UNIVERSITAT
ROVIRA I VIRGILI
DEPARTAMENT D’ECONOMIA

WORKING PAPERS
Col·lecció “DOCUMENTS DE TREBALL DEL
DEPARTAMENT D’ECONOMIA - CREIP”

Seasonality of hotel demand in the main Spanish provinces:
measurements and decomposition exercises

Juan Antonio Duro

Document de treball n.21- 2015

DEPARTAMENT D’ECONOMIA – CREIP
Facultat d’Economia i Empresa

UNIVERSITAT
ROVIRA I VIRGILI
DEPARTAMENT D’ECONOMIA

Edita:
Departament d’Economia
www.fcee.urv.es/departaments/economia/publi
c_html/index.html
Universitat Rovira i Virgili
Facultat d’Economia i Empresa
Avgda. de la Universitat, 1
43204 Reus
Tel.: +34 977 759 811
Fax: +34 977 300 661
Email: sde@urv.cat

CREIP
www.urv.cat/creip
Universitat Rovira i Virgili
Departament d’Economia
Avgda. de la Universitat, 1
43204 Reus
Tel.: +34 977 558 936
Email: creip@urv.cat

Adreçar comentaris al Departament d’Economia / CREIP

Dipòsit Legal: T - 1032 - 2015
ISSN edició en paper: 1576 - 3382
ISSN edició electrònica: 1988 - 0820

DEPARTAMENT D’ECONOMIA – CREIP
Facultat d’Economia i Empresa

Seasonality of hotel demand in the main Spanish provinces:
measurements and decomposition exercises*

Juan Antonio Duro
Department of Economics and CREIP, Universitat Rovira i Virgili, Av. de la
Universitat, 1, 43204 Reus, Spain. Tel: 0034 977759816. Fax: 0034
977300661. Email: juanantonio.duro@urv.net.

Abstract
This paper analyses the seasonal concentration of tourist activity in the main
Spanish provinces for the period 1999-2012, taking hotel nights as the indicator
of reference. We propose using several standard summary measures in order
to evaluate the level, evolution and some decompositions. Our main results can
be summarized as follows: first, across the whole country and especially since
2007, there is a growth in seasonality; second, seasonal concentration is
greatest in the Balearic Islands and two of the Catalan provinces, and least in
Madrid and the Canary Island provinces; third, although the overall patterns
typically agree, nevertheless, in some provinces the indexes we deal with show
some discrepancies; fourth, the decomposition of the monthly concentration by
major markets typically indicates the main role played by the foreign
component; finally, the overall evidence does not support the thesis that the
domestic market offsets the foreign one.

Keywords: seasonality; concentration indices; decomposition analysis;
Spanish provinces

*

The author acknowledges support from project ECO2013-45380-P.

1. Introduction
Baron (1975) highlights seasonality, or rather the seasonal concentration of
tourist activity, as one of the main problems facing mature tourist destinations.
Consequently, it is predictable that most of their medium- and long-term
development and consolidation strategies would include a significant reduction
in seasonality as one of their main policy goals, or, to put it another way, a
greater balance in the distribution of tourist flows throughout the year. The
reasons for this concern are well known (Baum, 1999). Firstly, there are
concerns about its impact on social and environmental sustainability (Manning
and Powers, 1984). The environmental effects include the effects on wildlife,
erosion, noise and environmental pollution, both directly and indirectly through
the consumption of inputs that have significant impacts. The social effects have
to do with the effects of this concentration on the welfare of the residents –
these arise from traffic problems, security, parking, queues, etc. In fact, these
impacts affect the welfare, not only of the residents, but also of the tourists
themselves Secondly, there is concern about the repercussions in terms of
economic inefficiency and the deficient use of both public and private resources
(Roselló et al., 2004). Indeed, seasonality gives rise to resource-use saturation
at some times of the year, with consequent effects on satisfaction levels and
service quality while, at other times, resources are under-used, with consequent
problems associated with revenues and private profitability. Finally, there are
worrying implications in terms of the labour market, salaries and disincentives
for investment in human capital (Krakover, 2000). So, seasonality creates
difficulties in recruiting and retaining workforce, can disincentive investment in
human capital and, finally, will result in reduced labour qualification and/or over
qualification.

For all these reasons it is logical that the academic literature has devoted
attention to studying seasonality from various different approaches. In this
respect, Koenig-Lewis and Bischoff (2005), in their well-known survey, identified

2

the six main topics of interest in relation to the analysis of seasonality, with
significant shortfalls in each of them. These were: the definitions of seasonality;
its causes and its impacts; the policy implications; studies into consumer
behaviour and approaches to measuring seasonality. This paper focuses
fundamentally on the last of these aspects.
To deal with the question of measuring this phenomenon consistently, it first
needs to be clarified what exactly is understood by seasonality. In this respect,
the literature uses slightly different connotations, although in most cases they
refer to seasonality as the emergence in a given destination of a systematic
pattern of tourist flows during the year (Baron, 1975; Allock, 1994; Butler 2004,
among others). On this regard, the definition that it seems most acceptable in
general terms is the one established by Butler (1994), who describes it as “a
temporal imbalance in the phenomenon of tourism which can be expressed in
terms of different indicators”. Based on this last definition, therefore, seasonality
would essentially be a distributional imbalance, which can be measured
synthetically.
Taking the above definition as a reference, it would therefore be necessary to
take synthetic distribution indices and apply them to the annual distribution of
tourist activity. Surprisingly, few papers in the literature focus on this type of
measurement of seasonality (Wanhill, 1980; Lundtorp, 2001; FernándezMorales, 2003; Fernández-Morales and Mayorga-Toledano, 2008 or Martín et
al., 2014). In particular, the literature on the measurement of inequality
specifically indicates a set of satisfactory synthetic measures like Gini
coefficient (widely used in seasonality applications until now), Theil indexes or
the Coefficient of Variation (CV hereafter) (Cowell, 1995). The problem,
however, is that there is no single measurement preferable than the rest. This is
mainly because each of them has a different weighting on the changes in the
different months (i.e. basic analysis units) (Duro, 2012). In this case, the
researcher either needs to explain his or her evaluation in this respect or
otherwise deal with a broad set of indicators to obtain a comprehensive
overview of the situation. So far, the literature on measurement of seasonality
does not seem to have devoted much attention to this.
3

In addition to the global measurement, an important analytical aspect to be
taken into account is the decomposition possibilities of the indexes. In this
respect, the literature on inequality measurement (Cowell, 1995) essentially
emphasizes the usefulness of two possibilities: firstly, the indices, or some of
them, may be decomposed by groups, identifying an inter-group component
and other intra-group ones (Shorrocks, 1984). In the seasonality case, the
groups would be formed by consecutive months or, to put in another way,
tourism seasons. This analysis, for example, is interested in terms of the
reliability of monthly aggregates as an instrument for explaining global
concentrations and as a tool for public planning; secondly, the literature has
highlighted the interest in using decompositions when the factors can be
expressed additively (also a common example in indicators of tourist demand)
(Shorrocks, 1980 and 1982). In this sense, it would be necessary to address the
role of each factor (source markets for instance) in measuring annual
seasonality. Regarding inequality decomposition analyses in the international
context we are only aware of the studies by Fernández-Morales (2003) and
Fernández-Morales and Mayorga-Toledano (2008). The former made an
analysis of the monthly concentration of hotel nights in three southern Spanish
provinces (Almeria, Granada and Malaga), using the Gini coefficient and its
decomposition by groups (i.e. seasons) as a sole indicator. In the latter study,
the authors decomposed by source (e.g. markets) the concentration of tourist
activity, this being measured analogically using again the Gini coefficient, for the
provinces in the south of Spain.
Consequently, this paper aims firstly to highlight the importance of
methodological elements of interest in relation to the measurement and
decomposition of seasonality, one of the main research lines suggested by
Koenig-Lewis and Bischoff (2005) but little investigated in the literature. In
particular, in the first case, and given the differential characteristics of the
measures, this paper proposes the joint use of the Gini coefficient (the index
typically used in the seasonality applications), the Theil Index and CV in order to
provide a sufficiently broad qualitative range and avoid over-extending the
number of measures. In this way, it hopes to overcome the academic literature’s
obsession with the Gini coefficient. Meanwhile, the paper also reviews the main
4

