Environmental Rebound Effects from Green Consumption Choices

This paper examines rebound effects from cost-saving 'green' consumption choices using Australian data. It estimates the size of rebound effects for reduced vehicle use, reduced electricity use, changing to smaller passenger vehicles, and utilizing fluorescent lighting. The study finds that rebound effects are higher when more efficient vehicles or lighting are utilized rather than simple conservation actions of forgoing use. Additionally, lower income households have higher rebound effects, suggesting that environmental policy directed at changing consumer behavior is most effective when targeted at high income households.

MPRAMunich Personal RePEc Archive
What if consumers decided to all ‘go
green’ ? Environmental rebound effects
from consumption decisions
Cameron K. Murray
University of Queensland
July 2012
Online at https://mpra.ub.uni-muenchen.de/40405/
MPRA Paper No.40405, posted 1.August 2012 01:09 UTC

What if consumers decided to all ‘go green’ ?
Environmental rebound effects from consumption decisions
Cameron K. Murray1,∗
University of Queensland,St Lucia, Queensland, Australia, 4067
Abstract
Shifting consumer preferences towards ‘green’consumption is promoted by many governments
and environmentalgroups.Rebound effects,which reduce the effectiveness ofsuch actions,are
estimated for cost-saving ‘green’ consumption choices using Australian data.
Cases examined are:reduced vehicle use,reduced electricity use,changing to smaller passenger
vehicles, and utilising fluorescent lighting.It is found that if rebound effects are ignored when eval-
uating ‘green’ consumption, environmental benefits will be overstated by around 20% for reduced
vehicle use,and 7% for reduced electricity use.Rebound effects are higher,and environmental
benefits lower,when more efficient vehicles or lighting are utilised rather than simple conserva-
tion actions offorgoing use.In addition,lower income households have higher rebound effects,
suggesting that environmentalpolicy directed at changing consumer behaviour is most effective
when targeted at high income households.Additionally,an inherent trade-off between economic
and environmental benefits of ‘green’ consumption choices is demonstrated.
The size of the rebound effect, and the observed variation with household income, is attributed to
life-cycle analysis (LCA) methodologies associated with the calculation of embodied GHG emis-
sions ofconsumption goods.These results should be therefore be interpreted as the minimum
rebound effect to include in policy evaluation.
Keywords: Rebound effect; conservation; household consumption; greenhouse emissions
∗Corresponding author
Email address:[e-mail] (Cameron K. Murray )
1ph. [phone], m.+61422 144 674
Preprint submitted to Elsevier July 31, 2012

1. Background
More sustainable consumption patterns are promoted by governments, environmental groups and
internationalagencies as a measure to combat environmentaldegradation (UN,1992;OECD,
2002).Efforts to reduce resource consumption, including energy consumption and the associated
negative externality of greenhouse gas (GHG) emissions, through household consumption choices
are attractive due to the ability for win-win outcomes;where cost saving ‘green’behaviour si-
multaneously leads to environmentalbenefits.If consumer preferences can be nudged towards
less every and resource intensive consumption,through information and marketing campaigns,
then environmentalexternalities can be party corrected by being brought into consumer utility
functions.
Before promoting an agenda to nudge consumer preferences towards ‘green’consumption,the
scale of potential environmental benefits should be understood.It is commonly assumed that high
rates of adoption of win-win ‘green’ consumption choices will significantly reduce GHG emissions.
However,this assumption is typically made using incomplete engineering-type analysis,where
many little actions are expected to add up to significant economy wide changes.This type of
calculation ignores economic rebound effects.
Rebound effects describe the flow-on effects from technology and consumption pattern changes that
offset intended environmentalbenefits.In the context of cost-effective new technology,rebound
effects are generally classified as direct,indirect,or economy wide (Sorrelland Dimitropoulos,
2008).Direct effects (or price effects) occur when new technology decreases the effective price of a
good or service, and consumers adapt by consuming more of that good or service.Indirect effects
(or incomes effects) occur when reduced costs of a good or service lead to increased consumption of
other goods and services, which themselves have embodied GHG emissions.Finally the economy-
wide effect considers these two effects,plus changes to the scale and composition of production
economy-wide, including the emergence of new products and services.
One widely held view is that the indirect effect with respect to GHG emissions is smalldue
to energy inputs comprising a smallcomponent ofhousehold expenditure (Lovins et al.,1988;
Schipper and Grubb,2000). This view is gradually being eroded.Recent studies utilising life-
cycle assessment (LCA) of embodied GHG emissions show that the amount of energy consumed
indirectly by households is often higher than energy consumed directly through electricity,gas,
and motor fuel,and is a growing proportion (Vringer and Blok, 1995, 2000; Vringer et al., 2007;
Lenzen, 1998; Lenzen et al., 2004; Weber and Perrels, 2000; Reinders et al., 2003)
Few studies explicitly or implicitly estimate the magnitude of the indirect rebound effect (Chalkley
et al., 2001;Lenzen and Dey,2002;Alfredsson,2004;Brannlund et al.,2007;Mizobuchi,2008;
2

Druckman et al.,2011).Since the rebound effect is expressed in terms ofa particular resource
or externality,estimates ofthe indirect effect require an estimate ofthe embodied resources in
household consumption.The scarcity of embodied resource data is one reason for slim body of
research,a point emphasised in Kok et al.(2006) reviewed 19 studies ofembodied energy and
GHG emissions from consumption patterns,finding only three that provided sufficient detailto
allow econometric estimation of the indirect effect.Given the variation in the embodied energy
and GHG emissions due to different composition of national energy sources, one would expect the
some cross-country variation of indirect rebound effects.
The rebound effect literature is also heavily focused on improvements in energy-efficient technology
and centres on the possibility of an economy wide backfire, where rebound effects are larger than
engineering estimates ofenvironmentalbenefits (Saunders,2000;Inhaber,1997;Alcott, 2005;
Hanley et al., 2008).
The most promising area for demand-side environmental policies to have short-run pay-offs is not
new technology,but the adoption of ‘green’consumption choices in the absence of any changes
to technology.In this context, better targeted nudging of consumer preferences may improve the
environmentalpay-off of such policies.But a better understanding of nature of rebound effects,
and therefore the potential size of environmental benefits, is required to achieve this aim.
2. Rebound effects from pure consumption choices
This paper considers the scenario where technology and product choice are fixed,and only con-
sumer preferences change.In the absence of technology changes, cost-saving ‘green’ consumption
choices are subject to rebound effects when liberated purchasing power is utilised for additional
consumption.A household with new ‘green’ preferences choosing a smaller but more fuel-efficient
car may be tempted to drive further,and will spend the resulting cost savings elsewhere in the
household budget.
Specific definitions ofdirect and indirect rebound effects are required in the context ofpure
consumption choices with fixed technology.The direct effect is the offsetting environmental impact
that occurs when cost savings are spent on the commodity from which they are saved.For example,
changing to a fuelefficient smallcar will generate savings for the household from fuelexpenses,
but driving further willincur some offsetting fuelcosts. The indirect effect is the offsetting
environmental impact from the increased spending in other areas of the household budget.
In this paper,consumption choices that involve new household capital,such as new appliances
or vehicles, are referred to as because the household service, of lighting or transport, is produced
3

