ArticlePDF Available

Abstract and Figures

To date, the majority of the research literature on the impacts of climate change has addressed the negative aspect, i.e. the risks associated with a future permanent modification of climate. Potential opportunities have received much less attention and are rarely transformed into monetary values. Furthermore, manufacturing is one of the economic sectors where the influ-ence of climate change remains practically unknown, although the economic performance of some industrial activities depends directly on climatic conditions – bottled water and non-alcoholic beverages (i.e. soft drinks and fruit juices) are among these. This paper aims to explore the link between weather and product sales in these sectors, and estimate in quantitative terms the potential impact of future climate change on their revenues. Historic data were explored through statistical analysis and appropriate regression models were developed. Models were applied for the historic (1961–90) and future climate (2021–50) and the difference in sales forms the expected quantified impact of climate change. The results indicate that significant opportunities may arise for some sectors from modifica-tions in climate, provided their production infrastructure can meet the expected demand and their management strategies can successfully adapt to altered climatic conditions.
Content may be subject to copyright.
The Impact of Climate Change on the Pattern of
Demand for Bottled Water and Non-Alcoholic
Beverages
S. Mirasgedis,
1
*E. Georgopoulou,
1
Y. Saradis,
1
K. Papagiannaki
1
and D. P. Lalas
2
1
National Observatory of Athens, Athens, Greece
2
FACE
3
TS S.A., Athens, Greece
ABSTRACT
To date, the majority of the research literature on the impacts of climate change has addressed
the negative aspect, i.e. the risks associated with a future permanent modication of climate.
Potential opportunities have received much less attention and are rarely transformed into
monetary values. Furthermore, manufacturing is one of the economic sectors where the inu-
ence of climate change remains practically unknown, although the economic performance
of some industrial activities depends directly on climatic conditions bottled water and
non-alcoholic beverages (i.e. soft drinks and fruit juices) are among these. This paper aims
to explore the link between weather and product sales in these sectors, and estimate in
quantitative terms the potential impact of future climate change on their revenues. Historic
data were explored through statistical analysis and appropriate regression models were
developed. Models were applied for the historic (196190) and future climate (202150)
and the difference in sales forms the expected quantied impact of climate change. The
results indicate that signicant opportunities may arise for some sectors from modica-
tions in climate, provided their production infrastructure can meet the expected demand
and their management strategies can successfully adapt to altered climatic conditions.
Copyright © 2013 John Wiley & Sons, Ltd and ERP Environment.
Received 7 September 2012; revised 18 December 2012; accepted 10 January 2013
Keywords: climate change; impacts; industry; Greece; water; beverage
Introduction
ANUMBER OF STUDIES HAVE INVESTIGATED IN QUANTITATIVE TERMS THE RELATIONSHIP BETWEEN TEMPERATURE OR
other meteorological parameters and the level of economic activity at either the sectoral (see for example
Bahng and Kincade, 2012) or the whole economy level (e.g. Xu and Sun, 2011). In that respect, although
a lot of research on the adverse effects of future climate change on physical and economic systems has
been carried out and published, not much has been said to date on the economic opportunities that may arise from
such a change. The consumption of number of products especially in the manufacturing sector is greatly inuenced
by weather conditions and thus their sales will probably be favorably affected by a future modication in climate.
*Correspondence to: S. Mirasgedis, National Observatory of Athens, Lofos Nimfon, Thesion, Athens 11810, Greece. E-mail: seba@noa.gr
Copyright © 2013 John Wiley & Sons, Ltd and ERP Environment
Business Strategy and the Environment
Bus. Strat. Env. (2013)
Published online in Wiley Online Library
(wileyonlinelibrary.com) DOI: 10.1002/bse.1782
Thus, while for power companies, households and most activities in the tertiary sector, heat waves represent a clear
risk, for example by increasing further peak electricity demand for space cooling, for others, including manufac-
turers of beer, ice cream, soft drinks, juices, and bottled mineral water, these weather conditions are highly bene-
cial as high temperatures (usually during summer) are in most of cases associated with increased product sales.
The importance of weatherand climateon business activities and the need to develop appropriate methods and tools
for companies to deal with the potential risks but also opportunities associatedwith climate change have been identied
and emphasized recentlyin several studies. Winn et al. (2011) have presented a conceptual framework for organizations
and businesses to deal with climate change that takes in account the management of sustainability, crisis, risk, resil-
ience and adaptive organizational change. Weinhofer and Busch (2012) in an exploratory case study in 11 electric utilities
showed that while climate change is perceived as a material issue for their business, the implemented climate risk man-
agement follows the same approach used for other business risks because of incomplete knowledge of the effects of
climate change. Wedawatta et al. (2011) explored the vulnerability of construction small and medium sized enterprises
(SMEs) in the UK to weather extremes and potential responses and identied a lack of appropriate coping strategies.
Acknowledging the importance of weather for their activity, a number of companies and industry associations
have also analyzed the annual variation of sales in relation to climatic variables and include this kind of analysis into
their annual reports at either company or sectoral level (e.g. British Soft Drinks Association, 2008; BCI, 2009).
For example, in the UK Soft Drinks Report for the year 2008, the section analyzing the weather conditions dur-
ing that year starts as follows: It was a promising start. In the absence of April showers, the UK basked in a balmy
sun-lled month, fuelling expectations of heat wave conditions to come. Similarly, the Irish Soft Drinks Report for
the same year states that ...May set expectations soaring with the lowest rainfall and highest temperatures recorded
for the ve years from 2003.
With regard to soft drinks in particular, the weather is considered a decisive factor for the industrys sales and the
correlation between sales and temperature has been demonstrated in several studies (see for example Murray et al.,
2010 and Dubé, 2004). An additional critical parameter is precipitation; as soft drinks are usually consumed out-
doors during the summer, continuous or frequent rain events will decrease the number of people out and about.
Sunshine duration may also represent a complementary variable that may be examined (see for example British Soft
Drinks Association, 2009; BCI, 2009).
A large seasonal increase in consumption as a result of weather conditions may represent not only an opportunity
but also a risk for a company if it is not well-prepared for this increase. For example, the existing production capacity
and safety stock of a company may not be sufcient to satisfy the higher demand, resulting in a temporal loss of
market share that may have further consequences (WTI, 2009).
It should also be noted that reports on soft drinks sales nationally reveal that although favorable weather condi-
tions during the summer signicantly affect seasonal consumption, their effect on the total annual consumption is
much less pronounced or even negligible, as non-climatic factors (e.g. household expenditure, availability of new or
specically targeted products) seem to play a more determining role. For example, although the summer of 2008 in
Ireland was very rainy compared with 2006 (August 2008, precipitation 150 mm; August 2006, 60 mm), the
annual consumption of soft drinks was more or less the same (2008, 858 million liters; 2006, 839 million liters).
Thus, in order to better estimate risks and opportunities associated with weather conditions, it seems important
to develop a better understanding of the role of weather on sales by developing a quantitative model that can forecast
the level of sales/ consumption of beverages as a function of climatic as well as non-climatic parameters. Although
large manufacturers may have developed such models for their own use, the information published on this issue is
scant. Regression analysis seems to be the main methodological tool applied in order to explore the link between
weather variables and consumption/sales. Typically, correlation is determined on a month-by-month basis. In the
cases of other beverage products (e.g. beer), this was based on a temperature threshold (usually 15 C) above which
an increase in consumption is triggered (Blom, 2009). Thus, an index called beverage degree days(BDD) was for-
mulated on the basis of maximum temperature and used similar to the way heating degree days (HDD) or cooling
degree days (CDD) are used in energy analysis. By developing a linear regression model involving BDD for a large
Norwegian brewery, Blom (2009) found that the determination coefcient (R
2
) of the model was 64%, i.e. 64% of
the changes in sales are associated with changes in accumulated BDDs. In addition, the model developed showed
that a 1 C of increase in maximum temperature during the summer season (MaySeptember) results in a 3.4%
increase in the industrys sales during this season.
S. Mirasgedis et al.
Copyright © 2013 John Wiley & Sons, Ltd and ERP Environment Bus. Strat. Env. (2013)
DOI: 10.1002/bse
Statistical analysis needs to treat different product families separately since the inuence of weather conditions
on sales is not the same. For example, in a recent study (Ramanathan and Muyldermans, 2010) analyzing and quan-
tifying the relative inuence of a number of factors on the sales of a large UK soft drinks company, temperature was
found to be an important factor for the 2 liter and 500 ml bottles but not for the 330 ml cans. Such differences are
associated not only with market factors such as where each product family is sold (supermarkets, outdoor stores,
etc.) but also with regional/ national particularities (e.g. types and locations of tourism activities).
It also needs to examine the suitability of various measures of the same variable, as for example maximum, mean
or apparent temperature (which incorporates somewhat the effect of humidity), but also the absolute level of average
local temperatures which varies substantially from location to location.
In view of the above, this paper aims to explore and quantify the link between climatic conditions and the sales of
bottled water and other non-alcoholic beverages. Alcoholic beverages have not been included as their sales are af-
fected by social factors that cannot be easily quantied and disaggregated. For the purposes of this study, a southern
European country (Greece) where high temperatures are often reached and their effects should be clearly felt has
been chosen. A second reason for choosing a southern country is the recent increase of heat waves there, occur-
rences of which are expected to become the norm in the future (Hansen et al., 2012), based on all predictions carried
out by numerous research groups worldwide of impending changes in climate. The possible inuence of climate
change on sales affects long-term business development planning, especially in choices of location of plants and
warehouses. To address these aspects, one needs to rst develop appropriate methodology to specify relevant forms
and their annual variations of the climatic variables affecting consumption over decades in the future. This method-
ology can also be useful in similar studies of the impact of climate change on other consumer goods, thus providing
useful information for relevant business risks and related costs.
The structure of this paper is as follows: the next section presents the methodological framework developed for
estimating the dependence of bottled water and other non-alcoholic beverage sales on weather variables, then we
describe the data used in the study. Following this we address the possible future implications on demand for
these products in Greece under climate change. Finally, we summarize the main ndings of the study and draw
our conclusions.
Methodological Framework
Overview
The methodological framework developed and implemented in this study comprises three basic stages. First, a
demand model for sales is constructed that aims to quantify the extent to which the demand for several products
classied under the beverage sector (e.g. juices, soft drinks, bottled mineral water and carbonated water) is a func-
tion of climatic and socio-economic factors. For this, a statistical approach (multiple regression models) based on
historical monthly data for the last 10 years (i.e. 200211) for Greece is applied. Second, the results from a regional
climate model are used to specify the changes in climatic conditions in Greece in the upcoming decades (i.e. the
period 202150) compared to historic climate (196190). The A1B scenario of the Intergovernmental Panel for
Climate Change (IPCC) Special Report on Emission Scenarios (SRES) has been chosen for the present analysis,
which is considered as a scenario of medium climate change intensityand matches with all EU policies on
emission restrictions until 2020. Third, the regression models developed previously are used to estimate future
demand of the examined products in Greece that can be attributed to climate change. Details for the rst two
parts follow.
The Demand Model
In the context of this study, several multiple regression models have been developed and tested in order to quantify
the inuence of various climatic and socio-economic factors on the monthly demand for various products of the
beverage sector in Greece.
Climate Change and Demand for Bottled Water and Non-Alcoholic Beverages
Copyright © 2013 John Wiley & Sons, Ltd and ERP Environment Bus. Strat. Env. (2013)
DOI: 10.1002/bse
Several factors, internal and external parameters to the industry (Agnew and Thornes, 1995), inuence the sales
of specic products of the beverages industry in a region. The former comprise the product prices (in general, as the
prices go down the demand goes up), the total amount of money spent in advertising campaigns, the in-store
promotions, the availability of competitive products, the level of development of distribution networks, etc. The
latter include the number of inhabitants and visitors (dening the size of the market), general economic condi-
tions inuencing the available income for spending, consumer preferences and social trends. They also comprise
weather variability (e.g. one is more likely to buy a soft drink in hot weather), holiday periods, other special conditions
(e.g. bottled water sales are higher in areas with degraded water resources), etc.
The development of analytical models for simulating the demand for specic beverages in relation to the above-
mentioned parameters requires a signicant amount of data in appropriate temporal analysis, which, in general are
not always available. In order to deal with this difculty, for each product examined in the context of this analysis we
have developed and tested several models, which incorporate a limited number of robust economic drivers, as well
as several meteorological parameters, as explanatory variables. The models nally selected, incorporate only the
statistically signicant independent parameters according to the results of the statistical analysis.
Regression models developed in the context of this study in order to simulate the demand of bottled water and
non-alcoholic beverages have the following general form:
log DEt
ðÞ¼cþX
i
ailog PMi;t

þX
j
bjlog SMj;t

þX
k
dklog Ek;t

þft þX
4
l¼2
plSlt þqSEtþet(1)
where c,a
i
,b
j
,d
k
,f,p
l
(lfrom 2 to 4) and qare the coefcients to be estimated from the regression analysis and e
t
is
the residual term.
The dependent variable (total amount of sales of a specic product in month t) appears in the above equation in
logarithmic form in order to reduce the serial correlation and the heteroscedasticity of the models developed
(Makridakis et al., 1998). Thus, possible loss of forecast accuracy, which is sometimes caused when the independent
variables of a forecast model take in the future values that differ substantially from those in the past on which the
regression of the model has been based, is avoided. We have examined four products, namely juices, soft drinks,
mineral bottled water and carbonated water, considering the total volume of sales, as well as the demand per product
and packaging size.
The explanatory variables used in the model comprise:
A number of primary meteorological parameters (PM
i,t
), namely the maximum, mean and minimum temperature
(T
max
,T
mean
and T
min
correspondingly), precipitation (Pr), sunshine duration (SD) and relative humidity (RH).
Secondary meteorological parameters (SM
j,t
) aiming at describing in a more effective way the meteorological
parameters which inuence consumer preferences and behavior. To this end, we have used the apparent temper-
ature (T
app
), which is estimated on a monthly basis from the corresponding values of maximum temperature and
relative humidity using the following formula (Steadman, 1984):
Tapp ¼1:3þ0:92Tmax þ2:2e
e¼RH
100 6:105 exp 17:27Tmax
237:7þTmax

101(2)
A number of selected economic variables (E
k,t
) aiming at describing the exogenous economic environment.
To this end, we have used the number of overnight stays during month t, which serve as a measure of the tourist
activity in the region, and the household expenditure in month twhich serve as a measure of the overall economic
environment and the willingness of the permanent population to spend money on consumer goods.
The time (t) in months aiming at taking into account the potential long-term trends in the monthly sales of the
product in question. This trend can be related to social, economic or demographic factors that result in a systematic
change to the total volume of sales in case the analysis undertaken on a sub-sectoral level or to a variety of parameters
which inuence the market share of a specic enterprise in case the analysis is undertaken at a company level.
S. Mirasgedis et al.
Copyright © 2013 John Wiley & Sons, Ltd and ERP Environment Bus. Strat. Env. (2013)
DOI: 10.1002/bse
Three dummy variables (S
l,t
), which aim at representing the potential seasonal variability in the sales of each prod-
uct that does not relate to specic weather conditions. An index ltaking values within the interval [2,4] has been
used, to represent three seasons of the year (i.e. l= 2 for spring, l= 3 for summer and l= 4 for autumn). The con-
sumption during winter is the baseline consumption in the model. It should be noted that any season of the year
could have been selected as the baseline season for the model without affecting the results of the analysis. Each
dummy variable has only two allowable values, 0 or 1. The variable S
l,t
takes the value of 1 if the t-observation
belongs to season I and 0 otherwise. We use one dummy variable less than the number of periods (four seasons
per year) to avoid multi-collinearity problems (Makridakis et al., 1998).
One dummy variable (SE
t
), with a view to cover months of the year when special events (e.g. holidays) inuence
the demand for a product. This variable takes the value of 1 if the month tincludes special events such as Christmas
and the mid-August holiday period and 0 otherwise.
Again, a logarithmic form was preferred in order to express the meteorological and economic parameters incor-
porated in the models developed, in order to avoid serial correlation and heteroscedasticity. Furthermore, it should
be pointed out that these models, though able to produce (given the simplications adopted and the uncertainty in
the future inputs) only approximate estimates of the future long-term demand of bottled water and non-alcoholic
beverages, can provide considerable insight on the potential changes attributable to weather variability.
Estimation of Future Regional Climate
Projections of anthropogenic emissions of greenhouse gases (GHGs) and other constituents are necessary to predict
future climate change. Since the long-term evolution of GHG emissions is characterized by large uncertainties, a
range of emissions scenarios has been generated by the IPCC on the basis of explicitly different packages of
socio-economic and technological conditions (storylines). Four qualitative storylines are usually considered that
yield four sets of scenarios called families: A1, A2, B1 and B2. They cover in a consistent way a wide range of
key future characteristics such as demographic change, economic development and technological change, resulting
in a substantial variation of GHG emissions (IPCC, 2000).
In the present analysis, the A1B scenario of the IPCC/SRES has been chosen, as it is considered a scenario of
medium climate change intensityand is consistent with EU policies on emission restrictions until 2020. In order
to specify the changes of climatic conditions in Greece in the upcoming decades (i.e. during 202150) compared
with historic climate (196190) the results of a regional climate model (RCM), namely the RACMO2 climatic model
developed by the Royal Dutch Meteorological Institute (Meijgaard et al., 2008), utilizing the A1B scenario emissions
pathway have been used.
The use of a RCM rather than the use of a global climate model (GCM) has been dictated by the considerable
variation in terrain characteristics (long coastlines in close proximity to mountainous regions as is the case in the
Mediterranean) which results in considerable spatial climatic variations that cannot be discernible with the low
resolution (a few hundred kilometers) of GCMs (CCSP, 2008).
In an attempt to construct comparable results on a monthly basis for a 30-year period, it was decided that
simple mean values of meteorological parameters for each month averaged over the whole period would not
capture the true variability present in both the past and the future 30-year period. Instead, the well-
established approach for long-term analysis of making use of a typical meteorological year (TMY) was preferred,
which, however, would be a special purpose TMY taking into account the ranking of the variables as identied in
the demand model.
The development of the special purpose TMY is based on the selection, following statistical criteria, of actual sin-
gle months from a multi-year database that are then merged into a TMY (Argiriou et al., 1999). Among the different
methods available for TMY generation, the Sandia National Laboratories method using FilkensteinSchafer (FS)
statistical method (Hall et al., 1978) is selected, as it is one of the most common methodologies used for generating
a TMY (Argiriou et al., 1999; Sawaqed et al., 2005; Skeiker, 2007; Jiang, 2010). The daily values computed by the
RACMO2 climatic model are utilized for the selection of a typical month on the basis of seven daily meteorological
parameters, namely the maximum, mean and minimum temperature, precipitation, sunshine duration and relative
humidity. The procedure is as follows:
Climate Change and Demand for Bottled Water and Non-Alcoholic Beverages
Copyright © 2013 John Wiley & Sons, Ltd and ERP Environment Bus. Strat. Env. (2013)
DOI: 10.1002/bse
Step 1: For each meteorological parameter mentioned above and each year of the selected 30-year period (i.e. 196190
or 202150), the daily values of the parameter during each month of the year are sorted (short-term sorting) in increas-
ing order (x
1
,x
2
,...,x
n
). In addition, for the same meteorological parameter and each month of the year, the daily
values of the parameter during that month and all years of the selected 30-year period are sorted in increasing order
(long-term sorting). Next, for each daily variable xin the sorting, the cumulative distribution function CDF
n
(x)is
calculated with the following function:
CDFnxðÞ¼
0forx<x1
k0:5
nfor xkx<xkþ1
1forxxn
8
>
<
>
:
(3)
where CDF
n
(x) is the value of the cumulative distribution function for daily variable x,nis the total number of sorted
elements and kis the rank order number. From its denition, CDF
n
(x) is a monotonically increasing step function
with steps of size 1/noccurring at x
i
and is bounded by 0 and 1. The calculation of CDF
n
(x) is performed both over
short-term and long-term sorting.
Step 2: The FS statistics for each month mof a specic year yare calculated according to the following equation:
FSxy;mðÞ¼
1
NX
N
i¼1
CDFmxi
ðÞCDFy;mxi
ðÞ
(4)
where FS is the FS statistics, CDF
m
is the long-term (30 years in this study) and CDF
y,m
is the short-term (for the year y)
cumulative distribution function ofthe daily variable xfor month m,x
i
is a sorted element in a set of Nobservations, and
Nis the number of daily data for month m.
Step 3:AweightingfactorWF
x
is applied to the FS statistics for each meteorological parameter x(FS
x
). In this way, a
weighted sum, WS, is derived for each month of each year in the selected 30-year period (196190 or 202150):
WS y;mðÞ¼
1
mX
m
i¼1
WFxFSxy;mðÞ (5)
where WF
x
are the weighting factors, one for each variable, and mis the number of meteorological parameters consid-
ered (seven in this study). For each month of the year, 5 months having the smallest weighted sum of the FS statistics of
the seven meteorological parameters are selected.
Step 4:Tonally select the typical meteorological months from the ve candidate months identied at the end of Step 3,
the simpler selection process introduced by Pissimanis et al. (1988), which uses the root mean square difference
(RMSD) of the key parameter (solar radiation in the original method) as the selection criterion, is applied:
RMSD ¼
ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
X
N
i¼1
KPy;m;iKPm
N
v
u
u
u
u
t(6)
where Nis the number of daily readings for the month, KP
y,m,i
is the daily values of the key meteorological parameter of
year y, month mand day i,andKP
m
is the long term mean daily values of the key meteorological parameter.
S. Mirasgedis et al.
Copyright © 2013 John Wiley & Sons, Ltd and ERP Environment Bus. Strat. Env. (2013)
DOI: 10.1002/bse
Data
The food and beverage industry constitutes one of the most signicant and dynamic manufacturing sectors of the
Greek economy. In 2009, the sector accounted for 25.3% of the total turnover of manufacturing, produced 31%
of the relevant overall gross added value and employed approximately 22.9% of workers in industry. The sub-
sectors of bottled water and soft drinks (NACE Rev.2 code 1107) and juices (NACE Rev.2 code 1032) account for
15% of the total sales in the food and beverage industry.
The regression models presented in this paper were developed utilizing historical data for sales for specic
beverage products provided by two large Greek enterprises, covering a 10-year period from 2002 to 2011. These
companies account for a signicant percentage of the market for bottled water and non-alcoholic beverages and
therefore can be considered to be a guidefor the behavior of the whole sector against weather and future climate
change. More specically, the company that provided sales data for bottled mineral water and carbonated water has a
large distributional network covering almost all the Greek territory. The second company that provided data for
juices and soft drinks covers mainly the metropolitan area of the capital of Greece (i.e. Athens, where approximately
40% of the permanent population as well as a large number of visitors reside). The data set used comprises monthly
sales data (in volume), disaggregated per product and type of selling pack (e.g. 500 ml bottle, 330 ml can etc.).
Figure 1 shows the sales prole of the four products examined (bottled mineral water, carbonated water, juices
and soft drinks) over the 10-year period (200211). As shown, the sales of all beverages in question exhibit a signif-
icant seasonal variation. Specically, the demand for bottled mineral water and carbonated water increases signi-
cantly during the summer. The sale of juices follows a similar prole; however, the demand rst peaks at the end of
spring/early summer, then declines and recovers temporarily in September. As the data on juices derive from a
company mainly active in the Athens area, the above-mentioned seasonal salesprole can be explained by the fact
that many permanent residents leave the capital during the summer vacation period (particularly in August). The
sales of soft drinks follow a trend similar to that of juices, with an additional peak in December probably due to
the Christmas holiday. Figure 1 also shows that for all products examined, the seasonal and hence weather-
sensitive effect is superimposed on a long-term increase in sales. This is particularly noticeable for juices and soft
Figure 1. Monthly demand for bottled mineral water and other non-alcoholic beverages in Greece (200211)
Climate Change and Demand for Bottled Water and Non-Alcoholic Beverages
Copyright © 2013 John Wiley & Sons, Ltd and ERP Environment Bus. Strat. Env. (2013)
DOI: 10.1002/bse
drinks and can possibly be attributed either to a substantial expansion of these two sub-sectors during the reference
period or/and a dynamic development of the specic company, which provided the data on sales.
Necessary historical weather data for the period 200211 were derived from a rst-class meteorological station
(so classied according to the criteria set by the World Meteorological Organization) located in the center of Athens.
Since almost half of the total Greek population lives in the greater Athens area, and consequently consumes a
relevant portion of bottled mineral water and non-alcoholic beverages, the meteorological data utilized constitute
a representative basis for analyzing the behavior of consumers in Greece in response to weather conditions.
The experience of many companies worldwide indicates that several weather elements can inuence the demand
for specic products. Thus, in the context of this analysis we have tested the inuence of maximum, mean and
minimum temperature, precipitation, sunshine duration and relative humidity on the sales of bottled water and
non-alcoholic beverages. The monthly values of these parameters are the average (for temperature, relative humidity
and sunshine duration) or the sum (for precipitation) of the corresponding daily gures.
To investigate the impact of climate change for the period 202150, the daily meteorological data at the same
location were derived by interpolating the results provided by the regional climate model for the four nearer grid points
of the computational grid of the model (grid analysis 25 km25 km). Then, a typical meteorological year for the period
202150 was formulated on the basis of the approach presented in Estimation of Future Regional Climate and nally
the monthly values of the selected parameters were calculated from the corresponding daily gures.
The socio-economic parameters included in the developed regression models comprise private consumption of
households and the number of overnight tourist stays; the relevant data derived from the National Statistical Service
of Greece. The number of overnight tourist stays was available on a monthly basis for the entire period under exam-
ination. Data on the private consumption of households were available on a quarterly basis. To disaggregate them on
a monthly basis, rst it was assumed that consumption is equally spread within each 3-month period. On the
assumption that the monthly values obtained stand for the private consumption at the second month of each quarter,
the consumption of the third month of the quarter nand the rst month of the quarter n+ 1 derived from a linear
interpolation of the consumption estimated for the second month of the quarters nand n+1.
Results
Sensitivity of Product Sales to Weather Conditions
Using n= 114 monthly observations (covering the period from January 2002 up to June 2011) several multiple re-
gression models were developed, correlating the volume of sales for each product category in question with a num-
ber of explanatory variables presented in The Demand Model. Table 1 shows, for each product category examined,
the models presenting the highest adjusted coefcient of determination (R
2
), the signicant independent variables
included and other indices related to the performance of the models. It is worth mentioning that the selected models
present a satisfactory to very high predictive power, with R
2
ranging from 70 to 90%, with somewhat lower values
(c. 55%) for small containers of less than 500 ml, meaning that the independent variables included in each model
explain a corresponding percentage of sales uctuations for the relevant product category. According to the results
of the analysis, in all models the private consumption of households is statistically very signicant. In some models,
additional economic parameters are statistically signicant, namely the overnight stays of tourists as well as the
linear time variation (see for example the equations modeling the sales of soft drinks and juices), identifying a long-
term development of the corresponding markets and/or the activity of the specic company which provided the relevant
sales data. In addition, some models selected incorporate as statistically signicant independent variables one or more
weather parameters, namely the mean monthly maximum temperature, the mean relative humidity or the mean
apparent temperature. Finally, for all the products examined, the corresponding models include a number of dummy
variables, identifying a seasonal or monthly variation of product sales not attributed to weather variability.
Despite the fact that the multiple regression models developed in the context of this study present a satisfactory to
high R
2
, many of them still show large serial correlation, violating one of the basic assumptions of the regression
analysis. The DurbinWatson statistic is widely used for testing the serial correlation or the independence of
S. Mirasgedis et al.
Copyright © 2013 John Wiley & Sons, Ltd and ERP Environment Bus. Strat. Env. (2013)
DOI: 10.1002/bse
residuals at lag 1 in regression models; it ranges in value from 0 to 4, with an intermediate value of 2, which
indicates the absence of autocorrelation at lag 1. As clearly depicted in Table 1, for several models developed in
the context of this study the DurbinWatson statistic deviates signicantly from this central value, which casts doubt
on the validity of the F- and t-tests and condence intervals (Makridakis et al., 1998). One of the common methods
for reducing the serial correlation observed in econometric regression is the incorporation of an autoregressive
structure in the error term (Makridakis et al., 1998).
In the context of this paper, autoregressive structures were incorporated in all models presenting a DurbinWatson
index higher than 2.3 or lower than 1.7 (facts that show a signicant serial correlation problem in the corresponding
models). All the other independent parameters remained unchanged. Tables 25presentthenal models selected
for each product category and the corresponding estimated coefcients. Notably, the incorporation of autoregressive
structures in the models resulted in increased R
2
values, indicating an improvement of their predictive power. The
inclusion of autoregressive structures in the models examined does not affect signicantly the importance of weather
parameters. In all cases of products examined, the maximum ambient temperature is a statistically signicant inde-
pendent variable, affecting positively the volume of product sales. The humidity, either directly or in combined form
(T
app
) has been included only in a limited number of models and seems to be less signicant. As regards the economic
parameters, the inclusion of autoregressive structures seems to weaken the signicance of the private consumption of
households in some models. This is reasonable as the autoregressive structures reect in a way the economic activity.
The additional presence of tourists in the models does not seem to inuence sales either of water (except for carbon-
ated water in large containers) or of the other non-alcoholic beverages. Possible explanations are the relatively small
percentage of their numbers in comparison to the total population in the Athens area and the similarity of their choice
behavior to the local residents.
As the maximum temperature was found to be the meteorological parameter with the most signicant inuence
on the level of sales of non-alcoholic beverages, Figure 2 shows the changes in total sales for all products examined
attributable to a uniform increase of the mean monthly maximum temperature by 1 C for all months of the year.
For the case of soft drinks in large containers where T
app
is included instead of T
max
, the value of T
app
to be used in
the demand equations resulted from Eq. (2) with a 1 C increase in T
max
and the relative humidity constant (RH).
Model Signicant Independent Variables Adjusted R
2
Signicance FDurbinWatson
Bottled mineral water
Total volume of sales T
max
, PC, S
2
,S
3
0.724 8.1 10
29
2.397
Sales in packs <500 ml T
max
, PC, S
2
,S
3
0.538 2.1 10
17
2.227
Sales in packs >500 ml T
max
, PC, S
2
,S
3
0.717 3.2 10
28
2.475
Carbonated water
Total volume of sales T
max
, PC, S
2
,S
3
0.733 1.5 10
29
2.206
Sales in packs <500 ml T
max
, PC, S
2
,S
3
0.722 1.2 10
28
2.075
Sales in packs >500 ml T
max
, PC, TRSM, S
4
0.759 7.9 10
32
2.208
Soft drinks
Total volume of sales T
app
, PC, t,M
Aug
,M
Dec
0.706 3.4 10
28
0.755
Sales in packs <500 ml T
max
, RH, PC, t,M
Aug
,M
Dec
0.587 8.5 10
20
0.545
Sales in packs >500 ml T
app
, PC, t,M
Aug
,M
Dec
0.730 3.4 10
30
0.889
Juices
Total volume of sales T
max
, RH, PC, t,M
Aug
0.901 1.3 10
53
1.390
Sales in packs <500 ml T
max
, PC, t,M
Aug
0.847 2.1 10
44
1.120
Sales in packs >500 ml T
max
, RH, PC, t,M
Aug
0.894 5.3 10
52
1.562
Table 1. Independent variables and overall performance of multiple regression models selected for analyzing the sales of bottled
mineral water and non-alcoholic beverages. All models are structured on a monthly basis, incorporating the following independent
variables: T
max
, mean maximum temperature; RH, mean relative humidity; T
app
, mean apparent temperature; PC, private
consumption of households; TRSM, number of tourist overnight stays; S
2
,S
3
and S
4
, dummy variables indicating the season of the
year (i.e. 2 for spring, 3 for summer and 4 for autumn); M
Aug
and M
Dec
, dummy variables indicating correspondingly the months
August and December; t, time in months showing the long-term trends in sales
Climate Change and Demand for Bottled Water and Non-Alcoholic Beverages
Copyright © 2013 John Wiley & Sons, Ltd and ERP Environment Bus. Strat. Env. (2013)
DOI: 10.1002/bse
As a general remark, the results obtained indicate that as the temperature increases, the demand for bottled mineral
water and other non-alcoholic drinks tends to be higher; on an annual basis this increase ranges from approximately
0.9% for bottled mineral water and soft drinks up to 2.2 and 2.3% for carbonated water and juices, respectively. Yet,
the estimated changes attributable to a +1 C increase in the mean monthly maximum temperature are seasonally not
uniform. Specically, on a percentage basis, the projected increases are more signicant in the winter period (i.e.
December to February), which is above 1.2% for bottled mineral water and soft drinks, reaching up to 3.7% for carbon-
ated water and 4% for juices. On the other hand, the level of sales during the summer period (i.e. May to September)
seems to be less affected by an increase in mean monthly maximum temperature by 1 C and ranges between 0.5 and
0.7% for bottled mineral water, 1.3 and 1.8% for carbonated water, 0.4 and 0.7% for soft drinks and 0.8 and 3.1% for juices.
Variable
Total volume of sales Sales in packs <500 ml Sales in packs >500 ml
Coefcient Sig. Coefcient Sig. Coefcient Sig.
c2.409 0.016 4.171 0.000 6.364 0.000
T
max
0.427 0.000 0.462 0.000 0.432 0.002
RH
T
app
PC 1.825 0.000 2.235 0.000 0.671 0.007
TRSM 0.153 0.002
t
S
2
0.106 0.000 0.099 0.000
S
3
0.150 0.000 0.130 0.000
S
4
0.200 0.000
M
Aug
M
Dec
R
2
0.733 0.722 0.759
Table 3. Values of the independent coefcients of the nal models selected for analyzing the sales of carbonated water. See text
and Table 1 for denitions of variables; c, constant of the model
Variable
Total volume of sales Sales in packs <500 ml Sales in packs >500 ml
Coefcient Sig. Coefcient Sig. Coefcient Sig.
c1.659 0.028 0.089 0.949 1.770 0.017
T
max
0.163 0.011 0.206 0.061 0.144 0.023
RH
T
app
PC 1.227 0.000 1.442 0.000 1.177 0.000
TRSM
t
S
2
0.116 0.000 0.125 0.000 0.111 0.000
S
3
0.240 0.000 0.212 0.000 0.251 0.000
S
4
M
Aug
M
Dec
AR (1) 0.213 0.033 0.251 0.011
R
2
0.746 0.538 0.744
Table 2. Values of the independent coefcients of the nal models selected for analyzing the sales of bottled mineral water. See text
and Table 1 for denitions of variables; c, constant of the model; AR(1), autoregressive structure at lag 1
S. Mirasgedis et al.
Copyright © 2013 John Wiley & Sons, Ltd and ERP Environment Bus. Strat. Env. (2013)
DOI: 10.1002/bse
A possible explanation for this differentiation is that the consumption of bottled mineral water and non-alcoholic bev-
erages during the summer period is alreadyhigh and therefore there is not much room for a substantial further growth
in sales as a result of higher temperatures. Also, at extremely high temperatures people limit their outdoor activities,
remaining at home.
Variable
Total volume of sales Sales in packs <500 ml Sales in packs >500 ml
Coefcient Sig. Coefcient Sig. Coefcient Sig.
c9.763 0.000 8.208 0.000 10.399 0.000
T
max
0.243 0.000
RH 0.077 0.171
T
app
0.164 0.000 0.136 0.000
PC 0.214 0.472 0.021 0.924 0.391 0.204
TRSM
t0.001 0.000 0.001 0.203 0.001 0.000
S
2
S
3
S
4
M
Aug
0.059 0.000 0.052 0.000 0.061 0.000
M
Dec
0.199 0.000 0.103 0.000 0.218 0.000
AR (1) 0.733 0.000 0.638 0.000 0.664 0.000
AR (2) 0.186 0.095
AR (3) 0.474 0.000
R
2
0.846 0.881 0.838
Table 4. Values of the independent coefcients of the nal models selected for analyzing the sales of soft drinks. See text and Table 1
for denitions of variables; c, constant of the model; AR(1) - AR(3), autoregressive structures at lag 13
Variable
Total volume of sales Sales in packs <500 ml Sales in packs >500 ml
Coefcient Sig. Coefcient Sig. Coefcient Sig.
c9.325 0.000 9.186 0.000 9.562 0.000
T
max
0.401 0.000 0.254 0.000 0.420 0.000
RH 0.108 0.079 0.179 0.005
T
app
PC 0.227 0.199 0.385 0.166 0.281 0.132
TRSM
t0.002 0.000 0.003 0.000 0.002 0.000
S
2
S
3
S
4
M
Aug
0.096 0.000 0.137 0.000 0.087 0.000
M
Dec
AR (1) 0.219 0.035 0.382 0.001 0.140 0.150
AR (2) 0.057 0.566 0.157 0.136 0.101 0.274
AR (3) 0.391 0.000 0.339 0.001
R
2
0.923 0.886 0.916
Table 5. Values of the independent coefcients of the nal models selected for analyzing the sales of juices. See text and Table 1 for
denitions of variables; c, constant of the model; AR(1) - AR(3), autoregressive structures at lag 13
Climate Change and Demand for Bottled Water and Non-Alcoholic Beverages
Copyright © 2013 John Wiley & Sons, Ltd and ERP Environment Bus. Strat. Env. (2013)
DOI: 10.1002/bse
Climate Projections for Greece
As already mentioned, changes in climatic conditions in Greece during the upcoming decades (i.e. 202150) com-
pared with historic climate (196190) are examined in the context of the A1B Scenario of the IPCC. The data set
developed for the historic and future climate comprises daily values for the selected meteorological parameters
for a 30-year period (196190 and 202150). TMY are dened on the basis of the above-mentioned data sets
according to the Sandia National Laboratories method.
In view of the results of the previous section, after the FS statistics were computed for all 30 years of the historical
(196190) and future (202150) periods as described in Step 2 of the TMY methodology above, the following
weighting factors (Table 6) have been utilized to compute the weighted sum of FS statistics (Step 3 of the TMY
methodology). The weighting factors are in line with the sales sensitivity for non-alcoholic beverages to the meteo-
rological parameters identied in the regression analysis. The regression analysis shows that maximum tempera-
ture and in some cases relative humidity are the most relevant meteorological parameters. Therefore, weights
assigned to temperature-related parameters were set to 50%, for relative humidity to 30% while the remaining
20% accounts for the other parameters (sunshine duration, wind speed and precipitation).
As mentioned, the nal selection of typical months is based on the simpler process introduced by Pissimanis
et al. (1988). Since maximum temperature represents the key determining meteorological parameter, the RMSD
of maximum temperature is calculated for the ve candidate years (Table 7). Years nally selected per month are
presented in bold.
Figure 2. Estimated change of sales of bottled mineral water and non-alcoholic beverages as a result of a +1 C change in the
average monthly maximum temperature
Maximum
temperature
Minimum
temperature
Mean
temperature
Precipitation Relative
humidity
Wind
speed
Sunshine
duration
Weight 30% 10% 10% 10% 30% 5% 5%
Table 6. Weighting factors used for the development of the typical meteorological years for historic and future climate
S. Mirasgedis et al.
Copyright © 2013 John Wiley & Sons, Ltd and ERP Environment Bus. Strat. Env. (2013)
DOI: 10.1002/bse
Figure 3 presents the monthly values of maximum temperature and relative humidity for the historic and future
climate calculated from the corresponding daily values of the TMYs developed. On average, mean annual maximum
temperature is expected to increase under the future climate by 1.4 C compared with the historic climate. On a
monthly basis, temperature increases are more pronounced during autumn periods and particularly in October,
when the maximum temperature is projected to increase by 3.0 C. During winter, spring and summer, maximum
monthly temperature increases, on average, by 0.8, 1.5 and 1.5 C, respectively. Changes of minimum and mean
temperature between historic and future climate follow a similar prole, while minor changes are observed for
the remaining parameters (e.g. relative humidity shown in Figure 3b).
Impacts of Climate Change on Product Sales
The empirical relationships developed using historic data in order to estimate the effect of weather conditions on the
demand for bottled water and the other non-alcoholic beverages in question have been utilized to estimate the
potential implications of future climate change for the level of product sales.
Table 8 presents the projected future monthly and annual changes in sales of bottled mineral water and other
non-alcoholic beverages attributable to the anticipated climate change in Greece for the period 202150 compared
to sales under the historic climate (196190). For all products examined, future climate (as simulated under the A1B
global GHG emissions scenario) will result in an increase in annual sales of all products examined. Specically, the
total volume of sales is projected to increase by 1.1% for bottled mineral water, 2.9% for carbonated water, 1.2% for
soft drinks and 3.1% for juices. On a seasonal basis, the most signicant increases in demand are expected in
autumn for bottled mineral water, carbonated water and soft drinks, and in spring for juices. On the other hand,
the total volume of sales in summer seems to be less affected by climate change.
Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec
Historic climate
3.2371 2.5820 3.5492 3.4458 2.6304 2.4000 3.6287 2.9602 3.1059 4.0684 2.3279 3.2379
1984 1982 1977 1988 1984 1988 1977 1986 1977 1970 1965 1974
2.5232 3.0443 3.3555 2.9314 3.1291 3.1661 2.9129 3.6161 2.5628 3.9097 2.6612 2.5408
1964 1983 1976 1986 1977 1981 1967 1963 1969 1975 1969 1978
3.4405 2.8763 2.3954 3.4494 2.7303 3.2687 2.8282 2.8647 3.4863 3.8832 1.8644 2.2017
1972 1988 1969 1990 1966 1975 1974 1987 1967 1980 1987 1971
2.6263 2.9217 3.3276 3.3719 3.3759 2.8757 2.7767 2.7039 2.6516 3.0020 2.6280 2.3136
1961 1987 1963 1968 1979 1984 1985 1964 1975 1976 1961 1986
4.0935 1.8709 3.7459 2.9051 2.6128 1.8589 2.4831 3.5020 3.6046 2.8898 2.6772 4.5677
1978 1973 1982 1976 1986 1990 1975 1972 1963 1965 1968 1984
Future climate
2.7317 3.1920 3.0589 1.9417 1.7007 3.0831 3.2969 1.9610 2.7650 2.0328 2.6879 4.0741
2047 2048 2025 2045 2036 2030 2040 2036 2041 2046 2031 2048
2.7038 4.2281 3.5996 4.5810 3.0275 2.8154 2.5816 2.9921 2.0654 2.1960 2.2300 3.3410
2039 2021 2034 2022 2041 2024 2033 2041 2047 2047 2044 2038
3.2718 4.2537 3.0287 2.4963 4.0578 2.6121 2.5964 3.8993 2.7687 2.4307 1.9985 3.1612
2025 2047 2036 2021 2046 2045 2039 2034 2027 2036 2036 2025
2.2027 2.7438 3.0285 3.6412 1.8015 2.8977 2.3518 2.7192 3.4011 2.0899 3.7143 1.8509
2036 2040 2028 2042 2050 2032 2048 2039 2028 2030 2049 2024
2.3067 3.1585 3.7659 2.8572 3.3510 2.2411 3.4347 2.5085 2.4827 4.1670 2.7890 2.6672
2048 2033 2031 2037 2039 2042 2025 2033 2023 2045 2024 2029
Table 7. Root mean square difference values for maximum temperature in historic and future climate and years selected per month
Climate Change and Demand for Bottled Water and Non-Alcoholic Beverages
Copyright © 2013 John Wiley & Sons, Ltd and ERP Environment Bus. Strat. Env. (2013)
DOI: 10.1002/bse
Regarding the behavior of different product types, the results of the analysis show that the sales of bottled mineral
water and soft drinks available in packages of less than 500 ml are most sensitive to weather conditions; this leads to
a higher expected growth of sales under future climate than for larger packages of the same products. This trend can
be explained by the fact that future climate is very likely to favor more outdoor activities, while the expected higher
temperatures will increase the demand for bottled mineral water and soft drinks in individual packs. Consumption
of bottled mineral water and soft dinks in family packages is also positively related to temperature increases, but to a
lesser extent than individual packs. On the other hand, the increased sales of carbonated water under the future cli-
mate seem to be irrelevant to the type of packaging. Finally, sales of juices in family packs seem to be more sensitive
to weather conditions compared with individual packs. As juice consumption mainly takes place within the home,
future climatic conditions will result in increased consumption of juices mainly in the form of family packs.
Some Conclusions
In this study, the potential impacts of the upcoming climate change in 202150 on consumer goods sales, and
specically on the sales of bottled water and other non-alcoholic beverages, has been investigated and applied
to Greece as a case study. To this end, a methodological framework was developed comprising demand equations
derived from long-term existing data that incorporate economic, social and climate parameters, and a special pur-
pose TMY for future and historic climate constructed from the outcome of a high-spatial-resolution regional cli-
mate model taking into account the signicance of various meteorological parameters as per the demand
equations. The use of special purpose TMY and the estimation of monthly values from the corresponding daily
Figure 3. Monthly maximum temperature (a) and relative humidity (b) under historic and future climate on the basis of the typical
meteorological year developed
S. Mirasgedis et al.
Copyright © 2013 John Wiley & Sons, Ltd and ERP Environment Bus. Strat. Env. (2013)
DOI: 10.1002/bse
Bottled mineral water Carbonated water Soft drinks Juices
Total
volume
of sales
Sales in
packs
<500 ml
Sales in
packs
>500 ml
Total
volume
of sales
Sales in
packs
<500 ml
Sales in
packs
>500 ml
Total
volume
of sales
Sales in
packs
<500 ml
Sales in
packs
>500 ml
Total
volume
of sales
Sales in
packs
<500 ml
Sales in
packs
>500 ml
Jan 0.8 1.0 0.7 2.1 2.2 2.1 0.7 1.6 0.5 2.6 1.2 3.1
Feb 2.0 2.5 1.8 5.3 5.8 5.4 2.1 3.0 1.8 5.0 3.1 5.2
Mar 1.4 1.8 1.2 3.7 4.0 3.7 1.4 2.4 1.2 4.0 2.2 4.4
Apr 1.3 1.6 1.1 3.4 3.7 3.4 1.3 2.2 1.1 3.5 2.0 3.9
May 1.1 1.4 1.0 2.9 3.1 2.9 0.9 2.7 0.7 4.2 1.7 5.3
Jun 0.7 0.9 0.6 1.8 1.9 1.8 0.9 0.6 0.7 1.2 1.1 0.9
Jul 0.8 1.0 0.7 2.1 2.2 2.1 1.0 0.9 0.8 1.6 1.2 1.4
Aug 0.9 1.1 0.8 2.3 2.5 2.3 0.8 1.9 0.7 3.0 1.4 3.7
Sep 1.3 1.7 1.2 3.6 3.9 3.6 1.6 1.9 1.3 3.2 2.1 3.3
Oct 2.3 2.9 2.0 6.2 6.7 6.3 2.6 3.6 2.1 6.0 3.6 6.4
Nov 0.6 0.8 0.6 1.6 1.8 1.7 0.6 1.1 0.5 1.8 1.0 2.0
Dec 0.4 0.5 0.4 1.1 1.2 1.1 0.4 0.6 0.4 1.0 0.6 1.1
Total 1.1 1.4 1.0 2.9 3.1 2.9 1.2 1.9 1.0 3.1 1.8 3.3
Table 8. Estimated percentage increase of sales of bottled mineral water and non-alcoholic beverages attributable solely to climate change under A1B scenario for the
period 202150 compared with historic climate (i.e. 196190)
Climate Change and Demand for Bottled Water and Non-Alcoholic Beverages
Copyright © 2013 John Wiley & Sons, Ltd and ERP Environment Bus. Strat. Env. (2013)
DOI: 10.1002/bse
gures rather than using average monthly values for the entire period under consideration also enhanced the
accuracy of the results.
The results of the analysis clearly show that future sales for bottled water and non-alcoholic beverages in Greece for
the period 202150 could increase considerably under the future warmer climate associated with the A1B global GHG
emissions scenario of IPCC. Specically, on an annual basis, the increase in the total sales volume attributed solely to
climate was estimated at 1.1% for bottled mineral water, 2.9% for carbonated water, 1.2% for soft drinks and 3.1% for
juices. It isalso worthmentioning that for all products examined these changes are not seasonally uniform, and in some
months exceed 6%. The increase is more pronounced in autumn and spring rather than summer, possibly because dur-
ing summer the consumption of bottled water and non-alcoholic beverages is already high and therefore there is not
much room for substantial further growth in sales. The expected rate of sales change depends also on the type of bev-
erage packaging. Thus, the sales of bottled mineral water and soft drinks in smaller packaging (i.e. 330 ml cans) seem to
be more weather-sensitive compared with family packs, while the opposite was found for juices. From the results of the
analysis itis obvious that future climate change is expected toaffect positively the demand for bottled mineral water and
non-alcoholic beverages, creating new business opportunities. Compared with historic climate (196190), future
warming in Greece during 202150 under the A1B global GHG emissions scenario is expected to average over 2C
and could result in an increase due to solely climatic conditions of the present (2010) annual turnover by 3.0 million
for bottled mineral water, 0.4 million for carbonated water, 6.8 million for soft drinks and 8.5 million for juices.
A large number of retail sales as well as other commercial activities will be inuenced, mostly adversely, by climate
change. Others, such as the water and soft drinks industry, could be affected positively. It is important to continue exam-
ining and quantifying theeffect of climate changein all business activities, above and beyond the clearly important ones
of energy, if we want to provide better estimates of the impact in economic terms of climate change, to be juxtaposed
with the cost of mitigation. To this end, it is important for enterprises and business associations to develop:
capacity for understanding the diverse impacts of climate change on the economy at both micro- and macro-level.
The identication of the climatic parameters that inuence a specic activity, the nature of impacts, their severity,
the temporal and spatial scale at which they will occur, the way they inuence businesses (directly or indirectly),
the uncertainties associated with climate research, etc., are important aspects that businesses need to understand
before developing adaptation strategies as well as short-to-medium business plans.
analytical historic data collection and recording capacity at company level and appropriate tools that will allow
investigation of the potential impacts of climatic parameters on businesses in quantitative terms and at an appro-
priate level of analysis. In this context, the existence of short term or even daily data regarding the level of sales for
different products per size of package, region of sale, etc., together with local meteorological parameters as well as
other economic parameters (e.g. expenditures in advertising campaigns) could provide a sound basis for improv-
ing and taking advantage of models aiming at quantifying the inuence of various parameters on the demand for
the products in question. This could result in adjusting the levels of production for specic products, enhancing
the stocks, implementing appropriate pricing policies, production of new products, etc.
Acknowledgement
The authors would like to thank the two (anonymous) Greek companies which provided the sales data for bottled mineral water
and non-alcoholic beverages.
References
Agnew MD, Thornes JE. 1995. The weather sensitivity of the UK food retail and distribution industry. Meteorological Applications 2: 137147.
Argiriou A, Lykoudis S, Kontoyiannidis CA, Asimakopoulos D, Petrakis M, Kassomenos P. 1999. Comparison of methodologies for TMY generation
using 20 years data for Athens, Greece. Solar Energy 66(1): 3345.
Bahng Y, Kincade D. 2012. The relationship between temperature and sales: Sales data analysis of a retailer of branded womens business wear.
International Journal of Retail & Distribution Management 40(6): 410426.
S. Mirasgedis et al.
Copyright © 2013 John Wiley & Sons, Ltd and ERP Environment Bus. Strat. Env. (2013)
DOI: 10.1002/bse
Beverage Council of Ireland [BCI]. 2009. The glass is half full - The 2009 Republic of Ireland Soft Drinks Market Review. BCI, Dublin, 40 p.,
available at: www.beveragecounciloreland.ie
Blom J-E. 2009. Hedging revenues with weather derivatives - A literature review of weather derivatives & a case study of Ringnes AS. MSc. Thesis,
Norges Handelshøyskole, Bergen, June, 114 p.
British Soft Drinks Association. 2008. A changing climate - The 2008 UK Soft Drinks Report. BSDA, London, 31 p., available at: www.
britishsoftdrinks.com
British Soft Drinks Association. 2009. Trusted Innovation - The 2009 UK Soft Drinks Report. BSDA, London, 36 p., available at: www.
britishsoftdrinks.com
CCSP. 2008. Climate Models: An Assessment of Strengths and Limitations. A Report by the U.S. Climate Change Science Program and the
Subcommittee on Global Change Research [Bader D.C., C. Covey, W.J. Gutowski Jr., I.M. Held, K.E. Kunkel, R.L. Miller, R.T. Tokmakian
and M.H. Zhang (Authors)]. Department of Energy, Ofce of Biological and Environmental Research, Washington, D.C., USA
Dubé J-P. 2004. Multiple discreteness and product differentiation: demand for carbonated soft drinks. Marketing Science 23(1): 6681.
Hall I, Prairie R, Anderson H, Boes E. 1978. Generation of typical meteorological years for 26 SOLMET stations. SAND78-1601, Sandia National
Laboratories, Albuquerque, NM.
Hansen J, Sato M, Ruedy R. 2012. Perception of climate change, PNAS early edition, 29 March 2012.
IPCC. 2000. Special report on emissions scenarios. Intergovernmental Panel of Climate Change. Cambridge University Press.
Jiang Y. 2010. Generation of typical meteorological year for different climates of China. Energy 35: 19461953
Makridakis S, Wheelwright S, Hyndman R. 1998. Forecasting methods and applications, 3rd Edition. John Wiley & Sons Inc.: New York
Meijgaard E, Ulft B, Berg WJ, Bosveld FC, Hurk B, Lenderink G, Siebesma AP. 2008. The KNMI regional atmospheric climate model RACMO,
version 2.1, KNMI technical report, TR-302.
Murray KB, Finn A, Leszczyc P, Di Muro F. 2010. The Effect of weather on consumer spending. Journal of Retailing and Consumer Services 17:512520.
Pissimanis D, Karras G, Notaridu V, Garva K. 1988. The generation of a typical meteorological year for the city of Athens. Solar Energy 40:405411.
Ramanathan U, Muyldermans L. 2010. Identifying demand factors for promotional planning and forecasting: A case of a soft drink company in
the UK. International Journal of Production Economics 128: 538545.
Sawaqed NM, Zurigat YH, Al-Hinai, H. 2005. A step-by-step application of Sandia method in developing typical meteorological years for different
locations in Oman. International Journal of Energy Research 29: 723737.
Skeiker K. 2007. Comparison of methodologies for TMY generation using 10 years data for Damascus, Syria. Energy Conversion and Management
48: 20902102.
Steadman R. 1984. A universal scale of apparent temperature. Journal of Climate and Applied Meteorology 23: 16741687.
Weather Trends International [WTI]. 2009. Identifying and Quantifying How Weather Inuences Consumer Behavior. WTI, April, 13 p.,
available at: www.wxtrends.com/wp-content/uploads/2012/03/identify_and_quantify_weather_inuences_on_human_behavior.pdf
Wedawatta G, Ingirige B, Jones K, Proverbs D. 2011. Extreme weather events and construction SMEs: vulnerability, impacts, and responses.
Structural Survey 29(2): 106119.
Weinhofer G, Busch T. 2012. Corporate strategies for managing climate risks. Business Strategy and the Environment, Published online in Wiley
Online Library. DOI: 10.1002/bse.1744
Winn M, Kirchgeorg M, Grifths A, Linnenluecke M, Gunther E. 2011. Impacts from climate change on organizations: a conceptual foundation.
Business Strategy and the Environment 20: 157173. DOI: 10.1002/bse.679
Xu J, Sun N. 2011. The impact of meteorological factors on manufacturing economy: a quantitative analysis based on trans-log production fuction.
Paper presented in International Conference on Remote Sensing, Environment and Transportation Engineering (RSETE), 2011.
Climate Change and Demand for Bottled Water and Non-Alcoholic Beverages
Copyright © 2013 John Wiley & Sons, Ltd and ERP Environment Bus. Strat. Env. (2013)
DOI: 10.1002/bse
... The author shows that GDP per capita is reduced by at least 0.6% due to the occurrence of climate-related disasters. Cavallo et al (2013) show that natural disasters affect the economic growth of countries in the short and long term. Nakamura et al (2013) estimate the economic cost of disasters in 24 countries over 100 years. ...
... Cavallo et al (2013) show that natural disasters affect the economic growth of countries in the short and long term. Nakamura et al (2013) estimate the economic cost of disasters in 24 countries over 100 years. The authors find that disasters increase the volatility of consumption 94 growth, specifically, consumption falls by 30% in the short term before half of this decline is recovered in the long term. ...
... Bouchrika. et Issaoui (2013) state that in Tunisia, climate change has created an imbalance between water supply and demand. Figure 1 shows the monthly volume of bottled water consumption per capita in 2020 and 2021. ...
Article
This study aims to explore the link between weather and bottled water consumption in Tunisia using the Autoregressive Distributed Lag model (ARDL) between 1995 and 2020. Ours results show that the precipitation and labor rates in the three sectors have an impact in the short and long term. An increase of 1°C in temperature in the short term leads to an increase in consumption of more than 4 liters of bottled water. However, 1 % more rainfall means a decrease in long-term bottled water consumption of about a quarter of a liter. While in the short term the effect is mixed (both positive and negative). Temperature further increases bottled water consumption in rural areas and among climate-exposed professions.
... For instance, the studies show a negative correlation between sales of jacket sales and average temperature while there is a positive correlation between ice cream and average temperature. It had been highly used in the studies that discusses about FMCG and clothing products (Mirasgedis et al., 2014;Ramanathan and Muyldermans, 2011;Žliobaite et al., 2012). ...
... The study further discusses how events affect the sales. As an example, clothes with seasonal themes reports better sales during the tagged seasons than on the other seasons (Badorf and Hoberg, 2020;Mirasgedis et al., 2014). ...
Thesis
Full-text available
in collaboration with Robert Gordon University Aberdeen , UK 2021 ii Declaration I declare that the work presented in this dissertation is my own work and to best of my knowledge acknowledgement is made for all sources of information used in this dissertation. Further, this as a whole or as parts has not been submitted previously or concurrently for a degree or any other qualification at any University or institutions of Higher Learning. …………………………. ……………………. Signature of the student Date Full Name of Student: Student Registration No: The above student carried out his/her research project under my supervision. ………………………… ……………………. Signature of the supervisor Date Name of the Supervisor: iii Abstract Predicting sales has become a vital role in consumer durable retail industry with the increased competition. The importance of this has increased due to its direct and indirect affect to the company's profitability. Managing the right inventory would reduce inventory holding costs while overstocking brings repercussions such as increasing the inventory holding cost, occurring promotional costs to flush out the excess stock etc. In this study, several machine learning techniques were used to predict the sales using the factors affecting sales. The techniques namely; K-nearest neighbor (KNN), Artificial neural network (ANN), Decision tree, Random Forest, Linear Regression and Bayesian Applied Regression. This study was conducted on four main product categories namely; Air conditioners, Fans, Refrigerator and audios. The algorithms were applied on the preprocessed sales data captured from the ERP system of the company, promotional data, event data and weather data. Once applied the algorithms, the best model was selected comparing RMSE and MAPE. KNN was identified to be the best fitting model of the seven models used for all the product lines. The analysis were done using R programming language and then the predicted values are then visualized through Power BI dashboards to present to the management. These predictions visualized through dashboards allows the retailer to take correct decisions at the right time while monitoring the progress of the actions. Moreover, the finalized data model allows the retailer to gain the cutting edge over the competitors while increasing the profits. iv
... It is estimated that the soft drinks market in the USA will reach in 2028 the 320 billion dollars mark, from 253 billion in 2016 [9]. Similar trends are observed for the European market [10], especially for Greece, which is the focus of this study [11,12]. Today, all over the world, the majority of soft drinks are distributed and available in PET bottles because PET is rugged, clear, durable and odor-resistant, and also has low carbon dioxide permeability [14]. ...
Preprint
Full-text available
This paper adopts a systems firm-centered perspective on the environmental assessment of recyclable glass and PET soft drink containers. We employ LCA and discrete-event simulation modelling for the environmental assessment of the two soft drinks packaging alternatives in operational terms, over the entire supply chain, over a period of three years. The assessment is based on real data collected from a large soft-drinks producer and its suppliers. The research and practice contribution of the paper is twofold: first, it introduces a methodological framework for environmental assessment of companies’ packaging environmental impact under different product and operations strategies, and secondly, it provides a holistic environmental assessment of the two packaging materials (PET and glass) taking into account specific operational issues, such as product mix and recycling and reuse options, and activity interdependences and stochasticity. The results of the simulation experiments confirm at system level, for glass, the importance for sustainability, to increase the number of reuse cycles and the percentage of used bottles collected for refilling, whereas for PET, to increase the percentage of recycled PET in new bottles.
Article
This study examines the influence of eight key weather factors, including atmospheric pressure (ATPRS), clear sky (CLSKY), precipitation (PRECP), relative and specific humidity (RHUMD and SHUMD, respectively), as well as temperature (TEMPR), wind direction and speed (WNDR and WNSP, respectively), on the returns and volatilities of the seven Moroccan sectors most sensitive to the weather (Agri-food/Production, Beverages, Distributors, Electricity, Leisure and Hotels, Oil and Gas, and Transport) from January 4, 1993, to October 30, 2022. The objective was to analyze how weather affects the fluctuations of sectoral stock indices and assess sectoral vulnerability. Utilizing daily frequency weather and stock market data and implementing a GJR-GARCH (1,1) model combined with an autoregressive (AR) process, the results demonstrate that all sectors are influenced by weather, particularly the Agri-food/Production and Beverages sectors. These insights are vital for the stakeholders in Morocco. Investors can make informed decisions, policymakers can create targeted support policies, and businesses can develop risk management strategies to handle weather-related challenges, ensuring economic stability and competitiveness. JEL Classification G10, G41
Article
Full-text available
This paper adopts a systems firm-centered perspective on the environmental assessment of recyclable glass and PET soft drink containers. We employ LCA and discrete-event simulation modeling for the environmental assessment of the two soft drinks packaging alternatives in operational terms over the entire supply chain over a period of three years. The assessment is based on real data collected from a large soft drink producer and its suppliers. The research and practice contribution of the paper is twofold: first, it introduces a methodological framework for environmental assessment of companies’ packaging environmental impact under different product and operations strategies; and secondly, it provides a holistic environmental assessment of the two packaging materials (PET and glass) taking into account specific operational issues, such as product mix and recycling and reuse options, as well as activity interdependences and stochasticity. The results of the simulation experiments confirm at the operations system level, for glass, the importance for sustainability, to increase the number of reuse cycles (for the particular case, for significant improvement, seven reuses) and the percentage of used bottles collected for refilling (80% recovery rate), whereas for PET, to increase the percentage of recycled PET in new bottles (towards 30%).
Article
Full-text available
Scientists extensively use mathematical models of Earth’s climate, executed on the most powerful computers available, to examine hypotheses about past and present-day climates. Development of climate models is fully consistent with approaches being taken in many other fields of science dealing with very complex systems. These climate simulations provide a framework within which enhanced understanding of climate-relevant processes, along with improved observations, are merged into coherent projections of future climate change. This report describes the models and their ability to simulate current climate. The science of climate modeling has matured through finer spatial resolution, the inclusion of a greater number of physical processes, and comparison to a rapidly expanding array of observations. These models have important strengths and limitations. They successfully simulate a growing set of processes and phenomena; this set intersects with, but does not fully cover, the set of processes and phenomena of central importance for attribution of past climate changes and the projection of future changes. Following is a concise summary of the information in this report, organized around questions from the “Prospectus,” which motivated its preparation, and focusing on these strengths and weaknesses.
Article
Full-text available
Purpose The influence of weather on business activities and human behaviour has been explored in several fields (e.g. finance and psychology), but little research about weather and retail sales is found in the retail or fashion literature. The purpose of the study is to analyse the relationship between temperature, one aspect of weather, and retail sales of seasonal garments. Design/methodology/approach The researchers collected sales data from a retailer of branded women's business wear in the Seoul‐Kyunggi area in South Korea. Along with the sales data for seasonal basic styles, corresponding daily and weekly average temperature data were collected and evaluated. The analysis for the study was drawn using descriptive statistics including graphical evaluations, correlation analysis and paired samples t ‐test. Interviews with the retailer's merchandisers were used to supplement interpretation of the statistical data. Findings Results of this study provide strong evidence that fluctuations in temperature can impact sales of seasonal garments. During sales periods when drastic temperature changes occurred, more seasonal garments were sold. However, the temperature changes from day to day or week to week did not affect the number of garments sold for the whole season. Of the seasonal garments expected to sell within the same season, the selling periods of each product category differed depending on type of fabric and design. For some seasonal garments, the actual sales dates were one week to two weeks in variance from the merchandisers' forecasts. Research limitations/implications Limitations in the sample (i.e. product category) and location of stores (i.e. geographic region) prevent the generalization of results to all seasonal garments or retailers. In spite of these limitations, this study can be a pilot study that supports the significant relationship between temperature and sales of seasonal basic products by quantifying the temperature effects on sales of particular products. Therefore, future studies are needed to establish generalized conclusions with a larger sample. Originality/value As little academic research is available about weather's effect on sales of garments, the present study contributes to the field of clothing and retail distribution by providing evidence of significant relationships between temperature and sales of seasonal clothing.
Article
Full-text available
Purpose The UK experienced a number of Extreme Weather Events (EWEs) during recent years and a significant number of businesses were affected as a result. With the intensity and frequency of weather extremes predicted in the future, enhancing the resilience of businesses, especially of Small and Medium‐sized Enterprises (SMEs), who are considered as highly vulnerable, has become a necessity. However, little research has been undertaken on how construction SMEs respond to the risk of EWEs. In seeking to help address this dearth of research, this investigation sought to identify how construction SMEs were being affected by EWEs and the coping strategies being used. Design/methodology/approach A mixed methods research design was adopted to elicit information from construction SMEs, involving a questionnaire survey and case study approach. Findings Results indicate a lack of coping strategies among the construction SMEs studied. Where the coping strategies have been implemented, these were found to be extensions of their existing risk management strategies rather than radical measures specifically addressing EWEs. Research limitations/implications The exploratory survey focused on the Greater London area and was limited to a relatively small sample size. This limitation is overcome by conducting detailed case studies utilising two SMEs whose projects were located in EWE prone localities. The mixed method research design adopted benefits the research by presenting more robust findings. Practical implications A better way of integrating the potential of EWEs into the initial project planning stage is required by the SMEs. This could possibly be achieved through a better risk assessment model supported by better EWE prediction data. Originality/value The paper provides an original contribution towards the overarching agenda of resilience of SMEs and policy making in the area of EWE risk management. It informs both policy makers and practitioners on issues of planning and preparedness against EWEs.
Article
Because of the need for a common meteorological data base for use in solar energy systems studies, a group has undertaken the task of developing a method for generating a typical meteorological year (TMY). The developed method has then been used to generate a TMY for each of the 26 SOLMET Rehabilitation Stations that have hourly data. The meteorological measures used in the present study to select the TMY were: dry bulb temperature, dew point temperature, wind velocity, and solar radiations on a horizontal surface--standard year corrected. Most SOLMET Stations have data available over a 23-year period beginning in 1953 and extending through 1975. The process used to select a TMY for a given station involves selecting, by statistical methods, one typical meteorological month (TMM) for each of the twelve calendar months from the 23 year period and catenating the twelve months to form a TMY. Thus, the TMY for each station consists of 12 months of actual meteorological data selected from the long term data base from that station, 23 years in most cases. A TMY at a given station could, for example, consist of January 1955, February 1966, March 1962, ... ... December 1973.
Article
This paper focuses on organizations and their management of climate risks. Climate risks stem from continued changes in climate means and the increase in frequency and intensity of extreme weather events. We ask whether companies also apply the usual process of corporate risk management to climate risks. In seeking to answer this question, we review several literature streams in order to set out an initial theoretical reflection. Based on this we conducted an exploratory case study with 11 electric utilities. Our results illustrate that these companies perceive climatic changes as a material issue for their business. However, management has restricted knowledge about such climatic changes and thus cannot precisely determine the potential negative impacts on business activities. As a consequence, the companies have implemented a climate risk management that does not differ from the usual process of managing other business risks. Our results further illustrate that there is some variation in how individual firms manage climate risks: While risk identification and risk assessment are equally important for all electric utilities, there are differences in how management determines the direction of the individual response to climate risks. Copyright © 2012 John Wiley & Sons, Ltd and ERP Environment.
Article
In recent years, the evaluation of the efficiency of the performance of solar energy units is not done by using long-term averages of weather data as input but preferably by using data sets representative of the climatological features of the site that are generated for this purpose. Such data sets, which are usually called “Test Reference Year” or “Short Reference Year,” consist mainly of solar radiation data, but they may also include other meteorological data, like temperature wind velocity etc, which may affect the response of the units. In this article, an attempt is made for the generation of such a representative data set for the city of Athens mainly by following a method that has been proposed by Hall et al. This data set, which includes global solar radiation data and six other meteorological parameters referring to temperature, dew point, and wind velocity, has been characterized by Hall as a “Typical Meteorological Year.”
Article
Based on the total thermal resistance required by a human model to effect equilibrium, a scale is prepared showing apparent temperature for any combination of dry-bulb temperature, vapor pressure, wind speed and extra radiation likely to be encountered meteorologically. Application to normal midday climates of the United States shows that dry-bulb temperature is modified by the three other variables by from 5 to +7 K. However, multiple linear regression indicates that dry-bulb temperature correlates most strongly with apparent temperature, and provides simple computing formulas.
Article
The impact of temperature and precipitation to Nanjing manufacturing economy was analyzed by using TRANSLOG model. The results of model estimated by randge regression method show that the average boundary output elasticity of temperature and precipitation factor to the manufacturing undulation output value are -0.1813 and -0.0683 respectively, which means both of temperature and precipitation have negative impact on Nanjing manufacturing. Over the long term, negative impact of temperature presents a weakening trend, while the one of precipitation presents a further increasing trend. In general, the proportion of temperature influence on Nanjing manufacturing total output is 1.7456%, while the proportion of precipitation is 0.6493%.
Article
The generation of a typical meteorological year (TMY) is of great importance for calculations concerning many applications in the field of thermal engineering. The need of an accurate TMY for simulations has been well recognized over the years. Various methods for deriving TMYs have been developed, but their final results can be significantly different. In this paper, the major methodologies reported in the literature were applied to 10 year hourly measurements of weather data from Damascus, Syria. The TMYs obtained were evaluated according to their impact on the typical Syrian building’s thermal system in order to decide which method should be recommended for generating typical meteorological years and for predicting the performance of thermal systems in buildings. Based on simulation results for seasonally, monthly and daily building thermal loads, three widely used statistical estimators, namely, root mean square difference RMSD, total standard error SEE and chi square χ2 were calculated to assess the performance of each TMY. The findings showed that the TMY giving the closest performance to the average performance of the building’s thermal system as predicted using the 10 year weather data is the one generated by using the modified Sandia method. This method gives sufficiently accurate results compared with the other methods reported in the literature.