|
Article Excerpt
Marketing literature has long recognized that brand price elasticity need not be monotonic and symmetric, but has yet to provide generalizable market-level insights on threshold-based price elasticity, asymmetric thresholds, and the sign and magnitude of elasticity transitions. This paper introduces smooth transition regression models to study threshold-based price elasticity of the top 4 brands across 20 fast-moving consumer good categories. Threshold-based price elasticity is found for 76% of all brands: 29% reflect historical benchmark prices, 16% reflect competitive benchmark prices, and 31% reflect both types of benchmarks. The authors demonstrate asymmetry for gains versus losses on three levels: the threshold size and the sign and the magnitude of the elasticity difference. Interestingly, they observe latitude of acceptance for gains compared to the historical benchmark, but saturation effects in most other cases. Moreover, category characteristics influence the extent and the nature of threshold-based price elasticity, while individual brand characteristics impact the size of the price thresholds. From a managerial perspective, the paper illustrates the sales, revenue, and margin implications for price changes typically observed in consumer markets.
Key words: kinked demand curve; smooth-transition regression models; time-series analysis; asymmetric price thresholds; empirical generalizations
1. Introduction
Marketing researchers and practitioners have long acknowledged that price response functions need not be monotonic and symmetric (e.g., Gutenberg 1976, Simon 1969). Kinked demand curves (Puffer 1992) imply that brand price elasticity might be subject to price benchmarks or thresholds. For example, shallow discounts might fail to generate consumer response and thus have underproportional effects on market performance compared to deep discounts (Gutenberg 1976, Hruschka 2000, Gilbride and Allenby 2004). At the same time, consumers might react strongly to even relatively minor price increases while habituation/adaptation (Kahneman 1991) leads to saturation effects for major price increases (van Heerde et al. 2001). Managerial interest in this topic is twofold: (1) to predict the sales and profit impact of different levels of price increases and decreases, and (2) to identify the category and brand characteristics that affect price elasticity thresholds (Han et al. 2001). As managers typically assess threshold effects by simple methods based on a cross-tabulation of sales versus price points across stores, Bucklin and Gupta (1999) call for more academic research on price threshold analysis. In this context, while complex threshold effects have been widely discussed (e.g., Moran 1978, Simon 1989), they have often escaped explicit modeling and empirical observation.
From a research perspective, there have been two sophisticated approaches to the problem of estimating price thresholds. First, individual-level analyses showed asymmetric thresholds around a reference price, with a "latitude of acceptance" region or region of indifference such that changes in price within this region produce no changes in perception (Monroe 1990). However, their focus remained restricted to the specific behavioral phenomenon of interest: historical or competitive reference prices and assimilation/contrast effects or saturation effects (Gupta and Cooper 1992, Thaler 1985, Han et al. 2001). Second, completely data-driven approximation of the effect curve offered more flexible estimation approaches to capture a wide variety of price threshold phenomena (van Heerde et al. 2001, Kalyanam and Shively 1998). Unfortunately, this flexibility comes at the expense of severe data requirements and difficult interpretation of the parameters, especially across categories to generate guidelines for retail pricing.
Thus, while research points to the existence of brand price thresholds and kinked demand curves, the extant marketing literature lacks a large-scale econometric investigation of this phenomenon across product categories in retail markets. In particular, retail pricing managers need insights into the moderating factors of threshold-based price elasticity at the aggregate level, where they have to set prices and are accountable for the sales results. A systematic comparison across brands and categories is therefore needed to uncover empirical generalizations, to offer concrete managerial guidelines (Shugan 2003), and to identify important areas for future research. As a result, we seek to address the following research questions: (i) Is there time-series evidence of thresholds in price elasticity across a wide variety of fast-moving consumer good categories? (ii) To what extent are such deviations from constant price elasticity driven by historical versus competitive benchmark prices (hereafter HBP versus CBP)? (iii) Is there time-series evidence for asymmetric thresholds and slope changes (latitude of acceptance versus saturation effects) for gains and losses? (iv) Do these characteristics of price elasticity vary across categories and brands? We apply the methodology of logistic smooth-transition regression (STR) models (see Van Dijk et al. 2002, among others) to assess the impact of price thresholds on price elasticities.
