ArticlePDF Available

Penny Wise and Pound Foolish: The Left-Digit Effect in Price Cognition

Authors:

Abstract and Figures

Through five experiments, we provide a cognitive account of when and why nine-ending prices are perceived to be smaller than a price one cent higher. First, this occurs only when the leftmost digits of the prices differ (e.g., $2.99 vs. $3.00). Second, the left-digit effect also depends on the numerical and psychological distances between the target price and a competing product's price. The closer the two prices being compared, the more likely is the left-digit effect. Third, the left-digit effect is not restricted to the domain of prices; it also manifests with other multidigit numbers. (c) 2005 by JOURNAL OF CONSUMER RESEARCH, Inc..
Content may be subject to copyright.
54
2005 by JOURNAL OF CONSUMER RESEARCH, Inc. Vol. 32 June 2005
All rights reserved. 0093-5301/2005/3201-0005$10.00
Penny Wise and Pound Foolish: The Left-Digit
Effect in Price Cognition
MANOJ THOMAS
VICKI MORWITZ*
Through five experiments, we provide a cognitive account of when and why nine-
ending prices are perceived to be smaller than a price one cent higher. First,this
occurs only when the leftmost digits of the prices differ (e.g., $2.99 vs. $3.00).
Second, the left-digit effect also depends on the numerical andpsychological dis-
tances between the target price and a competing product’s price. The closer the
two prices being compared, the more likely is the left-digit effect. Third, the left-
digit effect is not restricted to the domain of prices; it also manifests with other
multidigit numbers.
Do consumers perceive a nine-ending price to be sig-
nificantly lower than a price one cent higher (e.g.,
$3.99 vs. $4.00)? This question has attracted researchers’
attention as early as 1932. Past research (Monroe 2003) and
conventional wisdom suggest that consumers do not respond
to very small price changes. Since nine endings change the
price of a product by just one cent (e.g., from $4.00 to
$3.99), several early researchers were skeptical about the
effects of nine endings on magnitude perceptions (Bader
and Weinland 1932; Gabor 1977; Gabor and Granger 1964;
Knauth 1949). However, recent research suggests that the
last digit of a price can have a significant impact on firms’
revenues (Anderson and Simester 2003; Blattberg and Nes-
lin 1990; see Monroe 2003 and Stiving and Winer 1997 for
a summary of research on nine-ending prices). One com-
monly cited explanation for the popularity of nine-ending
prices is that consumers underestimate the magnitude of
such prices. Although evidence gathered from econometric
analysis of UPC retail scanner data (Stiving and Winer1997)
and surveys of retailers’ pricing practices (Schindler and
Kirby 1997) support the underestimation hypothesis, ex-
perimental evidence has been elusive (Lambert 1975; Schin-
dler and Kibarian 1993). More important, it is not clear why
*Manoj Thomas is a doctoral candidate in marketing at the Leonard N.
Stern School of Business, New York University, 44 West4th Street, Room
9-175, New York, NY 10012 (mthomas@stern.nyu.edu). Vicki Morwitz
is professor of marketing and the Robert Stansky Faculty Research Fellow
at the Leonard N. Stern School of Business, New York University, 44 West
4th Street, Room 9-71, New York, NY 10012 (vmorwitz@stern.nyu.edu).
The authors gratefully acknowledge helpful suggestions on earlier versions
of the manuscript from Sucharita Chandran, Eric Greenleaf, Geeta Menon,
Robert Schindler, Russell Winer, the editors, and three reviewers.This work
has also benefited by comments from participants at the 2003 Association
of Consumer Research Conference, Fordham Pricing Conference, the Uni-
versity of Illinois Pricing Camp, and seminar participants at Columbia
University, INSEAD, London Business School and New York University.
nine endings affect a price’s perceived magnitude or what
factors moderate the effect (Monroe 2003; Monroe and Lee
1999).
In this article, we develop a conceptual framework that
draws on the analog model of numerical cognition (Adaval
and Monroe 2002; Dehaene, Dupoux, and Mehler 1990;
Hinrichs, Yurko, and Hu 1981; Monroe and Lee 1999) to
provide a cognitive account of why and when the perceived
magnitude of a nine-ending price is lower than a price one
cent higher. The results of five studies provide support for
this framework. We find that nine-ending pricesaffect mag-
nitude perceptions only in certain specifiable situations.
First, not all nine endings affect magnitude perceptions;they
affect magnitude perceptions only if the leftmost digit
changes. Second, we find that left-digit effects are more
likely to manifest when the internal discriminabilitybetween
the two numbers being compared is poor (i.e., when the
activated analog magnitudes are close together). Finally,
contrary to past suggestions (Gabor and Granger 1964), our
results suggest that these effects may not be limited to certain
types of prices or products. The effect of a left-digit change
on magnitude perception seems to be a consequence of the
way the human mind converts numerical symbols to analog
magnitudes on the mental scale. From a theoretical per-
spective, our results explicate how consumers encode and
compare multidigit numbers, in general, and prices, in par-
ticular. Our results also have implications for pricing practice
and public policy.
CONCEPTUAL BACKGROUND
Following Monroe and Lee (1999), we use the analog
model (also known as the holistic model) of numerical cog-
nition to explain how nine-ending prices are encoded and
evaluated. The analog model (Dehaene 1997; Hinrichs et
LEFT-DIGIT EFFECT IN PRICE COGNITION 55
al. 1981) suggests that, when presented with two multidigit
numbers to be compared, we assess the quantitative meaning
of the numbers by spontaneously mapping them onto an
internal analog magnitude scale. This numerical symbol to
magnitude conversion affects the precision of the numbers
being encoded (Dehaene 1997). Our basic proposition is
that the effect of price ending on magnitude perception oc-
curs during this conversion from numerical symbols to men-
tal magnitudes. Left-to-right processing of numerical sym-
bols affects this magnitude conversion process and distorts
the price magnitude toward the leftmost digit. We discuss
three effects that support this proposition: the left-digit ef-
fect, the distance effect, and domain invariance.
Left-Digit Effect
The left-digit effect refers to the observation that using
a nine ending versus a zero ending, for example, $2.99
versus $3.00, changes the leftmost digit (i.e., the dollar digit
changes from three to two) and that it is this change in the
left digit, rather than the one cent drop, that affects the
magnitude perception. The analog model suggests that per-
ceivers convert multidigit numbers into mental magnitudes.
1
We build on this model to argue that under specifiable con-
ditions, the leftmost digit can exert a relatively greater in-
fluence than the other digits on the encoded magnitude.
For example, consider a consumer who is comparing the
prices of two pens: a target pen, priced at $3.00, and another
pen, priced at $4.00. Although our interest is in the mag-
nitude perception of the target price, the comparison process
plays an important role. When presented with these prices,
this consumer automatically encodes them into mental mag-
nitudes on an internal analog scale. The $3.00 priceis likely
to be mapped onto the lower end of this scale while $4.00
will be mapped onto the relatively higher end of the scale.
How would the encoding process differ if the target pen,
instead of being priced at $3.00, were priced at $2.99? As
stated earlier, multidigit numbers are encoded holisticallyas
one analog representation (Dehaene 1997; Hinrichs et al.
1981; Monroe and Lee 1999). Even though we read three
separate digits in $2.99, these digits would be represented
as one analog quantity on the internal scale. However, be-
cause of left-to-right processing, the encoded magnitude of
$2.99 could, at least in some situations, be significantly
lower than that of $3.00. Note that this difference in mag-
nitude is because the target price dollar digit changes from
$3 to $2 and not because of the one cent price difference.
We do not predict a discernible change in magnitude per-
ception if the target price changes from $3.60 to $3.59 be-
cause in this case the leftmost digit remains the same.
One possible explanation for this left-digit effect is that
encoding the magnitude of a multidigit number begins even
before we finish reading all the digits. Dehaene (1997) pos-
1
The digital model (Poltrock and Schwartz 1984; Stiving and Winer
1997) suggests that multidigit numbers are compared digit by digit rather
than holistically. See Hinrichs et al. (1981) for a more detailed discussion
of analog versus digital models of multidigit numerical comparison.
tulated that the process of numerical symbol to magnitude
conversion occurs very rapidly and beyond consciousness.
