Effects of base price upon search behavior of consumers in a supermarket: An operant analysis

Journal of Economic Psychology 24 (2003) 637–652 www.elsevier.com/locate/joep

Effects of base price upon search behavior of consumers in a supermarket: An operant analysis Jorge M. Oliveira-Castro

*

Departmento de Processos Psicol ogicos Basicos, Instituto de Psicologia, Universidade de Brasılia, Campus Universit ario Darcy Ribeiro, Asa Norte, Brasılia, DF 70910-900, Brazil

Abstract The effects of product base price upon the duration of search behavior of consumers was investigated in a supermarket using an operant framework. Searching was interpreted as a pre-current behavior, influenced by the consequences for buying and consuming. Search duration was measured while consumers selected two cleaning products (Experiment 1) and two food products (Experiment 2), differing in base price. Search duration was significantly larger for the more expensive products. Consistent individual differences in search duration were also observed across products. These results, obtained from direct observation of search behavior, corroborate those found in the literature which used laboratory simulations and surveys, and illustrate the feasibility of an operant analysis of consumer behavior. Ó 2003 Elsevier B.V. All rights reserved. PsycINFO classification: 3920 JEL classification: M30 Keywords: Consumer behavior; Search behavior; Base price; precurrent behavior; Operant psychology; Consumer psychology; Consumer economics

1. Introduction Consumers usually gather information before purchasing products. They may search for a favorable price among different stores or brands, examine product quality, *

Tel.: +55-61-9143074; fax: +55-61-2730203. E-mail address: [email protected] (J.M. Oliveira-Castro).

0167-4870/$ - see front matter Ó 2003 Elsevier B.V. All rights reserved. doi:10.1016/S0167-4870(03)00006-0

638

J.M. Oliveira-Castro / Journal of Economic Psychology 24 (2003) 637–652

try out new products or brands, or investigate payment conditions. This search may be brief and effortless, as when the person chooses a soft drink on a supermarket shelf, or long and costly, as when one is looking for an apartment. The identification of the variables that influence pre-purchase behavior may be important to our understanding of store and brand choice, sale promotion effectiveness, reactions to new products or packages, effects of store and product location, and such like. The amount of money that can be saved as search increases has been often cited as one of the variables that might influence search behavior. Some authors have suggested that perfectly rational consumers should continue to search until the expected gain from more search is less than its cost (Stigler, 1987). According to this analysis, consumers should know the distribution of prices in the market and the costs of searching, in order to maximize utility by balancing the amount of money saved from more search with the costs associated with that search. However, consumer behavior is not always optimal as described by traditional economic theory. It has been shown to be influenced by several other factors beyond the absolute amount of money saved and search costs (cf. Kahneman & Tversky, 1984; Thaler, 1985). Results stemming from experimental investigations, for instance, have shown that more participants said that they would make a trip to another store when they could get a discount of 33% than when the discount was equal to 4%, despite the fact that in both conditions they would save the same amount of money ($5) (cf. Tversky & Kahneman, 1981; Kahneman & Tversky, 1984). This inconsistency in their decisions may be explained, according to the authors (cf. Kahneman & Tversky, 1979), when one considers that financial transactions might be framed at different levels, namely, minimal, topical, or comprehensive. The minimal frame would consider only the absolute amount of money gained or lost, whereas a topical frame would be more inclusive, comparing, for example, the current price with the last price paid for the same product. A comprehensive frame would be even more inclusive and would consider, for instance, oneÕs monthly budget. In the case of the above mentioned results, participants would have been influenced by the percentage off the price, which would be a topical frame, rather than by the absolute amount of money saved, a minimal frame. Findings such as these led Kahneman and Tversky (1984) to conclude that sale prices are framed according to the percentage off rather than in terms of absolute amount of money saved. However, this conclusion has not been completely supported by subsequent research. Darke and Freedman (1993, Experiment 1), using a laboratory simulation, found that when the percentage to be saved by visiting another store was low (1% or 5%), the absolute amount of money ($5 or $25) to be saved influenced participantsÕ decision to extend price search. When the percentage of the base price to be saved varied more widely (5–25%), both variables influenced participantsÕ decisions (Darke & Freedman, 1993, Experiment 2). Another experiment corroborated and extended such results by showing that the percentage of discount influenced participants decision of ending price search for a product with low base price ($100) but not for a product with high base price ($300) (Darke, Freedman, & Chaiken, 1995). The authors explained these results by proposing a heuristic–systematic model of price search, which distinguishes between systematic and heuristic processing of

