Motivation crowding in real consumption decisions: who is messing with my groceries?
Perino, Grischa ; Panzone, Luca A. ; Swanson, Timothy 等
I. INTRODUCTION
Intrinsic motivation, the desire to perform a task or contribute to
a public good for its own sake rather than out of narrow material
interests, has been found important in the private provision of public
goods, for example, in the form of charitable giving, volunteering, and
many other contexts. (1) There is also substantial evidence that
intrinsic motivation interacts with regulatory interventions in the form
of monetary rewards, fines, or minimum contributions, either positively
(crowding in) or more commonly negatively (crowding out). (2) Crowding
effects have been observed either when moving from a nonmarket to a
market situation, (3) or in principal-agent settings. (4) The latter
does not involve provision of a public good but of effort that benefits
the principal at a cost to the agent. Whether motivation crowding exists
within purely market contexts has been controversial (Ariely, Bracha,
and Meier 2009; Fehr and Falk 2002) but has remained an untested
hypothesis. However, its presence can interfere with the effectiveness
of regulatory interventions intended to stimulate pro-social behavior,
making this issue relevant for both researchers and policymakers.
We provide the first experimental evidence of motivation crowding
in real purchasing decisions. (5) Using food purchases of more than 500
customers at a leading UK supermarket, we compare the impact of
different regulatory interventions on grocery choices. Climate change
mitigation, in the form of a product's carbon footprint, serves as
the public good. For the intrinsic motivation to privately contribute to
its provision to be behaviorally relevant (or "activated") in
a particular context requires information about the public good
dimension of available choices. In this experiment we choose the impure
public good context of grocery shopping (cola, milk, meat, and
butter/margarine) with products differing in their lifecycle carbon
footprints.
Crowding out occurs if an explicit regulatory intervention reduces
the willingness of an individual to buy a "clean" product
compared to an equivalent change in incentives without an explicitly
interventionist character. (6) For marketed goods prices change as a
matter of routine and a small change in price would not give rise to
crowding effects in the absence of information on how it came about.
Crowding out, however, may occur when a government deliberately alters
the incentive system of consumers. Explicit interventions are a key
component of most crowding theories.
We analyze how consumers respond to such government interventions
within the marketplace. Our experimental treatments focus on the context
of otherwise equivalent changes in prices. The context for changes in
prices is important for two reasons. The first is the information
transmitted on the intervention, which might trigger a strategic
response by consumers (Benabou and Tirole 2003; Ellingsen and
Johannesson 2008; Sliwka 2007). The second reason is a pure
"framing effect." Knowing that a change in price is caused by
a deliberate intervention might trigger psychological mechanisms such as
over-justification (tangible extrinsic motivation crowds out intangible
intrinsic motivation), (7) or a reduction in perceived
self-determination (deliberate interference by a third party reduces an
individuals' autonomy which in turn reduces its intrinsic
motivation, see Deci and Ryan 1985).
We are not the first to investigate how context affects responses
to changes in prices or choice sets. Kahneman, Knetsch, and Thaler
(1986) use survey data on the perception of price changes and do not
refer to intrinsic motivation. Eckel and Grossman (2003) compare two
functionally equivalent subsidy schemes for charitable giving, both are
explicit interventions. Kallbekken, Kroll, and Cherry (2011) provide
experimental evidence that acceptance of a Pigouvian tax is reduced when
it is called "tax" rather than a "fee." Falk and
Kosfeld (2006) and Schnedler and Vadovic (2011) investigate crowding
effects in a principal-agent model where the principal can restrict the
agent's choice set. The closest study to ours is the classroom
experiment in Goeschl and Perino (2012) comparing price and
quantity-based instruments with neutrally framed controls. The main
differences to the current paper are that they used direct payments as
the private and purchases of EU ETS emission allowances as the public
good, while we use real grocery products bought by normal customers in a
supermarket. Moreover, they compare a tax with a neutrally framed price
change and an emission standard enforced by a fine, while we compare a
subsidy with a neutrally framed price change and a labeling treatment.
However, both studies find clear evidence of crowding out of intrinsic
motivation to reduce greenhouse gas (GHG) emissions as a result of an
explicit monetary incentive compared to a noninterventionist price
change.
Our fundamental results can be summarized as follows: First, we
find that information provided via a label activates intrinsic
motivation to contribute to a public good in the context of grocery
choices. Labeling induces a significant shift toward cleaner consumption
decisions. On the other hand, we find that an explicit subsidy which
combines information on relative GHG emissions with a monetary incentive
results in a smaller change in behavior than labeling. This is an
example of consumption-based crowding out: the impact of norm activation
and the response to a relative price change are sub-additive. More
striking is the fact that the subsidy is having less of an impact than a
neutrally framed price change, a phenomenon we label over-crowding. This
indicates that consumers object to be regulated in traditionally private
realms, such as grocery shopping.
The context of the experiment, the capacity of instruments to
activate intrinsic motivation of consumers to contribute to climate
change mitigation, is highly relevant for policy making. Food production
significantly contributes to GHG emissions and diet choices are an
important determinant of a household's carbon footprint. However,
little work has been done on how to induce consumers to switch to less
GHG intensive diets. We investigate the ability of instruments to
complement rather than substitute intrinsic motivation in public goods
provision.
The remainder of the paper is structured as follows. Section II
describes the experiment. Section III derives testable hypotheses on the
effect of different regulatory interventions. Section IV presents the
results. The last section discusses the policy implications of our
findings and concludes.
II. THE EXPERIMENTAL SETUP
The experiment was conducted in seven Sainsbury's supermarket
(8) stores in the Greater London area (9) in February and March 2010.
(10) The experiment had two parts. The computer-based first part took
place in the entrance area of a store and participants made purchasing
decisions with carbon footprint information and prices controlled by
randomly assigned treatments. This part mimicked online grocery
shopping, a commonplace phenomenon in the United Kingdom. (11) The
second part of the experiment was the actual shopping trip in the main
part of the supermarket. Decisions made in the first part of the
experiment were enforced by making payment conditional on matching
purchases being made in the second part.
A table with laptop computers was placed in the entrance area of
the stores and a sign announced the opportunity to earn a 5 [pounds
sterling] voucher for participation in a university-sponsored grocery
shopping study. Anyone interested in participating was given a leaflet
stating the eligibility criteria and the payment procedure (see Appendix
SI). The purpose of the experiment was described as "studying how
people make REAL LIFE grocery shopping decisions." No other
information on the purpose of the experiment was provided. In
particular, environmental motivations were not mentioned at any point
during the recruitment phase. Interested shoppers were only turned down
if they (a) did not intend to buy any of the product categories listed
(cola in cans, butter, beef, or milk), (b) did not speak or read
English, (c) were not at least 21 years of age, or (d) had participated
in the experiment previously. (12) While our sample is not a random
sample, participants span a diverse set of socioeconomic backgrounds.