Since we read numbers from left to right, while evaluating
“2.99,” the magnitude encoding process starts as soon as
our eyes encounter the digit “2.” Consequently, the encoded
magnitude of $2.99 gets anchored on the leftmost digit (i.e.,
$2) and becomes significantly lower than the encoded mag-
nitude of $3.00. It may be argued that the leftmost digit
exerts a primacy effect on magnitude encoding of multidigit
numbers. Thus:
H1: Nine-ending prices will be perceived to be smaller
than a price one cent higher if the leftmost digit
changes to a lower level (e.g., $3.00 to $2.99) but
not if the leftmost digit remains unchanged (e.g.,
$3.60 to $3.59).
Distance Effect
The left-digit effect does not always manifest. Perceivers
tend to anchor magnitudes of multidigit numbers on the left
digit only when the internal discriminability between the
two numbers being compared is poor. Price evaluation usu-
ally involves comparing two numerical stimuli, a target price
and a comparison standard (Adaval and Monroe 2002; Jan-
iszewski and Lichtenstein 1999; Niedrich, Sharma, and
Wedell 2001). Before two numbers can be compared, the
numerical symbols must be mapped onto the internal analog
scale. The closer the perceived distance between the two
analog magnitudes, the greater the difficulty in discrimi-
nating them on this scale. Consequently, the time required
for comparing them is greater. This phenomenon has been
labeled the “distance effect” (Moyer and Landauer 1967).
Hinrichs et al. (1981) showed that, when asked to judge
whether a given two digit number is higher or lower than
55, participants took significantly more time to judge num-
bers in the 40–70 range than in the 10–40 or 70–100 ranges.
Thus, the distance between 55 and the target number mod-
erated the cognitive difficulty in judging the magnitudes of
numbers being compared. The distance effect is a very ro-
bust phenomenon and has stood up very well to systematic
replication (Shepard and Podgorny 1978).
We draw on this research to propose that the perceived
distance between the two prices being compared will mod-
erate the left-digit effect. Our brain is more likely to use a
heuristic involving anchoring the magnitude on the leftmost
digit when the comparison process makes the magnitude
encoding task relatively difficult. When the magnitude en-
coding is relatively easy, then the left-digit effect should
diminish. So, the farther (closer) the two prices being com-
pared, the greater the ease (difficulty) in encoding the mag-
nitude of the nine-ending price. Consequently, the farther
(closer) the two prices, the less (greater) the distorting in-
fluence of the leftmost digit. Stated simply, $4 versus $5 is
not quite the same as $3.99 versus $5, but $4 versus $10
may not be perceptibly different from $3.99 versus $10.
Formally:
56 JOURNAL OF CONSUMER RESEARCH
TABLE 1
STIMULI USED IN STUDY 1A
Nine-ending
condition ($) Zero-ending
condition ($)
Pair 1 (nine-ending target has lower
left digit):
Ballpoint (target) 2.99 3.00
Ballpoint (standard) 4.00 4.00
Pair 2 (nine-ending target has same
left digit):
Fountain (target) 3.59 3.60
Fountain (standard) 4.00 4.00
H2: A left-digit change caused by a nine-ending price
is less (more) likely to affect the price’s magnitude
perception when the comparison standard is per-
ceived to be far away (close).
It needs to be underscored that our focus is on perceived or
psychological distance. The distance as perceived on the
internal analog scale, rather than the nominal distance,mod-
erates the left-digit effect. Sometimes, nominal distance may
not reflect the psychological distance. The experiments pre-
sented in this article examine the effect of nominal as well
as psychological distance. Further, in order to gain insight
into the underlying process, we also examine how distance
affects response latencies for numerical judgments.
Domain Invariance
Domain invariance refers to the property that the left-digit
and distance effects are not restricted to the domain of prices;
they also manifest with other multidigit numbers. Past re-
search has often attributed the popularity of nine-ending
prices to perpetuated retailing practices (Gabor 1977; Gabor
and Granger 1964; Schindler 1991). Based on a survey of
published material and informal conversations with consum-
ers and retailers, Schindler (1991) proposed a list offourteen
meanings that price endings are likely to communicate to
consumers (e.g., price-related meanings, such as “low
price,” “discount price,” or meanings concerning nonprice
attributes of the product or retailer, such as “low quality”).
If consumers’ favorable responses to nine-ending prices
are based solely on such images, then these effects should
be confined to the domain of prices. However, if theseeffects
are, at least partly, due to left-to-right processing,then these
effects should be invariant to domain. (Note, we are not
ruling out image effects; rather, we suggest that nine-ending
numbers can be perceived to be smaller than a number one
unit high even when image effects are absent.) Drawing on
the premise that the left-digit effect is a characteristic of the
multidigit encoding process, we predict that this effect and
its interaction with the distance effect will manifest for most
types of nine-ending numbers. Thus:
H3: Decreasing the distance between the numbers be-
ing compared will increase the left-digit effect not
only in the domain of prices but also in other types
of nine-ending numbers.
STUDY 1A: LEFT-DIGIT EFFECT
If the effect of nine endings on magnitude perceptions is
a consequence of the primacy of the leftmost digit, then it
should manifest only when the left digit changes. We test
this (hypothesis 1) using a between-subjects experimental
design in which we manipulated two orthogonal factors:
whether a target product’s price ending was nine or zero
and whether its dollar digit remained the same or was
changed by the price-ending manipulation. Participantswere
also introduced to a comparison standard, the price of a
comparable product, that remained unchanged across con-
ditions. Using a comparison standard was expected to ini-
tiate the number comparison process and thus facilitate the
conversion of these numerical symbols into magnitudes on
an internal analog scale. Further, it also ensured that partic-
ipants always evaluated the magnitude of the target price
with respect to a common reference point.
Method
Participants. Fifty-two undergraduate students from a
large northeastern university participated in the experiment
in return for partial course credit.
Design. This study employed a mixed factorial2#2
design; the effect of the nine ending (nine vs. zero) was
examined between subjects while the effect of the left digit
of the target price (same vs. different) was examined within
subjects. The stimuli for this study were pens, and each
participant saw prices for four pens: first two ballpoint pens
and then two fountain pens (see table 1). In each category,
one brand was the target and other the comparison standard
pen. Price endings for the target pens were manipulated to
either have a zero or a nine ending. Half the participants
saw target prices that ended in the digit nine ($2.99 and
$3.59) and half in the digit zero ($3.00 and $3.60). The
price of the target ballpoint pen was chosen such that the
price ending manipulation resulted in a nine-ending price
with a lower dollar-digit ($3.00 vs. $2.99), while that of the
target fountain pen was chosen such that the nine-ending
manipulation did not affect the dollar-digit ($3.60 vs. $3.59).
The comparison standards were always held constant at
$4.00. The dependent variable was the magnitude perception
of the target price. Since the comparison standard was held
constant across conditions, the target price manipulations
were not expected to have any effect on the magnitude
perceptions of the comparison standard.
Procedure. Participants were told that Aprilla and Av-
alon are two brands of pens being sold online and that they
should compare these brands within each product category.
Participants were given a booklet with advertisements for
all four pens (first for the two brands of ballpoint pens and
then for the two brands of fountain pens). The pictures of
LEFT-DIGIT EFFECT IN PRICE COGNITION 57
FIGURE 1
LOWERING THE PRICE BY ONE CENT TO A NINE-ENDING
PRICE AFFECTS MAGNITUDE PERCEPTION ONLY WHEN THE
LEFT DIGIT CHANGES
the pens were similar and the pen descriptions were short
and nondiagnostic (e.g., “Avalon ballpoint pen, black-laser
engraved, solid brass cap and barrel, state-of-the-art laser-
engraved logo, twist action mechanism, ink color: black”).
Details concerning the size and imprint area were also pro-
vided. Below each pen’s description was its price, including
delivery charges. The target pen prices were the only ele-
ments manipulated between conditions; all other information
remained the same.
Dependent Variable. Participants reported their price
magnitude perceptions for each brand on five-point Likert
scales with responses, “Strongly disagree” and1p5p
“Strongly agree.” Specifically, participants indicated the de-
gree to which they agreed or disagreed with the statement
“___ pen’s price is high” for each brand and type of pen
they reviewed.
