## Abstract

This paper investigates the dynamic effects of price promotions in a retail setting through the use of a large-scale field experiment varying the promotion depths of 170 products across 17 categories in 10 supermarkets of a major retailer in Chile. In the intervention phase of the experiment, treated customers were exposed to deep discounts (approximately 30%), whereas control customers were exposed to shallow discounts (approximately 10%). In the subsequent measurement phase, the promotion schedule held discount levels constant across groups. We find that treated customers were 22.4% more likely to buy promoted items than their control counterparts, despite facing the same promotional deals. Strikingly, the magnitude of the dynamic effects of price promotions (when promotional depths are equal across conditions) is 61% of the promotional effects induced by offering shallow vs. deep discounts during the intervention phase. The result is robust to other concurrent dynamic forces, including consumer stockpiling behavior and state dependence. We use the experimental variation and historical promotional activities to inform a demand-side model in which consumers search for deals, and a supply-side model in which firms compete for those consumers. We find that small manufacturers can benefit from heightened promotion sensitivity by using promotions to induce future consideration. However, when unit margins are high, heightened promotion sensitivity leads to fierce competition, making all firms worse off.

This is a preview of subscription content, log in to check access.

## Notes

- 1.
The marketing literature distinguishes between several price promotion instruments, including temporary price reductions (TPRs), coupons, promotion packs, rebates, among others. Our focus is on TPRs – the most frequently used type of price promotion (Gedenk et al. 2010).

- 2.
Other dimensions of the effects of promotions have also been studied. For instance, Kopalle et al. (1999), using a descriptive dynamic sales model, find that the high- and low-share brands tend to over and underpromote, respectively. Erdem et al. (2008) develop a structural model to show that, since price is an important signal of quality, frequent price cuts can have an adverse effect on brand equity, while Chan et al. (2008) use a structural model to show that future reactions to current price promotions depend on whether consumers are brand loyal or not. See Neslin and Van Heerde (2009) for an extensive review of promotion dynamics.

- 3.
Anderson and Simester (2004) suggest our focus as a possible avenue for future research: “We cannot say how customers would have responded to [..] a subsequent discount. Investigating these issues would require different studies in which the experimental manipulations were [..] repeated in a subsequent catalog.” In addition to solving the selection issue, our context and focus display other differences. In terms of the context, we work in different institutional settings (e.g., durable goods sold by catalog vs. non-durable goods sold in supermarkets), implying that the findings from one environment need not necessarily extend to the other. For example, the presence of immediate competitors in a limited amount of shelf space, or the limited amount of consumer learning in retail settings (Tuchman et al. (2017)), may lead to different forces and different results. Regarding the focus, we aim to analyze the implication of dynamic consumer behavior for the supply-side, as rationalized by consumers searching for deals, rather than focus on the extent to which different mechanisms may contribute to the main effect.

- 4.
- 5.
- 6.
While we performed the same experimental manipulation in both retail chains, we only report the results for the first chain, which is the larger one. The intervention did not produce statistically significant results for stores of the smaller chain. However, all results are directionally consistent across chains and dependent variables. We believe the smaller sizes of stores in the second chain are responsible for the absence of statistically significant effects. The results of the intervention in the smaller chain are available from the authors.

- 7.
We detail the criteria used for category selection in Appendix A.

- 8.
Sku, or

*stock keeping unit,*is a unique identifier of the product at the retailer. - 9.
Regular prices at treated and control stores are very similar across store pairs since the retailer sets regular prices based on pricing “zones.” The retailer defines pricing “zones” according to geographic locations, demographics and competition intensities. Each pair of treated/control stores always belongs to the same pricing zone.

- 10.
We were able to get buy-in for our manipulation during the first half of the experiment. However, the retailer viewed the initial promotion schedule as too heavy-handed to be implemented for the whole 10 weeks, and so an approximate alternative promotion schedule was agreed upon for the second half. The reasons for the changes were mainly due to previous agreements with manufacturers regarding sales targets and the planned promotion activity for our promoted products and their respective competitors. The changes were at a national level and were not related to local demand conditions. Moreover, they are constant across control/treated store pairs, as we further describe in Section 2.3, and so are unlikely to affect the validity of our estimates.

- 11.
Throughout the paper, we use the symbol ‘$’ to denote US Dollars, which were calculated by converting from the local Chilean Peso currency, based on the exchange rate of 0.0016 USD/CLP obtained from Google on May 27th, 2015.

- 12.
All of the results of interest survive the inclusion of the problematic categories, although the significance of the treatment effects decreases. Moreover, a few confounding results arise, related to the unexpected significance of a few variables with no apparent explainable pattern.

- 13.
Our baseline specification pools across all purchases made by a given consumer, hence the use of a single sub-index

*i*in Eq. 1. - 14.
See Cameron and Miller (2015) and Imbens and Kolesar (2015) for relevant discussions and practical guidance. Importantly, the wild bootstrap procedure is valid in the presence of right-hand indicator variables (see Cameron and Miller (2015), pp. 345-346). Previous versions of this manuscript employed bootstrap-t standard errors, which suffer from this limitation.

- 15.
Results on the heterogeneity of treatment effects are available from the authors. We find treatment effects are positively correlated with the number of promoted items bought in the past, controlling for average basket sizes. This finding is consistent with the idea that consumers who value promotions more, or equivalently, face lower costs when searching for deals, also show the most substantial increases in promotion sensitivities.

