
-.14
.00 -.10
.00

-.05 -.05 -.25

Expansion Contraction Expansion Contraction Expansion Contraction
Value PRICE Premium PRICE Low AD High AD Short LL Long LL

d. Select. vs. Ext. Distribution e. Single-category vs. Umbrella f. Follower vs. Leader Position
.25 .20
.60 .44 .25
.39 .30
.20 .19
.40
Predicted SBBE

Predicted SBBE
Predicted SBBE

.15
.20 .15
.10
.10
.01
.00
-.20 .05 -.01 -.06
-.25 .00
-.20
.00

-.40 -.05 -.15

Expansion Contraction Expansion Contraction Expansion Contraction
Selective DIST Extensive DIST Single-category ARCH Umbrella ARCH Follower POS Leader POS
Note. We set EXP and CON to their maximum observed values (.032 and .059 respectively). The error bars of predicted value represent one SE range.

28
In expansions, the most important factors are strategic decisions made with regard to line length

and distribution. Their effects can be categorized as large, according to Cohen (1988): Cohen’s d

of .87 and 1.29, respectively. 18 Next is market position (d = .58), while strategic decisions made

with respect to price also play a role, although only modest in size (d = .27). The outcome of the

strategic decisions regarding distribution is the single most important factor by far in determining

how brand equity will hold up (or not) in contractions (d = 1.10). Other factors that matter are

brand architecture (d = .39), market position (d = .33), and advertising (d = .29).

Validation Checks

Relation with Consumer-Based Brand Equity

Datta, Ailawadi, and van Heerde (2017) showed that SBBE and CBBE are moderately correlated

with each other. If we have correctly followed the SBBE procedure in estimating brand equity, our

estimates should show similar correlations with CBBE values. We obtained Young & Rubicam’s

Brand Asset Valuator (BAV) scores in the UK. For the period of our study, Young & Rubicam

collected BAV data in 1997, 2000, 2002, 2005, 2006, and 2008. We calculated correlations

between our SBBE estimates and BAV’s aggregate score (see Table 5). The correlations range

between .27 and .35 across years and are significant and comparable in magnitude to correlations

Table 5: Correlation between our SBBE Estimates and BAV’s Brand Equity Scores
Year Overall 1997 2000 2002 2005 2006 2008
r (SBBE, BAV) .31 †
.30 .35 .30 .27 .31 .34
r (within category SBBE rank, .58 .60 .66 .60 .56 .57 .52
within category BAV rank)
Number of Observations 847 125 129 135 149 153 156
All correlations are significant at p < .001. Following Datta, Ailawadi, and van Heerde (2017, p. 10), to allow for
comparability, we standardize SBBE estimates and BAV scores across brands in each product category.
† To the best of our knowledge, Datta, Ailawadi, and van Heerde (2017) did not report the correlation between their
SBBE estimates and BAV’s Brand Asset score. Instead, they reported correlations between their SBBE estimates and
the four dimensions of BAV’s Brand Asset score. The four correlations were .39, .35, .53, and -.14, suggesting an
average (unweighted) correlation of .28 which is comparable with our .31 correlation.

18
Benchmarks are: small effect d = .2; medium effect d = .5; large effect d = .8.

29
reported by Datta, Ailawadi, and van Heerde (2017). Moreover, correlation of within category

rankings of SBBE and BAV values range from .52 to .66. These observations provide evidence for

the validity of our SBBE measures.

Relation with Revenue Premium

Ailawadi, Lehmann, and Neslin (2003) proposed revenue premium – operationalized as the

differential revenue that a brand generates compared to that of a baseline private label product in

its category – as measure of SBBE. We assess how well our intercept-based SBBE measure aligns

with Ailawadi et al.’s revenue premium measure. To measure revenue premium, we considered

quarterly sales of an average private label brand in the product category as our benchmark (i.e.,

total sales of all private labels in the category divided by the number of private labels in the

category). The resulting correlation between our SBBE estimates and the revenue premium

measure is .34. By-category correlation between SBBE estimates and the revenue premium

measure has a median of .47, 10th percentile of .16, and 90th percentile of .63. Moreover, the rank

correlation between SBBE and revenue premium is .70. These results provide evidence for

convergent validity of our measure.

Stability of Brand Equity Estimates

Following Ailawadi, Lehmann, and Neslin (2003), we calculated the correlation between brand

equity estimates and their first lag to assess the relative stability of our equity estimates overtime.

The correlation is .96 in our sample, which is highly similar to the values reported by these

authors: .96 (local sample) and .98 (national sample). In Web Appendix G, we report correlations

between brand equity estimates and their first lag separately for each category. The correlations

are above .88 across all 35 product categories. These findings suggest that our estimates do not

exhibit erratic changes.

30
Other Robustness Checks

We also conduct a series of additional robustness checks and report the results in Web Appendix

L. We briefly mention the nature of these analyses but refer for details to Web Appendix L. We

include the following second-stage robustness checks:

• Operationalizing CONq via a different time-series filtering approach.

• Specifying cluster-adjusted standard errors at different levels of aggregation.
• Accounting for category-specific and brand-specific seasonal patterns.
• Controlling for marketing mix activities in the current time period.
• Using category medians to operationalize the first four SBF variables.

We include these first-stage robustness checks:

• Controlling for lagged effects of marketing mix instruments.
• Allowing the effects of marketing variables to vary across the business cycle.
• Using value (instead of volume) market share as the dependent variable.
• Removing lagged market share as an independent variable.
• Removing Gaussian Copulas from the first-stage.

Our results are mostly robust across the 12 analyses that we report in Web Appendix L.

Discussion

Our paper straddles the brand equity and business cycle literatures. We proposed a framework for

examining the impact of macroeconomic expansions and contractions on brand equity, analyzed

through the lens of strategic brand factors. Using a utility-based framework, we developed specific

hypotheses that underlie this framework. We tested these hypotheses using household panel data

on 325 CPG national brands in 35 categories across almost two decades in the UK. We found

evidence that the effect of economic conditions on brand equity is systematically moderated by six

strategic brand equity factors.

Managerial Implications

For many firms, brands constitute one of their most valuable assets. Edeling and Fischer (2016)

reported that a 1% change in brand equity translates into .33% change in market capitalization.

Our study documents that macroeconomic conditions affect brand equity and that the effect

31
depends on the strategic positioning of the brand. Kantar (2021, p. 6) maintained that “In good

times and tough times, strong brands win.” In their work, strong brands are brands that are high on

differentiation (captured by our strategic brand factors premium priced and high advertising), high

on meaningfulness (captured by long line length and high advertising), and high on salience

(extensive distribution, umbrella brand architecture, and leading market position). Table 6

summarizes our long-term findings (Figure 3), taking into account both main effects and

interaction effects, organized along Kantar’s three components of strong brands.

Table 6: Aligning our Findings with Kantar’s Three Components of Strong Brands
Kantar Component Level of Strategic Do Strong Brands Win?
of Strong Brands Brand Factor Expansion Contraction
High Differentiation Premium Priced Yes No effect
High Advertising No effect Yes

High Meaningfulness Long Line Length Yes No effect a

High Advertising No effect Yes

High Salience Extensive Distribution Yes Yes

Umbrella Brand No effect Yes
Market Leader Yes Yes
a
The strong negative interaction effect and the strong positive main effect cancel each other out. Large effects (as
determined by Cohen’s d) are underlined.

Our findings provide broad support for Kantar’s claim. Table 6 shows that in expansions as

well as in contractions, strong brands do indeed win in terms of creating more brand equity than

weak brands, at least if we take the aggregate of the strategic brand factors for each Kantar

component.

Yet, the overall support for Kantar’s sweeping claim disguises the fact that various strategic

brand factors have a notably different effect on brand equity. Some strategic brand factors matter

much more than others. In particular, the outcomes of strategic decisions with respect to

distribution and line length emerge as the key factors to consider.

In contractions, the effect of distribution is the largest contributor to brand equity by far. It is

32
important to keep this in mind given current economic turmoil. Further, distribution has a large

effect in expansions as well. In short, in good times and bad times, extensively distributed brands

win. Managers of brands that have a selective distribution need to consider whether this is a

strategic choice or the unwanted result of bad implementation of strategies to expand distribution.

If it is a strategic choice, our findings point to the consequences. If it is an unwanted outcome,

they may need to either increase investments in channel incentives (Ailawadi and Farris 2020) or,

if the firm already spends a lot on trade marketing, examine why channel incentives do not result

in expanded distribution.

In expansions, a wide assortment is also a strong contributor to brand equity, while it does not

destroy brand equity in contractions. Given this finding, expanding the assortment should be a

priority for brand management, unless there are other overriding considerations (e.g., lack of

resources). As mentioned before, strategic brand factors are sticky but not immutable. It is

possible to change the brand’s competitive positioning from a limited variety brand to a broad

assortment brand, if brand management so decides. However, this will take time. A starting point

is to invest more in R&D. With the elevated risks of a recession in 2023-2024 (Kiley 2022; Torry

and DeBarros 2022), managers planning for the long term, might want to go against the general

practice of cutting R&D expenditures during recessions (Barlevy 2007; Steenkamp and Fang

2011), and instead invest more on R&D. Given the time it takes to develop new products, they

might be ready to launch just when the economy bounces back, reaping full benefits of assortment

expansion.

Further, a premium price position and market leadership build brand equity in expansions

while advertising, using an umbrella brand architecture, and market leadership contribute to brand

equity in contractions, but the effect of management decisions with respect to these factors has

only a modest effect on brand equity. Thus, while these factors do matter, they are of secondary

33
importance when it comes to growing brand equity. The key takeaway is that if the brand manager

wants to grow brand equity for the long term, expanding distribution and line length are the two

strategic brand factors to concentrate on.

To further illustrate the role of strategic brand factors in influencing SBBE during the business

cycle, Figure 4 presents SBBE of four brands that had ‘successful or ‘unsuccessful strategic brand

factors during the 2008 financial crisis and the expansionary period that followed. We define

successful and unsuccessful strategic brand factors based on our results regarding the strategic

brand factors that significantly help or hurt brands during contractions or expansions. As depicted

in Figure 4, successful strategic brand factors led to growth in SBBE of Johnsons (Fairy; known as

Dawn in the US) during the 2008 financial crisis (the expansionary period after the financial

crisis). On the other hand, brands with unsuccessful strategic brand factors, Mornflakes and Heinz,

lost SBBE during the global recession and the subsequent expansion, respectively.

Limitations

This study has several limitations that future research should address. Our study focused on the

CPG industry in the UK. Future research should examine other industries in different countries to

generalize or nuance our findings and uncover additional patterns regarding how different types of

brands are affected by macroeconomic fluctuations. Further, we focused on examining the equity

of national brands. It could be argued that in the current marketplace private labels do command

considerable equity (Keller, Dekimpe, and Geyskens 2016). Since the distribution of private labels

is typically restricted to the retailer’s own stores and product level advertising is limited, current

brand equity methods are not ideal for the estimation of private label brand equity. We need new

methods for the measurement of brand equity of private labels.

34
Figure 4: Example Brands with Different Strategic Brand Factors in Expansions and Contractions

35
 Our research examined the overall patterns across brands in 35 CPG categories and did not

examine category-specific patterns. Product categories vary along many dimensions such as

consumer involvement (Zaichkowsky 1985), brand relevance (Fischer, Völckner, and Sattler

2010), perceived risk (Bettman 1973), and complexity (Agustin and Singh 2005). Future research

should examine heterogeneity in our results across product categories in function of these (and

other) important category-level characteristics.

In this research, we focused on sales-based brand equity. While sales-based brand equity

captures observed value added by the brand in the marketplace, it does not say anything about

consumers’ attitudes and thought processes. To better understand why and how consumer attitudes

towards different brands change during expansions and contractions, future research could also

consider consumer-based brand equity measures. Finally, we conceptually linked the six strategic

brand factors in our framework to Kantar’s three components of brand strength. Future research

should conduct in-depth conceptual and empirical examination of the relationships between

different elements in the two frameworks.

Conclusion

Our research documents the role that management decisions with respect to strategic brand factors

play in helping (or hurting) a brand during macroeconomic expansions and contractions. We show

that a premium price position and market leadership build brand equity in expansions while

advertising, using an umbrella brand architecture, and market leadership contribute to brand equity

in contractions. However, two factors dominate: distribution and line length. A wide assortment

plays a key role in growing brand equity in expansions, and extensively distributed brands win in

expansions and contractions. If the brand manager wants to grow brand equity for the long term,

expanding distribution and line length are the two strategic brand factors to concentrate on.