methodological elements relating to the implementation of decomposition
exercises, highlighting, for example, the advantages of the Theil Index over the
Gini coefficient in some cases. This being the case, it examines the possibilities
associated with the use of decomposition by groups (by monthly periods) and
by sources (markets). Indeed, both these methodological elements stem from
extending the instruments currently available in the field of inequality
measurements, which, up to now, have received little attention in the academic
literature on tourism. Secondly, the paper makes an analysis of tourist
seasonality based on the above mentioned instrumental methodology for the
main Spanish provinces, which make up 83% of the total demand in Spain, thus
providing quite comprehensive territorial coverage. The analysis is conducted
using hotel nights as an indicator, based on the figures extracted from the Hotel
Occupancy Survey conducted monthly by the Spanish National Institute and for
the longest period available, i.e. 1999-2012. The results given in the text
concentrate on the years at each end of the period for reasons of space.
Therefore, we believe that this analysis and the work is interesting or makes
contributions of interest for the following specific reasons: first, it makes a
methodological proposal based on the characteristics of the different indices
and in particular, suggests their combined use for measuring seasonality in a
comprehensive and more robust way: second, the measurement for Spain
allows the analysis of this tourist dimension in some areas which typically
exhibit high seasonality and thus permits verification of the overall effectiveness
of policies; thirdly, the decomposition analysis used allows the investigation of
explanatory factors as a step towards guiding possible control strategies. Here,
the inequality decomposition by groups allows exploration of whether the
different monthly grouping proposals can be used as a reference for policy.
Furthermore, the additive decomposition of markets points to some strategic
directions for reducing seasonality.
The paper is thus structured as follows: the second section reviews the main
methodological aspects relating to the measurement of seasonal concentration
of tourist activity, its factorial decomposition and the main data used for the
implementation in the Spanish provinces cases. The third section reproduces
5

the main results obtained. The final section contains the main considerations
deriving from this work.

2. Methodology and data
Taking Butler’s (1994) definition as a reference, seasonality should be
measured by different intra-annual temporal imbalance – or inequality –
indicators. In this respect, an immediate approach would be to extract some of
the learning in the literature on measuring inequality. Authors such as
Chakravorty (1990) and Cowell (1995) have put forward some interesting views
in this respect. Essentially, this literature has produced certain basic axioms,
which all measures should satisfy, and has analysed the characteristics
associated with the use of different measures. Among these, some of the most
well-known are the Gini coefficient, the Theil family of indices and the CV. The
essential point here is that the use of any one of these can generate different
results in terms of how each one weights the distances of the different
observations. This section will review the specific characteristics of the above
list of measures.
First and foremost, it should be pointed out that this paper is examining the
seasonal distribution of tourist activity throughout the year in the region and/or
destination in question. In this respect, temporal observations, which are the
basic units for the analysis, may be varied. A reasonable option would be to
consider the months of the year as basic units, which would give the analysis an
interesting level of detail, with the data available, and would match the typical
seasonal monitoring unit of tourism companies and public administrations. In
this respect, we know that it is not possible to provide even more detail on
tourist activity in a homogenous way in Spain. However, it is possible to choose
a more aggregated temporal base for the analysis, such as quarters, fourmonthly periods or seasons. In this respect, a certain amount of analytic
precision would be lost, but on the other hand it would avoid the consequences
of the calendar effect in some years, such as the positioning of Easter week.

6

The key point is thus to establish the limits of these sub-periods, which is no
trivial matter.
Let us suppose, for example, that the seasonal concentration of activity takes a
calendar month as the unit of reference. Therefore, the inequality measures
would actually be inter-monthly inequality measures of the activity. In particular,
one of the most widely recommended measures (even in the literature on
tourism) in this respect has been the Gini coefficient (Gini, 1912), based on the
well-known Lorenz curve, which is defined as twice the area between the curve
and the line of perfect equality (Wanhill, 1980; Lundtorp, 2001; Roselló et al,
2004; Fernandez-Morales, 2003 and Martin et al, 2014). One of the most
conventional formulas for this measurement is the following (already adapted to
the analysis of seasonal tourist concentrations):

G

1
 pi p j yi  y j
2 i j

(1)

where pi and pj are the relative weights of the observations (months or
seasons); yi is the variable for observations (in this case, tourist demand) and
is the annual mean.
It should be noted that i and j could be any two months in the year. It should
also be noted that the weightings of each observation, in their application to the
analysis of seasonality, are equal for all the compared periods and equivalent to
1/n, where “n” is the number of periods under consideration (if the periods are
symmetrical). This being the case, if the analysis focuses on a monthly
comparison, “n” is equal to 12, while if it focuses on a four-monthly comparison,
it would be 3.
In particular, three features have been highlighted on this measure (Wanhill,
1980 or Lundtorp, 2001). First, its reduced dependence on the changes in the
peak months; second, its high stability compared with other measures and third,
its low sensitivity to extreme values. However, this index has a curious property:
it gives greater importance to distributional changes taking place in the centre
7

(mode) of the distribution (in our case, in the months of annual average
demand), and gives a symmetrical weight to changes in the tails (i.e. months
with higher and lower demand). In this sense, it seems reasonable not
necessarily agree with this automatic behaviour. In fact, for example, in the
typical income analysis, often the researchers would preferably use more
"sensitive" to what occurs in the lower part of the distribution (progressive
measures). If this is so also in this field, alternative measures would probably
have to focus our attention, such as the Theil family of indices, which also
provide a more general framework for the treatment of this aspect. In this
sense, Theil (1967) proposed a family of inequality measures based on the
concept of entropy in information theory. Its general expression is given by:
 y
T   
 pi  i

   1 i


1




  1   0,1





(2)

Indeed, the parameter  captures the sensitivity of the measurement to the
place where the distribution changes. Specifically, the lower this value is then
the measure will be more sensitive to changes in the lower ranking of
observations (more progressive is the index). In the limit as tends to –∞ the
index only focuses on the lower end. In contrast, the higher is this parameter
the measures is more sensitive to changes at the top of the ranking, in our case
of months depending on their tourism demand. In fact, for values greater than 2
 only seems to be sensitive to the changes among the months of highest
demand (in fact, T(2) is a neutral index and ordinally equivalent to the CV).
Among this family, individual indices that have enjoyed the most attention are
T(0) and T (1), the algebraic expressions are:


T (   0)   p p log 
y 
i
 i

(3)

y  y 
T (   1)   p p  i  log i 
 
i
   

(4)

8

Finally it should be noted the advantages of the CV as a summary measure of
dispersion. The CV is constructed by dividing the standard deviation by the
mean and succeeds to fulfil the basic axioms required of any satisfactory
measure. If we square and divided it in half the resulting measure is ordinally
equivalent to T(2). An important feature of their behaviour is that the CV values
evenly redistributive changes within the distribution. Therefore, it is irrelevant in
relation to the specific place of the observations (the months in our case). This
distributive neutrality appears quite useful for the analysis of tourism seasonality
from a methodological point of view and also practical.

Therefore, each inequality index has its own characteristics associated with its
sensitivity. The Gini index is sensitive to central observations; the Theil to those
in the last positions of the ranking and the CV is neutral. The researcher should
choose, when applicable, the measure that best suits their preferences and
value judgments. Since there is no unanimous preference for one over the
others, we recommended looking at all of them and checking whether or not
they produce discrepancies in patterns, or at least in the general measurement.
This paper has chosen this path.

Beyond meeting the basic properties an additional attractive of some measures
is its ability to break down into parts. With regard to this possibility the inequality
measurement literature has focused, until now, in two different approaches: the
group decomposition and the source decomposition. In the first case, the
distributive observations (in our case, the different months) are divided in
groups to analyse which part of the total inequality (seasonality) can be
attributed to each group. Thus, an aggregate index that can be additively
decomposable could be expressed as the sum of a between-group component,
which would measure the average dissimilarity among the groups, and a withingroup component, which would reflect the internal differences and, in fact, is a
weighted average of the inequalities in the interior of the groups. Thus, and
based on the literature on the index decomposability, an index I is additively
decomposable over this decomposition format if:

9

I   wg I g  I b

(5)

g

where wg are the weights attached to each group, Ig is the inner-inequality for
each group and Ib is the inequality among groups.
In the case of the Gini coefficient it can be shown (Zagier (1983) that the former
inequality decomposition does not generally hold. Only in the cases where the
groups do not overlap, or in which the inequality within the group is zero, nontrivial case in the case of the seasonality analysis, the Gini coefficient is
additively decomposable in the sense pointed out by (5). This decomposition
implies obtaining in the expression an additional component, which is referred
as transvariation component, to solve the inequality in (5), which can be
interpreted, as a test of the cohesion degree of the groups. In fact, Zaiger
(1983) had already shown that the Gini coefficient can only be additively
decomposable if the group distributions do not overlap among them, a very
unlikely assumption. However, the test of the degree of group cohesion and the
appropriateness of its definition can be directly done by the own within-group
component, whose relative weight can be interpreted in terms of the error made
to do the month grouping.
The family of Theil Index is always additively decomposable in the sense of (5).
Shorrocks (1984) shown that the group decomposition of T(0) generates less
ambiguity, so that:

1
1 
T (0)  T (0) w  T (0) B   T 0g   ln
y
g g
g g
 i






(6)

where T(0)w is the aggregate between-groups inequality component; T(0)B
denotes the internal inequality present in group “g”. Finally, yi represents the
tourism activity in month “i”; and  is the average annual average on tourism
activity indicator.

