Download as doc, pdf, or txt
Download as doc, pdf, or txt
You are on page 1of 22

ESTIMATING THE EFFECT OF ODD PRICING

Philip Gendall, Michael F Fox and Priscilla Wilton


Department of Marketing, Massey University, New Zealand

KEY WORDS: odd pricing, choice modelling, demand curve

ABSTRACT

This study used choice modelling to estimate the effect of odd pricing on the demand for three products:

a $4 can of flyspray, a $7 block of cheese, and a $50 electric kettle, with each product represented by

three different brands. For cheese and flyspray a significant odd-price effect was observed at 99 cents

but not at 95 cents, whereas for the electric kettles a significant odd-price effect occurred at 95 cents but

not at 99 cents. The estimated value of the odd-pricing effect ranged from four cents for flyspray, to

six cents for cheese, to $3.18 for the kettles. The results of the study provide empirical support for the

assumption that odd pricing generates greater-than-expected demand, at least at the individual brand

level, and for the common practice of setting retail prices that end in 95 cents or 99 cents.

AUTOBIOGRAPHICAL NOTE:

Philip Gendall is Professor of Marketing, Michael Fox was a Lecturer and Priscilla Wilton a Graduate
Student, in the Department of Marketing, Massey University, New Zealand.

Contact Author:

Professor Philip Gendall Phone: 64 6350 5582


Department of Marketing Fax: 64 6350 2260
Massey University E-mail: P.Gendall@massey.ac.nz
Private Bag 11-222
Palmerston North
New Zealand

1
ESTIMATING THE EFFECT OF ODD PRICING

INTRODUCTION

The term "odd pricing" is used in several ways. It can refer to the practice of ending prices in odd

numbers (1, 3, 5, 7, 9); that of ending prices in a number other than zero; or that of pricing just below

a zero (for example, 99 cents, $4.99, or $19.95). The latter is the practice most commonly referred to

when the term is used and is the subject of the study reported here.

The notable characteristic of odd pricing is its prevalence in retailing. Surveys of the incidence of

various price endings show that odd prices, particularly those ending in the digit 9, represent between

80% and 90% of retail prices in countries where odd pricing is common (Friedman, 1967; Holdershaw,

Gendall and Garland, 1997; Kreul, 1982; Twedt, 1965). This prevalence of odd pricing has been

enough to convince some authors and many practitioners of its effectiveness. However, there is

relatively little reported evidence to support the assumption that odd prices produce higher-than-

expected demand.

This paper reports the results of a study designed to estimate the effect of odd pricing on respondents'

choice behaviour for three brands of three products: a $4 can of fly spray, a $7 block of cheese, and a

$50 electric kettle. The study involved estimating demand curves for each product category using

choice modelling and testing whether the predicted demand at prices that ended in 95 cents or 99 cents

was significantly higher than expected.

2
THE EFFECT OF ODD PRICING ON DEMAND

The assumption underlying the practice of odd pricing is that it creates greater-than-expected demand at

these prices. In other words, odd pricing is assumed to produce a "kink" in the expected demand curve

for the produce concerned (see Gendall, Holdershaw and Garland (1997) for a discussion of the origins

of odd pricing and explanations of its effect). Repeated calls for research to validate this assumption

have been made over the past three decades (Georgoff, 1971; Harper, 1966; Holloway, 1973; Twedt,

1965). However, the results of attempts to validate the assumption that odd prices lead to greater-than-

expected demand have been mixed and inconclusive.

Ginzberg (1936) reported the earliest documented study into the effect of odd pricing on demand. This

experiment was conducted by a large American mail order company to determine whether its customary

odd pricing policy was effective. To investigate this question, the sales responses to two versions of

the retailer's catalogue were compared. In one version selected items were offered at odd prices and in

the other at even prices. It was found that the even prices had halved the sales of some items, created

disproportionately large sales increases for others and left sales of the rest of the items unchanged.