The age range was 21-80 years (mean: 37) and included a wide range of
incomes, educational backgrounds, and political, ethnic, and religious
groups.
In each of four categories (cola drinks, butter/margarine spreads,
fresh meat products, and milk) products with the largest market shares
available in the stores were included in the experiment. The prices,
sizes, and photos of the products were identical to those available in
the store, and displayed in a manner as close as possible to the
store's own online portal. The objective was to create a real but
controlled and incentive compatible choice environment as close as
possible to consumers' normal shopping experience.
The purchasing decision was a real one. Incentive compatibility was
achieved by conferring a 5 [pounds sterling] voucher only upon those who
in the second part purchased the goods chosen in the first part (13)
(see the "Consumer Contract" attached in Appendix SI).
Participants were informed about this procedure beforehand. The
compliance rate was 96%.
In the computerized part participants could select one or more of
the four product categories set out in Table 1. In each of these
categories they were presented with a range of varieties/brands. These
alternatives provided the participant with a range of options catering
to different tastes, but also with very different public good
components. The cola category for example contained six different
varieties (differentiated by brand and nutritional characteristics). The
public good dimension was added by making each variety available in two
different forms of packaging: six-pack of aluminum cans and a single PET
bottle, both containing the same total amount of cola. Cola purchased in
cans has twice the carbon footprint of cola in a bottle.
On the basis of the photographs of all of the products listed in
Table 1 and their current in-store prices (see Appendix SI) participants
made an initial choice in each product category selected. This first
choice constitutes their baseline which was identical for all
participants.
After that, all participants could obtain more information on
environmental and nutritional matters. These explained in general terms
the information that is provided on food labels--the meaning of
nutritional or of carbon footprint information--but did not give any
product-specific details. The information provided was the same across
all treatments. (14)
Then participants were randomly assigned to one of three
treatments. (15) There are two instrument (Labeling, Subsidy) and one
neutrally framed control (Neutral Price Change) treatments. This allowed
consumers to revise their initial choices based on the new
information/incentives imposed by the treatment. Having a baseline and a
treatment choice for each participant is essential as it allows
isolating the impact of treatments on behavior from any unobserved
differences in treatment groups. Treatments can then be compared using a
difference-in-difference analysis focusing on differences in behavioral
change across treatments. The order (baseline--treatment) is driven by
the desire to understand the effect of introducing (rather than
removing) a regulatory intervention. However, previous contributions
(Frohlich and Oppenheimer 2003; Goeschl and Perino 2012; Meier 2007)
showed that motivation crowding can be short-term persistent and hence
we would expect similar results if the order were reversed.
The Labeling treatment was intended to be as simple and non-leading
as possible and presented consumers with two sets of information: a
nutritional and a carbon footprint label. The former was a replication
of the fine print that appears on these goods giving information on
calories, protein, carbohydrate, fat, and salt content. The nutritional
information was intended to reduce the risk of participants guessing the
purpose of the experiment and then trying to please the experimenter
(experimenter demand effect). Since the information was the same as on
the packages, this was expected to have only a small effect on choices
as consumer caring about these characteristics would be already aware of
them. Moreover, variations in nutritional characteristics and carbon
footprints were perfectly orthogonal for one product category (Cola)
which allows to test for confounding effects. The carbon footprint label
followed the design provided by the Carbon Trust UK. It is used by
another major supermarket chain in the United Kingdom for own brand
products but not for the products in this experiment. The label has the
form of a stylized footprint and shows the amount of carbon dioxide
equivalent GHG emissions caused over the life cycle of the product in
grams. (16)
The Subsidy treatment stated that the government intervened by
subsidizing products with lower carbon footprints. The instructions
(here for cola) read "There has been a price change. Products in
plastic bottles have a 5p discount due to a GOVERNMENT SUBSIDY received
on account of its low carbon footprint." This is to be contrasted
with the noninterventionist Neutral Price Change treatment that was
identical to the Subsidy treatment with the exception of the motivation
given for the price change (market conditions rather than government
intervention). The instruction for the Neutral Price Change treatment
read (again for cola) "There has been a price change. Products in
plastic bottles have a 5p discount because of a change in the price of
materials."
Note that there are two differences between the Subsidy treatment
and its control: (a) the explicit interventionist character and (b)
information about the relative ranking of products with respect to their
carbon footprint. (17) The control treatment has neither while the
Subsidy treatment has both (intervention and information).
Both before and after the price reduction the cheapest product was
always a clean one (with the exception of milk where all option cost the
same in the baseline) and the most expensive option was the same before
and after the price change. A participant regarding clean and dirty
options as perfect substitutes and without intrinsic motivation should
not change her mind between baseline and treatment choice in any of the
treatments.
Following the experiment, socio-demographic details on the
respondents were collected. A total of 570 shoppers completed the task
and complied with all terms and conditions of the experiment, and are
included in the sample. The number of participants and hence independent
observations per treatment were: 237 for Subsidy, 248 for Neutral Price
Change, and 85 for Labeling. A total of 333 participants chose only one
product category, 154 made choices in two, 66 in three, and only 17 made
purchases from all four categories. This resulted in 907 pairs of
purchasing decisions.
III. SWITCHING BEHAVIOR WITH INTRINSIC MOTIVATION
In this section we derive testable hypotheses of the switching
behavior of intrinsically motivated consumers who use private
consumption decisions to determine their contribution toward public
goods. To obtain a clear benchmark, the activation of intrinsic
motivation is assumed to be independent of the regulatory instrument
(Heyes and Kapur 2011). Hence hypotheses are based on the assumption
that there is no motivation crowding.
Goods with lower carbon emissions (i.e., the "carbon
footprint" of the product concerned) are referred to as the
"clean" options, goods with higher carbon footprints as
"dirty." In order for intrinsic motivation to be behaviorally
relevant, participants require information allowing them to rank private
goods with respect to the levels of the public good provided with each.
The carbon footprint does exactly this. (18)
As participants are choosing between everyday products that were on
their shopping list for that day, it seems reasonable to assume that
they were familiar with them. We also assume that providing information
on either absolute or relative carbon footprints does not decrease
participants' intrinsic motivation to buy clean products. This
holds unless the distance between rather than the ranking of products
matters and participants initially overestimate the advantage of clean
products and update them based on the information provided in
treatments. We will consider this when discussing the results. Most
likely consumers were not aware of products' (relative) carbon
footprints at the start of the experiment as for the products considered
carbon footprints were neither in use in the United Kingdom nor widely
publicized. (19) For uninformed participants the information provided in
the Labeling and Subsidy treatments is essential for intrinsic
motivation to be behaviorally relevant. Carbon footprints can induce
participants who are intrinsically motivated to mitigate climate change
(and who initially bought a dirty product) to switch to a clean one,
either because they, for the first time, have the information required
to discriminate between clean and dirty products or because preexisting
information has been updated or reinforced. Because the information
provided by the label is the only difference between the two choice
situations, intrinsic motivation is the only systematic driver of any
switches observed. (20)
Hypothesis 1: Information on carbon footprints of products will
induce some participants to switch from a dirty to a clean product.
