An examination of stockpiling behavior in response to price deals.
There is evidence that price deals can cause dramatic short-term increases in the sales of consumer nondurables (e.g., Guadagni and Little, 1983; Blattberg et al., 1981). One potential cause of the spike in sales during promotional periods is stockpiling. That is, a price deal may cause consumers to buy larger than normal quantities of a product.
The phenomenon of stockpiling is important to both retailers and manufacturers. One concern is that consumers, by purchasing larger quantities, may simply be loading up on a brand that they would have bought at its regular price. However, stockpiling may also be beneficial if it leads to an increase in consumers' consumption rates or if it prevents or delays consumers from switching to competitors' brands.
A number of studies, using a number of different types of products and promotions, have examined the phenomenon of stockpiling (e.g., Gupta, 1988, 1991; Schneider and Currim, 1990; Krishna and Shoemaker, 1992). Most studies have found that stockpiling occurs, however, frequently the evidence has been quite weak (e.g. Neslin et al., 1985; Gupta, 1988, 1991). Moreover, some studies, particularly those using store-level sales data, have not found the effect of stockpiling on store sales to be large enough to be noticeable (e.g., Walters and Rinne, 1986; Moriarty, 1985). Apart from the finding that stockpiling exists, very little can be said with confidence about the phenomenon.
The purpose of this study is to increase our understanding of the phenomenon of stockpiling. Specifically, using household-level grocery purchase data, the study examines the impact of household characteristics, product-related factors, and purchase environment characteristics on the incidence of stockpiling in four product categories. The study uses a rather strict definition of stockpiling--a purchase quantity on a deal purchase occasion that is greater than the largest quantity purchased by the household on any non-deal purchase occasion during a 70-week period.
The cost-benefit framework of the household inventory model (Blattberg et al., 1978, 1981; Jeuland and Narasimhan, 1985, Krishna, 1994) provides a theoretical foundation that can be used to link household characteristics and other factors to stockpiling behavior. The household inventory model proposes that households seek to meet their demand for products while at the same time minimizing their total costs. These costs which households seek to minimize might include ordering costs, holding/storage costs, stock-out costs, promotion search costs (e.g., the costs of scanning newspapers for coupons and advertised price specials, cutting and sorting coupons etc.), brand substitution costs (i.e., the costs of purchasing less-preferred brands), and future ordering costs (Bawa and Shoemaker, 1987; Schneider and Currim, 1990; Krishna et al., 1991).
The conceptual framework of the inventory model can be used to derive hypotheses regarding the impact of household, product-related, and situational factors on the incidence of stockpiling. The implied causal structure is that these household, product-related and situational factors affect one or more components of a household's purchasing costs, which in turn affect the purchase quantity decision (see Figure 1 for a summary of the hypothesized factors).
Household Characteristics: Deal Proneness, Brand Loyalty, and Consumption Rate
The incidence of stockpiling may be affected by the extent to which households search for deals (Schneider and Currim, 1990). Households that search for deals incur search costs (i.e., the cost of time spent scanning advertisements and in-store displays). In order to help allay these costs, these households may be more likely to want to receive the financial and perhaps psychological benefits derived from stockpiling a product.
Also, consumers incur substitution costs when they purchase less-preferred brands when they are on sale. Consumers who are more loyal to a particular brand may be more likely to purchase more of that brand, when given a price deal, because they would incur no substitution costs.
A third household characteristic that may affect stockpiling behavior is a household's product consumption rate. The lower the consumption rate of a household in a product category the slower inventory levels will dwindle, resulting in higher holding costs (i.e., the amount of space occupied and the amount of capital tied up in inventory). Thus, there is expected to be a positive relationship between household product consumption rate and the incidence of stockpiling. Low consumption rate households should be less likely to accelerate their purchases because of the holding costs that would be incurred. Thus:
H1: Deal prone consumers are more likely than other consumers to stockpile.
H2: In response to a price deal for their preferred brand, brand loyal consumers are more likely than other consumers to stockpile.
H3: There will be a positive relationship between household product consumption rate and the incidence of stockpiling.