Results
Target pen perceived price magnitude was submitted to
a mixed factorial ANOVA. Nine endings increased2#2
the difference in perceived price magnitude between the zero
and the nine-ending prices only when the dollar digit
changed, supporting hypothesis 1 (see fig. 1). The dollar
digit by price ending interaction was significant
( , , ). When the left digits
2
F(1,50) p4.27 p!.05 hp.08
differed (i.e., $3.00 vs. $2.99 for ballpoint pens), then the
mean magnitude perception was lower when the price had
a nine versus a zero ending ( vs. ;Mp2.76 Mp2.07
09
, , ). However, when the
2
F(1,50) p9.57 p!.01 hp.16
leftmost digits were the same (i.e., $3.59 vs. $3.60 for foun-
tain pens), then the effect of price ending on price magnitude
was not significant ( vs. , ). NoMp2.65 Mp2.61 F!1
09
other effect reached significance. As expected, magnitude
perceptions of the comparison standards, which remained
constant at $4.00 across conditions, were not affected by
the target price ending manipulation ( ).F!1
STUDY 1B
Method
Study 1B was similar to the previous study except for the
following. First, the numerical stimuli in the same-left-digit
conditions were changed, since in study 1a, the distance
between the target and the comparison was confounded with
whether the left digits differed. Second, we use a different
scale to measure perceived magnitude. Participants indicated
their magnitude perceptions for each brand by placing an
“X” on an uncalibrated 110 mm horizontal line anchored at
“low” and “high.” Responses were recorded by measuring
the distance from the left end of the line to the “X” using
a standard ruler calibrated in millimeters, and thus ranged
from zero to 110. Third, the same- and lower-left-digit prices
were counterbalanced across product types, and the order
of price presentation was also manipulated between subjects.
The study employed a mixed factorial2#2#2#3
design. Sixty-three undergraduate students from a large
northeastern university were randomly assigned to one of
the between-subjects conditions: target price ending (nine
vs. zero), product counterbalancing, and order of price ex-
posure. Participants were told that they have to compare the
brands Avalon and Aprilla in three different categories of
writing instruments: fountain pens, ballpoint pens, and pen-
cils. Target price level was manipulated within subjects at
three different levels ($3.20 vs. $3.00 vs. $2.80); these target
price levels were chosen such that the price ending manip-
ulation changed the leftmost dollar digit only when theprice
level was $3.00/$2.99; at the other two price levels, the
dollar digit remained unchanged in both price ending con-
ditions ($3.20/$3.19 and $2.80/$2.79). The comparison stan-
dard was $4.00 across conditions. Thus in each condition,
the participants saw six different prices (see table 2).
Results
Since the order manipulation and product counterbalanc-
ing effects were not significant ( ), the data were col-F!1
lapsed across these manipulations. Target pen magnitude
58 JOURNAL OF CONSUMER RESEARCH
TABLE 2
STIMULI USED IN STUDY 1B
Nine-ending
condition ($) Zero-ending
condition ($)
Pair 1 (nine-ending target has same
left digit):
Target 2.79 2.80
Standard 4.00 4.00
Pair 2 (nine-ending target has lower
left digit):
Target 2.99 3.00
Standard 4.00 4.00
Pair 3 (nine-ending target has same
left digit):
Target 3.19 3.20
Standard 4.00 4.00
perceptions were submitted to a mixed factorial3#2
ANOVA with target price levels ($3.20 vs. $3.00 vs. $2.80)
as the within subjects factor and target price ending (nine
vs. zero) as the between-subjects factor. There was a sig-
nificant target price level by target price ending interaction
( , , ). The results were con-
2
F(2,122) p5.29 p!.01 hp.08
sistent with the left-digit-effect hypothesis. When the target
level was $3.00 such that the nine-ending condition (i.e.,
$2.99) resulted in a lower dollar digit, then the nine-ending
price was perceived to be significantly lower (Mp55.8
0
vs. ; , , ).
2
Mp35.6 F(1,122) p20.92 p!.01 hp.15
9
However, when the target level was $2.80 such that the nine-
ending condition (i.e., $2.79) did not change the dollar digit,
then the nine-ending price had no effect ( vs.Mp44.4
0
; ). Similarly, when the target level wasMp42.8 F!1
9
$3.20 such that the nine-ending condition (i.e., $3.19) did
not change the dollar digit, then, again, the nine-ending price
had no effect ( vs. ; ). TheseMp43.7 Mp47.7 F!1
09
results are similar to those obtained in study 1a. The left-
digit manipulation had no effect on the magnitude percep-
tions for the comparison standards ( ).F!1
Discussion
Results from studies 1a and 1b support the left-digit effect
hypothesis. They show that lowering a price by one cent to
a 99 ending affects magnitude perceptions when the left
digit changes (e.g., $3.00 to $2.99) but does not affect mag-
nitude perceptions when the left digit is unchanged ($3.20
to $3.19 or $2.80 to $2.79). These studies, contrary to some
of the earlier views (e.g., Gabor 1977; Knauth 1949), pro-
vide experimental evidence that nine-ending prices are per-
ceived to be smaller than a price one cent higher. These
experimental results also corroborate Stiving and Winer’s
(1997) finding, using scanner data, that the left digit exerts
a stronger influence than the right digits in price evaluation.
Study 1b also showed that distance between target number
and the comparison standard has no effect on magnitude
perceptions when the left digit remains unchanged. In the
following study we test the effect of distance between two
numbers being compared when the left digit of one of the
two numbers changes.
STUDY 2: ANALOG MAPPING AND THE
DISTANCE EFFECT
The process of mapping from numerical symbols to men-
tal magnitudes imposes a cost on the speed of mental cal-
culations (Shepard and Podgorny 1978). The closer the num-
bers being compared, the greater the effort required for their
comparison. This ease of comparison manifests in the re-
sponse latency for these comparisons. Moyer and Landauer
(1967) measured the time participants took in comparing
two Arabic numbers and found that as the numerical distance
between them decreased, the response time for the com-
parison task increased, a phenomenon which has come to
be known as the distance effect. The distance effect has
been cited as evidence for holistic or analogical encoding
of numbers. Dehaene (1997, 76) wrote, “The only expla-
nation I can think (for the distance effect) is that our brain
apprehends a two-digit numeral as a whole and transforms
it mentally into an internal quantity or magnitude. At this
stage, it forgets about the precise digits that led to this quan-
tity.” The distance effect has been shown to be a robust
phenomenon not only in humans but also in chimpanzees
and pigeons. Further, this effect extends to multidigit nu-
merals, resists training and is present at 6 yrs. of age, the
earliest age at which it has been tested (cf. Dehaene 1997).
The distance effect suggests that encoding the magnitude
of a price is more cognitively taxing when an available
comparison standard is closer to the target price. The dis-
tance effect should then exacerbate the primacy effect of
left digits. The closer the prices being compared, the higher
the cognitive load, and therefore the greater would be the
error in encoding their magnitudes. This argument is also
consistent with the notion that under higher cognitive load,
individuals will be more likely to rely on a simplifying
heuristic for relative magnitude judgments. Study 2 tests
whether numerical distance moderates the underestimation
effect caused by a lower left digit (hypothesis 2).
Method
Design. This study employed a fully fac-2#2#2
torial design. Distance between the target and comparison
standard ($1 vs. $2), comparison standard level (higher vs.
lower) and the price ending of the target price (zero vs. nine)
were manipulated. The stimuli were a subset of those used
in studies 1a and 1b. Each participant saw two ballpoint
pens (see table 3). One pen served as the target while the
other served as the comparison standard. We manipulated
the target brand’s price ending ($3.99 or $4.00) and the
comparison standard’s price level ($2.00, $3.00, $5.00 or
$6.00). The comparison standards were selected such that
they were either $2 higher ($6) or lower ($2) or $1 higher
($5) or lower ($3) than the target price. This resulted in two
LEFT-DIGIT EFFECT IN PRICE COGNITION 59
TABLE 3
STIMULI USED IN STUDY 2
Target price higher
than standard ($) Target price lower
than standard ($)
$2 higher $1 higher $1 lower $2 lower
Zero-ending target
price conditions:
Target 4.00 4.00 4.00 4.00
Standard 2.00 3.00 5.00 6.00
Nine-ending target
price conditions:
Target 3.99 3.99 3.99 3.99
Standard 2.00 3.00 5.00 6.00
FIGURE 2
DISTANCE MODERATES THE EFFECT OF LEFT DIGIT ON
PRICE MAGNITUDE PERCEPTIONS
levels of distances between the comparison standards and
the target price ($1 or $2).