more cheaply (ignoring any reductions in quality).In these scenarios one would expect both direct
and indirect effects.
A household that goes ‘green’ without changing its capital stock, but instead chooses a conservation
approach, such as replacing driving with cycling or utilising electrical appliances more sparingly,
the indirect rebound effect will be the total rebound effect.These scenarios are pure conservation
choices.Distinguishing between the two types of behaviour provides a clearer picture of net effect
of household choices,at least in the short run,and the types of consumption choices that have
better associated environmental benefits.
Existing estimates ofrebound effects from consumption choices alone are limited.Alfredsson
(2004) estimates rebound effects from ‘green’ consumption choices of 14% for transport abatement,
and 20% for ‘green housing, and a back-fire (approximately 200%) for a ‘green diet and a rebound
effect for these combined actions of20% in terms ofGHG emissions.Additionally,the impact
of increasing incomes was shown to offset any benefits made by consumption pattern changes.
Specifically,exogenous income growth of 1% per year offsets allbut 7% of the decrease in GHG
emissions from the combination of changes by 2020,while income growth of 2% willmore than
compensate for consumption pattern changes,and lead to a 13% increase in GHG emissions by
2020.
Lenzen and Dey (2002) account for the rebound effect from a change to a cheaper low carbon diet,
with estimates around 50%.Druckman et al. (2011) estimate the direct and indirect rebound effect
for three abatement actionshousehold energy reduction,more efficient food consumption (less
throw-away food), and reduced vehicle travelwith results showing a 7%, 59% and 22% rebound
effects respectively in terms of GHG emissions.
One common feature of these studies is that the rebound effect model allows for re-spending on the
goods from which the saving where made,meaning in allscenarios have a direct rebound effect.
For example,Alfredsson (2004),Lenzen and Dey (2002) and Druckman et al. (2011) use models
where households who adopt a green diet then proceed to spend a portion of the cost savings on
the previous diet.Whether this has a materialimpact on the estimates is uncertain,but it is
one area where this paper improves the estimation of rebound effects, and enables differentiation
between pure conservation choices and efficient consumption choices.
Of particular interest is the potentialfor variation in the magnitude ofthe rebound effect for
consumption choices across the different household incomes.One might expect that since there is
a trade-off between direct and indirect effects,and direct effects have been observed to diminish
with rising incomes that indirect rebound effects may increase with rising income levels (Baker
et al., 1989; Milne and Boardman, 2000; Roy, 2000).Yet LCA data of embodied GHG emissions
4

suggests that the opposite might be true due to the decrease in GHG intensity of luxury goods
(Lenzen et al., 2004, 2006; Hertwich, 2005).
Existing studies ofhousehold energy use suggest that rebound effects may be much higher in
households,and in countries,with low incomes,due to energy fuels comprising a larger share of
the household budget (Baker et al.,1989;Milne and Boardman,2000;Roy, 2000;Hong et al.,
2006).This evidence points to indirect effects becoming more significant than direct effects over
time and with increasing incomes.
Girod and De Haan (2010) suggest that evaluation of GHG emissions from consumption may over-
state emissions at high income levels due to increasing quality.The basic argument is that twice
the expenditure on a product does does purchase twice to physicalquantity.Yet these elevated
prices are either reflective ofincreased inputs to production,or some form ofrent transferred
to the producer,meaning that this assertion remains unclear at a macro levelVringer and Blok
(1995).
Indeed,as a backdrop to this literature,there remains a major concern about the limitations
of LCA data due to necessary boundary specifications ofinputs and outputs ofthe economy.
Variation in GHG intensity of consumption goods ultimately determines the size of the rebound
effect and the net environmentalbenefit ofconsumption choices.Typically LCA data shows a
trade-off between labour and energy intensity (Maddala,1965;Karunaratne,1981;Lenzen and
Dey, 2002),yet the supply oflabour into the production process requires the consumption of
other commodities.
To overcome these truncation errors, with the assumption that labour input is merely a transfer,
Costanza (1980) estimated the embodied energy of a number of economic outputs with alternative
system boundaries.This method greatly reduced variation in energy intensity across outputs,
leading to the observation that ”there is a strong relationship between embodied energy and
dollar value”.Within Costanza’s (1980) framework, consumption pattern changes would provide
no net changes to energy consumption or GHG emissions.Indeed,the only way for household
to reduce their GHG emissions would be to reduce their income at the same time as reducing
expenditure through conservation behaviour, as suggested by Madlener and Alcott (2009).
With this in mind, this paper builds on these handful of studies of rebound effects from consump-
tion choices.In particular,pure conservation choices by households are modelled with indirect
effects only,and the sensitivity of the scale of the rebound effect to household incomes is exam-
ined. The differential benefits of ‘green’ consumption across the income range may provide clues
to better targeted policy,particularly in light of the relative attractiveness of cost-saving ‘green’
choices for low income households.Indeed,any difference in effectiveness between efficiency and
5

conservation behaviour,and the additive effects ofmultiple ‘green’choices,are also important
policy considerations.
3. Methodology
Rebound effects are generally expressed as the amount ofenergy,resources or externality,gen-
erated by offsetting consumption,as a percentage ofpotentialreductions where not offsetting
consumption occurs (Berkhout et al.,2000). Measuring this baseline potentialreduction is a
critical factor in determining the scale of the rebound effect.
For example,an engineering estimate that converts per unit ofservice reductions in electricity
consumption of a more efficient appliance to kWh, then converts that into GHG emissions based
on transmission loss, electricity generation efficiency and the emissions per unit of coal combusted,
is flawed.(Lovins et al.,1988;Weizacker et al.,1998) The totalembodied energy in the more
efficient appliance (replacement capital) should be subtracted from the potential energy reductions
to determine the baseline,as this embodied resource consumption is necessary and inseparable
from the new appliance itself.This contrasts the position ofSorrelland Dimitropoulos (2008)
who propose that the embodied resource requirements of replacement capital comprise part of the
rebound effect.In this paper, the baseline potential reduction of GHG emissions is calculated as
the cost saving multiplied by the GHG intensity of that expenditure.Offsetting GHG emissions
are those embodied in the consumption of goods enabled by the cost savings.
3.1. Data
The 2003-4 Australian Bureau of Statistics (ABS) Household Expenditure Survey (HES) of 6,957
households aggregated into 36 commodity groups is used in this paper (ABS,2004).The corre-
sponding embodied GHG emissions for each commodity group, calculated using an input-output
based hybrid method, was made available from the Centre for Integrated Sustainability Analysis,
Sydney (Dey, 2008) (Appendix A).
Matching the two data sets shows decreasing emissions intensity with household income level,
but increasing quantity of emissions.No evidence of an environmentalKuznets curve for GHG
emissions is observed, which corresponds with the macroeconomic relationship between energy, or
greenhouse emissions,and gross domestic product typically seen in household emissions studies
(Holtz-Eakin and Selden, 1995; Schipper and Grubb, 2000; Greening, 2001; Lenzen et al., 2004).
3.2. Household demand models
The rebound effect model is based on a system of household demand equations where expenditure
on each commodity is dependent on total expenditure as a proxy for the household income level, as
6

is common in household demand studies (Deaton and Muellbauer, 1980; Haque, 2005; Brannlund
et al., 2007).Housing expenditure is excluded from data, as is expenditure on tobacco and health
services, which are not expected to be susceptible to the incorporation of environmental damage
into the utility function.Additionally,savings rates are assumed to be constant (saving reduced
costs would lead to greater future consumption in any case).
Selection ofa functionalform of the household demand system requires the ability to assess
the potentialvariation in the rebound effect at different income levels,and as such,the system
should allow for the possibility of threshold or saturation levels,where goods become inferior at
particular income levels.For completeness,four different household demand models are used to
generate parameters that feed into the estimation of the rebound effect;
1. a basic double semi-log (DSL) specification,
2. a DSL specification with non-income explanatory variables (DSL2),
3. the Working-Leser (WL) model of budget shares, and
4. a linear model.
Utilising this array of common functional forms also sheds some light on the sensitivity of estimates
to the statistical methods applied, which may be an important practical consideration for policy
assessment.The basic DSL is of the form
qi = αi + βi Y + γi ln Y (1)
where qi is the expenditure in on each of the 36 i commodities, and Y is total expenditure.The
extended DSL model in Equation 2 includes non-income explanatory variables;age of household
reference person, A, number of persons in the household, N , state, S, degree of urbanity, U , and
dwelling type,D, each ofwhich have been previously shown to have an impact on household
emissions (Lenzen et al.2004; Vringer et al.2007).
qi = αi + βi Y + γi ln Y + θ1i N + θ2i U + θ3i S + θ4i A + θ5i D (2)
The WL model relates budget shares, rather than expenditure, linearly with the logarithm of total
expenditure.The budget share, w, of each i commodity is calculated by
wi = qi
Y (3)
then the relationship
7