J.M. Oliveira-Castro / Journal of Economic Psychology 24 (2003) 637–652

639

information. Systematic processing would involve the examination of a large amount of information and would tend to be relatively effortful, whereas heuristic processing would involve simple decision rules and would require less effort. In the case of price search, a systematic processing would consist of visiting a large number of stores or examining a large number of brands in an attempt to find the best possible price. This type of search would occur when the consumer is highly motivated, as when a large amount of money can be saved (e.g., high base price, around $300). When the amount of money to be gained is small (e.g., because the base price is low), consumers would use the percentage discount of a sale price as a heuristic cue to decide to continue or stop searching. These different ways of cognitively processing information would, according to the authors, explain the results from their experiments and some apparently inconsistent results found in the literature (i.e., Blair & Landon, 1981; Della Bitta, Monroe, & McGinnis, 1981; Urbany, Bearden, & Weilbaker, 1988). Taken together, results from the above-cited experiments suggest that both percentage off the initial price and the amount of money that can be saved may influence consumer price search, depending upon the product base price and the size of the discount. Other investigations using data from surveys have also suggested the influence of product base price, or price variation, upon the behavior of searching for Christmas gifts (Laroche, Saad, Browne, Cleveland, & Kim, 2000) and price search in the retail grocery market (Urbany, Kalapurakal, & Dickson, 1996). All the investigations cited above illustrate the kind of theoretical explanation that has prevailed in the field for quite some time. In 1990, Foxall in his critique of the cognitive paradigm identified this tendency where he asserts that ‘‘the most widely-accepted and influential models of consumer behaviour derive from cognitive psychology which is rapidly assuming the status of a dominant, though not exclusive, paradigm for psychological research in general’’ (p. 8). Consumer choice, according to such models, is interpreted as a problem-solving activity, determined by rational and intellectual processing of information. Upon being presented with information concerning products, brands, prices and such like, consumers analyze, classify and interpret such information, transforming it, through more processing, into attitudes and intentions that ultimately produce choices and actual purchases (cf. Foxall, 1990, 1997, 1998). This type of explanation has not been exempt from theoretical, conceptual and empirical criticism. Theoretically, one of the major critics of such theories was Skinner (1953, 1969, 1977, 1985), who described them as incomplete, for they do not explain the presumed inner causes of behavior, fictional, for they are re-descriptions concealed as explanations, and unnecessary, for they could be replaced by much simpler environment-based theories (cf. Foxall, 1990). In the context of consumer behavior theory, such models have not emphasized the possible and probable effects of situational variables about which there is little systematic information (Foxall, 1997, 1998). Empirically, several assumptions of cognitive theories have not been supported by data from consumer behavior, since, as Foxall (1990) points out, consumers do not seem to be as rational, motivated and information-seeking as predicted by cognitive models. Moreover, a growing body of research has demonstrated

640

J.M. Oliveira-Castro / Journal of Economic Psychology 24 (2003) 637–652

that the prediction level of cognitive models of attitude is increased when they include situational variables, such as context specificity and previous behavior (cf. Foxall, 1997; Foxall, 2002b; Davies, Foxall, & Pallister, 2002). Independently of such criticisms, however, the present proposal can be justified as an attempt to extend the scope of consumer research beyond the prevailing cognitive paradigm. According to a relativistic perspective, the presentation of an antithetical theoretical position may, by itself, be salutary to the growth of the field (cf. Foxall, 1990, 1998, 2002a; Foxall & Greenley, 1998). Based upon these assumptions, the present work presents an operant analysis of consumer search behavior and an empirical investigation of the effects of base price on the duration of search behavior in a supermarket.

2. Consumer search behavior: An operant analysis In operant terms, pre-purchasing search may be interpreted as a kind of precurrent (or mediating) behavior. Skinner (1953, 1957, 1968, 1969) used this concept to refer to responses that increase the likelihood of other response (current) occurring or being reinforced. Looking up a phone number in the directory (precurrent) may increase the probability of correct dialing (current), which in turn might function as a precurrent to talking to someone. Precurrent contingencies may differ with respect to several characteristics (cf. Polson & Parsons, 1994). Precurrent responses may be described as signaled, when they produce stimulus changes (e.g., the number in the directory) correlated with changes in the reinforcement parameters for the current response (e.g., dialing the number). They can also be described as required or not required by the programmed contingencies. Getting food from the refrigerator (current) could not be reinforced without opening the refrigerator door (i.e., required precurrent response), whereas dialing a phone number could be reinforced without the response of looking up the number in the directory, as it happens when the person already knows the number (cf. Oliveira-Castro, Coelho, & Oliveira-Castro, 1999; Oliveira-Castro, Faria, Dias, & Coelho, 2002). The occurrence of precurrent responses, therefore, is influenced by the consequences for the current response in the sequence. If, for instance, the numbers in the directory were incorrect, as it may happen with old directories, the looking-up response would not be associated with increased probability of correct dialing. In this case, the response of looking up the number would eventually stop occurring, not because the person already knows the number, as mentioned above, but because the response lost its function. Fig. 1 displays a schematic representation of operant relations involving search and buying behavior. In the figure, S D , R, S
, and S þ stand for discriminative stimuli (i.e., events in the presence of which certain responses were reinforced in previous occasions), operant responses (i.e., responses influenced by their consequences), aversive stimuli (i.e., events that, when produced by a response, decrease its future probability) and reinforcing stimuli (i.e., events that, when produced by a response, increase its future probability), respectively. Arrows with solid lines point to the consequences of a response, whereas a colon indicates what response-consequence rela-

J.M. Oliveira-Castro / Journal of Economic Psychology 24 (2003) 637–652

641

Fig. 1. Schematic representation of operant relations involved in search behavior of consumers.