Procedure. One hundred and fifty-four undergraduate
students participated in this study. They were told toevaluate
two brands of pen sold by an online company. Participants
were given a booklet that showed advertisements for both
pens and the response scales. We used the same two fictitious
brands of ballpoint pens, Avalon and Aprilla, as in study 1.
Avalon served as the target brand and Aprilla as the com-
parison standard. The same dependent measure of perceived
price magnitude employed in study 1a was used in this study.
Results
Perceived price magnitude was subjected to three-way
ANOVA with comparison standard level, price ending, and
distance as between-subject factors. There was a main effect
of price ending ( , , ). For
2
F(1,145) p8.09 p!.01 hp.05
all levels of comparison standard, nine-ending target prices
were perceived to have lower magnitude than zero-ending
ones ( vs. ). This main effect was qual-Mp2.73 Mp3.24
90
ified by a significant price ending by distance interaction
( , , ), supporting hypoth-
2
F(1,145) p4.67 p!.05 hp.03
esis 2. The effect of price endings on perceived price mag-
nitude was greater when distance was small. When the dis-
tance between the target and the comparison standard was
$1, there was a significant difference in the magnitude per-
ceptions of nine- and zero-ending prices ( vs.Mp2.50
9
; , , ). How-
2
Mp3.39 F(1,145) p12.81 p!.01 hp.08
0
ever, there was no significant difference between these prices
when the distance was $2 ( vs. ;Mp2.96 Mp3.09
90
). There was also a main effect of comparison standardF!1
level such that the target price was perceived to be smaller
when it was lower than the comparison standard (Mp
low
vs. ; , ,
2
2.59 Mp3.38 F(1,145) p19.2 p!.01 hp
high
). No other effect reached significance. Wealso estimated.12
the mean perceived price magnitude separately for the four
levels of comparison standard (see fig. 2). The effect of
price ending was significant only when the target price was
$1 lower or higher than the comparison standard and not in
the other two conditions.
Discussion
The results of this study are consistent with the predictions
of the analog model and with our assertion that the under-
estimation caused by the left-digit effect occurs during the
magnitude encoding process. When the magnitude encoding
was made easier by increasing the numerical distance be-
tween the two prices, then the effect of left-digit change on
magnitude encoding was weakened.
STUDY 3: RESPONSE LATENCY
PATTERNS
This study was designed to (1) gain insight into the cog-
nitive encoding process underlying the left-digit effect using
response latencies and (2) test whether the left-digit effect
manifests in nonprice domains. Thus this study seeks sup-
port for both hypothesis 2 and hypothesis 3.
Since self-reports about encoding and processing of nu-
merical stimuli are not reliable, numerical cognition re-
searchers have traditionally relied on response time patterns
to make deductions about the underlying cognitive process
(Dehaene 1997; Hinrichs et al. 1981; Moyer and Landauer
1967). In this study we adopt the experimental paradigm
used by Hinrichs et al. (1981), with minor modifications.
Participants judged whether a given three-digit number, be-
tween 1.00 and 9.00, was lower or higher than 5.50. (Hin-
richs et al. 1981 used numbers between 10 and 100 with
55 as the comparison standard.) Drawing on past findings,
60 JOURNAL OF CONSUMER RESEARCH
we predicted that participants would take significantlymore
time to make magnitude judgments when the target number
was close to 5.50 than when it was farther away from 5.50.
More importantly, we also examined how the response times
varied for nine-ending numbers.
Method
Design. Sixteen numbers were chosen as target num-
bers, half with nine and half with zero endings. The chosen
target numbers were 1.99, 2.00, 2.99, 3.00, 3.99, 4.00, 4.99,
5.00, 5.99, 6.00, 6.99, 7.00, 7.99, 8.00, 8.99, and 9.00. These
numbers were symmetric around the comparison standard,
5.50, such that eight of the target numbers were lower and
eight were higher than it.
Participants. Fifty-three undergraduate students from
a large northeastern university, all with normal or corrected
vision, served as participants in partial fulfillment of course
requirements.
Procedure. Participants judged whether a target num-
ber presented on a computer screen was higher or lower
than the comparison standard 5.50. The comparison standard
was not presented on the screen; only the target numbers
were. Thus participants encoded each target number relative
to a memory based comparison standard. The target numbers
were displayed at the center of the screen. Below the target
number were two buttons labeled “higher” and “lower.”
The computer recorded the time participants took to click
on one of these buttons with a mouse after the target number
was flashed on the screen.
Participants were told “accuracy and speed are equally im-
portant.” Further, in order to ensure that they responded fast,
a small clock appeared at the bottom of the screen for each
trial. The clock completed one cycle in 10 sec.; if participants
did not respond in 10 sec., then they missed the chance to
respond to that number and the next screen was displayed.
In order to avoid demand effects, the 16 target numbers
were embedded in 16 filler numbers. The filler numbers were
three-digit numbers that had .25 or .75 endings (1.25, 1.75,
2.25, 2.75, 3.25, 3.75, etc.). To avoid order effects, eachblock
of numbers was presented twice to each participant, withtwo
different orderings. The computers were programmed to ran-
domize the sequence of numbers within each block and also
ensure that no subject saw the same number consecutively.
These three steps (i.e., inclusion of filler numbers, complete
randomization of presentation order, and recording two re-
sponses for each number) together ensured that no systematic
errors distorted the response latency pattern.
Before starting the trials, participants familiarized them-
selves with the task by responding to 20 practice trials. The
position of the response buttons were counterbalanced such
that half the subjects saw the “higher” button to the left
of “lower” button, and vice versa. To ensure that the mouse
position for the previous response did not influence response
times, after each trial a blank screen showed up with a
continue” button in the center of the screen.
Results
Participant’s response times for the target numbers were
submitted to an within subjects’ ANOVA.
2
The first8#2
factor, target level had eight levels: 2, 3, 4, 5, 6, 7, 8, and
9. The second factor was target number ending: zero versus
nine ending.
There was a main effect of target level (F(7,364) p
, , ). Participants took more time to
2
14.23 p!.01 hp.21
make comparisons when the target numbers were close to
the comparison standard. For numbers lower than 5.50, as
the magnitude increased toward 5.50, the response latency
(in milliseconds) also increased: ,Mp781 ms Mp
23
, , . A linear contrast of818 ms Mp838 ms Mp935 ms
45
these four means was significant confirming a systematic
pattern of increase in response latencies (F(1,364) p
, , ). For numbers higher than 5.50,
2
68.15 p!.01 hp.16
the response latency systematically decreased as the mag-
nitude increased away from 5.50: ,Mp888 ms Mp
67
, , . Again, a linear con-853 ms Mp793 ms Mp822 ms
89
trast of these four means was significant (F(1,364) p
, , ). These observations suggest that
2
11.37 p!.01 hp.03
the closer the number to the comparison standard, the greater
the difficulty in magnitude comparisons.
More interesting was the significant interaction between
distance and number ending ( , ,F(7,364) p4.36 p!.01
). The pattern of means supported the hypothesis
2
hp.08
that nine-ending numbers tend to affect response times only
when the distance between the target number and the com-
parison standard is small. First consider the numbers lower
than the comparison standard. When the magnitudes of the
target numbers were four or lower, then nine endings did
not affect response times ( ). However in the casep1.25
of 4.99 versus 5.00, a change in left digit significantly re-
duced the response time ( to ,Mp1,067 ms Mp903 ms
09
, , ). The response time
2
F(1,364) p22.80 p!.01 hp.06
for the nine-ending number was lower because its left digit
led to a perception that it was farther from the comparison
standard (see fig. 3). A similar pattern emerged fornumbers
higher than the comparison standard. When the magnitudes
of target numbers were seven or higher, then nine endings
did not affect response times ( ). However in the casep1.31
of 5.99 versus 6.00, response time was significantly higher
for the nine-ending number ( toMp853 ms Mp
09
, , , ). In this case,
2
923 ms F(1,364) p4.12 p!.05 hp.01
the response time for the nine-ending number was higher
because its left digit led to a perception that it was closer
to the comparison standard.
Discussion
The results of this experiment show that the left-digit
effect on response time is more likely to manifest when the
nine-ending numbers are close to the comparison standard.
This provides support for our proposition that nine endings
2
We repeated this analysis with log transforms of response times, and
obtained similar results (interaction p!.05).