- 16.
We have simplified the comparison of coefficients by verifying whether the matched estimates fall within the confidence intervals of the unmatched ones. This method is used as an approximation, since the full analysis is complicated by the fact that the matched sample is necessarily correlated with the original one. Theoretically, without information about the joint distribution of the matched and unmatched coefficients, an implementation of non-nested tests à la Vuong (1989) is unavailable. While it is possible to use a bootstrap approach that performs the matching procedure on each bootstrap sample of the original dataset, the requirement of nesting the standard error correction implies a complex procedure as well as an impractical amount of computation time.

- 17.
There exists one limitation of this analysis, nonetheless, which is that the matching procedure was decided during the research program, and so even if the matching procedure was performed independently of empirical analyses, the matching sample represents an ‘additional draw’ made available to the researchers.

- 18.
- 19.
In other words, consumers can distinguish between products being sold at regular prices and in promotion on shelves immediately because of the seller’s different price tags. Given the lack of regular price variation, search informs consumers of idiosyncratic fit across all products, and also of promotional price levels for promoted products.

- 20.
It is possible that consumers can track promotion depths at the product level, especially for products they buy regularly. While promotion depths are fairly highly correlated within products of the same category at each point in time, we simplify the estimation by considering category-level beliefs. Note that we allow for beliefs at the individual product level in the counterfactual analysis. Finally, in the historical dataset, each category almost always has at least one product being promoted during each promotional period. We do not use the infrequent week-category combinations that do not feature any promotions in the estimation of the belief process.

- 21.
Index

*t*is used to denote a consumer visit, except if noted otherwise. - 22.
Because in our experiment regular prices were not manipulated, and moreover because we observe little variation in past regular prices, we assume that \(p_{jt}=p_{j}\). In the model, \(\alpha _{j}=\alpha _{j}^{\prime }-\beta _{0}p_{j}\). Also, we simplify notation \(d_{jt}^{Deep}\) and \(d_{jt}^{Shallow}\) to indicate the type of discount product

*j*was sold with to consumer*i*during her \(t^{th}\) purchasing occasion. - 23.
Equations 4 and 5 present standard expected-utility maximizers. In contrast with reference-dependent models, where the utility specification depends on assumptions made on how reference prices are formed (see the meta-analysis in Briesch et al. (1997)), our model builds on minimal assumptions about consumer beliefs about prices. Moreover, given our experimental results and the nature of beliefs informed by historical promotion activities, reference-dependent models would predict results opposite to our experimental findings, due to treated consumers incurring losses when faced with shallow discounts during the second half of the experiment.

- 24.
Note that all three methods are relatively efficient in calculating reservation values. Hence, the choice of method depends on factors such as the specific search context and methodological convenience.

- 25.
Our approach to recover consumer beliefs is consistent with most of the scanner panel literature. In contrast, imposing the Bayesian equilibrium concept would entail having consumers form beliefs based on firm-side fundamentals. Such equilibrium outcomes are much more challenging to investigate, as they require the researcher to 1) impose a rationale on the distributions of random variables affecting firms’ willingness to offer promotions, and 2) later check possible promotional deviations based on each set of potential consumer beliefs.

- 26.
Because our intervention is relatively short, it is unlikely that it affected consumer beliefs at a fundamental level. Rather, the experimental effect is captured by the transition matrix. The recovered historical promotion transitions are used as the beliefs held by consumers for estimating the model, as well as in the counterfactual analysis.

- 27.
We do not discount second-period payoffs, given the weekly timing of promotions.

- 28.
In our context, most promotions are initiated by manufacturers. We assume 100% passthrough in the analysis, although the effective rate is slightly lower. While we condition on different opportunity costs, in reality, they are likely to change over time, leading to different equilibrium promotion profiles in different weeks.

- 29.
Theoretically, firms could observe individual consumers’ buying behaviors, and hence have a precise idea of the realized state dependence in the market at the individual level. In reality, firms may not always take all consumer behavior into account, and thus have to integrate over the distribution of state dependence in order to optimize their promotional offerings. In our case, our homogeneous demand structure allows us to summarize state dependence effects through past market shares.

- 30.
Given the number of players and the size of the action space, finding mixed strategy equilibria can be tedious. Whenever a pure strategy equilibrium was not found, our estimation code prepared a file to be read by the Gambit software (www.gambit-project.org), which was then used to run all algorithms available to generate an exhaustive list of mixed strategy outcomes. These algorithms always found a unique mixed-strategy outcome.

- 31.
This matching procedure is standard when randomization is at the cluster level, but statistical analysis is at the individual level (Imbens 2011).

- 32.
Unlike our paper, the matching technique is typically used with non-experimental data in order to balance relevant pre-treatment covariates between treatment and control groups (Imbens and Rubin 2015).

- 33.
The assignment problem is a mixed-integer programming (MIP) problem where some of the decision variables are constrained to be integer values at the optimal solution.

- 34.
Note that, for the assignment problem, the labeling of treatment and control units is irrelevant for its optimal solution. We relabel some stores to increase the sample size of the matched sample.

- 35.
We thank an anonymous referee for this suggestion.