10

Results derived from this decomposition analysis may be used for testing the
usefulness of the selected temporal aggregation criterion. If we observe that
most of the annual seasonality inequalities were attributable to intra-groups
disparities we might derive that informative relevance of the month-partitions
would be limited. This is because internal cohesion within the groups would be
small, losing the own significance of these groupings.
In order to perform the monthly groupings there are several available options:
quarters or four-month periods, so that the length of the periods is symmetric; or
periods of different lengths to take into account the actual activity concentration
in the different destinations. In any case, the choice of the longest periods
would tend to increase the inequality inside the groups and to reduce the
between groups inequality. In this paper, rather than grouping by quarters it was
decided to use a 5-4-3 rule, where the first period running from January to May
includes Easter and the months leading to higher demand; the second period
covers the typically high demand months and runs from June to September and
the final period consists of the last three months of the year1.
Also the CV is not decomposable in these terms. In particular, and if we take
the square of CV the sum of the inter-group and intra-group components do not
add to the global inequality, since the weights do not sum to unity. Further
decomposition is ambiguous, because the weights depend, in this case, on the
own weight of the demand of each month (Goerlich, 1998).
On the other side, another index decomposition that has received attention is
associated to the source decomposition. In this case, it should be assessed the
contribution over the total inequality by additive parts of the global component
(for example in our case typically incoming markets). The methodology
employed up to know in tourist applications (Fernandez-Morales and MayorgaToledano, 2008) was based on the decomposition technique for the Gini
suggested by Lerman and Yitzaki (1985), which allows to decompose this
The aggregate monthly periods can be established exogenously and globally as it is done in
this paper. Therefore, it would also be possible to establish endogenous aggregates for the
months, individualized for each province. Whatever the case, we believe that the four-monthly
groups and the 5+4+3 breakdown are intuitive and sufficiently illustrative at a general level.
1

11

measure as the sum for each source of the product of three factors: the
inequality arisen from each source (in that case the incoming market), its annual
weight, and a correlative component. However, and according to Shorrocks
(1982), there is no unique rule to do a factor distribution of the correlation
effects and thus, the contribution of the factors to the total inequality. As a
result, the contribution assigned to each component strictly depends on the way
the interaction effects are allocated among contributions. This is why the
literature does not consider Gini as a decomposable index in these terms
(Goerlich, 1998). Then, this kind of decomposition, contrary to what happens in
the group decomposition, is not quite clear. For the case of the Theil family of
indexes Shorrocks (1982) derived also precise source decompositions. CV also
allows it through what is called the natural decomposition of the variance. This
one basically implies the quantification of the contributions to the global
(seasonal) inequality that, basically, implies the quantification of the
contributions to global (seasonal) inequality of each additive factor from its own
dispersion and half of the affected covariances (Shorrocks, 1982). In fact,
Shorrocks (1982) proves that, under some very plausible axioms2, the previous
natural decomposition of the CV2, is the only unambiguous decomposition
method independently of the index used to measure whole inequality. Thus, a
rule can be established where the relative contributions of each factor can be
derived through the following expression (already adapted to the case of
incoming markets):

Ck 

Var M k    Cov M k , M j 
jk

Var ( M )

  Cov M k , M j   CovM k , M 

(7)

j

where Ck is the relative contribution made by incoming market k; Mk refers to
the vector of monthly market demand k; Var is the simple variance (or weighted)
and cov is the covariance.

2

The conditions are: a) the inequality index and the sources are continuous and symmetric. b)
The contributions do not depend on the aggregation level. c) The contributions of the factors
add the global inequality. d) The contribution of source k is zero if factor k is evenly distributed.
e) With two only factors, where one of them is a permutation of the other, the contributions must
be equal.

12

Thus, the rule which states that the weight of each source (market)
concentration (monthly) tourism demand depends on a direct component,
associated with the monthly market concentration itself, and an indirect
component associated with its correlation with other markets. In these terms, for
example, we can think hypothetically that the correlation between domestic
market and foreign market can be negative, in the sense that given the
concentration of the foreign market domestics may decide to go in the months
of lower foreign demand, thus avoiding congestion costs.
The basic data used for doing the application to Spain comes from the Hotel
Occupancy Survey of the Spanish Institute of Statistics (INE). In particular, they
refer to the indicator of hotel nights, which is a reasonable and operative
reflection of the impact of tourism on a region. For example, this indicator has
been used by Fernandez-Morales (2003), Fernandez-Morales and MayorgaToledano (2008), Cuccia and Rizzo (2011) or Martín et al (2014), among
others3. Furthermore, the volume of overnights generated by the hotels is far
greater than the one resulting from other accommodations options. In 2012,
280.7 million hotel overnights were generated nationwide – the corresponding
figures were 31.3 million for camping, 7.5 million for rural tourism and 63 million
for regulated apartments.
The data are distributed by months and by Spanish provinces and cover the
period of 1999-2012. In the empirical analysis, the main 14 provinces were
selected in terms of their annual hotel demand. In 2012, these provinces racked
up over 4 million hotel nights, which as a whole represents 83% of the overall
hotel demand in Spain. In order of demand, the following provinces were
included: the Balearic Islands, Las Palmas, Barcelona, Santa Cruz de Tenerife,
Madrid, Malaga, Alicante, Girona, Tarragona, Cadiz, Valencia, Granada, Seville
and Almeria. These provinces make up the bulk of the Mediterranean coastal
provinces and the islands, which typically have a strong demand for sun-sea-

3

Obviously it is plausible and interesting to extend the analysis to other tourist activity indicators
– for example, that of hotel supply, as was done by López and López (2006) and Capó et al.
(2007).

13

sand type vacations, plus the Spanish capital. The attached map shows their
specific locations4.

Map 1: Geographical location of Spanish provinces considered in the
empirical analysis

Note: Al is Alicante; AM is Almeria; BA is Baleares; BC is Barcelona; CA is Cadiz; GI is Girona;
GR is Granada; MA is Malaga; MD is Madrid; PA is Las Palmas; SC is Santa Cruz de Tenerife;
SE is Seville; TA is Tarragona and VA is Valencia
Source: Derived by present authors.

3. Empirical results for the main Spanish tourist provinces
Before examining the seasonal factor (i.e. the concentration) at a provincial
level, it is worth contextualizing these patterns in relation to the general
framework of the country. For this purpose Figure 1 is attached, which
represents the total annual evolution in Spain of hotel demand (hotel nights)
and their monthly concentration in terms of the four inequality indices proposed
in the previous section. In this respect, an average ascending pattern can be
observed in the total demand, although there are variations. For example, from
2002 to 2007, hotel demand clearly grew throughout the country, subsequently
4

Mainly to save space and to facilitate obtaining relevant patterns, it was decided to focus the
analysis on the major tourist provinces. Although seasonality in some of the excluded provinces
may be relevant, their overall explanatory power for seasonality in Spain is small given their low
relative weight on overall demand. Of the 10 provinces with the highest tourism seasonality, five
are studied in the paper; in particular, the top three (the Balearics, Tarragona and Girona) are
included. The author can provide all the data upon request.

14

stabilizing before showing a marked regression in 2009 (the large tourist crisis)
and then recovering through to 2011. In overall terms during this period, hotel
demand in Spain grew by 22%, or an annual cumulative average of 1.5%. At
the same time, it can be seen that the monthly concentration shows greater
stability, with a smaller seasonal variation range (mainly in the case of Gini and
CV). Whatever the case, and for simplification purposes, the concentration
recorded a downward path up to 2007, when it started a fairly clear upward
trajectory. For example, according to the CV or the Gini coefficient (G), the
monthly concentration increased from 2007 by almost 10%. Indeed, as a result
of this last evolution, and when compared with the overall evolution across the
whole period, a worsening in the concentration would have been recorded,
irrespective of the inequality index under consideration. In particular, increases
in concentration in the period as a whole would fluctuate between 5% and 9%,
according to the index. Therefore, at a domestic global level the context has
been one of a worsening in the concentration, this worsening being centred on
the period after 2007, which coincides with a typical phase of an increase in
global demand except for the economic (tourist) crisis of 2009.

Figure 1: Changes in yearly hotel demand and its monthly concentration
in Spain, 1999-2012

Note: data are on index numbers (1999=100): Demand is the total annual demand; CV is the
coefficient of variation; T0 and T1 are the Theil indexes with this sensivity parameters and G is
de Gini coefficient
Source: Derived by present authors from Hotel Occupancy Survey (INE) data.