tions prevail in the presence of a given stimulus situation. Arrows with dotted lines show events that become associated as the sequence is repeated, that is to say, antecedent events in the sequence may, after repetitive pairing, predict the occurrence of consequent events. Brackets indicate events in the sequence that may be skipped (i.e., Rsearch may not occur). In the case of search behavior of consumers, responses such as looking for product information and prices may increase the likelihood (or the magnitude) of reinforcement for the response of buying the product. At least two types of consequences, reinforcing and aversive, should be considered in a functional analysis D of search behavior. Searching is reinforced by producing events (Sbuying ) associated with increases in the probability of ‘‘successful’’ buying, that is, the probability of buying products with high quality (i.e., strong reinforcing effects when consumed) and low prices. Searching may produce, for example, a product with a better price, which functions as a discriminative stimulus to buying. At the same time, searching
produces aversive consequences (Scost ), for it is associated with response costs, which can involve time, money and effort, spent in the activity. Among other things, the D
proportion between such reinforcing and aversive consequences (Sbuying /Scost ) will probably influence the probability, duration and intensity of search behavior, as predicted by classical economic theory (Stigler, 1987). Searching would be a precurrent response to buying (Rbuying ), which, in turn, also produces reinforcing and aversive consequences, being interpreted as an approach– avoidance behavior (cf. Alhadeff, 1982; Foxall, 1990, 1997, 1998). Purchasing is folD lowed by reinforcing events (Sconsumption ), such as the opportunity to consume or use the product (Rconsuming ). In this sense, buying may function as a precurrent response to consuming the product, since it is maintained by the increased probability of consuming or using it. This is shown by the fact that the probability of buying a product should decrease if the product cannot be consumed (e.g., if it is spoiled or broken), or cannot be consumed immediately (e.g., when reinforcing consequences are postponed as in delayed product delivery). In both cases there would be a decrease in þ the reinforcing magnitude of consumption (i.e., decrease Sproduct ). This type of conþ sequence (Sproduct ), associated to consumption or use, is similar to what Foxall (1990, 1997, 1998) described as utilitarian reinforcement. It may be useful to notice that buying need not be, and usually is not, followed immediately by consumption and its direct reinforcing consequences. Most of the time, acquired products are consumed or used some time after the purchase or are even consumed or used by others. In the latter case, buying is probably maintained by social events in which reinforcer delivery is mediated by other people (cf. Skinner, 1957), as when one buys a gift to a

642

J.M. Oliveira-Castro / Journal of Economic Psychology 24 (2003) 637–652

child or buys a cleaning product to be used by an employee. In such situations, the product will not be used or consumed directly by the buyer, but buying, and consequent gift-giving, may be followed by reinforcing reactions, commentaries, and such like, from the child or the employee. These social events should therefore be also inþ cluded in the reinforcing consequences for ‘‘consumption’’ (i.e., Sproduct should include such events). As mentioned before, buying has also aversive consequences
(Smoney ) since it implies a loss of money, that is, it implies a loss of something that usually functions as a generalized positive reinforcer. The probability of buying would therefore depend, among other things, on the proportion of such conseþ
quences (Sproduct /Smoney ). Consequences for consuming that have low reinforcing magnitude associated with high price will decrease the probability of buying that product or brand (the level of stimulus generalization may vary) on future occasions. By the same token, praises concerning the acquired product received from the employee who uses it, associated with low prices, will probably increase the likelihood of buying that product or brand in later occasions. These aversive and reinforcing consequences of buying may become associated to the events produced by searching (indicated by the dotted-lined arrow a). If, for example, after searching for a given product, one finds a new brand with a low price and buys it, the consequences associated with consumption will become associated to that brand. In future occasions, that brandÕs features (e.g., name, package, color) will be predictive of the consequences of previous consumption, that is, the brand whose product showed good quality (i.e., strong reinforcing effects) when consumed will acquired the function of indicating a high probability of reinforcement for consumption, having therefore higher chances of being purchased. Brands whose products showed bad quality (i.e., low reinforcing effects) will signal low reinforcement probability and, consequently, would not be bought on subsequent occasions (cf. Foxall, 1990, 1997, 1998). Pre-purchase searching, as a precurrent behavior, is usually not required by the programmed contingencies, since one can, in most situations, buy without any search. This may happen, however, for several reasons, each one of which involving different functional relations. Search behavior may not occur due to a lack of alternatives. If there were only one brand of a given product in the market, then there would be no possibility that product being substituted by another, and the product would be sold at the same price everywhere: there would be no need for search on the part of potential consumers. Paying taxes would be another example where search is absent due to lack of options. In FoxallÕs (1990, 1997, 1998) terminology, such consumer behaviors would take place in ‘‘closed’’ settings, those which preclude or greatly restrict choice. In these cases, there is no reinforcing consequence for searching since such behavior does not produce events associated with increases in product quality or decreases in price. Search behavior may also be absent when response cost is too high compared to the product value, as when one needs to travel to another city to find another brand. Here, an increase in the magnitude of aversive consequences for searching decreases the response probability. Little search would also occur in situations where oneÕs need (this expression avoids the use of ‘‘deprivation level’’) for things such as a can opener is too intensive or too urgent, as in the case

