The Price Elasticity of Selective Demand: A Meta-Analysis of Sales Response Models
Jan 1, 1988
Price elasticity of selective demand.
Type of Report
Meta-analysis of 367 econometric models of price elasticity of sales and market share.
To provide a mean estimate of the price elasticity reported in previous studies and explain its variation.
Develops a meta-analytic model; defines independent variables and hypotheses; summarizes results of the meta-analysis and highlights implications for further research.
Modelers, marketing researchers, and promotion and pricing managers.
The Meta-Analytic Approach
For more than 30 years, researchers have been estimating the sensitivity of selective demand to changes in competitors’ prices. Even though these studies have ranged over disparate environments and methods, they reveal an underlying similarity. All have involved econometric models of the response of actual market sales or market share to market prices. In particular, most of the studies estimated the response sensitivity in the form of price elasticity. This price elasticity is a pure number, not bound by units, and reflects the percentage change in sales or market share for a one percent change in pace.
The question that naturally arises is: What have we learned from this body of research? What lessons can we derive from these disparate studies? The meta-analytic technique is a useful tool for answering these questions. In this case, meta-analysis applies the same econometric method used in the primary studies to the collected results. Specifically, the author has pooled estimates of price sensitivity from each of the studies, explaining their variation as a function of characteristics of the models which generated the estimates.
Price elasticity has been heavily researched, especially in marketing. This study includes 367 usable estimates of price elasticity from more than 220 different brands or markets. About 90 percent of the authors of these studies were marketers; the rest were economists. Most of the research was published in marketing journals, primarily the Journal of Marketing Research.
The uncorrected arithmetic mean estimate of price elasticity is -1.76, with a very narrow distribution around the mean. So as one would normally assume, price elasticity is significantly negative. However, this estimate contrasts dramatically in sign and size with the estimate of advertising elasticity of 0.22 obtained from a similar meta-analysis of advertising models by Assmus, Farley, end Lehmann (1984).
The price elasticity differs systematically by method in directions that are consistent with expectations from econometric theory. For example, if researchers did not control for distribution, the elasticity was significantly more negative or negatively biased, and if they did not control for quality, it was significantly less negative or positively biased. The most severe positive bias was due to the estimation of the elasticity with only cross-sectional data, due probably to capturing competitive positions or market share strategies with the latter data. Models that were estimated with data that was temporally aggregated above the weekly level also led to a positive bias, due to the loss of the time varying fluctuations and an emphasis on the cross-sectional variation.
The price elasticity also differs systematically by environment. It is less negative in the early stages of the life cycle, probably because of inadequate consumer information about prices. Similarly, it is less negative for pharmaceutical products, perhaps because pharmaceutical items are purchased on prescription or under conditions of some urgency.
The most important implication of the study is that sales and market share are very responsive to changes in price. After accounting for various biases in estimation, the absolute value of the price elasticity is about 10 times higher than that of advertising. While this study is the first to integrate all the diverse evidence to document this point, marketers have been increasingly aware of the phenomenon. That is probably why marketing expenditures have steadily been shifting from advertising to various forms of pricing and promotional discounts.
To obtain unbiased estimates of price elasticity, researchers should be careful about several aspects of the model building approach. Most importantly, time-series data eliminates potential biases from using only cross-sectional data. In addition, data at as disaggregate a level as possible helps capture the full dynamics of response to price changes and avoids biases from temporal aggregation. Attention to proper specification of the model is also essential for consistent estimates. In particular, the omission of quality or distribution from the model may bias the estimates.
The mean price elasticity obtained here over numerous past studies provides a reasonable estimate for managers unable to assess the sensitivity of their markets to price changes. The results can also be used as good benchmarks or null hypotheses against which to compare results from new models or studies specially focused on particular problems. But in doing so, researchers need to consider the systematic differences that arise because of environmental differences such as stages of the product life cycle, product categories, and countries.
Professor Tellis notes:
“The dramatic difference between price and advertising sensitivities should reassure critics of marketing activity who feel that marketers have undue influence over consumers’ minds. The results of the meta-analysis suggest that, on the average, consumers are reasonably informed and responsive to differences in prices, especially over time. However, caution about specific environments may be in order. For example, price sensitivity appears significantly lower during the early stages of the life cycle, as well as for pharmaceutical products. Closer scrutiny of pricing behavior in these markets may be productive.
“The overall high price sensitivity of selective demand, especially relative to advertising sensitivity, should not prompt managers into any drastic reallocation of expenditures from advertising to price discounting. Indeed, the danger of heavy price discounting is that it may decrease loyalty, while advertising may increase it. Both of these long-term effects have not yet been adequately captured in sales and market-share models of price and advertising.
“Another intriguing question is why markets are more responsive to changes in price than to changes in advertising intensity. An immediate answer would be that consumers derive more benefit from a price cut than they do from an increase in advertising. At a deeper level, the finding means that price cuts are immediately perceived as benefits, while advertising is not effective enough in attracting consumers’ attention or providing credible information or achieving attitude changes through repetition.
“The narrow distribution of the numerous estimates of price elasticity and their systematic variation as predicted by theory provide general support for this paradigm of research. The biases uncovered suggest directions for future research. For example, the use of scanner panel data and its analysis by choice models could avoid biases because of cross-sectional data or temporal aggregation; such an approach also provides researchers an opportunity to integrate behavioral theory in building models at the individual household level.”
About the Author
Gerard J. Tellis is Associate Professor of Marketing at the College of Business Administration at the University of Iowa
This resource is not available online.