36
 References
Aaker, David A. (1991), Managing Brand Equity, New York: The Free Press.

—— (1996), “Measuring Brand Equity across Products and Markets,” California Management
Review, 38 (3), 102-20.

—— (2007), “Innovation: Brand it or Lose it,” California Management Review, 50 (1), 8-24.

—— and Erich Joachimsthaler (2000), “The Brand Relationship Spectrum: The Key to the
Architecture Challenge,” California Management Review, 42 (Summer), 8–23.

Agustin, Clara and Jagdip Singh (2005), “Curvilinear Effects of Consumer Loyalty Determinants
in Relational Exchanges,” Journal of Marketing Research, 42 (1), 96-108.

Ailawadi, Kusum L. and Paul W. Farris (2020), Getting Multi-Channel Distribution Right,
Hoboken, NJ: Wiley.

——, Donald R. Lehmann, and Scott A. Neslin (2003), “Revenue Premium as an Outcome
Measure of Brand Equity,” Journal of Marketing, 67 (4), 1-17.

Ataman, M. Berk, Carl F. Mela, and Harald J. van Heerde (2008), “Building Brands,” Marketing
Science, 27 (6), 1036–1054.

——, Harald J. van Heerde, and Carl F. Mela (2010), “The Long-term Effect of Marketing
Strategy on Brand Sales,” Journal of Marketing Research, 47 (5), 866-82.

Ba, Sulin and Paul A. Pavlou (2002), “Evidence of the Effect of Trust Building Technology in
Electronic Markets: Price Premiums and Buyer Behavior,” MIS quarterly, 26 (3), 243-68.

Barlevy, Gadi (2007), “On the Cyclicality of Research and Development,” American Economic
Review, 97 (4), 1131-64.

Bettman, James R. (1973), “Perceived Risk and its Components: A Model and Empirical Test,”
Journal of Marketing Research, 10 (May), 184–90.

Bezawada, Ram and Koen Pauwels (2013), “What is Special bout Marketing Organic Products?
How Organic Assortment, Price, and Promotions Drive Retailer Performance,” Journal of
Marketing, 77 (1), 31-51.

Bijmolt, Tammo H.A., Harald J. van Heerde, and Rik G.M. Pieters (2005), “New Empirical
Generalizations on the Determinants of Price Elasticity,” Journal of Marketing Research, 42
(2), 141-56.

Bronnenberg, Bart J., Vijay Mahajan, and Wilfried R. Vanhonacker (2000), “The Emergence of
Market Structure in New Repeat-Purchase Categories: The Interplay of Market Share and
Retailer Distribution,” Journal of Marketing Research, 37 (February), 16–31.

Chaudhuri, Arjun and Morris B. Holbrook (2001), “The Chain of Effects from Brand Trust and

37
 Brand Affect to Brand Performance: The Role of Brand Loyalty,” Journal of Marketing, 65
(2), 81-93.

Clarke, Darral G. (1976), “Econometric Measurement of the Duration of Advertising Effects on

Sales,” Journal of Marketing Research, 13 (4), 345–57.

Cohen, Jacob (1988), Statistical Power Analysis for the Behavioral Sciences, Hillsdale, NJ:
Lawrence Erlbaum, 2nd ed.

Cooper, Lee G. and Masao Nakanishi (1988), Market-Share Analysis. Boston: Kluwer Academic
Publishers.

Datta, Hannes, Kusum L. Ailawadi, and Harald J. van Heerde (2017), “How Well Does
Consumer-Based Brand Equity Align with Sales-Based Brand Equity and Marketing-Mix
Response?,” Journal of Marketing, 81 (3), 1-20.

——, Harald J. van Heerde, Marnik G. Dekimpe, and Jan-Benedict EM Steenkamp (2022),
“Cross-National Differences in Market Response: Line-Length, Price, and Distribution
Elasticities in Fourteen Indo-Pacific Rim Economies,” Journal of Marketing
Research, Forthcoming.

Dekimpe, Marnik G. and Barbara Deleersnyder (2018), “Business Cycle Research in Marketing:
A Review and Research Agenda,” Journal of the Academy of Marketing Science, 46 (1), 31-
58.

Deleersnyder, Barbara, Marnik G. Dekimpe, Miklos Sarvary, and Philip M. Parker (2004),
“Weathering Tight Economic Times: The Sales Evolution of Consumer Durables over the
Business Cycle,” Quantitative Marketing and Economics, 2(4), 347-83.

Du, Rex Y. and Wagner A. Kamakura (2008), “Where Did All that Money Go? Understanding
How Consumers Allocate Their Consumption Budget,” Journal of Marketing, 72 (6), 109-31.

Edeling, Alexander and Marc Fisher (2016), “Marketing’s Impact on Firm Value: Generalizations
from a Meta-Analysis,” Journal of Marketing Research, 53 (4), 515-534.

—— and Alexander Himme (2018), “When Does Market Share Matter? New Empirical
Generalizations from a Meta-analysis of the Market Share–Performance
Relationship,” Journal of Marketing, 82 (3), 1-24.

Erdem, Tülin (1998), “An Empirical Analysis of Umbrella Branding,” Journal of Marketing
Research, 35 (3), 339-51.

—— and Sue Ryung Chang (2012), “A Cross-category and Cross-country Analysis of Umbrella
Branding for National and Store Brands,” Journal of the Academy of Marketing Science, 40
(1), 86-101.

—— and Michael P. Keane (1996), “Decision-making Under Uncertainty: Capturing Dynamic

Brand Choice Processes in Turbulent Consumer Goods Markets,” Marketing Science, 15 (1):
1-20.

38
—— and Baohong Sun (2002), “An Empirical Investigation of the Spillover Effects of
Advertising and Sales Promotions in Umbrella Branding,” Journal of Marketing Research, 39
(4), 408-20.

——, Ying Zhao, and Ana Valenzuela (2004), “Performance of Store Brands: A Cross-country
Analysis of Consumer Store-brand Preferences, Perceptions, and Risk,” Journal of Marketing
Research, 41 (1), 86-100.

Farquhar, Peter H. (1989), “Managing Brand Equity,” Marketing Research, 1 (3), 24-33.

Fischer, Marc, Franziska Völckner, and Henrik Sattler (2010), “How Important are Brands? A
Cross-category, Cross-country Study,” Journal of Marketing Research, 47 (5), 823-39.

Geyskens, Inge, Katrijn Gielens, and Els Gijsbrechts (2010), “Proliferating Private-label
Portfolios: How Introducing Economy and Premium Private Labels Influences Brand
Choice,” Journal of Marketing Research, 47 (5), 791-807.

Gielens, Katrijn (2012), “New Products: The Antidote to Private Label Growth?,” Journal of
Marketing Research, 49 (3), 408-23.

Gordon, Brett R., Avi Goldfarb, and Yang Li (2013), “Does Price Elasticity Vary with Economic
Growth? A Cross-Category Analysis,” Journal of Marketing Research, 50 (1), 4-23.

Hellofs, Linda L. and Robert Jacobson (1999), “Market Share and Customers’ Perceptions of
Quality: When Can Firms Grow their Way to Higher versus Lower Quality?,” Journal of
Marketing, 63 (1), 16-25.

Higgins, E. Tory (2006), “Value from Hedonic Experience and Engagement,” Psychological
Review, 113 (3), 439-60.

Jacoby, Jacob and Leon B. Kaplan (1972), “The Components of Perceived Risk,” in Proceedings
of the Third Annual Conference of the Association for Consumer Research, M. Venkatesan,
ed. Chicago: Association for Consumer Research, 382–93.

Jindal, Pranav, Ting Zhu, Pradeep Chintagunta, and Sanjay Dhar (2020), “Marketing-mix
Response across Retail Formats: The Role of Shopping Trip Types,” Journal of Marketing, 84
(2), 114-32.

Kamakura, Wagner A. and Rex Yuxing Du (2012), “How Economic Contractions and Expansions
Affect Expenditure Patterns,” Journal of Consumer Research, 39 (2), 229-47.

Kantar (2021), A Comprehensive Guide to Brand Equity and Growth.

Keller, Kevin (1993), “Conceptualizing, Measuring, and Managing, Customer-Based Brand

Equity,” Journal of Marketing, 57 (1), 1-22.

Keller, Kristopher O., Marnik G. Dekimpe, and Inge Geyskens (2016), “Let Your Banner Wave?
Antecedents and Performance Implications of Retailers’ Private-Label Strategies,” Journal of
Marketing, 80 (4), 1-19.

39
Kihlstrom, Richard E. and Michael H. Riordan (1984), “Advertising as a Signal,” Journal of
Political Economy, 92 (3), 427-50.

Kiley, Michael T. (2022), “Financial and Macroeconomic Indicators of Recession Risk,” Board of
Governors of the Federal Reserve System Finance and Economics Discussion Series, No.
2022-06-21-1.

Kirmani, Amna and Peter Wright (1989), “Money Talks: Perceived Advertising Expense and
Expected Product Quality,” Journal of Consumer Research, 16 (3), 344-53.

Klein, Benjamin and Keith B. Leffler (1981), “The Role of Market Forces in Assuring Contractual
Performance,” Journal of Political Economy, 89 (4), 615-41.

Lamey, Lien, Barbara Deleersnyder, Marnik G. Dekimpe, and Jan-Benedict E.M. Steenkamp
(2007), “How Business Cycles Contribute to Private-Label Success: Evidence from the United
States and Europe,” Journal of Marketing, 71 (1), 1-15.

——, ——, Jan-Benedict E.M. Steenkamp, and Marnik G. Dekimpe (2012), “The Effect of
Business-Cycle Fluctuations on Private-Label Share: What Has Marketing Conduct Got to Do
With It?,” Journal of Marketing, 76 (1), 1-19.

——, ——, ——, ——, (2018), “New Product Success in the Consumer Packaged Goods
Industry: A Shopper Marketing Approach,” International Journal of Research in Marketing,
35 (3), 432-452.

Lodish, Leonard M. and Carl F. Mela (2007), “If Brands Are Built over Years, Why Are They
Managed over Quarters?” Harvard Business Review, 85 (7/8), 104-12.

Ma, Yu, Kusum L. Ailawadi, Dinesh K. Gauri, and Dhruv Grewal (2011), “An Empirical
Investigation of the Impact of Gasoline Prices on Grocery Shopping,” Journal of Marketing,
75 (2), 18-35.

Maslow, Abraham H. (1943), “A Theory of Human Motivation,” Psychological Review, 50 (4),

370-96.

Mela, Carl F., Sunil Gupta, and Kamel Jedidi (1998), “Assessing Long-Term Promotional
Influences on Market Structure,” International Journal of Research in Marketing, 15 (May),
89–107.

Meyer, Robert J. (1981), “A Model of Multiattribute Judgments Under Attribute Uncertainty and
Informational Constraint,” Journal of Marketing Research, 18 (4), 428-41.

Miklós-Thal, Jeanine (2012), “Linking Reputations through Umbrella Branding,” Quantitative

Marketing and Economics, 10 (3), 335-74.

Milgrom, Paul and John Roberts (1986), “Price and Advertising Signals of Product
Quality,” Journal of Political Economy, 94 (4), 796-821.

Millet, Kobe, Lien Lamey, and Bram Van den Bergh (2012), “Avoiding Negative vs. Achieving

40
 Positive Outcomes in Hard and Prosperous Economic Times,” Organizational Behavior and
Human Decision Processes, 117 (2), 275-84.

Millward Brown (2017), “Establish KPIs that Grow Brand Equity,” (accessed May 6, 2020),
[available at https://tinyurl.com/y8f7ashb]

Montgomery, Cynthia A. and Birger Wernerfelt (1992), “Risk Reduction and Umbrella
Branding,” Journal of Business, 65 (1): 31-50.

Myers, James H. and Allen D. Shocker (1981), “The Nature of Product-Related Attributes,” in
Research in Marketing, Vol. 5, Jagdish N. Sheth, ed. Greenwich (CT): JAI Press, 211-236.

Nevo, Aviv (2001), “Measuring Market Power in the Ready‐to‐Eat Cereal Industry,”
Econometrica, 69 (2), 307-42.

Papies, Dominik, Peter Ebbes, and Harald J. van Heerde (2017), “Addressing Endogeneity in
Marketing Models,” in Advanced Methods in Modeling Markets, Peter S.H. Leeflang, Jaap E.
Wieringa, Tammo H.A. Bijmolt, and Koen H. Pauwels, eds. Cham, CH: Springer, 581–627.

Park, Sungho and Sachin Gupta (2012), “Handling Endogenous Regressors by Joint Estimation
Using Copulas,” Marketing Science, 31(4), 567–86.

