-1Tradeoffs between costs and benefits:
Lessons from “the price of 0”
Dan Ariely & Kristina Shampan’er
MIT
Abstract
The paper suggests that cost benefit decisions might be more complex than simply
subtracting costs from benefits. In a series of experiments subjects in one condition were
offered to choose between a more attractive candy for a reasonable positive price and a less
attractive candy for a price close to zero (they also had an option of choosing neither). In
another condition both prices were reduced by a constant such that the smaller price became
zero. The standard theoretical model where the option with the highest benefit cost difference
is chosen, predicts that both demands should weekly increase when prices are reduced. In
contrast the results of six experiments show that while the demand for the less attractive
chocolate increases, the demand for the more attractive chocolate decreases. Three general
psychological theories are proposed to account for this effect: uniqueness of 0, difficulty in
mapping pleasure onto money, and relative aversion to transactions that involve both benefits
and costs. The results support the tradeoff aversion type of mechanism.
-2Tradeoffs between costs and benefits:
Lessons from “the price of 0”
One of the most generic and common choices that consumers face are choices of one
option among competing alternatives that vary on quality, quantity, and price. Examples for
such decisions range from selecting one’s morning coffee, to picking vacations and cars and
even larger decisions such as selecting a house to live in. When researchers conceptualize how
consumers make such choices the standard perspective common to both economic theory and
behavioral decision theory is that consumers behave as decision analysts – trading off costs and
benefits. In particular, when making a choice of a single item out of many substitutes, the
standard perspective prescribe that consumers assess the benefits of each product, subtract the
cost from these benefits and select the option that has the highest positive net value.
When considering such a process from a psychological perspective, one may wonder
whether the need to engage in a cost benefit tradeoff could by itself have an impact on the
overall outcome.
In particular, to the extent that such tradeoffs have a utility in and of
themselves, this could color the entire transaction. Compelling evidence that such tradeoff
conflicts can create some negative utilities was provided by Tversky and Shafir (1992), who
showed that including additional options in the choice set can increase the tradeoff conflict and
hence make the choices themselves less attractive (see also Diederich, 2003; Houston & Doan,
1996; Luce, 1998). As another example, Iyengar and Lepper (2000) demonstrated that as the
size of the choice set increases the propensity to act on one of the options decreases. In one of
their experiments a set of 6 or 24 jams were presented at a tasting booth in a supermarket, and
while more shoppers approached the tasting booth when the larger set was displayed, many
less ended up purchasing a jam in this condition. Thus, different streams of evidence suggest
that it might be useful to rethink the ways in which economic and behavioral decision theories
conceptualize tradeoffs, and in particular to consider the possibility that transactions that
require a low or no level of tradeoffs might become more attractive (and transactions that
-3require tradeoffs might become less attractive) – an idea we will refer to here as “tradeoff
aversion.”
In the current work we focus on the most extreme case involving easy tradeoffs – ones
that involve no cost and thus no tradeoffs. The experiments presented have the same general
structure: n the cost conditions respondents are offered a choice between a moderately priced
high-end candy (e.g. a Lindt truffle), a low priced lower-end candy (e.g. a Hershey's kiss), or
neither. In the free conditions the prices of both goods are discounted by an amount equal to
the price of the lower priced candy in the cost condition.
Thus in the free condition
respondents are given a choice between a discounted moderately priced high-end candy (e.g. a
Lindt truffle), a free lower-end candy (e.g. a Hershey's kiss), or neither.
The predictions regarding the results are different for the standard model and for the
model that includes tradeoff aversion.
In the standard model, where the option with the
highest benefit cost difference is chosen, demands for both goods should weakly increase in
the free condition compared with the cost condition.