- 36.
See the Appendix in Kehoe and Midrigan (2015) for a detailed description of the method.

- 37.
Despite its merits, one limitation of this procedure is that, due to the data requirements involved, it does not provide a complete sequence of promotion depths for all products in all stores in all periods.

- 38.
We explored different values for the tolerance parameters \(\varepsilon _{j}\) for each covariate \(j\in \{1,..,P\}\), to account for the trade-off between the proximity measures of the paired customers and the resulting sample size. On the one hand, large values of \(\varepsilon _{j}\) lead to poorly matched pairs, while on the other, smaller values of \(\varepsilon _{j}\) reduce the sample size. In fact, some combinations of small values of \(\varepsilon _{j}\) imply no feasible solutions, i.e., no assignment meets the desired levels of balance on covariates. Importantly, no outcome analyses were performed during this stage.

- 39.
To be clear, the effect of state dependence can go either way: customers may also return in a week where their previously purchased product is no longer promoted, buy it (due to state dependence), and as a result, buy fewer (other) promoted goods during the second half of the experiment.

- 40.
It is worth clarifying the way the sample underlying these estimates was constructed. Following our choice of including in the analysis purchases (and non-purchases) by all consumers who visited a store and made a purchase in the first-half of the experiment, we included in the above estimation the purchase decisions of all individuals who made a purchase in the first-half of the experiment regardless of whether they bought a “new” product in promotion during the second-half. In the estimation reported above, purchases of the products previously purchased during the first half of the experiment were included as zeros in this analysis.

- 41.
After this manuscript was accepted, we learned about a result by Choi et al. (2018) that substantially simplifies the estimation of search-based demand models.

- 42.
See Train (2009) (Sec. 5.6.2) for a careful exposition of this issue.

- 43.
See Honka and Chintagunta (2017) for an application within the search framework.

## References

Anderson, E.T., & Simester, D.I. (2004). Long-Run Effects of promotion depth on new versus established customers: Three field studies.

*Marketing Science*,*23*(1), 4–20.Becker, G.S., Grossman, M., Murphy, K.M. (1991). Rational addiction and the effect of price on consumption.

*American Economic Review*,*81*(2), 237–241.Becker, G.S., & Murphy, K.M. (1988). A theory of rational addiction.

*Journal of Political Economy*,*96*(4), 675–700.Bertrand, M., Duflo, E., Mullainathan, S. (2004). How much should we trust Differences-In-Differences estimates?.

*The Quarterly Journal of Economics*,*119*(1), 249–275.Blattberg, R.C., Briesch, R., Fox, E.J. (1995). How promotions work.

*Marketing Science*,*14*(3), G122–G132.Bolton, R.N. (1989). The relationship between market characteristics and promotional price elasticities.

*Marketing Science*,*8*, 153–159.Boulding, W., Lee, E., Staelin, R. (1995). Mastering the mix: do advertising, promotion and sales force activities lead to differentiation.

*Journal of Marketing Research*,*31*(2), 159–172.Briesch, R.A., Krishnamurthi, L., Mazumdar, T., Raj, S.P. (1997). A comparative analysis of reference price models.

*Journal of Consumer Research*,*24*(2), 202–214.Bronnenberg, B.J., Kim, J.B., Mela, C.F. (2016). Zooming in on choice: How do consumers search for cameras online?.

*Marketing Science*,*35*(5), 693–712.Cabral, L.M., & Villas-Boas, M. (2005). Bertrand supertraps.

*Management Science*,*51*(4), 599–613.Cameron, A.C., Gelbach, J.B., Miller, D.L. (2008). Bootstrap-based improvements for inference with clustered errors.

*Review of Economics and Statistics*,*90*(3), 414–427.Cameron, A.C., & Miller, D.L. (2015). A practitioner’s guide to cluster-robust inference.

*Journal of Human Resources*,*50*, 317–372.Chan, T., Narasimhan, C., Zhang, Q. (2008). Decomposing promotional effects with a dynamic structural model of flexible consumption.

*Journal of Marketing Research*,*45*(4), 487–498.Choi, M., Dai, A.Y., Kim, K. (2018). Consumer search and price competition.

*Econometrica*,*86*(4), 1257–1281.Davis, S.J., Inman, J., McAlister, L. (1992). Promotion has a Negative Effect on Brand Evaluations-Or Does it? Additional Disconforming Evidence.

*Journal of Marketing Research*,*29*(1), 143–148.Dodson, J.A., Tybout, A.M., Sternthal, B. (1978). Impact of deals and deal retraction on brand switching.

*Journal of Marketing Research*,*15*(1), 71–81.Dube, J. -P., Hitsch, G.J., Rossi, P.E. (2010). State dependence and alternative explanations for consumer inertia.

*The RAND Journal of Economics*,*41*(3), 417–445.Ehrenberg, A.S.C., Hammond, K., Goodhardt, G.J. (1994). The After-Effects of Price-Related consumer promotions.

*Journal of Advertising Research*,*34*(4), 11–21.Erdem, T., Imai, S., Keane, M.P. (2003). Brand and quantity choice dynamics under price uncertainty.

*Quantitative Marketing and Economics*,*1*(1), 5–64.Erdem, T., Keane, M.P., Sun, B. (2008). A dynamic model of brand choice when price and advertising signal product quality.