Pointer Media Network (2009), “Losing Loyalty: The Consumer Defection Dilemma,” CMO
Council and Pointer Media Network.

Pras, Bernard and John O. Summers (1978), “Perceived Risk and Composition Models for
Multiattribute Decisions,” Journal of Marketing Research, 15 (3), 429-37.

Rajavi, Koushyar, Tarun Kushwaha, and Jan-Benedict E.M. Steenkamp (2019), “In Brands We
Trust? A Multicategory, Multicountry Investigation of Sensitivity of Consumers’ Trust in
Brands to Marketing-Mix Activities,” Journal of Consumer Research, 46 (4), 651-70.

Rao, Akshay R. and Humaira Mahi (2003), “The Price of Launching a New Product: Empirical
Evidence on Factors Affecting the Relative Magnitude of Slotting Allowances,” Marketing
Science, 22 (2), 246-68.

Rosenthal, Robert (1991), Meta-Analytic Procedures for Social Research. Newbury Park, CA:
Sage.

Scholdra, Thomas P., Julian RK Wichmann, Maik Eisenbeiss, and Werner J. Reinartz (2022),
“Households Under Economic Change: How Micro-and Macroeconomic Conditions Shape
Grocery Shopping Behavior,” Journal of Marketing, 00222429211036882.

Seiler, Stephan, Anna Tuchman, and Song Yao (2021), “The Impact of Soda Taxes: Pass-through,
Tax Avoidance, and Nutritional Effects,” Journal of Marketing Research, 58 (1), 22-49.

Sharp, Byron (2010), How Brands Grow, Melbourrne: Oxford University Press.

Sheppard, Blair H., Jon Hartwick, and Paul R. Warshaw (1988), “The Theory of Reasoned Action:

41
 A Meta-analysis of Past Research with Recommendations for Modifications and Future
Research,” Journal of Consumer Research, 15 (3), 325-43.

Srinivasan, Raji, Gary L. Lilien, and Shrihari Sridhar (2011), “Should Firms Spend More on R&D
and Advertising during Recessions?” Journal of Marketing, 79 (2), 49-65.

Srinivasan, Shuba, Marc Vanhuele, and Koen Pauwels (2010), “Mind-set Metrics in Market
Response Models: An Integrative Approach,” Journal of Marketing Research, 47 (4), 672-84.

Sriram, S., Subramanian Balachander, and Manohar U. Kalwani (2007), “Monitoring the
Dynamics of Brand Equity Using Store-Level Data,” Journal of Marketing, 71 (2), 61-78.

Steenkamp, Jan-Benedict E.M. (2014), “How Global Brands Create Firm Value: The 4V
Model,” International Marketing Review, 31 (1), 5-29.

____ and Eric Fang (2011), “The Impact of Economic Contractions on the Effectiveness of R&D
and Advertising: Evidence from US Companies Spanning Three Decades,” Marketing Science,
3 (4), 628-45.

Tamir, Maya, Chi-Yue Chiu, and James J. Gross (2007), “Business or Pleasure? Utilitarian Versus
Hedonic Considerations in Emotion Regulation,” Emotion, 7 (3), 546-54.

Torry, Harriet and Anthony DeBarros (2022), “Recession Probability Soars as Inflation Worsens,”
Wall Street Journal, June 19 [available at http://tinyurl.com/2s3f9vn2]

Van Ewijk, Bernadette J., Els Gijsbrechts, and Jan-Benedict E.M. Steenkamp (2022), “The Dark
Side of Innovation: How New SKUs Affect Brand Choice in the Presence of Consumer
Uncertainty and Learning,” International Journal of Research in Marketing (in press).

Van Heerde, Harald J., Maarten J. Gijsenberg, Marnik G. Dekimpe, and Jan-Benedict E.M.
Steenkamp (2013), “Price and Advertising Effectiveness Over the Business Cycle,” Journal of
Marketing Research, 50 (2), 177-93.

Van Herpen, Erica, Rik Pieters, and Marcel Zeelenberg (2009). “When Demand Accelerates
Demand: Trailing the Bandwagon,” Journal of Consumer Psychology, 19 (3), 302-12.

Yoo, Boonghee, Naveen Donthu, and Sungho Lee (2000), “An Examination of Selected
Marketing Mix Elements and Brand Equity,” Journal of the Academy of Marketing Science, 28
(2), 195-211.

Zaichkowsky, Judith Lynne (1985), “Measuring the Involvement Construct,” Journal of

Consumer Research, 12 (3), 341-52.

42
 WEB APPENDIX

Brand Equity in Good and Bad Times:

What Distinguishes Winners from Losers in CPG Industries?

Koushyar Rajavi (Koushyar.rajavi@scheller.gatech.edu)

Assistant Professor of Marketing, Scheller College of Business,
Georgia Institute of Technology

Tarun Kushwaha (tkushwah@gmu.edu)

Professor of Marketing, School of Business,
George Mason University

Jan-Benedict E.M. Steenkamp (jbs@unc.edu)

C. Knox Massey Distinguished Professor of Marketing, Kenan-Flagler Business School,
University of North Carolina

Table of Contents

Section Title Page

Web Appendix A Research on the Impact of Macroeconomic Fluctuations on ii
Marketing-Related Phenomena
Web Appendix B Comparing Utility Functions for Different Conditions iii
Web Appendix C Market Share Statistics Across Categories iv
Web Appendix D Sample Statistics Across Different Categories and Brands v
Web Appendix E Product Attributes across Different Categories vii
Web Appendix F By-Category Summary of First-stage Results ix
Web Appendix G SBBE Estimates by Category x
Web Appendix H Operationalizing Business Cycles Using Time-Series xi
Filtering
Web Appendix I Correlation Table for Focal Predictor Variables in the xiii
Second-stage
Web Appendix J Model-free Evidence for Differences in SBBE of Brands in xiv
Expansions and Contractions Depending on Their SBF
Web Appendix K Adding Blocks of Predictors to Build the Final Model xv
Web Appendix L Additional Robustness Checks xvi
References Not in the Manuscript xxix

i
 WEB APPENDIX A – RESEARCH ON THE IMPACT OF MACROECONOMIC FLUCTUATIONS ON MARKETING-
RELATED PHENOMENA
Paper Outcome Variable Moderating Effects Level of Key Findings
Analysis
Deleersnyder et Sales of Durables Product Type, Product Industry Durables are very sensitive to business-cycle fluctuations. Nature of the durable and the
al. 2004 Life Cycle, etc. stage in a product’s life cycle moderate the extent of sensitivity in durable sales
patterns.
Lamey et al. Share of Private Product Private label share behaves cyclically and business cycles have temporary and
-
2007 Labels Category permanent impacts on private label share.
Deleersnyder et Advertising Spending National Culture Advertising Advertising is sensitive to business-cycles. Advertising behaves less cyclically in
al. 2009 Media- countries high in long-term orientation and power distance and low in uncertainty
Country avoidance.
Kamakura and Customer Preferences Type of Goods and Household For any given consumption budget, expenditure shares for positional goods/services
Du 2012 for Categories Services will decrease during a recession, while shares for non-positional goods/services will
increase.
Srinivasan, Effectiveness of Market share, Financial Firm The greater the firm’s market share, the more an increase in R&D spending during
Lilien, and Advertising and leverage, and Product- recessions increases its profits. The greater the firm’s financial leverage, the more an
Sridhar (2011) R&D market profile increase in advertising spending in recessions increases profits.
Steenkamp and Effectiveness of Industry cyclicality Firm Increasing advertising and R&D in downturns have a stronger effect on profit and
Fang 2011 advertising and R&D market share than increasing advertising or R&D in upturns. Advertising effectiveness,
especially in downturns, in particular, is systematically moderated by the degree of
cyclicality of the industry.
Lamey et al. Share of Private National Brands’ Product Private-label share behaves countercyclically. Brands’ procyclical behavior regarding
2012 Labels Marketing Category new product introductions, advertising, and promotions is associated with more
pronounced cyclical changes in PL share.
Gordon, Price Elasticity Category’s Price Household Price sensitivity is countercyclical and rises when the economy weakens. The
Goldfarb, and Li Sensitivity relationship between price sensitivity and business cycles correlates strongly with the
2013 average level of price sensitivity in a category.
van Heerde et al. Advertising and Price Brand Segments, Brand Long-term price sensitivity decreases during expansions, whereas long-term advertising
2013 Elasticity Product Type elasticities increase. These patterns vary across different product categories and brands.
This Study Brand Equity Strategic Brand Factors Brand In expansions, premium brands, brands with long line length, extensively distributed
brands, and market followers perform better on brand equity, whereas in contractions,
high ad spender, low line length, extensively distributed, and umbrella brands fare better
than other brands.

ii
 WEB APPENDIX B – COMPARING UTILITY FUNCTIONS FOR DIFFERENT CONDITIONS

Elements of Utility Function Strategic Brand Factors

Differentiation Meaningfulness Salience
Price Positioning Advertising Spending Line Length Distribution Intensity Brand Architecture Market Position
(Value vs. Premium) (Low vs. High) (Short vs. Long) (Selective vs. Extensive) (Single vs. Umbrella) (Follower vs. Leader)

Price PPRM > PVAL (a) PEXT < PSEL

Functional Attributes Xf,PRM > Xf,VAL (a) Xf,HI-AD > Xf,LO-AD Xf,EXT > Xf,SEL Xf,UMB > Xf,SIN Xf,LEA > Xf,FOL
Functional Risk σf,PRM < σf,VAL σf,LNG > σf,SHR σf,UMB < σf,SIN

Emotional Attributes Xe,PRM > Xe,VAL Xe,HI-AD > Xe,LO-AD Xe,LNG > Xe,SHR Xe,EXT > Xe,SEL Xe,LEA <or> Xe,FOL

Emotional Risk σe,HI-AD < σe,LO-AD σe,UMB > σe,SIN

Net Effect of Relevant Components (all else equal)
(c)
UHI-AD,EXP = CHI-AD,EXP
UPRM,EXP = CPRM,EXP UUMB,EXP = CUMB,EXP
+ ωe,EXP*Xe,HI-AD ULNG,EXP = CLNG,EXP UEXT,EXP = CEXT,EXP ULEA,EXP = CLEA,EXP
on (b)
–αEXP*PPRM – re,EXP*σe,UMB
– re,EXP*σe,HI-AD + ωe,EXP*Xe,LNG + ωe,EXP*Xe,EXT + ωe,EXP*Xe,LEA
Utility αEXP < αCON +ωe,EXP*Xe,PRM <<
>> >> >> < OR >
During ωe,EXP > ωe,CON >> USIN,EXP = CSIN,EXP
Expansions re,EXP > re,CON ULO-AD,EXP = CLO-AD,EXP USHR,EXP = CSHR,EXP USEL,EXP = CSEL,EXP UFOL,EXP = CFOL,EXP
UVAL,EXP = CVAL,EXP – re,EXP*σe,SIN
+ωe,EXP*Xe,LO-AD + ωe,EXP*Xe,SHR +ωe,EXP*Xe,SEL + ωe,EXP*Xe,FOL
–αEXP*PVAL
+ωe,EXP*Xe,VAL – re,EXP*σe,LO-AD
UPRM,CON = CPRM,CON
UUMB,CON = CUMB,CON
– αCON*PPRM UEXT,CON = CEXT,CONT
+ωf,CON*Xf,UMB
+ ωf,CON*Xf,PRM UHI-AD,CON = CHI-AD,CON ULNG,CON = CLNG,CON – αCON*PEXT ULEA,CON = CLEA,CON
on – rf,CON*σf,UMB
αEXP < αCON – rf,CON*σf,PRM + ωf,CON*Xf,HI-AD – rf,CON*σf,LNG + ωf,CON*Xf,EXT + ωf,CON*Xf,LEA
Utility >>
ωf,CON > ωf,EXP < OR > >> << >> >>
During USIN,CON = CSIN,CON
Contractions
rf,CON > rf,EXP UVAL,CON = CVAL,CON ULO-AD,CON = CLO-AD,CON USHR,CON = CLNG,CON USEL,CON = CSEL,CONT UFOL,CON = CFOL,CON
+ωf,CON*Xf,SIN
– αCON*PVAL + ωf,CON*Xf,LO-AD – rf,CON*σf,SHR – αCON*PSEL + ωf,CON*Xf,FOL
– rf,CON*σf,SIN
+ωf,CON*Xf,VAL + ωf,CON*Xf,SEL
– rf,CON*σf,VAL
To be read as
(a) Premium brands have higher prices (PPRM > PVAL) and also provide higher functional attributes (Xf,PRM > Xf,VAL) than value brands.
(b) During expansions (vs. contractions) consumers are less price sensitive (αCON > αEXP) and assign greater importance to emotional attributes (ωe,EXP >
ωe,CON).
(c) Thus, all else equal, during expansions, the utility that consumers derive from premium brands (UPRM,EXP = CPRM,EXP – αEXP*PPRM + ωe,EXP*Xe,PRM) will
be more than that they will derive from value brands (UVAL,EXP = CVAL,EXP –αEXP*PVAL + ωe,EXP*Xe,VAL). Here, C is the weighted sum of utility components
not affected, in this case CPRM, EXP = ωf,EXP*Xf,PRM - re,EXP*σe,PRM – rf,EXP*σf,PRM and CVAL, EXP = ωf,EXP*Xf, VAL – re,EXP*σe, VAL – rf,EXP*σf, VAL