Overall, the company judged that the losses were balanced by the gains. Unfortunately, Ginzberg's

one-page report did not describe these results in any more detail, nor did it provide any tests of

significance. Consequently, no generalisable results could be drawn from this study. A repetition may

have yielded more conclusive results, however, a $50,000 sales loss produced by a one cent increase in

the price of one item diminished the firm's interest in further testing.

A subsequent study by Georgoff (1971) proved just as inconclusive. Using a quasi-experimental

design in a field setting Georgoff examined ten products in a six-store chain of department stores.

Retail price endings were manipulated over a four-week period, alternating between 00-ending prices

and 99-ending prices, and the effect on sales monitored. The results showed only random variations

3
between price policy and sales, and no statistically significant sales differences between the two price-

ending conditions.

Dalrymple and Haines (1970) also attempted to model sales in a field setting (the junior sportswear

department of a Southern Californian department store). Dalrymple and Haines' model included a

variable indicating whether an item's price ended in the digits 00. The model produced a coefficient for

the price ending variable that was in the direction of a sales advantage for non-00 ending prices,

however, this coefficient was not statistically significant, nor were all the non-00 ending prices "odd".

Furthermore, the direction of the odd-pricing effect indicated by the model depended on how its

parameters were estimated. Log transformation of the data indicated that odd prices were associated

with increased product demand. But estimating the parameters on untransformed data indicated that

the odd pricing was associated with decreased product demand.

Blattberg and Wisniewski (1987) developed a model relating price and other factors to sales of

supermarket items. They discounted popular brands of products and advertised these in a supermarket

as weekly specials. Estimating the model's parameters on store-level scanner data for a large

supermarket chain, Wisniewski and Blattberg found that 13 of 21 brands tested showed substantially

higher sales when their prices ended in the digit 9, and that using a 9-ending price for promotions

generated an average 10% sales increase. However, because no attempt was made to examine

promotional and competitive activities alongside the increases in sales, these results cannot be

considered conclusive.

Little and Ginese (1987; cited in Stiving and Winer, 1977) and Stiving and Winer (1997) also analysed

scanner data to examine the effect of odd prices on the demand for low-priced supermarket items, using

logit regression models. Both studies found significant 9-ending effects on brand choice, but in Stiving

and Winer's study the effects were in different directions for the two products studied (tuna and

4
yoghurt).

In a field experiment in Germany, Diller and Brielmaier (1995) compared the sales of detergents and

health-care products in four pairs of matched drugstores. In the experimental stores, all odd prices in

the two product groups were rounded up in the nearest 10-pfennig for a period of four weeks, whereas

this was not done in the four control stores. A before and after comparison revealed a 6.8% increase in

the number of articles sold and a 4.8% increase in the value of sales in the even-priced stores. Diller

and Brielmaier also questioned customers about their price consideration, recall, and perceptions, and

their attitudes towards even prices and store image, obtaining responses which supported their

conclusions that consumers prefer even prices to odd prices and that German retailers should review

their odd-pricing policy.

However, Diller and Brielmaier's results imply upward sloping demand curves for detergents and

health-care products, or demand curves kinked upwards at even prices. The former seems very

unlikely, but since odd pricing is ubiquitous in Germany, it is possible that Diller and Brielmaier's

results reflect the novelty value of a short period of even prices in a predominantly odd-priced

environment. Nevertheless, the study does suggest that some of the odd pricing effects detected at the

brand level may disappear if all brands in a product category are odd-priced or even-priced.

Like Ginzberg forty years earlier, Schindler and Kibarian (1996) used a direct mail catalogue to test the

sales effect of odd prices. Three randomly-selected samples of 30,000 customers received one of three

versions of a winter clearance catalogue from a women's clothing retailer. The catalogues were

identical except for the selling prices, which ended in 00, 99 or 88, respectively, in each catalogue.

The 99-ending version of the catalogue produced 8% more sales revenue than both 00-ending or 88-

ending catalogues (which produced similar results). Not only did the 99-ending version generate more

5
purchasers than did the other catalogues but those who purchased from the 99-ending catalogue spent

more on their purchases. A similar study reported by Holdershaw (1995) produced a similar result.

Two samples of consumers were exposed to a mail order catalogue which differed only in the pricing of

the 42 products included. Half of the catalogues featured even-dollar prices, the other half showed the

same products at slightly cheaper odd prices with 95 cent price endings. The proportion of consumers

ordering was 9.8% from the sample which received the even-priced catalogue, but 10.3% from those

who received the odd-priced catalogue.

However, despite the fact that these were both well controlled experiments, we cannot assume that such

efforts will always occur. In Schindler and Kibarian's study each item's listing included a selling price

and a reference price, with emphasis on the fact that the selling price had been discounted (e.g., "was

$40 .... now only $23.99"). Given that 99-price endings may normally carry the connotations of a sale

or discount price (Kreul, 1982), this emphasis on price discounting may have enhanced consumers'

response to the 99-price endings. Furthermore, it has been claimed that women are more responsive to

odd-pricing than men (Georgoff, 1971), and given the nature of the catalogue used, it seems reasonable

to assume that most of the purchasers in Schindler and Kibarian's study were women. And, although

the study reported by Holdershaw supported Schindler and Kibarian's conclusions, the difference

observed was not significant and no attempt was made to compare the sales revenue generated by the

two catalogues tested.

An indirect experimental study conducted in England by Gabor and Granger (1964) attempted to assess

the efficacy of fixing prices just below certain found figures by estimating quasi-demand curves for two

types of products. For one type of product (Product A) the price setters of practically all the competing

brands followed a practice of setting a price one penny below a round figure. For the other type of

product (Product B), the competing brands were priced without regard to this principle. The

methodology involved calling out selected prices to subjects who responded by saying either "Buy",

6
"No, too expensive", or "No, too cheap."

The quasi-demand curves estimated for the two types of product were quite different. For Product A,

the curve was kinked upwards at prices which ended in 11 pence, whereas for Product B the percent

who would buy decreased roughly monotonically as price increased, giving a downward-sloping

"demand" curve. Gabor and Granger concluded that the odd price effects observed for products for

which prices were traditionally set at "penny-below" levels occurred because consumers had become

conditioned to expect these prices.

Lambert (1975) also used a laboratory method to estimate the monetary effect of odd and even price

endings. Respondents were asked to imagine they were participating in a television game show where

they had to choose between the chance of winning a set of products or the certainty of winning some

money. After respondents were shown the set of products (with either odd or even prices), they were

asked to specify the amount of money that would make them indifferent to the choice between the

certain money and the gamble.

Lambert hypothesised that respondents would specify smaller amounts of money as equivalent to odd-

priced product sets than to the even-priced ones. However, his results were inconclusive. For two of

the five product sets there was either no difference between the odd - and even-priced sets or the

difference was in the direction opposite to that expected. Lambert concluded that odd price effects

(which he called lower price illusions) may occur under certain conditions for some products, but not

for all products under all conditions.

Another indirect experimental attempt to estimate the effect of odd pricing was conducted by Dodds and

Monroe (1985), who investigated students' willingness to buy an FM stereo cassette player headset,

using a factorial design with three price levels and an even and 95-cent-ending odd price at each level.

7
They found no differences in willingness to buy between odd and even prices.

More recently, Gendall, Holdershaw and Garland (1997) tested the effect of odd prices on consumers'

purchase probabilities for a range of products at various price levels and at varying price points. The

products tested were a block of cheese, a frozen chicken, a box of chocolates, a hair dryer, an electric

kettle and a food blender. From purchase probabilities derived using the Juster scale, the demand at

each odd and even test price was determined for each product. For all six products, demand was higher

than expected at one or both of the odd price points tested. This effect was particularly marked for the

lower-priced food items (cheese, chicken and chocolates) and for prices ending in the digit 9.

However, the authors pointed out that respondents were presented with a single brand in several product

categories, whereas, in reality, consumers are confronted with several brands in each product category.

Together, the empirical studies reviewed above present an inconsistent and sometimes contradictory

picture. Some studies appear to provide evidence of odd pricing effects, while others do not. While

field experimentation in real-shopping situations allows for the manipulation of prices under actual

market conditions, it does not allow for complete control of external conditions. Thus environmental

influences on demand, such as competitive action and advertising, may have weakened or disguised odd

pricing effects in such studies. By contrast, while they may have been free from the uncontrollable

external factors which can confound field studies, laboratory studies inevitably suffer from the fact that

they use surrogate measures of purchase behaviour and less-than-realistic shopping situations. Thus,

despite more than 60 years of research, empirical support for the efficacy of odd-pricing is still

relatively weak.

The study reported here was a further attempt to estimate the effect of odd pricing. Though it involved

an indirect experimental approach, the use of choice modelling and the inclusion of three brands in each

of three product categories provided a more realistic surrogate purchase situation than in previous

8
laboratory studies.

METHODOLOGY

As discussed in the previous section, the assumption underlying the practice of odd pricing is that odd

prices will result in greater-than-expected demand, creating a "kink" in the traditional, downward-

sloping demand curve. The aim of this study was to test this assumption in a way that emulated a

realistic purchase situation but was not affected by uncontrollable external factors. To achieve this, a

stated-preference choice modelling experiment was conducted for three product categories; fly spray,

cheese and electric kettles. Each product was represented by three brands and five prices, including

prices ending in 95 cents and 99 cents. Demand curves estimated for each product were analysed to

determine whether higher demand than expected occurred at the odd price points tested, and, if so, the

extent and significance of this effect.

Three criteria were used to select the products tested. First, the products needed to have a reasonably

broad appeal to minimise the chance of a high proportion of respondents not purchasing the product

category. Second, there needed to be at least three comparable brands. And, finally, the products were

required to represent three different product categories: retail food items, retail non-food items and

consumer durables, each at a different price level.

The brands and price levels chosen are shown in Table I.

(Take in Table I about here)

For each product-brand combination five different prices were set: a high "anchor" price, a low "anchor"

price, and three "test" prices between these anchor prices. The test prices were set at .95, .99, and .00

9
price endings, and the anchor prices were chosen equidistant from each even (00-ending) test price. A

main effects, fractional factorial design was used to estimate linear and quadratic price effects, as well

as the effects of two "odd" prices (95 cents and 99 cents).

The choice sets constructed contained the three brands of each product, each at a specified price. The

same design, with 18 choice sets, was used for each product category, representing a total of 54 choice

sets [1]. To overcome the potential problem of respondent fatigue, the 54 choice sets were split into

three groups. Thus, three versions of the questionnaire were produced, with 18 choice sets in each; six

choice sets for each product, presented sequentially.

The sample for the study consisted of 300 household shoppers, selected by mall intercept in a

Palmerston North shopping centre in November 1996. Respondents were randomly allocated to one of

the three questionnaire versions and asked to view 18 showcards. On each showcard were three brands

of fly spray, cheese or electric kettle, each brand accompanied by one of the test or anchor prices. The

experimental design used meant that each test price-brand combination was presented to 200

respondents.

Respondents were told to assume they were shopping for the particular product shown and that the

brands on the showcard were the only ones available, and asked to choose one of the three brands. A

small number of respondents (three in total) declined to select a brand from one or more choice sets;

these observations were excluded from the analysis.

The effect of odd pricing on respondents' choice behaviour was calculated by estimating a multinomial

logit model.

THE MULTINOMIAL LOGIT MODEL

10
The multinomial logit model is one of several probabilistic, discrete choice models. The model

assumes that consumer choice depends on the relative utility of the alternatives offered and that the

utility of each alternative can be calculated by summing its attributes weighted by their multinomial

logit regression coefficients.

The MNL model takes the following form:


eßxi
P (ic) = _______
eßxj
j

where P(ic) is the probability of choosing alternative i in choice set C, ß is the vector of regression
coefficients, i is the vector of attribute descriptions for alternative i and j is the vector of attribute
descriptions for alternative j. For each of the alternatives in choice set C the utility, ßx i, is found. The
exponential of this utility is then divided by the sum of the exponentials, over all choice sets, to
estimate the choice probability for each alternative.

To estimate the model, respondents' choices were combined to produce aggregate frequencies for each
brand in each choice set. These aggregate frequencies were then regressed against the matrix of
attribute variables, which included brand, linear price, quadratic price, and the two odd prices (using a
maximum likelihood estimation procedure). Brands were entered into the analysis as dummy
variables. Dummy variables were also created to capture the effect of prices ending in 95 cents and 99
cents compared with those ending in 00 cents. The linear and quadratic price variables, which defined
the traditional downward-sloping demand curve, were specified as orthogonal polynomial codes [2].

RESULTS

All the models fitted were highly significant and explained between 58% and 74% of the variation in
the dependent variable (choice frequencies). For each of the three products, utility was higher than
expected at both of the odd-price points tested. For cheese and fly spray a significant odd-price effect
was observed at 99 cents but not at 95 cents, whereas for the electric kettles a significant odd-price
effect occurred at 95 cents but not at 99 cents (see Table II).

(Take in Table II about here.)

11
To illustrate the effect of odd pricing on demand (strictly speaking, on utility), "demand" curves were
estimated for each brand, using the regression coefficients estimated for the generic main effects model.
These demand, or utility, curves are shown in Figures 1, 2 and 3. As previously explained, the MNL
model assumes that choice probability is a function of the utility of the alternatives presented. The
utility of each alternative can be calculated as a linear combination of the alternative's attributes
weighted by their MNL regression coefficients. Thus, in this case:

Utility = ß1 (Brand1) + ß2 (Brand2) + ß3 (Linear price) + ß4 (Quadratic price)


+ ß5 (odd price 95 cents) + ß6 (odd price 99 cents)

(Take in Figures 1, 2 and 3 about here.)

The estimated demand curves are negatively-sloped and concave between the two anchor prices. The
fact that the shape of these curves is consistent with that traditionally assumed gives some reassurance
about the validity of the experimental method used (though, of course, the same conclusion could have
been reached by simply observing the signs of the linear price and quadratic price coeffients). The
effect of odd pricing is clearly shown in all three figures: the utility values at the odd price points are to
the right of the estimated demand curves, illustrating the upward "kink" expected if odd pricing had the
effect which is normally assumed.

The height of each demand curve reflects the relative market share of the brand concerned. Thus, for
flyspray, Blackflag is the preferred brand, followed by Raid, followed by Mortein. Discussion with the
manager of a local supermarket confirmed that actual sales of flyspray and cheese were consistent with
the rank ordering of brands shown in Figures 1 and 2.

The multinomial logit models estimated were also used to calculate the estimated value of the odd-
pricing effect; that is, the amount by which the product price would need to be reduced to achieve the
same level of demand as that produced by an odd price.

According to the multinomial logit model the probability of selecting an alternative P(ic) is directly
related to the utility of that alternative (Ui), and this utility is a linear combination of ß iXi. Thus, for a
given level of utility, a change in the level of one attribute can be compensated for by a change in the
level of another attribute. By removing the beta coefficients for prices ending in 99 cents and 95 cents
(setting them to zero) the price level which provides the same utility as these test prices was calculated

12
for each product. This price level was then subtracted from the odd price to determine the dollar value
of the odd pricing effect [3].

The results for each test price are presented in Table III.

(Take in Table III about here)

For effective odd prices (i.e., those with significant coefficients) the value of the odd-pricing effect
ranged from four cents for flyspray, to six cents for cheese, to $3.18 for the electric kettles. In
percentage terms, the size of these effects ranges from approximately 1% for flyspray and cheese to
6.4% for the kettles.

CONCLUSIONS

There is widespread use of and belief in odd pricing, and no shortage of anecdotal evidence to support
it. However, there is relatively little empirical evidence of its effectiveness. The results of this study
provide empirical support for the assumption that odd pricing generates greater-than-expected demand,
and for the common practice of setting retail prices that end in 95 cents or 99 cents. It confirms
Gendall et al's finding that the most effective odd price may depend on the price level concerned. For
low-priced, fast-moving consumer goods 99-cent price endings appear to be more effective in
generating demand than either even or 95-cent price endings. However, at $50, a 95-cent ending may
be more effective than a 99-cent-ending.

Previous studies have reported increases in sales revenue from odd pricing of between 8% and 10%
(Schindler and Kibarian, 1996; Blattberg and Wisniewski, 1987). For one of the products tested in
this study the odd pricing effect was of a similar order (6.4% for the electric kettles); but for the other
two products the effect was much smaller (around 1%). An explanation for the difference in effect
sizes in these studies is that in both of the other studies product prices were promoted to consumers.
Presumably this could have created greater price awareness and an enhanced price response.

Even if the smaller effects reported in this study are those which could be expected in the absence of
price promotions, they are nevertheless non-trivial from a managerial perspective. A 6% increase in
sales at a cost of a 0.1% reduction in price (i.e., $50 to $49.95) is clearly an attractive proposition.
Though less attractive, the equivalent relationships for flyspray and cheese also suggest that higher
revenue can be achieved by using 99-ending prices than slightly higher even prices.

13
However, these effects were estimated at the brand level, and may not be replicated at the product
category level. In other words, while odd pricing may increase demand for an individual brand, if all
the brands in a category are odd-priced, the overall effect may be weaker or even non-existent. This
would explain some of the apparently contradictory results in studies which have examined the
phenomenon of odd-pricing at different levels. (For example, it would explain why Schindler and
Kibarian (1996) and Holdershaw (1995) found increased sales as a result of odd pricing in direct-mail
catalogues, but Diller and Brielmaier (1995) found the opposite in German drugstores.)

In this study only a small range of products and price levels were tested. Thus the extent to which the
results can be generalised to other products and other price levels is uncertain. Furthermore, the study
analysed choice behaviour in an experimental setting rather than actual behaviour and assumed that
utility was a surrogate for demand. Finally, the study did not attempt to explain the phenomenon of
odd pricing (for discussion of this see Brenner and Brenner, 1982; Schindler and Kibarian, 1993 and
1996; and Schindler and Kirby, 1997). Greater understanding of the underlying process which produces
the odd pricing effect would help to determine the circumstances in which it is likely to be most
effective. Nevertheless, this study provides support for the assumption that odd pricing generates
greater-than-expected demand, at least at the individual brand level, and for the common practice of
setting retail prices that end in 95 cents or 99 cents.

NOTES

1. A complete factorial design involving three brands and five prices would require 5x5x5 = 125
choice sets. However, by using an appropriate experimental design the number of choice sets
which needs to be evaluated can be reduced to only 18. This "fractional factorial design"
ensures that each brand is presented at each price to the same number of respondents.

2. The formula used to calculate the linear price codes was as follows:

PL = 2*Actual Price - (Minimum Priced Tested + Maximum Price Tested)


(Maximum Price Tested - Minimum Price Tested)

This formula centres the prices tested and scales them in relation to the mean, producing codes ranging
from -1 to 1, including 0. The following table summarises the calculation of the linear price codes for
each of the three products investigated.

14
Product Price Calculation Linear Price Code
Fly Spray $3.90 2*3.90 - 8/0.2 -1
$3.95 2*3.95 - 8/0.2 -0.5
$3.99 2*3.99 - 8/0.2 -0.1
$4.00 2*4.00 - 8/0.2 0
$4.10 2*4.10 - 8/0.2 1

Cheese $6.80 2*6.80 - 14/0.4 -1


$6.95 2*6.95 - 14/0.4 -0.25
$6.99 2*6.99 - 14/0.4 -0.05
$7.00 2*7.00 - 14/0.4 0
$7.20 2*7.20 - 14/0.4 1

Kettles $45.50 2*45.50 - 100/9 -1


$49.95 2*49.95 - 100/9 -0.011
$49.99 2*49.99 - 100/9 -0.002
$50.00 2*50.00 - 100/9 0
$54.50 2*54.50 - 100/9 1

15
The formula used to calculate the quadratic price codes was as follows:

PQ = (PL*PL) *3 - 2

This formula squares, means centres, and then scales the linear price codes resulting in codes ranging
from -2 to 1. The following table summarises the calculation of the quadratic price codes.

Product Price Calculation Linear Price Code


Fly Spray -1 (-1*-1) *3-2 1
-0.5 (-0.5*-0.5) *3-2 -1.25
-0.1 (-0.1*-0.1) *3-2 -1.97
0 (0) *3-2 -2
1 (1*1) *3-2 1

Cheese -1 (-1*-1) *3-2 1


-0.25 (-0.25*-0.25) *3-2 -1.81
-0.05 (-0.05*-0.05) *3-2 -1.99
0 (0) *3-2 -2
1 (1*1) *3-2 1

Kettles -1 (-1*-1) *3-2 1


-0.011 (-0.011*-0.011) *3-2 -1.9996
-0.002 (-0.002*-0.002) *3-2 -1.9999
0 (0) *3 - 2 -2
1 (1*1) *3-2) 1

16
3. The easiest way to see how the odd pricing effect was estimated is to observe figures 1, 2 and 3.
In figure 1, for example, the point on the utility curve for each brand which is equal to the
utility at $3.95 is roughly equivalent to a price of $3.94. More precise estimates of the odd-
pricing effect can be calculated using the regression coefficients from Table II and the values of
the variables which define brand, price effect and the shape of the demand curve in the equation
on page 11. By substituting appropriate linear price and quadratic price values for different
prices, the utility which is equivalent to that for the test prices can be calculated.

17
REFERENCES

Blattberg, R.C. and Wisniewski, K.J. (1987), How retail price promotions work : Empirical results.
Marketing Working Paper No. 42, Graduate School of Business, Chicago, IL : University of
Chicago.

Brenner, G.A. and Brenner, R. (1982), Memory and markets, or why are you paying $2.99 for a
widget? Journal of Business, Vol. 55, No. 1, pp. 147-158.

Dalrymple, D.J. and Haynes, G.H. Jr. (1970), A study of the predictive ability of market period
demand - supply relations for a firm selling fashion products. Applied Economics, Vol. 1, No. 4,
pp. 277-285.

Diller, H. and Brielmaier, A. (1995), The impact of rounding-up odd prices : results of a field
experiment in German drugstores. Pricing Strategy & Practice, Vol. 3, No. 4, pp. 4-13.

Dodds, W.B. and Monroe, K.B. (1985), The effect of brand and price information on subjective
product evaluations. Advances in Consumer Research, Vol. 12, pp. 85-90.

Friedman, L. (1967), Psychological pricing in the food industry. In Phillips, A. and O.E. Williamson
(eds), Prices: Issues in theory, practice, and public policy, University of Pennsylvania Press,
Philadelphia, PA.

Gabor, A. and Granger, G.W.J. (1964), Price sensitivity of the consumer. Journal of Advertising
Research, Vol. 4, December, pp. 40-44.

Gendall, P., Holdershaw, J. and Garland, R. (1997), The effect of odd pricing on demand, European
Journal of Marketing, Vol. 31, No. 11 (in press).

Georgoff, D.M. (1971), Odd-even retail price endings. Michigan State University Press, Ann Arbor,
MI.

Ginzberg, E. (1936), Customary prices. American Economic Review, June 26, p. 296.

Guadagni, P.M. and Little, J.D.C. (1983), A logit model of brand choice calibrated on scanner data.
Marketing Science, Vol. 2, Summer, pp. 203-238.

Harper, D.V. (1966). Price policy and procedure. Harcourt, Brace and World, New York, NY.

Holdershaw, J. Gendall, P. and Garland, R. (1997), The widespread use of odd pricing in the retail
sector. Marketing Bulletin, Vol. 8, pp. 53-58.

Holdershaw, J.L. (1995), The validity of odd pricing. Unpublished MBS Thesis, Massey University,
Palmerston North, New Zealand.

Holloway, R.J. (1973), Odd-even price endings. Journal of Retailing, Vol. 49, Spring, pp 77-78.

Kreul, L.M. (1982), Magic numbers: Psychological aspects of menu pricing. Cornell Hotel and
Restaurant Administration Quarterly, Vol. 23, No. 1, pp. 70-75.