Product-Related Characteristics: Deal Frequency and Product Price
Households are expected to be more likely to stockpile when a brand is infrequently promoted than when it is frequently promoted. When deal frequency in a product category is perceived to be high, consumers may be less likely to stockpile because of the feeling that another deal will soon be available (Helsen and Schmittlein, 1992; Krishna et al., 1991). There is less incentive to incur the costs associated with stockpiling (e.g., holding costs) when future ordering costs are also expected to be low.
A second product-related characteristic that may affect stockpiling behavior is product price. A number of studies have examined the impact of a product's regular price on purchase behavior (e.g., Guadagni and Little, 1983; Gupta, 1988). There is expected to be a negative relationship between product price level and the incidence of stockpiling because price obviously affects holding, ordering and brand substitution costs. Thus:
H4: In response to a price deal, households are more likely to stockpile when the brand is infrequently promoted than when it is frequently promoted.
H5: There will be a negative relationship between product price and the incidence of stockpiling.
Situational Factors: Level of Inventory and Depth of Price Cut
One situational factor that may affect the incidence of stockpiling is a household's current inventory level of the product that is being promoted. Studies have found that a household's level of product in inventory affects purchase quantity decisions (e.g., Neslin et al., 1985; Gupta, 1988). Consumers with a large inventory at home tend to wait before purchasing and, if a purchase is made, purchase smaller quantities. The level of current inventory affects a household's holding costs, therefore, level of inventory is expected to have a negative effect on the incidence of stockpiling.
A second situational factor that may affect the incidence of stockpiling is the depth of the price cut for a promoted product. Previous research has generally found a positive relationship between the depth of any price cut and the response of consumers to it (e.g., Blattberg et al., 1981; Moriarty, 1983). Thus, larger price cuts should lead to more stockpiling (Helsen and Schmittlein, 1992). Thus:
H6: There will be a negative relationship between level of household inventory and the incidence of stockpiling.
H7: There will be a positive relationship between the depth of a price cut and the incidence of stockpiling.
The scanner data consist of a 70 week record of grocery purchases made by households in a store of a national supermarket chain. Purchase records were available for all households that were members of the grocery store's frequent shopper's club. These purchase records were combined with price and promotional activity information. Four product categories were selected for analysis -ground caffeinated coffee, paper towels, toilet paper, and canned tuna. All of these product categories have been examined in prior consumer promotion studies.
To insure that complete purchase histories of households were obtained, only households that met a number of selection criteria were included in the study1. The number of households in the study ranges from 395 in the coffee category to 756 in the toilet paper category (see Table 1).
The number of purchases in the four categories ranges from 3,462 (coffee) to 7,719 (toilet paper) (see Table 1). In each of the four product categories, more than half of all purchases are deal purchases.
Determining and Modeling the Incidence of Stockpiling
A logit model is used to examine the effect of household characteristics and other factors on stockpiling behavior. The logit model has been discussed and used extensively to study the effect of promotions on purchase behavior (for example, see Blattberg and Neslin (1990) or Guadagni and Little (1983) for details on the use of logit models).
Previous studies have usually examined stockpiling by comparing purchase quantities on deal and off-deal purchase occasions or by using regression or some similar analysis in which purchase quantity is dependent on a deal status variable (i.e., deal/off-deal). Rather than modeling purchase quantity as most studies have done (Neslin et al., 1985; Krishanmurthi and Raj, 1988, 1991), this study specifically models stockpiling. In this study, stockpiling is said to have occurred if the purchase quantity on a deal purchase occasion is greater than the largest quantity purchased by the household on any non-deal purchase occasion during the 70-week period. Purchase quantity is measured as the amount, in a relevant unit of measurement (e.g., ounces), of a particular brand that a household purchases on a purchase occasion.
Thus, stockpiling is operationalized as a binary variable--given the value one if the quantity purchased on deal is greater than the quantity purchased on any off-deal purchase occasion, or given the value zero otherwise. In the logit models (one for each product), stockpiling is a binary (0/1) dependent variable and the independent variables are the household characteristics, product-related factors and situational factors. The logit models are estimated using the maximum likelihood approach. The first 30 weeks of purchase data are used to initialize the brand loyalty and inventory level variables for each of the household (see "Independent Variables in the Stockpiling Model") and the next 40 weeks of data are used to estimate the logit models in the four product categories.