iii
 WEB APPENDIX C – MARKET SHARE STATISTICS ACROSS CATEGORIES
Brands with Lowest Brands with Highest Avg
Category # Brands
Avg MS Avg MS HHIψ
Artificial Sweeteners 5 Fuisana, Sucron Hermesetas, Sweetex .136
Bath Additives 8 E45, Badedas Radox, Johnsons .029
Bathroom Tissue 4 Izal, Nouvelle Velvet, Andrex .070
Breakfast Cereals 9 Scotts, Ready Brek Weetabix, Kelloggs .116
Butter 7 President, Kerrygold Lurpak, Anchor .161
Canned Fruit 9 Trout Hall, Bridge House Del Monte, Princes .020
Weight Watchers,
Canned Soup 4 Baxters, Heinz .230
Covent Garden
Cat Food 11 Purina, Friskies Whiskas, Felix .111
Cereal Bars 3 Tracker Jordans .044
Cleansers (Facial) 13 Mudd, Ponds Clean & Clear, Clearasil .031
Conditioners 11 Vitapointe, Nicky Clarke Alberto, Pantene .031
Cooking Sauces 18 Heinz, Encona Dolmio, Homepride .037
Deodorants 13 Old Spice, Amplex Lynx, Sure .069
Dog Food 16 Frolic, Masterchoice Winalot, Pedigree .053
Dry Pasta 3 Marshalls Buitoni .003
Frozen Fish 8 Kershaws, Lyons Seafoods Birds Eye, Youngs .150
Fruit/Yoghurt Juice 18 Yoplait, Roses Tropicana, Robinsons .076
Ground/Bean Coffee 4 Rombouts, Lavazza Douwe Egbert, Lyons .021
Household Cleaners 8 Ecover, Stardrops Cif, Flash .028
Red Mountain,
Instant Coffee 4 Maxwell House, Nescafe .283
Mellow Birds
Laundry Detergents 8 Ecover, Dreft Ariel, Persil .082
Margarine 8 Summer County, Willow Stork, Flora .060
Mineral Water 10 San Pellegrino, Malvern Highland Spring, Evian .010
Potato chips 7 KP Brannigans, Highlander Walkers, Kettle Foods .223
Razor Blades 5 Personna, Laser Bic, Gillette .213
Sanitary Protection Products 9 Interlude, Helen Harper Tampax, Always .062
Shampoo 13 Gliss Corimist, Simple Pantene, Head & Shoulders .023
Shower Prod. 11 Badedas, Simple Radox, Imperial Leather .045
Soft Drinks 29 Ben Shaw, Appletiser Coca Cola, Pepsi .025
Stout 3 Mackeson Guinness .579
Table Sauces 5 Hammonds, C&B Branston H.P. Sauces, Heinz .157
Tea 12 Glengettie, Nambarrie Tetley, P.G.Tips .070
Toothpaste 13 Oral B, Euthymol Colgate, Aquafresh .145
Washing Up Products 6 Surcare, Ecover Finish, Fairy .095
Yoghurt 10 Longley Farm, Nestle Ski, Muller .106
ψ
Based on average HHI in each category across 203 months. Monthly HHI is calculated based on square of market
share of the top three national brands in the category in terms of monthly volume market share.

iv
WEB APPENDIX D – SAMPLE STATISTICS ACROSS DIFFERENT CATEGORIES
AND BRANDS
Table WA.D1 – Marketing Mix Instruments across Categories
# Price Promo. Advertising Distribution Line
Category
Brands Depth (%) (000 pounds) (%) Length
Artificial Sweeteners 5 1.4 (2.7) 21.4 (84.9) 67.6 (29.6) 5.0 (3.4)
Bath Additives 8 4.6 (6.0) 28.5 (141.4) 68.9 (25.9) 8.9 (8.6)
Bathroom Tissue 4 3.1 (4.8) 198.0 (330.9) 81.8 (21.2) 19.3 (14.3)
Breakfast Cereals 9 2.5 (3.6) 676.0 (1,448.7) 93.3 (7.6) 23.1 (24.6)
Butter 7 1.7 (3.4) 97.9 (255.8) 73.2 (33.2) 4.2 (3.5)
Canned Fruit 9 3.8 (5.2) 5.3 (50.5) 52.4 (31.3) 12.7 (13.4)
Canned Soup 4 2.6 (3.8) 59.9 (203.1) 90.7 (8.8) 40.6 (26.0)
Cat Food 11 2.1 (3.3) 137.5 (324.2) 85.4 (23.2) 44.2 (58.2)
Cereal Bars 3 2.6 (4.6) 10.9 (54.6) 89.3 (6.5) 9.7 (5.8)
Cleansers (Facial) 13 3.8 (5.0) 51.7 (13.6) 69.8 (25.3) 6.5 (5.3)
Conditioners 11 4.6 (6.6) 21.9 (100.5) 56.3 (31.0) 9.3 (11.2)
Cooking Sauces 18 3.2 (4.5) 99.3 (258.7) 84.6 (18.0) 23.7 (20.4)
Deodorants 13 4.0 (4.8) 138.1 (383.5) 78.7 (22.9) 19.5 (15.5)
Dog Food 16 1.6 (3.2) 76.6 (235.7) 70.3 (33.8) 21.5 (30.5)
Dry Pasta 3 4.4 (6.5) 7.8 (35.6) 68.1 (29.4) 11.9 (8.3)
Frozen Fish 8 3.3 (4.2) 40.5 (173.5) 60.7 (34.5) 19.4 (24.8)
Fruit/Yoghurt Juice 18 3.0 (4.9) 87.7 (254.9) 67.4 (31.7) 11.5 (12.3)
Ground/Bean Coffee 4 3.6 (5.2) 40.1 (137.5) 83.2 (15.9) 9.4 (6.7)
Household Cleaners 8 3.0 (4.3) 137.2 (263.4) 71.5 (32.9) 11.5 (11.3)
Instant Coffee 4 4.0 (6.0) 351.5 (732.2) 85.2 (17.4) 16.9 (21.5)
Laundry Detergents 8 2.7 (3.6) 677.2 (728.1) 88.3 (19.3) 23.0 (19.3)
Margarine 8 2.6 (4.7) 157.1 (345.0) 85.9 (26.9) 5.2 (3.6)
Mineral Water 10 2.6 (5.2) 42.3 (153.1) 58.7 (33.2) 7.6 (5.9)
Potato chips 7 3.9 (6.5) 209.7 (487.6) 78.0 (26.6) 23.6 (22.4)
Razor Blades 5 2.7 (3.9) 209.5 (447.9) 65.9 (38.5) 19.0 (17.0)
Sanitary Protection Prod. 9 3.2 (3.9) 173.5 (325.5) 74.3 (36.0) 14.9 (7.4)
Shampoo 13 4.3 (6.2) 125.8 (296.6) 66.0 (34.1) 11.2 (12.2)
Shower Prod. 11 5.9 (7.5) 85.5 (232.5) 73.2 (26.8) 12.7 (10.2)
Soft Drinks 29 3.3 (4.9) 155.4 (480.6) 73.3 (30.7) 15.3 (10.8)
Stout 3 2.9 (4.9) 438.9 (759.3) 83.5 (13.5) 6.6 (6.2)
Table Sauces 5 2.4 (3.7) 89.2 (256.1) 85.6 (19.9) 10.3 (8.2)
Tea 12 3.3 (5.0) 123.8 (324.4) 81.8 (19.2) 10.5 (8.5)
Toothpaste 13 3.1 (4.3) 169.9 (351.1) 83.6 (15.7) 11.8 (13.4)
Washing Up Prod. 6 2.4 (3.6) 213.6 (369.2) 82.4 (22.2) 14.7 (16.8)
Yoghurt 10 2.4 (3.5) 175.0 (415.3) 79.4 (23.8) 16.6 (14.5)_
Average (standard deviation) of marketing mix instruments across the whole time period of 203 months
and across all the brands in a category reported. We do not report summary statistics for regular price
(PRICE) as that variable depends on unit of measurement in each category (which we report in Web
Appendix D) and hence difficult to interpret. The advertising columns describe raw advertising
expenditures and not the advertising stock which we use in our first-stage estimation.

v
 Table WA.D2 – Within-brand Variation in Marketing Mix Instruments
Price Promo. Advertising Distribution Line
Depth (%) (000 pounds) (%) Length
Within-brand average (averaged across 325 brands) 3.2 143.9 74.7 15.4
Within-brand average (25th percentile) 2.1 2.0 61.2 4.6
Within-brand average (median) 3.1 30.4 85.8 9.2
Within-brand average (75th percentile) 4.0 140.8 93.7 20.8
Within-brand std. dev. (averaged across 325 brands) 4.4 173.2 11.3 .6
Within-brand std. dev. (25th percentile) 2.9 9.7 2.6 .1
Within-brand std. dev. (median) 4.0 111.2 8.6 .4
Within-brand std. dev. (75th percentile) 5.4 263.9 18.7 .7
Overall standard deviation (across all observations) 4.9 434.7 28.8 20.1

vi
 WEB APPENDIX E – PRODUCT ATTRIBUTES ACROSS DIFFERENT CATEGORIES
Category Unit of Sales Attributes
Artificial Sweeteners Grams Multi-Pack, Large, Tablets, Granules / Powders
Bath Additives Milliliters Multi-Pack, Large, Liquid, Salts, Baby, Aromatherapy
Bathroom Tissue Count (Packs) Multi-Pack, Large, White, Quilted, Peach, Pink, Green, Moist, Soft
Breakfast Cereals Grams Multi-Pack, Large, Crispy, Crunchy, Flakes, Crunchy, Oat, Fruit, Nut
Butter Grams Multi-Pack, Large, Spreadable, Light, Salted
Canned Fruit Grams Multi-Pack, Large, Slices, Halves, Chunk, Syrup, Juice, Segments, Pieces
Canned Soup Milliliters/Grams Multi-Pack, Large, Wet, Fresh, Vegetable, Broth, Organic
Cat Food Grams Multi-Pack, Large, Jelly, Adult, Chunks, Kitten, Canned, Dry, Chicken
Cereal Bars Grams Multi-Pack, Large, Chewy, Crunchy, Berry
Cleansers (Facial) Milliliters/Grams Multi-Pack, Large, Facial, Scrub, Wipes, Medicated, Mask, Lotion
Conditioners Milliliters Multi-Pack, Large, Normal, Dry, Damaged, Frizz, Perm
Cooking Sauces Milliliters/Grams Multi-Pack, Large, Additive, Pour-Over, Jelly, Pasta, Jar
Deodorants Milliliters Multi-Pack, Large, Bodyspray, Sensitive, 24h, Dry, Sport, Men, Women
Dog Food Grams Multi-Pack, Large, Dry, Beef, Vegetable, Puppy, Canned, Soft / Moist, Biscuit
Dry Pasta Grams Multi-Pack, Large, Wheat, Verdi, Shapes, Twirl
Dry Soup Grams Multi-Pack, Large, Dry, Instant, Quick, Veg, Noodle, Sachet, Packet
Frozen Fish Grams Multi-Pack, Large, Filet, Pie, Prawn, Breaded, Salmon, Scampi, Haddock
Fruit/Yoghurt Juice Milliliters Multi-Pack, Large, Pure, Juice Drinks, High Juice, Added Sugar, Yoghurt, Low Calorie
Ground/Bean Coffee Grams Multi-Pack, Large, Filter, Medium, Decaf, Single, Espresso, Pod
Household Cleaners Milliliters/Grams Multi-Pack, Large, Wipes, Kitchen, Spray, Bath, Bleach, Cream
Instant Coffee Grams Multi-Pack, Large, Blend, Decaf, Cappuccino, Powder, Unsweetened, Pure
Laundry Detergents Milliliters/Grams Liquid, Large, Tabs, Caps, Powder, Concentrate
Margarine Grams Multi-Pack, Large
Mineral Water Milliliters Multi-Pack, Large, Glass Bottle, Plastic Bottle, Fruit, Spring, Carbonated, Flavored
Peanut Butter Grams Multi-Pack, Large, Crunchy, Organic, Smooth
Potato chips Grams Multi-Pack, Large, Assorted, Salted, Roasted, Vinegar, Cheese
Razor Blades Count Multi-Pack, Large, Sensitive, Fixed, Women, Cartridge
Sanitary Protection Products Count Multi-Pack, Large, Digital, Wing, Applicator, Night, Ultra, Normal, Super
Shampoo Milliliters Multi-Pack, Large, Frequent, Herbal, Dry, Damaged, Fine, Perm, Volume
Shower Prod. Milliliters Multi-Pack, Large, Gel, Fresh, Cream, Women, Active, Sport
Soft Drinks Milliliters Multi-Pack, Large, Cola, Lemon, Diet, Cherry, Canned, PET Bottle, Glass Bottle
Stout Milliliters Multi-Pack, Large, Can, Bottle, Draught
Table Sauces Grams Multi-Pack, Large, Glass Bottle, Plastic Bottle, Chili, Sweet, BBQ, Tomato, Brown
Tea Grams Multi-Pack, Large, Specialty, Round, Pyramid, One Cup, PMP
Toothpaste Milliliters Multi-Pack, Large, Pump, Whitening, Mint, Gel, Cool, Sensitive, Paste
Washing Up Products Milliliters/Grams Multi-Pack, Large, Lemon, Liquid, Tablet, Concentrated, Powder, Dishwash
Yoghurt Grams Multi-Pack, Large, Strawberry, Raspberry, Greek, Natural, Diet, Cherry, Vanilla

vii
Notes on our approach in identifying product attributes:

Some of the product attributes such as product size were easily derived from the “barcode
description” file that was available to us. Regarding size, in each category, we calculated average
SKU size and defined SKUs that were above average size to be “large” and the rest to be “not
large” SKUs. The same goes for the “Multi-pack” attribute that we have in most categories; the
data specifies whether a SKU has one unit of product or multiple units.

As for the remaining attributes, we applied text mining techniques to the SKU “description”
column that we had in our “barcode description” file (on a total of 26,914 SKUs that were
marketed from 1994 to 2010 in 37 product categories). For example, a breakfast cereal SKU by
Alpen is described in the following way “ALPEN NUTTY CRUNCH 500GM”. Our algorithm
allowed us to define “nutty” and “crunchy” attributes based on this description. We did the text
mining to hundreds or thousands of SKUs in each category and based on each SKU description,
we detected different attributes. Once we discovered all possible attributes across all SKUs in a
category, we counted the frequency of each attribute amongst the SKUs in the category. To keep
things manageable, we only focused on the most important attributes in each category; i.e.,
attributes with most frequency across all the SKUs in each category. Thereby, we limited number
of attributes to a maximum of nine in each product category.

viii
 WEB APPENDIX F – BY-CATEGORY SUMMARY OF FIRST-STAGE RESULTS
Advertising (AdStock) Price Promotion Depth (Promo) Distribution Intensity (Dist) Product Line Length (LL) Regular Price (Price)
Category #Sig #Sig #Sig #Sig #Sig #Sig #Sig #Sig #Sig #Sig
B Mean† Med. >0 <0 λ B Mean Med. >0 <0 B Mean Med. >0 <0 B Mean Med. >0 <0 B Mean Med. >0 <0
Artificial Sweeteners 5 .01 .02 0 0 .50 5 .17 .19 3 0 5 .57 .58 5 0 5 .45 .63 1 0 5 -.70 -.92 0 2
Bath Additives 8 .00 -.01 1 0 .90 8 .33 .28 6 0 8 .40 .37 6 0 8 .80 .92 3 0 8 -.38 -.48 0 1
Bathroom Tissue 4 .04 .06 2 0 .70 4 .08 .07 1 0 4 .66 .66 4 0 4 .35 .44 1 0 4 .22 .18 1 0
Breakfast Cereals 9 .01 .01 4 1 .00 9 .24 .19 9 0 9 .90 .89 5 1 9 .46 .49 3 0 9 -.15 -.12 1 3
Butter 7 .02 .01 2 0 .50 7 .26 .26 6 0 7 .37 .34 4 0 7 .13 -.04 1 0 7 -3.23 -3.20 0 6
Canned Fruit 9 -.03 -.01 0 1 .80 9 .15 .12 4 0 9 .20 .18 2 0 9 1.92 2.30 7 0 9 -1.25 -1.13 0 6
Canned Soup 4 .02 .03 1 0 .90 4 .26 .32 4 0 4 .35 .61 1 0 4 .24 .29 1 0 4 -.01 -.15 1 0
Cat Food 11 .03 .02 3 1 .90 11 .19 .23 8 0 11 .45 .34 6 0 11 .30 .27 4 1 11 -.54 -.51 0 4
Cereal Bars 3 .02 .02 0 0 .90 3 .17 .17 3 0 3 .48 .51 0 0 3 .37 .34 1 0 3 -.25 -.22 0 0
Cleansers (Facial) 13 -.01 -.02 1 5 .40 13 .22 .14 6 0 13 .63 .61 10 0 13 .96 .97 7 0 13 -.60 -.50 0 9
Conditioners 11 .14 .07 3 0 .90 11 .15 .10 4 0 11 .54 .57 9 0 11 1.05 .78 6 0 11 -1.02 -1.02 0 8
Cooking Sauces 18 .01 .01 3 2 .40 18 .25 .21 11 0 18 .18 .08 5 2 18 .81 .75 12 1 18 -.98 -.90 1 10
Deodorants 13 .03 .03 1 0 .70 13 .15 .18 7 0 13 .12 .17 5 2 13 .68 .69 7 0 13 -.43 -.58 0 4
Dog Food 16 .01 .00 1 2 .90 16 .20 .19 13 0 16 .23 .28 8 1 16 .16 .15 3 1 16 -.68 -.27 0 6
Dry Pasta 3 -.01 -.03 0 0 .90 3 .26 .25 3 0 3 .12 .10 0 0 3 1.61 1.63 3 0 3 -3.77 -3.63 0 3
Frozen Fish 8 .06 -.01 2 1 .90 8 .21 .23 6 0 8 .49 .51 8 0 8 1.60 1.70 7 0 8 -.13 -.10 2 2
Fruit/Yoghurt Juice 18 .00 .01 3 3 .80 18 .20 .16 11 0 18 .12 .06 4 0 18 .62 .53 8 1 18 -.79 -1.02 2 7
Ground/Bean Coffee 4 .04 .04 1 0 .90 4 .29 .28 4 0 4 .44 .45 3 0 4 .95 1.04 4 0 4 -.69 -.63 0 3
Household Cleaners 8 .00 .00 1 1 .10 8 .32 .43 6 0 8 .29 .29 5 0 8 1.25 1.19 5 0 8 -.47 -.40 1 4
Instant Coffee 4 .00 .00 0 1 .00 4 .08 .07 1 0 4 .51 .52 4 0 4 .97 1.05 3 0 4 .26 -.04 1 0
Laundry Detergents 8 .01 .02 0 0 .30 8 .41 .41 7 0 8 .57 .54 8 0 8 .34 .39 4 0 8 -.56 -.52 0 1
Margarine 8 .00 -.02 2 2 .90 8 .20 .20 7 0 8 .54 .48 5 0 8 .82 .94 4 0 8 -.89 -.78 0 3
Mineral Water 10 -.03 -.04 0 1 .80 10 .39 .33 9 0 10 .37 .34 5 0 10 .86 .85 3 0 10 -1.60 -1.51 0 6
Potato chips 7 .01 .00 2 2 .40 7 .25 .24 5 0 7 .92 1.10 5 0 7 .95 1.13 5 0 7 -.46 -.60 0 7
Razor Blades 5 .03 .02 1 1 .80 5 .11 .08 1 0 5 .40 .40 4 0 5 .16 .24 2 0 5 -.34 -.36 0 4
Sanitary Prot. Prod. 9 .02 .03 1 1 .70 9 .10 .09 2 0 9 .14 .20 4 1 9 .50 .70 6 0 9 -.35 -.27 1 4
Shampoo 13 .01 .01 4 1 .30 13 .15 .12 7 1 13 .53 .69 10 0 13 .71 .69 7 0 13 .23 .15 4 2
Shower Prod. 11 .02 .00 3 1 .90 11 .24 .24 6 0 11 .59 .57 8 0 11 1.05 1.28 6 0 11 .24 .40 1 0
Soft Drinks 29 .01 .01 5 4 .40 29 .19 .19 18 1 29 .22 .21 10 2 29 .47 .50 13 1 29 -.88 -1.35 6 13
Stout 3 .02 .03 1 0 .90 3 .08 .11 3 0 3 .24 .33 3 0 3 .41 .55 2 0 3 .08 .15 0 0
Table Sauces 5 .10 .11 3 0 .90 5 .22 .31 4 0 5 .15 .18 0 0 5 .36 .25 1 0 5 -.78 -.68 0 1
Tea 12 .03 .04 6 0 .40 12 .26 .25 9 0 12 .56 .67 9 0 12 1.02 1.35 8 1 12 -.89 -.92 1 7
Toothpaste 13 .00 .01 3 1 .00 13 .12 .10 7 0 13 .42 .39 9 0 13 .66 .29 6 0 13 -.49 -.43 1 9
Washing Up Prod. 6 .05 .05 1 0 .90 6 .26 .23 6 0 6 .68 .68 5 0 6 .15 .06 1 0 6 -.16 -.23 1 1
Yoghurt 10 .06 .07 7 0 .70 10 .17 .18 7 0 10 .31 .34 4 0 10 .33 .52 5 0 10 -.11 -.04 0 0
† Meta-analytic weighted means reported

ix
 WEB APPENDIX G – SBBE ESTIMATES BY CATEGORY
Brands with Lowest Avg Brands with Highest Corr(SBBE,
Category # Brands
SBBE † Avg SBBE † l1.SBBE)
Artificial Sweeteners 5 Sucron, Canderel Hermesetas, Sweetex .917
Bath Additives 8 E45, Matey Dove, Radox .891
Bathroom Tissue 4 Izal, Velvet Nouvelle, Andrex .958
Breakfast Cereals 9 Mornflakes, Kelloggs Scotts, Jordans .912
Butter 7 St. Ivel, Wheelbarrow Butter Lurpak, Anchor .989
Canned Fruit 9 Trout Hall, Valfruta Princes, Fruitini .934
Canned Soup 4 Weight Watchers, Heinz Baxters, Covent Garden .948
Cat Food 11 Arthurs, Katkins Go-Cat, Felix .975
Cereal Bars 3 Tracker Jordans .883
Cleansers (Facial) 13 Anne French, Oxy Clean & Clear, Johnsons .926
Conditioners 11 Revlon, Nicky Clarke Alberto, Pantene .935
Crosse & Blackweel,
Cooking Sauces 18 Sacla, Amoy .927
Napolina
Deodorants 13 Mum, Arrid Sure, Adidas .970
Dog Food 16 Tex, Chappie Hi-Life, Bakers Dog Food .980
Dry Pasta 3 Napolina Marshalls .943
Frozen Fish 8 Macrae, Lyons Seafoods Kershaws, Youngs .962
Southern Delight, Tropicana,
Fruit/Yoghurt Juice 18 .956
Sunpride Ocean Spray
Ground/Bean Coffee 4 Rombouts, Lyons Douwe Egbert, Lavazza .925
Household Cleaners 8 Ajax, Domestos Dettol, Mr Muscle .960
Red Mountain,
Instant Coffee 4 Maxwell House, Nescafe .928
Mellow Birds
Laundry Detergents 8 Daz, Ariel Fairy, Bold .966
Margarine 8 Summer County, Vitalite I C B I N B, St Ivel .987
Mineral Water 10 Abbey Well, Malvern San Pellegrino, Evian .967
KP Brannigans,
Potato chips 7 KP, Kettle Foods .979
Golden Wonder
Razor Blades 5 Personna, Bic Gillette, Wilkinson Sword .945
Sanitary Protection Products 9 Interlude, Tampax Carefree, Always .973
Shampoo 13 Timotei, Wash & Go Head & Shoulders, T/Gel .941
Shower Prod. 11 Badedas, Nivea Imperial Leather, Johnsons .888
Soft Drinks 29 Ben Shaw, Carters Dr. Pepper, Coca Cola .962
Stout 3 Mackeson Guinness .919
Daddies Sauce,
Table Sauces 5 H.P. Sauces, Heinz .915
C&B Branston
Tea 12 Lift, Tetley Yorkshire Tea, R. Twining .941
Toothpaste 13 Mentadent, Thera-Med Colgate, Sensodyne .973
Washing Up Products 6 Persil, Morning Fresh Finish, Fairy .953
Yoghurt 10 Longley Farm, Nestle Yoplait, Rachels .963
† Brands with lowest and highest average SBBEs in a product category are determined based on average of a brand’s
SBBE estimates during all time periods, weighted by inverse of the standard error for each estimate.