15

3.1. Results of overall concentration in the country

In view of the above, what is the comparative seasonality of the provinces in
question? Is there any difference in the measurement in terms of the
concentration measure used? What has been the temporal evolution of this
concentration in recent years? For this purpose, Table 1 provides the data for
2012 as well as the global evolution from 1999-2012 according to the indices.
These calculations are made on a monthly basis. As a complementary exercise,
and to examine the possible ‘calendar impact’, for instance, the results are also
provided using a four-monthly base unit in Table 2 and by asymmetric monthly
periods in Table 3, coinciding approximately with the general low season
periods in these types of provinces (January to May and October to December)
and the high season period (June to September).
Firstly, Table 1 reproduces the information relating to the concentration of hotel
demand by provinces on a monthly basis. This is hypothetically the most
accurate measure, but can be affected by the ‘calendar impact’ in some
particular years and particularly by the dates of the Easter holiday.
Nevertheless, we believe that it provides a useful start to the analysis.
Furthermore, it should be noted that this element has no significant effect on
territorial comparisons (i.e. cross-sections). In this respect, the results indicate
certain interesting points. For example, the rankings typically do not vary very
much when using the different concentration indices. In 2012 the most
unbalanced provinces (seasonally) were unanimously the Balearic Islands,
Tarragona and Girona (typical sun-sand-and-sea destinations) while the least
unbalanced were the provinces of the Canary Islands (lower monthly climate
variations) and Madrid, which specializes in business and cultural tourism, as
well as Granada and Seville which both have a strong cultural component (e.g.
World Heritage Sites) and tourist diversification. Meanwhile, in terms of
evolution there was a worsening on average, although with different patterns
across the provinces. For example, the Balearic Islands and Almeria were the
provinces that witnessed the highest growth in monthly hotel demand
concentration for the period 1999-2012 (particularly problematic in the case of
16

the Balearics, given the high absolute level). Although in most cases the
concentration worsened, in some cases the seasonal pattern is one of
improvement. This is the case, for example, of Girona (which is following a
strategy of diversification towards a more cultural tourism offering), Barcelona
(the effect of being a European tourism capital and its diversification), Granada
and Santa Cruz de Tenerife. Finally, in terms of the direction of the evolution,
and although the indices essentially point a broadly similar direction, they
typically diverge in the scale of the variation, given their different properties, and
even in some specific cases even in the direction of change. For example, the
latter was the case of Tarragona, whose concentration rose slightly based on
the CV but dropped according to the Theil Index and slightly increased
according to the Gini coefficient. Therefore, we must be careful to establish firm
conclusions on the basis of a single index, not only in terms of the magnitude of
evolution but sometimes even in terms of the direction of change itself.
Nevertheless, and in comparative terms, for instance, Fernandez-Morales
(2003) conducted an analysis of the monthly concentration of three of our
detailed provinces (Almeria, Granada and Malaga), for an earlier period (19902000), using the same demand indicator (overnights) and the Gini index.
Among its key findings, the author found a reduction in concentration during this
period in the case of Almeria (sun and beach) and Granada (diversity of
product) and stabilization in the case of Málaga (sun and beach). In contrast,
our results indicate a worsening for Almeria and Malaga, both highly specialized
products sun and sand, but a new improvement in the case of Granada.
In addition, Figure 2 reproduces the entire time series of monthly concentration
of demand for provinces with a higher than national average seasonality in
2012; Figure 3 concerns those with below average seasonality. To save space,
we only offer results from the CV, the neutral index.5 Firstly, we note that in
general the range of variation of the concentration is not particularly high over
time. In fact, the variation in concentration is clearly higher if we compare
5

Because this is a complementary analysis and it is necessary to save space, we have decided
only to offer the results from CV. Nevertheless, If we extend the temporal (yearly) analysis to the
other concentration measures, there would not be significant overall qualitative differences in
the selected provinces. Results available on request.

17

provinces. This stability can be attributed partly, but only partly, to the index
itself. In fact, it is not easy to change the concentration much in the short term.
Secondly, if we focus on the provinces with the highest concentration, a
monotonic increasing comparative pattern for the Balearic Islands can be seen
(again really a disturbing pattern), a clear progression may be observed (mainly
since 2007) for Cadiz and Almeria, both specialized mainly in sun-sand-beach.
The decrease in concentration experienced by Girona, especially until 2008, is
noteworthy. In the cases of Málaga, Alicante or Valencia (all coastal provinces
and specialized in sun and beach) there is a slightly increase and, finally, in
Tarragona the pattern is fundamentally stable, albeit including a record high.
Thirdly, in regard to the provinces with lower monthly concentration (relative to
the Spanish mean), no very clear patterns emerge. However, for the two
Canary provinces, Las Palmas increased until 2010 and then declined while, in
Santa Cruz de Tenerife there was an increase until 2005 and a subsequent
decrease which emerged at the end of the period with the lowest tourist
concentration.
So, monthly results indicate that coastal provinces typically specializing in sun
and beach, with the exception of Catalan provinces, recorded a worsening of
seasonality in the period. This seems to indicate the relevance of a regional
factor in explaining the difference and might, in part, be attributable to policy.

18

Table 1: Monthly concentration of hotel demand in major Spanish tourist
provinces according to different measures, 1999 and 2012

Balearic Islands
Tarragona
Girona
Almeria
Cadiz
Malaga
Barcelona
Alicante/Alacant
Valencia/València
Seville
Granada
Palmas, Las
Madrid
Santa Cruz Tenerife
SPAIN

CV.
1999
0.6556
0.8595
0.8556
0.4167
0.4765
0.3293
0.4206
0.2321
0.2091
0.2208
0.1817
0.0768
0.1174
0.0952
0.3821

2012
0.8824
0.8692
0.7596
0.6738
0.5976
0.4343
0.3662
0.3352
0.3250
0.2194
0.1572
0.1151
0.1084
0.0744
0.4037

Theil-0
1999
0.3036
0.6604
0.4937
0.0874
0.1268
0.0589
0.0964
0.0268
0.0228
0.0267
0.0174
0.0029
0.0070
0.0044
0.0775

2012
0.7210
0.5708
0.3473
0.2223
0.2136
0.1091
0.0728
0.0559
0.0548
0.0241
0.0133
0.0064
0.0059
0.0028
0.0831

Theil-1
1999
0.2375
0.4211
0.3739
0.0849
0.1153
0.0555
0.0900
0.0267
0.0222
0.0253
0.0168
0.0029
0.0069
0.0045
0.0737

2012
0.4402
0.4029
0.2867
0.2098
0.1830
0.0983
0.0684
0.0551
0.0528
0.0239
0.0128
0.0065
0.0059
0.0028
0.0804

Gini
1999
0.3650
0.4783
0.4706
0.2338
0.2672
0.1872
0.2391
0.1311
0.1167
0.1247
0.1036
0.0430
0.0663
0.0514
0.2167

2012
0.4906
0.4800
0.4194
0.3619
0.3367
0.2469
0.2085
0.1877
0.1840
0.1245
0.0881
0.0629
0.0614
0.0419
0.2273

Note: provinces sorted in descending order based on the value of CV (neutral index) at 2012.
Source: Derived by present authors from Hotel Occupancy Survey (INE) data.

Figure 2: Changes in yearly hotel demand and its monthly concentration
(for Spanish provinces that had above average seasonality in 2012), 19992012

Note: Bal is Balearic Islands; Tgn is Tarragona; Gir is Girona; Alm is Almeria; Cad is Cadiz; Mal
is Malaga; Bcn is Barcelona; Ali is Alicante and Val is Valencia.
Source: Derived by present authors from Hotel Occupancy Survey (INE) data.

19

Figure 3: Changes in yearly hotel demand and its monthly concentration
(for Spanish provinces that had below average seasonality in 2012), 19992012

Note: Sev is Seville; Gra is Granada; Pal is Palmas; Mad is Madrid; Sta is Santa Cruz de
Tenerife.
Source: Derived by present authors from Hotel Occupancy Survey (INE) data.

Table 2 shows the same analysis than in Table 1 but based on four-monthly
periods. Consequently the results are less detailed yet involve two additional
factors. Firstly, calculations on a four-monthly basis can avoid the impact of the
measurement of holiday periods and their different positions depending on the
year in question. Secondly, given that the analysis units are more aggregate,
the concentration index values generate lower figures. The context is once
again one of an increase in global concentration over the period. With regard to
the rankings, most of the positions are generally maintained when compared to
those computed on a monthly basis. For example, the Balearic Islands,
Tarragona and Girona emerge as the most unbalanced provinces while the
Canary Islands (Las Palmas and Santa Cruz) are the least unbalanced.
However, there are some specific differences; for example, Madrid’s position
worsened. In terms of evolution, if a comparison is made between the indices
they are broadly similar, although it is possible to find discrepancies in a
detailed scale. Meanwhile, a temporal comparison of the evolution of the indices
on a four-monthly and monthly basis generally points also same direction, with
20

certain exceptions such as Granada, whose imbalance is smaller on a monthly
basis yet increases with the four-monthly one.