These increases in demands would
presumably be caused by the consumers who buy nothing in the cost condition, but buy
something in the free condition due to price drop. Moreover, because both candies’ prices are
reduced by the same amount, the relative comparison of the two items remains constant, which
implies that no consumer should switch candies across these two conditions. To express these
predictions more formally, suppose a person has three options: either to buy a “low value”
product which she values at L for a small price ε, or to buy a “high value” product which she
values at H > L for the price p + ε, or buy neither. According to the standard theory the person
will choose the low value object, if [L object if [H – p -
ε ≥ H – p - ε (i.e. H - L ≤ p) and L ≥ ε], the high value
ε ≥ L -ε (i.e. H - L ≥ p), and H ≥ p + ε], and neither otherwise. Now, suppose,
both prices went down by ε, i.e. the low value product is now free and the high value product
now costs p. Now the person will choose the low value object, if [L ≥ H-p (i.e. H-L ≤ p)], the
-4high value object if [H-p ≥ L (i.e. H-L ≥ p), and H ≥ p], and neither otherwise.
Thus, as
depicted in Figure 1, the rational model predicts that as prices decrease: the behavior of those
who were buying one of the objects would not change and demands for both objects would
weakly increase. The people whose behavior changes are precisely those with 0 ≤ L < ε and H
< p +
ε and these people will switch from nothing to one of the objects. It follows that if ε is
small (say 1¢) then the changes in behavior should be negligible1.
In contrast to this model the notion of tradeoff aversion suggests, that free options are
more attractive than one would predict from the preferences revealed when all prices are
positive, because a free option does not include a tradeoff between benefits and price. Thus,
we expect that in the free condition the market share of the low value good will increase and
the market share for the high value good will decrease beyond the theoretical predictions. This
prediction will be supported in the following six experiments. Experiment 1 is a paper and
pencil survey showing that our intuition coincides with the intuition of the subjects about what
they would do facing the choices described above. Experiment 2 replicates these findings in
real transactions. Experiment 3 eliminates transaction cost as an explanation for the effect.
Experiment 4 generalizes the result from zero to non-positive costs. Finally experiments 5 and
6 contrast the “mapping difficulty” and tradeoff aversion mechanisms – providing support for
the trade off aversion hypothesis.
••• Figure 1 •••
Experiment 1: Hypothetical Experiment
This experiment was conducted to test whether intuitions agreed with ours, that
reducing a cost to zero would disproportionably increase the attractiveness of the free option,
1
Even if people differ on how they rank the attractiveness of the two goods, it is still true that when both prices go
down by the same amount no fall in any of the demands is theoretically possible.
-5shifting demand from the high-end product whose benefit cost difference was presumably
higher before the discount. In addition to the two main conditions described earlier, a third
condition was included in order to examine the effect of price changes that maintain positive
(non 0) prices. By comparing this condition to the cost condition, we could isolate the effect of
price change from the effect of price change to 0. The contrast of this condition with the cost
condition has another advantage; namely that this effect is predicted to be the same by both
the standard model and the tradeoff aversion model (as prices go down demands weakly go up)
so we could use it to make sure that there is nothing inherently wrong with using this
procedure, stimuli, etc. and that the standard model predicts the results in non-zero cases.
Method. Sixty subjects took part in the experiment, randomly assigned to one of the 3
conditions. The two products used were Hershey's kisses (Hershey's) as the good product and
Ferrero Rocher chocolates (Ferrero) as the better product. There were one free condition and
two cost conditions, which were labeled 0&25, 1&26, and 2&27 (see Table 1). The prices for
Hershey’s in the conditions were 0¢, 1¢ and 2¢ respectively. The prices for Ferrero were 25¢,
26¢, and 27¢ respectively. Participants read a printed questionnaire accompanied by the
pictures of the chocolates and were asked to circle the option that they would have preferred. In
1&26 [2&27] condition the question was:
“Imagine that you had a choice of buying a Hershey’s kiss for ¢1 [2¢] or a
Ferrero Rocher for ¢26 [27¢], or neither of those. Circle the option that you
would choose.” The choice options at the bottom of the page were: “Pay 1¢ [2¢]
for a Hershey’s kiss, Pay 26¢ [27¢] for a Ferrero Rocher, Neither”
In 0&25 condition the question was:
“Imagine that you had a choice of getting a free Hershey’s kiss or buying
a Ferrero Rocher for ¢25, or neither of those. Circle the option that you would
-6choose.”