Participants initially opting for a clean product are not affected.
The Neutral Price Change treatment tests whether participants
respond to small changes in relative prices. For that purpose all clean
products received a reduction in prices. The price reductions are given
in Table 1. Participants were not made aware that the price change had
anything to do with environmental characteristics of the product. Hence,
there is no reason why intrinsic motivation should be affected in the
Neutral Price Change treatment.
Hypothesis 2: A price change will induce some participants to
switch from a dirty to a clean product. Participants initially opting
for a clean product are not affected.
In the Subsidy treatment participants learned that the government
has introduced a subsidy which reduces the price of clean products. Note
that the clean products were already cheaper (with the exception of milk
where all varieties had the same price) before the price change.
Consumers faced a new set of relative prices, were able to rank products
in each category according to their carbon footprint (only the cleaner
options qualify for the reward), and were aware of the interventionist
character of the price change. Except for the latter two aspects which
can affect activation of intrinsic motivation but nothing else, the
Subsidy treatment is equivalent to the Neutral Price Change treatment.
The monetary reward increased the attractiveness of clean products as
does any activation of intrinsic motivation.
Comparing switching behavior in the Subsidy and the Neutral Price
Change treatments allows isolating the effect of an explicit
intervention on intrinsic motivation. Compared to the Labeling
treatment, the change in relative prices creates an additional incentive
to switch. This gives rise to the following hypotheses:
Hypothesis 3:
(a) The subsidy (for clean products) will induce some participants
to switch toward clean products. Participants initially opting for a
clean product will be unaffected by the subsidy.
(b) The subsidy will induce more participants to switch from dirty
to clean varieties than in either the Labeling or the Neutral Price
Change treatments.
If Hypothesis 3b were not confirmed this would indicate crowding
out of intrinsic motivation by the explicit price intervention in the
Subsidy treatment. Hypotheses 1-3a make within-subject comparisons and
conjecture a systematic impact of treatments on participants'
behavior compared to the baseline (choice 1). Hypothesis 3b. central to
detecting crowding out, makes between-subject comparisons and
predictions about systematic differences between impacts of treatments.
All hypotheses make predictions about changes in the behavior induced by
treatments and not about initial choices allowing us to abstract from
issues like which product is more expensive in absolute terms and focus
on changes in relative prices instead.
[FIGURE 1 OMITTED]
IV. RESULTS
This section presents the switching pattern for all treatments and
tests the hypotheses formulated in Section III. We verify whether a
carbon label activates intrinsic motivation, if a change in relative
prices has the expected substitution effect, and we test the interaction
of price and information in the form of a subsidy. By making comparisons
across treatments we identify whether and to what extent there is
motivation crowding.
Switching behavior is presented by treatment in Figure 1 and by
treatment and product in Table 2. Labeling and Neutral Price Change had
roughly the same impact, but this varied significantly across products,
which is most likely because of the differences in elasticities of
substitutions. For example, the rate of switching between "cola in
cans" and "cola in plastic" is higher than between
"beef' and "chicken."
However, product categories differ in more than one dimension and
hence a definitive answer cannot be given.
Most striking is the response to the subsidy. It induces fewer
switches than both the Neutral Price Change and the Labeling treatments
in all product categories.
A. Activation of Intrinsic Motivation
The Labeling treatment provided two sets of information to
participants: nutritional information on calories, fat, carbohydrates,
and salt copied from the products' packages and information about
carbon footprints. We first analyze the impact of nutritional
information before we turn to carbon footprints.
As expected, the nutritional information had only a small impact on
participants' choices. This is most obvious for the cola category
where nutritional and environmental variables were perfectly orthogonal.
All nutritional aspects were driven by the variety (e.g., Coke vs. Diet
Coke) consumed while the carbon footprint was determined by the
packaging (bottles vs. cans). All varieties were offered in both forms
of packaging. As many as 36.4% of participants with a dirty baseline
choice (4 of 11) changed to the cleaner packaging option while only 8.7%
of participants (2 of 23) changed cola variety after information on
nutrition and carbon footprints was provided. For all product categories
pooled together, the only significant difference between nutritional
aspects of first and second choices exists for calories and fat (which
are correlated with carbon footprints for all product categories except
cola) for the subgroup that moved from dirty to clean options. This is
despite there being substantial variation (up to two orders of
magnitude) between nutritional aspects of varieties within both the
clean and dirty options. The irrelevance of nutritional information is
intuitive as it is available on product labels and hence participants
that strongly care about these attributes already took them into account
in their baseline choice.
Participants who are intrinsically motivated to make a private
contribution to climate change mitigation by buying products with a low
carbon footprint (compared to substitute goods) may switch from a dirty
to a clean product in the Labeling treatment (Hypothesis 1). (21) To
test for such a pattern we compare the choices for milk and the other
three products separately. The reason for this split is twofold.
Firstly, milk has three different categories (whole/dirty,
semi-skimmed/medium, and skimmed/clean), while the other products have
only two (dirty and clean). (22) Secondly, participants were preselected
based on whether they were intending to buy milk, cola in cans, butter,
or beef (i.e., the dirty variety in each category with the exception of
milk). (23)
Of the 57 participants buying milk, 6 switched from dirty to clean
while no one changed from clean to dirty in the Labeling treatment (see
Table A2). The null hypothesis that the carbon label had no impact on
choices is rejected but only at the 10% level (Bhapkar
(24)/Stuart-Maxwell tests for marginal homogeneity yield p = .0528Ip =
.0695, respectively).
For the other three product categories, 37.2% (16 of 43) of initial
dirty choices but only 10% (3 of 30) of initial clean choices were
reversed after the label was introduced (see Table A3). This is
significant at the 1% level (Bhapkar/McNemar tests for marginal
homogeneity, p = .0015/jo = .0029). This confirms Hypothesis 1 and hence
gives rise to the following first result.
Result 1: Providing information on the relative environmental
performance allows participants to act on or increases preexisting
intrinsic motivation and induces switching toward clean products.
B. Response to a (Pure) Price Change
We proceed with testing Hypothesis 2, which states that a change in
relative prices induces switching toward products that became cheaper.
For the milk subsample (175 participants) 21 changed as expected and 7
changed in the opposite direction (see Table A4). This is only
marginally significant (Bhapkar/Stuart-Maxwell tests, p = .0958/. 1019).
However, for the other product categories (233 observations, see Table
A5), there were 45 switches from dirty to clean and 19 from clean to
dirty. The null hypothesis that the Neutral Price Change treatment had
no systematic effect is clearly rejected at the 1% level
(Bhapkar/McNemar tests for marginal homogeneity, p = .0009/ p = .0012).