wi = αi + βi ln Y (4)
is estimated.The functional from of the Engel curve from the WL model is then determined by
substituting equation (3) into (4) to get
qi = αi Y + βi Y ln Y (5)
Appendices B through E provide results of these model regressions.In both DSL models, Whites
heteroskedasticity consistent method of calculating standard errors and covariance is used, while
for the linear and WL model, ordinary least squares is used with no further statistical adjustment.
It is important to note that total expenditure is a significant variable for every commodity group.
This validates to some degree the income determinism assumption underpinning these models.
The significance levels observed for the non-income explanatory variables in the extended DSL2
modelalso provide evidence that these household characteristics are important determinants of
expenditure choices.In the domestic fuel and power and vehicle fuel commodity groups, the most
GHG intensive expenditure groups,almost allof the non-income variables are significant.Most
other results follow intuitive logic.
3.3. Rebound effect model
The marginalbudget share (MBS),or the amount ofextra expenditure on commodity i for an
increase in totalexpenditure of one dollar,is utilised in the rebound estimation modelbased on
estimated coefficients from the household demand model.For each of the functionalforms used
in this study, the MBS for each i commodity is as follows:
DSL MBS i = βi + γi
Y
Linear MBS i = βi
WL MBS i = αi + β log Y + βi
Two alternative models are used for estimating the rebound effect.The first applies to efficient
consumption choices,an efficiency model,where although technology is fixed there are cheaper
energy efficient alternatives currently available for providing similar household services.Given that
technology is unchanged, there is a sacrifice in the quality of service, such as passenger kilometres,
that accompanies the reduction in price.In such cases,the direct effect,caused by the income
effect but excluding the substitution effect, will be considered.New ’green’ consumer preferences
cause the change in household capital, so estimating prices effects based on the ’ungreen’ sample,
and ignoring quality changes, would be misleading to some extent.
8

The second applies to conservation choices, a conservation model, which only allows increases in
expenditure on the goods or services from which cost savings were not made.For example,an
individual who chooses to cycle instead of drive is unlikely to use any cost savings to drive further.
And even ifthey did,it would simply be a reduction in the conservation measure,and not a
rebound effect.Existing studies typically do not controlfor this in their models,meaning that
unlikely behaviour such as cost savings from electricity conservation being spent on more electricity
is a common outcome (Alfredsson, 2004; Brannlund et al., 2007; Druckman et al., 2011).
Denoting cost savings from ‘green’consumption choices X,then for commodity s from which
savings are made, the new expenditure in the efficiency model is
Qsnew = Qsold − X + X.MBS s (6)
while for all other for other i commodities in the household budget the new expenditure, Qi new , is
Qi new = Qi old + X.MBS i . (7)
In the conservation model the new expenditure on the conserved commodity is
Qsnew = Qsold − X (8)
while for allother i commodities the new expenditure levelensures Walras’Law by reallocating
the expected M BSs across all other commodities.
Qi new = Qi old + X.MBS i +
∞X
n=1
X.MBS n
s .MBS i . (9)
To estimate the change in GHG emissions from the change in consumption patterns,the expen-
diture in each commodity group is multiplied by the GHG intensity ofthat commodity.Since
there are no technology changes applicable to production stages of the economy, the same embod-
ied emissions data can be used in both the before and after scenario without concerns regarding
changing production patterns in the economy.
In the resource generic form of Lenzen and Dey (2002),if the overallembodiment of resource f
(in this case GHG emission),for commodity i,is Rf,i , then the totalembodiment off for all
consumption is
f = X Qi Rf,i (10)
9

The potentialresource savings,or the denominator ofthe rebound effect,are calculated as X
multiplied by the embodied factor Rf for commodity s.The rebound effect for resource f can
then be expressed as a percentage of the potential resource savings, as
RE = (X.R f,s ) − (
P Qi old Rf,i − P Qi new .Rf,i )
X.R f,s
(11)
which simplifies to
RE = 1 −
P Qi old Rf,i − P Qi new .Rf,i
X.R f,s
(12)
Conservation and efficiency model are generated by using the two alternative Qnew calculations.
Further, each model is estimated using the four functional forms of the household demand system.
Importantly,in this modelthe rebound effect is a function ofthe totalexpenditure level(as a
proxy for income) and it is expected that a degree of variation will be observed across the income
range.
3.4. Cases
3.4.1. Vehicle fuel
Driving less,or choosing a smaller fuelefficient vehicle,are widely promoted choices households
can make to reduce GHG emissions (Foundation,2007;Government,2007). Both vehicle fuel
cases (conservation and efficiency) have been developed to represent the same baseline reductions
in fuel use and GHG emissions.
To ensure feasibility at all income levels, the efficiency case allows for the replacement of passenger
vehicles (household capital) with no change in capital cost.Evidence suggests that replacing the
average Australian passenger vehicle on the second hand market with one that uses 4L/100Kms
less fuelis possible without increased capitalcosts,by sacrificing size and quality (:20,2008;
of Consumer and Protection, 2008).Other input includes include the average number of kilometres
driven by Australian household per year,at approximately 13,900kms in 2003-04,and the price
of fuel at $0.90 per litre (ABS, 2006; of Consumer and Protection, 2008).
Further to the savings on motor fuelitself,there are cost savings on complementary goods such
as vehicle registration,tyres and servicing.The registration cost difference between a four and
six cylinder car (the most likely vehicle substitute) in Queensland is $111.95 (Transport,2008).
A saving of $50 has been assumed for the reduction in associated servicing and running costs per
year. Combining these figures to construct the cost savings for the efficiency case is shown in
Table 1.
10

Table 1:Vehicle fuel case details
Case study changes Old New Consumption category Annual saving
Fuel economy (L/100km)11 7 Motor vehicle fuel $500
Annual Kms travelled 13,900 13,900 Vehicle reg.and insurance $111
Registration costs $363 $251 Parts and Accessories $50
Servicing costs $250 $200 Total $661
The conservation case has the same fuel use reduction as the efficiency case.This could occur by
replacing driving with cycling,car pooling,or any other means with which a household reduces
driving from the Australian average of 13,900kms per year to 8,720kms per year.Reducing vehicle
mileage willreduce non-fuelrunning costs and improve the economic pay-off for this household
choice.It is therefore of interest to estimate the rebound effect with and without additional cost
savings, which are assumed for the sake of the exercise to be equal those from the efficiency case.
In both cases,the reduced motor fuel use gives a baseline potential GHG emissions reduction in
both the efficiency and conservation cases of 1,300kg CO2−e per year.
3.4.2. Household electricity
Numerous behaviouralchanges,including changing the stock of household electricalappliances,
can save money and decrease electricity use.In terms of conservation choices,cost savings from
behaviouralchanges such as shorter showers (where there is electric water heating),turning off
lights when leaving a room, and turning off stand-by appliances can save a typical household $100
per year (Foundation, 2007).
For an efficiency case that involves replacing household capital, this analysis uses the replacement
of incandescent light bulbs with compact fluorescent lightbulbs (CFL), which are a cost effective
option.CFLs can produce the equivalent lighting of an incandescent bulb that requires five times
more power.In Australian supermarkets incandescent bulbs cost between $0.39 and $0.59 for a
75W globe while CFLs cost between $4.49 and $6.29 for a 15W bulbs .For simplicity,a cost of
$0.50 and $5.00 is assumed in this case for incandescent and CFLs respectively.The increased
lifespan of CFLs must be considered, which is widely claimed to be around ten times longer than
incandescent bulbs (Mirabella, 2008).A 10,000 hour life is assumed for compact fluorescents, and
1,000 for incandescent bulbs in this case study.Residential electricity price adopted is 17.10c per
kilowatt-hour for tariff 11,which was the rate for generalpower and lighting in Queensland in
2003 (Lucas, 2003).Finally, it is assumed that ten 75W bulbs are replaced by the household and
that each bulb is used for 2 hours per day.Taken together these assumptions generate a scenario
where that capitalcost oflighting per period is equal,and the cost savings arise from $75 less
11

electricity use per year with potential GHG emissions reductions of 550kg CO2−e.
A scenario where a household adopts both of the vehicle fuel and electricity cases, in either their
efficiency or conservation form, was also estimated.
4. Results
In the following sections,rebound effect estimates are presented graphically across a $300 to
$1,200 per week household income range.All DSL2 model results are with mean values for other
non-income household explanatory variables
4.1. Vehicle fuel
Rebound effect estimates for the vehicle fuelcases are in Figure 1.Depending on the household
demand model applied, the rebound effect in the conservation case is somewhere between 12 and
17% at the median household income level($37,400 per year,or $717 per week),reducing the
expected environmental benefit of 1,330kg CO2−e per year to the range of 1,090 -1,100kg CO2−e.
The size ofthe rebound effect varies significantly by choice ofhousehold demand model.In all
non-linear models the rebound effect is lower at higher income levels.
Figure 1:Rebound effects from vehicle fuel cases
12