The choice options at the bottom of the page were: “Get a free
Hershey’s kiss, Pay 25¢ for a Ferrero Rocher, Neither”
••• Table 1 •••
Results & Discussion: As can be seen in Figure 2 the results were not in line with the
standard model. In the 0&25 condition the demand for Hershey’s increased substantially in
comparison with 1&26 condition [t(31)=3.8, p<0.001] while, more importantly, the demand for
Ferrero decreased substantially [t(31)=-2.3, p<0.03] 2. Note that while the increase in demand
as a function of price decrease is in principle consistent with the standard model, the decrease
in the demand for Ferrero suggests that the increase in demand for Hershey between the cost
conditions and the free condition was beyond the standard effect of price reduction.
The
decrease in demand for the Ferrero is inconsistent with the standard model, but is consistent
with the hypothesis of tradeoff aversion3. Finally the fact that the difference in demands
between the 1&26 and 2&27 conditions was unnoticeable (for Hershey’s [t(38)=-0.3, p=0.76],
for Ferrero [t(38)=0, p=1]) demonstrates that once all the prices are positive, a 1¢ change in
prices does not have a large effect on the demands, and it is only when one of the prices
becomes zero, that the observed perturbations take place.
In summary, the standard cost benefit model predicts that the preferences between
Hershey’s and Ferrero should not change as long as the price difference is constant, because in
either condition this preference is determined by whether or not H – L > p (the only possible
preference change is in the preferences between buying nothing and something). Our results
show, however, that in a hypothetical setting preferences between Hershey’s and Ferrero do
change: a substantial amount of subjects are willing to forgo the opportunity to buy Ferrero
2
All the p-values reported are for two-tailed tests.
-7once the Hershey’s becomes free.
••• Figure 2 •••
Experiment 2: Real Purchasing Experiment
While the results of the hypothetical experiments provided support for the tradeoff
aversion hypothesis, it is still important to test whether they will hold with real transactions. In
addition, in order to test the robustness of this effect the current experiment also included a
condition that compared the reduction in price of the low-end candy to a larger nominal
reduction in price in the high-end candy.
Method. Three hundred ninety eight subjects took part in the experiment. The two
products used were Hershey's kisses as the low value product and Lindt truffles as the high
value product. There were two free conditions and one cost condition, which were labeled
0&14, 0&10, and 1&15 (see Table 1). In the 0&14, and 0&10 conditions the price of Hershey's
was 0¢ and the price of Lindt was 14¢ and 10¢ respectively. In the 1&15 condition the price of
the Hershey's was 1¢ and the price of the Lindt was 15¢.
A booth at MIT’s student center was set up with two cardboard boxes full of chocolates
and a large upright sign that read "one chocolate per person." Next to each box of chocolates
was a sign lying flat on the table that indicated the price of the chocolate in that condition. We
used the flat signs because we wanted to measure the demand distributions including the
number of people who considered the offer and decided not to partake – we coded each person
who looked at the prices but did not stop or did not purchase as a "no."
3
A second version of this hypothetical experiment was conducted in a within subjects design where each
participant saw and reacted to both conditions side by side. The results replicated those of the between subjects
-8Although this study was designed as a field experiment, one of the limitations of the
setup was that the experimental conditions could not be randomized for each subject and
instead the signs (conditions) were replaced approximately every 30 minutes. When replacing
the signs we wanted to reduce the chance that students would notice the change (thus mixing a
within and between subjects designs) and took a 15 minute break between each of the 30
minute experimental sessions.
Results & Discussion. As can be seen in Figure 3, the results were in line with the
tradeoff aversion hypothesis.
In the 0&14 condition the demand for Hershey’s increased
substantially [t(263)=5.6, p<0.001] compared with the 1&15 condition, while the demand for
Lindt decreased substantially [t(238)=-3.2, p=0.002].
