Costly search and consideration sets in storable goods markets

Abstract

Costly search can result in consumers restricting their attention to a subset of products–the consideration set–before making a final purchase decision. The search process is usually not observed, which creates econometric challenges. I show that inventory and the availability of different package sizes create new sources of variation to identify search costs in storable goods markets. To evaluate the importance of costly search in these markets, I estimate a dynamic choice model with search frictions using data on purchases of laundry detergent. My estimates show that consumers incur significant search costs, and ignoring costly search overestimates the own-price elasticity for products more often present in consideration sets and underestimates the elasticity of frequently excluded products. Firms employ marketing devices, such as product displays and advertising, to influence consideration sets. These devices have direct and strategic effects, which I explore using the estimates of the model. I find that using marketing devices to reduce a product’s search cost during a price promotion has modest effects on the overall category revenues, and decreases the revenues of some products.

This is a preview of subscription content, access via your institution.

Fig. 1

Notes

  1. 1.

    I refer to a consideration set as the optimal subset of products searched by a consumer and within which the consumer makes an explicit utility comparison before choosing the product to purchase.

  2. 2.

    The unit of observation in my data is a household and this is the relevant consumer in my application. I use “consumers” and “households” interchangeably.

  3. 3.

    See Baye, Morgan and Scholten (2007) for an overview of the literature on consumer costly search. See also Hong and Shum (2006), Moraga-Gonzalez, Sandor and Wildenbeest (2009), Gentry (2012), De los Santos, Hortaçsu and Wildenbeest (2012), Honka (2014), and Koulayev (2014).

  4. 4.

    See Blattberg and Neslin (1990) for a survey of the literature on demand accumulation.

  5. 5.

    Movement data for South Carolina were also used in some of the analysis performed in Section 3.

  6. 6.

    As pointed out by Seiler (2013), laundry detergent is a suitable product category choice because it is storable and purchased infrequently. Thus, consumer search behavior is likely to be important, and promotions are not expected to have a large effect on weekly consumption. Laundry detergent comes in two main forms: liquid and powder. Most of the quantity sold is from liquid detergent. For this reason I restrict the analysis to the liquid form.

  7. 7.

    All prices are in dollars per pound hereafter.

  8. 8.

    A potential complementarity between laundry detergent and other goods could also explain the positive correlation between the likelihood of buying detergent and the non-detergent expenditure. I, however, believe that this is not the explanation for the observed correlation, because there are very few complementarities between laundry detergent and the other goods that can be purchased during a shopping trip. Even if there was such complementarity, there is no reason for purchases of laundry detergent and of potential complementary goods to occur during the same shopping trip, because laundry detergent is a storable good.

  9. 9.

    Neither a standard discrete choice model nor a non-search based inventory model with independent random shocks to preferences and products that are perfect substitutes in consumption predicts an effect of inventory holdings on the choice of the brand to buy conditional on purchasing detergent. In a standard discrete choice model, preferences are not affected by inventory, so the choice of the product is always independent of inventory. If products are perfect substitutes in consumption and the random shocks to consumer choices are independent and identically distributed extreme value type I, the probability of choosing a brand conditional on size is also independent of inventory (see Hendel and Nevo 2006b, for a discussion). Pires (2013) discusses how this result can be extended to different assumptions.

  10. 10.

    The most frequently purchased brand is calculated over the whole time period of the data. I checked whether the most frequently purchased brand varies over time by evaluating the proportion of households with the same most frequently purchased brand in the first and in the last 3 years of the sample. In the overall sample, 83 percent of the households who made purchases in both halves of the sample have the same most frequently purchased brand in both halves. This suggests that the most frequently purchased brand does not change over time for most of the households.

  11. 11.

    I assume that consumers know their taste for a product and thus do not exhibit a learning behavior regarding product quality. In my empirical application the same products have been available for a long time and therefore I expect that this assumption is satisfied.

  12. 12.

    The cost of searching includes, among others, the time spent to find and collect information about a product, mental storage, cost of finding a product in the shelves, and processing costs (e.g., reading ingredients). Hence, the definition of search costs includes the cost of including a product at any given purchase occasion and an evaluation cost (see Hauser and Wernerfelt 1990).