This confirms Hypothesis 2.
Result 2: Participants respond to a change in relative prices by
substituting toward products for which prices were reduced.
C. Subsidies and Intrinsic (De-) Motivation
Having established how participants respond to information on
relative environmental performance of products and to relative price
changes
we now investigate the effect of combining the two drivers of
behavior. The Subsidy treatment provides exactly the same monetary
incentive as the Neutral Price Change treatment but combines it with an
explicit rationale (framing). The subsidy is paid to increase the
attractiveness of products with low carbon footprints. Participants
hence learned the relative ranking of products with respect to their
carbon footprint. Both effects, price change and intrinsic motivation,
work in the same direction and are expected to reinforce each other
(Hypothesis 3b). But first we test Hypothesis 3a by checking whether a
subsidy has a significant effect on behavior.
Surprisingly, the null hypothesis that the subsidy had no
systematic impact on choices cannot be rejected for either sample
(Bhapkar/Stuart-Maxwell tests for marginal homogeneity, no milk: p =
.4342/.4349 and milk: p = .4346/ .4403). For cola, butter/margarine, and
meat, only 18.4% (23 of 125) of dirty initial choices but 19.8% (18 of
91) of clean initial choices were reversed (see Tables A6 and A7). The
subsidy hence has no statistically significant effect on purchasing
behavior. This result is not driven by differences in initial choices
between the Labeling, Neutral Price Change, and Subsidy treatments. A
Kruskal-Wallis test on equality of populations in choice 1 yields a p
value of .5267 (.4068 for milk subsample).
We investigate this further by testing Hypothesis 3b, which
requires a differences-in-differ ences analysis and a focus on dirty
initial choices. Are people more likely to switch from dirty to clean
products in the Subsidy treatment than in the Labeling or Neutral Price
Change treatments? We first perform pair-wise Fisher's exact tests
and two-sample tests of equality of proportions using the null
hypothesis that the frequency of switching is the same in the Subsidy
and the Neutral Price Change/Labeling treatments. All of them reject
equality of frequencies (Table 3).
Regressions (1) and (2) in Table 4 present results from logit
regressions that control for observable differences, for example,
product types and socio-demographic characteristics of participants.
(25) The coefficients confirm that the likelihood of switching from a
dirty to a clean product is significantly lower in the Subsidy treatment
than both in the Neutral Price Change and the Labeling treatments. (26)
This finding not only rejects Hypothesis 3b, but confirms the exact
opposite. The instrument that combines both drivers of switching
behavior performs worse than if each driver is applied separately.
Regression (3) in Table 4 includes an interaction term between the
Subsidy dummy and Log of Income. The results indicate that the crowding
out in the Subsidy treatment is mainly due to participants with average
and above average income. For them, the negative interaction term
dominates the positive direct effect (the mean of Log of Income is 3.2).
The effect of income on crowding is intuitive since poorer consumers
face a tougher trade-off between money and an urge to resist government
intervention.
Taken together, the tests of hypotheses 3 a and 3b provide evidence
of crowding out of intrinsic motivation. The preexisting intrinsic
motivation activated or reenforced by information on relative carbon
footprints is not inducing any additional changes. Quite the contrary:
the sensitivity to a change in relative prices is reduced if the price
change is framed as an explicit intervention in the form of a GHG
abatement subsidy. Hence, the combined effect of a price change and
information is not only sub-additive (crowding out) but also smaller
than each effect individually (over-crowding). While effects similar to
over-crowding have been covered in the theoretical literature (Bowles
and Hwang 2008), we are not aware of any other evidence of its
existence.
Result 3: A subsidy combining a price change and information on
relative environmental performance of products via an explicit
intervention is less effective in activating intrinsic motivation than
either a carbon label or a neutral equivalent price change (crowding
out).
This observation would be in line with "rational"
crowding based on signaling of the size of the environmental advantage
of clean products by the regulator (Benabou and Tirole 2003) assuming
that the regulator is better informed about the environmental damages
associated with different products and that consumers update their
beliefs based on the size of the subsidy. This can explain crowding out
under two conditions: A substantial share of participants (a) held prior
beliefs about the carbon footprints of products and (b) perceived the
level of the subsidy (but not the actual footprint provided in the
Labeling treatment) as a signal that they had overestimated the
advantage of clean varieties. As we have argued previously, information
on carbon footprints of the products used was not generally available
prior to our experiment. Moreover, if the above conditions would be met,
we would also expect to see a difference between the Subsidy and the
Neutral Price Change in the number of switches from clean to dirty. The
same updating mechanism would induce some participants to switch from
clean to dirty as dirty appears no longer as bad as it was when taking
the baseline choice in the Subsidy treatment. However, there is no
evidence of such an effect. The only product for which there is a
significant difference in the clean to dirty switching between the two
treatments (at the 10% level using Fisher's exact test and a test
of proportions) is cola but in the opposite direction. Hence,
participants do not appear to have arrived at the experiment with an
overestimation of the climate benefits of clean options. Vertical
signaling as proposed by Benabou and Tirole (2003) does not explain our
findings.
One potential problem is that the information provided in the
Labeling and Subsidy treatments is not exactly the same. The former
presents carbon footprints while the latter only identifies the relative
ranking of products. While this could explain sub-additivity, it cannot
be the sole reason of over-crowding and the failure of a subsidy to
induce switching. Furthermore, there is evidence that participants did
not pay close attention to the size of the difference in carbon
footprints. (27) It therefore seems reasonable to assume that
information on relative rankings of products is sufficient to activate
intrinsic motivation.
V. DISCUSSION: WHO IS MESSING WITH MY GROCERIES?
It is our hypothesis that the primary distinction between
treatments is the exposure of the consumer's private
decision-making to government intervention. In the Labeling treatment,
consumers were subject to an intervention but it neither changed prices
nor the choice set. Consumers could choose to ignore the information
entirely, and so exposure of the grocery-shopping decision-making was
entirely voluntary. In the Neutral Price Change treatment the price
change was involuntary, but it was framed as a market phenomenon, that
is, it was not an intervention imposed by some external (only faceless
market forces). Consumers responded in the expected ways.
The Subsidy treatment combined the involuntary nature of a price
change with an explicit intervention. Here, the consumer's formerly
private decision-making realm was being impacted without the consent of
the consumer. Although the consumer's welfare was weakly improved
by reason of both the increase in its private budget (the subsidy) and
the proposed increase in the public good (the public benefits from
reduced climate change), the reduction of the private realm of consumer
choice was rejected by participants.
It is interesting that consumers might be contesting the division
between the private and the public realms of choice. This is another,
potentially competing explanation for the phenomenon of crowding out. It
could be that intrinsic motivation to provide public goods is
complemented by an in-built belief that there should be a separation
between the private and the public realms of choice. Citizens might be
defending their boundaries around the private realm of behavior as much
as they are the public realm.