In addition, there was no significant
difference between the demands for chocolate in the 0&10 and 0&14 conditions [for Hershey’s
t(263)=0.5, p=0.64, for Lindt t(271)=1.5, p=0.13], suggesting that the reduction of price to 0
is a powerful effect that can overcome a reduction that is five times as large on the higher-end
chocolate.
In sum, the results of the current experiment demonstrate that when facing a choice
between two options where one is at price 0, its relative attractiveness becomes substantially
higher. This was shown with real transactions in a field setting and even in the case when the
price drop for the high value product was substantially larger than the price drop of the low
value product. The observed drop in demand for the high value good in such a case (the 0&10
condition in comparison to 1&15 condition) is theoretically even "more impossible" than in the
case when prices decrease by the same amount.
Note that one possible shortcoming of the current experiment is that the conditions also
differed in the transaction costs of options: taking a free Hershey’s or buying nothing means
design suggesting that even direct comparison of these two conditions provides the same intuitions to participants.
-9not only a zero monetary price, but also no hassle of looking for change in one's pocket. Thus,
transaction cost difference was in favor of the Hershey’s and no-choice options in the 0&14 and
0&10 conditions, while this difference only favored the no-choice option in the 1&15 condition.
The similarity in the proportions of participants who picked the no-choice option seems to
suggest that transaction costs are not driving the effect. Nevertheless we test this possible
alternative explanation more directly in the next experiment (transaction costs are also
controlled for in experiments 5 and 6).
••• Figure 3 •••
Experiment 3: No Transaction Cost Experiment
One possible alternative explanation to the results presented thus far concerns the
physical transaction cost that is involved when paying any amount even if it is very small for a
good. In fact one could argue that the hassle involved in looking for money in one’s pocket or
backpack is high and constant for most small payments, creating a stark contrast in total cost
(physical and financial) between the free and the 1¢ options.
From this perspective, the
observed effect of 0 in the previous experiments could be attributed to the elimination of
transaction costs in the free condition compared with the cost condition and not to the change
in price per se4. To investigate this possibility we conducted an experiment that was similar in
it’s general structure to Experiment 2 but where transaction cost inequality was eliminated. To
do so we conducted the experiment as one of the regular promotions in the Sloan cafeteria at
MIT, using their current customers as participants and adding the cost of the chocolate to their
bill as done in any other purchase in the cafeteria.
Method.
4
Two hundred thirty two subjects took part in the experiment. The two
Note that the hypothetical experiment could also be interpreted in this light if one assumes that people are aware
of the hassle of payment and take this into account in their hypothetical answers.
- 10 products used were Hershey's as the low value product and Lindt as the high value product.
There was one free condition and one cost conditions, which were labeled 0&13 and 1&14 (see
Table 1). In the 0&13 condition the price of the Hershey's was 0¢ and the price of the Lindt was
13¢. In the 1&14 condition the price of the Hershey's was 1¢ and the price of the Lindt was
14¢.
A box with two compartments, one containing Hershey's and the other containing Lindt,
was placed next to the cashier at the Sloan cafeteria at MIT.
A price (varying across the
conditions) was posted next to each chocolate type and in addition there was a large sign that
read “one chocolate per person.” Students who wanted one of the chocolates could take one
and the cost was added to their bill. Payments in this cafeteria are carried out mostly with a
pre-paid MIT dinning card, or with cash and very rarely with credit or debit cards. In addition,
there was a penny jar next to the cashier, which had at least a few pennies for the use of the
customers.
As in experiment 2, the experiment itself was conducted as a field experiment, which
meant that it was relatively difficult to randomize the prices for each subject. Instead the signs
(conditions) were replaced approximately every 30 minutes with a 10 minute break between
experimental sessions. In addition, the experimental setup made it difficult to perfectly count
those customers who decided not to buy any of the chocolates. We could not distinguish the
people who observed the signs and decided not to participate from the people who did not
notice the offer, so we coded all the customers who passed by the cashier and did not select
any of our chocolates (for whatever reason) as a "no purchase."