LEFT-DIGIT EFFECT IN PRICE COGNITION 61
FIGURE 3
CHANGE IN LEFT DIGIT AFFECTS RESPONSE LATENCY
ONLY WHEN THE DISTANCE FROM COMPARISON
STANDARD IS SMALL
N
OTE
.—Thenine-ending numbers are 1.99, 2.99, 3.99,4.99,5.99,6.99,7.99,
and 8.99. The corresponding zero-ending numbers are 2.00, 3.00, 4.00, 5.00,
6.00, 7.00, 8.00, and 9.00. These numbers were compared with 5.50.
affect magnitude perceptions because of the manner in
which they are encoded. Further, these effects manifested
with numbers with no domain specification, suggesting that
the left-digit and distance effects are not restricted to prices.
STUDY 4: REFERENCE FRAMES AND
PSYCHOLOGICAL DISTANCE
The numerical cognition literature suggests that the psy-
chological distance between numbers affects how they are
processed. The psychological distance between numerical
stimuli depends on the reference frame (Janiszewski and
Lichtenstein 1999; Niedrich et al. 2001; Parducci 1965; Rog-
geveen and Johar 2004). Volkmann (1951) suggested that
it is primarily the endpoints of the stimulus range that control
perceptions of magnitude and distance. The intermediate
points in the stimulus range are judged relative to these
endpoints. Sherif and Hovland (1961) also suggested that
when no explicit standard is introduced within a series of
stimuli then the endpoints are used as standards for judg-
ment. For instance, in the absence of accessible internal
standards, whether the number five is perceived to be high
or low will depend on whether the stimulus range is 0–6 or
4–10 (Lynch, Chakravarti, and Mitra 1991). It follows that
the psychological distance between stimuli will also depend
on the endpoints of the stimulus range.
The idea that reference frame can manipulate psycholog-
ical distance motivates two interesting propositions. First,
making a stimulus the upper endpoint of a series will cause
its perceived magnitude to be higher (than its magnitude in
the absence of the frame). For instance, when consumers
are comparing two products with quality ratings (QR) 2.99
and 3.50, introducing a product with a 3.25 QR will increase
the perceived distance between 2.99 and 3.50. Since con-
sumers tend to use the end stimuli as standards, they will
map 2.99 as the lowest and 3.50 as the highest standard on
their internal analog scale while making product quality
judgments (Sherif and Hovland 1961; Volkmann 1951). Sec-
ond, adding an upper endpoint to a series will cause the
perceived magnitude of the internal points to be lower (than
their perceived magnitude in the absence of an upper end-
point). For instance, when consumers compare two products
with QRs of 2.99 and 9.25, introducing a product with a
9.50 QR can decrease the perceived distance between 2.99
and 9.25. Thus, a small number can be framed as relatively
large by presenting that number as the highest endpoint in
the range, and a large number can be framed as relatively
small by introducing a larger endpoint in the range.
In the following experiment, we manipulate framing to
examine the effect of psychological distance (hypothesis2).
Further, we examine these effects in yet another non-price
domain, namely, product quality ratings (hypothesis 3).
Method
Design. Three factors—nine endings in QR ratings
(nine vs. zero), numerical distance (low vs. high), and psy-
chological distance (low vs. high)—were manipulated
within subjects. Participants saw quality ratings of three
different brands in each of four different product categories.
Product category presentation order was manipulated be-
tween subjects. In two product categories (web cameras and
refrigerators) the numerical distance between the quality rat-
ings was high, approximately six (2.99/3.00 to 9.50). In the
other two product categories (digital cameras and air con-
ditioners), the numerical distance between quality ratings
was low, approximately 0.50 (2.99/3.00 to 3.50). Within
each product category, the first and third brands (as shown
in table 4) served as comparison standards, while the second
brand served as the target brand. In each product category,
participants compared the target brand QR with both com-
parison brands. Note that, in all four product categories, the
first and the second brands had the highest and the lowest
values in the product category, respectively, and thus would
serve as endpoints of the internal analog scale for the cat-
egory (see table 4). The second (i.e., target) brand had either
a zero or nine-ending QR (3.00 or 2.99). Since the first and
second brands’ QRs served as endpoints of the internal mag-
nitude scale for that category, the psychological distance
between them was always higher than the psychological
distance between the second brand and the third brand. The
analog model suggests that QR comparisons will happen on
the internal magnitude scale relevant to that product cate-
62 JOURNAL OF CONSUMER RESEARCH
TABLE 4
STIMULI USED IN STUDY 4
Product category Comparison standard 1 Target brand Comparison standard 2
High numerical distance comparisons:
Web cameras Brand X p9.50 Brand Y p3.00 Brand Z p9.25
Refrigerators Brand A p9.50 Brand B p2.99 Brand C p9.25
Low numerical distance comparisons:
Digital cameras Brand X p3.50 Brand Y p3.00 Brand Z p3.25
Air conditioners Brand A p3.50 Brand B p2.99 Brand C p3.25
N
OTE
.—Comparison standard 1 1comparison standard 2 1target brand. Since comparison standard 1 and the target brand were the endpoints of the stimulus
range in each product category, these comparisons served as the high psychological distance comparisons. Comparisons between comparison standard2 andthe
target brand were the low psychological distance comparisons.
gory. Therefore, drawing on hypothesis 2 we predicted that
psychological distance would moderate the effect of nine
endings on magnitude comparisons of QRs.
Procedure. Twenty-seven undergraduate students par-
ticipated in the experiment in return for partial coursecredit.
The stimuli were presented in a booklet. Quality ratings and
dependent measures for each category were presented on
separate pages, to ensure participants used the relevant ref-
erence frame for each product category. Participants, who
were randomly assigned to first see the low numerical dis-
tance categories, began by examining the quality ratings for
digital cameras that had a zero-ending target brand QR. The
quality ratings for the three brands were presented on a
single line and in the same order as shown in table 4. The
dependent variables were the perceived differences between
the target and the two comparison standard brands. After
responding to the quality evaluation questions for digital
cameras, participants turned to the next page to see the
quality ratings for air conditioners that had a nine-ending
target brand QR. Next they saw the two categories with high
numerical distances. The remaining participants who were
first exposed to high numerical distance categories sawqual-
ity ratings for web cameras and refrigerators and then for
the other two categories.
Dependent Variables. For each product category, par-
ticipants reported two dependent measures: the perceived
QR differences between the target brand and both compar-
ison standards. Participants’ responses were recorded on
seven-point semantic differential scales anchored at “low”
and “high” in response to the statement: “The difference
between Brand X’s (Z’s) and Brand Y’s Quality Ratings is
___.”
Results
The perceived difference perceptions were subjected to a
mixed factorial ANOVA with QR ending2#2#2#2
(nine vs. zero), numerical distance (low vs. high) and psy-
chological distance (low vs. high) as within subject factors
and product presentation order as a between-subjects factor.
Since the main effect of order, and the order withQR ending
interaction were not significant, we collapsed across pre-
sentation order. As expected, the psychological distance by
QR ending interaction was significant ( ,
F(1,26) p15.04
, ). When the psychological distance was
2
p!.01 hp.37
low, then a nine ending in the target QR significantly in-
creased the difference perception ( vs.
Mp4.33 Mp
09
, , , ). However, when
2
4.85 F(1,26) p17.96 p!.01 hp.41
the psychological distance was high, then a nine ending in
the target QR did not affect the difference perception
( vs. , ).
Mp5.48 Mp5.46 F!1
09
We analyzed the moderating effect of psychological dis-
tance on the underestimation of nine-ending numbers sep-
arately for high and low numerical distance. For low nu-
merical distance, when the comparison standard level was
3.25, then a change in target QR from 3.00 to 2.99 caused
a significant change in the perceived difference between the
target and the comparison standard QR ( vs.
Mp2.70
0
; , , ). However,
2
Mp3.37 F(1,26) p14.84 p!.01 hp.36
9
when the comparison standard level was 3.50, the effect of
nine ending was not significant ( vs.
Mp4.22 Mp
09
, ). Similarly, for high numerical distance, when
4.25 F!1
the comparison standard level was 9.50, the effect of left-
digit change was not significant ( vs.
Mp6.74 Mp
09
, ). However, when the comparison standard level
6.66 F!1
decreased to 9.25, then a change in target QR from 3.00 to
2.99 caused a significant change in the perceived difference
between the target and comparison standard QR (Mp
0
vs. ; , , ).