J.M. Oliveira-Castro / Journal of Economic Psychology 24 (2003) 637–652

643

of selecting a restaurant after a long time without a meal or choosing for a hospital in an emergency (these are sometimes called ‘‘impulsive buying’’). Searching may also be minimal or inexistent when alternative brands and prices remain unchanged, which allows consumers to learn the alternative brands and prices. In such situations, products and brands are the same for a long period of time, their prices remain the same also for a long time, and the person has a long experience buying and consuming such products. In these cases, consumers would probably purchase the same product or brand without much search, for the person already ‘‘knows’’ all the alternatives. In other words, the duration of search would decrease due to a transference of stimulus function between the events produced D by search (Sbuying ), such as products characteristics and prices, and the events in the initial setting (Sinitial ), such as the available alternatives of brands or stores. Certain aspects of the initial setting would acquire the functions of the events produced by search, that is, aspects of the initial setting may predict the number of alternative brands, price range, product quality, and such like. This decrease is similar to the decrease in duration of the response of looking up frequently used phone numbers in the directory. As each name–number pair is repeated and the person looks up the number, the function of the number in the directory, which allows the person to dial correctly, is transferred to the name. That is, after several pairings the name acquires the function of allowing correct dialing (cf. Oliveira-Castro et al., 1999; Oliveira-Castro et al., 2002). In the case of searching for product quality and price, aspects of the initial setting acquire the function of making possible a successful purchase without searching. According to this line of reasoning, search duration in fixed settings may be indicative of brand association learning. By ‘‘fixed’’ setting it is meant a setting in which alternative prices, products, brands, and such like, remain unchanged for long periods of time. This would not be the same as closed settings, which would imply a lack of alternatives (cf. Foxall, 1990, 1997, 1998). The initial setting (Sinitial ), where the purchase occurs, contains a number of events that influence search and buying probabilities. Alternative brands, products or stores, are some obvious sources of influence, since each brand (or product or store) may have its own history of association with strongly (or frequently) or weakly (or rarely) reinforced consumption. In this sense, different brands (or product or store) may have different discriminative ‘‘power’’ in the sense that each may have different degrees of association to past reinforcement. The more frequently and systematically the event is associated to reinforcement (or punishment), higher would be its discriminative positive (or negative) power. Such history may be very specific, as when the specific product from a specific brand has been previously consumed or used, or very general, as when other products of that brand have been tried out but not that specific one (the effect in this case is due to stimulus generalization). Moreover, each brand may also have a different history of association with instructions from friends or relatives and with advertisement. In this case, some instructions and advertisements may function as discriminative stimuli to ‘‘doing what they say’’ kind of behavior, which has been successful (i.e., reinforced) in the past. Some advertisements may exert some influence based on other kinds of association, such as Pavlovian conditioning, according to which some brand features may elicit reactions that

644

J.M. Oliveira-Castro / Journal of Economic Psychology 24 (2003) 637–652

typically accompany ‘‘pleasurable’’ events (e.g., Kim, Lim, & Bhargava, 1998). Therefore, the initial setting may contain different products or brands, and their specific features, each one of which has a different discriminative power over searching and buying, in the sense that each has a different history of association. Searching and buying will depend upon the interactions of all these sources plus the other variables already analyzed, such as prices and search costs. Sale promotions may influence the discriminative power distribution of the setting by, for example, indicating price reduction of a less powerful brand. The variables that influence the discriminative power of brands and products, must, of course, be identified empirically and could be analyzed for individual consumers or groups of consumers. The present interpretation can be viewed as an alternative to more cognitively oriented theories of brand association learning (e.g., van Osselaer & Alba, 2000; van Osselaer & Janiszewski, 2001). Other variables not directly related to past associations between the initial setting and consumption may also influence search and buying. One example would be the intensity of oneÕs need, as mentioned above, or time pressure. Searching for a restaurant, for example, may be reduced if one is too hungry or in a hurry. Characteristics of the shopping environment, such as music (e.g., North, Hargreaves, & McKendrick, 1999), could also be mentioned as examples, since they may also influence searching and buying despite the fact that they are not predictors of product quality. The present functional analysis would also be applicable to store selection, which would be analogous to an operant analysis of brand choice. The initial setting would contain store alternatives, their distances, parking facilities, and such like, which may influence the probability and duration of searching and going to the store (analogous to buying), which would give the opportunity of shopping in the store (analogous to consumption), which would be followed by reinforcing or aversive consequences. Such consequences, which would probably include finding the desired product, store atmosphere, staff attendance, and so on, would become associated to the store name, location, and such like, which would become predictive of reinforcing or aversive consequences.

3. Effects of base price on search duration in a supermarket As mentioned previously, there are some results in the literature suggesting that consumers engage in more search for items with higher base prices than for items with lower base prices (Darke et al., 1995). Such results were obtained in shopping simulations and surveys, and, according to the authors, are due to the fact that ‘‘higher base prices are associated with greater potential gains and losses’’ (p. 581), since more expensive items usually present wider price variation than less expensive ones. According to the model presented here, the probability and duration of search would be a direct function of the potential increase in the probability of a successful D
purchase (Sbuying ) and an inverse function of search costs (Scost ). Increases in the probability of a successful purchase are associated to increases in the quality of the prod-

