CHAPTER 2 REVIEW OF LITERATURE Reference price studies

CHAPTER 2
REVIEW OF LITERATURE
Reference price studies since the early 1990s (Mayhew and Winer, 1992; Rajendran and Tellis,
1994; Briesch, Krishnamurthi, Mazumdar and Raj, 1997) have focused primarily on
understanding which of the two reference price constructs dominate the consumer’s brand choice
decisions. Studies by Mayhew and Winer (1992) and Rajendran and Tellis (1994) were the first
step in this direction. Findings from these studies indicated that statistical models with reference
prices, which include both internal reference price (memory based) and external reference price
(stimulus / external information based) perform significantly better than models that include
only an IRP term or have no reference price term. However, the degree of usage of these two
types of reference prices may vary (Briesch, Krishnamurthi, Mazumdar and Raj, 1997). Briesch
et al. (1997) were of the view that if reference price was a continuum, then one end of the
continuum would be IRP while the other extreme would be ERP where the users require no
memory of past prices. Consumers in between these two extremes may use reference prices in
varying degree. Thus, consumers could be divided into three broad segments based on the type
of reference price they use- IRP users only, ERP users only and users of both IRP and ERP to a
varying degree. This segmentation of the consumers based on the type of reference price used is
of vital importance for both the marketers and the academicians. Two studies in behavioral
pricing (Mazumdar and Papatla, 2000; Moon et al. 2006) undertook the task of segmenting the
consumers depending on the type of reference price used by them. Findings from these studies
further suggest that both consumer and product characteristics have an impact on the type of
reference price used by the consumer in making brand choice decisions.
In the following sections, a thorough review of literature is presented wherein, firstly, the
definition of reference prices and its two conceptualizations are discussed. This is followed by a
section in which different models incorporating the IRP and ERP concepts, and their effect on
brand choice decisions is discussed. Next studies relating to segmentation of consumers based on
the type of reference price used is discussed. This is followed by two sections that discusse the
literature on antecedents of reference price usage and its consequences. Finally, the section on
Scale Development reviews the procedure of scale development in marketing research.
2.1 Research on Reference Prices
The seminal paper of Monroe (1973) is considered by academicians to have introduced the
concept of reference price in the marketing literature. Using the adaptation level theory (ALT) of
Helson (1964), Monroe in this conceptual paper defined reference price as a standard against
which the purchase price of a product is judged. The standard of the consumer is framed by
environmental factors like shelf prices and past prices encountered while dealing with products.
On similar note marketing researchers like Winer (1988), Rajendran and Telis (1994),
Janiszewski and Lichtenstein (1999) and Chandrasekharan (2012), have also considered
reference prices as a standard which may be a point estimate or a range with which observed
prices or actual prices are compared while making a purchase.
According to Winer (1988), reference price has multiple conceptualization like normal price, fair
or just price, last price paid, price observed in the market place and others. However, two
prominent conceptualizations of reference price, which have been extensively used in literature
are Internal Reference Price and External Reference Price. According to Winer (1988), internal
reference price is “a point on the internal judgment scale that is used as the standard to judge
offer prices.” The IRP is a memory-resident price formed by the repeated exposure of the
consumer to different or similar prices over the last few purchases. This price is used as a
standard to compare the observed prices. Since IRP is formed after some repeated exposure to a
product’s price some researchers (Mazumdar and Paptla, 1995; Mazumdar and Monroe, 1990)
also consider it to be a weighted average of past prices encountered. While the IRP is a memory
based price, External Reference Price is a stimulus based price (Mayhew and Winer, 1992;
Kopalle and Mullikin, 2003). The stimulus may be the current price of the brand last purchased,
the average price of different brands available on the shelf, retailer supplied comparative prices,
maximum retail prices printed on the packaging of the product and others or the lowest observed
shelf price in a product category. Table 2.1, 2.2 and 2.3 below present in brief some of the
definitions on reference prices, internal reference prices and external reference prices, used by
marketing researchers.
Table 2.1
Definitions of Reference Price
Year
Author/s
Reference price
1973
Monroe
standards against which the purchase price of a product is judged.
1988
Winer
price point on the subjective judgment scale above which all prices are
typically judged as high and below which they are judged as low.
1994
Rajendran and
Tellis
the standard price against which consumers evaluate the actual prices of
the products they are considering.
1995
Kalyanaram and
Winer
an internal standard against which observed prices are compared.
2012
Chandrashekaran
any price in relation to which other prices are seen.
Table 2.2
Definitions of Internal Reference Price (IRP)
Year
Author/s
IRP
a point on the internal judgment scale that is used as the standard
to judge offer prices.
