CARDIFF BUSINESS SCHOOL WORKING PAPER SERIES

Cardiff Marketing and Strategy Working Papers Gordon Foxall, Ji Yan, Victoria James and Jorge Oliveira-Castro Brand-related and situational influences on demand elasticity M2009/2

Cardiff Business School Cardiff University Colum Drive Cardiff CF10 3EU United Kingdom t: +44 (0)29 2087 4000 f: +44 (0)29 2087 4419 www.cardiff.ac.uk/carbs ISSN: 1753-1632 November 2009

This working paper is produced for discussion purpose only. These working papers are expected to be published in due course, in revised form, and should not be quoted or cited without the author’s written permission.

1

Brand-related and situational influences on demand elasticity Gordon Foxall1 Ji Yan1 Victoria James1 Jorge Oliveira-Castro2 1

Consumer Behaviour Analysis Research Group, Cardiff University 2 Consuma, University of Brasilia

Address correspondence to Gordon Foxall Cardiff Business School, Cardiff University, Aberconway Building, Colum Drive, Cardiff CF10 3EU, UK. [email protected]

2

Brand-related and situational influences on demand elasticity

Abstract The paper reports an investigation of variations in demand elasticity for foods, the aim of which was to investigate whether price elasticity varies from product to product, from brand to brand, with differing brand attributes that contribute to combinations of functional and symbolic benefit to consumers, and for varying price dynamics. Using panel data for 1500+ British consumers purchasing 4 food products with 2000+ brands over 52 weeks, the study also examines how factors other than price affect demand elasticity for brands. Price elasticity differs among products and brands, for price variations occurring at different speeds. Moreover, differing combinations of the informational and utilitarian benefits provided by brands influence the level of price elasticity exhibited by different brand types. These results differ significantly from those of an earlier study by Ehrenberg and England. The paper discusses reasons for this discrepancy in terms of brand-related and situational influences on consumer choice. Keywords: elasticity of demand; brand choice; utilitarian benefit; informational benefit; food products

3

Brand-related and situational influences on demand elasticity Marketing scholars and practitioners often claim that brands have a characteristic price-elasticity. Evidence for this comes from experimental investigations (e.g. Urban & Hauser, 1980, Narasimhan & Sen, 1983; Shoker & Hall, 1986; Mahajan & Wind, 1986) which report that a brand or product has its own elasticity, a finding that corroborates demand theory (Telser, 1962; Broadbent, 1980; Roberts, 1980; Nagle, 1987; Gabor, 1988). Price is not the sole factor affecting elasticity, of course (e.g. Scriven & Ehrenberg, 1999), but consumers’ price sensitivities make a central input to marketing strategy and tactics (Anderson & Simester, 2009; Ratchford, 2009). Any study which generates contrary results and which remains influential some 20 years after its initial publication therefore deserves respect and attention. Ehrenberg and England (1990) reported, on the basis of an experimental study, that elasticity for foods does not differ significantly across brands and products, even when prices are rising or falling and doing so at different speeds. This paper examines these claims by describing an investigation based on direct observation of consumer choice which assesses elasticity across food products and brands. The findings indicate that elasticity for such products and brands is dynamic to an extent not identified by Ehrenberg and England and that the functional and symbolic characteristics of brands are systematically related to patterns of elasticity. Most research into factors that influence elasticity focuses on consumer-related characteristics, demographics and psychographics, but shows little consensus (cf. Trier et al., 1960; Gabor & Granger, 1961; Coe, 1971; Murphy, 1978; Zeithaml & Fuerst, 1983; McGoldrick & Marks, 1987; Sirvanci, 1993; Hoch et al., 1995; Mazumdar & Papatla, 1995; Dillon & Gupta, 1996; George et al., 1996; Jones & Mustiful, 1996; Dhar & Hoch, 1997; Kallyanam & Putler, 1997; Ainslie & Rossi, 1998; Mulhern et al., 1998; Bell et al., 1999; Kim et al., 1999; Sethuraman & Cole, 1999; Erdem et al., 2001; Estelami & Lehmann, 2001; Kenesei and Todd, 2003; Ailawadi & Keller, 2004; Boatwright et al., 2004; Rosa-Diaz, 2004; Scriven & Ehrenberg, 2004; Rao, 2009). Few investigations consider brand-related characteristics as factors affecting elasticity. Foxall et al. (2004) present evidence that consumer behavior is influenced by utilitarian (functional) and informational (symbolic) benefits. Utilitarian benefits include functional outcomes of purchase and consumption, derived from the use of the product itself. Informational benefit, in contrast, is symbolic, social, and mediated by the actions and reactions of other people. While utilitarian benefit is related to economic and functional benefits of products or services, informational benefit is related to social status and prestige, associated with buying, owning, or using products or services. In the case of packaged goods, increased utilitarian benefits relate to product formulations that offer more or better functional attributes, such as baked beans with sausage vs. plain baked beans, or chocolate chips cookies vs. plain digestive cookies. These formulations usually have distinct brand identities and differ in price. Based on earlier research which has identified differences in elasticity between groups of consumers purchasing brands embodying various combinations of utilitarian and informational reinforcement (Oliveira-Castro et al., 2005, 2006, 2008a,

