A Meta-analysis of Loss Aversion in Product Choice

Highlights

• We investigate loss aversion in individual-level random utility models.

• On average, loss aversion seems manifest in product choice.

• We find substantial effect variation in loss aversion across research contexts.

• Methodological decisions play a crucial role in loss-aversion results.

• Loss sensitivity does not significantly differ for price and quality dimensions.

Abstract

Loss aversion is a behavioral phenomenon with game-changing implications for economic theory and practice. We conduct a meta-analysis of 33 studies (providing 109 observations) investigating loss aversion in random utility models of brand choice. Specifically, we use multilevel modeling techniques to examine potential moderators of preference asymmetries as well as the variability of loss-aversion effects within and between studies. We find that loss aversion is manifest in product choice, but that it exhibits substantial variation across research contexts. Product-related variables (e.g., the category type), consumer characteristics (e.g., reference-point mechanisms), and particularly methodological decisions (e.g., model specifications) determine the observed degree of loss–gain disparity. Practical implications of the specific findings and opportunities for future research are discussed.

Introduction

“Losses loom larger than gains” is a popular statement that summarizes the concept of loss aversion (Brenner et al., 2007, Kahneman and Tversky, 1979). This asymmetry in consumer behavior with regard to framed losses (e.g., prices above a reference brand) and gains (e.g., prices below a reference brand) has critical implications for competitive strategies of retailers. Because consumers weigh product attribute levels with disproportionate importance, the microeconomic rules of competition are subject to change; thus firms need to adjust their product positioning to maximize profits (Heidhues and Kőszegi 2008). On a tactical level, enhanced loss sensitivity could render promotional campaigns ineffective overall, as altering price-reference levels could change long-term behavior unfavorably (Erdem, Mayhew, and Sun 2001). Even for growth strategies, loss aversion can play a crucial role: managers considering entry into a new market [segment] should closely monitor the market leader brand, against which their product is likely to be compared (Hardie, Johnson, and Fader 1993).

Given the theoretical and practical relevance, it is not surprising that a great deal of research is devoted to loss–gain differences. In consumer choice, loss aversion has been found to explain preferences for products as diverse as eggs (Putler 1992), orange juice (Hardie, Johnson, and Fader 1993), and real estate (Genesove and Mayer 2001). Moreover, in the reference-price literature, loss aversion has been proposed as an empirical generalization (Kalyanaram and Winer, 1995, Meyer and Johnson, 1995).

However, despite the wide acceptance of loss aversion both as an effect and as a theory (Willemsen, Böckenholt, and Johnson 2011), the results across studies are quite mixed with respect to the significance and magnitude of the phenomenon (Bell and Lattin, 2000, Klapper et al., 2005). For example, in the area of price effects, some researchers find strong evidence for loss aversion in their data (Kalwani et al., 1990, Kalyanaram and Little, 1994), while other studies find none (Briesch et al., 1997, Chang et al., 1999) or only partial support for loss aversion (Klapper, Ebling, and Temme 2005). Similarly, when investigating choices of digital cameras, Ataman and Rooderkerk (2010) do not find higher loss than gain sensitivity for all tested quality features.

The large unexplained heterogeneity in loss aversion demonstrates that little is known about how the phenomenon varies across people, measures, and choice situations (Dhar and Wertenbroch, 2000, Klapper et al., 2005). Why do some categories seem to exhibit loss aversion and others not? Are there specific consumer characteristics or product attributes that drive loss aversion? What differences across the loss-aversion methods and modeling approaches could provide an explanation of the identified effect-size discrepancies?

In this work, we use a meta-analytic framework to address these questions. The advantage of a meta-analysis is that it allows conclusions that are more credible and accurate than can be obtained in a single primary study (Rosenthal and DiMatteo 2001). Specifically, the objectives of the meta-analysis in this research are: (1) to investigate the degree of loss-aversion heterogeneity in product choice, (2) to test the impact of potential moderators on the phenomenon in this context, and (3) to assess how much of the effect variation can be explained by the variables included in the model. To achieve this, we employ a multilevel random-effects technique, which permits us to examine the variability of effect sizes within and between studies (Hox 2010). Because meta-analyses are particularly useful in explaining effect heterogeneity based on between-study differences, the distinction between these two variance sources helps us assess the explanatory power of our model while simultaneously revealing a roadmap for future work. Conforming to meta-analysis practices, the moderators tested in our multilevel model comprise study and observation differences based on methodology decisions, consumer- and product-related characteristics, and other relevant variables (see Table 1).

The remainder of this paper is organized as follows. First we specify the research domain of this work. Next we discuss moderators that potentially could explain effect differences and are therefore tested in the meta-analysis. This is followed by a detailed description of the methodology. Then we present the results and conclude with a discussion of the implications and research directions emanating from this research.