In the efficiency case,the rebound effect is slightly larger,as expected,between 14 and 20% at
the median household income level,reducing the expected environmental benefit to the range of
1,065 - 1,140kg CO2−e.
Between a third and one fifth of the rebound effect is from direct effects due to increased vehicle
use in this scenario.The two DSL models show the direct effect falls as a proportion ofthe
indirect effect with increasing expenditure level, suggesting that it is more important to consider
the indirect effect at higher income levels, and vice-versa.
In the conservation case,adding additionalnon-fuelcost savings expected from reduced driving
increases the rebound effect significantly.Rather than a range of 12 to 17%,the range increases
to 15 to 22%.This supports the contention that a trade-off exists between the economic pay-off
of ‘green’ household consumption, and the environmental benefit.
Figure 2:Rebound effect for vehicle fuel conservation with/without additional cost savings
13

4.2. Household electricity
Rebound effect estimates for the ‘green’electricity cases are in Figure 3.The rebound effect in
the conservation case is between 4.5 and 6.5% at the median household income level, reducing the
expected environmental benefit from 550kg CO2−e per year to between 515 and 526kg CO2−e. As
with the vehicle fuel case, the size of the rebound effect varies significantly by choice of household
demand model, and with lower rebound effects at higher income levels for all non-linear models.
The electricity efficiency case has slightly higher rebound effects than the conservation case, in the
order of 5 to 7.5%.Only one quarter to one sixth of the rebound effect is from direct effects due
to increased electricity, with direct effects still clearly more important at lower income levels.
Figure 3:Rebound effects for electricity cases across household incomes
4.3. Combined case
For households undertaking combined conservation measures, the rebound effect is estimated at 12
and 14% near the median income (Figure 4).This reduces the expected GHG emissions benefits
from 1880kg CO2−e to between 1655 and 1615kg CO2−e per year.
In the efficiency case it is clear that direct effects are a larger proportion ofthe totalrebound
effect,at over half the indirect effect,especially compared to each case individually where direct
effects were a mere one fifth or sixth of the indirect effect.The direct effect is also more strongly
inversely related to household incomes in the combined case.
14

Figure 4:Rebound effects from combined case
The combined cases also offer some insight into the additive effects of green choices.The combined
conservation case shows greatly reduced variation ofthe rebound effect over the income range
compared to either case in isolation, while the combined efficiency case appears shows no change
in income variation.This is primarily due to the elimination ofthe two commodities with the
highest embodied GHG emissions, electricity and vehicle fuel, from the income effect.
Upon closer inspection, the rebound effect in the combined conservation case is slightly less than
one would expect from a simple addition ofthe individualcase results,meaning that the en-
vironmentalbenefits ofthe actions together is greater than in isolation (for example,the same
household making both choices, rather than one household doing each).The net effect of the com-
bined case, compared to the sum of each case, is shown in Panel (d) of Figure 4, with the combined
conservation case between 2 and 5% more effective for reducing environmental externalities than
the sum of each case.For the efficiency case however, the combined case has exactly the same net
environmental impact as the sum of the two individual cases.
5. Discussion
This rebound analysis of a series of ‘green’ household consumption case studies has demonstrated
that while consumption pattern changes can be an effective way for households to decrease their
GHG emissions, the results are lower than anticipated by engineering estimates.Depending on the
15

household demand models used and household income level,the highest rebound effect estimate
was around 20% in the case of adopting a more efficient vehicle,while estimates were as low as
4% in the electricity conservation case.
At the median household income level, the estimated rebound effect for vehicle fuel conservation
was approximately 12 to 17% (depending on the associated non-fuel cost savings), which is in line
with the 14% result of Alfredsson (2004),but not too far off the 22% result of Druckman et al.
(2011).One would expect that the variation in results has much to do with both the different
embodied GHG emissions pattern for the consumption goods across countries, and the structure of
the model.Australia’s electricity generation is almost purely from coal,for example,potentially
leading to higher embodied emissions per dollar ofelectricity consumption than the electricity
generation from more mixed sources in much ofEurope. For the electricity conservation cases,
the 5-8% result was far less than Alfredsson’s (2004) estimate,due to the fact that the current
conservation model does not allow responding electricity savings on more electricity consumption.
The result is more consistent with the 7% estimate of Druckman et al. (2011)..
The empirical results confirm that household income level is an important determinant of the scale
of the rebound effect.In both the conservation and efficiency models the total rebound effect, and
both the direct and indirect effects individually, were inversely related to household income level.
This is consistent with the findings in the literature of higher direct rebound effects for low income
households (Baker et al.,1989;Milne and Boardman,2000;Hong et al.,2006) and the implied
reduction in direct effects at high incomes due to a saturation ofdemand for household energy
services,as noted by many authors including Khazzoom (1980) and Wirl(1997).This suggests
that public policy to promote ‘green’ consumption choices might be more effective if focussed on
higher income households.
A second key finding regarding the impact of income levelis that the indirect effect becomes a
larger proportion ofthe totalrebound effect at higher income levels.This supports comments
made by others (Sorrelland Dimitropoulos,2008;Madlener and Alcott,2009) that a low direct
rebound effect should not be interpreted as indication ofthe scale ofthe totalrebound effect,
especially in high income countries.
Regarding the use ofthe two rebound-effect models,efficiency and conservation,some general
observations can be made.First, the conservation model,if indeed it is representative of house-
hold behaviour when household internalise environmental externalities into their utility function,
produces a much lower rebound effect than in cases where household choices contains an implied
price reduction due to capitalreplacement.Additionally,when conservation measures are com-
bined the environmentalbenefits are amplified,and the rebound effect reduced,compared each
case in isolation.
16

Moreover, when efficient choices are combined, the rebound effect becomes a mere average of the
two, with the environmental benefits equal to the sum of each case in isolation.One could easily
imagine that if price effects played a role,in addition to income effects,that the additive effects
of efficient technology would result in combined rebound effects being higher, and environmental
benefits lower, than the sum of each case.
A third key finding is that the greater the economic benefit ofhousehold ’green’consumption
choices,the larger the rebound effect and less effective the action.This is demonstrated in the
vehicle fuel case where two conservation options where estimated - one with cost savings on fuel
only,and one with associated reductions in vehicle maintenance costs.The added economic win
for the household, in terms of reduced vehicle running costs, greatly increased the rebound effect,
indicating an inherent trade-off between economic and environmentalbenefits. This supports
the finding of Carlsson-Kanyama et al. (2005) who find a negative rebound effect for households
adopting a green diet,due to the increased cost of the diet.As a rule of thumb it seems that in
the context of ‘green’ consumption, the greater the economic benefit, the lower the environmental
benefit.
Lastly, the choice of household demand model used in the rebound estimation was most important
at the high and low extremes of household income level.Near the average income, the linear model
seemed to overestimate the size of the rebound effect, while the WL model greatly overestimated
rebound effects at low income, and possibly underestimated at high incomes.The DSL specifica-
tion appears to offer a more accurate income-dependent specification to utilise in rebound effect
analysis at a household level.
The policy implications from these results are clear enough.Conservation,rather than more
efficient consumption,is a preferable household action.Indeed,combining conservation actions
improves outcomes,and is more effectively achieved by high income households who have lower
rebound effects.Also, it is clear that there is a trade-off between the cost benefits to the household,
and the environmental benefits.Thus, publicising the economic pay-off from ‘green’ consumption
choices, might lead to households adopting the least environmentally effective consumption choices.
To be clear, avoiding rebound effects from household consumption requires a corresponding reduc-
tion in household income (Madlener and Alcott, 2009).Robinson (2007) espoused the idea that if
households were to properly internalise the environmental externalities of their consumption, they
would actually work less and reduce total consumption.Yet, working less is not a well publicised
household conservation choice.
These results are also generally consistent with the idea that supply limitations on environmental
externalities such as cap-and-trade quotas, which force reduced consumption of goods whose pro-
17