2
5.96 Mp6.33 F(1,26) p4.58 p!.05 hp.15
9
These results support hypothesis 2 and confirm that psy-
chological distance moderated the left-digit effect.
There were also main effects of numerical distance
( vs. ; ,Mp3.63 Mp6.42 F(1,26) p125.82 p!
low high
, ), and psychological distance ( vs.
2
.01 hp.83 Mp4.59
low
; , , ). Further
2
Mp5.47 F(1,26) p28.33 p!.01 hp.52
high
the interaction between numerical distance and psycholog-
ical distance was significant ( , ,F(1, 26) p6.57 p!.01
). When the numerical distance was high and the
2
hp.20
comparison standard’s QR increased from 9.25 to 9.50, the
perceived difference between the comparison standard and
target brand QRs increased ( vs. ;Mp6.14 Mp6.70
low high
, , ). When numerical dis-
2
F(1,26) p20.62 p!.01 hp.44
tance was low, then an increase in the comparisonstandard’s
QR from 3.25 to 3.50 had a stronger effect (Mp3.04
low
vs. ; , , ).
2
Mp4.24 F(1, 26) p96.78 p!.01 hp.79
high
LEFT-DIGIT EFFECT IN PRICE COGNITION 63
These results served as manipulation checks to suggest that
both numerical distance and psychological distance affected
perceived distance. However, since quality comparisons
were done within the reference frame relevant to eachprod-
uct category, only psychological distance (manipulated
within reference frames) moderated nine-ending effects,
while numerical distance did not.
Discussion
These results have several important implications. First,
they add to the empirical observations from study 3 to show
that the left-digit effect is not restricted to the domain of
prices. Second, these results suggest that the distance effect
should be interpreted cautiously. In the context of left-digit
effects, the psychological distance as perceived on the in-
ternal analog scale is of greater relevance than objective
numerical distance.
GENERAL DISCUSSION
This research adds further evidence to the view echoed
by previous researchers (Blattberg and Neslin 1990; Monroe
2003; Stiving and Winer 1997) that the decision whether or
not to use nine-ending prices is an important one and de-
serves due attention. Importantly, we show that nine-ending
prices may sometimes but not always be perceived to be
lower than a price one cent higher. This perception is more
likely to occur when introducing a nine ending in the price
causes a change in the leftmost digit. Further,this perception
is more likely when the nine-ending price is perceived to
be close to the comparison standard price. Our studiesshow
left-digit effects manifest in the domain of quality ratings
and in the domain of unspecified general numbers. Thus
there seems to be a domain invariant cognitive phenomenon
behind the popularity of nine-ending prices.
A research question that remains unanswered is whether
the primacy effect of left digits will manifest whenthe right-
digits are not 99. Our studies examined only numbers that
ended with 99. Dehaene et al. (1990) found that repetition
of a digit in a number influenced the number comparison
process. Therefore, it is possible that the processing of num-
bers that end in 99 differs from numbers that end in 98, 96,
95 or other digits. Thus, a potential research question
emerges: will numbers such as 3.95 and 3.90 also be un-
derestimated in the same way as 3.99? A related question
is whether digits other than the leftmost in a multidigit num-
ber can influence that number’s magnitude perception. In
the studies examined in this research, there was only one
digit to the left of the decimal point. In a pricing context,
when there are two or more digits to the left of the decimal
point, a nine-ending that changes the dollar digit, may or
may not also change the 10’s digit (e.g., $19.99 vs. $20.00
or $22.99 vs. $21.99). Future research should examine
whether there are effects associated with such internal left-
digit changes.
Following the approach suggested by Monroe and Lee
(1999), we based our hypotheses on the analog model of
multidigit number cognition. Our findings add to the evi-
dence accumulating in favor of the analog model. However,
the objective of this article was more to examine cognitive
phenomena associated with nine-ending prices rather than
to defend the analog model. Several other models of nu-
merical cognition such as the digital model (Poltrock and
Schwartz 1984; Stiving and Winer 1997) and the semantic
coding model (Banks 1977) have been proposed. Some of
these models also can predict and explain the empiricalphe-
nomenon presented in this article, although many research-
ers (Dehaene 1997) believe that the analog model postulates
the most parsimonious explanation for the distance effect.
Which of these models offer the most convincing account
for the left-digit effect, the distance effect, and other effects
in price cognition is a question worthy of future research.
[Dawn Iacobucci served as editor and Kent Monroe
served as associate editor for this article.]
REFERENCES
Adaval, Rashmi and Kent B. Monroe (2002), “Automatic Con-
struction and Use of Contextual Information for Product and
Price Evaluations,” Journal of Consumer Research,28
(March), 572–87.
Anderson, Eric and Duncan Simester (2003), “Effects of $9 Price
Endings on Retail Sales: Evidence from Field Experiments,”
Quantitative Marketing and Economics, 1 (1), 93–110.
Bader, Louis and James D. Weinland (1932), “Do Odd PricesEarn
Money?” Journal of Retailing, 8 (January), 102–4.
Banks, William P. (1977), “Encoding and Processing of Symbolic
Information in Comparative Judgments,” in The Psychology
of Learning and Motivation, Vol. 11, ed. G. H. Bower, New
York: Academic, 101–59.
Blattberg, Robert C. and Scott A. Neslin (1990), Sales Promotion:
Concepts, Methods and Strategies, Englewood Cliffs, NJ:
Prentice Hall, 349–50.
Dehaene, Stanislas (1997), The Number Sense, New York: Oxford.
Dehaene, Stanislas, Emmanuel Dupoux, and Jacques Mehler
(1990), “Is Numerical Comparison Digital? Analog and Sym-
bolic Effects in Two-Digit Number Comparison,” Journal of
Experimental Psychology: Human Perception and Perfor-
mance, 16 (3), 626–41.
Gabor, Andre (1977), Pricing: Principles and Practices, London:
Heinemann, 200–201.
Gabor, Andre and Clive W. J. Granger (1964), “Price Sensitivity
of the Consumer,” Journal of Advertising Research, 4 (De-
cember), 40–44.
Hinrichs, James V., Dales S. Yurko, and Jing-Mei Hu (1981),
“Two-Digit Number Comparison: Use of Place Information,”
Journal of Experimental Psychology: Human Perception and
Performance, 7(4), 890–901.
Janiszewski, Chris and Donald R. Lichtenstein (1999), “A Range
Theory Account of Price Perception,” Journal of Consumer
Research, 25 (March), 353–68.
Knauth, Oswald (1949), “Considerations in the Setting of Retail
Prices,” Journal of Marketing, 14 (July), 1–12.
Lambert, V. Zarrel (1975), “Perceived Prices as Related to Odd
and Even Price Endings,” Journal of Retailing, 51 (3), 13–22.
Lynch, John G., Dipankar Chakravarti, and Anusree Mitra (1991),
“Contrast Effects in Consumer Judgments: Changes in Mental
64 JOURNAL OF CONSUMER RESEARCH
Representations or in the Anchoring of Rating Scales?” Jour-
nal of Consumer Research, 18 (December), 284–97.
Monroe, Kent B. (2003), Pricing: Making Profitable Decisions,
New York: McGraw-Hill/Irwin.
Monroe, Kent B. and Angela Y. Lee (1999), “Remembering versus
Knowing: Issues in Buyers’ Processing of Price Information,”
Journal of Academy of Marketing Science, 27 (2), 207–25.
Moyer, Robert S. and ThomasK. Landauer (1967), “TimeRequired
for Judgments of Numerical Inequality,” Nature, 215,
1519–20.
Neidrich, Ronald, Subhash Sharma, and Douglas Wedell (2001),
“Reference Price and Price Perception: A Comparison of Al-
ternative Models,” Journal of Consumer Research,28(De-
cember), 329–54.
Parducci, Allen (1965), “Category Judgment: A Range-Frequency
Model,” Psychological Review, 72 (November), 407–18.
Poltrock, Steven E. and David R. Schwartz (1984), “Comparative
Judgments of Multi-Digit Numbers,” Journal of Experimental
Psychology, Learning, Memory and Cognition, 10 (1), 32–45.
Roggeveen, Anne and Gita Venkataramani Johar, “Integration of
Discrepant Sales Forecasts: The Influence of Plausibility In-
ferences Based on an Evoked Range,” Journal of Marketing
Research, 41 (February), 19–30.