J.M. Oliveira-Castro / Journal of Economic Psychology 24 (2003) 637–652

645

uct and decreases in prices. Considering that prices for more expensive products usually vary more widely than those for cheaper ones, and assuming that product quality is kept constant, the probability of reinforcement for searching would be higher, based on larger possible savings, for products with higher base price. Based upon this increased reinforcement probability one can predict higher frequency and duration of searching for products with higher base price, when searching costs are kept constant. This increased probability of reinforcement would explain the ‘‘effects’’ of perceived price dispersion, often cited in economics as an indicator of search benefit (cf. Urbany, 1986). With the purpose of testing this prediction, the duration of searching for products with different base prices in a supermarket was examined in two experiments. Although supermarkets may be described as open settings (also low utilitarian, low informational; Foxall, 1990, 1997, 1998), an attempt was made to control, as much as possible, for some of the variables of the initial setting that could influence searching behavior. Differently from previous investigations of search behavior, the present research examined searching duration in real-life situations rather than in laboratory simulation, where participants pretended that they were purchasing products. Also different was the fact that, in the present investigation, shoppers chose among different brands and types of a product in the same store rather than choosing for the same product among different stores. Moreover, the experiments conducted by Darke and Freedman (1993) and Darke et al. (1995) were primarily concerned with the effects of percentage discount when compared to the effects of the amount of money that could be saved. Effects of sale promotions were not investigated here.

4. Experiment 1 With the purpose of testing the effects of base price upon search duration of products in a supermarket, searching time for two cleaning products was measured. Search duration was recorded while shoppers searched for and selected the target products. 4.1. Method 4.1.1. Participants Forty-nine shoppers (35 women and 14 men) in a large supermarket, located in Brasilia, Brazil, were observed. They were selected for observation in the sequence that they approached the shelves where the target products were displayed. They could be alone or accompanied, as long as conversations did not disrupt product searching. 4.1.2. Equipment and material Two cleaning products, washing-up liquid and fabric softener, were defined as the target products. They were selected on the basis of the following criteria: reasonable difference in price, similar space on the shelves, similar number of alternative brands

646

J.M. Oliveira-Castro / Journal of Economic Psychology 24 (2003) 637–652

and/or package sizes, proximal location, and proximity to aisles (in order to facilitate observation). Washing-up liquid, was located, across the aisle, right in front of fabric softener. Both products occupied, vertically, seven shelves, and, horizontally, five shelves with a width of 1.40 m each. Eight different brands of washing-up liquid were displayed, all of which having the same package size. Twelve different types of fabric softener were available, including nine different brands and two package sizes (three brands presented two package sizes). Washing-up liquid prices ranged from R$ 0.35 to R$ 0.70 (at the time of the experiments the exchange value of U$ 1.00 varied from R$ 2.30 to R$ 2.40 approximately), with a mean of R$ 0.55 (SD ¼ 0:12). Fabric softener prices ranged from R$ 1.49 to R$ 3.73, with a mean of R$ 2.55 (SD ¼ 0:74). For both products, different brands were interspersed on the shelves, each brand occupying more than one location. Near the cheapest brand of each product, there was a sale promotion sign containing the brandÕs name. 4.1.3. Procedure The observation procedure consisted in measuring the time shoppers spent looking at the product on the shelf. This was measured from the moment they started looking at the shelf until they put the product in the cart. Such time periods were measured by using a wrist watch containing a chronograph. Shoppers were observed while searching for washing-up liquid, fabric softener or both, depending on what the shopper did. Fifteen shoppers chose both products whereas the other 34 selected only one of them. Data were discarded in the presence of any interruption in searching time due to conversation, distraction or such things. The observer stood at a certain distance holding a shopping cart with some products in it, and acted as if he was also shopping or waiting for someone else. Three periods of observation, ranging in duration from 1 to 2 h approximately, were conducted. All such periods occurred either after 3:00 pm on weekdays or between 11:00 am and 1:00 pm on Saturday. The following information was recorded for each observation: search duration, number of items selected, sex, number and type of company (e.g., sex, adult or children), and approximate volume of products in the cart (empty, half full or full). This last measure was an attempt to control for antecedent shopping time, which could have some systematic effect related to fatigue or time pressure. 4.2. Results and discussion Considering that 15 shoppers (12 women and 3 men) selected both products, their data were analyzed, as a within-subject procedure, separately from those obtained from the other 34 consumers. For nine of these 15 shoppers, search duration (in seconds) was longer for fabric softener (M ¼ 26:73, SD ¼ 18:26) than for washing-up liquid (M ¼ 18:00, SD ¼ 13:30). A one-way repeated-measures Anova indicated that this difference was not significant (F ¼ 2:84, p ¼ 114). However, considering that the selected quantity of the two products varied considerably (washing-up liquid from 1 to 5, fabric softener from 1 to 3), and that it could confound possible effects of base price, search duration per selected unit was calculated by dividing search duration by the selected quantity. For 13 of 15 shoppers, this duration was longer for fabric soft-