*Marketing Science*,*27*(6), 1111–1125.Freimer, M., & Horsky, D. (2008). Try it, you will like it - Does consumer learning lead to competitive price promotions?.

*Marketing Science*,*27*(5), 796–810.Gedenk, K., & Neslin, S.A. (1999). The role of retail promotion in determining future brand loyalty: Its effect on purchase event feedback.

*Journal of Retailing*,*75*(4), 433–459.Gedenk, K., Neslin, S.A., Ailawadi, K.L. (2010).

*Sales Promotion in: Retailing in the 21st Century: Current and Future Trends*. Berlin: Springer-Verlag Eds.Guadagni, P.M., & Little, J.D. (1983). A logit model of brand choice calibrated on scanner data.

*Marketing Science*,*2*(3), 203–238.Hauser, J.R., & Wernerfelt, B. (1990). An evaluation cost model of consideration sets.

*Journal of Consumer Research*,*16*(4), 393–408.Honka, E. (2014). Quantifying search and switching costs in the US auto insurance industry.

*The RAND Journal of Economics*,*45*(4), 847–884.Honka, E., & Chintagunta, P. (2017). Simultaneous or Sequential? Search Strategies in the U.S. Auto Insurance Industry.

*Marketing Science*,*36*(1), 21–42.Imbens, G. (2011). Experimental design for unit and cluster randomized trials, 3ie, International Initiative for Impact Evaluation.

Imbens, G., & Kolesar, M. (2015). Robust Standard Errors in Small Samples: Some Practical Advice. Working Paper.

Imbens, G.W., & Rubin, D.B. (2015).

*Causal inference in statistics, social, and biomedical sciences*. Cambridge: Cambridge University Press.Jedidi, K., Mela, C.F., Gupta, S. (1999). Managing advertising and promotion for long-run profitability.

*Marketing Science*,*18*(1), 1–22.Jones, J.M., & Zufryden, F.S. (1980). Adding explanatory variables to a consumer purchase behavior model: an exploratory study.

*Journal of Marketing Research*,*17*(3), 47–53.Ke, T.T., Shen, Z. -J. M., Villas-Boas, J.M. (2016). Search for information on multiple products.

*Management Science*,*62*(12), 3576–3603.Kehoe, P., & Midrigan, V. (2015). Prices are sticky after all.

*Journal of Monetary Economics*,*75*, 35–53.Kim, J.B., Albuquerque, P., Bronnenberg, B.J. (2010). Online demand under limited consumer search.

*Marketing Science*,*29*(6), 1001–1023.Kopalle, P.K., Mela, C.F., Marsh, L. (1999). The dynamic effect of discounting on sales: Empirical analysis and normative pricing implications.

*Marketing Science*,*18*(3), 317–332.Lal, R., & Matutes, C. (1994). Retail pricing and advertising strategies.

*Journal of Business*,*67*(3), 345–370.McFadden, D. (1989). A method of simulated moments for estimation of discrete response models without numerical integration.

*Econometrica*,*57*(5), 995–1026.Mela, C.F., Gupta, S., Lehmann, D.R. (1997). The long-term impact of promotion and advertising on consumer brand choice.

*Journal of Marketing Research*,*34*(2), 248–261.Narasimhan, C., Neslin, S.A., Sen, S.K. (1996). Promotional elasticities and category characteristics.

*Journal of Marketing*,*60*(2), 17–30.Neslin, S.A., & Shoemaker, R.W. (1989). An alternative explanation for lower repeat rates after promotion purchases.

*Journal of Marketing Research*,*26*(2), 205–213.Neslin, S.A., & Van Heerde, H.J. (2009). Promotion dynamics.

*Foundations and Trends in Marketing*,*3*(4), 177–268.Pollak, R.A. (1970). Habit formation and dynamic demand functions.

*Journal of Political Economy*,*78*(4, Part 1), 745–763.Raju, J.S. (1992). The effect of price promotions on variability in product category sales.

*Marketing Science*,*11*(3), 207–220.Rubin, D.B. (1973). Matching to remove bias in observational studies.

*Biometrics*,*29*, 159–183.Rubin, D.B. (1979). Using multivariate matched sampling and regression adjustment to control bias in observational studies.

*Journal of the American Statistical Association*,*74*(366), 318–328.Scott, C.A. (1976). The effects of trial and incentives on repeat purchase behavior.

*Journal of Marketing Research*,*13*(1), 263–269.Seiler, S. (2013). The impact of search costs on consumer behavior: a dynamic approach.

*Quantitative Marketing and Economics*,*11*(2), 155–203.Seiler, S., & Pinna, F. (2017). Estimating search benefits from Path-Tracking data: Measurement and determinants, Marketing Science, Forthcoming.

Shoemaker, R.W., & Shoaf, R.F. (1977). Repeat rates of deal purchases.

*Journal of Advertising Research*,*17*(2), 47–53.Shugan, S.M. (1980). The cost of thinking.

*Journal of Consumer Research*,*7*(2), 99–111.Spinnewyn, F. (1981). Rational habit formation.

*European Economic Review*,*15*(1), 91–109.Train, K.E. (2009).