1988
Winer
1992
Mayhew and
Winer
memory resident prices based on actual, fair, or other price
concepts.
1995
Kalyanaram and
Winer
past prices as part of the consumer's information set.
1998
Grewal et al.
a price (or price scale) in buyers' memories that serves as a basis
for judging or comparing actual prices.
2012
Chandrashekaran
price a consumer expects to pay.
Table 2.3
Definitions of External Reference Prices (ERP)
Year
Author/s
ERP
1992
Mayhew and Winer
observed regular prices printed on shelf labels.
1995
Mazumdar and
Papatla
current price of a previously chosen brand.
1998
Grewal et al.
observed stimuli in the purchase environment.
2003
Kopalle and
Mullikin
retailer-supplied comparative prices because they provide an
external reference against which an offered price can be judged.
2012
Chandrashekaran
price information for similar products may be used as an external
standard of comparison to assess the value of the target price.
2.2 Effect of Internal Reference Price and External Reference Price on Brand Choice
Decision
Two studies on reference price research one by Mayhew and Winer (1992) in the yogurt
category and the other by Rajendran and Tellis (1994) in the saltline cracker category have
contributed significantly in understanding the role that IRP and ERP together have in making
brand choice decisions. These two studies made a departure from the previously held notion that
consumers use only IRP to make brand choice while externally available prices are used only for
updation of internal standards. As a matter of fact, in both the studies it was observed that ERP
played a more dominant role in making brand choice than IRP.
Using scanner panel data from the yogurt category, Mayhew and Winer investigated the relative
impact of IRP and ERP on making brand choice decisions. Additionally, the authors also
investigated whether the presence of an ERP discrepancy (i.e. ERP- Actual sales price) or the
signal of a deal, had as much effect on purchase probabilities as the actual size of the
discrepancy. Two multinomial logit models (following McFadden, 1974) were used in the study.
Each model included the IRP term but to account for the ERP term one model used the ERP
discrepancy variable while the other used an indicator variable to identify whether there was a
special price. The two models are given in equation 2.1 and 2.2:
Equation 2.1
ERP Discrepancy Model
Equation 2.2
ERP Indicator Model
where
and
were the IRP terms. These terms captured the negative and
positive transaction utilities, when the store prices (
) were higher and lower than
(which was a proxy for the IRP term, i.e. price paid or charged the last time in a category when a
purchase was made);
for the brand (
= the discrepancy between the price currently being charged
and the regular price (or ERP, i.e.
) . This variable measured the
reference price effects of advertisements (that mentions both the regular price and the sale price)
as well as the effects of the in-store shelf tags (that mentions both the regular price and the sale
price);
was an indicator variable indicating whether there was a special reduced
price or not.
The yogurt category data (only single size 5-8 ounces, which accounted for 84 percent of the
sales dollar in the yogurt category) used in this study was provided by A.C. Nielsen Co. The data
used was for one city across 22 stores. From the data two samples were taken- households with
single person and households without any size restrictions. Moreover, in the case of single size
households only those households which purchased 50 percent of the modeled brands were
considered for the study. This resulted into a sample size of 52 households. In the case of
households with no restrictions only those households which purchased 80 percent of the
modeled brands were considered for the study. This resulted into a sample size of 185
households. Further to initialize the lagged choice and the IRP variables, 60 weeks was available
while the model calibration was over a period of 52 weeks. The analysis of the data revealed that
coefficients of IRP, i.e., the loss and the gain terms were not significant for the small households
but were significant for the large households’ sample. Secondly, when a model was fitted in the
small household without any ERP discrepancy or indicator term, there was a sizeable drop in the
log likelihood values ( -928 and -922 for ERP discrepancy model and ERP indicator model to 941). In contrast when the two IRP terms, i.e. the loss and the gain variables were dropped from
the two models, it resulted in a small decrease in the log likelihood of the small household from 928 and -922 to -930 and -924. Thus, ERP had a much greater effect on brand choices than IRP
had in the small household sample. No such significant effects were observed when the IRP and
ERP terms were dropped from the models of the large household. Finally, though the difference
in the log likelihood values (-928 vs. -922 in the small household; and -6467 vs. -6463 in the
large household) were very small in both the samples still they indicated that the ERP indicator
model fitted as well as the ERP discrepancy model. Hence, the authors concluded that for a
consumer, a signal of a price reduction in this product category was sufficient for making a
choice rather than the magnitude of price reduction. Findings from the study clearly indicated
that both memory based reference prices (IRP), and stimulus based reference prices (ERP) were
important for making brand choice decisions. However, ERP was a stronger determinant in
making brand choice decisions across the two samples. Furthermore, it was observed by the
authors that consumers reacted more to signals of savings like price deals contained in ERP and
less to the magnitude or the actual amount of savings.