Table 2: Quarterly concentration of hotel demand in major Spanish tourist
provinces, according to different measures, 1999 and 2012
Balearic Islands
Tarragona
Girona
Cadiz
Almeria
Malaga
Barcelona
Alicante/Alacant
Valencia/València
Granada
Madrid
Seville
Palmas
Granada
Santa Cruz Tenerife
SPAIN

CV
1999
0.5191
0.6857
0.6708
0.3424
0.3073
0.2321
0.3361
0.1607
0.1028
0.0584
0.0334
0.0742
0.0150
0.0584
0.0097
0.2894

2012
0.7160
0.6935
0.5584
0.4430
0.4377
0.3160
0.2719
0.2421
0.2230
0.0693
0.0305
0.0286
0.0261
0.0693
0.0229
0.3059

Theil-0
1999
0.1416
0.2858
0.2284
0.0562
0.0438
0.0258
0.0577
0.0125
0.0053
0.0017
0.0006
0.0029
0.0001
0.0017
0.0000
0.0409

2012
0.3098
0.2590
0.1553
0.0924
0.0869
0.0463
0.0353
0.0276
0.0230
0.0024
0.0005
0.0004
0.0003
0.0024
0.0003
0.0446

Theil-1
1999
0.1335
0.2412
0.2160
0.0567
0.0451
0.0262
0.0563
0.0127
0.0053
0.0017
0.0006
0.0028
0.0001
0.0017
0.0000
0.0410

2012
0.2610
0.2354
0.1504
0.0934
0.0898
0.0477
0.0358
0.0283
0.0238
0.0024
0.0005
0.0004
0.0003
0.0024
0.0003
0.0452

Gini
1999
0.2764
0.3657
0.3453
0.1796
0.1558
0.1220
0.1813
0.0848
0.0559
0.0318
0.0182
0.0356
0.0074
0.0318
0.0052
0.1540

2012
0.3797
0.3615
0.2908
0.2274
0.2147
0.1602
0.1423
0.1246
0.1061
0.0372
0.0160
0.0156
0.0142
0.0372
0.0109
0.1598

Note: provinces sorted in descending order based on the value of CV (neutral index) at 2012.
Source: Derived by present authors from Hotel Occupancy Survey (INE) data.

Table 3 shows the analysis with a different aggregate separation, i.e.
considering three heterogeneous periods spanning a first period from January
to May, a second one typical of the high season, from June to September, and
a final low season period, from October to December. In this respect, the results
indicate that: first, overall, the ascending pattern of seasonal concentration
remains the same, except for the Gini; second, although there are certain
changes in the ranking, these are not dramatic ones. In any event, it is worth
highlighting the changes experienced by Seville, which comparatively is better
positioned using the periodic basis than in monthly terms, but the change is not
dramatic; third, in general evolutionary terms, there are no major discrepancies
between these calculations and those deriving from a monthly base unit.
Perhaps the most notable cases are those of Granada (whose concentration
drops with the monthly base unit and increases when the 5+4+3 breakdown is
used), Seville and Madrid (the same evolution as Granada). Thus, with these
calculations, the territorial image of a growth in seasonality is more general;

21

fourth, in this case, significant discrepancies in evolution can be observed in
some provinces, depending on the index chosen. For example, Cadiz, Malaga,
Alicante, Granada, Las Palmas and Seville show different types of variations
depending on the index used. Typically, these discrepancies are between the
Gini Coefficient and the others. So, in this case the particular choice of an index
is more important that in the previous cases.

Table 3: Concentration of hotel demand by seasons (5+4+3) in major
Spanish tourist provinces, according to different measures, 1999 and 2012
Tarragona
Balearic Islands
Girona
Almeria
Cadiz
Málaga
Barcelona
Alicante/Alacant
Valencia/València
Granada
Palmas
Seville
Madrid
Santa Cruz Tenerife
SPAIN

CV
1999
0.7558
0.5426
0.7566
0.3673
0.3754
0.2649
0.3542
0.1862
0.1519
0.0882
0.0259
0.0319
0.0090
0.0267
0.3199

2012
0.8016
0.7957
0.6739
0.6221
0.5196
0.3680
0.3048
0.2874
0.2819
0.1134
0.0613
0.0419
0.0153
0.0112
0.3574

Theil-0
1999
0.2682
0.1340
0.2782
0.0621
0.0640
0.0323
0.0569
0.0162
0.0109
0.0041
0.0003
0.0005
0.0000
0.0004
0.0466

2012
0.3048
0.2981
0.2066
0.1857
0.1200
0.0599
0.0417
0.0370
0.0353
0.0056
0.0020
0.0008
0.0001
0.0001
0.0570

Theil-1
1999
0.2644
0.1371
0.2680
0.0639
0.0663
0.0335
0.0591
0.0167
0.0112
0.0040
0.0003
0.0005
0.0000
0.0004
0.0484

2012
0.2932
0.2902
0.2069
0.1754
0.1227
0.0615
0.0432
0.0384
0.0358
0.0055
0.0020
0.0007
0.0001
0.0001
0.0592

Gini
1999
0.5384
0.4567
0.5979
0.3892
0.3640
0.3073
0.3108
0.2565
0.2294
0.2729
0.1563
0.1631
0.1580
0.1772
0.3266

2012
0.4980
0.4775
0.4150
0.4407
0.3304
0.2461
0.1892
0.1803
0.2010
0.0960
0.0386
0.0378
0.0101
0.0082
0.2205

Note: Provinces sorted in descending order based on the value of CV (neutral index) at 2012.
The monthly structuring 5 + 4 + 3 indicates three monthly groups: one from January to May (5
months); another from June to September (4 months) and, finally, October to December.
Source: Derived by present authors from Hotel Occupancy Survey (INE) data.

Nevertheless, despite the variations, when examining the changes in detail or
by scale, the bulk of the overall results do not seem to vary excessively
according to whether the seasonal concentration of hotel demand is ascertained
on a monthly, four-monthly or seasonal basis. It typically worsens in most of
Spain’s main tourist provinces. However, there are notable exceptions, such as
the provinces of Girona and Barcelona, both of which are coastal and yet
despite specializing in sun-sand-and-sea type tourism stand out for their strong
positioning in the cultural and heritage sector. Similarly, the ranking of provinces
in the last year for which reliable data are available is headed by the Balearic
Islands, Tarragona and Girona; in contrast, provinces such as Santa Cruz de
22

Tenerife and Las Palmas (i.e., the Canary Islands), with a year-round mild
climate, and Granada and Seville (with a strong cultural component) and Madrid
(business tourism and the capital city) show very much lower imbalances. On
the other hand, the choice of index typically does not yield very different results,
perhaps with the exception of seasonal analysis (5+4+ 3), where discrepancies
are observed mainly between the Gini and the other measures in the case of six
provinces.

3.2. Decomposing seasonality
Table 4 shows the decomposition by monthly groups (seasons) of the monthly
concentration of hotel demand in the main Spanish provinces based on the
results obtained from the Theil Index (T(0)). It should be remembered that the
choice of this index for this exercise was mainly based on its facility to be
decomposed and its lack of ambiguity with regard to this type of decomposition
(instead using, for example, the Gini coefficient. See section 2). In this respect,
the results reveal some interesting points:
Firstly, the monthly groupings are quite important as an explanatory element of
seasonal concentration. In the country as a whole, the four-monthly
segmentation explains more than half the global concentration. The explanatory
capacity of structuring by periods (5+4+3) is even higher, having increased its
weight by up to almost 70% of global inequalities in 2012.
Secondly, as is to be expected, there are regional differences. Typically, the
groups used are more important in the cases in which the global concentration
is more important, which is to be expected. This being the case, in most of the
provinces where seasonality is higher, the groupings are significant as an
element of synthesis. The case of Almeria, for example, stands out, where the
segmentation of 5+4+3 explains 84% of the monthly concentration. On the other
side of the coin, there are the provinces of Seville, Madrid and Santa Cruz,
where the groups barely have any explanatory capacity.
Thirdly, in relation to the variations in the concentration and in the role of the
group components, it is obvious that a single pattern cannot be established.
Whatever the case, it is true that typically the role of the between component,

23

which is the one that justifies the explanatory capacity of this decomposition, is
important. For example, in the four-monthly evaluation, the changes in the
between component explain more than 40% of the growth in concentration in
provinces such as the Balearic Islands, Cadiz, Malaga and Alicante. For
example, this group component explains half the reduction in the monthly
concentration that took place in Girona. Indeed, the inter-group component in
this case explains two-thirds of the overall increase in Spain. FernandezMorales (2003) also decomposed seasonality by groups but considered
differentially the possibility that months are not consecutive and used the Gini
index as a reference measure for three of our provinces (Malaga, Granada and
Almeria) and for an earlier period. In this case, given the flexibility of forming
groups, which are different in each province, the weight of between-inequality
component is typically higher. Nevertheless, we believe that, in terms of policy,
it is more reasonable that the months be consecutive in each group.