The rest of the paper is organized as follows. In [section] 2, we propose a research framework and hypotheses for both price discounts and price hikes on three dimensions: the nature of the benchmark (historical versus competitive), the size of the price threshold (small versus large), and the price slope difference (latitude of acceptance versus saturation effects). Next, we focus on the category and brand characteristics that might influence the presence, nature, and size of price thresholds and price elasticity differences. In [section] 3, we discuss the econometric representation of the model we use to examine threshold-based transitions in short-run price elasticity. Section 4 describes the data and operationalization of the variables, while [section] 5 reports the results. Finally, we formulate conclusions and future research avenues in [section] 6.
2. Thresholds in Short-Run Price Elasticity
Over the past decade, researchers have identified thresholds in price elasticity (for a review, see Kalyanaram and Little 1994, and Raman and Bass 2002) and have called for further exploration of this issue (Bucklin and Gupta 1999, Simon 1989). Remaining issues include (1) the nature of these price thresholds or benchmarks, (2) the size of the thresholds, and (3) the sign of the change to the price elasticity. Moreover, it is not clear to what extent brand and category moderators influence these characteristics of price thresholds in retail markets. We discuss these in turn.
2.1. Characteristics of Price Thresholds
First, researchers have typically assumed that consumers use either a historical (temporal) benchmark price (1) or a competitive (contextual) benchmark price in brand choice decisions (Briesch et al. 1997). The former view argues that consumers remember the prices encountered on past purchase occasions while the latter view argues that a benchmark price is formed during the purchase occasion on the basis of the prices observed (e.g., shelf prices of competing products). This distinction in benchmark price formation is important for market-level price setting. Historical benchmark prices imply that managers should beware of own past discounting as brand price should compare favorably with past own prices, whereas competitive benchmark prices focus management attention on current competitive prices as brand price should compare favorably with those at the point of purchase (Mazumdar and Papatla 2000, Rajendran and Tellis 1994). Interestingly, the few papers that analyzed both historical and competitive benchmark prices find that both benchmark types matter (Kumar et al. 1998, Mayhew and Winer 1992, Rajendran and Tellis 1994, Mazumdar and Papatla 2000). Because these studies analyzed one or a few product categories, we do not yet know under which circumstances either type is more important.
Second, the observed threshold size is important for the interpretation and managerial implications of threshold-based price elasticity. Smaller thresholds, typically less than 15%, have been interpreted as an assimilation effect in consumer price perception and encoding (Kalyanaram and Winer 1995). Instead, larger thresholds could reflect intentional consumer behavior of lie in wait for even better deals (Mela et al. 1997). Moreover, threshold size could be asymmetric to gains (price decreases) versus losses (price increases) (Kalyanaram and Little 1994, Moran 1978). Recently, Han et al. (2001) found larger thresholds for gains versus losses in the coffee category. It is currently unclear whether this finding generalizes to other categories. A second asymmetry has been found for the magnitude of the elasticity difference, as consumers react more to perceived price losses than to price gains (Kalyanaram and Winer 1995) or vice versa (Greenleaf 1995, Krishnamurthi et al. 1992).
Finally, most researchers have focused on demonstrating a latitude of price acceptance, implying amplification of the price elasticity beyond a threshold (Sherif et al. 1965). In contrast, recent research has shown the possibility of saturation effects, implying attenuation of the price elasticity beyond a threshold (van Heerde et al. 2001). The distinction is crucial for pricing managers, as it implies either larger or smaller bang for the buck once the price change exceeds the threshold. Table 1 juxtaposes "latitude of acceptance" and "saturation" effects for negative price gaps (gains to the consumer; price discounts to the manager) and positive price gaps (losses to the consumer, price hikes to the manager).