4

2008b, 2010), we expected inclusion of these variables in the calculation of elasticity coefficients would produce results rather different from those reported by Ehrenberg and England. A consideration that arises from this involves the rather different contexts in which our research and that of Ehrenberg and England took place. Utilitarian and informational outcomes of consumer choice are sources of benefit that have been conceptualized as post-behavioral controlling consequences of choice within the Behavioral Perspective Model (BPM; Foxall, 2004) which also includes pre-behavioral mechanisms for the integration of consumption history and current stimuli that predict behavioral outcomes. This complex of predictive stimuli is known as the consumer behavior setting and its scope is determined by the number of competing choices it signals as available to the consumer. A setting that permits only a single behavior to be enacted within it or, at best, a few behaviors is known as a closed setting and is exemplified by being a dental patient: although one is at liberty to leave the surgery at any time, most people feel constrained to follow the single program of behaviors that define being a patient. By contrast, an open setting permits numerous alternative behaviors and is exemplified by a buffet at which the consumer is able to select among many foods and drinks and combinations of foods and drinks, to move around more or less at will, to speak to whomever she chooses, and to leave at any time. The idea of a continuum of consumer behavior settings defined in terms of their scope is of particular interest in the quest for factors that determine the elasticity of demand for food products and brands, for elasticity varies with the number of substitute behaviors (hence, products or brands) available to the consumer. Elasticity is higher when more rather than fewer substitutes are available. We would predict, therefore, that elasticity would be more dynamic in the case of the relatively open settings in which our investigation occurred than for the relatively closed experimental settings in which Ehrenberg and England’s research was conducted. The first question is whether there are consistent patterns in elasticity for products and brands, as prices rise and fall, and as they do so at different speeds. Eqn (1) measures the relationship between the quantity a consumer buys and the price paid (Kagel et al., 1995): Log Quantity = a – b (log Price), in which b represents the price elasticity coefficient. The second question is whether brand related characteristics affect price elasticity, in particular, considering utilitarian benefit and informational benefit of brands. Such patterns are measured by decomposing the elasticity into three coefficients: brand price, informational coefficients and utilitarian coefficients. Eqn (2) measures the relationships among quantities bought, prices paid, utilitarian benefits of brands and informational benefits of brands: Log Q = a – b1 (log P) – b2 (log INF) – b3 (log UTIL), where Q is quantity purchased, P is price, INF is informational benefit level, and UTIL is utilitarian benefit level. Based on the conceptual proposal and results in the literature, and especially to compare levels of elasticity in natural purchase settings with the experimental results reported by Ehrenberg and England, the following hypotheses were tested: H1: Price elasticity varies across different brands and products. H2: Price elasticity varies across products and brands when prices are changing at different speeds. H3: Informational and utilitarian benefits are