VI. CONCLUSIONS
Our findings provide the first direct evidence of motivation
crowding in real consumer purchasing decisions. We find an explicit
price intervention crowds-out (and even "overcrowds")
intrinsic motivation.
To summarize the key results, information provision through
labeling is able to induce shoppers to switch to more climate friendly
product varieties by activating or reenforcing their preexisting
intrinsic motivation to contribute to climate change mitigation. It is
interesting to note that approximately 20% of consumers require only
information to give effect to socially beneficial behavior. (28)
However, combining information on the relative environmental performance
of products with a monetary reward for switching is less effective than
information alone. Moreover, using a subsidy as an explicit regulatory
intervention performs worse than an equivalent but neutrally framed
price change. This result indicates that the framing of otherwise
equivalent changes in the incentive structure matters for intrinsic
motivation and is in line with findings by Goeschl and Perino (2012) who
observe crowding out of intrinsic motivation by an environmental tax
compared to a neutrally framed price change. Hence, the negative effects
of a price intervention seem to be broadly similar for taxes and
subsidies.
From a policy perspective we are investigating the ability of
various instruments to complement intrinsic motivation in public goods
provision within the context of private consumption decisions. In other
words, what works to activate consumer's public-spiritedness within
the realm of private decision-making?
Our findings represent clear evidence in support of the standard
"crowding out" effect, that is, the effect that adding the
subsidy to the information does not improve performance of the
instrument. More practically, it means that a governmental intervention
that is highly visible to consumers might be less effective than a more
"hidden" tax or subsidy. (29) More interestingly, it might
also mean that there is a more fundamental cause of some of these
observed "crowding" effects. The resistance of consumers to
forced interventions, even those representing unambiguous increases in
welfare such as a subsidy, is indicative of a resistance of consumers to
altering the line between private and public decision making--in either
direction. It could be that consumers are as interested in keeping
governmental interventions outside of certain realms of normally private
decision-making, as they are in keeping market-based interventions
outside of certain realms of normally social decision-making.
Taken together, the results from this experiment suggest that a
government working in an area of consumer choice imbued with both public
and private good characteristics should proceed carefully in its
selection of instruments. Not all instruments have the same effect when
operating around the boundaries between private and public goods, simply
because citizens (and consumers) might resist the alteration of that
boundary.
A number of caveats are in order. Firstly, by design the experiment
can only capture short-term effects. Switching patterns might be
different in the long run as previous studies have found diverging
results on long-term persistence of crowding out (Goeschl and Perino
2012; Meier 2007). Secondly, the credibility of the framings can be
questioned on the grounds that participants were fully aware that they
were participating in an experiment and that any subsidy or price change
would not be permanent, and that any government involvement is
hypothetical. However, if participants would indeed have found the
framing unconvincing, we should not have found any significantly
different effects between the Subsidy and the Neutral Price Change
treatments. The fact that we did, indicates that the setup did work as
intended.
ABBREVIATION
GHG: Greenhouse Gas
SUPPORTING INFORMATION
Additional Supporting Information may be found in the online
version of this article:
Appendix S1: Recruitment and Experimental Instructions.
doi:10.1111/ecin.12024
APPENDIX A
TABLE A1
Variable Descriptions
Variable Description
Treatment dummies
Labeling Dummy: 1 if participant was subject to Labeling
treatment, 0 otherwise
Subsidy Dummy: 1 if participant was subject to Subsidy
treatment, 0 otherwise
Neutral Price Dummy: 1 if participant was subject to Neutral
Change Price Change treatment, 0 otherwise
Product dummies
Milk Dummy: 1 for milk, 0 otherwise
Meat Dummy: 1 for meat, 0 otherwise
Butter Dummy: 1 for butter/margarine, 0 otherwise
Location dummies
L.NewBamet Dummy: 1 if observation is from store in New
Barnet, 0 otherwise.
L.Edgware Dummy: 1 if observation is from store in Edgware,
0 otherwise.
L.Chiswick Dummy: 1 if observation is from store in
Chsiwick, 0 otherwise.
L.Merton Dummy: 1 if observation is from store in Merton,
0 otherwise.
L.Walthamstow Dummy: 1 if observation is from store in
Walthamstow, 0 otherwise.
Socio-demographic
variables
Envinfo Dummy: 1 if participant opts to read the
environmental information sketching the problem
of climate change and the meaning of a carbon
label, 0 otherwise.
Nutrinfo Dummy: 1 if participant opts to read the
nutritional information sketching the health
impact of energy, proteins, carbohydrates, fat,
and salt referred to on food labels,
0 otherwise.
Number of children Number of children (below 12 years old) in the
in household household
Number of persons Number of persons living in a household
in household
Male Dummy: 1 if participant is male, 0 otherwise.
Age Age of participant
Student Dummy: 1 if participant is a student, 0 otherwise.
Ncsec Socioeconomic class, based on participants'
occupation in accordance with UK Office for
National Statistics 2005 guidelines.
Unemployed Dummy: 1 if participant is unemployed,
0 otherwise.
Retired Dummy: 1 if participant is retired, 0 otherwise.
Edul Dummy: 1 if participant does not have any
university-level education, 0 otherwise.
Log of Income Log of household income (in thousand GBP)
TABLE A2
Choices in Labeling Treatment (Milk)
Choice 2
Whole Semi- Skimmed
skimmed
Choice 1 Whole 15 2 2 19
Semi-skimmed 0 29 2 31
Skimmed 0 0 7 7
15 31 11 57
TABLE A3
Choices in Labeling Treatment
(Cola, Butter/Margarine, and Meat)
Choice 2
Dirty Clean
Choice 1 Dirty 27 16 43
Clean 3 27 30
30 43 73
TABLE A4
Choices in Neutral Price Change Treatment (Milk)
Choice 2
Whole Semi- Skimmed
skimmed
Choice 1 Whole 45 11 1 57
Semi- 3 88 9 100
skimmed
Skimmed 2 2 14 18
50 101 24 175
TABLE A5
Choices in Neutral Price Change Treatment
(Cola, Butter/Margarine, and Meat)
Choice 2
Dirty Clean
Choice 1 Dirty 105 45 150
Clean 19 64 83
124 109 233
TABLE A6
Choices in Subsidy Treatment (Milk)
Choice 2
Semi-
Whole skimmed Skimmed
Choice 1 Whole 54 5 0 59
Semi- 3 72 7 82
skimmed 1 2 9 12
Skimmed 58 79 16 153
TABLE A7
Choices in Subsidy Treatment
(Cola, Butter/Margarine, and Meat)
Choice 2
Dirty Clean
Choice 1 Dirty 102 23 125
Clean 18 73 91
120 96 216
TABLE A8
Summary Statistics
Standard
Variable Mean Deviation
Labeling 0.14 0.35
Subsidy 0.41 0.49
Neutral Price Change 0.45 0.50
Milk 0.42 0.49
Meat 0.18 0.38
Butter 0.23 0.42
L.NewBarnet 0.06 0.23
L.Edgware 0.17 0.38
L.Chiswick 0.19 0.39
L.Merton 0.30 0.46
L.Walthamstow 0.11 0.31
Envinfo 0.37 0.48
Nutrinfo 0.58 0.49
Number of children in 0.64 1.02
household
Number of persons in 2.94 1.65
household
Male 0.35 0.48
Age 37.4 12.13
Student 0.08 0.27
Ncsec 8.53 4.22
Unemployed 0.04 0.21
Retired 0.04 0.20
Edul 0.40 0.49
Income (in '000 GBP) 30.45 18.14
REFERENCES
Andreoni, J. "Privately Provided Public Goods in a Large
Economy: The Limits of Altruism." Journal of Public Economics, 35,
1988, 57-73.