Results & Discussion. As can be seen in Figure 4, the results were in line with the
tradeoff aversion hypothesis. In the condition where the Hershey’s was free, the demand for
Hershey’s increased substantially [t(189)=4.7, p<0.001] while the demand for Lindt decreased
substantially [t(206)=-3.2, p=0.001]. These results join the partial evidence from the previous
- 11 experiments to suggest that the effect of the “price of 0” is most likely not due to
considerations such as physical transaction costs: getting out cash, opening one’s wallet, or
looking for change.
••• Figure 4 •••
Experiment 4: Negative Price Experiment
The next question we ask is whether this effect is due to the unique nature of the price
of 0 or weather it is generalizable to all prices that imply no downside to the transaction. This
distinction could be of importance given that there have been numerous studies that have
demonstrated that not paying (or not being paid) can create a very different incentive structure
by changing the mindset of the players. For example, Gneezy and Rustichini (2000) show that
introducing a small payment for each correct answer on an IQ test decreases performance in
comparison to the case where no payment is mentioned.
Following Fisk’s rational theory
(1992) Heyman and Ariely (2004) argue further that when no monetary outcomes are involved,
individuals consider a transaction in terms of social-market exchange are thus less sensitive to
the magnitude of an exchange. This implies that participants in our experiments might have
viewed a free Hershey’s kiss as a part of social market, placing higher value on it, much like
they would have one Mom’s dinner or friend’s help.
Heyman and Ariely (2004) further demonstrated that once a financial exchange is
mentioned (for example when mentioning the cost of a gift), the perception of the exchange
changes from social market to financial exchange.
Using this distinction, the goal of the
current experiment was to set a price that is non-zero (maintaining the financial aspect of the
exchange) and at the same time has no downside. To achieve these two goals we selected a
small negative price (-1¢). The prediction is that to the extent that the effect noted in the
- 12 previous experiments is due to the unique nature of the price of 0, a small negative price
should have a very different effect from a price of 0. On the other hand, to the extent that the
effect noted in the previous experiments is due to tradeoff aversion, a small negative price
should have a very similar effect to the price of 0 since they both have no downside (in
particular, eliminating the need to figure out whether the chocolate is worth its price or not).
Method. Three hundred forty two subjects took part in the experiment. The experiment
replicated the structure of Experiment 2 but with different prices in the three conditions (see
Table 1). In the 1&14 condition the price of Hershey's was 1¢ and the price of Lindt was 14¢.
In the 0&13 condition the price of Hershey's was 0¢ and the price of Lindt was 13¢. In the 1&12 condition the price of the Hershey's was -1¢ (participants were offered a Hershey’s kiss
plus a penny) and the price of Lindt was 12¢.
Results & Discussion. As can be seen in Figure 5, the comparison of 1&14 condition
with 0&13 condition replicates previous findings.
In the 0&13 condition, the demand for
Hershey’s increased substantially [t(193)=3.4, p<0.001] compared with the 1&14 condition
while the demand for Lindt decreased substantially [t(212)=-3.8, p<0.001].
More centrally to the current experiment, the pattern of choices in the -1¢&12 condition
show that the observed effect is most likely not due to some special role of 0 but generalizes to
all options that involve no downside. The pattern of data shows that when Hershey’s price
dropped from 0¢ to -1¢, the demand for Hershey’s further increased [t(223)=2.5, p=0.01]
while the demand for Lindt did not change significantly [t(220)=-0.04, p=0.97]. This means
that a small negative price did not influence the choice like a positive price, but like the zero
price, with the demands going from the cost condition to negative cost condition in the same
direction as to the free condition, but slightly more so.
Thus price of 0 is not a unique
attractor of the demand for Hershey’s, and we can relabel the “effect of price 0” to the “effect of
no cost”. The effect for which the theoretical model cannot account (switching between
- 13 chocolates, and thus the drop in demand for Lindt) happens only when prices change from
1&14 to 0&13. However when prices drop from 0&13 to -1&12 the demand for Lindt does not
change and the additional Hershey’s buyers are not switching from Lindt, but switching from
the “neither” option.