Independent Variables in the Stockpiling Model
The study uses an exponentially weighted average of a household's past purchases of a brand as a brand loyalty measure (see Gupta, 1988; Papatla, 1995; Papatla and Krishnamurthi, 1996). The measure allows for differences in loyalties across households and changes in loyalties within a household over time. The brand loyalty of a particular household i with respect to the purchased brand j at purchase occasion n is as follows:
BL(i,j,n) = BL(i,j,n-1)*a + (1--a) If household i purchased brand j at n-1 = BL(i,j,n-1)*a [Otherwise.sup.2]
The deal proneness measure, developed by Webster (1965), is obtained using all the purchases of a household during the 70 week period. It is as follows:
[Ds.sub.i] = [Sum.sub.j] [[PPA.sub.ij]/[PPB.sub.j]] * [PPC.sub.ij] where [PPA.sub.ij] = Percentage of household i's purchases of brand j made on deal [PPB.sub.j] = Percentage of all purchases of brand j (all households) that are made on deal [PPC.sub.ij] = Percentage of household i's purchases in the product category in which brand j was purchased
Deal frequency is measured as the number of weeks during the 70-week period that the brand-size was on deal. Both advertised and unadvertised price cuts were considered to be deals.
The depth of any price discount is defined as the regular price minus the shelf price (in cents/ounce or cents/sq.ft.). The price of a brand is determined to be discounted if the brand is featured in a weekly newspaper insert or if a price reduction is greater than 30 cents or 15 percent of the price in the prior week.
The estimate of the level of product category inventory for a household in a specific week is obtained from the measurement procedure employed by Gupta (1988, 1991) and Krishnamurthi et al. (1992):
Estimated Estimated Estimated Quantity Bought Inventory = Inventory -- Weekly + in Week t-1 in Week t in Week t-1 Consumption (if any) (beginning) (beginning)
Average weekly purchase quantity during the 70-week period is used as a surrogate for weekly product consumption rate, and the initial level of inventory at the first purchase occasion is set to zero (Bucklin and Lattin, 1991).
Table 2 presents parameter estimates and measures of fit for the four stockpiling logit models. As seen in Table 2, the log-likelihood chi-square statistic, not surprisingly, is significant in all four of the stockpiling models. That is, the null hypothesis that all parameter estimates are zero is rejected in all of the models.
However, the predictive ability of the models is rather low. The likelihood ratio index (p2), a "pseudo-[r.sup.2]" measure of the proportion of the log-likelihood explained by the model, ranges from .1 to .16. Although the values are low, the levels are consistent with the results from previous efforts modeling purchase quantities (e.g., Neslin et al., 1985; Gupta, 1988, 1991). The results indicate that a large proportion of the variability in purchase quantity is not explained by the models. Results relating to specific hypotheses are as follows:
Among the three household characteristics variables, deal proneness is most strongly and consistently related to the incidence of stockpiling. As hypothesized (H1), in all four product categories, there is a significant positive relationship (p < .01) between household deal proneness and the incidence of stockpiling.
Surprisingly, brand loyalty is not significantly related to stockpiling behavior in any of the four product categories. Thus, hypothesis 2 is not supported. The absence of any relationship between brand loyalty and the incidence of stockpiling may be due to the fact that deal purchase occasions were found to be dominated by households that make the majority of their purchases on-deal and have low levels of brand loyalty. It might also be the case that brand preferences in the four product categories are not as strong or important as size preferences or product-type preferences (e.g., light vs. chunk white tuna).
Hypothesis 3 is also not strongly supported because a significant negative relationship (p < .05) between household consumption rate and the incidence of stockpiling is found in three of the four of the product categories. The hypothesis is supported (p < .01) only in the canned tuna product category. The negative relationship between consumption rate and stockpiling in three of the four product categories may have occurred because heavy users are already buying large quantities of product and they may find it difficult to respond to promotions by buying larger quantities on deal purchase occasions. Heavy users of canned tuna, compared to heavy users in the other categories, would appear to be more easily able to increase their purchase quantities on deal purchase occasions because the additional quantity can be more quickly consumed and more easily transported and stocked.
Of the two product-related variables, product price is more strongly related to the incidence of stockpiling. There is a significant negative relationship (p < .01) between the regular price of a product and the incidence of stockpiling in three of the four product categories (toilet paper is the exception). Thus, H5 is generally supported.
