Consumers tend to be loyal to products they have purchased in the past, which can confer market power on firms able to obtain higher prices from their captive customers (Dubé et al. 2010). To understand how customer habits change and how they shape firm prices and consumer well-being, it is important to understand how easily habits can be disrupted and behaviors changed. Recently, stores have been experiencing “out of stock” due to COVID-related purchases in categories such as ready meals and toilet paper. Supply chain disruptions have also led to shortages of some products, forcing consumers to switch to different brands than what they usually buy.
Two factors can cause customers to buy the same product repeatedly: preferences and state dependence (Heckman 1981). The explanation of the first is intuitive: customers have different preferences and always buy the products they like. Alternatively, product loyalty may be driven by state dependence. This explanation suggests that customers’ purchase histories affect their subsequent choices, i.e., a customer is more likely to purchase a product because they purchased it in the past. While both of these factors likely affect product loyalty to some extent, understanding their relative magnitude is important for business pricing and consumer welfare. If the effect of government dependency is very strong, this suggests that firms could capture customers with temporary price discounts and rely on their loyalty even after raising prices. Alternatively, if unobserved preferences primarily explain brand loyalty, a temporary price discount would have less long-term effects, as customers would continue to buy their preferred product after the price reinstatement.
A natural experience
In a recent paper (Levine and Seiler 2021), we develop a state dependency presence test with a new natural experiment. The ideal experiment would be to randomly assign a group of customers to switch from their usual brand (brand A) to another brand of equal quality (brand B). We would then observe their subsequent choices. A complete return to brand A would suggest that there is no state dependency effect and that persistent choices of brand A exist because people prefer that brand. A complete conversion to brand B would suggest that there is no effect of unobserved preferences and that brand loyalty exists simply because of state dependence.
Instead of such an experiment, we use natural stock-outs that cause some households to drop their usual brand, allowing us to study the dynamics following this random variation of choice sets. We identify areas and times when stockouts are likely to occur due to hurricane preparations using a publicly available hurricane tracking database maintained by the US National Hurricane Center. We use the Nielsen Retail Scanner dataset to select counties close enough to the hurricane that also experienced spikes in demand for staples (canned soup, batteries, bottled water) just before the hurricane. Using the Nielsen HomeScan panel, which documents customer purchases over time, we identify all households that live in these counties as having been at risk of experiencing a stockout. We compare the buying habits of these households (the treated group) to a control group of households that live far from the path of the hurricane and therefore did not experience a random shock in their choice sets.
We focus our analysis on bottled water, something many people stockpile in anticipation of a hurricane. The upper panel in Figure 1 shows the average weekly spending by week against the hurricane separately for the treatment and control groups. The gray region covers weeks 0 and 1, with week 0 representing the week before the hurricane. We observe an increase in spending on bottled water for treated households in the weeks surrounding the hurricane, validating our selection criteria: treated households increased demand for a hurricane staple in the weeks leading up to the hurricane. hurricane.1
This increase in demand increases the likelihood of stock-outs of individual brands, which is the variation we exploit for our analysis. The bottom panel of Figure 1 shows evidence of this, plotting the share of “new” brands over time, on average across treatment and control group households. This measure divides the number of brands purchased that were new to a household by the number of unique brands purchased.2 The increase in the share of new brands for the treated group in weeks 0 and 1 suggests that increased demand has resulted in a higher likelihood of a brand being out of stock, causing some households to to switch to a new brand.
Figure 1 Spending on bottled water and buying new brands
Methodology and construction of variables
By construction, the treated and control households live in different geographical regions; the treated households live near a hurricane and the control households do not. As a result, we likely find that there are temporal differences in purchasing behavior between groups. We therefore construct a synthetic control group, following Xu (2017). This method uses control group data to estimate time-varying factors that are common to all households. It then uses the pre-hurricane data to estimate how these factors should be weighted for each treated household, to minimize prediction errors. Finally, we construct a synthetic control household for each treated household, which allows us to compare the loyalty of the treated households to what their loyalty would have been without the hurricane.
Our analysis focuses on a variable that describes brand (and UPC) loyalty. This variable is the share of unique brands (or UPCs) purchased in the bottled water category that were purchased during the last purchase occasion. A value of 1 (0.5) means that all (half) of the brands purchased were purchased on the customer’s last purchase occasion. On the first trip after the storm, the variable would measure brand loyalty that customers purchased during the hurricane. Even in the presence of state dependency, this value would be lower for treated customers than for control customers because some of them are doomed to return to their preferred brand. Therefore, we make a modification to this variable for ease of interpretation: on the first trip after the hurricane, this variable represents the share of unique brands (or UPCs) purchased that were purchased on the last purchase occasion of the customer before the hurricane. If this modified measure, hereinafter referred to as “fidelity”, is not significantly different between the treated households and the control households after the stock-out, this means that the treated households are also faithful to their choices before the hurricane than control households, despite the shock suffered by their choice of hurricane. This would suggest that there is no substantial effect of state dependency. However, if this variable is lower for the treated households than for the control households, it would suggest that some households have changed their loyalty to their hurricane choices.
Figure 2 shows fidelity over time, averaged between households within the treated and synthetic control group. We find a significant decrease in brand (and UPC) loyalty for households treated during the stock-out (weeks 0 and 1). This mirrors the increase in the share of the new brand shown in the bottom panel of Figure 1. Importantly, there is no difference between the treated and control households in loyalty following breaks in stock. On the contrary, treated households seem to immediately revert to their pre-hurricane choices.
Figure 2 Average fidelity of treated households and synthetic controls
We test this more formally by estimating average treatment effects (ATE) over time. The ATE is the average difference between the treated households and the synthetic control households at a given time. The synthetic control method allows us to generate standard errors, using bootstrap draws. We find that although there is a large and significant decrease in loyalty during the out-of-stock weeks, there is no effect in the post-hurricane period.
Contrary to previous research (Dubé et al. 2010, Simonov et al. 2020), we find no evidence of state dependency in the choices of bottled water brands or UPCs. One possible explanation for this is that our natural experience forces consumers to choose the last remaining options when out of stock, which may be lower quality or more expensive than their usual choices. However, we see no change in product popularity or prices during the hurricane. Regardless, we estimate treatment effects for several subsets of treated households: those who purchased more/less popular products and those who purchased more/less expensive products. We find that there is no long-term effect of stockout for any of these subgroups, suggesting that the null effect is not due to unusual switching behavior.
Another explanation for our contrary findings is that we are studying the bottled water category, where customers may have less defined preferences. However, we find that bottled water purchases exhibit similar levels of loyalty to other commonly studied GIC categories, such as margarine and orange juice.
Authors’ note: The researchers’ own analyzes were calculated (or derived) in part based on data from Nielsen Consumer LLC and marketing databases provided by NielsenIQ datasets at the Kilts Center for Marketing Data Center at the University of Chicago Booth School of Business. Conclusions drawn from NielsenIQ data are those of the researchers and do not reflect the views of NielsenIQ. NielsenIQ is not responsible, played no role, and was not involved in the analysis and preparation of the results reported here.
Dubé, JP, GJ Hitsch and PE Rossi (2010), “State Dependence and Alternative Explanations for Consumer Inertia”, The RAND Journal of Economics 41(3): 417–445.
Heckman, JJ (1981), “Heterogeneity and State Dependence”, in S Rose (ed.), Labor market studiess, University of Chicago Press.
Levine, J and S Seiler (2021), “Identifying State Dependence in Brand Choice: Evidence from Hurricanes», SSRN Electronic Journal.
Simonov, A, JP Dubé, G Hitsch and P Rossi (2020), “State-dependent demand estimation with correction of initial conditions”, Marketing Research Journal 57(5): 789–809.
Xu, Y (2017), “Generalized Synthetic Control Method: Causal Inference with Interactive Fixed-Effect Models”, Political analysis 25(1): 57–76.
1 There is seasonality in spending on bottled water, likely due to seasonal demand. Week 0 usually falls between June and November, so the decrease in spending after week 0 coincides with colder temperatures.
2 Brands are defined as “new” if they have not been purchased by the household during a six-month period preceding the main sample.