Thus, the results of the current experiment replicated the previous findings by showing
that offers that did not require out of pocket expenses were viewed more favorably then
predictable from the positive price condition. The results also showed that this effect was not
unique to the price of 0 and that it applied to negative prices as well. Thus the results did not
support the idea that a free Hershey’s was seen as social market good (Heyman & Ariely, 2004),
but were more consistent with the trade off aversion hypothesis.
••• Figure 5 •••
Experiment 5: Halloween Experiment
One possible explanation for the results presented thus far is based on the difficulty of
mapping and evaluating different goods in monetary terms (Ariely, Loewenstein, & Prelec, 2003;
Hsee, Yu, Zhang, & Zhang, 2003; Nunes & Park, 2003). The idea is that to the extent that
evaluating the utility of a piece of chocolate in monetary terms is difficult; individuals might
resort to a strategy that assures them at least some positive surplus. Specifically, while getting
a piece of good chocolate for free has to be a positive net gain, paying for a piece of better
chocolate is less clearly so. In line with this idea, the main goal of this experiment was to
examine the "effect of no cost" outside of the domain of monetary transactions by asking
respondents to trade candy for candy – under the assumption that it is simpler to map the
utility of consumption of one type of chocolate into the utility of consumption of another type
of chocolate. The prediction is that to the extent that the mapping explanation is correct, the
effect of the “price of no cost” will not show up when trading off candy for candy. On the other
- 14 hand, to the extent that the effect of the “price of no cost” will be present when trading off
candy for candy, we would increase our confidence in the main hypothesis – namely that
options that do not involve tradeoffs are viewed more favorably than their mere benefit cost
difference would imply. Finally, the current experiment utilized experimental procedures with
almost no difference in transaction costs across conditions.
Method. On Halloween thirty-four trick-or-treaters at Ariely’s house were given three
Hershey's kisses (each weighing about 0.2 oz.) and then were offered a choice between a small
(1 oz) and a large (2 oz) Snickers bar. In the 0&1 condition the kids were told that they could
get the small Snickers bar or exchange one of their Hershey's kiss for the large Snickers bar. In
the 1&2 condition the kids were told that they could exchange one of their Hershey's kiss for
the small Snickers bar or exchange two for the large Snickers bar (see Table 1).
Before
mentioning the Snickers the experimenter asked the kids to hold the Hershey’s they just
received in their open hand in front of them. This was done in order to eliminate any hassle of
transacting – after a candy goes down to a kids candy bag it is much harder to later find it.
Once the kids voiced their choice the experimenter took the specified amount of Hershey’s
kisses from their hand and gave them the Snickers bar of their choice. No one declined the
additional Snickers bar, thus from standard theory prospective (but not from tradeoff aversion
perspective) the demands were expected to be the same in both conditions.
Results & Discussion. As can be seen in Figure 6 the results supported the tradeoff
aversion hypothesis even when the tradeoffs did not involve money. In the condition where the
small Snickers bar was free, the demand for it increased substantially [t(31)=4.9, p<0.001].
These results generalize the results of the "effect of no cost" in a few ways. First and foremost,
they demonstrate that the attractiveness of the zero cost is not limited to monetary
transactions. There seems to be a general increase in attractiveness to options that do not
require giving up anything. Second, the results hold when goods and exchange currency are
commensurate – both being chocolate based candy [for other results regarding
- 15 commensurability see (Ariely et al., 2003; Hsee et al., 2003; Nunes & Park, 2003)]. Third, while
1¢ prices are not very common in the marketplace, the choice of candy is more common and
moreover it is particularly common in the context of getting Halloween candy and in trading
them between friends and siblings later, adding ecological validity to the finding. Fourth, the
results provide further support that the physical hassle involved in transactions (taking money
out of one's wallet for example) cannot account for the results. Finally the results show that
this effect also holds with kids.