24

Table 4: Decomposing of monthly concentration of hotel demand by
subgroups (Theil index)
Fourmonths
seasons
5+4+3
1999
2012
1999
2012
Between Within Between Within Between Within Between
0.1416
0.1620
0.3098
0.4112
0.1340
0.1696
0.2981
Balearic Islands
(47%)
(53%)
(43%)
(57%)
(44%)
(56%)
(41%)
0.2858
0.3746
0.2590
0.3118
0.2682
0.3922
0.3048
Tarragona
(43%)
(57%)
(45%)
(55%)
(41%)
(59%)
(53%)
0.2284
0.2653
0.1553
0.192
0.2782
0.2155
0.2066
Girona
(46%)
(54%)
(45%)
(55%)
(56%)
(44%)
(59%)
0.0438
0.0436
0.0869
0.1354
0.0621
0.0253
0.1857
Almeria
(50%)
(50%)
(39%)
(61%)
(71%)
(29%)
(84%)
0.0562
0.0706
0.0924
0.1212
0.0640
0.0628
0.1200
Cadiz
(44%)
(56%)
(43%)
(57%)
(50%)
(50%)
(56%)
0.0258
0.0331
0.0463
0.0628
0.0323
0.0266
0.0599
Málaga
(44%)
(56%)
(42%)
(58%)
(55%)
(45%)
(55%)
0.0577
0.0387
0.0353
0.0375
0.0569
0.0395
0.0417
Barcelona
(60%)
(40%)
(48%)
(52%)
(59%)
(41%)
(57%)
0.0125
0.0143
0.0276
0.0283
0.0162
0.0106
0.0370
Alicante/Alacant
(47%)
(53%)
(49%)
(51%)
(60%)
(40%)
(66%)
0.0053
0.0175
0.0230
0.0318
0.0109
0.0119
0.0353
Valencia/València
(23%)
(77%)
(42%)
(58%)
(48%)
(52%)
(64%)
0.0029
0.0238
0.0004
0.0237
0.0005
0.0262
0.0008
Seville
(11%)
(89%)
(2%)
(98%)
(2%)
(98%)
(3%)
0.0017
0.0157
0.0024
0.0109
0.0041
0.0133
0.0056
Granada
(10%)
(90%)
(18%)
(82%)
(24%)
(76%)
(42%)
0.0001
0.0028
0.0003
0.0061
0.0003
0.0026
0.0020
Palmas, Las
(3%)
(97%)
(5%)
(95%)
(12%)
(88%)
(31%)
0.0006
0.0064
0.0005
0.0054
0.0000
0.0070
0.0001
Madrid
(9%)
(91%)
(8%)
(92%)
(1%)
(99%)
(2%)
0.0000
0.0044
0.0003
0.0025
0.0004
0.0040
0.0001
Santa Cruz Tenerife
(0%)
(100%)
(11%)
(89%)
(8%)
(92%)
(4%)
0.0409
0.0366
0.0446
0.0385
0.0466
0.0309
0.0570
SPAIN
(53%)
(47%)
(54%)
(46%)
(60%)
(40%)
(69%)
Note: The monthly structuring 5 + 4 + 3 indicates three monthly groups: one from January to
May (5 months); another from June to September (4 months) and, finally, October to December.
Source: Derived by present authors from Hotel Occupancy Survey (INE) data.

Meanwhile, Table 5 shows the main results emerging from the decomposition of
the monthly concentration by additive sources; in this case, by major markets
(i.e. domestic as opposed to international). The distinction is interesting
because the behaviour in terms of seasonality and the overall explanatory role
may be different. For instance, domestic tourists have more access to houses,
more knowledge about national destinations, typically domestic prices are lower
than the international ones and also we can find differences in terms of the
average stay and daily expenditures. Ex-ante, we may expect less seasonality
among domestic tourists that for international ones.

25

Within
0.4229
(50%)
0.2660
(47%)
0.1407
(41%)
0.0366
(16%)
0.0936
(44%)
0.0492
(45%)
0.0311
(43%)
0.0189
(34%)
0.0195
(36%)
0.0233
(97%)
0.0077
(58%)
0.0044
(69%)
0.0058
(98%)
0.0027
(96%)
0.0261
(31%)

The results stem from the use of the natural variance decomposition rule, which
provides the total contribution of each factor as the addition of its individual
variance and the cross-covariance. It can be seen that the decomposition could
have been specified much more at the level of source markets, but our
understanding was that as a first analysis it could be sufficient. In fact it can be
surmised a priori that the majority of the differences in the monthly
concentration of markets would be attributable to the domestic versus
international market binomial. Table 6 also details the role of the direct
component (the individual variation) and the indirect component (the covariance
between the two large markets) in explaining the relative contribution of each
major market. It can be seen that, in this case, it is interesting to know the sign
and relevance of the correlation and, consequently, to verify whether the two
markets strengthen the monthly concentration or not. Table 7, meanwhile,
reproduces two of the main factors that would be behind the direct component,
i.e. the weighting and the individual variation, in this case approximate, for the
CV. In particular, the main results obtained can be summed up in the following
points:
Firstly, seasonality can be explained nationwide by the foreign component.
Thus in the case of Spain as a whole, the foreign market explains approximately
70% of the seasonal concentration. This is due to its greater global weighting
and its higher monthly concentration as compared to the domestic market.
Secondly, the main factor behind these weightings, especially if they are large,
is typically the direct component. Whatever the case, the indirect component
always has a positive value, which is indicative of the fact that both markets
strengthen each other in determining the concentration, so without any
significant compensatory effect between the domestic and international markets
being observed. Indeed, in seven of the Spanish provinces analysed, the
domestic market concentration is even higher that the international one (Table
7), in contrast to the previous standard view. In particular, this is the case in
coastal sun and beach provinces (except the Catalan ones) and the Canary
Islands provinces. Therefore, the increase in the domestic market role in the
provinces would not necessarily reduce seasonality in all of them. It depends on
the particular province.
26

Thirdly, focusing now on the direct effect, the documented increase in the
country’s seasonal concentration would be attributable, above all, to the
increase in the monthly concentration experienced in the two major markets, i.e.
the individual seasonal variation (indicated in Table 7 by the CV).
Fourthly, focusing now on the territorial analysis, in most of the provinces under
consideration the foreign market is the main generator of concentration, in line
with the results obtained at an aggregate level. Only in the cases of Almeria,
Cadiz, Alicante and Valencia is the domestic market the main protagonist. In all
these cases, this role is attributed both to the yearly weight of the domestic
market as to the differential seasonality of this market regarding international
one.
As point five, when the evolution is examined, there is a very significant
increase in the weighting of the foreign component in regions such as
Tarragona, Alicante, Valencia, Seville, Madrid and Santa Cruz de Tenerife. In
all these regions, the weighting of the foreign market increases and also it is
observed a worsening of their individual concentrations, except in case of
Madrid (see Table 7).
In sixth place, in the provinces of Almeria and Las Palmas it is the domestic
component whose participation increases, in both cases due to the combined
effect of its weighting and individual concentration.
Table 5: Decomposing monthly concentration by relative contribution of
large markets, 1999 and 2012

Balearic Islands
Tarragona
Girona
Almeria
Cadiz
Málaga
Barcelona
Alicante/Alacant
Valencia/València
Seville
Granada
Palmas, Las
Madrid
Santa Cruz Tenerife
SPAIN

1999
Domestic
0,0308
0,3458
0,1782
0,5686
0,5906
0,3491
0,0993
0,8558
0,7595
0,3592
0,3951
0,2672
0,2998
0,4462
0,2845

2012
International Domestic
0,9692
0,0524
0,6542
0,2993
0,8218
0,2062
0,4314
0,8294
0,4094
0,6319
0,6509
0,3897
0,9007
0,0798
0,1442
0,6083
0,2405
0,6397
0,6408
0,2821
0,6049
0,4252
0,7328
0,3887
0,7002
0,2459
0,5538
0,1859
0,7171
0,3028

International
0,9476
0,7007
0,7938
0,1706
0,3681
0,6103
0,9202
0,3917
0,3603
0,7179
0,5748
0,6113
0,7541
0,8141
0,6978

Source: Derived by present authors from Hotel Occupancy Survey (INE) data.