*Discrete choice methods with simulation*. Cambridge: Cambridge University Press.Tuchman, A., Nair, H., Gardete, P.M. (2017). Complementarities in consumption and the consumer demand for advertising, Working Paper.

Villas-Boas, S.B., & Villas-Boas, J.M. (2008). Learning, forgetting, and sales.

*Management Science*,*54*(11), 1951–1960.Vuong, Q.H. (1989). Likelihood ratio tests for model selection and non-nested hypotheses.

*Econometrica*,*57*(2), 307–333.Weitzman, M.L. (1979). Optimal search for the best alternative.

*Econometrica*,*47*(3), 641–654.Zenor, M.J., Bronnenberg, B.J., McAlister, L. (1998). The impact of marketing policy on promotional price elasticities and baseline sales.

*Journal of Retailing and Consumer Services*,*5*(1), 25–32.Zou, D. (2014). Intertemporal pricing of new products, Incentivizing Consumer Learning and Inertia, Working Paper.

Zubizarreta, J.R. (2012). Using mixed integer programming for matching in an observational study of kidney failure after surgery.

*Journal of the American Statistical Association*,*107*(500), 1360–1371.

## Author information

### Affiliations

### Corresponding author

## Additional information

We thank Eva Ascarza, Bryan Bollinger, JP Dubé, Wes Hartmann, Avery Haviv, Przemyslaw Jeziorski, Carl Mela, Ilya Morozov, Harikesh Nair, Duncan Simester, Miguel Villas-Boas, and the review team for helpful comments, as well participants in research seminars at Pontificia Universidad Católica de Chile, Santa Clara University, UC Berkeley, Universidad Diego Portales, and at the Marketing Science Conference, QME Conference, SICS, the UCSD Rady Field Experimentation Conference, and the Workshop in Consumer Analytics. The outstanding research assistance by Claudio Palominos and Breno Vieira is gratefully acknowledged. Noton acknowledges financial support from the Institute for Research in Market Imperfections and Public Policy, ICM IS130002.

## Appendices

### Appendix A: Criteria for category selection

We selected categories with the goal of providing the maximum amount of useful variation. First, we wanted to limit the influence of stockpiling behavior on the response to the promotion stimulus. If consumers respond to promotions by anticipating purchases, then severe post promotion dips could affect our estimates. On this basis, we excluded a few categories for which households’ inventory costs were deemed to be very low (e.g., soups) and others for which consumers could keep the product in inventory for a period of time, well beyond the post-promotion period (e.g., coffee). A second related consideration for including a category was the length of the typical interpurchase time observed in the category. In particular, we excluded those categories for which typical interpurchase times exceeded 5 weeks on average. Third, we only included categories that had already been promoted on a regular basis. Since our focus is on the effects of changes in promotion depth, we wanted to keep the frequency with which products were placed on promotion as constant as possible. This requirement led us to exclude categories such as “baked goods” which were rarely, if ever, placed on promotion. Fourth, we included categories that were purchased across different demographic segments (i.e., heterogeneous in terms of socioeconomic groups and ages). By imposing this requirement, we wanted to ensure that the same categories would be relevant across all stores included in the experimental design. Fifth, we chose categories in which consumers were unlikely to use the presence of promotion as an input in their assessment of a product’s quality. For instance, the presence of promotions in specific categories (e.g., fresh produce) can be interpreted as a negative quality signal, e.g., the product is about to expire or does not sell well, and the promotion is seen as an attempt to sell it rapidly. Sixth, we chose categories with different degrees of brand loyalty, e.g., soft drinks are well-known for having a few star brands with very loyal consumers, whereas milk exhibits more generic products, likely to be considered close substitutes by more consumers. Other considerations that played a role in our choice of categories were avoiding categories in which stockouts were known to occur more frequently and avoiding categories with a small number of brands.

### Appendix B: Robustness/alternative mechanisms

### B.1 Continuous treatment variable

One of the consequences of the experimental protocol not having been implemented by store managers exactly as designed is that different stores effectively received different treatment intensities (see Table 3). Therefore, it is useful to verify whether the same results hold when the empirical treatment, rather than the theoretical one, is introduced into the econometric specification. We consider two approaches. First, we calculate the average promotional product price for each store during the first half, across the experimental goods. Using this statistic as the treatment variable takes into account that consumers who shop at different stores may have been effectively exposed to different treatment intensities.

Table 16 summarizes the results (even-numbered columns) and compares them with the main regression results based on the treatment indicator variable (odd-numbered columns). We find that an *increase* in the discount depth (i.e., lower prices through promotions) during the intervention phase leads to a *higher* purchase likelihood during the second half of the experiment. Consumers are 10.4% more likely to buy a promoted item during the second half of the experiment for each dollar received through discounts during the first half. The same qualitative results are found in terms of the composition of consumers’ basket, in terms of the percentage of items and of expenditure.

In order to provide further robustness of our results, we also perform the same analysis at the category level. In this case, the regressor of interest is the average promotional price across experimental goods of each category. In this case, an observation is an individual/category combination. Because most individuals do not buy from most categories, this leads to many zeros on the left-hand side.

The results are presented in Table 17. All results remain statistically significant, and consistent with the original directions, although a slight loss in power may have been induced by the number of zeros introduced by this approach. Overall, our results are robust to using the empirical treatment intensities.