The study of Mayhew and Winer (1992) was followed by a seminal study of Rajendran and
Tellis (1994). In this study, the authors investigated the strength of contextual reference price
(ERP) vis-a-vis temporal reference price (IRP) in making brand choice decisions. Another
objective of the authors was to investigate the importance of price cues that affected reference
prices. As regards to the IRP component the authors tested two measures or price cues - single
price for all brands based on some average of past prices, and a single price for each brand based
on some average of each brand’s past prices. For the ERP component, the authors tested three
measures- the highest or lowest or mean of the prices of all brands in the store at the purchase
occasion. Finally, the authors also tested the effect of consumer characteristics like brand
preference strength, brand sampling and frequency of purchases on the type of reference price
used. Following the standard brand choice modeling procedure (as recommended by Gudagni
and Little, 1983), the authors proposed a model where utility of a brand to a consumer at the
point of purchase was a function of the difference between the weighted average of IRP and the
actual price of the brand, the difference between ERP and the price of the brand, brand loyalty,
lagged choice, display, feature and coupon usage. The utility function was next estimated with
errors that followed a gumbel distribution. The authors used multinomial logistic to derive the
probability of brand choice. The authors next defined a base model which included all variables
except the price term. Additionally, to test the importance of reference price three other models
were tested - 1) the base model plus “price” alone, 2) the base model plus “temporal reference
price- price” 3) the base model with both “temporal reference price- price” plus “contextual
reference price- price.”
In this study, 16 oz size saltline crackers data for a period of two years provided by IRI was used.
The data covered three U.S cities and had a total of 42,180 purchases. The salted and unsalted
categories accounted for 85 % and 15% of the purchases respectively. A fourth market along
with the three cities was also considered. This market was a pool of the purchases in the unsalted
categories in all the three cities. Panelists were filtered based on a certain minimum number of
purchases during both the calibration and estimation period. The two year data was divided into a
calibration period of 35 weeks and an estimation period of 69 weeks. Data for the calibration
period was used to develop measures of loyalty and reference price. The authors defined brand
loyalty as “the share of a household's purchases of each brand in the calibration period of 35
weeks.” Further the weighted mean of prices of the past three purchase occasions during the
calibration period was used to measure IRP. The highest current price, the lowest, and the mean
of current prices of brands were used to measure the contextual reference prices.
An analysis of the data yielded the following results. Out of the three measures or price cues of
ERP, the lowest price of the brand at the point of purchase yielded the best result with the
highest t value and higher model chi square. According to the authors, this result was expected as
the most available and often advertised price was the lowest price. Secondly, of the two price
cues of temporal reference prices, the past price of each brand was always significant with the
right sign and higher log likelihood values. Thirdly, the temporal and contextual reference prices
across the four markets (three cities where purchases for the salted categories were considered
plus the pooled effect of the unsalted category across all the markets) were significant (t values
ranging between 2.4 to 11) and had the right sign. The results further substantiated the fact as
similar to the Mayhew and Winer (1992) study, that the model which included both the ERP and
IRP component had the highest log likelihood values across all the markets. As regards to the
effect of consumer characteristics and their effect on the type of reference price used, it was
observed by the authors that when brand preference strength was weaker, the ERP comparison
was higher. When the brand preference strength was stronger, the IRP comparison was stronger
in only two markets. As regards to the frequency of purchase the contextual component was
stronger for infrequent consumers in only two of the markets while the temporal component was
stronger for the frequent consumers in three markets.