27

Table 6: Decomposing monthly concentration by relative contribution of
large markets, 1999 and 2012 and by direct and indirect components

Balearic Islands
Tarragona
Girona
Almeria
Cadiz
Málaga
Barcelona
Alicante/Alacant
Valencia/València
Seville
Granada
Palmas, Las
Madrid
Santa Cruz Tenerife
SPAIN

Dom.
Direct
0,0023
0,1307
0,0390
0,3829
0,3897
0,1476
0,0130
0,7477
0,5959
0,1652
0,3191
0,1427
0,1682
0,4232
0,0932

1999
Int.
Direct
0,9407
0,4392
0,6825
0,2458
0,2084
0,4493
0,8144
0,0361
0,0768
0,4468
0,5289
0,6084
0,5687
0,5308
0,5258

Indirect
0,0285
0,2151
0,1393
0,1856
0,2010
0,2015
0,0863
0,1081
0,1637
0,1940
0,0760
0,1244
0,1315
0,0230
0,1913

Dom.
Direct
0,0042
0,0968
0,0509
0,6940
0,4309
0,1945
0,0092
0,3962
0,4209
0,1302
0,3316
0,2996
0,1115
0,4487
0,1016

2012
Int.
Direct
0,8993
0,4982
0,6385
0,0352
0,1671
0,4150
0,8496
0,1796
0,1415
0,5660
0,4813
0,5222
0,6197
1,0770
0,4965

Indirect
0,0482
0,2025
0,1553
0,1354
0,2010
0,1953
0,0706
0,2121
0,2188
0,1519
0,0936
0,0891
0,1344
-0,2629
0,2012

Note: relative weights of total variance. In the case of the indirect component is the total weight,
which would be symetrically allocated by each market.
Source: Derived by present authors from Hotel Occupancy Survey (INE) data.

Table 7: Decomposing monthly concentration by large markets, 1999 and
2012. Explanatory Factors.
Balearic Islands
Tarragona
Girona
Almeria
Cadiz
Málaga
Barcelona
Alicante/Alacant
Valencia/València
Seville
Granada
Palmas, Las
Madrid
Santa Cruz Tenerife
SPAIN

1999
%Dom
0,0960
0,3945
0,2508
0,4790
0,5249
0,2899
0,3116
0,4384
0,7749
0,5144
0,5407
0,1099
0,5237
0,2011
0,3541

CV Dom
1,1447
2,7287
2,3329
1,8650
1,9631
1,5121
0,5331
1,5860
0,7214
0,6043
0,6575
0,9151
0,3186
1,0664
1,1411

%Int
0,9040
0,6055
0,7492
0,5210
0,4751
0,7101
0,6884
0,5616
0,2252
0,4856
0,4593
0,8901
0,4763
0,7989
0,6474

CV Int
2,4367
3,2585
3,2682
1,3735
1,5864
1,0769
1,9100
0,2719
0,8914
1,0526
0,9965
0,2332
0,6441
0,3007
1,4825

2012
%Dom
0,0885
0,3789
0,2900
0,7850
0,5678
0,3371
0,2219
0,5429
0,6188
0,4651
0,5990
0,1343
0,5009
0,1809
0,3642

CV Dom
2,2282
2,4731
2,0461
2,4771
2,3931
1,9680
0,5495
1,3462
1,1801
0,5897
0,5235
1,6252
0,2504
0,9544
1,2242

%Int
0,9115
0,6211
0,7100
0,2150
0,4322
0,6629
0,7781
0,4571
0,3812
0,5349
0,4010
0,8657
0,4991
0,8191
0,6368

Source: Derived by present authors from Hotel Occupancy Survey (INE) data.

Finally, Table 8 reproduces in detail the decomposition of the concentration by
individual markets, apart from the major markets shown in Table 5. In relation to
the domestic market, its weighting in the concentration stands out, providing

28

CV Int
3,1802
3,4214
2,9617
2,0383
1,9582
1,4618
1,5027
1,0767
1,1107
1,0690
0,9422
0,3328
0,5924
0,3266
1,5474

30% of the monthly concentration. This is followed by the British market (20%)
and the German market (17%), in that order. Therefore these three markets
account for two-thirds of the global concentration in Spain, and must be given
priority in the strategy of fighting this inequality on a global scale. Obviously,
looking down the list different patterns can be seen, fundamentally attributable
to the weighting of these markets in the annual global demand. To mention a
few of the cases that differ from the overall pattern, it is first worth highlighting
the case of the Balearic Islands, where the German and British markets are the
main generators of concentration (jointly explaining 66% of the monthly
concentration in this province) while the domestic market has a relatively low
weighting. Indeed, Roselló et al. (2004) analysed the pattern of seasonality of
the British and German markets for the period 1982-2001 in the case precisely
of the Balearic Islands. They found, for example, a growing seasonality for the
British market in the islands, but a decreasing one in the German case.
Moreover, in both cases it was found that income and prices acted to reduce
and increase seasonality, respectively and also that that Germans are less price
sensitive than the English and also, in the short term, with regard to income. On
the other hand, in the case of the province of Tarragona, the rise in Russian
tourism has led to this particular market generating 30% of the global
concentration, an almost identical weight to that of the domestic market. In
Girona, the northernmost Mediterranean province, Russian tourism also plays
an important role (16%), similar to that of the French market and close to the
21% of the domestic market. In the case of Almeria, above all, but also Cadiz,
Alicante and Valencia, it is once again the domestic market that is the keydetermining factor of the situation. In the case of Barcelona, there is no
particular bias towards any one market. In the other provinces, which are
typically those with a comparatively lower concentration, the domestic market is
generally the main provider.

29

Table 8: Decomposing hotel demand monthly concentration by markets
(detailed), 2012
Balearic Islands
Tarragona
Girona
Almeria
Cadiz
Málaga
Barcelona
Alicante/Alacant
Valencia/València
Seville
Granada
Palmas, Las
Madrid
Santa Cruz Tenerife
SPAIN

Dom
0,052
0,299
0,206
0,829
0,632
0,390
0,080
0,608
0,640
0,282
0,425
0,389
0,246
0,186
0,303

Ger
0,334
0,014
0,107
0,000
0,231
0,054
0,108
0,010
0,009
0,074
0,127
0,200
0,061
0,219
0,167

UK
0,334
0,189
0,061
0,035
0,016
0,217
0,114
0,177
0,034
0,073
0,030
0,193
0,053
0,096
0,206

Fra
0,031
0,070
0,165
0,011
0,017
0,071
0,079
0,032
0,058
0,149
0,144
0,044
0,042
0,019
0,055

Ita
0,056
0,004
0,036
0,006
0,016
0,022
0,038
0,015
0,088
0,062
0,065
0,041
0,017
0,089
0,039

Rus
0,020
0,300
0,164
0,007
0,003
0,025
0,135
0,022
0,011
0,004
0,006
0,000
0,022
0,016
0,060

Net
0,022
0,026
0,037
0,009
0,014
0,041
0,082
0,004
0,026
0,043
0,045
0,067
0,023
0,050
0,029

Bel
0,013
0,026
0,074
0,024
0,016
0,022
0,035
0,004
0,010
0,020
0,028
0,030
0,013
0,037
0,020

Rest
0,138
0,072
0,150
0,079
0,055
0,159
0,330
0,128
0,125
0,293
0,130
0,037
0,524
0,288
0,122

Note: Dom is domestic demand; Ger is Germany. Fra is France; Ita is Italy; Rus is Russia; Net
is Netherlands; Bel is Belgium; Rest is rest of markets
Source: Derived by present authors from Hotel Occupancy Survey (INE) data.