Schindler, Robert M. (1991), “Symbolic Meaning of a Price End-
ing,” in Advances in Consumer Research, Vol. 18, ed.Rebecca
H. Holman and Michael R Solomon, Provo, UT: Association
of Consumer Research, 794–801.
Schindler, Robert M. and Thomas Kibarian (1993), “Testing for
Perceptual Underestimation of Nine Ending Prices,” in Ad-
vances in Consumer Research, Vol. 20, ed. Leigh McAlister
and Michael L. Rothschild, Provo, UT: Association of Con-
sumer Research, 580–85.
Schindler, Robert M. and Patrick N. Kirby (1997), “Patterns of
Rightmost Digits Used in Advertised Prices: Implications for
Nine Ending Effects,” Journal of Consumer Research,24(2),
192–201.
Shepard, Roger N. and Peter Podgorny (1978), “Cognitive Pro-
cesses that Resemble Perceptual Processes,” in Handbook of
Learning and Cognitive Processes, Vol. 5, ed. W. Estes, Hills-
dale, NJ: Erlbaum, 189–237.
Sherif, Muzafer and Carl I. Hovland (1961), Social Judgment, New
Haven, CT: Yale University Press.
Stiving, Mark and Russell S. Winer (1997), “An Empirical Anal-
ysis of Price Endings with Scanner Data,” Journal of Con-
sumer Research, 24 (June), 57–67.
Volkmann, John (1951), “Scales of Judgment and their Implica-
tions for Social Psychology,” in Social Psychology at the
Crossroads, ed. John H. Rohrer and Muzafer Sherif, New
York: Harper, 273–96.
... Thus, $4.99 is perceived to be significantly lower than $5.00, but the latter is not perceived to be any different from $5.01 (Manning & Sprott, 2009;Sokolova et al., 2020;Thomas & Morwitz, 2005). Similarly, research also suggests that price differences are perceived to be larger when the arithmetic differences are easier to compute (Biswas et al., 2013;Thomas & Morwitz, 2009a, 2009b. ...
Article
Full-text available
Economic price theory assumes that consumers' responses to prices can be characterized by stable demand curves and price elasticities. The author posits that this assumption lacks descriptive validity because the demand curve is rather unstable; subtle changes in framing and contextual cues can change the demand curve. The article outlines heuristic price theory, which posits that price evaluations are pluralistic in nature. Each price evaluation entails several heuristic decision rules (or decision criteria) that are activated by conscious and unconscious evaluative responses to price and the contextual cues. The extant literature identifies six types of evaluative responses that influence these heuristic decision rules: the pain of paying, price comparisons, price–quality inferences, price negotiability judgments, price fairness judgments, and price–feature tradeoff. To predict how prices influence consumer behavior in a particular context, it is important to identify the heuristic decision rules being used in that context. This implies that managers and researchers, instead of focusing only on estimating price elasticities using stylized demand curves, should also study the heuristic decision rules that shoppers use to evaluate prices.
... Studies in consumer research such as Thomas and Morwitz [12] suggest that as consumers read from left to right, they will often focus on the first digits in deciding how reasonable the price is. But there is evidence reported by Schindler [13,14] and others that as consumers inspect goods prices, they process the right-most digit information, particularly 9-endings, as a signal for low prices. ...
Article
Using daily unleaded gasoline data for almost the totality of Western Australian retail outlets over twenty years, we find that retail prices are most rigid when they are 9-ending as opposed to other price endings. Upward rigidity from a 9-ending retail price is found to be greater than downward rigidity in terms of a lower number of price movements. Irrespective of whether or not a 9-ending price is being charged, an upward gasoline price movement is likely, in absolute terms, to lead to a larger size of price change than a downward movement. In sharp contrast, we find that wholesale gasoline prices are not characterised in such ways and that irrespective of whether or not taxes are included, there is uniformity across the frequency distribution of price endings. The presence of 9-ending pricing affects the nonlinear response of retail gasoline prices to wholesale price movements.
... This may be due to the rounding down of 9-ending prices 8 because rounding up is cognitively more demanding (Freling et al. 2010;Gabor and Granger 1964;Schindler and Kirby 1997;Manning and Sprott 2009). Alternatively, it may be because of lower attention to the price's rightmost digits due to L-to-R processing of price information, causing shoppers to interpret 9-ending prices as lower than they actually are (Poltrock and Schwartz 1984;Thomas and Morwitz 2005). Snir et al. (2017) survey shoppers in Israel and find that they perceive 9-endings as a signal for low prices. ...
Preprint
Full-text available
We assess the role of cognitive convenience in the popularity and rigidity of 0 ending prices in convenience settings. Studies show that 0 ending prices are common at convenience stores because of the transaction convenience that 0 ending prices offer. Using a large store level retail CPI data, we find that 0 ending prices are popular and rigid at convenience stores even when they offer little transaction convenience. We corroborate these findings with two large retail scanner price datasets from Dominicks and Nielsen. In the Dominicks data, we find that there are more 0 endings in the prices of the items in the front end candies category than in any other category, even though these prices have no effect on the convenience of the consumers check out transaction. In addition, in both Dominicks and Nielsens datasets, we find that 0 ending prices have a positive effect on demand. Ruling out consumer antagonism and retailers use of heuristics in pricing, we conclude that 0 ending prices are popular and rigid, and that they increase demand at convenience settings, not only for their transaction convenience, but also for the cognitive convenience they offer.
... This may be due to the rounding down of 9-ending prices 8 because rounding up is cognitively more demanding (Freling et al. 2010;Gabor and Granger 1964;Schindler and Kirby 1997;Manning and Sprott 2009). Alternatively, it may be because of lower attention to the price's rightmost digits due to L-to-R processing of price information, causing shoppers to interpret 9-ending prices as lower than they actually are (Poltrock and Schwartz 1984;Thomas and Morwitz 2005). Snir et al. (2017) survey shoppers in Israel and find that they perceive 9-endings as a signal for low prices. ...
Article
Full-text available
We assess the role of cognitive convenience in the popularity and rigidity of 0-ending prices in convenience settings. Studies show that 0-ending prices are common at convenience stores because of the transaction convenience that 0-ending prices offer. Using large store-level retail CPI data, we find that 0-ending prices are popular and rigid at convenience stores even when they offer little transaction convenience. We corroborate these findings with two large retail scanner price datasets from Dominick's and Nielsen. In Dominick's data, we find that there are more 0-endings in the prices of the items in the front-end candies category than in any other category, even though these prices do not affect the convenience of the consumer's check-out transaction. In addition, in both Dominick's and Nielsen datasets, we find that 0-ending prices have a positive effect on demand. Ruling out consumer antagonism and retailers' use of heuristics in pricing, we conclude that 0-ending prices are popular and rigid, and that they increase demand in convenience settings, not only for their transaction convenience but also for the cognitive convenience they offer.
... Lacetera, Pope, and Sydnor (2012) show evidence of left digit bias in the American used car market in the form of discontinuities in average prices of $150 to $200 at 10,000 mile markers. There have been many papers documenting the presence of consumer left digit bias in a variety of markets, such as Basu (1997), Anderson and Simester(2003), Thomas and Morwitz (2005), and Bray and Harris (2006). This left digit bias can manifest in somewhat surprising surprising ways, such as in Reiley and Samek (2019) where find that changing suggested donations from $100 to $95 reduces donations over $90 by 30%, suggesting that individuals feel differently about rounded numbers. ...