J.M. Oliveira-Castro / Journal of Economic Psychology 24 (2003) 637–652

647

ener (M ¼ 23:09, SD ¼ 19:37) than for washing-up liquid (M ¼ 9:90, SD ¼ 10:24). A one-way repeated-measures Anova indicated that this was a significant difference between the two products (F ¼ 9:64, p ¼ 0:008). These results indicate that search duration per purchased unit was larger for the more expensive item. These data, obtained for the two products from the same consumers, may also be used to examine the consistency of individual differences in search duration across products, that is, to examine if those consumers that spent more time searching for washing-up liquid also tended to spend more time searching for fabric softener. Pearson correlation analysis yielded a significant positive coefficient between search duration per unit for washing-up liquid and fabric softener (r ¼ 0:53, N ¼ 15, p ¼ 0:043), indicating consistent individual differences in search duration across products. Data from the other 34 shoppers (23 women and 11 men), who selected either washing-up liquid (10 women and 9 men) or fabric softener (13 women and 2 men) were analyzed as a between-subject procedure. Mean search duration was larger for fabric softener (M ¼ 30:47, SD ¼ 19:15) than for washing-up liquid (M ¼ 26:53, SD ¼ 15:37), although the difference was not statistically significant (F ¼ 0:44, p ¼ 0:510). Search duration per quantity selected was also larger for fabric softener (M ¼ 27:96, SD ¼ 25:52, quantity varied from 1 to 3) than for washingup liquid (M ¼ 14:43, SD ¼ 9:61, quantity varied from 1 to 10), such difference being statistically significant (F ¼ 4:55, p ¼ 0:041). Such results corroborated those obtained from the within-subject analysis, indicating that search duration per product unit was larger for the product with higher base price. These results should be interpreted with caution, however, considering the lack of control over several variables in the situation. One of these variables was related to the patterns of search behavior observed for the two different products. Half the women who selected fabric softener opened the plastic bottles of different brands and smelled the product. This pattern was not observed for washing-up liquid and may have biased the results, since it may have increased search duration for fabric softener. Another difference between the two products that may have confounded the present results was the number of alternative brands and sizes displayed for each product. More alternatives were displayed for fabric softener (12) than for washing-up liquid (8) which may have increased search duration for fabric softener. In addition to this, the proportion of women, in the between-subject sample, who selected fabric softener (i.e., 86.67%) was larger than that for washing-up liquid (i.e., 52.63%), which may also have biased the present findings.

5. Experiment 2 Considering the possible confounding effects of the above-mentioned variables, two other target products were selected for observation in Experiment 2. An attempt was made to select products that: (a) belonged to a different category (i.e., food) from those in Experiment 1, which might increase the generality of the results;

648

J.M. Oliveira-Castro / Journal of Economic Psychology 24 (2003) 637–652

(b) would generate similar patterns of search behavior; and (c) would have the same number of alternatives or a larger number for the product with lower base price. 5.1. Method 5.1.1. Participants Thirty-six shoppers (27 women and 9 men) in a large supermarket, located in Brasilia, Brazil, were observed. They were selected for observation in the sequence that they approached the shelves where the target products were displayed. They could be alone or accompanied, as long as conversations did not disrupt product searching. 5.1.2. Equipment and material Two food products, tomato puree and green olives (sold in glass jars), were defined as target products. They were selected on the basis of the following criteria: reasonable difference in price, similar space in the shelves, similar number of alternative brands and/or package sizes, proximal location, and proximity to aisles (in order to facilitate observation). Tomato puree was located, across the aisle, right in front of green olives. Both products occupied, vertically, seven shelves, and, horizontally, three shelves with a width of 1.40 m each. Nineteen different alternatives (13 brands and three package sizes) of tomato puree were displayed. Fourteen alternatives (eight brands and three package sizes) of green olives were available. Right beside the shelves where the green olives were displayed there was a space, width of one shelf and height of seven shelves, displaying black olives. Data from customers that looked at black olives at any moment were discarded. Tomato puree prices ranged from R$ 0.41 to R$ 1.35, with a mean of R$ 0.82 (SD ¼ 0:26). Green olives prices ranged from R$ 1.25 to R$ 6.47, with a mean of R$ 3.82 (SD ¼ 1:35). For both products, each brand occupied only one location on the shelves. Near the cheapest brand of each product (largest package size), there was a sale promotion sign containing the brandÕs name. For these same products, there was an end-of-aisle promotion for tomato puree and a promotion for green olives announced in the store magazine. 5.1.3. Procedure Identical to Experiment 1, except for the fact that data were recorded during one 2-h period of observation. 5.2. Results and discussion Eighteen shoppers (14 women and 4 men) selected tomato puree whereas the other 18 selected green olives (13 women and 5 men). During the observation period, no customer selected both products. Mean search duration was larger for green olives (M ¼ 31:67, SD ¼ 31:94) than for tomato puree (M ¼ 15:61, SD ¼ 9:70), a statistically significant difference (F ¼ 4:16, p ¼ 0:049). Search duration per product unit was also larger for green olives (M ¼ 30:14, SD ¼ 32:24, quantity varied from

J.M. Oliveira-Castro / Journal of Economic Psychology 24 (2003) 637–652

649

1 to 2) than for tomato puree (M ¼ 12:29, SD ¼ 11:44, quantity varied from 1 to 4), such difference being statistically significant (F ¼ 4:90, p ¼ 0:034). These results indicate that both total search duration and search duration per product unit were larger for the product with higher base price. Although in Experiment 1 only search duration per unit showed significant differences, this discrepancy may be due to the variation in selected quantity. Whereas in Experiment 1 this varied from 1 to 10, in the present experiment selected quantity varied from 1 to 4. This narrower range of variation approximates the measures of total search duration and search duration per unit, suggesting that the effect of base price on search duration per product unit is more systematic than that on total search duration.