### B.2 Improving efficiency through matching

We complement our analysis with a matching procedure. Much like the use of control regressors, a matching method reduces the variance of the unobserved error term by taking advantage of the correlation between observable and unobservable characteristics. This approach shifts the emphasis from the cluster to the *individual* level by matching, within pairwise randomized stores, those individuals who have balanced covariates before the experimental period (Rubin 1973, 1979; Imbens and Rubin 2015).

In our case, the matching technique allows us to construct a large subsample of *statistical twins* (one twin buying at a control store and the other at a treated one) to ensure identical pre-treatment purchasing behavior between treatment and control *at the individual level*.^{Footnote 31} Notice that in our setting, the exogeneity of the treatment is guaranteed by the experimental design and matching is only needed to identify similar customers based on historical data.^{Footnote 32}

#### B.2.1 Matching framework

To construct our sample of statistical twins, we introduce a recent matching technique developed by Zubizarreta (2012). This matching technique takes advantage of new developments in optimization to match individuals in multiple aspects, which until recently was an unfeasible task due to the large dimensionality of the problem. Matching individuals on several dimensions encompasses other matching techniques, such as propensity score, by creating a superior and easily interpretable matching sample.

Formally, let \(\mathcal {T}=\{t_{1},\ldots ,t_{T}\}\) be the set of treated units, and \(\mathcal {C}=\{c_{1},\ldots ,c_{C}\}\), the set of potential controls. Without loss of generality, suppose \(T\leq C\). Each treated unit \(t\in \mathcal {T}\) has a *P* dimensional vector of observed covariates **x**_{t} = {*x*_{t,1},...,*x*_{t,P}}, and each control \(c\in \mathcal {C}\) has a similar vector \(\mathbf {x}_{c}=\{x_{c,1},...,x_{c,P}\}\). Let the assignment indicator \(a_{t,c}\) be equal to 1 if treated unit *t* is assigned to control *c*, and 0 otherwise; and denote the entire assignment matrix by **a**.^{Footnote 33} The optimal assignment problem is given by:

where \(\delta _{t,c}\in \left [0,\infty \right )\) is a distance function between treated and control units (e.g. Euclidean distance), \({\sum }_{t\in \mathcal {T}}{\sum }_{c\in \mathcal {C}}\frac {x_{c,j}a_{t,c}}{T}\) denotes the average covariate *j* of assigned controls and \(\overline {x}_{\mathcal {T},j}\) denotes the average covariate *j* across all treated individuals.

The goal of the matching program is to minimize the total sum of distances between treated units and matched controls as stated in expression (14). The first three constraints describe the integer nature of the assignment problem: Each treatment unit is paired with one control unit (Eq. 15) and not all control units should be used (Eq. 16).^{Footnote 34} The set of constraints given by expression (18) introduces an upper bound on the difference allowed between treatment and control individuals for each covariate, according to \(\varepsilon _{j}>0\), the pre-determined tolerance level for covariate \(j\in \{1,..,P\}\). This last set of constraints is a distinctive feature of the mixed-integer programming (MIP) matching approach proposed by Zubizarreta (2012).

#### B.2.2 Matching supermarket customers

We use the individual-level data available through the retailer’s loyalty card to match pairs of control and treated consumers based on the pre-experimental records according to the procedure described above. To construct statistical twins, we consider demographic and behavior-based covariates: Age, gender, the weekly average of total expenditure, the weekly average of total spending in the experimental categories, and the frequency of trips to the store. We then applied the criteria that customers are required to buy an item in at least one of the 31 main categories during the first half of the experiment and not visit multiple stores during the experimental period. The historical dataset used for this task covers a 46 week period, ending approximately one year before the experiment.

Consumers’ historical behavior toward promoted goods is a natural variable to include in this analysis.^{Footnote 35} In our case, the supermarket chain does not hold records that allow us to identify promotions during the pre-experimental period directly. We construct the promotional variable for the pre-experimental period using the algorithm proposed by Kehoe and Midrigan (2015). This algorithm identifies regular prices in a posted price time series as the modal price within a rolling window of five weeks centered on the current week.^{Footnote 36} We define a promotional event for a given price trajectory as a period when the posted price for a given UPC-store lies below the regular price (for the same UPC-store) derived from the Kehoe and Midrigan (2015) price filter.^{Footnote 37} We use the depth measurements to create two new variables: the share of promoted items bought by each consumer during the historical dataset, and the number of promoted items bought per consumer-visit during the same period.

Table 18 presents the sample sizes of the universe of customers before matching and those who were matched by the MIP matching procedure. Columns (1) and (2) present the number of customers who faced the experimental promotional activity in each store pair and for whom we have historical data. Overall, as shown in columns (3), (4) and (5), the MIP matching generated 13,482 one-to-one customer pairs (one control and one treated customer), distributed across 10 stores of our retail chain. Table 19 presents the resulting covariates of the final matched sample.^{Footnote 38} The last column reports the p-value for the null hypothesis of identical means. We obtain close averages in total expenditure, expenditure in promoted categories, and age between treatment and control matched individuals. Given the large sample size of individuals, the tests reject nearly same means in gender and number of trips, although the table shows that the actual values are quite similar. As for the purchase behavior metrics related to promoted goods (last two variables in in Table 19) added later per the review team’s suggestion, we do not reject that those variables have the same means across conditions. Moreover, the absolute differences for these variables are slim, if anything, favoring the control group towards being more active in purchasing promoted goods (thus making our results potentially conservative).