The deal frequency of a brand is found to have a significant negative effect (p < .05) on stockpiling in two of the four product categories (paper towels and toilet paper), and a significant positive effect on stockpiling in the canned tuna category. Thus, H4 is supported in only two of the four product categories.
The absence of a stronger relationship between deal frequency and purchase acceleration may be due to characteristics of the promotional environment. If many households purchasing on deal have no strong preferences toward any one brand, there is less motivation to monitor or develop perceptions about the deal frequency of any one brand. This should be particularly true when there are frequent promotions in the product category.
Among the situation-related factors, both level of inventory and depth of price cut are consistently related to the incidence of stockpiling. In three of the four categories, household inventory level has a significant negative effective (p < .01) on the incidence of stockpiling, thus supporting H6. In all four of the product categories, there is a significant positive relationship (p < .05) between the depth of the price cut on deal purchase occasions and stockpiling behavior. Thus, H7 is supported.
Using the cost-benefit framework of the household inventory model by linking certain household, product-related, and purchase-situation characteristics to the components of a household's total costs, this study derived and tested hypotheses regarding the incidence of stockpiling.
Deal proneness is consistently found to be positively related to the incidence of stockpiling. Consumers who search for promotions incur costs (i.e., the cost of time spent scanning advertisements and in-store displays). To help allay these costs, they are more likely to purchase more of promoted products in order to receive the additional financial and perhaps psychological benefits.
In the four product categories, brand loyalty is not related to the incidence of stockpiling. Deal purchase occasions seem to be dominated by households with rather low levels of loyalty toward the promoted brand. If many consumers are largely indifferent between certain brands and willing to switch among certain brands if given a promotion, then perceived substitution costs may be low and perhaps not large enough to affect stockpiling behavior.
These two findings (regarding deal-prone and brand loyal households) should provide some comfort to manufacturers who are concerned that deals primarily attract consumers who stockpile brands that they would have bought at their regular price. The results from this study indicate that deal prone consumers, not brand-loyal consumers, are likely to be stockpilers. It may actually be beneficial to manufacturers to have deal prone consumers stockpile because it delays them from switching to some other brand that is on sale. In product categories where household penetration is either high or not easily increased, using deals to obtain brand switchers and defend against brand switching becomes a prominent objective, particularly if brand loyalty and perceived product differentiation are low.
Brand deal frequency was not consistently related to stockpiling behavior. This finding differs from Krishna (1994) who, in a laboratory experiment, found that if deals occur frequently there is less stockpiling and less purchasing of less preferred brands on deal. The results of this study suggest that many consumers have no salient deal frequency expectations and are willing to stockpile in response to the deals of many brands. In addition, the deal frequency measure is really only a surrogate for consumers' deal frequency perceptions and expectations. Future research should attempt to combine scanner-based behavioral data with perceptual and attitudinal information.
The regular price of a product and the depth of a price cut appear to be important in inducing stockpiling. Consistent with the theoretical framework of the inventory model, households are found to be more likely to stockpile if the brand's regular price is not high or if the depth of any price cut is high. The results also imply that consumers do attend to pricing and promotional information when making purchase quantity decisions. If an objective of price deals is to induce stockpiling, then it appears that steep price cuts will be influential in achieving that objective.
Finally, level of inventory was shown to have an effect on the incidence of stockpiling. As expected, lower inventories are associated with more stockpiling. This finding suggests that consumers do consider their holding costs when making purchase quantity decisions.
One limitation of the study is that information about coupon usage in the four product categories is not available. Although manufacturer couponing activity in the product categories is believed to be low, coupon redemptions by households may have affected purchase quantity decisions.
The results of the study suggest the need for research specifically examining the extent to which stockpiling affects household consumption rates (e.g. Wansink and Deshpande, 1994). Previous studies have often examined products which would seem to have relatively stable and not easily expanded consumption rates (e.g., toilet paper) and have assumed, at least implicitly, that household consumption rates are in fact stable. This assumption should be tested.
Differences in the findings across the four product categories illustrate the importance of continuing to broaden the scope of products used in sales promotion studies. Fader and Lodish (1990) have addressed this need in their efforts to discriminate among product categories based on the structure of the product category and its promotional activity. Further across-category research would strengthen confidence in the generalizability of sales promotion models and theories of the effect of deals on consumer purchasing behavior.