4. Concluding Remarks
As Koenig-Lewis and Bischoff (2005) demonstrated in an interesting survey on
research into seasonality, the aspects relating to its measurement represent an
area of particular interest. In this respect, the scant attention paid to this area by
the literature is fairly surprising, particularly the lack of interest shown in
adapting the instruments for measuring inequality to analyse tourist seasonality.
In this respect, this work highlights two basic issues of interest in the
methodological field.
Firstly, there are different measures for approximating the level of seasonal
concentration of tourist activity, including, but not only, the Gini coefficient. The
important point is that every index makes a different weighting of the changes
that occur in the different units of time used (e.g. months). In particular, the Gini
coefficient is very sensitive to the changes that take place in the months with
demand close to the average. Thus researchers have no reason to share this

30

benign behaviour. In this respect, researchers might prefer other behaviour or,
perhaps even better manage a set of indices to obtain a more comprehensive
overview of their measurement. This paper proposes the use, apart from the
Gini coefficient, of the Theil Index (which is more sensitive to changes in the
months with lower demand) and the Coefficient of Variation (CV), which is
neutral.
Secondly, in this measurement context, it is interesting to use decomposition
techniques by factors. In this respect, two approaches have been taken:
decomposition by groups and decomposition by additive sources. The former
involves observing the role of groups of months, or seasons, in explaining the
global concentration. The latter makes an evaluation of the role of additive
sources (for example, source markets) in generating the annual concentration.
With regard to group decomposition, of the three indices used the best one in
methodological terms is the Theil Index. The Gini coefficient can be
decomposed but only roughly, and the CV poses problems apart from not being
perfectly decomposable (the sum of the group components does not give a
total). With regard to additive decomposition, things are more complex.
Whatever the case, the natural decomposition variance rule would also apply to
CV and in certain eventualities it would even be a single rule irrespective of the
index.
Given the above methodological points, this paper has made an empirical study
of the main Spanish tourist provinces in the period 1999-2012, which should be
sufficient for observing significant variations in seasonal concentration, given its
limited variability in the short term. Hotel nights have been used as an indicator
of tourist activity along with the abovementioned indicators and methods. In this
respect, some of the main results obtained are as follows:
Firstly, the global context, across the whole country, is one of growth in
seasonality, especially since 2007. In this respect, the growth of Spain as an
essentially international destination has occurred without any significant
correction of the seasonal concentration. This result is irrespective of the index
and the time base used (i.e. monthly, four-monthly or by season). This result

31

casts doubt on the sustainability of growth and is important for prioritizing
corrective policies.
Secondly, although in a significant part the provincial rankings remain the same
when we compare the different concentration measures (Gini, Theil and CV),
there are, however, differences in some cases. Nevertheless, the concentration
is typically led by the Balearic Islands and two Catalan provinces (Girona and
Tarragona). On the other side of the coin are the Canary Island provinces and
Madrid. Factors relating to the tourism product itself and seasonal climate
variations would be behind these behavioural extremes.
Thirdly, the comparison of results in terms of the time base under consideration
does not typically throw up many substantial changes but also we can find, in
the detail, some divergences in some provinces.
Fourthly, the choice of index measuring seasonality at the provincial level
typically does not produce significantly different results, except perhaps for the
analysis on a periodic base (5+4+3), where the Gini behaves differently than the
others in many provinces.
Fifthly, the decomposition of the concentration by monthly groupings produced
a high general explanatory capacity, especially with regard to the 5+4+3 division
and the provinces with higher concentrations (typically sun-sand-beach
territories). So, in these cases, it seems that these seasons can be used as
references for policy monitoring and evaluation.
Fifthly, the decomposition of the monthly concentration by additive factors (by
major markets) typically indicates the foreign component weighting in the level
and evolution. This increase in weighting is due to both the individual share of
demand and the increase in its own monthly concentration.
Finally, meanwhile, although typically the direct effect dominates the relative
contribution of the major markets to the global concentration, the interrelation
factor is positive, which demonstrates that the two main markets – domestic and
international–strengthen each other in the explanation of concentration, without,
for example, any generally significant evidence of the domestic market
offsetting the foreign one up to now. In fact, the seasonality of the domestic
market is even higher than that of the international market in nearly half of the
provinces evaluated, and this would also grow almost half of them.

32

Therefore in terms of policy (and academic) implications, these results throw
some basic points:
First of all, the recent quite global unbalanced evolution, especially in the last
few years, underlines the need to intensify efforts – whether this is in terms of
products, markets or prices – to attenuate this seasonality.
Second, these efforts should be primarily concentrated on the markets that
make the biggest contribution. In this respect, on a global level there is a need
to take action, especially in international markets, and detail the strategies for
each region in terms of the weighting of each of them.
Third, overall, it does not appear that the national component compensated
much for the increased seasonality characteristic of the international market. In
fact, seasonality in the domestic market is particularly high in many provinces,
even higher than the international market. In any case, given the potential for
reducing the seasonality of this market (by proximity and degree of knowledge,
for example) it would seem reasonable that specific measures should be
applied, especially in provinces with high weight and high seasonality in this
market. It must be noted that, if this differential seasonality of domestic market
is not countering an automatic policy of increasing the weight of the domestic
market, it may even be counterproductive in terms of seasonality for these
provinces.
Finally, although many of the results do not change when using different
concentration indices, the evidence that, in some cases, discrepancies appear
implies that one should move cautiously in drawing general conclusions based
on a single index. An obsessive use of the Gini coefficient, on its own, is
inadvisable.

33

References
Allock, J. (1994). Seasonality. In S.F. Witt & L. Motuinjo (Eds), Tourism
Marketing and Management Handbook (pp. 86-92). New York: Prentice Hall.
Baron, R.V. (1975). Seasonality in Tourism-A Guide to the Analysis of
Seasonality and Trends for Policy Making. Technical Series, 2, Economist
Intelligence Unit, London
Baum, T. (1999). Seasonality in Tourism: understanding the challengesintroduction. Tourism Economics, 5 (1), 5-8.
Butler, R. (1994). Seasonality in Tourism: issues and problems. In A. Seaton
(Ed.) Tourism, The State of the Art (pp. 332-340), Chichester: Wiley.
Capó, J., Riera, A. & Rosello, J. (2007). Accommodation determinants of
seasonal patterns. Annals of Tourism Research, 34, 422-436.
Chakravorty, S.R. (1990). Ethical Social Index Numbers, Berlin: Springer
Verlag,.
Cowell, F. (1995). Measuring Inequality, (2nd ed.). New York: Prentice Hall.
Cuccia, T. & Rizzo, I. (2011). Tourism seasonality in cultural destinations.
Empirical evidence from Sicily, Tourism Management, 32, 589-595.
Duro, J.A. (2012), On the automatic application of inequality indexes in the
analysis of international distribution of environmental indicators. Ecological
Economics, 76, 1-7.
Fernandez-Morales, A. (2003). Decomposing seasonal concentration. Annals of
Tourism Research, 30 (4), 942-956.
Fernandez-Morales,

A.

&

Mayorga-Toledano,

M.C.

(2008).

Seasonal

concentration of the hotel demand in Costa del Sol: A decomposition by
domesticities. Tourism Management, 29, 940-949.
Gini, C. (1912). Variabilità e mutabilità, contributo allo studio delle distribución e
relazioni statistiche. Studi Economico-Giuridici dell’ University di Cagliari. 3, part
2, 1-158.
Goerlich, F. (1998). Desigualdad, Diversidad y Convergencia: (algunos)
instrumentos de medida. monografía 1998-01, IVIE.
Koenig-Lewis, N. & Bischoff, E. (2005). Seasonality research: the state of the
art. International Journal of Tourism Research, 7, 201-219.

34

Krakover, S. (2000). Partitioning seasonal employment in the hospitality
industry. Tourism Management, 2, 461-471.
Lerman, R. & Yitzaki, S.(1985). Income inequality effects by income. The
Review of Economic and Statistics, 67, 151-156.
López, J.M. & López, L.M. (2006). La concentración estacional en las regiones
españolas desde una perspectiva de la oferta turística. Revista de Estudios
Regionales, 77, 77-104.
Lundtorp, S. (2001). Measuring tourism seasonality. In T. Baum and S.
Lundtorp (Eds.). Seasonality in tourism (pp. 23-50). Oxford: Pergamon.
Manning, R.E. & Powers, L.A. (1984). Peak and off-peak use: redistributing the
outdoor recreation/tourism load. Journal of Travel Research, 23 (2), 25-31.
Martin, J.M., Jiménez, J. & Molina, V. (2014). Impacts of seasonality on
environmental sustainability in the tourism sector based on destination type: an
application to Spain’s Andalusia region. Tourism Economics, 20(1), 123-142.
Rosello, J.A., Riera, A. & Sanso, A. (2004). The economic determinants of
Seasonal patterns. Annals of Tourism Research, 31, 697-711.
Shorrocks, A. F. (1980). The class of additively decomposable inequality
measures. Econometrica, 48(3), pp. 613-625.
Shorrocks, A. F. (1982). Inequality decomposition by factor components.
Econometrica, 50(1), pp. 193-211.
Shorrocks, A. F. (1984). Inequality decomposition by population subgroups.
Econometrica, 52(6), pp. 1369-1385.
Theil, H. (1967). Economics and Information Theory. Amsterdam: North
Holland.
Wanhill, S. (1980). Tackling seasonality: A technical note. International Journal
of Tourism Management, 1(4), 84-98.
Zaiger, D. (1983). Inequalities for the Gini coefficient of composite populations.
Journal of Mathematical Economics, 12.

35