--."Giving with Impure Altruism: Applications to Charity and
Ricardian Equivalence." Journal of Political Economy, 97(6), 1989,
1447-58.
Arce, D. G. "Is Agency Theory Self-Activating?" Economic
Inquiry, 45(4), 2007, 708-20.
Ariely, D., A. Bracha, and S. Meier. "Doing Good or Doing
Well? Image Motivation and Monetary Incentives in Behaving
Prosocially." American Economic Review, 99(1), 2009. 544-55.
Bern, D. J. "Self-Perception: An Alternative Interpretation of
Cognitive Dissonance Phenomena." Psychological Review, 74(3), 1967,
183-200.
Benabou, R., and J. Tirole. "Intrinsic and Extrinsic
Motivation." Review of Economic Studies, 70, 2003, 489-520.
Bhapkar, V. P. "A Note on the Equivalence of Two Test Criteria
for Hypotheses in Categorical Data." Journal of the American
Statistical Association, 61(313), 1966, 228-35.
Bjomer, T. B., L. G. Hansen, and C. S. Russell. "Environmental
Labelling and Consumers' Choice--An Empirical Analysis of the
Effect of the Nordic Swan." Journal of Environmental Economics and
Management, 47(3), 2004, 411-34.
Bowles, S. "Policies Designed for Self-Interested Citizens May
Undermine The Moral Sentiments': Evidence from Economic
Experiments." Science, 320, 2008, 1605-09.
Bowles, S., and S.-H. Hwang. "Social Preferences and Public
Economics: Mechanism Design When Social Preferences Depend on
Incentives." Journal of Public Economics, 92(8-9), 2008, 1811-20.
Charness, G., and U. Gneezy. "Incentives to Exercise."
Econometrica, 77(3), 2009, 909-31.
Chetty, R., A. Looney, and K. Kroft. "Salience and Taxation:
Theory and Evidence." American Economic Review, 99(4), 2009,
1145-77.
Corgnet, B. "Peer Evaluations and Team Performance: When
Friends Do Worse Than Strangers." Economic Inquiry, 50(1), 2012,
171-81.
Deci, E., and R. M. Ryan. 1985. Intrinsic Motivation and
Self-Determination in Human Behaviour. New York: Plenum Press.
Eckel, C. C., and P. J. Grossman. "Rebate Versus Matching:
Does How We Subsidize Charitable Contributions Matter?" Journal of
Public Economics, 87(3-4), 2003, 681-701.
Ellingsen, T" and M. Johannesson. "Pride and Prejudice:
The Human Side of Incentive Theory." American Economic Review,
98(3), 2008, 990-1008.
Falk, A., and M. Kosfeld. "The Hidden Costs of Control."
American Economic Review, 96(5), 2006, 1611-30.
Fehr, E., and A. Falk. "Psychological Foundations of
Incentives." European Economic Review, 46, 2002, 687-724.
Fehr, E., and K. M. Schmidt. "Fairness, Incentives, and
Contractual Choices." European Economic Review, 44, 2000, 1057-68.
Frey, B. S. 1993. "Does Monitoring Increase Work Effort? The
Rivalry with Trust and Loyalty." Economic Inquiry, 31:663-670.
--. Not Just for the Money: An Economic Theory of Personal
Motivation. Cheltenham, UK: Edward Elgar Publishing, 1997.
Frey, B. S., and F. Oberholzer-Gee. "The Cost of Price
Incentives: An Empirical Analysis of Motivational Crowding-Out."
American Economic Review, 87(4), 1997, 746-55.
Frey, B. S., and R. Jegen. "Motivation Crowding Theory."
Journal of Economic Surveys, 15(5), 2001, 589-611.
Frohlich, N., and J. Oppenheimer. "Optimal Policies and
Socially Oriented Behavior: Some Problematic Effects of an Incentive
Compatible Device." Public Choice, 117, 2003, 273-93.
Gneezy, U., and A. Rustichini. "A Fine Is a Price."
Journal of Legal Studies, 29(1), 2000a, 1-17.
--. "Pay Enough or Don't Pay at All." Quarterly
Journal of Economics, 115(2), 2000b, 791-810.
Goeschl, T" and G. Perino. "Instrument Choice and
Motivation: Evidence from a Climate Change Experiment."
Environmental and Resource Economics, 52(2), 2012, 195-212.
Harrison, G. W., and J. A. List. "Field Experiments."
Journal of Economic Literature, 42, 2004, 1009-55.
Heyes, A., and S. Kapur. "Regulating Altruistic Agents."
Canadian Journal of Economics, 44(1), 2011, 227-46.
Heyman, J., and D. Ariely. "Effort for Payment: A Tale of Two
Markets." Psychological Science, 15(11), 2004, 787-93.
Kahneman, D., J. L. Knetsch, and R. Thaler. "Fairness as a
Constraint on Profit Seeking: Entitlements in the Market." American
Economic Review, 76(4), 1986, 728-41.
Kallbekken, S., S. Kroll, and T. L. Cherry. "Do You Not Like
Pigou, or Do You Not Understand Him? Tax Aversion and Revenue Recycling
in Lab." Journal of Environmental Economics and Management, 62(1),
2011, 53-64.
Landry, C. E., A. Lange, J. A. List, M. K. Price, and N. G. Rupp.
"Is There a 'Hidden Cost of Control' in Naturally
Occurring Markets? Evidence from a Natural Field Experiment." NBER
Working Paper 17472, 2011.
Meier, S. "Do Subsidies Increase Charitable Giving in the Long
Run? Matching Donations in a Field Experiment." Journal of the
European Economic Association, 5(6), 2007, 1203-22.
Meier, S., and A. Stutzer. 2008. "Is Volunteering Rewarding in
Itself?" Economica, 75(297):39-59.