duction involves that externality, are more effective than demand side measures, as they eliminate
rebound effect types of outcomes.
A few criticalbut unanswered questions remain about conservation choices and rebound effects,
and it is worth briefly commenting on them.
1. Do household income variations in environmentalimpacts of consumption choice have im-
plications for redistribution policies?
2. How might one better address the boundary problems of LCA methodologies for determining
embodied environmental externatlities?
3. What general equilibrium effects are being overlooked?
Regarding household income variations, if one does believe the data, then clearly yes, there will be
significant environmental costs from income redistribution.While a lot of research has uncovered
the non-linear relationship between GHG emissions and household income, no one is yet to use the
data to examine environmental costs of income redistribution through the tax and welfare system.
This may not be a concern if one actually believes that the boundary specifications required to
generate the LCA data of embodied GHG emissions are not theoretically sound.Ignoring inputs
into labour supply in LCA analysis leads to a divergence of environmental externalities associated
with different goods,that simply does not exist in reality.While authors have suggested that
higher quality goods have lower environmental intensity (Girod and De Haan, 2010),the typical
finding is that lower energy intensity of goods it mostly due to a labour-energy trade-off that is
eliminated when one considers the inputs required for labour supply (Costanza,1980;Maddala,
1965;Karunaratne,1981;Lenzen and Dey,2002). This issue has routinely been overlooked in
studies of rebound effects and in analysis of benefits of demand-side environmental policies more
generally.
Finally, regarding generalequilibrium effects from household conservation behaviour,one must
consider the likely producer reactions to any genuine reduction in demand that results from wide
adoption of‘green’choices. If conservation ofelectricity actually reduces electricity use in a
connected area,suppliers are likely to decrease prices to encourage higher use from the non-
conserving portion of the population to maintain optimal returns.
In terms of fuel use, will the small marginal reductions in fuel demand have a price impact, however
minor, that increases demand by non-conservers elsewhere in the economy? The answer to these
question is clearly yes, but the degree is uncertain.
‘Green’ household consumption choices, including conservation, are very indirect ways to promote
environmentalaims. Regulation of environmentalexternalities at their source,such as enacting
18

(tradable) quota systems,will not incur rebound effects,and are preferred to consumer-focussed
actions.Indeed, facilitation of a transition towards energy sources or technologies with fewer envi-
ronmental externalities is another way to directly change the environmental burden of consumption
at the broadest level.
6. Conclusion
It was found that rebound effects occur from the type of ‘green’consumption choices promoted
as ways for households to decrease their environmentalimpact. If rebound effects are ignored
when evaluating the environmentalbenefits from ‘green’consumption,then they willoverstate
the actuallikely benefits by at least 15% in the case of vehicle fuelconservation,and 6% in the
case electricity conservation.
Indeed,the cost-effective nature of these consumption choices means that they willbe more at-
tractive to household with lower incomes,but these households willhave the highest rebound
effects, and lowest resulting environmental improvement.Environmental policy directed at chang-
ing consumer behaviour is therefore best targeted at high income households,with conservation
measures rather than efficiency measures being promoted as more effective environmentalcon-
sumption choices.Indeed,reduced work and increased leisure time should also be promoted as
effective environmentalconsumption choices for households.Rebound effects estimated in this
paper should be considered conservative minimum estimates,given the limitations oflife cycle
analysis.
19

References
Research and reviews, 2008.URL carsales.com.
ABS. Household expenditure survey.Technical report, Australian Bureau of Statistics, 2004.
ABS. Survey of motor vehicle use - 9208.0, 2006.
Blake Alcott.Jevons’ paradox.EcologicalEconomics, 54(1):9–21, 2005.
Eva Alfredsson. ‘green’ consumption–no solution for climate change. Energy, 29(4):513–524, 2004.
Paul Baker, Richard Blundell, and John Micklewright.Modelling household energy expenditures
using micro-data.Economic Journal, 99(397):720–738, 1989.
Peter H. G. Berkhout,Jos C. Muskens,and Jan W. Velthuijsen.Defining the rebound effect.
Energy Policy, 28(6-7):425–432, 2000.
Runar Brannlund,Tarek Ghalwash,and Jonas Nordstrom.Increased energy efficiency and the
rebound effect:Effects on consumption and emissions.Energy Economics, 29(1):1–17, 2007.
Annika Carlsson-Kanyama, Rebecka Engstrm, and Rixt Kok.Indirect and direct energy require-
ments of city households in sweden:Options for reduction,lessons from modeling.Journal of
IndustrialEcology, 9(1/2):221–235, 2005.
Ann Marie Chalkley, Eric Billett, and David Harrison.An investigation of the possible extent of
the re-spending rebound effect in the sphere of consumer products.The Journalof Sustainable
Product Design, 1(3):163–170, 2001.
Robert Costanza.Embodied energy and economic valuation.Science, 210(4475):1219–1224, 1980.
Angus Deaton and John Muellbauer.Economics and consumer behaviour.Cambridge University
Press, Melbourne, 1980.
Christopher Dey.Embodied energy of 2003-4 abs hes commodity groups.2008.
Angela Druckman,Mona Chitnis,Steve Sorrell,and Tim Jackson. Missing carbon reductions?
exploring rebound and backfire effects in uk households.Energy Policy, 39(6):3572–3581, 2011.
Australian Conservation Foundation.The green home guide, 2007.
Bastien Girod and Peter De Haan.More or better? a model for changes in household greenhouse
gas emissions due to higher income.Journal of IndustrialEcology,14(1):31–49,2010. ISSN
1530-9290.
Australian Government.Be climate clever., 2007.
Lorna A. Greening.Effects of changes in residentialend-uses and behavior on aggregate carbon
intensity:comparison of 10 oecd countries for the period 1970 through 1993. Energy Economics,
23:153–178, 2001.
Nick Hanley,Peter G. McGregor,J. Kim Swales,and Karen Turner. Do increases in energy
efficiency improve environmentalquality and sustainability?EcologicalEconomics,In Press,
Corrected Proof, 2008.
M. Ohidul Haque. Income elasticity and economic development:Methodsand applications.
Springer, Dordrecht, 2005.
E. G. Hertwich.Life cycle approaches to sustainable consumption:A critical review.Environ-
mentalScience and Technology, 39(13):4673–4684, 2005.0013-936x.
20