In order to further test the effects of the pre-experimental promotion-related variables, we include the historical share of promoted products bought as well as the historical mean number of promoted products purchased per visit in the main analysis as control variables. We report the results in Table 20. We find virtually no differences to the treatment effects (compare to Table 8). We interpret these findings, also in connection with those reported in Table 19, as the matching procedure having performed well in matching consumers overall, through the five dimensions mentioned above, such that the remaining dimensions had little effect on the treatment estimates.

### B.3 State dependence

To understand how state dependence could explain our results, consider a pair of similar customers who only differ on the experimental condition they were exposed to in the intervention phase. Assume that the treated customer bought a promoted product because of the deep discount, whereas the control customer decided not to purchase it, given its lower promotional discount (10%). It is possible that, during the second half of the experiment, both customers visited the store on the week that the same product was on promotion once again, at a shallow level. In this case, the treated customer may be more likely to buy the product in the second half, not because of heightened promotion sensitivity, but rather because of state dependence.^{Footnote 39}

We repeat the main analysis, but now only consider, for each consumer, purchases of goods that were not bought during the first half of the experiment. While this procedure is expected to mechanically decrease the treatment effect due to ignoring relevant data, it has the merit of parsing out the effect of state dependence. Table 21 summarizes the results for the full sample: the treatment effect in column (1) remains statistically insignificant, and all treatment effect estimates decrease slightly. However, statistic \(E\left [\left .\widehat {y}_{i}\right |X_{i},T_{i}= 1\right ]\div E\left [\left .\widehat {y}_{i}\right |X_{i},T_{i}= 0\right ]\) produces an estimate of a 21.2% relative increase of purchases of promoted products, which is similar to the one of 22.4% found before, when state dependence was not controlled for.^{Footnote 40}

Given the slight changes in significance and minimal changes to treatment effect estimates, we believe the new results are due to an overly-stringent test rather than state-dependence being responsible for the results. In order to investigate this issue further, we consider the same analysis on the matched sample, discussed in the previous section, and present the results in Table 22. All results of interest (columns (1)–(3)) remain statistically significant in this case, with the treatment estimates falling only slightly. Taken together, the results imply that state dependence may play a role in our measurement, but is unlikely to be responsible for the finding of heightened promotion sensitivity.

### B.4 Stockpiling behavior

Our results are likely to be made conservative by consumer stockpiling behaviors. The reason is that treated consumers are likely to hold higher inventories than control ones, due to the exposure to deep promotions during the intervention phase. As a result, stockpiling is expected to dampen treated consumer purchases during the second half of the experiment, including those of promoted products.

### B.5 Placebo test

In this section, we introduce a placebo test designed to assess whether our experimental intervention is likely to be effectively responsible for the differences in consumer behavior across treated and control pools. We focus the analysis on customers who did not visit the supermarket, or alternatively, did not use their loyalty cards during the first half of the experiment. Since these customers are less likely to have been exposed to the differential treatment conditions, we expect to find lower magnitudes and very few statistical significance of treatment effects (i.e., less than 5% of measured effects should be significant).

The results of the analysis are presented in Table 23 and are quite different from those in Table 6. First, only two behaviors are found to be marginally significant across different dependent variables and methods of calculating standard errors. Second, all point estimates fall below the original ones, often by an order of magnitude. If taken ‘as is’, the point estimate in column (1) would set a lower bound on the real treatment effect of 0.024, or a 10.7% increase \(\left (0.024\div E\left [\left .\widehat {y}_{i}^{full sample}\right |X_{i},T_{i}= 0\right ]\right )\) in relative terms.

These results are positive in the sense that consumers who are less likely to have been exposed to the intervention exhibit lower treatment effects, generally without statistical significance. The results are reassuring in terms of assessing experimental validity, but not without limitations. In particular, the choice of not purchasing from a major category during the first half of the experiment is unlikely to be exogenous. Because of this, the lack of a statistically significant treatment effect can source from selection, i.e., this test may sample from consumers who respond less to price promotions in the first place, and who exhibit lower promotion sensitivity effects as well. However, a countervailing force is the fact that it is impossible to rule out that some of these customers were exposed to our intervention but purchased without using their loyalty card during the first half of the experiment. This force would contaminate the results in the opposite direction, making the treatment effects in Table 23 conservative (in other words, the analysis did not focus on truly Placebo consumers).