Finally, other research issues suggested by the findings of the study include the examination of other consumer characteristics (e.g., lifestyle characteristics) that might affect stockpiling behavior, and the effect of stockpiling in one product category on sales in other product categories.
(1) The selection criteria were as follows. First, households must have stated that the majority of their grocery shopping occurred at the store. Second, only households who made trips to the store throughout the entire 70 week period were included in the study. Households averaging less than one trip per week to the store were also excluded from the analysis because their purchase histories were believed to be incomplete. Third, only households that made seven or more purchases in the product categories during the 70 week period were included in the study.
(2) The brand carry-over constant (a) is set at .8. Other studies have found optimal values of this constant to be in the range of .8 to .875 and, furthermore, have found that parameter estimates and model fit are insensitive to small changes in the value of the constant (Gupta, 1988; Ortmeyer et al., 1991). At the first purchase occasion, the value of the brand loyalty variables are set to either (1--a)/(number of available brands--1), if the brand is not purchased on that occasion, or set to the constant (a), if the brand is purchased on that occasion.
Bawa, K. & R.W. Shoemaker (1987). "The Coupon-Prone Consumer: Some Findings Based on Purchase Behavior Across Product Classes," Journal of Marketing, (October), 99-110.
Blattberg, R. C., T. Buesing, P. Peacock, & S.K. Sen (1978). "Identifying the Deal Prone Segment," Journal of Marketing Research, (August), 369-377.
Blattberg R. C., G.D. Eppen, & J. Lieberman (1981). "A Theoretical and Empirical Evaluation of Price Deals for Consumer Nondurables," Journal of Marketing, (Winter), 116-129.
Blattberg, R.C. & S.A. Neslin (1990). Sales Promotion: Concepts, Methods, and Strategies,(Englewood Cliffs NJ: Prentice-Hall).
Bucklin, R.E. & J. M. Lattin (1991). "A Two-State Model of Purchase Incidence and Brand Choice," Marketing Science, (Winter), 24-39.
Fader, P.S. & L. M. Lodish (1990). "A Cross-Category Analysis of Category Structure and Promotional Activity for Grocery Products," Journal of Marketing, (October), 52-65.
Guadagni, P.M. & J. D. C. Little (1983). "A Logit Model of Brand Choice Calibrated on Scanner Data," Marketing Science, (Summer), 203-238.
Gupta, S. (1988). "Impact of Sales Promotions on When, What, and How Much to Buy," Journal of Marketing Research, (November), 342-355.
Gupta, S. (1991). "Stochastic Models of Interpurchase Time With Time-Dependent Covariates," Journal of Marketing Research, (February), 1-15.
Helsen, K.& D.C. Schmittlein (1992). "How Does a Product Market's Typical Price-Promotion Pattern Affect the Timing of Households' Purchases? An Empirical Study Using UPC Scanner Data," Journal of Retailing, 68 (3), 316-338.
Jeuland, A.P. & C. Narasimhan (1985). "Dealing--Temporary Price Cuts--by Seller as a Buyer Discrimination Mechanism," Journal of Business, 58 (3), 295-308.
Krishna, A., I. S. Currim, & R.W. Shoemaker (1991). "Consumer Perceptions of Promotional Activity," Journal of Marketing, (April), 4-16.
Krishna, A. (1994). "The Effect of Deal Knowledge on Consumer Purchase Behavior," Journal of Marketing Research, 31, 76-91.
Krishna, A. & R.W. Shoemaker (1992). "Estimating the Effects of Higher Coupon Face Values on the Timing of Redemptions, the Mix of Coupon Redeemers, and Purchase Quantity," Psychology and Marketing, 9 (6), 453-467.
Krishnamurthi, L. & S.P. Raj (1988). "A Model of Brand Choice and Purchase Quantity Price Sensitivity," Marketing Science, (Winter), 1-20.
Krishnamurthi, L. & S. P. Raj (1991). "An Empirical Analysis of the Relationship Between Brand Loyalty and Consumer Price Elasticity," Marketing Science, (Spring), 172-183.