The studies by Mayhew and Winer (1992) and Rajendran and Tellis (1994) led to a spurt in
reference price based marketing research using scanner panel data. However, there are two
challenges of using of using scanner data. Firstly, IRP is developed using a proxy and is not a
true reflection of the IRP which a consumer has in her memory. Secondly, the process of IRP
updation could not be understood. Grewal, Monroe and Krishnan (1998), came up with a
behavioral research based on experimental design to solve these problems. The authors
investigated the effects of advertised reference prices (ARP) and advertised selling prices (ASP)
on shoppers IRP (which was directly obtained from the consumer and was not a proxy),
transaction and acquisition value and willingness to buy and search. Specifically, the objectives
of this study were 1) to observe how consumers’ perceptions of value were influenced by price
comparison advertising and 2) to distinguish between acquisition value and transaction value. By
price comparison advertisements were meant those situations where the shopper compared the
advertised selling price with the advertised reference price. The advertised selling price was
always set by the retailer lower than the ARP to induce a sense of savings or value to the
consumer. The model tested by Grewal et al. is given in Figure 2.1 below:
Figure 2.1
Price Comparison Advetising on Perceptions of Value (Grewal et al. 1998)
In this study, a 2X2 between- subject experimental design was used with two ASPs ($249.95 and
$349.95) and two ARPs ($400 and $500). The advertisement used was on bicycles of a known
U.S.A brand. The ARPs were obtained by a market survey. In study 1, the respondents were 361
university undergraduate students with a mean age of 23 years. In Study 2, in order to assess the
replicability and boundary condition of Study 1’s results, 328 staffs of the university with a mean
age of 41 were used. Apart from ARP, ASP and IRP, all the constructs were measured using
scale items. IRP was operationalised by the following two questions- “What is your estimate of
the average market price of this bicycle?” and “what do you think will be a fair price for this
bicycle?” The data from the two studies were analyzed in two stages. The measurement model
was tested to obtain whether the scales were uni-dimensional and reliable. After reliability was
tested the structural model was tested to determine the strength of the individual relationships,
the model’s goodness of fit and the hypothesized paths. The outcome of the study was as
follows-
a) The advertised reference price and advertised selling price did not affect buyer’s
perception of quality significantly
b) The buyer’s internal reference price was a function of perceived quality, advertised
selling price and advertised reference price
c) Perceived acquisition value was a positive function of subject’s perception of quality.
d) In both the models Advertised selling price had a favorable influence on perceptions of
acquisition value
e) There was a significant negative relationship between the actual selling price and the
subjects’ perceptions of transaction value. Furthermore, there is a positive relationship
between the consumer’s IRP and perceptions of transaction value.
f) There was a significant positive relationship between consumer’s perceived acquisition
value and willingness to buy. There was a perceived negative relationship between
perceived acquisition value and search intentions in both the studies.
This study by Grewal, Monroe and Krishnan, however, was limited by the fact that external
reference price like advertisement reference price was used only to update the internal
reference price but was not directly used by the consumer to make a purchase decision. The
experiment undertaken was based on the belief that consumers use only IRP to make a
purchase. This may not always be true as is evident from the previous studies of Mayhew and
Winer (1992) and Rajendran and Tellis (1994) were consumers were shown to use both
external reference price and internal reference price while making a purchase decision.
2.3 Segmentation of Consumers on the basis of Internal Reference Price Usage and
External Reference Price Usage
Segmentation of the consumers on the basis of the type of reference price first appeared in the
study of Mazumdar and Papatla (2000) and Moon et al. (2006). The primary objective of the
study by Mazumdar and Papatla (2000) was to segment the consumers based on the type of the
reference price used in making brand choice decisions. The authors segmented the customers as
IRP using only, ERP using only and consumers using both IRP and ERP but to a varying degree.
Secondly, the authors also investigated the effects that ‘consumer characteristics’ like number of
brands purchased, sensitivity to in-store information and coupon usage and the ‘product
characteristics’ like price levels, promotional frequencies and purchase frequencies have on the
type of reference price used by the consumer. Assuming that there exists k=1,...,K segments in
the population the authors propose the utility function for a brand to a consumer which is given
in Equation 2.3 below:
Equation 2.3
Utility Function
The relative weights placed by a segment on the gains and the loss terms based on IRP was
The authors assumed that if
was smaller than 1 –
(i.e.,
< .5), consumers in segment k
used ERP to a larger extent than IRP in evaluating gains and losses. Conversely, if
greater than 1 –
(i.e.,
.
was
> .5), they relied more on IRP than ERP in assessing potential gains
and losses. Moreover, if
= 0, consumers in segment k only used information available
externally in evaluating prices. However, if
= 1, they only used previous purchase price
information. The authors developed the utility function to derive the probability of choice for a
brand given that a consumer was in a specific segment. For this, multinomial logistic function
was used. Segmentation of the consumers was done by using a latent class methodology which is
a type of clustering algorithm in which the likelihood function is used to determine which
household belongs in each group. The authors then used different response models in different
segments. The algorithm then fitted a household’s data to each of the segment models, and then
allocated the household to the segment whose model yielded the best fit. To test whether the
segmentation was done appropriately Bayesian analysis was done.
The authors next compared the proposed model with three benchmark models namely the no
reference price model, the IRP only model (where
). Also a fourth model was considered in which
) and an ERP only model (where
was constrained to be equal across all
segments. The authors wanted to compare the performance of their proposed model with the
fourth model to test whether the weights (
) which were ascribed to IRP with respect to ERP
does vary across segments. AC Nielsen scanner panel was used in this study and the product
categories involved were liquid detergent, ketchup, toilet tissue and yogurt. The scanner data was
available for a period of 138 weeks starting from January, 1986. Panel data for the first half of
the year 1986 was used as a calibration period while the rest was used for the estimation period.