6. General discussion The results obtained in Experiments 1 and 2 suggest that the duration of consumer search behavior per product unit is larger for products with higher base price. These findings were replicated across experiments despite several differences among target products. The two pairs of supermarket products investigated here belonged to different categories (i.e., cleaning products and food), had different numbers of alternative brands and package sizes (varying from 8 to 19), produced different search patterns (i.e., looking and smelling or just looking), occupied different space on the shelves (width equal to 4.20 or 7.00 m), and were displayed differently on the shelves (i.e., side by side, Experiment 2, or mixed, Experiment 1). Taken together, these results suggest that search duration per unit increases with product base price. This relation can be seen in Fig. 2, which shows mean search duration per product unit as a function of product base price for the data obtained in Experiments 1 and 2. Despite all the above-mentioned differences between the two

Search Duration per Unit (s)

40

30

20

10 .5

1.0

1.5

2.0

2.5

3.0

3.5

4.0

Mean Product Price (R$)

Fig. 2. Search duration per unit of selected product as a function of product base price in Experiments 1 and 2.

650

J.M. Oliveira-Castro / Journal of Economic Psychology 24 (2003) 637–652

pairs of products, Pearson correlation analysis indicates significant increases in search duration per product unit as a function of increases in product base price (r ¼ 0:34, N ¼ 70, p ¼ 0:004). These systematic results were obtained despite wide individual differences in search duration, indicated by the large values of standard deviations, proportionally to the means, found in both experiments. Results from Experiment 1 also suggest that such individual differences in search duration are consistent across products, that is, customers that spend more time searching for one product also tend to spend more time searching for other products. According to the interpretation proposed here, the probability and duration of searching would be a direct function of the potential increase in the probability of D
successful purchase (Sbuying ) and an inverse function of search costs (Scost ). Assuming that, in each experiment, costs associated with searching were almost constant across products, the duration of searching would depend primarily upon the probability of successful purchase, which, in turn, would depend upon potential increases in product quality and potential decreases in price. Considering that prices for more expensive products usually vary more widely than those for cheaper ones, the probability of reinforcement for searching would be higher, based on larger possible savings, for products with higher base price. This increased probability of reinforcement for searching was expected to increase search duration. This effect, according to the present interpretation, would be dependent upon consumersÕ experience with price variation in the market. When prices do not vary, for example, when paying government taxes or electric energy bills, search should be reduced. Moreover, if price variation for expensive products were equal, in terms of absolute amount of money, to those for cheap products, one would not predict more extended search for more expensive items. In such cases there would not be an increased reinforcement probability for searching for more expensive items. In the present experiments, means and standard deviations of product prices were highly positively correlated (r ¼ 0:99, N ¼ 4, p ¼ 0:008), which serves to illustrate the wider price variation for more expensive products so common in the market place. The present results corroborated the above predictions, for they suggest that both potential increases in product quality and potential decreases in price may influence search duration in a supermarket. The searching pattern of several customers observed in Experiment 1, characterized by smelling, indicates that search duration was influenced by potential increases in product quality. The larger search duration for products with higher base price, observed in both experiments, indicates that potential decreases in price also influenced search duration. However, as base price and price variation of the products investigated here were highly correlated, it would be necessary to test such predictions with products that have similar base prices and different price variations or different base prices and similar price variations. Results obtained here corroborate and expand data reported in previous studies using different methodologies. Evidence of extended pre-purchase search for products with higher base price can also be found in experimental investigations of price search in laboratory simulations (e.g., Darke & Freedman, 1993; Darke et al., 1995), in a survey based on questionnaire data about searching for Christmas gifts (Laroche

J.M. Oliveira-Castro / Journal of Economic Psychology 24 (2003) 637–652

651

et al., 2000), and in a survey, based on telephone interviews, about grocery shopping (Urbany et al., 1996). In these previous investigations, the main results were based on what individuals said they would do if they were really shopping or what they said they do when they go shopping. In both cases the results were based on what individuals say rather than on what they actually do. The present results suggest that consumers in fact extend search for more expensive items, although one would have to test if the same results are obtained for other types of search (e.g., searching for different stores or other kinds of products) and search measures (e.g., patterns instead of duration). The functional analysis advanced here may serve to illustrate the possibility of developing a coherent operant framework that explains and predicts consumer behavior, as defended by Foxall (1990, 1997, 1998, 2002a) and Foxall and Greenley (1998). This approach emphasizes the effects of situational variables rather than ‘‘intra-personal’’ factors as is usually the case with cognitively oriented theories (cf. Foxall, 1999). The results demonstrate systematic effects of base price on search duration despite the fact that they were obtained in a natural shopping environment, in which several variables remained uncontrolled. These orderly functional relations were observed independently of consumersÕ awareness of their behavior, since no record of consumersÕ verbal reports about their search patterns was obtained. Whether or not consumers are capable of accurately describing such differences in search duration of their own behavior is an empirical question that may be worth pursuing. The type of field study adopted here may be even used to investigate the level of correspondence between what consumers say about what they do and what they actually do. This kind of information has become particularly relevant in view of the fact that a great portion of research concerned with consumer behavior has been based on data from questionnaires and interviews, that is, on what consumers say about what they do. The present procedure may encourage the investigation of some aspects of what they actually do. Acknowledgement The author thanks Eileen P. Flores for helpful comments on previous versions of this manuscript. References Alhadeff, D. A. (1982). Microeconomics and human behavior: Toward a new synthesis of economics and psychology. Berkeley, CA: University of California Press. Blair, E. A., & Landon, E. L. (1981). The effects of reference prices in retail advertisements. Journal of Marketing, 45, 61–69. Darke, P. R., & Freedman, J. L. (1993). Deciding whether to seek a bargain: Effects of both amount and percentage off. Journal of Applied Psychology, 78, 960–965. Darke, P. R., Freedman, J. L., & Chaiken, S. (1995). Percentage discounts, initial price, and bargain hunting: A heuristic–systematic approach to price search behavior. Journal of Applied Psychology, 80, 580–586.