Mellstrom, C., and M. Johannesson. "Crowding Out in Blood
Donation: Was Titmuss Right?" Journal of the European Economic
Association, 6(4), 2008, 845-63.
Mintel. "Many Consumers Yet to Click with Online Grocery
Shopping." Press Release, 2009. Accessed March 6, 2011.
http://www.mintel.com/press-centre/pressreleases/400/
many-consumers-yet-to-click-withonline-grocery-shopping.
Noussair, C., S. Robin, and B. Ruffieux. "Do Consumers Really
Refuse to Buy Genetically Modified Food?" Economic Journal, 114,
2004, 102-20.
Schnedler, W., and R. Vadovic. "Legitimacy of Control."
Journal of Economics and Management Strategy, 20(4), 2011, 985-1009.
Sliwka, D. "Trust as a Signal of a Social Norm and the Hidden
Costs of Incentive Schemes." American Economic Review, 97(3), 2007,
999-1012.
UK Office for National Statistics. Socio-economic
Classification-User Manual. Basingstoke, UK: Palgrave Macmillan, 2005.
(1.) See for example, Andreoni (1989), Frey (1997), Meier and
Stutzer (2008), and Corgnet (2012).
(2.) For surveys of the literature on motivation crowding see
Bowles (2008) and Frey and Jegen (2001).
(3.) See for example, Frey and Oberholzer-Gee (1997), Gneezy and
Rustichini (2000a, 2000b), Heyman and Ariely (2004), Meier (2007),
Mellstrom and Johannesson (2008), Charness and Gneezy (2009) and Ariely
et al. (2009).
(4.) See for example, Frey (1993), Fehr and Schmidt (2000), and
Falk and Kosfeld (2006) and Arce (2007) for an evolutionary explanation.
(5.) Landry et al. (2011) tested for crowding out in naturally
occurring labor markets and do not find crowding out.
(6.) As we study only one context (grocery shopping) we are unable
to distinguish between a reduction in the underlying intrinsic
motivation to mitigate climate change and a reduction in the extent to
which this becomes activated in a particular context. Previous
contributions did not require this distinction as they studied direct
contributions to the public good (donations, fundraising effort,
purchases of emission allowances).
(7.) See Bern (1967).
(8.) Sainsbury's has a market share of 27% in the study area
(http://www.j-sainsbury.co.uk/index.asp?pageid=451), and 16% in the
United Kingdom. It has a well-developed internet shopping facility that
reaches 88% of the UK population, with over 500 millions [pounds
sterling] worth of sales in 2009 (http://www.j-sainsbury.co.uk/files/
reports/ar2009_report.pdf, page 5. This value only includes food and
grocery products, as the nonfood area has only been launched in 2010).
(9.) The stores were chosen by moving along the circumference of a
radial approximately 7 miles from the center of London, and provide a
geographically dispersed sample from the London metropolitan area.
Stores were located in Walthamstow, New Barnet, Edgware, Chiswick,
Merton, and Lewisham. Each was surveyed for 8 hours a day for 2 days,
with the exception of Edgware, where the experiment lasted 4 days.
(10.) In the classification of Harrison and List (2004) the
experiment is between a framed and a natural field experiment. It
involves a non-standard subject pool (supermarket customers), field
context in commodity (grocery products), task (purchasing these
products) and information (on prices, availability and characteristics
of products) and is conducted in an environment where subjects usually
undertake this task (in the supermarket during the actual shopping
trip). However, subjects were aware that they were participating in an
experiment.
(11.) Online grocery shopping is prevalent in the United Kingdom
with over a third of adults shopping for food online in 2009, and a
significant proportion of UK consumers do most of their grocery shopping
in this manner (Mintel 2009).
(12.) The same experimenter was present during the entire
experiment which together with the frequent change of location minimized
the risk of the same person participating more than once.
(13.) Compensation was paid only if participants completed the
computer-based experiment, had their choices recorded on a paper slip,
signed off by the experimenter, and returned to the experimenter
(stationed at the store's exit) with their slip, purchasing
receipt, and shopping basket at the end of the shopping trip. In the
case of the Subsidy and the Neutral Price Change treatments,
participants also received the difference between the price announced in
the computer-based part of the experiment and the real (unaltered)
in-store prices if they purchased any subsidized option.
(14.) The dummy variables Envinfo and Nutrinfo recording whether a
participant has opted to see the corresponding type of information have
been used as controls in the regressions reported in Section IV but
coefficients are not reported as they are not significant at the 10%
level.
(15.) Participants who selected more than one product category
completed the first choices for all categories before moving to the
second choices and were subject to the same treatment throughout.
(16.) A graphical representation of the label can be found in the
experimental instructions provided in Appendix SI. We wish to thank the
Carbon Trust for their cooperation and assistance.
(17.) In contrast to the Labeling treatment the Subsidy explicitly
provided only information on the relative ranking of products (clean vs.
dirty) as would be the case in a real world subsidy. However, the size
of the subsidy was proportional to the difference in carbon footprints
within each product category.
(18.) Note that for the carbon footprint of a product to have an
impact on the consumption choice of a consumer, she needs to care not
only about the environmental public good affected (mitigation of climate
change) but also has to have at least some form of intrinsic motivation
to contribute to this cause. Intrinsic motivation is a necessary
condition as the relevant population of the climate mitigation game is
in the order of billions and hence close enough to infinity to drive the
Nash contribution down to zero (Andreoni 1988).
(19.) Although the carbon footprints for milk were taken from Tesco
(another major supermarket in the United Kingdom) they only used
footprint labels (similar to those in this study) on the following four
product categories: orange juice, potatoes, light bulbs, and laundry
detergents. No carbon footprint labels were in use at Sainsbury's
at the time the study was conducted in any of the product categories
included in the experiment.
(20.) Note that it does not matter whether consumers perceive the
clean product to become better or the dirty product to become worse. All
that matters is the change in relative terms.
(21.) The carbon label might induce an experimenter demand effect.
To reduce confounding information on nutritional characteristics was
introduced at the same time as the carbon footprint. Moreover, the tests
of Hypothesis 3 suggest that experimenter demand effects are not driving
our results.
(22.) Here and elsewhere we ignore changes from one dirty (clean)
to another dirty (clean) variety, e.g., from Coca Cola in cans to Pepsi
Cola in cans. Only switches that have an impact on carbon footprints are
considered as switching.
(23.) The large number of participants (177 across all three
treatments) choosing clean products in the first choice (for cola, meat,
and butter/margarine) is most likely owing to participants who entered
the experiment because they were willing to purchase one of the above in
exchange for the 5 [pounds sterling] voucher. While this suggests that
some participants might have pursued a strategy that minimized
expenditure, this cannot explain any of our results as there is no price
change in the Labeling treatment, and the Neutral Price Change and
Subsidy treatments did not move the cheapest item from dirty to clean
(or vice versa).