x
 WEB APPENDIX H – OPERATIONALIZING BUSINESS CYCLES USING TIME-
SERIES FILTERING

Hodrick and Prescott (1997) filter (hereinafter, HP filter) has been widely used in marketing
research on business cycles (e.g., Lamey et al 2007; 2012; Deleersnyder et al 2009; Steenkamp
and Fang 2011). The HP filter breaks down a time-series into (1) a gradually evolving trend
component that represents long-term changes in a series and (2) cyclical fluctuations around the
trend component that represent short-term changes in a series. In HP filters, the trend component
(Xtr) is extracted by minimizing the following formula:

(1) ∑𝑄𝑄 � � 𝑡𝑡𝑡𝑡 2 𝑄𝑄−1 � 𝑡𝑡𝑡𝑡 � 𝑡𝑡𝑡𝑡 � � 𝑡𝑡𝑡𝑡

𝑞𝑞=1�𝑋𝑋𝑖𝑖𝑖𝑖𝑖𝑖 − 𝑋𝑋𝑖𝑖𝑖𝑖𝑖𝑖 � + 𝜆𝜆 ∑𝑞𝑞=2 ��𝑋𝑋𝑖𝑖𝑖𝑖𝑖𝑖+1 − 𝑋𝑋𝑖𝑖𝑖𝑖𝑖𝑖 � − �𝑋𝑋𝑖𝑖𝑖𝑖𝑖𝑖 − 𝑋𝑋𝑖𝑖𝑖𝑖𝑖𝑖−1 ��
2

with λ being the smoothing parameter. Following past research, for quarterly data, we use λ =
1600 (Hodrick and Prescott 1997; Ravn and Uhlig 2002; Kesavan and Kushwaha 2014).
Consistent with past research (Lamey et al. 2007), we use inflation-adjusted GDPPC as the proxy
for economic activity. Since SBBE estimates are at a quarterly level, we use quarterly GDPPC
(GDPPCq). For log-transformed GDPPCq (i.e., lnGDPPCq), we extract its cyclical component
(𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝐶𝐶𝑞𝑞𝑐𝑐𝑐𝑐𝑐𝑐 ), which is a measure of business cycles: 1
𝑐𝑐𝑐𝑐𝑐𝑐
(2) 𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝐶𝐶𝑞𝑞 = 𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝐶𝐶𝑞𝑞 − 𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝐶𝐶𝑞𝑞𝑡𝑡𝑡𝑡

Next, we use 𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝐶𝐶𝑞𝑞𝑐𝑐𝑐𝑐𝑐𝑐 to operationalize the extent of expansions (EXPq) and contractions
(CONq) at any point in time. We follow van Heerde et al. (2013) and define the magnitude of
expansion (contraction) as the difference between the actual level of the cyclical component of the
macroeconomic fluctuations at quarter q and the prior trough (peak):
𝑐𝑐𝑐𝑐𝑐𝑐 𝑐𝑐𝑐𝑐𝑐𝑐 𝑐𝑐𝑐𝑐𝑐𝑐
𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝐶𝐶𝑞𝑞 − �𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝 𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡ℎ 𝑖𝑖𝑖𝑖 𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝐶𝐶𝑞𝑞 � ; 𝑖𝑖𝑖𝑖 ∆𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝐶𝐶𝑞𝑞 >0
(3) 𝐸𝐸𝐸𝐸𝐸𝐸𝑞𝑞 = � 𝑐𝑐𝑐𝑐𝑐𝑐
0 ; 𝑖𝑖𝑖𝑖 ∆𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝐶𝐶𝑞𝑞 ≤0
𝑐𝑐𝑐𝑐𝑐𝑐
0 ; 𝑖𝑖𝑖𝑖 ∆𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝐶𝐶𝑞𝑞 >0
(4) 𝐶𝐶𝐶𝐶𝐶𝐶𝑞𝑞 = � 𝑐𝑐𝑐𝑐𝑐𝑐 𝑐𝑐𝑐𝑐𝑐𝑐 𝑐𝑐𝑐𝑐𝑐𝑐
�𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝 𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝 𝑖𝑖𝑖𝑖 𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝐶𝐶𝑞𝑞 � − 𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝐶𝐶𝑞𝑞 ; 𝑖𝑖𝑖𝑖 ∆𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝐶𝐶𝑞𝑞 ≤0

EXPq (CONq) takes positive values during economic upturns (downturns) and 0 during downturns
(upturns). This operationalization allows us to capture the magnitude of expansions or slowdowns,
with the value of EXPq (CONq) capturing the percentage improvement (decline) in the economy

1
For a detailed discussion on the rationale behind business cycle filtering and methodological details, see
Deleersnyder et al. (2004) and Lamey et al. (2007, 2012).

xi
during expansions (contractions).

We note that van Heerde et al. (2013) use Christiano-Fitzgerald (CF) filtering approach when
applying filters to their GDP variable. We followed the CF filtering procedure which led to a
revised 𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝐶𝐶𝑞𝑞𝑐𝑐𝑐𝑐𝑐𝑐 (and subsequently a revised CONq) which were strongly correlated with the
𝑐𝑐𝑐𝑐𝑐𝑐
𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝐶𝐶𝑞𝑞 (and CONq) that were based on HP filtering approach (r>.9). Moreover, as we show

in the robustness section, majority of our findings remain substantively unchanged when we
follow the CF filtering approach (see Web Appendix L).

xii
 WEB APPENDIX I – CORRELATION TABLE FOR FOCAL SECOND-STAGE PREDICTOR VARIABLES

(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) (13) (14) (15) (16) (17) (18) (19)
1) PRICE 1.00
2) AD .13 1.00
3) LL -.01 .49 1.00
4) DIST .04 .39 .48 1.00
5) ARCH -.01 .05 .02 .00 1.00
6) POS .08 .59 .64 .45 .01 1.00
7) EXP .00 -.01 -.01 .01 .00 .00 1.00
8) CON .00 .00 .00 .01 .00 .00 -.31 1.00
9) EXP*PRICE .57 .09 .01 .03 .00 .06 .00 .00 1.00
10) EXP*AD .08 .54 .29 .20 .01 .33 -.35 .11 .16 1.00
11) EXP*LL .01 .30 .56 .26 .01 .36 -.21 .06 .02 .58 1.00
12) EXP*DIST .03 .21 .26 .55 .00 .25 .26 -.08 .06 .25 .39 1.00
13) EXP*ARCH .00 .02 .01 .00 .49 .00 -.52 .16 .01 .19 .12 -.13 1.00
14) EXP*POS .05 .32 .35 .23 .00 .53 -.37 .12 .10 .66 .67 .30 .20 1.00
15) CON*PRICE .41 .06 .00 .02 .00 .03 .00 -.02 .00 .00 .00 .00 .00 .00 1.00
16) CON*AD .06 .38 .20 .14 .02 .22 .12 -.38 .00 -.04 -.02 .03 -.06 -.04 .16 1.00
18) CON*LL .00 .21 .40 .19 .02 .26 .07 -.23 .00 -.02 -.01 .02 -.04 -.03 .04 .60 1.00
17) CON*DIST .02 .15 .19 .39 .01 .17 -.09 .31 .00 .03 .02 -.02 .05 .03 .08 .21 .38 1.00
19) CON*ARCH .01 .02 .01 .00 .34 .01 .18 -.56 .00 -.06 -.04 .05 -.09 -.07 .03 .26 -.15 .17 1.00
20) CON*POS .03 .22 .24 .16 .01 .38 .13 -.41 .00 -.04 -.03 .03 -.07 -.05 .09 .67 .25 .67 .27

xiii
 WEB APPENDIX J – MODEL-FREE EVIDENCE FOR DIFFERENCES IN SBBE OF
BRANDS IN EXPANSIONS AND CONTRACTIONS DEPENDING ON THEIR
STRATEGIC BRAND FACTORS
We compare average SBBE of observations representing different strategic brand characeristics.
We do this separately for “regular times”, expansions, and contractions. We define regular times
as quarters in which magnitude of expansions or contractions were smaller than .5%
(0<EXPq≤.005 or 0<CONq≤.005) which accounts for roughly one-third of the quarters in the time
period of our data. The differences due to SBFs during regular times provides the baseline effect
and corresponds to the main effects in our model (α2 through α7). In this analysis, expansions are
the time periods with EXPq>.005 and contractions are time periods with CONq>.005. Majority of
the observed patterns are in line with our main findings:

Regular Times (Baseline) Expansions Contractions

Value Premium Value Premium Value Premium
-.0660 .1357 -.1113 .2042 -.0815 .1527
Δ = .2017 Δ = .3155 Δ = .2342
Low AD High AD Low AD High AD Low AD High AD
-.1105 .3637 -.1635 .5290 -.1920 .6068
Δ = .4742 Δ = .6925 Δ = .7988
Short LL Long LL Short LL Long LL Short LL Long LL
-.1559 .3418 -.1955 .4425 -.1295 .2929
Δ = .4977 Δ = .6380 Δ = .4224
Selective Dist. Extensive Dist. Selective Dist. Extensive Dist. Selective Dist. Extensive Dist.
-.3754 .2592 -.5266 .3813 -.5741 .3617
Δ = .6346 Δ = .9079 Δ = .9358
Single-category Umbrella Single-category Umbrella Single-category Umbrella
-.0084 .1952 -.0121 .2359 -.0163 .2485
Δ = .2036 Δ = .248 Δ = .2648
Follower Leader Follower Leader Follower Leader
-.1086 .4006 -.1612 .5858 -.1362 .4717
Δ = .5092 Δ = .7470 Δ = .6079

xiv
 WEB APPENDIX K – ADDING BLOCKS OF PREDICTORS TO BUILD THE FINAL
MODEL

M1: M2: M1+ M3: M2+

Expected M0: Only M0+ EXP & Its CON & Its
Predictors Effect Controls SBFs Interact. Interact.
Intercept .3207*** .3071*** .3083*** .3095***
SBBEijq-1 .8802*** .8657*** .8634*** .8626***
PRICEijq .0042 .0007 -.0012
ADijq .0110 *
.0107 .0083
Strategic LLijq .0372 ***
.0307 ***
.0341***
Brand
Factors DISTijq .0529*** .0466*** .0418***
ARCHij NA† NA† NA†
POSijq .0280** .0265** .0246**
EXPq -.0602 -.0949
EXPq * PRICEijq H1EXP: + .5173* .6165**
Differential
Effect of EXPq * ADijq H2EXP: + .0164 .1655
Expansions EXPq * LLijq H3EXP: + 1.1843*** .9600**
for Different EXPq * DISTijq H4EXP: + 1.3898* 1.6825**
Brands
EXPq * ARCHij H5EXP: - .1718 .3898
EXPq * POSijq .4584* .5672*
CONq -.1011
CONq * PRICEijq .2116
Differential
Effect of CONq * ADijq H2CON: + .3602*
Contractions CONq * LLijq H3CON: - -.5414***
for Different CONq * DISTijq H4CON: + .6786**
Brands
CONq * ARCHij H5CON: + .4897***
CONq * POSijq H6CON: + .2307
OTHERSPPijq .0288 .0183 .0186 .0197
OTHERSADijq .0001 -.0001 -.0001 -.0001
OTHERSLLijq -.0058 -.0055 -.0050 -.0053
Control OTHERSDISTijq -.2962*** -.2861*** -.2874*** -.2866***
Variables PLMSjq -.0679 -.0592 -.0609 -.0589
Brand FEs Included Included Included Included
Year FEs Included Included Included Included
Quarter FEs Included Included Included Included
*** p<.01; ** p<.05; * p<.10 (one-sided p-values for hypothesized effects and two-sided for others). Significance
assessed using 2-way cluster-adjusted SEs (at brand and quarter levels). N=20,800 (due to the nature of
operationalization of most SBF variables that utilize past four quarters of data, we do not use first year of data [1994]
in our second-stage analysis).
† Main effect of ARCH is not identified since it is a time-invariant characteristic and hence the effect is subsumed
within brand fixed effects.

xv
 WEB APPENDIX L – ADDITIONAL ROBUSTNESS CHECKS

We conduct a series of robustness checks to assess the sensitivity of our findings with respect to
different choices in our first- and second-stage models. Below we describe these checks:

Second-stage robustness checks:

1- Operationalizing EXPq and CONq via another time-series filtering approach: we follow van
Heerde et al. (2013) and construct EXPq and CONq using the Christiano-Fitzgerald (CF)
filtering approach. Results are reported in Table WA.K1. All our results are substantively
unchanged, with the exception being CONq * DISTijq which is no longer significant.
Moreover, EXPq * ARCHjq that was not significant in our main analysis was found to be
positive and significant in this analysis (in line with H5e).
2- Specifying cluster-adjusted standard errors at different levels of aggregation: our main
analysis utilized a rigorous two-way clustering approach for the standard errors (at brand
and quarter levels) that accounts for within-time (cross-brand) and within-brand
correlations across observations. Following Seiler, Tuchman, and Yao (2021), we show
robustness of our results to alternative standard error specifications at different levels of
aggregation. These results are reported in Tables WA.K2a (clustered SEs at brand and year
levels), and WA.K2b (clustered SEs at brand and quarter*category levels).
3- Accounting for category-specific and brand-specific seasonal patterns: in our main
analysis, we include quarter fixed effects to account for seasonal patterns in SBBE. But
perhaps the seasonal patterns in SBBE are category-specific or brand-specific. We address
such concerns by including quarter*category and quarter*brand fixed effects in analyses
which we report in WA.K3a and WA.K3b, respectively.
4- Controlling for marketing mix activities in the current time period: we add five marketing
mix variables that represent quarterly advertising, regular price, price promotion, line
length, and distribution intensity of the focal brands to equation 6. We create these
variables by averaging monthly values of these variables that we use in our first-stage
model. Upon adding these variables, we realized that inclusion of distribution intensity led
to serious multicollinearity issues and maximum VIF value rose to 34.52. We therefore
removed distribution intensity from our model. Results are reported in Table WA.K4.
5- Using category medians to operationalize the first four SBF variables: in our main
analysis, we used category means in the past four quarters to operationalize four SBF

xvi
 variables (i.e., value vs. premium pricing, low vs high ad spenders, short vs. long line
length, and selective vs. extensive distribution). In WA.K5 we present results after using
category medians to operationalize these variables. All of our findings remain unchanged.

First-stage robustness checks:

6- Controlling for lagged effects of marketing mix instruments: we add first lag of the
marketing mix instruments to the market share attraction model that we utilize to obtain
SBBE estimates. Since advertising stock already incorporates previous advertising
expenditures, we only include lagged values of the other marketing mix instruments. The
new results are reported in Table WA.K6.
7- Allowing the effects of marketing variables to vary across the business cycle: in our first-
stage analysis, we add interactions of the five marketing mix instruments with the variables
representing expansions (EXPq) and contractions (CONq). These interactions would
account for the possibility that the effects of marketing mix instruments on market share
might be different in expansions and contractions. We report the results in Table WA.K7.
Most of our findings are replicated. It is also worth noting that unlike our main analysis
and in line with H6c, we find support for CONq * POSijq.
8- Using value (instead of volume) market share as the dependent variable: in our first-stage
analysis, we used volume market share as our dependent variable. We redo our analyses by
using value market share in the first-stage analysis (see Table WA.K8).
9- Removing lagged market share: in the first-stage model, to account for dynamics and state
dependence in market share, we included lagged market share as an independent variable.
We follow Datta, Ailawadi, and van Heerde (2017) by specifying a first-stage model
without lagged market share. In Table WA.K9 we report the results of our analysis if
lagged market share is not added in the first-stage model. All our findings are replicated.
10- Removing Gaussian Copulas from the first-stage: it could be argued that with the inclusion
of lagged dependent variable, multiple marketing mix instruments, and product attributes
there is little endogeneity concerns in our first-stage model. In WA.K10 we report the
results of analysis without Gaussian Copulas in our first-stage model.

xvii
Overall, our results remain generally robust across the 12 analyses presented in Web Appendix L.
In the table below we summarize how many times we find support for the focal interactions that
were significant in our main analysis:

Effect EXPq* EXPq* EXPq* EXPq * CONq * CONq* CONq* CONq*

PRICEijq LLijq DISTijq POSijq ADijq LLijq DISTijq ARCHij
Supported in # Analyses 10/12 12/12 10/12 8/12 11/12 11/12 9/12 11/12

xviii
Table WA.L1 – Using Christiano-Fitzgerald (CF) Filtering to Operationalize EXP and CON
Expected Main
Predictors Effect Finding Estimate
Intercept .3094***
SBBEijq-1 .8649***
PRICEijq -.0024
ADijq .0103*
Strategic Brand LLijq .0348***
Factors DISTijq .0391***
ARCHij NA†
POSijq .0245**
EXPq -.3551
EXPq * PRICEijq H1EXP: + + .7475**
Differential EXPq * ADijq H2EXP: + NS -.1947
Effect of
EXPq * LLijq H3EXP: + + .8093**
Expansions for
Different Brands EXPq * DISTijq H4EXP: + + 2.1197***
EXPq * ARCHij H5EXP: - + 1.1570***
EXPq * POSijq + .6620*
CONq -.0328
CONq * PRICEijq NS .2479
Differential CONq * ADijq H2CON: + + .4324*
Effect of
CONq * LLijq H3CON: - - -.3604**
Contractions for
Different Brands CONq * DISTijq H4CON: + + .4621
CONq * ARCHij H5CON: + + .8381***
CONq * POSijq H6CON: + NS -.0571
OTHERSPPijq .0142
OTHERSADijq .0001
OTHERSLLijq -.0046
Control OTHERSDISTijq -.2836***
Variables PLMSjq -.0537
Brand Fixed Effects Included
Year Fixed Effects Included
Quarter Fixed Effects Included
*** p<.01; ** p<.05; * p<.10 (one-sided p-values for hypothesized effects and two-sided for others). Significance
assessed using 2-way cluster-adjusted SEs (at brand and quarter levels). N=20,800 (due to the nature of
operationalization of most SBF variables that utilize past four quarters of data, we do not use first year of data [1994]
in our second-stage analysis).
† Main effect of ARCH is not identified since it is a time-invariant characteristic and hence the effect is subsumed
within brand fixed effects.

19
 Table WA.L2 – Clustering Standard Errors at Different Levels of Aggregation
a. Clustered b. Clustered
Expected Main SEs at Brand SEs at Brand
Predictors Effect Finding & Year & Qtr*Cat
Intercept .3095*** .3095***
SBBEijq-1 .8626*** .8623***
PRICEijq -.0012 -.0001
ADijq .0083 .0090
Strategic Brand LLijq .0341 ***
.0336***
Factors DISTijq .0418*** .0425***
ARCHij NA† NA†
POSijq .0246** .0261**
EXPq -.0949 -.2220
EXPq * PRICEijq H1EXP: + + .6165* .5489*
Differential EXPq * ADijq H2EXP: + NS .1655 .7763
Effect of
EXPq * LLijq H3EXP: + + .9600** .8138*
Expansions for
Different Brands EXPq * DISTijq H4EXP: + + 1.6825 1.4986**
EXPq * ARCHij H5EXP: - NS .3898 .2757
EXPq * POSijq + .5672* .4677
CONq -.1011 -.2215
CONq * PRICEijq NS .2116 .0781
Differential CONq * ADijq H2CON: + + .3602 .8860**
Effect of
CONq * LLijq H3CON: - - -.5414** -.6282**
Contractions for
Different Brands CONq * DISTijq H4CON: + + .6786** .4491
CONq * ARCHij H5CON: + + .4897* .3943*
CONq * POSijq H6CON: + NS .2307 .1698
OTHERSPPijq .0197 .0209
OTHERSADijq -.0001 -.0001
OTHERSLLijq -.0053 -.0046
Control OTHERSDISTijq -.2866*** -.2867***
Variables PLMSjq -.0589 -.0654
Brand Fixed Effects Included Included
Year Fixed Effects Included Included
Quarter Fixed Effects Included Included
*** p<.01; ** p<.05; * p<.10 (one-sided p-values for hypothesized effects and two-sided for others). Significance
assessed using 2-way cluster-adjusted SEs (at brand and year levels on the “a” column, and brand and
quarter*category levels on the “b” column). N=20,800 (due to the nature of operationalization of most SBF variables
that utilize past four quarters of data, we do not use first year of data [1994] in our second-stage analysis).
† Main effect of ARCH is not identified since it is a time-invariant characteristic and hence the effect is subsumed
within brand fixed effects.

xx
 Table WA.L3 – Accounting for Category-specific and Brand-Specific Seasonal Patterns
a. Adding b. Adding
Expected Main Cat.*Qtr Brand*Qtr
Predictors Effect Finding Fixed Effects Fixed Effects
Intercept .3094*** .2605***
SBBEijq-1 .8632*** .8850***
PRICEijq -.0014 -.0001
ADijq .0083 .0079
Strategic Brand LLijq .0339 ***
.0260***
Factors DISTijq .0418*** .0334***
ARCHij NA† NA†
POSijq .0242** .0178**
EXPq -.0865 -.1653
EXPq * PRICEijq H1EXP: + + .6237** .5934*
Differential EXPq * ADijq H2EXP: + NS .1634 .7930
Effect of
EXPq * LLijq H3EXP: + + .9415** .7093*
Expansions for
Different Brands EXPq * DISTijq H4EXP: + + 1.6520** 1.1053
EXPq * ARCHij H5EXP: - NS .4064 .2353
EXPq * POSijq + .5859* .4110
CONq -.0906 -.2104
CONq * PRICEijq NS .2192 .2041
Differential CONq * ADijq H2CON: + + .3400* .6553*
Effect of
CONq * LLijq H3CON: - - -.5251*** -.5540***
Contractions for
Different Brands CONq * DISTijq H4CON: + + .6556** .4248*
CONq * ARCHij H5CON: + + .5153*** .3868***
CONq * POSijq H6CON: + NS .2369 .1418
OTHERSPPijq .0197 .0191
OTHERSADijq -.0001 .0001
OTHERSLLijq -.0053 -.0031
OTHERSDISTijq -.2866*** -.2457***
Control PLMSjq -.0589 -.0518
Variables Brand Fixed Effects Included Included
Year Fixed Effects Included Included
Quarter Fixed Effects Included Included
Category*Quarter FEs Included
Brand*Quarter FEs Included
*** p<.01; ** p<.05; * p<.10 (one-sided p-values for hypothesized effects and two-sided for others). Significance
assessed using 2-way cluster-adjusted SEs (at brand and quarter levels). N=20,800 (due to the nature of
operationalization of most SBF variables that utilize past four quarters of data, we do not use first year of data [1994]
in our second-stage analysis).
† Main effect of ARCH is not identified since it is a time-invariant characteristic and hence the effect is subsumed
within brand fixed effects.

xxi
 Table WA.L4 – Controlling for Marketing Mix Activities in the Current Period
Expected Main
Predictors Effect Finding Estimate
Intercept .2849***
SBBEijq-1 .8620***
PRICEijq -.0026
ADijq .0066
Strategic Brand LLijq .0248***
Factors DISTijq .0399***
ARCHij NA†
POSijq .0248**
EXPq -.1386
EXPq * PRICEijq H1EXP: + + .6065*
Differential EXPq * ADijq H2EXP: + NS .2344
Effect of
EXPq * LLijq H3EXP: + + .6684*
Expansions for
Different Brands EXPq * DISTijq H4EXP: + + 1.6846**
EXPq * ARCHij H5EXP: - NS .3423
EXPq * POSijq NS .2805
CONq -.1425
CONq * PRICEijq NS .1971
Differential CONq * ADijq H2CON: + + .3307*
Effect of
CONq * LLijq H3CON: - - -.7225***
Contractions for
Different Brands CONq * DISTijq H4CON: + + .6757**
CONq * ARCHij H5CON: + + .4881**
CONq * POSijq H6CON: + NS .0423
OTHERSPPijq .0207
OTHERSADijq -.0001
OTHERSLLijq -.0150**
OTHERSDISTijq -.2823***
PLMSjq -.0596
Control Brand Fixed Effects Included
Variables Year Fixed Effects Included
Quarter Fixed Effects Included
Ad Expenditures .0001**
Regular Price -.0193
Price Promotion .2060**
Line Length .0012***
*** p<.01; ** p<.05; * p<.10 (one-sided p-values for hypothesized effects and two-sided for others). Significance
assessed using 2-way cluster-adjusted SEs (at brand and quarter levels). N=20,800 (due to the nature of
operationalization of most SBF variables that utilize past four quarters of data, we do not use first year of data [1994]
in our second-stage analysis). † Main effect of ARCH is not identified since it is a time-invariant characteristic and
hence the effect is subsumed within brand fixed effects.