652

J.M. Oliveira-Castro / Journal of Economic Psychology 24 (2003) 637–652

Davies, J., Foxall, G. R., & Pallister, J. (2002). Beyond the intention-behaviour mythology: An integrated model of recycling. Marketing Theory, 2, 29–113. Della Bitta, A. J., Monroe, K. B., & McGinnis, J. M. (1981). Consumer perceptions of comparative price advertisements. Journal of Marketing Research, 18, 416–427. Foxall, G. R. (1990). Consumer psychology in behavioral perspective. New York: Routledge. Foxall, G. R. (1997). Marketing psychology: The paradigm in the wings. London: Macmillan. Foxall, G. R. (1998). Radical behaviorist interpretation: Generating and evaluating an account of consumer behavior. The Behavior Analyst, 21, 321–354. Foxall, G. R. (1999). The contextual stance. Philosophical Psychology, 12, 25–46. Foxall, G. R. (2002a). Consumer behaviour analysis: Critical perspectives. London and New York: Routledge. Foxall, G. R. (2002b). MarketingÕs attitude problem – and how to solve it. Journal of Customer Behaviour, 1, 19–48. Foxall, G. R., & Greenley, G. E. (1998). The affective structure of consumer situations. Environment and Behavior, 30, 781–798. Kahneman, D., & Tversky, A. (1979). Prospect theory: An analysis of decision under risk. Econometrica, 47, 263–291. Kahneman, D., & Tversky, A. (1984). Choices, values, and frames. American Psychologist, 39, 341–350. Kim, J., Lim, J. S., & Bhargava, M. (1998). The role of affect in attitude formation: A classical conditioning approach. Academy of Marketing Science, 26, 146–152. Laroche, M., Saad, G., Browne, E., Cleveland, M., & Kim, C. (2000). Determinants of in-store information search strategies pertaining to a Christmas gift purchase. Canadian Journal of Administrative Sciences, 17, 1–19. North, A. C., Hargreaves, D. J., & McKendrick, J. (1999). The influence of in-store music on wine selections. Journal of Applied Psychology, 84, 271–276. Oliveira-Castro, J. M., Coelho, D. S., & Oliveira-Castro, G. A. (1999). Decrease of precurrent behavior as training increases: Effects of task complexity. The Psychological Record, 49, 299–325. Oliveira-Castro, J. M., Faria, J. B., Dias, M. B., & Coelho, D. S. (2002). Effects of task complexity on learning to skip steps: An operant analysis. Behavioural processes, 59, 101–120. Polson, D. A., & Parsons, J. A. (1994). Precurrent contingencies: Behavior reinforced by altering reinforcement probability for other behavior. Journal of the Experimental Analysis of Behavior, 61, 427–439. Skinner, B. F. (1953). Science and human behavior. New York: The Macmillan Company. Skinner, B. F. (1957). Verbal behavior. Englewood Cliffs, NJ: Prentice Hall. Skinner, B. F. (1968). The technology of teaching. Englewood Cliffs, NJ: Prentice Hall. Skinner, B. F. (1969). Contingencies of reinforcement: A theoretical analysis. Englewood Cliffs, NJ: Prentice Hall. Skinner, B. F. (1977). Why I am not a cognitive psychologist. Behaviorism, 5, 1–10. Skinner, B. F. (1985). Cognitive science and behaviorism. British Journal of Psychology, 76, 291–301. Stigler, G. J. (1987). The theory of price. New York: Macmillan. Tversky, A., & Kahneman, D. (1981). The framing of decisions and the psychology of choice. Science, 211, 453–458. Thaler, R. (1985). Mental accounting and consumer choice. Marketing Science, 4, 199–214. Urbany, J. E. (1986). An experimental examination of the economics of information. Journal of Consumer Research, 13, 257–271. Urbany, J. E., Bearden, W. O., & Weilbaker, D. C. (1988). The effect of plausible and exaggerated reference prices on consumer perceptions and price search. Journal of Consumer Research, 15, 95–110. Urbany, J. E., Kalapurakal, R., & Dickson, P. R. (1996). Price search in the retail grocery market. Journal of Marketing, 60, 91–101. van Osselaer, S. M. J., & Janiszewski, C. (2001). Two ways of learning brand associations. Journal of Consumer Research, 28, 202–223. van Osselaer, S. M. J., & Alba, J. W. (2000). Consumer learning and brand equity. Journal of Consumer Research, 27, 1–16.