Purchases from the calibration period were used to operationalize the reference price variables
and loyalty parameters. IRP for each brand was calculated by an exponentially smoothed
composite of the previous prices for the brand. ERP was operationalized as the loyalty-weighted
average of the current prices of all the brands. Across all the four product categories, some
filtration was done by the authors. This was necessary so as to incorporate only those consumers
who qualify a certain minimum to cut off regarding the number of purchases made in a product
category in both the calibration and the estimation period.
The analysis of the results revealed some interesting findings. The BIC of the proposed model
was the least and hence was better than the BIC of the other benchmark models in three out of
the four product categories. In the ketchup category, however, both the proposed model and the
no reference price model performed equally well. This meant that different segments of
consumers did use both the reference prices at least across the three product categories.
However, the degree of use of IRP and ERP varied depending on the emphasis placed on the two
reference price terms. Contrary to the studies of Rajendran and Tellis (1994) and Mayhew and
Winer (1992), in this study the IRP only model performed better than the ERP only model. The
parameter estimates of the proposed model obtained across the four product categories also
indicated on some interesting findings. In the liquid detergent category, four segments were
obtained of which the first segment had a lambda of .22 signifying that this segment was
dominated by the ERP users. The other three segments had lambdas greater than .5 signifying
that IRP users dominated these segments. When this result was combined with the IRP carryover
parameter of .53 (t= 13.54), it became evident that this category was dominated by the IRP users.
Nevertheless, a very high loyalty carryover parameter was obtained in the liquid detergent
category indicating that brand loyal customers dominated this category. For segment 1 in the
liquid detergent category, the gain coefficient was smaller than the loss coefficient implying that
this segment which relies on ERP was loss averse. The converse was true for the other three
segments. Similar analysis of the other three categories revealed that the toilet tissue category
was dominated by the ERP users as the IRP carry over parameter was only .22 (t= 7.73), while in
the Ketchup category, no such domination was established as the IRP carry over parameter was
.51(t=3.76). Finally, the yogurt category was dominated by the IRP users as the IRP carry over
parameter was .59 (t=25.71). The analysis of the data also revealed that across the four product
categories higher propensity to use IRP was associated with the purchases of fewer brands,
greater use of coupons and fewer purchases of a brand under store promotions. When the
consumers were ERP oriented, their proneness to display promotions increased. Similarly
Mazumdar and Papatla, tested for product characteristics like the average per unit price in the
category, category level display activity and average inter-purchase time in the category, on the
type of reference price used in a category. It was observed by the authors that in high-priced
categories like yogurt and detergent, IRP was more frequently used as these categories demand
higher attention and processing of information. Another interesting observation was that as the
frequency of in-store promotions increased the consumers used more of ERP and less of IRP.
Finally, in categories with shorter inter-purchase time, like yogurt category, IRP users dominated
the ERP users.
The study by Mazumdar and Papatla (2000) was significant because it not only showed which
product category was dominated by which type of reference price user, it also showed that every
category had a mix of different types of reference price users. Retailers should not only consider
the most dominant segment but as well should look into the functioning of the other segments in
a product category as some of the segments may be as good as the dominant segment. However,
this study also suffered from a lacuna similar to the previous studies using scanner panel data.
The IRP and ERP terms were operationalized using a proxy and not through direct measures.
2.4 Antecedents of Internal Reference Price Usage and External Reference Price Usage
The extant literature on reference price indicates that consumer characteristics like Perceived
Quality (Grewal, Monroe and Krishnan, 1998), type of learning (Mazumdar and Monroe, 1990),
involvement of the consumer (Chandrasekharan, 2012), number of brand purchased or brand
loyalty (Mazumdar and Papatla 1995, 2000) and certain product characteristics like promotion
frequency and purchase frequency (Mazumdar and Papatla, 2000; Monroe, 2003) influence the
usage of internal reference price and external reference prices. In reference price research, it
becomes vitally important to consider the heterogeneity of the consumer (Mazumdar et al. 2005)
as different consumers may have different characteristics based on which they may use IRP or
ERP or both or neither while making a purchase decision in a product category.
2.4.1 Intentional and Incidental Learning
Mazumdar and Monroe (1990) in a seminal study investigated the type of learning a consumer
undertook to encode prices. By price encoding, is meant how consumers meaningfully receive
and retain price information. Since price encoding could not be observed directly, the authors
used different price memory tests to estimate the actual encoding process. They conducted a
study (memory test) across 90 adult women shoppers who were randomly divided into two
groups with one group of the respondents being asked to memorize the prices of brands while the
other group not being given any such instructions. It was observed that intentional price learning
resulted from an active search and memorization of exact prices, typically for specific brands.