### Appendix C: Search model

### C.1 Proposition: Logistic uncertainty

The logistic p.d.f. and c.d.f. with unit scale parameter are given by \(f\left (x\right )=\frac {e^{-\left (x-\mu \right )}}{\left (1+e^{-\left (x-\mu \right )}\right )^{2}}\) and \(F\left (x\right )=\frac {1}{1+e^{-\left (x-\mu \right )}}\) respectively. We seek the solution to equation

with respect to *z*. Plugging in the expressions above yields

Integration by parts yields

and the reservation value Eq. 19 becomes

and the solution is unique for

### C.2 Theorem: Contraction mapping

Let *u* be a random variable with continuous p.d.f. and c.d.f. \(f\left (\cdot \right )\), \(F\left (\cdot \right )\) respectively. The indifference condition is given by

Define \({\Gamma }\left (z\right )=-c+{\int }_{z}^{\infty }uf\left (u\right )du+zF\left (z\right )\). Under standard continuity assumptions, \({\Gamma }\left (z\right )\) is a contraction mapping if \({\Gamma }^{\prime }\left (z\right )\in \left [0,1\right )\). In our case,

which is bounded between zero and one. The use of the Leibniz integral rule implies integration must be interchangeable with differentiation. The contraction mapping applies for a large class of differentiable distributions, as long as \({\int }_{z_{n}}^{\infty }uf\left (u\right )du\) is finite , which is also implied by the original Weitzman (1979) model. So, the theorem applies to most distributions used in empirical work.

Using the proposition above, it is easy to show that in the case of mixture of logistics given by Eq. 8, the reservation value can be found through contraction

where \(\omega _{ijt}^{\kappa }\) is consumer *i*’s belief associated with finding a deep discount for a given promotional history \(\kappa \in \left \{ S,D\right \} \).

### C.3 Likelihood and estimation

We now characterize the likelihood of an alternative being chosen, which involves adding over search sequences. First, we rank the inside alternatives by their reservation values such that \(z_{1}>z_{2}>...>z_{n}\), where *n* is equal to the number of inside alternatives in the choice set. We depict the potential search paths consistent with a choice of alternative *j* in the diagram of Fig. 7, where searching an additional option corresponds to a lateral movement, and a downward one depicts the purchase of alternative *j*.^{Footnote 41} For a consumer to be willing to search option *j* with reservation value \(z_{j}\), she must have inspected options with higher reservation values before and have found that it was worthwhile searching option *j* nonetheless. The reason is that options are ordered by their reservation values, and so if a consumer did not search option \(j-1\) then she prefers not to search option *j* either. The sequence of events leading the consumer to arrive to node *j* is given by

The identity above can be shown by induction. If a consumer searched option 2 for example, then \(z_{2}>\max \left \{ u_{0},u_{1}\right \} \). This implies the consumer also searched option 1 because

since \(z_{2}<z_{1}\). For the consumer to prefer option *j* to the options searched before, we require \(u_{j}>\max \left \{ u_{0}..u_{j-1}\right \} \), and so a consumer searches alternative *j* and considers it the best option up to that stage if and only if

Conditional on searching option *j* and preferring it up to that stage, many subsequent search paths can lead to a final choice of *j*. For example, the consumer may choose alternative *j* without searching any further, or do so after searching option \(j + 1\), options \(j + 1\) and \(j + 2\), etc. Let \(Buy_{j\left |k\right .}\) be each of such *subsequent* paths, where *j* is the chosen product and \(k\geq j\) is the last product searched by the consumer. Then, the probability of choosing option *j*, which informs our likelihood function, is equal to

We now characterize each of the paths, where movements referred to as ‘down’ and ‘right’ are related to the ones in the Fig. 7:

We have characterized the likelihood function. It remains to maximize it with respect to parameters, conditional on the data. Because utilities are probabilistic, we use simulation to generate *u*’s and construct the likelihood. Moreover, the need to investigate multiple search paths led us to employ 10,000 draws per choice-alternative.

In order to account for heterogeneity in search sequences, we add a noise parameter \(\eta \sim N\left (0,1\right )\) to the reservation values. For example, in some circumstances, consumers may not include some products in their consideration sets, which is equivalent to those products featuring very low reservation values. This assumption also provides the demand function with smoothness for purposes of the counterfactual analysis.

An additional difficulty with ‘accept/reject choice simulation’ is that small changes in parameters do not affect simulated outcomes, even for large sets of draws.^{Footnote 42} Moreover, the log-likelihood function exhibits saddle points that make finding the global maximum challenging.

We implement a patterned grid search across a wide range of parameter values, and ensure that the bounds set for the parameters were never achieved during the estimation procedure. Calculation of the standard errors required additional smoothing. For this purpose, following McFadden (1989), we smoothed out the likelihood function by use of a kernel function, which in our case is analogous to adding a low-variance extreme-value noise to each *u* and *z* component.^{Footnote 43} For illustration purposes, suppose we observe option \(n-1\) being chosen. The probability of this choice is

Given a parameter guess, we generate *R* sets of simulations of *u*’s. For each set *r*, we calculate

where

is the logistic kernel with smoothing parameter \(\sigma = 0.001\). We used the smoothing parameter to calculate standard errors. During estimation, we used \(K\left (x\right )= 1\left (x>0\right )\) instead, because the grid search algorithm does not require smoothing out the objective function.

Finally, we average across simulation results to calculate the choice probability, i.e.

McFadden (1989) characterizes the estimator above as well as its consistency in detail.

## Rights and permissions

## About this article

### Cite this article

Elberg, A., Gardete, P.M., Macera, R. *et al.* Dynamic effects of price promotions: field evidence, consumer search, and supply-side implications.
*Quant Mark Econ* **17, **1–58 (2019). https://doi.org/10.1007/s11129-018-9205-5

Received:

Accepted:

Published:

Issue Date:

### Keywords

- Price promotions
- Dynamic effects
- Consumer search
- Bertrand supertraps
- Field experiments