Douglas Holtz-Eakin and Thomas M.Selden. Stoking the fires?co2 emissions and economic
growth.Journal of Public Economics, 57(1):85–101, 1995.
Sung H. Hong, Tadj Oreszczyn, and Ian Ridley. The impact of energy efficient refurbishment on the
space heating fuelconsumption in english dwellings.Energy and Buildings,38(10):1171–1181,
2006.
Herbert Inhaber.Why energy conservation fails.Quorom Books, Westport, Connecticut, 1997.
Neil Dias Karunaratne.An input-output analysis ofaustralian energy planning issues.Energy
Economics, 3(3):159–168, 1981.
J. DanielKhazzoom.Economic implications for mandated efficiency in standards for household
appliances.The Energy Journal, 1(4):21–40, 1980.
Rixt Kok, Rene M. J. Benders, and Henri C. Moll. Measuring the environmental load of household
consumption using some methods based on input-output energy analysis:A comparison of
methods and a discussion of results.Energy Policy, 34(17):2744–2761, 2006.
Manfred Lenzen.Primary energy and greenhouse gases embodied in australian final consumption:
an input-output analysis.Energy Policy, 26(6):495–506, 1998.
Manfred Lenzen and Christopher J. Dey.Economic, energy and greenhouse emissions impacts of
some consumer choice,technology and government outlay options.Energy Economics,24(4):
377–403, 2002.
Manfred Lenzen, Christoper Dey, and Barney Foran.Energy requirements of sydney households.
EcologicalEconomics, 49:375–399, 2004.
Manfred Lenzen,Mette Wier, Claude Cohen,Hitoshi Hayami,ShonaliPachauri,and Roberto
Schaeffer.A comparative multivariate analysis of household energy requirements in australia,
brazil, denmark, india and japan.Energy, 31(2-3):181–207, 2006.
A. B. Lovins, J. Henly, H. Ruderman, and M. D. Levine. Energy saving resulting from the adoption
of more efficient appliances:another view; a follow-up.The Energy Journal, 9(2), 1988.
Paul Lucas. Retail electricity prices for non-contestable customers.Queensland Government
Gazette, 47, 2003.
G.S. Maddala. Productivity and technologicalchange in the bituminous coalindustry,1919-54.
Journal of PoliticalEconomy, 73(4):352–365, 1965.
R. Madlener and B.Alcott. Energy rebound and economic growth:A review of the main issues
and research needs.Energy, 34(3):370–376, 2009.0360-5442 doi:10.1016/j.energy.2008.10.011.
Geoffrey Milne and Brenda Boardman.Making cold homes warmer:the effect of energy efficiency
improvements in low-income homes a report to the energy action grants agency charitable trust.
Energy Policy, 28(6-7):411–424, 2000.
Mirabella.Energy saver globes.Brochure, 2008.
KenichiMizobuchi.An empiricalstudy on the rebound effect considering capitalcosts.Energy
Economics, 30(5):2486–2516, 2008.
OECD. Towards sustainable consumption?Technicalreport,Organisation for Economic Co-
operation and Development, 2002.
Department of Consumer and Employment Protection.Fuelwatch, 2008.
A. H. M. E. Reinders,K. Vringer,and K. Blok. The direct and indirect energy requirement of
households in the european union.Energy Policy, 31(2):139–153, 2003.
21

T. Robinson.Work, leisure and the environment.Edward Elgar, Cheltenham UK, 2007.
Joyashree Roy.The rebound effect:some empirical evidence from india.Energy Policy, 28(6-7):
433–438, 2000.
Harry D. Saunders. A view from the macro side:rebound, backfire, and khazzoom-brookes. Energy
Policy, 28(6-7):439–449, 2000.
Lee Schipper and MichaelGrubb. On the rebound?feedback between energy intensities and
energy uses in iea countries.Energy Policy, 28(6-7):367–388, 2000.
Steve Sorrell and John Dimitropoulos.The rebound effect:Microeconomic definitions, limitations
and extensions.EcologicalEconomics, 65(3):636–649, 2008.
Queensland Transport.Registration fees, 2008.
UN. Report of the un conference on environment and development - chapter 4 - changing con-
sumption patterns.Technical report, United Nations, 1992.
Kees Vringer and Kornelis Blok.The direct and indirect energy requirements of households in the
netherlands.Energy Policy, 23(10):893–910, 1995.
Kees Vringer and Kornelis Blok.Long-term trends and direct and indirect household energy
intensitites:A factor in dematerialisation? Energy Policy, 28:713–727, 2000.
Kees Vringer, Theo Aalbers, and Kornelis Blok. Household energy requirement and value patterns.
Energy Policy, 35:553–566, 2007.
Christoph Weber and Adriaan Perrels.Modelling lifestyle effects on energy demand and related
emissions.Energy Policy, 28(8):549–566, 2000.
Ernst Von Weizacker,Amory B. Lovins,and L. Hunter Lovins.Factor four: Doubling wealth,
halving resource use.Earthscan, London, 1998.
Franz Wirl. The economics ofconservation programs.Kulwer Academic Publishers,Dortrecht,
1997.
22

Appendix A
Broad commodity group Detailed commodity group GHG intensity (kg CO2−e /$)
Domestic fuel and power Domestic fuel and power 7.33
Food and beverages Bakery products 0.40
Condiments 0.44
Dairy products 1.16
Fish 0.51
Fruit and nuts 0.39
Meals out 0.39
Meat 1.71
Non-alcoholic beverages 0.28
Vegetables 0.40
Alcoholic beverages Alcohol 0.30
Clothing and footwear Clothing 0.31
Clothing services 0.14
Footwear 0.30
Household furnishings Appliances 0.74
Blankets, linen and furniture 0.35
Furniture and flooring 0.30
Glass and tableware 0.61
Tools 0.24
Household services Household services 0.21
Medical care & health Health fees 0.26
Health insurance 0.02
Transport Freight 0.75
Vehicle fuel 2.60
Motor vehicle purchase 0.29
Vehicle parts and accessories 0.29
Public transport 0.54
Vehicle charges 0.15
Registration & insurance 0.02
Recreation Holidays 0.85
Pets 0.36
Recreational goods 0.41
Recreational services 0.13
Personal care Personal care 0.22
Miscellaneous Miscellaneous goods 0.31
Miscellaneous services 0.16
Source:Centre for Integrated Sustainability Analysis, Sydney (Dey, 2008)
23

Appendix B
***1% significance, **5% significance, *10% significance.
Double Semi Log α β γ Adj. r 2
Alcohol -17.53 0.025*** 3.73 0.15
(11.91) (0.000) (2.41)
Appliances -12.53 0.016*** 2.18 0.05
(9.21) (0.004) (1.88)
Bakery -23.6*** 0.003*** 6.03*** 0.23
(1.98) (0.001) (0.39)
Blankets/linen 6.98 0.015*** -1.68 0.06
(8.44) (0.004) (1.73)
Clothing 47.6** 0.068*** -10.6** 0.24
(23.54) (0.01) (4.76)
Clothing services 1.77 0.002*** -0.38 0.04
(1.21) (0.001) (0.25)
Condiments -28.1*** 0.006*** 6.73*** 0.25
(3.57) (0.001) (0.72)
Dairy -16.4*** 0.001** 4.31*** 0.18
(1.40) (0.001) (0.28)
Domestic fuel & power -0.98 0.009*** 3.13*** 0.18
(4.72) (0.002) (0.94)
Fish -2.92* 0.002*** 0.81** 0.05
(1.58) (0.001) (0.32)
Footwear 3.13 0.012*** -0.88 0.08
(5.01) (0.002) (1.02)
Freight 8.12** 0.007*** -1.66** 0.04
(3.28) (0.002) (0.68)
Fruit and nuts -12.73*** 0.003*** 3.24*** 0.14
(1.65) (0.001) (0.33)
Vehicle fuel -72.15*** 0.008*** 15.79*** 0.21
(6.29) (0.002) (1.27)
Furniture/ flooring 5.88 0.042*** -2.49 0.09
(17.65) (0.008) (3.59)
Glass/tableware 0.18 0.007*** -0.15 0.07
(5.24) (0.002) (1.06)
Health fees -14.64** 0.015*** 2.85* 0.09
(6.91) (0.003) (1.41)
Health Insurance -34.70*** 0.008*** 7.52*** 0.19
(3.23) (0.001) (0.65)
Holidays -15.50 0.068*** 1.58 0.21
(17.91) (0.008) (3.66)
24

α β γ Adj. r 2
Household services -57.20*** 0.031*** 14.51*** 0.25
(9.79) (0.004) (1.98)
Meals out -50.20*** 0.044*** 9.68*** 0.36
(10.21) (0.004) (2.07)
Meat -30.96*** 0.004*** 7.70*** 0.17
(2.92) (0.001) (0.58)
Miscellaneous goods 21.96 0.031*** -4.71 0.13
(23.31) (0.009) (4.68)
Miscellaneous services 133.18*** 0.17*** -29.81*** 0.31
(46.14) (0.019) (9.34)
Motor vehicle purchase 206.36*** 0.20*** -47.45*** 0.24
(52.30) (0.022) (10.61)
Non-alcoholic beverages -21.15*** 0.004*** 4.91*** 0.25
(1.74) (0.001) (0.35)
Vehicle parts/ accessories -12.21 0.021*** 2.45** 0.04
(7.33) (0.003) (0.98)
Personal care -6.10 0.021*** 1.39 0.20
(7.33) (0.003) (1.48)
Pets 1.03 0.012*** -0.15 0.03
(10.14) (0.005) (20.8)
Public transport -6.99*** 0.001 1.57*** 0.02
(1.43) (0.001) (0.28)
Recreational goods 60.79* 0.083*** -13.13* 0.22
(34.35) (0.014) (6.92)
Recreational services -13.29 0.025*** 2.40 0.12
(10.67) (0.005) (2.16)
Tools -8.83** 0.008*** 1.60** 0.05
(4.18) (0.002) (0.085)
Vegetables -15.52*** 0.002*** 3.97*** 0.18
(1.31) (0.000) (0.026)
Vehicle charges 18.55 0.036*** -4.42 0.09
(17.98) (0.007) (3.63)
Vehicle registration -40.62*** 0.007*** 9.44*** 0.31
(3.29) (0.001) (0.656)
25