Krishnamurthi, L., T. Mazumddar, & S.P. Raj (1992). "Asymmetric Response to Price in Consumer Brand Choice and Purchase Quantity Decisions," Journal of Consumer Research, 19, 387400.
Moriarty, M. (1983). "Feature Advertising-Price Interaction Effects in the Retail Environment," Journal of Retailing, (Summer) 80-98.
Moriarty, M. (1985). "Retail Promotional Effects on Intra- and Interbrand Sales Performance," Journal of Retailing, (Fall), 27-47.
Neslin, S.A., C. Henderson & J. Quelch (1985). "Consumer Promotions and the Acceleration of Product Purchases," Marketing Science, (Spring), 147-165.
Ortmeyer, G., J. M. Lattin, & D.B. Montgomery (1991). "Individual Differences in Response to Consumer Promotions," International Journal of Research in Marketing, 8, 169-186.
Papatla, P. (1995). "A Dynamic Model of the Advertising-Price Sensitivity Relationship for Heterogeneous Consumers," Journal of Business Research, 33, 251-260.
Papatla, P. & L. Krishnamurthi (1996). "Measuring the Dynamic Effects of Promotions on Brand Choice," Journal of Marketing Research, 33, 20-35.
Schneider, L. & I. Currim (1990). "Consumer Purchase Behaviors Associated with Active and Passive Deal Proneness," International Journal of Research in Marketing, 8, 205-222.
Walters, R. G. & H. J. Rinne (1986). "An Empirical Investigation into the Impact of Price Promotions on Retail Store Performance," Journal of Retailing, (Fall), 237-266.
Wansink, B. & R. Deshpande (1994). "Out of Sight, Out of Mind: Pantry Stockpiling and BrandUsage Frequency," Marketing Letters, 5(1), 91-100.
Webster, F. E. (1965). "The Deal-Prone Consumer," Journal of Marketing Research, (May), 186189.
Fred M. Beasley,
Northern Kentucky University
Figure 1 Summary of Hypothesized Factors Affecting Stockpiling Behavior Household Characteristics Deal Proneness (H1) (+) Brand Loyalty (H2) (+) Product Consumption Rate (H3) (+) Product-Related Factors Deal Frequency (H4) (-) ======> Incidence of Stockpiling Price (H5) (-) Purchase Situation Characteristics Product Inventory Level (H6) (-) Depth of Price Discount (H7) (+) Table 1 Purchases of Households Making 7+ Purchases Estimation Period Purchases Coffee Paper Toilet Canned (40 weeks) Towels Paper Tuna Off-Deal 1,668 3,247 3,575 1,274 Deal 1,794 3,708 4,144 3,278 Number of Households 395 680 756 553 Table 2: Parameter Estimates for the Stockpiling Models (Standard errors in parentheses) Coffee Paper Toilet Canned Towels Paper Tuna Deal Proneness .319 ** 3.299 ** 1.902 ** 4.520 ** (.112) (.184) (.104) (.254) Brand Loyalty -0.181 0.068 -0.088 -0.264 (.179) (.204) (.152) (.166) Consumption Rate -.067 ** -.003 * -.005 ** .047 ** (.013) (.001) 0 (.013) Deal Frequency -0.01 -.037 * -.026 ** .072 ** (.011) (.015) (.007) (.014) Product Price -.291 ** -2.673 ** 0.177 -.102 ** (.030) (.259) (.338) (.018) Inventory Level -.033 * -.033 ** -.048 ** -0.009 (.012) (.009) (.008) (.008) Price Cut .176 ** 1.817 ** .917 * .245 ** (.026) (.508) (.446) (.034) * Significant at .05 level ** Significant at .01 level N 1794 3708 4144 3278 -2LL (model) 1932.92 3381.67 4858.57 3743.5 Model chi-square 230.71 ** 656.26 ** 578.63 ** 703.94 ** (7 d.f.) [p.sup.2] .106 .162 .106 .158
|Printer friendly Cite/link Email Feedback|
|Author:||Beasley, Fred M.|
|Publication:||Academy of Marketing Studies Journal|
|Date:||Jan 1, 1998|
|Previous Article:||Enhancing the delivery of quality education: a Services Marketing perspective.|
|Next Article:||The market orientation performance relationship in minority and woman-owned small firms.|