5

elements of brands that are embodied in their buyers’ sensitivity to price changes. H4: Informational benefits are positively correlated with the quantity consumers buy. H5: Utilitarian benefits are positively correlated with the quantity consumers buy. H6: Informational benefits account for more variance in quantity bought than utilitarian benefits. H7: Own price elasticity of a specific brand should account for more variance in the quantity consumers buy than informational and utilitarian benefits. Method TM Data from the ACNielsen Homescan panel, based on 10,000+ UK households that used home barcode scanners, including information about four product categories during 52 weeks from July 2004 were obtained for baked beans, biscuits, fruit juice, and yellow fats; there was data for 832, 1594, 895, and 1354 households, respectively, for these products including, for each purchase, the brand, store, item characteristics, pack size, amount spent, number of items, and date. The methods used to assess brand characteristics were as follows (Oliveira-Castro et al., 2005, 2008a, 2008b). Informational benefit (INF) offered by each brand was measured by a questionnaire on which a convenience sample (n=33) rated each brand in terms, first, of how well-known they judged it to be: (0 not at all, 1 a little, 2 quite well known, 3 very well known); and, second, the brand’s perceived quality (0 unknown, 1 low, 2 medium, 3 high). Higher levels of utilitarian benefit (UTIL) manifest in additional attributes which are considered to have value-adding qualities for the product or its consumption; they are visibly declared on the package or are part of the product name, and ultimately justify higher prices. Moreover, in most cases, several general brands offer product varieties with and witho8ut these attributes. UTIL was assessed by adopting the same ranking procedure used in previous studies (Oliveira-Castro et al., 2005). Plain formulations of items were ranked as having a UTIL of 1, whereas more sophisticated formulations were ranked as having a UTIL of 2). Brands were classified into groups, derived from the combination of 2 UTIL levels and 3 INF levels: (1) INF 1, UTIL 1; (2): INF 1, UTIL; (3) INF 2, UTIL 1; (4) INF 2, UTIL 2; (5) INF 3, UTIL 1; (6): INF 3, UTIL 2. Results and Discussion Results for Eqn (1) are shown in Tables 1 and 2. All regressions are significant; all elasticity coefficients for groups are negative; all fall between 0 and -1.0, indicating inelastic demand. Absolute coefficient values (Figure 1(a)) are lower for groups 1-3 than the others, indicating brands with lower functional and symbolic attributes show higher price responsiveness. Elasticity for each brand group within each product differs from the price elasticity for each brand group across all products (Table 2, Fig 1(b)). Absolute elasticity differs among brand groups within each product category, among the same brand groups across products, among different brand brands across product categories, and with the overall elasticity of each product. All but one of the absolute elasticity values across products and brand groups fall between 0 and 1. The highest inelasticity groups for each product are shown in Table 2. Absolute elasticity (Figure 2) varies substantially by group. H1 is accepted. Different groups show lowest inelasticity across brands within each product category. Unit prices for each product rose and fell at different speeds. Figure 2

6

indicates lack of consistency in price changes across the seasons for the products. All slopes within each product and for the brand groups differ, indicating that the price dynamics of each product and brand group differ with the season, i.e. average unit prices rose or fell over the seasons at different speeds. Chow-tests (Chow, 1960) for regressions for products based on Eqn (1) are shown in Tables 3; all are significant: the independent variables have different impacts on the dependent for the various subgroups. Unit price impacts differently on quantity bought in each season by products and brand groups. Elasticity for each product thus differs when prices rise and fall at different speeds. H2 is accepted. All Eqn (2) regressions are significant (Table 4). Adjusted-R2 is improved by adding UTIL and INF variables in all cases (Table 5), albeit marginally. Nevertheless, UTIL and INF enhance the explanation of price responsiveness. H3 is accepted. B1, B2, and B3 are significant for all products, indicating that consumers respond to all the benefit attributes in Eqn (2). Values of B1 are all negative indicating that increases in price were associated with decreases in quantity demanded. B3 is negative for baked beans, biscuits and juice; indicating that increases in UTIL are associated with decreases in quantity demanded for the three products. B2 values are all positive across products: increases in INF are significantly related to increases in quantity demanded. H4 is accepted. B3 values are positive in biscuits showing that UTIL positively influences Q. Hence, H5 is partially accepted. B1 is larger than B2 and B3 in all cases, so H7 is accepted. B2 values are larger than B3 for baked beans and biscuits; B2 smaller than B3 for biscuits. H6 is rejected since UTIL accounts for more variance in quantity bought than INF. Q varies with changes in prices and benefits that occur within and across products and within and across brands. Price elasticity varies across products and brands, contra Ehrenberg and England. These findings corroborate evidence that price elasticity differs among products and brands. Q varies also as price rises and falls, and with the speed with which it does so. Price elasticity explains a larger portion of variance in quantity bought than do informational and utilitarian benefits, and does so within and across products and across brands. However, UTIL and INF do contribute to the explanation of variance: they individually influence Q. Brand related characteristics can, therefore, contribute to consumers’ price responsiveness. Although an obvious source of discrepancy between Ehrenberg and England’s results and ours lies in the experimental vs. real-world contexts, there is a particularly significant element of this methodological diversity which bears on the discrepancy: the number of substitutes available in our study. In their experimental study, despite the fact that consumers could buy at any store, the closeness of the experimental shop, which offered few brands, could in part explain the constant elasticity found. One could argue, nevertheless, that their closed setting should decrease elasticity coefficients but not make them constant, but the scope of the consumer behavior settings involved in these studies appears to be an important source of the differences in elasticity (cf. Gijsbrechts, 1993). Economic theory suggests that the presence of substitutes for a target commodity increases the price elasticity of demand for it. Since open settings by definition permit a number of competing behaviors to be available to