Certain consumers who were very careful shoppers and strived to minimize costs adopted such
behavior. Intentional learning included explicit comparison of current prices with previous prices
stored in memory. In contrast, the group which was not asked to memorize the exact prices used
incidental price learning where in consumers compared prices across brands during buying,
without any clear intention or effort to memorize them. These repetitive price comparisons over
time lead to low involvement learning, but such learning was likely to be of the relative price
rank rather than of specific prices.
Because retailers frequently reduced the price of one brand in a choice set, incidental learning
could soon become obsolete, requiring consumers to keep comparing the prices across brands.
According to Rajendran and Tellis (1994), Mazumdar, Raj and Sinha (2005) these two price
encoding processes are major determinants of IRP and ERP usage. Since the intentional learners
specifically remember the prices a consumer possessing such a capability will use IRP while an
incidental learner who is not in the habit of memorizing exact prices but is more inclined to
compare prices across brands will be more prone to ERP usage.
2.4. 2 Involvement with the Product Category
The heterogeneity of the consumers in terms of the involvement level they had with a product or
product category influenced the type of reference price being used (Chandrasekharan, 2012).
Chandreshekaran following from the research of Bettman (1979) argued that in contradiction to
low involved consumers; highly involved consumers were motivated to search for relevant
information. Such high involved consumers also possessed better product and price knowledge
and hence were more likely to possess well defined internal standards than low involvement
consumers. These high involvement consumers were thus more comfortable to evaluate retail
prices against mental standards. Less involved consumers were, however, not motivated enough
to engage in extensive search behavior and neither, they were involved in detailed processing of
available information. Consequently such consumers were not likely to be confident about their
mental standards. Nevertheless, it was inferred that level of involvement acted as a major
determinant of consumers’ utilization of reference prices. The author was of the view that high
involved customers were more likely to use a brand’s past price while low involved customers
were more prone to use a brand’s current price. Another aspect of the high involved customers
was that as they were exposed to market information all the time the high involved customers
may continuously update their internal standards. The low involvement customers, however, did
not put much effort on their cognitive resource for the use of long term memory for price
information storage, retrieval and subsequent use and hence had lesser belief and confidence in
their internal standards. Low involvement customers were thus more likely to dependent on the
market based point of purchase cues. From an experimental study conducted, over two weeks
period across 200 respondents in the jeans product category, the author observed that
respondents who had higher personal involvement inventory scale (developed by Zaichkowsky,
,1985 and used in the study) scores and thereby categorized to have higher involvement levels
were more familiar/ knowledgeable about the product than the low involvement groups. In fact
the high involvement respondents were using their IRP to evaluate the market prices while the
low involvement respondents were using both the normal prices and the lowest available prices
available in the environment for evaluation. Thus, high involvement customers were more likely
to use IRP and low involvement customers were more prone to use ERP.
2.4.3 Promotion Frequency
Mazumdar and Paptla (2000) in their study observed that categories which had higher frequency
of promotion were associated with a greater use of ERP than IRP. In this study, they considered
four product categories namely: yogurt, toilet tissue, ketchup and liquid detergent. The authors
observed that in categories like yogurt where promotion frequency was low, and these categories
were dominated by IRP users. On the contrary, in categories like tissue paper were promotion
frequency was higher, ERP users dominated this category.
2.4.4 Perceived Quality
Perceived Quality was defined as the “buyer’s estimate of a product’s cumulative excellence”
(Zeithaml, 1988).In the studies of Grewal, (1989) and Urbany and Bearden, (1990) buyer’s IRP
was influenced by cues like perceptions of product quality. In a study by Grewal et al. (1998) it
was observed by the authors that perceptions about the product quality helped the buyers to
develop internal reference prices. So if a product was of high quality, it was assumed that over
time, the consumers were likely to frame positive perceptions about it and were bound to
remember its price, and this eventually led to the usage of IRP. However, the literature is devoid
of any such relationship between perceived quality and ERP usage. .
2.5 Consequences of Internal Reference Price Usage and External Reference Price Usage
In the Grewal et al. (1998) study the major consequence of IRP was perceived transaction value
which subsequently affected the perceived acquisition value, which in turn affected the purchase
intention or willingness to buy and search intention of a consumer. Perceived transaction value is
the perception of psychological satisfaction or pleasure obtained from taking advantage of the
financial terms of a price deal. Grewal et al were of the view that a consumers’ always compares
the merit of a deal at the point of purchase with some standard. This standard of the consumer
may be a memory- resident price or price range or it may be an external stimulus like a price cue.