In summary, the main reason to conduct the current experiment was to test whether the
difficulty of mapping money onto experiences such as eating chocolate could be the cause for
the no-cost effect (Ariely et al., 2003). To test this idea we replaced money as the exchange
medium with chocolates that are presumably more naturally mapped into other chocolates.
The results demonstrate that the no-cost effect is not limited to monetary exchanges –
supporting the tradeoff aversion hypothesis.
••• Figure 6 •••
Experiment 6: Adult Chocolate Experiment
The Halloween experiment provided evidence that the no-cost effect persisted in nonmonetary transactions and where the goods are commensurable. Yet, the main contribution of
this experiment is not in its ability to generalize the effects to non-monetary transactions, but
in its ability to distinguish between the two psychological theories that could underlie this
effect: difficulty in mapping pleasure onto money, and aversion to transactions that involve
both costs and benefits. Since the distinction between the mapping and the tradeoff aversion
theories is central to our understanding of this effect, it was important to replicate the basic
structure of the experiments with adults to make sure it does not depend on the particular
sample.
- 16 -
Method. One hundred students took part in the experiment. The experiment was carried
out at the student center at MIT in a setup that was similar to experiments 2 and 5 (see Table
1). The procedure itself was very similar to the one used in the Halloween experiment. Two
cardboard boxes were placed on the table, one containing Hershey kisses and the other
containing 1 and 2 oz Snickers bars. The first 100 students who approached the booth were
asked to put their hand forward and were given eight Hershey's kisses while being asked to
keep their hand open. At that point the experimenter showed them two Snickers bars, one
small (1 oz) and one large (2 oz). In the 0&4 condition participants were told that they could
get the small Snickers bar or exchange four of their Hershey's kiss for the large Snickers bar. In
the 1&5 condition participants were told that they could exchange one of their Hershey's kiss
for the small Snickers bar or exchange five of their Hershey's kisses for the large Snickers bar.
Once the participants made their choice the experimenter took the specified amount of Hershey
kisses from their hand and gave them the Snickers bar of their choice. Note that all participants
chose to get an additional Snickers bar.
Results & Discussion. As can be seen in Figure 7 the results replicated the Halloween
experiment providing additional support to the tradeoff aversion hypothesis. In the condition
where the small Snickers bar was free, the demand for it increased substantially [t(92)=4.5,
p<0.001]. These results strengthen the results of the Halloween experiment showing that the
no-cost effect holds in non monetary transactions, with commensurable goods, in simple
standard choices, when there is no physical hassle involved in transacting, and with relatively
smart young adults.
••• Figure 7 •••
- 17 General Discussion
Contrary to the standard rational model, free goods appear to be qualitatively distinct
from and much more attractive than goods that are merely cheap.
Experiment 1 provided
evidence that people have accurate intuition about their own behavior both in a between and
within respondent designs. Experiment 2 demonstrated this effect in a real buying setup and
Experiment 3 replicated the results while demonstrating that the effect cannot be due to
physical transaction costs. The next three experiments were aimed at shedding some light on
the psychological mechanisms that could underlie this phenomenon. Experiment 4 tested if the
results are limited to price of 0 or whether they generalize to transactions with non-positive
costs. The results show that the effects are generalized to transactions that have no downside.
Experiments 5 and 6 attempted to further distinguish between two psychological accounts: one
based on the difficulty of mapping experiences onto money and the second based on aversion
to tradeoffs.
These experiments were carried out with exchange units that were
commensurable (Hershey’s Kisses) making it simple for participants to directly compare the
benefits of the goods and the exchange units. These experiments showed “effects of no cost”,
providing support for the view that purchases that involve no tradeoffs between benefits and
costs receive some boost in their overall attractiveness (or alternatively that options that involve
tradeoffs between benefits and costs are negatively influenced).
At a minimum, the evidence presented here illustrates a utility discontinuity at 0 price.
A more bold interpretation would take these results as a perspective from which to view all
transactions, and suggest that the effect of the price of 0 reflects the psychology of exchange
more generally. Based on this perspective, we proposed two possible psychological theories:
one based on the difficulty of mapping experiences onto money and the second based on
aversion to tradeoffs. While we believe that both of these theories are likely to be valid, the
current evidence points more clearly toward tradeoff aversion.