Appendix C
Age - 1-30 for age of head of household with 30 categories from 15 to >80
Occupancy - 1-6 (where 6 is six or more persons per household)
State - 1-New South Wales, 2-Victoria, 3-Queensland,4-South Australia, 5-Western Australia,6-Tasmania,7-
A.C.T. and NT
Urbanity - 0-NA, 1-Capital City, 2-Balance of state
Dwelling - 1-Separate house, 2-Semi-detached one storey, 3-Semi-detached two or more storeys, 4-Apartment in 1
or 2 storey block, 5-Apartment in a 3 storey block, 6-Apartment in a 4 or more storey block, 7-Apartment attached
to a house, 8-Caravan, houseboat, improvised home
***1% significance, **5% significance, *10% significance.
DSL2 α β γ Occup. Urbanity State Age Dwelling Adj. r2
Alcohol -16.88 0.026*** 5.11** -3.65*** 2.33*** 1.19*** -0.47*** 0.35 0.17
(12.92) (0.005) (2.46) (0.45) (0.85) (0.27) (0.07) (0.39)
Appliances -18.47* 0.016*** 3.91** -2.45*** 0.52 0.55* -0.018 -0.67* 0.56
(10.17) (0.004) (1.89) (0.55) (1.00) (0.30) (0.075) (0.35)
Bakery -23.35*** 0.0020*** 3.49*** 4.45*** -0.035 -0.098 0.31*** -0.097 0.39
(2.17) (0.001) (0.39) (0.13) (0.22) (0.089) (0.018) (0.089)
Linen 1.24 0.015*** -0.69 -1.01*** 0.088 0.40* 0.059 -0.22 0.061
(9.14) (0.004) (1.71) (0.38) (0.64) (0.22) (0.051) (0.26)
Clothing 56.77** 0.068*** -13.01*** 2.90*** -0.32 -0.36 -0.092 1.09** 0.24
(24.09) (0.0095) (4.76) (0.70) (1.08) (0.35) (0.097) (0.54)
Cloth. serv. 0.72 0.002*** -0.19 -0.086* -0.29*** 0.000 0.022*** 0.085* 0.039
(1.30) (0.001) (0.25) (0.051) (0.092) (0.029) (0.008) (0.047)
Condiment -18.26*** 0.006*** 2.96*** 4.69*** 0.51* 0.19** 0.053** -0.042 0.36
(3.88) (0.001) (0.72) (0.18) (0.30) (0.09) (0.024) (0.13)
Power 5.28 0.009*** 0.65 2.70*** -0.041 0.91*** 0.084*** -1.25*** 0.24
(4.96) (0.002) (0.95) (0.17) (0.34) (0.094) (0.028) (0.16)
Fish -5.02*** 0.002*** 1.13*** 0.11 -0.69*** -0.28*** 0.11*** -0.035 0.065
(1.72) (0.001) (0.32) (0.093) (0.18) (0.051) (0.014) (0.067)
Footwear 3.11 0.012*** -1.27 0.71*** 0.023 -0.026 0.033 0.18 0.080
(4.96) (0.002) (0.99) (0.24) (0.42) (0.13) (0.035) (0.18)
Freight 7.80** 0.0065*** -1.17* -0.66*** -0.60*** -0.011 -0.061** 0.48*** 0.049
(3.50) (0.002) (0.65) (0.18) (0.20) (0.079) (0.025) (0.15)
Vehicle fuel -51.7*** 0.0093*** 11.32*** 3.19*** 3.54*** 0.61*** -0.22*** -2.36*** 0.24
(7.01) (0.0026) (1.31) (0.41) (0.75) (0.22) (0.057) (0.24)
Furniture -3.32 0.043*** 1.071 -5.46*** 2.75* 0.47 -0.26** -0.19 0.093
(18.97) (0.008) (3.64) (0.81) (1.45) (0.45) (0.13) (0.67)
Tableware -2.93 0.007*** 0.26 -0.17 0.008 0.003 0.057** 0.07 0.068
(5.86) (0.002) (1.11) (0.13) (0.25) (0.087) (0.026) (0.14)
Health fees -18.83** 0.014*** 3.85** 0.075 -3.63*** -0.66*** 0.29*** -0.15 0.095
(7.97) (0.003) (1.52) (0.39) (0.63) (0.23) (0.057) (0.28)
26

DSL2 α β γ Occup. Urbanity State Age Dwelling Adj.r 2
Health Ins. -54.43*** 0.007*** 9.31*** 0.34 -0.79* 0.12 0.59*** -0.74*** 0.22
(3.83) (0.001) (0.68) (0.23) (0.45) (0.13) (0.036) (0.17)
Holidays -56.53*** 0.066*** 10.23*** -7.08*** -5.31** 0.17 0.62*** 0.70 0.22
(19.05) (0.008) (3.66) (0.87) (1.74) (0.52) (0.15) (0.68)
House serv -20.72* 0.031*** 9.23*** 3.97*** -4.12*** -0.67** -0.27*** -1.81*** 0.27
(10.74) (0.004) (2.03) (0.55) (0.98) (0.31) (0.078) (0.36)
Meals out -28.14*** 0.044*** 9.26*** -0.56 -7.19*** -1.79*** -0.41*** 2.23*** 0.37
(10.74) (0.004) (2.14) (0.51) (0.83) (0.28) (0.079) (0.45)
Misc. goods 21.20 0.031*** -5.07 0.98** -1.26* 0.088 0.065 0.60* 0.13
(24.73) (0.009) (4.81) (0.42) (0.68) (0.20) (0.062) (0.34)
Misc. services 150.05*** 0.17*** -31.36*** 0.61 -0.14 -1.13 -0.70*** 4.48*** 0.31
(48.59) (0.019) (9.33) (1.44) (2.48) (0.73) (0.19) (1.21)
Vehicle purch. 159.39*** 0.20*** -38.09*** -12.80*** 16.90*** 2.04** -0.43* -0.34 0.25
(55.25) (0.022) (10.69) (1.64) (3.10) (0.94) (0.24) (1.21)
Beverage -12.39*** 0.004*** 2.85*** 2.24*** -0.49** -0.10 -0.043** 0.072 0.30
(1.89) (0.001) (0.35) (0.12) (0.22) (0.067) (0.017) (0.096)
Vehicle part -8.11 0.006*** 1.68* 0.19 1.44*** 0.27* -0.12*** -0.34* 0.043
(5.26) (0.002) (0.99) (0.30) (0.54) (0.15) (0.041) (0.20)
Personal care -11.10 0.02*** 1.87 0.33 -0.94* -0.19 0.14*** 0.48** 0.20
(7.72) (0.003) (1.52) (0.29) (0.54) (0.17) (0.045) (0.23)
Pets 1.59 0.013*** 0.46 -1.53*** 1.21 0.18 -0.052 -1.29*** 0.04
(10.63) (0.005) (2.05) (0.37) (0.74) (0.20) (0.072) (0.30)
Public trans. -0.97 0.000 1.27*** 0.49*** -2.20*** -0.95*** -0.052*** 0.89*** 0.068
(1.65) (0.001) (0.30) (0.13) (0.19) (0.071) (0.018) (0.14)
Rec. goods 80.93** 0.085*** -15.08** -0.097 0.55 -0.054 -0.66*** 1.54** 0.22
(36.35) (0.014) (6.98) (0.82) (1.43) (0.48) (0.11) (0.71)
Rec. serv. -20.06* 0.024*** 2.50 0.89 0.51 0.006 0.20*** 0.11 0.12
(11.91) (0.005) (2.22) (0.55) (1.06) (0.29) (0.075) (0.37)
Tools -9.52** 0.008*** 1.69* -0.36 0.72 0.32** -0.016 -0.42** 0.051
(4.67) (0.002) (0.89) (0.35) (0.54) (0.14) (0.040) (0.17)
Vegetables -21.16*** 0.001** 3.82*** 1.35*** -0.55*** -0.022 0.26*** -1.47*** 0.23
(1.56) (0.000) (0.27) (0.098) (0.19) (0.058) (0.016) (0.16)
Veh. charges 12.82 0.036*** -1.68 -3.64*** 0.91 -0.62* -0.023 -0.75* 0.093
(19.56) (0.007) (3.70) (0.64) (1.20) (0.36) (0.10) (0.43)
Vehicle reg -26.33*** 0.007*** 7.88*** 1.12*** -2.63*** -0.63*** 0.005 -1.47*** 0.33
(3.65) (0.001) (0.68) (0.19) (0.35) (0.11) (0.029) (0.16)
Dairy -13.74*** 0.001** 2.31*** 2.81*** 0.49*** 0.22*** 0.11*** -0.19*** 0.30
(1.57) (0.000) (0.28) (0.10) (0.18) (0.054) (0.014) (0.074)
Fruit / nuts -22.60*** 0.002*** 3.75*** 1.25*** -1.19*** -0.36*** 0.39*** 0.078 0.20
(1.87) (0.001) (0.33) (0.11) (0.21) (0.065) (0.019) (0.094)
Meat -36.37*** 0.003*** 5.78*** 4.17*** -0.089 0.20* 0.50*** -0.98*** 0.25
(3.31) (0.001) (0.59) (0.20) (0.39) (0.12) (0.031) (0.13)
27

