environmental impact, quality, and price: consumer trade-offs and the development of environmentally...

Technological Forecasting & Social Change 76 (2009) 552–566

Contents lists available at ScienceDirect

Technological Forecasting & Social Change

Environmental impact, quality, and price: Consumer trade-offs and thedevelopment of environmentally friendly technologies☆

Paul Windrum a,⁎,1, Tommaso Ciarli a,b,2, Chris Birchenhall c,3

a Manchester Metropolitan University Business School, Manchester, UKb Max Planck Institute for Economics, Jena, Germanyc University of Manchester, Manchester, UK

a r t i c l e i n f o

☆ The authors gratefully acknowledge funding throuand Humanities Programme of the EU 7th Frameworkimprove the original draft of the paper.⁎ Corresponding author. Manchester Metropolitan U

E-mail address: [email protected] (P. Windr1 Paul Windrum, PhD is a Reader at Manchester M

Germany.2 Tommaso Ciarli, PhD is a Post-Doctoral Research F3 Chris Birchenhall is a Senior Lecturer in Comput

Economics and is an associate editor of the Computat

0040-1625/$ – see front matter © 2008 Elsevier Inc.doi:10.1016/j.techfore.2008.04.012

a b s t r a c t

Article history:Received 14 September 2007Received in revised form 16 April 2008Accepted 22 April 2008

The paper examines the effect of heterogeneous consumer trade-offs – between environmentalperformance, quality of service characteristics, and price – on the generation and diffusion ofenvironmentally benign technology paradigms. We find that the direction, timing, andenvironmental impact of new paradigms is shaped by the distribution of consumer trade-offs.Of key importance are the initial distributions of consumer preferences, and how thosedistributions evolve over time. This has serious implications on environmental pollution, andfor policy makers seeking to influence the ‘greening’ of consumer demand.

© 2008 Elsevier Inc. All rights reserved.

Keywords:Heterogeneous consumer demandTrade-offs in environmental preferencesQuality and priceTechnological innovationParadigm substitutions

1. Introduction

What are the implications of the trade-offs we, as consumers, face when choosing between alternative technology products?Do we choose the environmentally cleaner product, even if it offers inferior service quality compared with a more pollutingalternative? What happens if the cleaner technology is more expensive than its more polluting rival? Such choices are familiar tous all. As consumers, we face them over a wide range of examples, from purchases involving large ticket items such as cars andholidays, to the everyday purchases we make in supermarkets. Each of us answers these questions in different ways, and ouranswers have a direct bearing on our individual carbon footprints, and on our collective environmental impact. In other words,heterogeneity of consumer preferences towards trade-offs matter. They have a direct effect on the consumer selections that wemake and on the resulting environmental impact of our selections. This is the focus of the current paper.

The paper is a companion to our other paper in this Special Issue. It complements and extends the discussion of howheterogeneous environmental preferences influence the development of cleaner designs within a given paradigm, and paradigmsubstitutions. By addressing the consequences of heterogeneous consumer attitudes to trade-offs between environmentalperformance, product quality, and price, this paper enriches and substantially develops the research instigated in the companionpaper in this issue.

gh the Public–Private Services Innovation (ServPPIN) project, funded through the Socio-Economic Sciences. We thank the two anonymous journal referees for their excellent comments and suggestions on how to

niversity Business School, Aytoun Building, Aytoun Street, Manchester M1 3GH, United Kingdom.um).etropolitan University Business School and a Visiting Professor Max Planck Institute for Economics, Jena

ellow at Manchester Metropolitan University Business School.ational Economics at the University of Manchester, founding member of the Society of Computationaional Economics journal.

All rights reserved.

,

l

mailto:[email protected]

http://dx.doi.org/10.1016/j.techfore.2008.04.012

http://www.sciencedirect.com/science/journal/00401625

553P. Windrum et al. / Technological Forecasting & Social Change 76 (2009) 552–566

The paper is structured in the following way. Section 2 discusses the empirically grounded model that captures a set of stylisedfacts regarding paradigm substitutions. Using this model, Section 3 investigates the effect of consumer trade-offs on paradigmsubstitutions and global pollution. First, we examine the effect of differences in mean consumer preferences for environmentalperformance and direct utility. This enables us to compare, for example, the environmental impact of distributions with a highmean preference for environmental performance and a lower mean preference for the quality of service characteristics, withdistributions containing high mean preferences for both environmental performance and service characteristics. Second, weconsider the environmental impact of trade-offs between environmental performance and indirect utility (i.e. price). Here wecompare distributions with different standard deviations in environmental preferences and in price preferences. Section 4concludes by bringing together the current set of findings with those established in the companion paper. This enables us to draw aset of broad results on the importance of heterogeneous consumer preferences for the development of environmentally friendlytechnologies.

2. The model

2.1. Stylised empirical facts in the model

In this sectionwe outline the key features of our empirically groundedmodel of sequential technology competitions. Themodelcaptures a set of stylised empirical facts on pollution, consumption, and paradigm substitutions gleaned from the history of urbantransport.4

1) Deep path dependence. Deep path dependency exists across successive environmental–technological paradigms. This is due tothe unfolding of new technological paradigms from old paradigms. While a new paradigm differs from the previous paradigmin some important respects, it also share one or more features. This is reflected in the technology designs themselves.

2) Environmental promise. The environmental ‘promise’ of a new paradigm is an important factor affecting the behaviour of firmsand consumers. It was not simply the service characteristics of the car (speed, flexibility, and private consumption) that led toits rapid adoption afterWWII. Demand for the car was closely tied to the demand for healthy suburban living, driven by a desireto escape the disease and filth of horse pollution in city centres. Cars enabled the middle class to truly explore a healthier set ofconsumption opportunities offered by suburban living.

3) Windows of opportunity. A pollution minimising design exists within each technology paradigm. Once this design has beenidentified, no further improvement is possible. As a consequence, pollution steadily increases as consumers continue to buy thetechnology. Since pollution is a negative externality, continued consumption eventually turns all consumers' utilities negative.This opens a ‘window of opportunity’ for new, more environmentally benign designs. Consumers are willing to experimentwith new technological solutions, and firms have economic incentives to experiment and develop novel technologies that arebased on these new scientific/engineering discoveries.

4) New technology firms. New technology start-ups champion the new technology. If successful, these start-ups replace the oldtechnology firms to become the dominant industry players. This was the case in the car industry [1], and the importance of newstart-ups has been identified in other industries [2].

5) Consumer preferences affect the timing, direction, and type of innovation. The direction of intra- and inter-paradigm innovation isdriven by the economic rewards to innovative firms. These are strongly influenced by consumer preferences. The decision todevelop less polluting designs or, alternatively, designs with high quality service characteristics and/or low price, depends on:(i) a technologically given relationship between alternative combinations of service characteristics and environmentalperformance (the position of the environmentally optimum design within the paradigm landscape), and (ii) consumers'preferences regarding environmental impact, service characteristics, and price. Heterogeneous environmental preferenceswere explored in the companion paper. In this paper we examine the relationship between preferences for environmentalutility, direct utility (product quality), and indirect utility (price).

Our model differs to other well-known models of sequential technology competitions, such as [3] and [4]. First theadoption, development and diffusion of new technology paradigms are endogenous within the model. The ‘window ofopportunity’ is operationalised in the following way. An environmental–technology paradigm is modelled as a pseudo-NKlandscape. Within each paradigm there exists an optimal global peak. This is the design that combines minimum en-vironmental impact with a relatively high product quality and low price. Once identified, no further improvements can bemade with respect to environmental pollution within the paradigm. If global pollution continues to rise over time, consumershave a real incentive to experiment with new, less polluting technology paradigms. Once the limits of pollution improvementhave been reached – i.e. once the optimal design of an existing paradigm has been identified – the search begins for a moreenvironmentally friendly technology, based on the latest scientific/engineering breakthroughs. New start-ups firms developthe commercial applications of these breakthroughs.

Second, Malerba et al. et al. treat quality as a simple integer value. We unpack performance quality, using the Lancaster [5]service characteristics approach, into a vector of service characteristics x→. This vector contains a set of complex, non-linearrelationships between service characteristics. Perceived quality depends on the valuations that are placed on these service

4 See the companion paper in this Special Issue for a detailed discussion.

554 P. Windrum et al. / Technological Forecasting & Social Change 76 (2009) 552–566

characteristics by the utility functions of an existing set of consumer classes. This has a radical implication. Namely, a universalmeasure of product quality does not exist. Perceived quality depends on the particular set of consumer preferences that are presentin the market at a given moment in time. Quality is therefore subjective, temporal, and subject to change as the population ofconsumer preferences changes over time (also see [6] on this point).

Third, we do not assume, as Malerba et al. et al. do, that new technology goods are always superior in quality/price performanceto old technology goods, or, hence, that new technology firms automatically have a performance advantage over old technologyfirms. The initial designs of new technology firms are randomly initialised. Hence, their designs may initially be inferior to oldtechnology firms' designs.

A final distinguishing feature of ourmodel is the existence of an environmental–technological ‘promise’. In deciding whether toexperiment with a new technology paradigm, consumers must weigh the environmental promise of a new paradigm against theexisting set of new technology designs. In the early days, the gap between these may be large. Furthermore, closing the gapdepends on a stream of R&D, which in turn requires a steady stream of consumer sales in order to raise the necessary funds forR&D. Whether or not firms reach this environmental optimum depends on the rewards to innovation, which are stronglyinfluenced by consumer preferences.

2.2. Consumer demand

Following Windrum and Birchenhall [7,8], we use a ‘fuzzy’ replicator algorithm to model the movement of individualconsumers across a population of consumer classes. A consumer class that is well catered by one or more technology goods (i.e.by goods that provide the combination of direct, indirect, and environmental utility which that particular consumer classprefers), will offer a higher average utility than a class that is not well catered for by the current set of goods. Classes withabove-average utility grow as a proportion of the total population, while classes with below-average utility decline. In the limit,a pure replicator dynamic algorithm tends to reduce the number of consumer classes to one. In our model this would invariablymean convergence to a single dominant design that is provided by a single firm. The fuzzy parameter f tempers the strength ofselection, allowing a number of classes, which attain a similar level of utility, to obtain the same share ψj,t of individualconsumers C.

The number of individual consumers Cj,t=ψj,tC within each class j is then computed as a ratio ψj,t of all individualconsumers:

ψj;t = ψj;t−1

Pu fj;t

PU f

t − 1

ð1Þ

Pufj;t = f

Plul; j;t + εu = Cj;tP

jψj;t

Plul;j;tεu = Cj;tð Þ = round f

Puj;tPUt

h iis the average ‘fuzzy’ utility of the class and U− f

t−1 is the global average. ul,j,t is theu

where

utility that a single consumer l in class j obtains through consumption at time t. ε is parameter that allows each class to surviveover time if its fitness, or the environment, changes.

In each period, all consumer classes access the market in a random order. Consumers from each class select the firm that bestsatisfies their utility and purchase one good from a firm's inventory of finished goods. To simplify, we assume that each consumerbuys just one unit. When the firm is out of stock, consumers move to the second best firm, and so on. In the unlikely event that twofirms attain the same utility for a specific class, one of the two is randomly picked with equal probability.

The class utility function uj that individual consumers use to select between alternative technologies is

ul;j = d Yxi� �

+ v mj − pj� �

+ e s Yxi� �

;G� � ð2Þ

x→i is the vector of service characteristics offered by a product design, mj is a given budget constraint of the consumer
whereclass, pj is product price, G is the global level of pollution, and s(x→i) the environmental impact of consuming one unit of a productdesign.
The class utility function thus comprises three distinct components. d(x→) is the ‘direct utility’, offered by a particular designx→, that a consumer gains by consuming the set of service characteristics, embodied in a good, over its lifetime [5,9]. Relativeproduct quality can therefore be defined as better/worse performance over a set of service characteristics.

V(m−p) is the ‘indirect utility’ a consumer obtains from spending residual income (income minus the price paid) on othergoods. The higher the price of a good, the less money the consumer can spend in other markets, and the lower is his/her indirectutility. An important link exists between indirect utility and indirect network effects. Indirect network effects are the staticeconomies of scale built up over time by old technology firms. Clearly, these are not initially available to new technology start-ups[10,11].

‘Environmental utility’ e(s(x→i),G) comprises two components. First, there is the utility associated with the environmentalsustainability of an individual consuming a good over its lifetime. This is determined by the pollution created by a particular designx→i. The second component is the global level of pollutionG that has built up as a consequence of past consumption by all consumers.

Different consumer classes place different weights on service characteristics x, on price p, and on the environmental impact ofconsumption s. A key advantage of this set up is that it allows one to consider different scenarios according to alternativedistributions of consumer classes.


The three components of a class utility function take the following form,

5 Para

dj =P

h∈zj βj;hffiffiffiffiffiffiffiffiffiffiffiffiffiffiffixt−1;h;i

pvj = αj

ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffimj − pt−1;i

p 8pt−1;i < mj

ej = ηjEj st−1;i

� �−—s

h i1−ρ

1− ρ8—s < E sð Þ

ð3Þ

αj is the weight (i.e. preference) that a consumer class places on the price of the good produced by firm i, and βj,h is the
whereweight it places on the quality of the h service characteristics x→i,h in paradigm Zj. We model environmental utility as a compositefunction that reflects the hyperbolic absolute risk aversion of consumers towards the global damage caused by environmentalpollution G.
Ej(si) is the consumer class' expectation of the environment impact of a specific firm's technology good. ρ is a parameter thatreflects the relative risk aversion of a class towards global pollution G, and ηj is the discount rate of a consumer class. s_ is theminimum level of environmental fitness (i.e. the maximum pollution) that a class is ready to accept from a firm. η j

p is a weightingbetween 0 and 1 (i.e., η j

p ∈ [0,1]) that consumers attach to the current sustainability of design i relative to the ‘technologicalpromise’ of the paradigm z.

Ej sið Þ = ηjps Yxi� �

1 + s zð Þ− s x→i

� � + 1− ηjp� �

s zð Þ ð4Þ

te that Ej(si) is a relative evaluation. Each technology is being compared to the sustainability of the best technology currently
Noavailable on the market s(z) at time t. Ej(si) is thus a combination of the sustainability of the technology good relative to thecleanest technology design available in the market s(z) at time t,5 and the gap between the current environmental performance ofthe technology s(x→i) and its environmental ‘promise’.
The minimum level of sustainability that consumers are ready to purchase in a design (s_) follows, with a logistic form, thechange in global pollution stocks:

Ps t =

Psτ +

st = 2− sτ

1 + st = 2− sτs0 − 1

� �e − r Gt − 1 −Gτð Þ½ �

ΔGt−1 > 0

Psτ

1 +sPτs0 − 1

� �e½− rGt− 1 �

ΔGt−1 < 0

8>>>>>>>><>>>>>>>>:

ð5Þ

τ is the period inwhich the boundaries of theminimum environmental sustainability change. This may either be due to (i) a
wherechange in the paradigm z (with a consequent change in st), or (ii) a change in the sign of global environmental pollution ΔG. r isthe rate of growth of the minimum level of global impact, and s0 is the lower asymptote.
This specification captures three key stylised facts about environmental pollution and consumer preferences. The first is theimportant role played by the environmental ‘promise’ of new technologies. Given lags in understanding and information about thetrue impact of new technology paradigms, it may be that early consumer adoption is largely based on its initial environmental‘promise’. The environmental promise is exogenously given, based on the current state of scientific knowledge. The second stylisedfact is the existence of a difference between the environmental ‘promise’ of a paradigm and the level of pollution produced by thecurrent generation of technology products. The third stylised fact is the very different weights that heterogeneous consumerclasses can place on pollution. This is captured in the relative risk aversion towards pollution, and the discount rate of eachconsumer class.

The latter point raises a core question about individual and collective rationality. Why should an individual conserve energy,recycle, or purchase more environmentally benign goods when our own personal contribution has such little overall impact? Ofcourse, if many individuals do actually conserve energy, recycle, and purchase cleaner products there will be a general greening ofour collective consumption lifestyles, resulting in a more environmentally sustainable economy. For a given distribution ofconsumer preferences, many people's actions can, together, make a difference. Thus, initial distributions of class preferences willhave a direct bearing on the global level of pollution. Further, the dynamics that determine the direction inwhich class preferencesevolve – captured, in our model, by the replicator algorithm and the opening up of new paradigms and associated consumerpreferences – is of central importance to the final environmental outcome.

Finally, note that a consumer may decide not to purchase any of the technology designs currently offered. The utility obtainedfrom not making a purchase is the null utility αj

ffiffiffiffiffiffimj

p Þ�. An individual prefers to hold on to themoney (or else buy a different good in

another market), if (i) the utility offered by the current set of designs is lower than the null utility, i.e. uj < αjffiffiffiffiffiffimj

p Þ�, (ii) if price is

greater than the consumer's budget constraint, or (iii) if firms are sold out and cannot supply this consumer in the current period.

digms z evolve endogenously in the model, and are assumed to improve exogenously as an outcome of basic research.


2.3. Supply

2.3.1. Production and pricingInitially, all firms in the model are endowed with identical levels of capacity and wealth. In each period, every firm has a

current design, a productive capacity (setting an upper limit on output), and a non-negative inventory of stock that is carriedover from the previous period. Product inventories set the maximum number of goods that a firm can currently sell in themarket.

Given total consumer demand Dt,i and the level of sales St,i=min(Dt,i,qt− l,i), firms determine their target level ofproduction:

and m

and ac

6 Fixeresearch

y⁎i = λyDi + 1− λy� �

Si ð6Þ

is the available stock of the firm, and λy∈ [0,1] is the speed at which the firm adjusts its level of planned output to the level of
qt−1,i
demand. The production target then defines the level of capital (dis-)investment It,1 that is required when the financial resourcesconstraint wt,i⁎ =wt−1,i+πt,i is positive:

It;i =λc min y⁎t;i − kt−1;i;w

⁎t;i

� �if y⁎t;i > kt−1;i

−λc min kt−1;i − y⁎t;i; kt−1;i

� �if y⁎t;i < kt−1;i

8<: ð7Þ

λc ∈ [0,1] represents the physical constraint of changing production levels in the short run. Accordingly, changes in capital
wherestock depend on the above investment rule, together with the financial resources available at the outset of period t,
kt;i =kt−1;i + It;i if w⁎

t;i > 0

max kt−1;i + wt;i;0� �

if w⁎t;i < 0

8<: ð8Þ

odify the stock of financial resources at the end of period t (which will be available at the outset of period t+1:

wt;i = w⁎t;i − kt;i − kt−1;i

� �: ð9Þ

fits are the difference between monetary sales and costs
Pro
πi = Sipi − ciyi: ð10Þ

e price of a given design is determined by a fixed mark-up on the unit cost of production (i.e. prices do not adjust to clear the
Thmarket).6 This is determined by the simple mark-up rule pi=(1+υ)ci. To simplify, we assume that the mark-up is the same for allfirms, and is fixed.
Costs ci depend on a fixed component F, and a variable component ch that is related to the combination of servicecharacteristics that comprise the current design.

ct;i =F

1 + yt−1;i

XhaZi

chx2t;i;h ð11Þ

Zi is the set of characteristics, and yt−1,i is the level of production.
whereFinally, firms produce each unit of output using a simple constant returns production function, with capital as the sole
input

yt;i = kt−1;i ð12Þ

cumulate inventories for the next period of sales qt,i=yt,i−St,i+qt−1,i.

2.3.2. Environmental impact of firms' designsFirms are heterogeneous with respect to the quality of service characteristics offered by their product designs, and the

consumer class that they target. These are randomly generated at the outset. The environmental impact of using a particular designis related to the environmental fitness that the producer's design attains. This is an average over the fitness of the Hi

z servicecharacteristics that form part of design x→i within a particular paradigm zi:

s Yxi� �

=

Ph∈zi

’z;i;h

Hzi

: ð13Þ

d mark-up pricing is a stylised fact found in many models of innovation (e.g. Nelson and Winter [12], and Malerba et al. [3]). The classic empiricaon mark-up pricing is Hall and Hitch [13], with more recent studies such as Blinder [14], and Hall et al. [15] finding widespread use of this cost-based

l

where

where

where

7 This


order to model the complex interaction between the service characteristics of a design, and their individual impact on the
Inenvironment, we assume that firms need to find the overall fitness of a design within a complex landscape. The shape of thislandscape is given by the interaction between the characteristics. That is, the environmental fitness of each characteristic dependson its own position on the multi-dimensional environmental–technological landscape, and on the position of the othercharacteristics that make up this landscape z:
’z;i;h =s zð Þ

1 + jxi;h − vi;h jð14Þ

s(z) is the maximum fitness (i.e. the global environmental peak) within paradigm z;

vi;h = χz;i +XgaZi

ag;h xi;g ≠ h ð15Þ

ag,h measures the strength and the direction of a change in xg on the fitness of xh;

χz;i = χˆ z +XhaZi

λzag;h ð16Þ

is the optimal point on the landscape for all the service characteristics xh within paradigm z. This is the environmentally
χz
optimal combination of service characteristics.Finally, environmental impact is a decreasing function of environmental fitness, with a steeper slope for intermediate levels of

fitness7:

1i =1

1 +s Yxi� �

− s0/

� �2 ð17Þ

e maximum environmental impact of goods, s0 the minimum level of fitness attainable, and ϕ a parameter that defines the
1 is thrate at which an improvement in environmental fitness reduces the impact on global pollution.
2.3.3. InnovationGiven that all firms in our model face the same underlying technology and same cost functions, and that there is a common,

fixed mark-up, innovation is the one means by which firms can improve their competitive position. Product innovation involvesthe creation and evaluation of new designs with different combinations of service characteristics, different prices, and differentialenvironmental performance (fitness). In effect, firms compete by offering consumers designs that represent distinct points withina multi-dimensional space of service characteristic/price/environmental performance.

In each time period, all firms perform product R&D. Modification of a service characteristic (i) changes the quality level of theservice, (ii) alters price in line with the change in quality of the service characteristics offered, and (iii) changes the overallenvironmental quality of the design. First, firms attempt to mutate all the product characteristics within their existing design.Thereafter they evaluate the results of this R&D against the utility function of their target consumer class. Following this evaluation,a firm decides whether to put the innovation(s) into production in the next period.

Mutation. In the manner of Nelson andWinter [12], there is a given probability l that R&D effort on each characteristic hwill besuccessful and will result in an innovation. When R&D on a characteristic xi,h is successful, innovation results in a variation of thequality of this service characteristic:

Δ xi;h = N 0;1ð Þ � ξ ð18Þ

N(0,1) is a random draw from a standard distribution, and ξ is a weighting factor that determines how localised the
whereinnovation process is.
Evaluation. If R&D is successful for at least one service characteristic, the firm evaluates the environmental fitness of theproposed new design (see Eqs. (13) and (14)), and the overall impact of this on the utility function of its target consumer class j. Itis assumed that each firm has perfect information about the utility function of its target consumer class (Eqs. (2)–(5)). Further, it isassumed that the class which a firm targets is fixed throughout the firm's lifetime.

The evaluation rule by which firms decide to implement a new design is:

xt;i;h =xt−1;i;h if PuΔx

j ≤ ―uj

min xt−1;i;h + Δxi;h;0� �

if PuΔxj > Puj

8h8<: ð19Þ

u jΔx is the utility that consumers belonging to class jwould attain, should they buy the new design in period t+1 from firm i.

is modelled as a modified truncated Cauchy function.


2.4. Paradigms

Environmental–technological paradigms affect the results of the model in a number of ways. First, consumers evaluate theenvironmental fitness of a firm with respect to the ‘environmental promise’ of a particular paradigm. Second, a firm targets oneclass within a particular paradigm and cannot switch to a consumer class in an alternative paradigm. Third, the full direct utilitythat a consumer class enjoys is limited by the particular combination of service characteristics, and the interactions between thesecharacteristics, as determined by the paradigm landscape.

From Eqs. (13) and (19) we know that firms attempt to reach the global environmental peak within a paradigm s (z), providedthis improves the utility of its target consumer class; i.e. depending on the preferences of its target consumer class towards trade-offs between environmental performance and price, and between environmental performance and quality of servicecharacteristics. When at least one firm reaches the environmental peak, a window of opportunity is opened and an exogenousresearch for a new, improved, paradigm commences. In practice, when s(z)− s(xYi)<N(0,σ z) for any firm i at time t, an exogenoussearch for a new paradigm commences. After τ z~U[τ z

Min, τ zMax] periods, exogenous scientific research identifies a new paradigm z

(i.e. a new environmental–technological landscape). New technology start-ups are able to explore this new paradigm throughtheir R&D activity.

In the model we assume that the landscape of the new paradigm contains an environmental peak that is higher than the peakof the previous paradigm. Hence, the introduction of a new paradigm induces an increase in maximum environmentalfitness

s zð Þt;z = s zð Þt−τz ;z + Δ’ ð20Þ

Δφ is a parameter.
whereOn the other hand, the introduction of a new paradigm changes the environmental landscape, altering the interactions between
service characteristics and their environmental impact:

χt;z = N χt−τz ;z;σχ

� �ð21Þ

σχ proxies the technological distance of the new paradigmwith respect to the preceding one. It should be noted that, while
wherethe characteristics in each paradigm can differ, the total number of service characteristics is assumed to be constant (Hz). Further, itis assumed that at least one characteristic is shared by two consecutive paradigms, zt∩zt−τz ≠ 0. Therefore, the number of newcharacteristics when a new paradigm is introduced is U(1,Hz−1).
Finally, the stock of global pollution depends on the environmental impact of firms' designs and their sales over time:

Gt = Gt−1 +Xi

1t;i � St;i: ð22Þ

te that a given rate of pollution decay is implicit in the relationship between environmental fitness and impact (Eq. (17)).
No
2.5. Market dynamics

Every τr time periods (where τr~U[τrmin, τrmax]), there is a replacement of less efficient firms by new technology start-ups, and of unpopulated consumer classes with new consumer classes that positively weight the characteristics of the newtechnology.

2.5.1. Consumer dynamicsAt τr, any consumer class j that is populated by less than a given level δc of individual consumers Cτr−1, j is replaced by a

new consumer class jτr. The new class is initiated by transferring these individual consumers. The new class has a new set of

preferences βτr, j,h~U[βMin, βMax] that positively weight the service characteristics of the new paradigm zτr> zt− τr. Followingτr, all new consumer classes are new technology users. These positively value the service characteristics βτr,j,h ~U[βMin, βMax]of the new technology paradigm. As a result, a number of new classes establish a market for the designs produced by newtechnology firms in the new paradigm.

2.5.2. Firms dynamicsAt τr all firms that have reduced their capital stock kτr−1,i belowa given level δk are replaced by newentrants. Each new firmbegins

with a capital stock and financialwealth equal to themarket average in period t−1 (i.e. kτr−1,i= kτr−1 andwτr−1,i=w—τr−1), and eachfirm has an initial inventory of finished products sufficient to satisfy consumer demand qτr−1,i=kτr−1,i. New firms also enter withdesigns containing ‘improved’ service characteristics xi,h that are modified with respect to the existing incumbent: Δxi,h=N(0,1) ξ.

Following τr, all new firms are new technology producers — provided there is at least one viable consumer class that canbe targeted. In this way, old paradigm service characteristics xi,h,zt− τr are gradually replaced by new paradigm characteristics xi,h,zτr~U[x_h, xh]; where x_h is the minimum, and xh the maximum value of service characteristics currently in the market. Finally, each newfirm randomly targets a (new technology) consumer class jτ

r

.


3. Results

The methodological approach that we adopt is expounded in detail in our companion paper. Briefly, we run batches of 10simulations for each configuration. For example, for a given distribution of environmental preferences and price preferences, werun 10 simulations using different random seedings for the stochastic variables of the model. Each individual simulation run lasts3000 periods and contains a population of 25 firms, and 500 consumers divided into 100 consumer classes. Given this largenumber of agents and the long time frame, each run effectively represents a time series. We average over the outputs of the 10simulation time series, and present them as a single ‘averaged observation’. By comparing the averaged observations of alternativeconfigurations, we evaluate the levels of global pollution generated by a set of alternative scenarios.

3.1. Scenario 1. trade-off between environmental performance and service characteristics

Distributions containing consumer classes that place a large weight on environmental performance and a (relatively) lowweight on the quality of service characteristics have a low trade-off preference. Ex ante, one would expect a consumer distributionof this type to create less pollution than a distribution containing consumer classes that place a high weight on the quality ofservice characteristics. This leads us to the first proposition to be examined;

Proposition 1. The higher the mean value of environmental preferences across a population of consumer classes, the lower the globalpollution, provided consumer preferences for direct utility are sufficiently low.

In order to test this proposition, we compare distributions with homogeneous environmental preferences η and character-istics preferences β. By altering these weights, we can compare the pollution generated by alternative distributions of consumerpopulations.

Fig. 1 shows the global level of pollution generated in the final period (i.e. in period 3000) for 10 alternative configurations. Themean values of environmental preferences are on the x-axis, while the mean values of characteristics preferences are on the y-axis.

The results presented in Fig. 1 support Proposition 1. Differences in theweight placed on environmental utility η directly affectsthe final level of environmental pollution, provided those consumer classes are relatively indifferent to the direct utility β providedby service characteristics. However, as predicted, as soon as direct utility preferences become relevant to the consumption choice, amuch higher preference for environmental utility is required if relatively low levels of pollution are to be achieved. When theweight placed on service characteristics is very high, even a large difference in the environmental preferences of a population ofconsumer classes will have little effect on the global level of pollution.

What is driving this result? Remember that average consumer utility will be high if consumer classes place a high meanweighton environmental preferences η, or, alternatively, a high mean weight on direct preferences β. Firms respond to the incentivesprovided by these consumer preferences, and focus their innovation effort accordingly — i.e. if a high mean value is placed onenvironmental utility across all consumer classes within the population, then firms will focus on the environmental performanceof their designs. There is a second condition under which firms prioritise environmental performance. This is when the global levelof pollution becomes so great that it turns average consumer utility negative. In this case, pollution becomes of critical concern toall consumers – regardless of the values of β and η – and firms have an incentive to develop designs that improve sustainability atthe expense of product quality.

Fig. 1. Aggregate pollution for different combinations of mean environmental preferences and mean characteristics preferences.

Fig. 2. Combined effect of environmental and direct preferences on average level of consumer utility and the evolution of paradigms.


Fig. 2 shows consumers' utility (Fig. 2a) and the number of paradigms (Fig. 2b) in period 3000 for different values ofenvironmental preference (x-axis) and service characteristic preference (y-axis). While the average utility enjoyed by aconsumer distribution with a high mean weight on environmental performance may be as high as a distribution with a highmean weight on service characteristics, there is a clear difference with regard to the rate at which new environmentaltechnologies are discovered. As one might expect, this is much lower for consumer populations with high averagecharacteristics preferences/low average environmental preferences. As a result, the final level of global pollution is muchhigher than it is for consumer populations with high average environmental preferences/low average characteristicspreferences.

Referring back to the historical case study of urban transport in our companion paper, early car designs were less polluting thanhorses and were slightly faster. Consumers were attracted to the car, leading to a switch in paradigm. However, the car was also aprivate consumption good. Once it became a status good, a way of distinguishing oneself, this affected the direction of innovationwithin this paradigm. Given consumer preferences, the economic incentives for firms were to develop new designs that improvedspeed, acceleration, and size. The onus therefore moved away from environmental performance to improving servicecharacteristics. Only recently, with increasing consumer concerns about the impact of intensive car use on the environmentand on people's health, has there been an impetus for firms to engage in R&D in hybrid electric–petrol engines andexperimentation with biofuels. Having said this, the current fashion for petrol guzzling SUVs serves as a warning that consumersare still only partly driven by environmental concerns.

3.2. Scenario 2. trade-off between environmental performance and price

Let us now consider a second potential trade-off; that between environmental preferences and price. Referring back to ourcompanion paper, price played an important role in the history of urban transport. It was the middle social classes that were ableand willing to buy cars, and who first enthusiastically took up new, car-based suburban lifestyles. Lower social classes could notafford this new lifestyle. As a consequence, the composition and the character of cities changed.

Fig. 3. Aggregate pollution for populations with different standard deviations environmental and price preferences.


The price preference of a consumer class is a proxy for the price elasticity of a group of individual consumers. This leads us to thesecond of our propositions;

Proposition 2. The level of global pollution depends on the distribution of consumer preferences with regard to trade-offs betweenenvironmental and price preferences. The larger the difference in trade-off, the higher the level of total pollution that is created.

Here we consider populations of consumer classes with heterogeneous environmental preferences and heterogeneous pricepreferences. By holding the mean averages constant and altering the standard deviations, we can evaluate the impact of opposingchanges in environmental and price preferences, i.e. where the direction of change in environmental preferences is offset by anequal and opposite change in price preferences.

In effect, we are setting up a scenario inwhich consumer distributions contain two distinct types of consumer class. At one end ofthe distribution there is a class containing individuals who are sensitive to environmental issues, and are wealthy enough to beunconcerned by the higher prices that firmsmay need to charge in order to produce cleaner designs. For this class, there is a low trade-off between price and environmental performance. At the other end of the distribution there is the opposite type of consumer class.This class has a high trade-off betweenprice andenvironmental preferences, i.e. it contains individualswhoareveryprice sensitive andare unconcerned about the environmental impact of their consumption (i.e. they place a very low weight on pollution).

Fig. 3 shows the time series of global environmental pollution generated between period 2000 and period 3000 fordistributions of consumer classes with different standard deviations of environmental preferences and price preferences.

The findings suggest that higher levels of global pollution are created as standard deviations increase. This supports Proposition2 that global pollution levels will be higher for consumer populations that contain some consumer classes that are highly pricesensitive and relatively unconcerned about the environmental impact of their consumption.

Fig. 4 helps us consider, in greater detail, the relationship between heterogeneous preferences for environmental performanceand price, and its impact on global pollution. Strikingly, the scatter plot of Fig. 4a indicates a positive relationship between theinitial degree of preference heterogeneity across classes and the final level of global pollution. Fig. 4b plots the results of a (verysimple) polynomial prediction of global environmental impact, given different standard deviations in environmental preferencesη. The findings are important because, in the absence of a price–environmental performance trade-off, a negative relationshipexists between the heterogeneity of environmental preferences and global environmental impact (as reported in our companionpaper in this Special Issue).

Finally, we report an unanticipated finding. Ex ante, one might reasonably expect the final level of consumer utility to remainconstant for distributions that contain offsetting variations in the standard deviations of environmental and price preferences. In fact,the final average level of consumer utility is slightly higher for distributions containing classes with very different trade-offs (i.e. withlarge standard deviations of around 0.3) than for distributions containing classes with equivalent trade-offs (i.e. with a zero standarddeviation). This is illustrated in Fig. 5a below. From Fig. 5bwe see the lower number of consumer classes and firms that can survive inthe casewhere there are large standard deviations in both environmental preferences and price preferences across consumer classes.

Both findings are driven by the market bifurcating into two distinct niches when there are large standard deviations inenvironmental and in price preferences. This is clearly indicated by Fig. 6a and b. The findings are analogous to those reported in[8]. In that paper the authors examined the implications of heterogeneous preferences in contemporaneous technologycompetitions. They found that strongly heterogeneous consumer preferences could lead to a bifurcation of the market into a set ofdistinct market niches (sub-markets), each supporting a competing variant of a new technology. In the case of sequential

Fig. 4. Relationship between heterogeneous preferences for environmental performance and price, and impact on global pollution.


technology competitions,wefind that strongly heterogeneous preferences support the co-existence of bothold andnew technologyparadigms. As Fig. 6b shows, individual consumers polarise into two distinct user groups. That is, a bi-modal distribution emerges.One group continues to support the old technologywhile the other champions the new technology. This prevents convergence to asingle consumer class, as would be the case if no trade-off existed between environmental performance and price (Fig. 6b).

On the one hand, sustained technological variety implies increased consumer choice, with slightly higher levels of final averageutility across the entire consumer population. On the other hand, the situation either prevents or slows down paradigmsubstitutions. The direct consequence of this is higher levels of global pollution. This is the downside of maintaining variety insequential technology competitions.

4. Conclusions

Let us summarise the findings of this paper, and position these with respect to the findings presented in the companion paper.In the companion paper we considered the environmental impact of variations in the standard deviations and in mean averagelevels of environmental utility amongst distributions of consumer classes. In this paper we have widened considerably thediscussion by examining the impact of consumer trade-offs between environmental utility and direct utility, and betweenenvironmental utility and indirect utility.

Taken together, the papers identify two general results. The first is the importance of distributions of consumer preferences.There are two aspects here. The first is the particular distribution that exists at a particular moment in time, such as when a new

Fig. 5. The effect of heterogeneous environmental and income preferences on consumers' utility and on market concentration of firms and consumer classes.


technology paradigm is introduced. This has a significant bearing on the direction of innovation, on paradigm substitutions and,hence, on the level of global pollution. The second aspect is how a distribution evolves over time. This also has a fundamentalimpact on global pollution. In part, the evolution is due to consumer learning, as captured by the replicator algorithm in our model.In part, it is due to the interaction between consumer learning and the innovative activities of firms.

This leads us to the second general finding. Firms respond to the incentive structures provided by the distribution of consumerpreferences within amarket. Firms have a strong incentive to improve the environmental performance of their products under oneof two conditions. First, if a high value is placed on environmental utility by the consumer classes within the market. Second, if thelevel of global pollution rises to such an extent that the average level of utility of all consumer classes becomes negative.

Taken together, these findings indicate that heterogeneous preferences are a major factor determining the direction ofinnovation and, hence, the global level of pollution that is generated over time. In this respect, the contention that competitivemarkets can, in principle, produce low-carbon and high-efficiency goods and services is supported. Having said this, the findings ofthis paper clearly indicate that significant variations in global pollution can be generated by relatively small differences in thedistributions of consumer preferences for trade-offs between environmental utility, direct utility, and indirect utility.Consequently, one should be cautious about solely relying on consumer induced innovation to achieve a particular (target)level of pollution reduction. Government intervention may be required.

Let us now consider the specific implications of our findings with regards to consumer trade-offs, paradigm substitutions, andglobal pollution. The first trade-off we considered was between environmental utility and direct utility. This depends on therelative weights placed on environmental performance and on service characteristics. In the companion paper we find that anegative relationship exists between global pollution and the mean preference for environmental utility, i.e. lower levels of global

Fig. 6. Distribution of consumers across classes over time: homogeneous versus heterogeneous environmental and income preferences.


pollution are created by consumer distributions with higher mean weights on environmental utility. In this paper we find theresult is strongly tempered by preferences for direct utility. A negative relationship will hold provided mean preferences for directutility are low. As soon as direct utility becomes relevant to the consumption choice, a much higher mean preference forenvironmental utility is required for relatively low levels of pollution. The situationworsens when a highmeanweight is placed onservice characteristics. Under these circumstances, firms have a strong incentive to focus innovation on improving servicecharacteristics, to the detriment of environmental performance. As a consequence, the rate of paradigm substitutions is very low,and there are high levels of global pollution.

The history of the car provides a particularly good example of this. The pollution and ill health associated with horse basedtransport was such that it negatively affected all urban dwellers' utility. This set up a strong desire for alternative, less pollutingtransport options. These were eventually provided by the car. However, as car-based transport systems replaced horses, the focusswitched to improving engine performance (i.e. speed and acceleration), rather than improving environmental performance. To befair, early adopters were unaware of the environmental impact of intensive car use on the environment and on people's health. Asthis has become better understood, so interest in alternatives to the petrol engine has grown. And yet, still, it is clear that manyconsumers are not willing to sacrifice too much, in terms of engine performance. Worse, the current fashion, amongst someconsumer types, for petrol guzzling SUVs provides a salutary warning that some consumers continue to discount theenvironmental impact of their consumption in favour of service characteristics. It appears, then, that car users are not yet willing totrade-off significant reductions in engine performance for dramatic improvements in environmental pollution, or to considerswitching from their cars to alternative modes of transport. This helps explain why car manufacturers have undertaken certaintypes of R&D and not others: hybrid electric–petrol engines and biofuel engines offer similar engine performance while delivering‘some’ environmental benefit.

The problem is further compounded by the fact that the car remains an important status good in our society. Consumersassociate themselves strongly with the characteristics of different types of cars. Owners of sports cars, SUVs, small family cars, andso on are signalling something about themselves — to themselves, and to others. In addition to use value, goods provide symbolic


value. Both types of value determine the direct utility that a consumer enjoys when purchasing and consuming a good over itslifetime. The issue of signalling has been addressed, among others, by the Douglas and Isherwood [16]. All goods carry ‘meanings’.Sometimes they are placed there by us and drawn from the repository of shared cultural practices in our society. At other timesthey are insinuated into goods by advertising, marketing, and promotion. These meanings enable people to make sense of theirworld, to communicate with others, and to express their collective and personal identities. The importance of the car, as asignalling good, is such that people in developed countries are not willing to give them up, and people in developing countriescontinue to aspire to car ownership as a means of signalling success and social status.

Let us now turn to the trade-off between environmental utility and indirect utility (price). In our companion paper we reportedthat a negative relationship exists between pollution and the standard deviation of environmental preferences, i.e. the higher thehigher standard deviations in environmental preferences, the lower the level of global pollution. This suggested a case could bemade for supporting the diffusion of the lifestyles developed by eco-warriors. The physicist Philip Anderson has observed that“much of the real world is controlled as much by the ‘tails’ of distributions as by means or averages: by the exceptional, not themean… by the very rich not the middle class” ([17, p.566]. The idea that one tail of the social distribution sets the agenda, andothers eventually follow their lead, is not new in innovation studies. Rogers' [18] work on the diffusion of innovations, dating backto the 1960s, has been highly influential in policy circles. In a nutshell, this states that pioneering consumer types are to be foundamongst the upper and the upper–middle classes. These are able and willing – financially and cognitively – to experiment withnew technology goods, and who have the requisite social standing and social capital to influence others by their choices. Rogersargues that, if a new technology proves successful amongst these social classes, then its use will diffuse downwards to lower socialclasses.

What if there is not a uni-directional, hierarchically structured set of social relationships? Note that in the current specificationof our model, consumer classes do not represent social classes. The effect of income distributions, which empirically tends to betightly correlated with education and social class, is work for future research. In this paper we considered distributions thatdiffered with respect to price elasticity and to their environmental views. At one end of this distribution lies a group that is willingto pay a higher price for environmentally friendly goods — the eco-warrior class. At the other end of the distribution lies a groupthat is not. This price sensitive consumer class may comprise individuals that are less well off, and so unable to pay the higher priceof more environmentally benign goods. Equally, however, theymay simply be ‘petrol heads’; individuals who discount informationabout the impact of their car consumption on the environment, who doubt discussions about global warming and, hence, aresimply unwilling to pay for an alternative technology. Such attitudes are not restricted to a particular social group. Hence, theability to explore a set of alternative distributions is a strength of the current model.

By increasing the standard deviation, we have considered what happens when both tails of this distribution becomeincreasingly pronounced. The results were striking. First, and most noticeably, the global level of pollution is higher when a trade-off exits. Second, a positive relationship exists between global pollution and the standard deviation of environmental and pricepreferences. This contrasts sharply the negative relationship, discussed above, between pollution and the standard deviation ofenvironmental preferences which exists in the absence of a trade-off.

We found that, for large standard deviations, the market can bifurcate into two distinct niches. Responding to the incentivesprovided by such a heterogeneous consumer population, some firms develop clean products that cater to environmentallysensitive/price inelastic consumers. Other firms cater to environmentally insensitive/price elastic consumers by developingdesigns with inexpensive service characteristics but poor environmental performance. The net result is a higher level of pollution.The unexpected finding was that the final average level of consumer utility is slightly higher for distributions with large standarddeviations. Variety, it seems, can be positive for consumer choice while being negative for the environment. In terms of policy, thefindings force one to reconsider the efficacy of a policy that focuses on an eco-warrior class. It now appears simplistic to assume theautomatic diffusion of the lifestyles and consumer choices developed by such a consumer class. Depending on the distribution ofconsumer classes in a market, policy may well need to consider, and deal with, both ends of the distribution if it is to support thegreening of demand and achieve an improvement in environmental sustainability.

To summarise, the findings of the paper highlight the importance of heterogeneous consumer preferences in the developmentof environmentally friendly technologies. Different distributions of consumer preferences for environmental sustainability,product quality, and price fundamentally influence the type of innovation undertaken by firms and the rate of paradigmsubstitution. The net consequence is a quantitative difference in the rate of global pollution. In future papers wewill examine otheraspects of this co-evolutionary model of paradigm substitutions. In particular, we will consider the effect of income distributionsand the implications of non-modular technologies for the current set of results. Wewill so seek to calibrate themodel for a numberof different markets. This will provide a secure platform on which to develop technological forecasts and to discuss the likelyimpact of alternative policy interventions.

References

[1] W.J. Abernathy, J.M. Utterback, A dynamic model of product and process innovation, Omega 3 (1975) 639–656.[2] S. Klepper, Entry, exit, growth and innovation over the product life cycle, Am Econ. Rev. 86 (3) (1996) 562–583.[3] F. Malerba, R.R. Nelson, L. Orsenigo, S.G. Winter, History friendly models of industry evolution: the computer industry, Ind. Corp. Change 8 (1) (1999) 3–41.[4] O. Shy, Technology revolutions in the presence of network externalities, Int. J. Ind. Organ. 14 (1996) 785–800.[5] K.J. Lancaster, A new approach to consumer theory, J. Polit. Econ. 14 (1966) 133–146.[6] P. Windrum, Neo-Schumpeterian Simulation Models, in: H. Hanusch, A. Pyka (Eds.), The Edward Elgar Companion to Neo-Schumpeterian Economics, Edward

Elgar, Cheltenham, 2007.


[7] P. Windrum, C. Birchenhall, Is life cycle theory a special case?: dominant designs and the emergence of market niches through co-evolutionary learning,Struct. Chang. Econ. Dyn. 9 (1998) 109–134.

[8] P.Windrum, C. Birchenhall, Structural change in the presence of networkexternalities: a co-evolutionarymodel of technological successions, J Evol. Econ.15 (2)(2005) 123–148.

[9] P.P. Saviotti, J.S. Metcalfe, A theoretical approach to the construction of technological output indicators, Res. Policy 13 (1984) 141–151.[10] W.B. Arthur, Competing technologies: an overview, in: G. Dosi, C. Freeman, R. Nelson, G. Silverberg, L. Soete (Eds.), Technical Change and Economic Theory,

Pinter, London, 1988.[11] P. David, S. Greenstein, The economics of compatibility standards: an introduction of recent research, Econ. Innov. New Technol. 1 (1990) 3–41.[12] R.R. Nelson, S.G. Winter, An Evolutionary Theory of Economic Change, Harvard University Press, Cambridge, Mass., 1982.[13] R.L. Hall, C.J. Hitch, Price theory and business behaviour, Oxford Economic Papers, 2 (1939) 12–45. Reprinted in: T. Wilson, P.W.S. Andrews (Eds.), Oxford

Studies in the Price Mechanism. Clarendon Press, Oxford (1951).[14] A.S. Blinder, Why are prices sticky? Preliminary results from an interview study, Am. Econ. Rev. 81 (2) (1991) 89–96.[15] S. Hall, M. Walsh, A. Yates, How do UK companies set prices? Bank of EnglandWorking Paper No. 67, 1997, available at < http://papers.ssrn.com/sol3/papers.

cfm?abstract_id=114948>.[16] M. Douglas, B. Isherwood, The World of Goods: Towards an Anthropology of Consumption, Routledge, London, 1979.[17] P. Anderson, Some thoughts about distributions in economics, in: W.B. Arthur, S. Durlauf, D. Lane (Eds.), The Economy as a Complex Evolving System II,

Addison-Wesley, New York, 1997.[18] E.M. Rogers, Diffusion of Innovations, 5th ed.Free Press, New York, 2004.

http://papers.ssrn.com/sol3/papers.cfm?abstract_id=114948

http://papers.ssrn.com/sol3/papers.cfm?abstract_id=114948

environmental impact, quality, and price: consumer trade-offs and the development of environmentally...

Documents