7

the consumer, we would expect the price elasticity of demand for any commodity for which the consumer is in a state of deprivation to be greater than would be the case in a closed setting where, at the extreme, only one reinforcer is available and there is only one behavioral route to its delivery. The scope of the consumer behavior setting is related to Hursh’s (1980) concept of open and closed experimental economies (Foxall and Schrezenmaier, 2003). In the former, participants are allowed to supplement the reinforcers they obtain during experiments by means of post-sessional consumption. In the latter, the experimental sessions are the only source of the reinforcer. Demand for a commodity shows greater price elasticity when the economy is open: the availability of an alternative source of the reinforcer makes the respondent less likely to work for it during the experiment. The discrepancy between Ehrenberg and England’s results and our findings thus elucidates the theoretical basis of the BPM. The more open setting, which featured a wider range of brands available in a range of retail settings at prices that differed according to the competitive pressures of natural marketing settings, notably an extensive spectrum of complete marketing mixes, exhibited a more dynamic interaction of price elasticity among products and brands than did the experimental context. Table 1. Parameters of Eqn (1) calculated for each brand group, the significance level of the regression (P), and the standard error of the estimate of b. Brand

R-Square

B

S.E.

P

VIF

D-W

N

Group 1

.245

-.527

.002

.000

1.000

.989

16466

Group 2

.302

-.511

.007

.000

1.000

1.207

5641

Group 3

.258

-.497

.010

.000

1.000

.828

39539

Group 4

.311

-.558

.006

.000

1.000

.925

21451

Group 5

.322

-.605

.005

.000

1.000

.981

33023

Group 6

.241

-.533

.006

.000

1.000

1.131

24790

Group

Table 2. Absolute values of the price elasticity of demand across products and brand groups All |Ed|

Beans

Butter

Biscuits

Juice

products

Overall

0.818

0.409

0.570

0.468

0.539

Group 1

0.309

0.525

0.587

0.529

0.527

Group 2

0.699

0.589

0.477

0.578

0.511

Group 3

0.348

0.493

0.584

0.437

0.497

Group 4

0.555

0.583

0.563

0.565

0.558

Group 5

1.86

0.476

0.569

0.5

0.605

Group 6

0.879

0.024

0.66

0.077

0.533

Largest absolute values

5 and 6

2 and 4

1 and 6

2 and 4

4 and 5

Smallest absolute values

1 and 3

5 and 6

2 and 4

3 and 6

2 and 3

Table 3. Results from performing Chow-test for regressions based on 4 product categories and 6 Brand Groups

8

Tests of Between-Subjects Effects Dependent variable: Log Quantity; Independent variable X: Log (Price) Source

Df

F

Sig.

df

F

Sig.

4

13761.924

.000

6

9064.843

.000

Intercept

1

39725.828

.000

Intercept

1

39667.845

.000

Log( Price)

1

23162.144

.000

Log (Price)

1

36338.542

.000

Product * X

3

277.942

.000

Group * X

5

71.593

.000

Error

140905

Error

140903

Total

140910

Total

140910

Corrected Total

140909

Corrected Total

140909

Corrected Model

Source Corrected Model

a. R Squared = .281

a. R Squared = .279

Table 4. Regression of quantity bought (Q) of products on unit price of products (P), informational benefit (INF) and utilitarian benefit (UTIL) for all purchases and shopping trips

All purchases on all shopping trips R2=.235

D-W= .561

n=13671

R2=.388

D-W=1.170 n=30538

Value

Error

Sig.

Value

Error

Sig.

Constant

4.744

.037

.000

Constant

4.792

.005