Appendix D
***1% significance, **5% significance, *10% significance.
WL model α β r2
Alcohol 0.026*** 0.001* 0.000
(0.006) (0.001)
Appliances -0.009** 0.004*** 0.004
(0.005) (0.001)
Bakery 0.11*** -0.013*** 0.18
(0.002) (0.000)
Blankets/linen -0.008*** 0.003*** 0.005
(0.003) (0.000)
Clothing -0.034*** 0.011*** 0.025
(0.005) (0.001)
Clothing services -0.001 0.000*** 0.002
(0.001) (0.000)
Condiments 0.091*** -0.009*** 0.077
(0.003) (0.000)
Dairy 0.087*** -0.010*** 0.18
(0.002) (0.000)
Domestic fuel / power 0.29*** -0.038*** 0.36
(0.004) (0.001)
Fish 0.019*** -0.002*** 0.017
(0.001) (0.000)
Footwear -0.008*** 0.002*** 0.006
(0.002) (0.000)
Freight 0.007*** -0.001** 0.001
(0.001) (0.000)
Fruit and nuts 0.059*** -0.007*** 0.071
(0.002) (0.000)
Vehicle fuel 0.084*** -0.005*** 0.006
(0.005) (0.001)
Furniture/ flooring -0.044*** 0.011*** 0.016
(0.007) (0.001)
Glass/tableware -0.000 0.001*** 0.002
(0.001) (0.000)
Health fees 0.010** 0.001** 0.001
(0.004) (0.001)
Health Insurance 0.045*** -0.003*** 0.003
(0.007) (0.001)
Holidays -0.049*** 0.016*** 0.025
(0.008) (0.001)
Household services 0.28*** -0.03*** 0.11
(0.007) (0.001)
Meals out 0.005 0.008*** 0.014
(0.005) (0.001)
Meat 0.12*** -0.014*** 0.09
(0.003) (0.001)
Miscellaneous goods -0.004 0.004*** 0.006
(0.004) (0.001)
28

WL model α β r 2
Miscellaneous services -0.091*** 0.027*** 0.045
(0.01) (0.002)
Motor vehicle purchase -0.24*** 0.046*** 0.089
(0.011) (0.002)
Non-alcoholic beverages 0.056*** -0.006*** 0.057
(0.002) (0.000)
Motor vehicle parts/accessories 0.003 0.001** 0.001
(0.003) (0.000)
Personal care 0.029*** -0.001 0.000
(0.003) (0.001)
Pets 0.015*** -0.000 -0.000
(0.003) (0.001)
Public transport 0.019*** -0.002*** 0.008
(0.002) (0.000)
Recreational goods -0.010 0.009*** 0.010
(0.007) (0.001)
Recreational services 0.001 0.004*** 0.005
(0.005) (0.001)
Tools -0.001 0.002*** 0.002
(0.003) (0.000)
Vegetables 0.074*** -0.009*** 0.13
(0.002) (0.000)
Vehicle charges -0.029*** 0.008*** 0.012
(0.005) (0.001)
Vehicle registration 0.10*** -0.01*** 0.059
(0.003) (0.000)
29

Appendix E
***1% significance, **5% significance, *10% significance.
Linear model α β r2
Alcohol 2.54*** 0.030*** 0.15
(0.78) (0.001)
Appliances -0.83 0.019*** 0.052
(0.88) (0.001)
Bakery 8.82*** 0.011*** 0.20
(0.23) (0.000)
Blankets/linen -2.061*** 0.013*** 0.059
(0.58) (0.001)
Clothing -9.57*** 0.054*** 0.24
(1.07) (0.001)
Clothing services -0.28*** 0.002*** 0.035
(0.087) (0.000)
Condiments 8.12*** 0.015*** 0.23
(0.30) (0.000)
Dairy 6.78*** 0.007*** 0.15
(0.18) (0.000)
Domestic fuel and power 15.85*** 0.013*** 0.17
(0.31) (0.000)
Fish 1.46*** 0.003*** 0.053
(0.15) (0.000)
Footwear -1.59*** 0.011*** 0.079
(0.39) (0.000)
Freight -0.80*** 0.004*** 0.038
(0.24) (0.000)
Fruit and nuts 4.68*** 0.007*** 0.13
(0.20) (0.000)
Vehicle fuel 12.82*** 0.028*** 0.19
(0.65) (0.001)
Furniture/ flooring -7.51*** 0.038*** 0.086
(1.38) (0.002)
Glass/tableware -0.60** 0.007*** 0.068
(0.26) (0.000)
Health fees 0.68 0.019*** 0.089
(0.66) (0.001)
Health Insurance 5.76*** 0.018*** 0.18
(0.42) (0.000)
Holidays -6.99*** 0.070*** 0.21
(1.52) (0.002)
Household services 20.16*** 0.049*** 0.24
(0.95) (0.001)
Meals out 1.86** 0.056*** 0.35
(0.84) (0.001)
Meat 10.49*** 0.014*** 0.15
(0.36) (0.000)
30

Linear model α β r 2
Miscellaneous goods -3.37*** 0.025*** 0.13
(0.72) (0.001)
Miscellaneous services -27.20*** 0.13*** 0.30
(2.19) (0.003)
Motor vehicle purchase -48.89*** 0.14*** 0.23
(2.83) (0.003)
Non-alcoholic beverages 5.25*** 0.01*** 0.23
(0.21) (0.000)
Motor vehicle parts/accessories 0.97*** 0.008*** 0.039
(0.46) (0.001)
Personal care 1.38*** 0.023*** 0.20
(0.51) (0.001)
Pets 0.25 0.012*** 0.036
(0.67) (0.001)
Public transport 1.45*** 0.003*** 0.017
(0.22) (0.000)
Recreational goods -9.84*** 0.066*** 0.22
(1.39) (0.002)
Recreational services -0.38 0.028*** 0.12
(0.85) (0.001)
Tools -0.20 0.010*** 0.050
(0.49) (0.001)
Vegetables 5.86*** 0.007*** 0.16
(0.18) (0.000)
Vehicle charges -5.22*** 0.03*** 0.087
(1.07) (0.001)
Vehicle registration 10.17*** 0.019*** 0.27
(0.34) (0.000)
31