In their study the authors observed that IRP had a positive influence on perceived transaction
value. However, the authors in their model did not incorporate the effect of external reference
price on perceived transaction value. Grewal et al. were further of the view that perceived
transaction value had a positive influence on perceived acquisition value, which is the perception
of net gains that a consumer may derive from acquiring a product or a service. In addition,
perceived acquisition value was also affected by the perceived quality of a product or a service
since a consumer’s perception of net gains was also based on the product’s perceived cumulative
excellence. Thus, perceived acquisition value in the Grewal et al. study was a function of both
perceived transaction value and perceived quality.
2.6 Scale of Internal Reference Price Usage and External Reference Price Usage
Development of a measurement tool in behavioral pricing has repeatedly been emphasized by
marketing authors like Monroe (1973), Biersch et al. (1997) and Mazumdar et al. (2005). The
primary use of such a quantitative scale could be that it can be employed to collect data through
survey method and reveal information about the consumer characteristics: like the opinions,
behavior, attitude, feelings and beliefs (Daniel, 1979).
According to Monroe (2003), and Mazumdar et al. (2005) most of the studies of reference prices
were based on either panel data or data collected through experiments. Winer (1985), Rajendran
and Tellis (1994), Briesch et al. (1997), Kopalle and Mullikin (2003), Moon et al. (2006) and
other marketing researchers, have applied statistical models like Latent Class Analysis, Bayesian
Analysis and Multinomial Logistic on panel data to estimate the effect of IRP or ERP on
purchase decisions of a consumer. Marketing researchers like Mazumdar and Papatla (2000) and
Erdem et al. ( 2001) had in addition undertaken studies wherein the consumers were segmented
based on their IRP and ERP usage. These segmentation studies also considered panel data.
The primary disadvantage of panel data is the use of proxies while operationalizing variables like
IRP and ERP. Thus, a true estimate of IRP or ERP and their usage can never be obtained using
panel data (Mazumdar et al. 2005; Monroe, 2003). Another disadvantage of panel data is that
companies maintaining panel data of organized retail sector purchases like A.C. Nielsen, IRI etc.
provide such panels at an exorbitant cost. To avoid such a cost researcher generally considers
panels, which are of a much earlier period and at times outdated. Using such data marketing
researchers have predicted the brand choice of consumers based on their IRP or ERP usages.
However, in organized retail sector trends and fashion keep on changing and using panel data
with five to ten years of lag will never be able to depict the right picture (Monroe, 2003). In a
different direction researchers like Mazumdar and Monroe (1990), Grewal et al. (1998)
Mazumdar and Papatla (1995), Chandrashekaran and Grewal (2006), Chandrashekaran (2012)
and others have undertaken experiments in their studies involving reference prices. The primary
limitation of using experimental design is that it suffers from the issue of demand artifacts
(Monroe, 2003). Demand Artifact includes all aspects of the experiment which cause the subject
to perceive, interpret, and act upon what she believes is expected or desired of her by the
experimenter (Dickson and Sawyer, 1990). In the light of these issues related with using panel
data and experimental designs, researchers (Monroe, 2003 and Mazumdar et al. 2005) have
called for studies where scale/s of IRP and ERP usage could be used to collect data from fieldbased surveys. A direct measurement procedure like a field survey, where scales of IRP and
ERP usage can be administered, at the point of purchase can additionally be used for classifying,
and profiling of the consumers based upon the type of reference price used (Mazumdar and
Papatla, 2000 and Chandrashekaran, 2012).
2.6.1 Scale Development Procedure
McMillan and Schumacher (1984) were of the view that a researcher should always aim to
develop such a scale that “no other more valid and reliable technique could be used” to measure
the same construct which a researcher is trying to measure. One of the most widely cited and
seminal work on scale development was published by Churchill in 1979 in Journal of Marketing
Research. Churchill recommended seven stages of developing a scale. Figure 2.2 below gives a
brief description of the seven stages of scale development recommended by Churchill.
Figure 2
Scale Development Procedure (Churchill, 1979)
Step1: Specify Domain of Construct
Step 2: Generate Sample of Items
Step 3: Collect Data
Step 4: Purify Measure
Step 5: Collect Data
Step 6: Assess Reliability
Step 7: Assess Validity
Stage one of Churchill’s recommendation on scale development is ‘specifying the domain of
construct’ which is to clearly delineate the construct which a researcher wants to measure. The
second stage requires the researcher to ‘generate a pool of items’ using exploratory research
literature reviews, focus group discussions, in-depth interviews. The third stage involves
collecting data from a sample of respondents. The fourth stage of Churchill’s recommendation
involves, undertaking measure for purification of the scale. In the fifth stage, the researchers
need to collect data again to test the scale further. The final two steps require assessing reliability
and validity of the scale. Over the period of last three decades, it has been observed that most of
the scales developed based on the recommendations laid down by Churchill (1979), have
resulted into development of better measures in marketing research (Finna and Kayande, 2007).
2.6.2 Item Development
An important issue that researchers should invariably consider before developing a scale is that a
scale should consist of items, which should at least possess content validity (Schriescheim,
Powers, Scandura, Gardnier and Lankau, 1993). Thus, marketing scholars have consistently
highlighted on the significant role of item generation in scale development. Schwab’s (1980)
proposed that items should be generated using both review of literature and focus group
discussions. Once the items are generated, the next step involved the development of a
questionnaire and testing its adequacy (Churchill, 1979). Testing a questionnaire for adequacy
implies that it should be tested using a pilot study. According to Bork and Francis (1985), the
respondents of the pilot study should be as similar possible to the target respondents of the study.
Several other uses of pilot studies are: checking for the a) ease of handling the questionnaire in
the field b) the lucidity of the instructions c) the efficiency of the presentation d) the quality of
the questions and e) the time and cost of the project.
2.6.3 Measurement Theory
Precise measurement, which allows the hypotheses to be adequately tested, is an essential aspect
of scientific enquiry. According to Kline (1998) most of the measurement tools can be
categorized as either technically poor or efficient. A measurement tool like a scale, should
possess certain necessary characteristics like reliability and validity so that sufficient trust could
be placed on the results which a researcher gets after using the tool (Baumgartner and Jackson,
1982). In psychology literature such essential characteristics are labeled as measurement theory.
2.6.3.1 Reliability
The reliability of a scale is an essential characteristic which an instrument should have “so that
an instrument produces consistent results if repeated measurements are made” (Malhotra, 2011).
Different approaches to measure reliability are present like the test-retest, alternative forms and
others. However, in marketing and psychological research the most dominant and often the most
used reliability measure is the internal consistency reliability. According to Malhotra (2011)
“internal consistency reliability is used to assess the reliability of a summated scale where
several items are summed to form a total score.” By internal consistency is meant that all the
items in the scale are highly correlated implying that the same concept is being measured by all
the items together. Although marketing researchers have a number of ways through which
internal consistency could be measured, the most common technique used to measure internal
consistency reliability is Cronbach’s Alpha (Devillis, 1991). The Cronbach’s alpha can lie in
between 0 and 1. However, a scale with a value of .60 and below is generally regarded as not
being reliable.
2.6.3.2 Validity
Another important characteristic that a scale should possess is the validity of the scale.
According to Zikmund (2003) “validity is the accuracy of a measure or the extent to which a
score truthfully represents a concept.” Different approaches which may be used to establish the
validity of a scale are: Face Validity, Content Validity, Criterion Validity, Convergent Validity
and Discriminant Validity.
2.6.3.2.1 Face Validity
Face validity is present when there is a subjective concord among judges who may be
academicians or professionals, that an instrument coherently estimates the measured concept
(Zikmund, 2003). Are the items making sense given a definition of a concept? Is an important
question asked by the judges under face validity? Clearly understood questions in a
questionnaire indicate towards face validity of the scale.
2.6.3.2.2 Content Validity
An important distinction between content and face validity is that under face validity, items
appear to reflect what is being measured while under content validity, items are about what is
being measured. Similar to face validity, content validity is also determined by a group of judges
who analyze the content of each item in a scale and whether those items are exactly reflecting the
measured concept. Another interesting aspect of content validity is that a scale should only
reflect upon the measure concept and should not measure any other concept that it is not
supposed to measure (Zikmund, 2003; Tyler and Walsh, 1979).
2.6.3.2.3 Construct Validity
When a researcher wishes to establish, what a scale or an instrument is truly measuring? she is
looking for construct validity (Churchill, 1979). Of all the validity approaches construct validity
is the most important for establishing the truthfulness of a scale. Construct validity is categorized
into two convergent and discriminant validity. Convergent validity is defined as “the extent to
which the scale correlates positively with other measures of the same construct” (Malhotra,
2011). On the contrary, discriminant validity is the lack of correlations between distinct
constructs. An important statistic used to estimate convergent and discriminant validity of a
construct is the Average Variance Extracted (AVE), which indicates to the “amount of variance
explained by a construct relative to the amount of variance that may be attributed to
measurement error” (Fornell and Larcker, 1981). The value of AVE should always be greater
than .50 indicating that it is greater than the unique variance of a construct.