If one were to take this
interpretation seriously, this implies that as tradeoffs become more salient and more complex
- 18 the likelihood of actions (such as purchasing) will be diminished. Moreover, this interpretation
also implies that consumers, for whom almost all transactions involve costs, consistently
under-buy in their day-to-day purchases, or alternatively, that occasionally when they have a
chance to get stuff for free they are too eager to get something that they have no value for.
- 19 -
References
Ariely, D., Loewenstein, G., & Prelec, D. (2003). "Coherent arbitariness": Stable demand curves
without stable preferences. The Quarterly Journal of Economics, 118(1), 73-105.
Diederich, A. (2003). Decision making under conflict: Decision time as a measure of conflict
strength. Psychonomic Bulletin and Review, 10(1), 167-175.
Fiske, A. P. (1992). The four elementary forms of sociality: Framework for a unified theory of
social relations. Psychological Review, 99(4), 689-723.
Gneezy, U., & Rustichini, A. (2000). Pay enough or don't pay at all. The Quarterly Journal of
Economics, 115(3), 791-810.
Heyman, J., & Ariely, D. (2004). Effort For Payment: A Tale of Two Markets.Unpublished
manuscript, Cambridge.
Houston, D. A., & Doan, K. (1996). Comparison of paired choice alternatives and choice conflict.
Applied Cognitive Psychology, 10(Spec Issue), S125-S135.
Hsee, C. K., Yu, F., Zhang, J., & Zhang, Y. (2003). Medium Maximization. Journal of Consumer
Research, 30(1), 1-14.
Iyengar, S. S., & Lepper, M. R. (2000). When choice is demotivating: Can one desire too much of
a good thing? Journal of Personality and Social Psychology, 79(6), 995-1006.
Luce, M. F. (1998). Choosing to avoid: Coping with negatively emotion-laden consumer
decisions. Journal of Consumer Research, 24(4), 409-433.
Nunes, J. C., & Park, C. W. (2003). Incommensurate resources: Not just more of the same. JMR,
Journal of Marketing Research, 40(4), 26-38.
Tversky, A., & Shafir, E. (1992). Choice under conflict: The dynamics of deferred decision.
Psychological Science, 3(6), 358-361.
- 20 Table 1
Experiment
Experiment 1
Experiment 2
Experiment 3
Experiment 4
Experiment 5
Experiment 6
Condition
Low value good ¢
High value good ¢
0&25
0¢
25¢
1&26
1¢
26¢
2&27
2¢
27¢
0&14
0¢
14¢
0&10
0¢
10¢
1&15
1¢
15¢
0&13
0¢
13¢
1&14
1¢
14¢
-1&12
-1¢
12¢
0&13
0¢
13¢
1&14
1¢
14¢
0&1
0 Hershey’s
1 Hershey’s
1&2
1 Hershey’s
2 Hershey’s
0&4
0 Hershey’s
4 Hershey’s
1&5
1 Hershey’s
5 Hershey’s
- 21 -
Figure captions
Figure 1: A graphical illustration of choosing between a low value product, a high value
product and neither as prices go from
ε and p + ε to 0 and p.
Figure 2: Results from the hypothetical experiment
Figure 3: Results from the real purchasing experiment. Panel a including those who preferred
not to get any chocolate and panel b without these participants.
Figure 4: Results from the no transaction cost experiment. Panel a including those who
preferred not to get any chocolate and panel b without these participants.
Figure 5: Results from the negative price experiment. Panel a including those who preferred
not to get any chocolate and panel b without these participants.
Figure 6: Results from the Halloween experiment
Figure 7: Results from the adult chocolate experiment
- 22 -
Figure 1:
- 23 -
Figure 2:
- 24 -
Figure 3:
- 25 -
Figure 4:
- 26 -
Figure 5:
- 27 -
Figure 6:
- 28 -
Figure 7: