The Effects of the Cost on the Opportunity and Probability of Purchase

The article discusses whether demographics can help identify consumers who are acceleration-prone in their purchase. Because of promotion, some consumers accelerate a purchase. Identifying such consumers is important for store and brand managers. Purchase acceleration means both the decrease of a household’s interpurchase times or the increase of the purchased quantity. Purchase acceleration, stockpiling and forward buying are basically the same. Theoretical studies have shown that purchase acceleration and stockpiling affect the profitability of promotions but such claims have not been systematically tested. The article discusses the types of household costs that affect accelerated behavior, select demographic variables linked to such costs, develop and test propositions about the relationship of demographic variables and purchase acceleration. The testing of these propositions is based on continuous-time individual-level models.

In the literature review section, the outcome of various other research papers is mentioned. An article written by Blattberg et al. (1978) indicates that differential consumer response to promotions is triggered by differential household costs. Households with low transaction, holding and unit costs are therefore expected to exhibit a more pronounced stockpiling behavior. Another article by Blattberg et al. (1981) conclude that purchase acceleration is the result of transferring holding costs from the retailer to the consumer. Retailers incur the cost of selling the product at a discount but are relieved from inventory costs. Assuncao and Meyer (1993) propose that lower household holding costs lead to stockpiling during promotional periods. (Rao, 1991) found that households with low search costs and costs of time are more prone to accelerate their purchases because of a promotion.

The article mentions that similar research uses the employment status of husband and wife and the presence of small children and to show household opportunity costs of time. The reason behind this assumption is that household without small children or unemployed spouse has less opportunity cost of time. Income is another variable because high income household has more space and less holding cost, but such household has more time constraints. Household size and home ownership variables is included to represent the holding cost. Car ownership is also included as a variable because a household will require extra time to search for low costs. The article concentrates on the household’s purchase timing decision described by an appropriate individual level model. The article uses use likelihood ratio tests to compare elasticities across demographic groups.

The article uses purchase data of 3,036 households who made at least one purchase during the entire observation period. The observation period of 132 weeks is divided into first 52 weeks for exploratory purposes and the last 80 weeks for model estimation. The article includes only the households that made at least five purchases during the first 52 weeks and at least ten purchases the rest of the period. That leaves only 844 qualified households. From the 844 households, a random subsample of 268 was selected to reduce computational burden. The article includes two groups of variables which are marketing variable and demographic variable. The marketing variable are price, feature and display. The demographic variables are annual household income, household size and female head employment status. Residence type, residence status and male head working status is ignored in the article because they do not vary considerably regarding these variables.

The writer noted that theoretical deliberation showed that households that differ in their cost structure also differ in their response to promotions. On the basis of the employment status of the female head of the family, the first hypothesis is “Households with the wife not working or employed part-time are expected to be more prone to accelerate a purchase because of a promotion than households with a female head full-time employed”. The second hypothesis is “Small households are more prone to accelerate their purchases because of a promotion than large households”. Because of the complexity of the effect of income on deal proneness, the third hypothesis is “Middle-income households are expected to accelerate their purchases because of a promotion more than low- or high-income households”. This is because both low-income household lack space and high-income households lack time. The study also examines the combined effects of demographic variables by examining effect of a particular variable while controlling for other factors.

The model used in the article is called Accelerated Failure Time model (AFT), which is a log-linear regression model capturing the effects of explanatory variables on purchase timing. This model was selected because the error component in this model is allowed to follow a distribution other than the standard normal distribution. In the article, the model is estimated separately for each demographic group and their combinations. The likelihood ratio test compares the log-likelihood obtained from estimating model with a set of coefficients common to all groups with the log-likelihood obtained from the estimation of separate sets of coefficients for each group.

When model is estimated separately for each of the four female head employment status groups, proposition 1 is supported. Wives that are not full-time employed are more price sensitive than wives that are full-time employed. The results in the article support proposition 2. Households of smaller size are more price and feature sensitive than larger size households for which the price feature and display coefficients are not significant. Proposition 3 is partially supported. The price and deal coefficients are not significant for the low-income group and are significant for the two middle income groups as predicted. However, the price variable is also significant for the high-income group, which is contrary to the expectations of the writer. Also, it is found that the overall direction of the female employment status effects is preserved even when adjusting for income effects. The effects persist even in the presence of household size effect. The results suggested that all three demographic variables have a significant effect on purchase acceleration.

In conclusion, the article summed up the effects of demographic variables on household purchase timing decisions. In particular, smaller households are more prone to accelerate their purchases due to deals than larger households. However, the impact of the household effect reduces when adjusting for income and female head employment status effects. The writer argued that the effect of household size has been overemphasized by some theoretical studies. Households with less than full-time employed wives tend to rely on prices for deciding whether to accelerate their purchases or not, while for households with a full-time employed wife the promotional signal is enough to trigger acceleration. The behavior of households with less than full-time employed wives suggests that such households evaluate promotions more carefully. Finally, the researcher recommends that identification of acceleration-prone customers should be used for segmentation purposes. The writer also indicated how to stop loyal customers from accelerating can be a topic of future research.