(24.) The Bhapkar (Bhapkar 1966) test of marginal homogeneity is a
more powerful alternative to the Stuart-Maxwell test (Keele 1982). They
are asymptotically equivalent.
(25.) The sample used in this regression is restricted to
observations with "dirty" baseline choices in all product
categories.
(26.) While there is no significant difference between switching
frequencies in the Neutral Price Change and the Labeling treatments, the
null hypothesis that the coefficients of the Labeling and the Subsidy
dummies are the same is rejected at the 1% level (p = .0088).
(27.) The Labeling treatment was run in three different versions
which differed in the carbon footprints provided (only the version with
the true footprints is reported in this paper). There was no systematic
effect of the scale of potential GHG reductions on participants
switching behavior. Furthermore, all results presented on the Labeling
treatment would hold if all three versions of it would have been
included.
(28.) This is in line with evidence on the effect of labeling in
other contexts (Bjdrner, Hansen, and Russell 2004; Noussair, Robin, and
Ruffieux 2004).
(29.) This complements findings by Kallbekken, Kroll, and Cherry
(2011) who establish an aversion against interventions labeled as a
"tax" compared to an equivalent with a different name. These
findings do not contradict results by Chetty, Looney, and Kroft (2009)
as they compare changes in taxes already included in the product's
price with taxes that are added at the check-out.
GRISCHA PERINO, LUCA A. PANZONE, and TIMOTHY SWANSON *
* We are grateful to Denise Leung (University College London), who
did excellent work in the data collection process, and the supermarket
chain Sainsbury's for permission to run the experiment in their
stores. We are indebted to the FP7 project EU-POPP (www.eupopp.net) for
providing the funding required for the field experiment. We gratefully
acknowledge comments from the editor, two anonymous referees,
participants of EU-POPP project meetings, seminar participants at the
universities of Beijing, East Anglia, Heidelberg, Luneburg, and
Manchester, and participants at the 2011 AERE and EAERE annual
conferences.
Perino: School of Business, Economics and Social Sciences,
University of Hamburg, Welckerstr. 8, Hamburg 20354, Germany. Phone +49
40 42838-8767, E-mail grischa.perino@wiso.uni-hamburg.de
Panzone: Sustainable Consumption Institute. University of
Manchester, 188 Waterloo Place, Oxford Road, Manchester, M13 9PL, UK.
Phone +44 (0)161 27 54275, Email luca.panzone@manchester.ac.uk
Swanson: Centre for International Environmental Studies, Graduate
Institute Geneva, PO Box 136, 1211 Geneva 21, Switzerland. Phone +41 22
908 62 17, E-mail tim.swanson@graduateinstitute.ch
TABLE 1
Products Offered in the Experiment
Product Carbon Footprint
Categories Quantity Options (in C[O.sub.2]e)
Cola 2 1 Cans (six pack) 1.020 g
PET bottle 500 g
Milk 2 pint Whole 1.800 g
Semi-skimmed 1,600 g
Skimmed 1,400 g
Meat Various Beef 16,000 g/kg of beef
weights
Chicken 5,000 g/kg of chicken
Spreads 500g Butter 11.900 g
Margarine 675 g
Product Original Subsidy/Price
Categories Price Reduction
Cola 2.63 [pounds sterling]-
2.69 [pounds sterling]
1.56 [pounds sterling]- 5 P
1.59 [pounds sterling]
Milk 0.86 [pounds sterling]
0.86 [pounds sterling] 3 P
0.86 [pounds sterling] 6 p
Meat 4.40 [pounds sterling]-
7.93 [pounds sterling]
per 1 kg
3.39 [pounds sterling]- 21 p/kg
9.16 [pounds sterling]
per 1 kg
Spreads 1.70 [pounds sterling]-
2.76 [pounds sterling]
1.00 [pounds sterling]- 43 p
2.38 [pounds sterling]
Product Brands/Varieties
Categories
Cola Coca Cola, Pepsi Cola, Diet Coke,
Diet Pepsi, Coke Zero, Pepsi Max
Milk Sainsbury's own brand fresh milk
Meat Minced meat, casserole steak,
Chicken breast, mini chicken
fillet, drumsticks
Spreads Lurpak, Anchor, Countrylife,
Kerry gold, Sainsbury's own brand
Lurpak, Anchor, Flora, Clover,
Sainsbury's own brand
TABLE 2
Relative Frequency of Switching from a Dirty to a Clean
Variety (by Product Category)
Neutral Price
Labeling (%) Subsidy (%) Change (%)
Cola 36.4 27.3 58.7
Butter/margarine 61.5 16.1 20.0
Meat 21.1 12.0 14.8
Milk 12.0 8.5 13.4
All products 23.7 13.2 21.5
TABLE 3
Test Results on Equality of Switching Frequency
across Treatments
Fisher's Test of
Exact Test Proportions
Subsidy versus Neutral 0.011 0.009
Price Change
Subsidy versus Labeling 0.021 0.017
TABLE 4
Logit Regressions on Drivers of Switching from Dirty to
Clean Products in Labeling, Neutral Price Change (baseline),
and Subsidy Treatments
(1) Twenty Controls (a)
Subsidy -0.687 ** (0.011)
Subsidy x Log of Income
Labeling 0.253 (0.433)
Milk -1.864 (0.000)
Meat -1.548 *** (0.000)
Butter -0.962 *** (0.004)
Number of children -0.295 ** (0.034)
in household
Male -0.0603 (0.808)
Unemployed -0.713 (0.286)
Retired 0.379 (0.627)
Log of Income 0.0979 (0.589)
Observations 666
Pseudo [R.sup.2] 0.133
[chi square] 79.08
(2) Eight
Controls (a)
Subsidy -0.759 *** (0.003)
Subsidy x Log of Income
Labeling
Milk -1.872 *** (0.000)
Meat -1.572 *** (0.000)
Butter -0.979 *** (0.004)
Number of children -0.225 ** (0.046)
in household
Male
Unemployed
Retired
Log of Income
Observations 666
Pseudo [R.sup.2] 0.126
[chi square] 72.86
(3) Interaction Term
Subsidy 1.409 (0.205)
Subsidy x Log of Income -0.669 * (0.052)
Labeling
Milk -1.899 *** (0.000)
Meat -1.583 *** (0.000)
Butter -1.050 *** (0.002)
Number of children -0.213 * (0.068)
in household
Male
Unemployed
Retired
Log of Income 0.268 (0.171)
Observations 666
Pseudo [R.sup.2] 0.132
[chi square] 75.75
Notes: Dependent variable: Switching dummy (1 if participant
switches from dirty to clean product), clustered at
participant level, 493 independent observations, p values
in parentheses.
(a) Count excludes treatment dummies and a constant. List
of variables and descriptive statistics in Appendix A.
* p < .10; ** p < .05; *** p < 0.01.