  13. 13.

    See De los Santos, Hortacsu and Wildenbeest (2012) and Honka (2014) for a discussion of the sequential and simultaneous search processes.

  14. 14.

    Pires (2013) shows that (1) the prices of brands that are not the favorite brand affect the likelihood of purchasing the favorite brand even when the favorite brand is on sale, (2) the likelihood of buying the favorite brand falls with an exogenous and unexpected decrease in the prices of all brands when the magnitude of the price discount of brands that are not the favorite is larger, (3) the effect of the prices of the favorite brand on the incentives to continue searching is not significant, and (4) there is a negative correlation between price dispersion and search costs. All of these patterns are expected when consumers follow a simultaneous search strategy but usually cannot be explained by a sequential search strategy. The results in Table 4 provide additional evidence in favor of a simultaneous search process. In a sequential search model a consumer stops searching and buys a product when she finds a price below the reservation value. Under certain assumptions, low inventory may increase that reservation value, thereby reducing search and increasing the likelihood of choosing the favorite brand.

  15. 15.

    The prices and the nonprice attributes are store specific. The idiosyncratic tastes and the random shocks can also be store specific. I omit the store subscript from those variables to simplify the notation. Likewise, in the specification of search costs, the Display and Feature variables are store specific but the subscript for the store is omitted.

  16. 16.

    This assumption implies that the intention to purchase detergent does not cause consumers to go shopping. This assumption is supported by the evidence that consumer decision making usually occurs in store (Hoch and Deighton 1989; Dreze, Hoch and Purk 1994). Also, according to Seiler (2013), this assumption is reasonable because detergents are a small fraction of the total expenditure on the typical shopping trip.

  17. 17.

    Products with size greater than 0 and lower than 4lb were assigned to the small size, products with size greater than 4lb and lower than 8lb were assigned to the medium size, and products with size greater than 8lb were assigned to the large size.

  18. 18.

    This specification ensures that storage costs are always positive.

  19. 19.

    In my model product displays and feature ads affect only search costs. The role of those variables is to create salience and to inform consumers about the characteristics and prices of a product. It is usually assumed that displays and feature ads reduce search costs to zero because they give price information. Nevertheless, as pointed out by Mehta et al. (2003), some consumers do not observe displays and ads, so one must model each consumer as being exposed to the stimuli created by displays and ads. Hence, at the aggregate level, one would only expect the search costs for a product’s posted price to reduce by a certain fraction (and not to zero) in the presence of displays or feature ads.

  20. 20.

    Types are characterized by whether (1) income is above 50k; (2) family size is more than 2; (3) consumption rate is higher than 1.5 lb.; and (4) the favorite brand is Tide.

  21. 21.

    As pointed out by Koulayev (2014), the monetary value of search costs can be interpreted as a measure of the benefits of searching predicted by the model. So, a large value of the marginal search costs is a way of explaining low search activity in a model that predicts large benefits. My estimates of search costs therefore include the opportunity cost of not searching (i.e., the amount lost when the consumer chooses not to search).

  22. 22.

    In the Online Appendix, I evaluate the importance of the instruments that firms can employ to influence search behavior. To perform the analysis, I study households’ behavior and consequent effects on revenues if displays and feature ads were not available. The goal of this counterfactual is to understand the effects on revenues of using displays and feature ads. The results show that total revenues of laundry detergent are lower when displays and feature ads are not available. Product displays and feature ads are therefore important instruments because they encourage consumers to search. The effects are nevertheless heterogeneous. Although for most of the products the market shares and the revenues are lower without displays and feature ads, a situation where marketing devices cannot be used is desirable for some products because it improves their revenues by leveling the search costs of all products. In particular, if displays and feature ads were not available, products for which less is invested in those activities would be in a more advantageous situation when competing against other products to influence consumers to search.

  23. 23.

    For these estimates I assume that all products are never displayed or featured.

  24. 24.

    See the Online Appendix for the estimated search probability of each product.

  25. 25.

    The results should therefore be interpreted as the short-term effects to the new policy, as consumers are likely to quickly learn about the perfect coordination between price promotions and marketing activities, and update their beliefs according to that.

  26. 26.

    In this counterfactual scenario I do not calculate a new price equilibrium. Hence, this counterfactual exercise should be seen as a comparative statics exercise. Also, I do not propose a specific model for the supply side, which limits the scope of my analysis. In particular, I ignore that a change in search costs will likely lead to a reaction of competing stores, making the change less effective.

  27. 27.

    The specific counterfactual proposed by Seiler (2013) is to accompany a promotion with a reduction in search costs. Without loss of generality I assume the reduction in search costs is due to a decrease in displays and feature ads.

References

  1. Baye, M. R., Morgan, J., & Scholten, P. (2007). Information, search and price dispersion. Handbook of Economics and Information Systems, I.

  2. Blattberg, R., & Neslin, S. (1990). Sales promotions. Englewood Cliffs. New Jersey: Prentice Hall.

    Google Scholar 

  3. Boizot, C., Robin, J., & Visser, M. (2001). The demand for food products: An analysis of interpurchase times and purchased quantities. Economic Journal, 111, 391–419.

    Article  Google Scholar 

  4. Bronnenberg, B., & Vanhonacker, W. (1996). Limited choice sets, local price response, and implied measures of price competition. Journal of Marketing Research, 33, 163–173.

    Article  Google Scholar 

  5. De los Santos, B., Hortacsu, A., & Wildenbeest, M. (2012). Testing models of consumer search using data on web browsing and purchasing behavior. American Economic Review, 102, 2955–2980.

    Article  Google Scholar 

  6. Dinerstein, M., Einav, L., Levin, J., & Sundaresan, N. (2014). Consumer price search and platform design in internet commerce. Mimeo: Stanford University.

    Google Scholar 

  7. Draganska, M., & Klapper, D. (2011). Choice set heterogeneity and the role of advertising: an analysis with micro and macro data. Journal of Marketing Research, 48, 653–669.

    Article  Google Scholar 

  8. Dreze, X., Hoch, S., & Purk, M. (1994). Shelf management and space elasticity. Journal of Retailing, 70, 301–326.

    Article  Google Scholar 

  9. Dube, J., Hitsch, G., & Rossi, P. (2010). State dependence and alternative explanations for consumer inertia. RAND Journal of Economics, 41, 417–445.

    Article  Google Scholar 

  10. Eliaz, K., & Spiegler, R. (2011). Consideration sets and competitive marketing. Review of Economic Studies, 78, 235–262.

    Article  Google Scholar 

  11. Erdem, T., Imai, S., & Keane, M. (2003). Brand and quantity choice dynamics under price uncertainty. Quantitative Marketing and Economics, 1, 5–64.

    Article  Google Scholar 

  12. Gentry, M. (2012). Displays, sales and in-store search in retail markets. Mimeo: London School of Economics.

    Google Scholar 

  13. Goeree, M. (2008). Limited information and advertising in the U.S. personal computer industry. Econometrica, 76, 1017–1074.

    Article  Google Scholar 

  14. Hartmann, W., & Nair, H. (2010). Retail competition and the dynamics of demand for tied goods. Marketing Science, 29, 366–386.

    Article  Google Scholar 

  15. Hauser, J., & Wernerfelt, B. (1990). An evaluation cost model of consideration sets. Journal of Consumer Research, 16, 393–408.

    Article  Google Scholar 

  16. Hendel, I., & Nevo, A. (2006a). Sales and consumer inventory. The RAND Journal of Economics, 37, 543–561.

  17. Hendel, I., & Nevo, A. (2006b). Measuring the implications of sales and consumer inventory behavior. Econometrica, 74, 1637–1673.

  18. Hoch, S., & Deighton, J. (1989). Managing what consumers learn from experience. Journal of Marketing, 53, 1–20.

    Article  Google Scholar 

  19. Hong, H., & Shum, M. (2006). Using price distributions to estimate search costs. RAND Journal of Economics, 37, 257–275.

    Article  Google Scholar 

  20. Honka, E. (2014). Quantifying search and switching costs in the U.S. auto insurance industry. RAND Journal of Economics, 45, 847– 884.