Lambert, Z.V. (1975), Perceived prices as related to odd and even price endings, Journal of Retailing,

18
Vol. 51, No. 3, pp. 13-22, 78.

Schindler, R.M. and Kibarian, T. (1993), Testing for perceptual underestimation of 9-ending prices.
Advances in Consumer Research, Vol. 15, pp. 348-352.

Schindler, R.M. and Kibarian, T. (1996), Increased consumer sales response through use of 99-ending
prices. Journal of Retailing, Vol. 72, No. 2, pp. 187-199.

Schindler, R.M. and Kirby, P.N. (1997), Patterns of rightmost digits used in advertised prices:
Implications for 9-ending effects. Journal of Consumer Research, Vol, 24, No. 2, (in press).

Stiving, M. and Winer, R.S. (1997), An empirical analysis of price endings using scanner data. Journal
of Consumer Research, Vol. 24, No. 1, (in press).

Twedt, D.W. (1965), Does the '9 fixation' in retail pricing really promote sales?, Journal of Marketing,
Vol. 29, No. 4, pp. 54-55.

19
Table I Product Attributes and Levels Tested

Attributes Levels

Fly Spray
Brand Blackflag Mortein Raid
Price ($) 3.90 3.95 3.99 4.00 4.10

Cheese
Brand Mainland Tararua Anchor
Price ($) 6.80 6.95 6.99 7.00 7.20

Electric Kettles
Brand Sunbeam Black'n'Decker Kambrook
Price ($) 45.50 49.95 49.99 50.00 54.50

Note: Within product categories, each brand is tested at each price level.

Table II MNL - regression results

Regression Coefficients and t values


_________________________________________________________
Variable1
Fly Spray Cheese Electric Kettle
___________ ___________ ___________

ß t ß t ß t

Brand 1 .260 4.71 .470 6.35 1.146 17.10


Brand 2 -.493 -7.22 1.177 17.12 .529 7.29
Linear price -.508 -13.24 -.720 -17.13 -.421 -10.78
Quadratic price .062 1.77 .060 1.71 .055 1.61
Odd price 95 cents .056 .49 .152 1.13 .382 2.93
Odd price 99 cents .251 2.06 .249 1.93 .053 .39

Likelihood ratio 354.9 686.3 488.6


Corrected 2 .59 .74 .67

Note 1. For fly spray: Brand 1 = Blackflag; Brand 2 = Mortein; For cheese: Brand 1 = Mainland;
Brand 2 = Tararua. For electric kettles: Brand 1 = Sunbeam; Brand 2 = Black'n'Decker.

20
2. -2 (L(0) - L(B)) with 6 d.f.

Table III Dollar value of odd pricing effects

Price with Odd Pricing Percentage


Equivalent Utility Effect Effect
to Odd Test Price

$ $ %
Fly Spray
$3.95 3.94 0.01 0.25
$3.99 3.95 0.04 1.00

Cheese
$6.95 6.91 0.04 0.58
$6.99 6.93 0.06 0.85

Kettles
$49.95 46.77 3.18 6.30
$49.99 49.45 0.54 1.10

21
Table 3 Generic main effects MNL estimates for cheese

Variable  s.e.  t
Mainland brand .470 .074 6.35
Tararua brand 1.177 .069 17.12
Linear price -.720 .042 -17.13
Quadratic price .060 .035 1.71
Odd price 95 cents .152 .134 1.13
Odd price 99 cents .249 .129 1.93

-2 (L(0) - L (B)) = 686.3; d.f.= 6

corrected 2 = .74

Table 4 Generic main effects MNL estimates for kettles

Variable  s.e.  t
Sunbeam brand 1.146 .067 17.10
Black 'n' Decker brand .529 .073 7.29
Linear price -.421 .039 -10.78
Quadratic price .055 .034 1.61
Odd price 95 cents .382 .130 2.93
Odd price 99 cents .053 .136 .39

-2 (L(0) - L (B)) = 488.6; d.f.= 6

corrected 2 = .67

22

You might also like