Thesis
This dissertation contains three essays that use reduced form techniques to examine how taxation shapes residential housing markets. Chapter I studies how homebuyers responded to changes in the US federal tax treatment of housing. Chapter II focuses on the capitalization of property taxes into housing prices. Chapter III investigates the presence of cognitive bias with respect to the size of recurring future payments in residential housing markets. The US tax code contains provisions that significantly reduce homeownership costs, including by allowing itemizing income tax payers to deduct property tax and mortgage interest payments from their taxable income. The Tax Cuts and Jobs Act of 2017 capped these deductions and raised the standard deduction, which increased the real cost of property taxes and mortgages for a subset of taxpayers. This shift in the tax law was the most significant change in the US federal tax treatment of homeownership since the Tax Reform Act of 1986. Chapter I and Chapter II of this dissertation study the impact of this change in the law on residential housing markets. In the first chapter, "The Impact of the Tax Cuts and Jobs Act on Residential Housing Choices", I examine how individual homebuyers responded to these changes. The data used in this chapter was constructed by matching of home loan records to deeds and mortgage documents in New Jersey's Middlesex County. Employing a continuous difference in differences estimation technique, this chapter shows that homebuyers responded by purchasing smaller homes with lower property tax burdens, with the level of response indicating that the price elasticity of housing demand is approximately unit elastic. Homebuyers also reduced the size of their home loans (relative to sale price) by the equivalent of the response to a two percentage point increase in interest rates. The second chapter, "Housing Prices and Deductibility of Property Taxes: Evidence from the Tax Cuts and Jobs Act", focuses on how residential housing prices responded to these changes. Using the universe of residential home sales in New Jersey and employing a repeat sales model, this paper estimates that home prices in high-property tax areas fell by an amount corresponding to 70 percent of the increase in property tax liabilities. Chapter III, "Left Digit Bias in Property Taxes", provides evidence of left digit bias in the housing market when it comes to anticipated future property tax payments. Left digit bias is a well documented cognitive bias wherein individuals overemphasize the left-most digit of a number. Left digit bias means that if one considers all possible pairs of numbers which differ by the same amount, the difference between pairs with different leftmost digits will be perceived as larger than the difference in pairs with identical leftmost digits. Using a regression discontinuity technique, this chapter shows that homes with property taxes just over a $1,000 threshold sell for 0.5% less than homes with property taxes just under a $1,000 threshold. This bias amounts to homeowners overpaying for homes by an average of $1,672. This chapter provides evidence that even in high-cost situations, individuals appear to exhibit bounded rationality.
Article
Does the power of reference points mean that minute differences in a purchase price then reverberate in future sales prices? In this research, I show that if previous sales prices are round numbers, defined as multiples of £1,000 (e.g. £231,000), subsequent sales prices entail a considerable premium relative to similar properties that were previously priced at charm numbers that are marginally below those round numbers (e.g. £230,999 or £230,950). Using a sample of repeat sales from the Greater London region from 1995 to 2017, I estimate the premium to be approximately 4 percent after controlling for property characteristics and a large set of fixed effects. Increasing public accessibility of information attenuates the effect. Tax considerations, financial constraints, and pricing errors cannot explain the result. I propose a framework of reference dependence and left-digit bias to explain the result, highlighting the presence of behavioural biases in household decisions, even when very high stakes are involved.
Article
We hypothesize that managers anticipate a disproportionately larger price increase associated with rounded earnings per share (EPS) and make additional effort to round EPS when they plan to sell shares after the earnings announcement. Consistent with this hypothesis, we find that managers who round diluted EPS have higher managerial insider sales following the earnings announcement compared to managers who do not. Furthermore, we find that the positive association between rounding of diluted EPS and subsequent stock sales undertaken by chief financial officers (CFOs) is stronger when the level of abnormal stock repurchases is higher, consistent with managers’ strategic behavior.
Article
In judgment and choice, consumers show a variety of biases, from the sunk cost fallacy and projection bias to usage frequency neglect and erroneous price–quality inferences. This article explains these seemingly disparate biases and predicts new biases using an overarching framework based on the relevance insensitivity theory proposed by Hsee et al. (2019). According to the theory, many biases arise because people are insufficiently sensitive to the relevance (i.e., weight) of a cue variable to the target variable (the dependent variable). The direction of the bias depends on the normative relevance of the cue—people over‐rely on the cue when it is normatively irrelevant and under‐rely on the cue when it is normatively highly relevant. We show that ostensibly unique and universal biases are neither unique nor universal: All are manifestations of relevance insensitivity, and each bias attenuates or reverses as the cue variable's relevance changes.
Article
Full-text available
This paper examines how people price the resale of durable goods in systematically biased ways. We show across four studies that the anchoring effect of durable goods’ prior sales prices on subsequent valuations is discontinuous at psychologically salient round number reference points (e.g., $10,000 increments) because these numbers create qualitative differences in how people perceive values below them versus values at/above them. Resellers set disproportionately larger subsequent prices when previous prices move from just below round number thresholds (e.g., $349,000) to those at or just above these thresholds (e.g., $351,000). The findings show that buyers who pay a price just below a round number, therefore, may sacrifice money because they receive disproportionately less when reselling the good. Market forces only partially attenuate this pricing bias, but valuator experience seems to play a moderating role. Archival data show that home buyers who previously paid just under a $10,000 reference point subsequently listed their homes for about 1.8% (over $3,700) less on average than did buyers selling comparable homes who previously paid at or above a round number threshold. This drop is observable controlling for home characteristics and the general relationship between previous and current prices. Three experimental studies looking at housing and used car markets replicate these findings, highlight the mechanism, and increase confidence in causality. Market mechanisms and the negotiation process attenuate discontinuities by about 30%, but lower initial listing prices persist to final sales prices. We find additional weak evidence suggesting that valuator experience may attenuate intergenerational pricing bias.
Article
Number line estimation tasks are frequently used to study numerical cognition skills. In a typical version, the bounded number line task, target numerals must be placed on a bounded line labeled only at its endpoints (e.g., with 0 and 100). Placements by adults, while highly accurate, reveal a cyclical pattern of over- and underestimation of target numerals. The pattern suggests use of proportion judgment strategies and is well-captured by cyclical power models. Another systematic number line bias that has recently been observed, but has not yet been considered in modeling efforts, is the left digit effect. Numerals with different leftmost digits (e.g., 39 and 41) are placed farther apart on a line than is warranted. In the current study (N = 60), adult estimates were obtained for all numerals on a 0-100 number line estimation task, and fit of the standard cyclical power model was compared with two modified versions of the model. One modified version included a parameter that underweights the rightward digit's place value (e.g., the ones digit here), and the other used the same parameter to underweight all digits' place values. We found that both modifications provided a considerably better fit for individual and median data than the standard model, and we discuss their relative merits and cognitive interpretations. The data and models suggest how a left digit bias might impact estimates across the number line.
Article
Full-text available
A traditional assumption concerning how prices influence buyers’ purchasing behaviors has been that buyers know the prices of the products and services that they consider for purchase. However, empirical research during the past four decades repeatedly has discovered that buyers often are not able to remember the prices of items they had recently purchased. One conclusion that has been drawn is that buyers often do not attend to price information in purchase decisions. The authors argue that this conclusion may be incorrect in that what consumers can explicitly remember is not always a good indicator of what they implicitly know. Price information not consciously remembered can still influence internal reference prices and product evaluations. In this article, the authors discuss the conceptual and methodological ramifications of the distinction between remembering and knowing to reassess and refine our understanding of how buyers process and use price information.
Article
This chapter presents a particular theoretical interpretation of the relationship for one area of experience. The area of experience is elicited by what are usually considered continuous attributes, such as brightness, size, and the theoretical approach that assumes process attributes in terms of discrete semantic codes. The chapter describes the comparative judgment paradigm and the most important effects obtained with it. It also describes and critically reviews the previous models of these results. The semantic coding approach and its application to data from comparative judgment experiments are also discussed. The chapter discusses the relationship between direct experience and the memory of it. The model presented in the chapter is designed to account for the experimental results pertaining to processing of attributes in comparative judgment tasks. It uses hypothetical processes of code manipulation to make the necessary predictions. The codes are called “semantic” because they are assumed to carry information in the same way as do other natural language codes, and they are considered to be discrete.
Article
Even the best advertising campaign may fail if the price of the product is not appropriate, yet retail prices are still determined more by rules of thumb than systematic study.
Article
Illustrating them with studies carried out in his laboratory and elsewhere, the author describes a number of different judgment-scales and discusses their strengths and the conditions under which they are appropriately used. (PsycINFO Database Record (c) 2012 APA, all rights reserved)
Article
The greater of 2 multidigit integers can be chosen by generating and comparing internal representations of the integers (holistic models), sequentially comparing corresponding digits in the integers (sequential place-value models), or simultaneously comparing corresponding digits (parallel place-value models). In Exps I and II, 44 undergraduates chose the greater of 2 4-digit or 2 6-digit integers. As predicted by sequential place-value models, latencies did not depend on the number of unequal digits in the 2 numbers and latencies increased linearly with the position of the leftmost unequal digits, except when only the rightmost digits were unequal. In Exp III (17 Ss), latencies increased linearly for all positions when 2 letters were presented to the right of the integers. Results imply that multidigit integers are compared by sequentially comparing digits. (15 ref) (PsycINFO Database Record (c) 2012 APA, all rights reserved)