xxii
Table WA.L5 – Using Category Medians to Operationalize the First Four SBF Variables

Expected Main
Predictors Effect Finding Estimate
Intercept .3107***
SBBEijq-1 .8632***
PRICEijq -.0083
ADijq .0126*
Strategic Brand LLijq .0360***
Factors DISTijq .0398***
ARCHij NA†
POSijq .0228**
EXPq .0146
EXPq * PRICEijq H1EXP: + + .4230*
Differential EXPq * ADijq H2EXP: + NS .4105
Effect of
EXPq * LLijq H3EXP: + + .5526*
Expansions for
Different Brands EXPq * DISTijq H4EXP: + + 1.5445*
EXPq * ARCHij H5EXP: - NS .3724
EXPq * POSijq + .7968***
CONq -.0481
CONq * PRICEijq NS .2364
Differential CONq * ADijq H2CON: + + .5797**
Effect of
CONq * LLijq H3CON: - - -.9597***
Contractions for
Different Brands CONq * DISTijq H4CON: + + .7332**
CONq * ARCHij H5CON: + + .5276***
CONq * POSijq H6CON: + NS .3679
OTHERSPPijq .0200
OTHERSADijq -.0001
OTHERSLLijq -.0056
Control OTHERSDISTijq -.2862***
Variables PLMSjq -.0595
Brand Fixed Effects Included
Year Fixed Effects Included
Quarter Fixed Effects Included
*** p<.01; ** p<.05; * p<.10 (one-sided p-values for hypothesized effects and two-sided for others). Significance
assessed using 2-way cluster-adjusted SEs (at brand and quarter levels). N=20,800 (due to the nature of
operationalization of most SBF variables that utilize past four quarters of data, we do not use first year of data [1994]
in our second-stage analysis).
† Main effect of ARCH is not identified since it is a time-invariant characteristic and hence the effect is subsumed
within brand fixed effects.

xxiii
 Table WA.L6 – Controlling for Lagged Effects of Marketing Mix Instruments
Expected Main
Predictors Effect Finding Estimate
Intercept .0108***
SBBEijq-1 .8594***
PRICEijq .0022
ADijq .0080
Strategic Brand LLijq .0345***
Factors DISTijq .0369***
ARCHij NA†
POSijq .0245**
EXPq .0313
EXPq * PRICEijq H1EXP: + + .6047**
Differential EXPq * ADijq H2EXP: + NS .3305
Effect of
EXPq * LLijq H3EXP: + + .5965*
Expansions for
Different Brands EXPq * DISTijq H4EXP: + + 1.5647**
EXPq * ARCHij H5EXP: - NS .2661
EXPq * POSijq + .7191**
CONq -.0722
CONq * PRICEijq NS .2107
Differential CONq * ADijq H2CON: + + .3848*
Effect of
CONq * LLijq H3CON: - - -.6697***
Contractions for
Different Brands CONq * DISTijq H4CON: + + .7245**
CONq * ARCHij H5CON: + + .5360**
CONq * POSijq H6CON: + NS .4357
OTHERSPPijq .0015
OTHERSADijq -.0001
OTHERSLLijq .0061
Control OTHERSDISTijq -.2251***
Variables PLMSjq -.0453
Brand Fixed Effects Included
Year Fixed Effects Included
Quarter Fixed Effects Included
*** p<.01; ** p<.05; * p<.10 (one-sided p-values for hypothesized effects and two-sided for others). Significance
assessed using 2-way cluster-adjusted SEs (at brand and quarter levels). N=20,800 (due to the nature of
operationalization of most SBF variables that utilize past four quarters of data, we do not use first year of data [1994]
in our second-stage analysis).
† Main effect of ARCH is not identified since it is a time-invariant characteristic and hence the effect is subsumed
within brand fixed effects.

xxiv
 Table WA.L7 – Allowing the Effects of Marketing Variables to vary across the Business
Cycle in the First-stage Model
Expected Main
Predictors Effect Finding Estimate
Intercept -.2343
SBBEijq-1 .3622**
PRICEijq -.0481
ADijq -.0037
Strategic Brand LLijq .1724***
Factors DISTijq .1916**
ARCHij NA†
POSijq .2185***
EXPq -.2826
EXPq * PRICEijq H1EXP: + + 4.1135**
Differential EXPq * ADijq H2EXP: + NS .6347
Effect of
EXPq * LLijq H3EXP: + + 2.3533*
Expansions for
Different Brands EXPq * DISTijq H4EXP: + + 7.3954*
EXPq * ARCHij H5EXP: - NS 2.5506
EXPq * POSijq + 2.5921*
CONq -.2880
CONq * PRICEijq NS 5.0389
Differential CONq * ADijq H2CON: + + 4.5851*
Effect of
CONq * LLijq H3CON: - NS 1.2563
Contractions for
Different Brands CONq * DISTijq H4CON: + + 8.5827**
CONq * ARCHij H5CON: + NS -4.7153
CONq * POSijq H6CON: + + 7.0962**
OTHERSPPijq -.0493
OTHERSADijq .0023
OTHERSLLijq .0239
Control OTHERSDISTijq -1.0064***
Variables PLMSjq -.2353
Brand Fixed Effects Included
Year Fixed Effects Included
Quarter Fixed Effects Included
*** p<.01; ** p<.05; * p<.10 (one-sided p-values for hypothesized effects and two-sided for others). Significance
assessed using 2-way cluster-adjusted SEs (at brand and quarter levels). N=20,800 (due to the nature of
operationalization of most SBF variables that utilize past four quarters of data, we do not use first year of data [1994]
in our second-stage analysis).
† Main effect of ARCH is not identified since it is a time-invariant characteristic and hence the effect is subsumed
within brand fixed effects.

xxv
 Table WA.L8 – Using Value Market Share in the First-stage Model

Expected Main
Predictors Effect Finding Estimate
Intercept .0144***
SBBEijq-1 .8644***
PRICEijq .0026
ADijq .0074
Strategic Brand LLijq .0294***
Factors DISTijq .0498***
ARCHij NA†
POSijq .0337***
EXPq -.0322
EXPq * PRICEijq H1EXP: + NS .2723
Differential EXPq * ADijq H2EXP: + NS .2035
Effect of
EXPq * LLijq H3EXP: + + .8385**
Expansions for
Different Brands EXPq * DISTijq H4EXP: + + 1.5654**
EXPq * ARCHij H5EXP: - NS .4064
EXPq * POSijq NS .5753
CONq -.0204
CONq * PRICEijq NS .0872
Differential CONq * ADijq H2CON: + + .5522***
Effect of
CONq * LLijq H3CON: - - -.4003**
Contractions for
Different Brands CONq * DISTijq H4CON: + NS .4307
CONq * ARCHij H5CON: + + .5255***
CONq * POSijq H6CON: + NS .0626
OTHERSPPijq -.0432
OTHERSADijq .0008
OTHERSLLijq .0003
Control OTHERSDISTijq -.2663***
Variables PLMSjq -.0553
Brand Fixed Effects Included
Year Fixed Effects Included
Quarter Fixed Effects Included
*** p<.01; ** p<.05; * p<.10 (one-sided p-values for hypothesized effects and two-sided for others). Significance
assessed using 2-way cluster-adjusted SEs (at brand and quarter levels). N=20,800 (due to the nature of
operationalization of most SBF variables that utilize past four quarters of data, we do not use first year of data [1994]
in our second-stage analysis).
† Main effect of ARCH is not identified since it is a time-invariant characteristic and hence the effect is subsumed
within brand fixed effects.

xxvi
 Table WA.L9 – Removing Lagged Market Share in the First-stage Model

Expected Main
Predictors Effect Finding Estimate
Intercept .3095***
SBBEijq-1 .8626***
PRICEijq -.0012
ADijq .0083
Strategic Brand LLijq .0341***
Factors DISTijq .0418***
ARCHij NA†
POSijq .0246**
EXPq -.0949
EXPq * PRICEijq H1EXP: + + .6165**
Differential EXPq * ADijq H2EXP: + NS .1655
Effect of
EXPq * LLijq H3EXP: + + .9600*
Expansions for
Different Brands EXPq * DISTijq H4EXP: + + 1.6825**
EXPq * ARCHij H5EXP: - NS .3898
EXPq * POSijq + .5672*
CONq -.1011
CONq * PRICEijq NS .2116
Differential CONq * ADijq H2CON: + + .3602*
Effect of
CONq * LLijq H3CON: - - -.5414***
Contractions for
Different Brands CONq * DISTijq H4CON: + + .6786**
CONq * ARCHij H5CON: + + .4897***
CONq * POSijq H6CON: + NS .2307
OTHERSPPijq .0197
OTHERSADijq -.0001
OTHERSLLijq -.0053
Control OTHERSDISTijq -.2866***
Variables PLMSjq -.0589
Brand Fixed Effects Included
Year Fixed Effects Included
Quarter Fixed Effects Included
*** p<.01; ** p<.05; * p<.10 (one-sided p-values for hypothesized effects and two-sided for others). Significance
assessed using 2-way cluster-adjusted SEs (at brand and quarter levels). N=20,800 (due to the nature of
operationalization of most SBF variables that utilize past four quarters of data, we do not use first year of data [1994]
in our second-stage analysis).
† Main effect of ARCH is not identified since it is a time-invariant characteristic and hence the effect is subsumed
within brand fixed effects.

xxvii
 Table WA.L10 – Removing Gaussian Copulas from the First-stage Model

Expected Main
Predictors Effect Finding Estimate
Intercept .0106***
SBBEijq-1 .8592***
PRICEijq -.0001
ADijq .0047
Strategic Brand LLijq .0387***
Factors DISTijq .0487***
ARCHij NA†
POSijq .0279**
EXPq .0606
EXPq * PRICEijq H1EXP: + NS .4827
Differential EXPq * ADijq H2EXP: + NS .5693
Effect of
EXPq * LLijq H3EXP: + + 1.1081**
Expansions for
Different Brands EXPq * DISTijq H4EXP: + + 1.4247*
EXPq * ARCHij H5EXP: - NS .4539
EXPq * POSijq + .5533*
CONq -.0898
CONq * PRICEijq NS .0653
Differential CONq * ADijq H2CON: + + .4380**
Effect of
CONq * LLijq H3CON: - - -.5650***
Contractions for
Different Brands CONq * DISTijq H4CON: + + .4799*
CONq * ARCHij H5CON: + + .3507*
CONq * POSijq H6CON: + NS .3077
OTHERSPPijq .0196
OTHERSADijq .0001
OTHERSLLijq .0099
Control OTHERSDISTijq -.2725***
Variables PLMSjq -.0670
Brand Fixed Effects Included
Year Fixed Effects Included
Quarter Fixed Effects Included
*** p<.01; ** p<.05; * p<.10 (one-sided p-values for hypothesized effects and two-sided for others). Significance
assessed using 2-way cluster-adjusted SEs (at brand and quarter levels). N=20,800 (due to the nature of
operationalization of most SBF variables that utilize past four quarters of data, we do not use first year of data [1994]
in our second-stage analysis).
† Main effect of ARCH is not identified since it is a time-invariant characteristic and hence the effect is subsumed
within brand fixed effects.

xxviii
 REFERENCES NOT IN THE MANUSCRIPT

Christiano, Lawrence J. and Terry J. Fitzgerald (1998), “The Business Cycle: It's Still a
Puzzle,” Economic Perspectives-Federal Reserve Bank of Chicago, 22, 56-83.

Cleeren, Kathleen, Harald J. van Heerde, and Marnik G. Dekimpe (2013), “Rising from the Ashes:
How Brands and Categories Can Overcome Product-harm Crises,” Journal of Marketing, 77
(2), 58-77.

Deleersnyder, Barbara, Marnik G. Dekimpe, Jan-Benedict E.M. Steenkamp, and Peter SH

Leeflang (2009), “The Role of National Culture in Advertising’s Sensitivity to Business
Cycles: An Investigation across Continents,” Journal of Marketing Research, 46 (5), 623-36.

Hodrick, Robert J. and Edward C. Prescott (1997), “Postwar US Business Cycles: An Empirical
Investigation,” Journal of Money, Credit, and Banking, 29 (1), 1-16.

Kesavan, Saravanan and Tarun Kushwaha (2014), “Differences in Retail Inventory Investment
Behavior during Macroeconomic Shocks,” Production and Operations Management, 23 (12),
2118-36.

Ravn, Morten O. and Harald Uhlig (2002), “On Adjusting the Hodrick-Prescott Filter for the
Frequency of Observations,” Review of Economics and Statistics, 84 (2), 371-76.

xxix