    Article  Google Scholar 

  21. Kim, J., Albuquerque, P., & Bronnenberg, B. (2010). Online demand under limited consumer search. Marketing Science, 29, 1001–1023.

    Article  Google Scholar 

  22. Koulayev, S. (2014). Search for differentiated products: Identification and estimation. RAND Journal of Economics, 45, 553–575.

    Article  Google Scholar 

  23. Manski, C. (1977). The structure of random utility models. Theory and Decision, 8, 229–254.

    Article  Google Scholar 

  24. Mehta, N., Rajiv, S., & Srinivasan, K. (2003). Price uncertainty and consumer search: A structural model of consideration set formation. Marketing Science, 22, 58–84.

    Article  Google Scholar 

  25. Moraga-Gonzales, J., Sandor, Z., & Wildenbeest, M. (2009). Consumer search and prices in the automobile market. Mimeo: Kelley School of Business, Indiana University.

    Google Scholar 

  26. Pesendorfer, M. (2002). Retail sales: A study of pricing behavior in supermarkets. Journal of Business, 75, 33–66.

    Article  Google Scholar 

  27. Pires, T. (2013). Empirical studies of search frictions and consumer choices: Northwestern University Doctoral Dissertation, Department of Economics, Northwestern University.

  28. Pires, T. (2014). Measuring the effects of search costs on equilibrium prices and profits. Mimeo: University of North Carolina.

    Google Scholar 

  29. Roberts, J., & Lattin, J. (1991). Development and testing of a model of consideration set composition. Journal of Marketing Research, 28, 429–440.

    Article  Google Scholar 

  30. Rust, J. (1987). Optimal replacement of GMC bus engines: An empirical model of Harold Zurcher. Econometrica, 55, 999–1033.

    Article  Google Scholar 

  31. Seiler, S. (2013). The impact of search costs on consumer behavior: A dynamic approach. Quantitative Marketing and Economics, 11, 155–203.

    Article  Google Scholar 

  32. Stigler, G. (1961). The economics of information. Journal of Political Economy, 69, 213–225.

    Article  Google Scholar 

Download references

Acknowledgments

I am especially indebted and grateful to Aviv Nevo for his advice, guidance and support. I am also indebted and grateful to Igal Hendel and Robert Porter for valuable comments and discussions. I would like to thank Alberto Salvo, Brian McManus, Andre Trindade, Mike Abito, Guillermo Marshall, Agnieszka Roy, Arkadiusz Szydlowski, Sergio Urzua, Esteban Petruzello, Fernando Luco, Eric Anderson, Robin Lee, Song Yao, David Henriques, Tiago Botelho, Claudia Alves, Ernesto Freitas, MathisWagner, Mike Powell, David Miller, Jose Espin-Sanchez, Maja Kos, and seminar participants at Northwestern University, University of Toronto, Northeastern University, Tilburg University, Einaudi Institute for Economics and Finance, Banco de Portugal, University of North Carolina, Oklahoma University, Bates White, and Compass Lexecon for their suggestions. I am thankful to IRI and particularly Mike Kruger for generously supplying the data. Financial support from Fundaca̧o para a Cîencia e Tecnologia under the scholarship SFRH/BD/43857/2008 is also gratefully acknowledged. This paper is a revised chapter from my dissertation at Northwestern University. All errors are my responsibility.

Author information

Affiliations

Authors

Corresponding author

Correspondence to Tiago Pires.

Additional information

Tiago Pires died after this article was conditionally accepted at QME but before he could complete all final edits. Brian McManus and Javier Donna assisted in the preparations of the article for print. Please contact Brian McManus (mcmanusb@unc.edu) with questions about the article.

Electronic supplementary material

Below is the link to the electronic supplementary material.

(PDF 192 KB)

Rights and permissions

Reprints and Permissions

About this article

Verify currency and authenticity via CrossMark

Cite this article

Pires, T. Costly search and consideration sets in storable goods markets. Quant Mark Econ 14, 157–193 (2016). https://doi.org/10.1007/s11129-016-9169-2

Download citation

Keywords

  • Search costs
  • Consideration set
  • Information
  • Storable goods
  • Dynamic discrete-choice models

JEL Classification

  • D12
  • D83
  • L81