Several consumer behavior theories are consistent with the four scenarios in Table 1.
* For negative price gaps (consumer gains), a latitude of acceptance is implied by adaptation level theory and assimilation-contrast theory (Kalyanaram and Little 1994, Kalwani et al. 1990): Before consumers can contrast the low price with their benchmark, the price must be perceived as different. Moreover, even when they perceive and recognize discounts, consumers may not react strongly if they are waiting for still better deals (Mela et al. 1997, Kopalle et al. 1999). Interestingly, both assimilation-contrast and "lie-in-wait" effects have been demonstrated only vis-a-vis an historical benchmark (i.e., the past price of the focal brand), not vis-a-vis competitive benchmarks.
* In contrast, saturation effects for gains are consistent with consumers engaging in "discounting of discounts" (Gupta and Cooper 1992). Intuitively, consumers do not fully consider that the price is that much lower than the benchmark and adjust their gain perception to more reasonable levels. Alternatively, saturation effects in retail markets may originate from consumer limits to purchasing, transporting, and stockpiling products (van Heerde et al. 2001). These physical limits may apply to discounts compared to both historical or competitive benchmarks.
* For positive price gaps (consumer losses), a latitude of acceptance is again consistent with adaptation level theory: A loss must also exceed a consumer's price threshold in order to be perceived. Instead, minor price hikes within the threshold are less likely to be noticed (Kalyanaram and Little 1994, Kalwani and Yim 1992).
* Saturation effects for consumer losses may represent a "discounting of price hikes," i.e., consumers may mentally adjust price increases to more "reasonable" levels. Such behavior might occur as a rationalization for buying products at higher prices, for instance, for indulgence products, or simply reflect a partial encoding of the price increase (Alba et al. 1991). Beyond perception, saturation effects are also consistent with the presence of a core loyal consumer base with a strong need or desire for the focal brand (Jacoby and Chestnut 1978). While these consumers might buy less quantity as brand price increases, they do not refrain altogether from buying the focal brand, even at very high prices.
In sum, empirical generalizations on price thresholds should consider both "latitude of acceptance" and "saturation" effects and allow for asymmetric thresholds for gains and losses. Figure 1 visualizes these different elements and provides definitions of key parameters.
[FIGURE 1 OMITTED]
2.2. Moderating Role of Category and Brand Characteristics
Because managers have a keen interest in which of the identified scenarios apply under which circumstances, we develop hypotheses on the drivers of (1) the relative importance of historical versus competitive price benchmarks, (2) the price slope (elasticity difference) beyond the price threshold (latitude of acceptance versus saturation effects), and (3) the location of these benchmarks (threshold size). Prior marketing theory drives our selection of the second-stage covariates, which include category/product and brand characteristics. The former are of key interest to retailers (and multicategory manufacturers) setting pricing guidelines across categories (Shankar and Bolton 2004), while the latter are important to both retailers and brand managers.
2.2.1. Historical or Competitive Benchmarks? Latitude of Acceptance or Saturation Effects? We gauge the likelihood for historical versus competitive benchmarks by adapting the accessibility-diagnosticity framework (Feldman and Lynch 1988). In particular, the prominence of historical price benchmarks increases with (1) how likely the consumer is to remember past prices (e.g., Biehal and Chakravarti 1983), and (2) how diagnostic this memory of past prices is in predicting current/future prices (e.g., Briesch et al. 1997). We expect these drivers to also affect the price elasticity beyond the gain threshold, i.e., whether large price discounts yield higher price sensitivity (2) (latitude of acceptance) or lower price sensitivity (saturation effects).
HYPOTHESIS 1A. Historical benchmarks are more prominent in expensive categories.
Expensive categories should draw greater attention to prices relative to less expensive categories purchased at the same outlet. As such, consumers are more likely to recall the price because it stands...
|
