tÍtulo de la comunicaciÓn (title): foreign direct ... · over the last forty years, foreign...
TRANSCRIPT
1
TÍTULO DE LA COMUNICACIÓN (Title): Foreign direct investment in Spain: regional distribution and determinants AUTOR 1 (Author 1): José Villaverde Email: [email protected]
AUTOR 2 (Author 2): Adolfo Maza Email: [email protected]
DEPARTAMENTO (Department): Departamento de Economía
UNIVERSIDAD (University): Universidad de Cantabria
ÁREA TEMÁTICA (Subject Area): Globalización y desarrollo regional
RESUMEN (Abstract The purpose of this paper is to empirically analyse the regional distribution of foreign direct investment (FDI) in Spain and its main determinants between 1995 and 2005/2008. By means of using different indicators, the paper reveals the main traits of FDI regional distribution in Spain, among which its highly geographical concentration in Madrid and, to a much lesser extent, in Cataluña stands out. Afterwards, the paper performs an explanatory factor analysis to reduce the vast array of FDI determinants considered in the literature to a manageable number of variables (factors). Finally, the paper considers these factors as independents variables in the FDI regression equation and estimates it in order to unveil the relevance of each factor. The main conclusion is that factors such as economic potential, labour conditions and competitiveness are important for attracting FDI; on the contrary, it seems that the market size is not relevant at all as booster of FDI.
PALABRAS CLAVE (Key words): Foreign direct investment; factor analysis; Spanish regions
2
1. Introduction
Over the last forty years, Foreign Direct Investment (FDI) has been a prominent
driver for the development of the Spanish economy.1 During the 1960’s FDI played
an important role in Spain’s transition to a much more open, market oriented
economy. Although during the 1970’s and early 1980’s this process continued at an
even higher pace, it was not until around mid eighties, as a result of the entry into the
EU and the parallel macroeconomic and political stability, that Spain became a really
highly attractive host country for FDI. Notwithstanding international comparisons
are always difficult, this attractiveness is clearly reflected in the UNCTAD’s Inward
FDI Performance Index (UNCTAD, 2002), as it happens that, relative to its (GDP)
size, the Spanish economy is, for instance, more successful than the French,
American or Italian economies in attracting FDI.
A number of studies have analysed the characteristics, dynamics, determinants and
economic implications of FDI in Spain (Bajo and Sosvilla, 1994; Martín and
Velázquez, 1997; Herce at al., 1998; Muñoz, 1999; Bajo and López, 2002; Barrios
and Strobl, 2002; Barrios et al., 2004; Bajo et al., 2007…). This strand of applied
research has mainly adopted a national perspective and paid scant or none attention
to the analysis of the spatial (notably regional) distribution of FDI, being some
exceptions Egea and López (1991), Fernández-Otheo (2000), Díaz (2002) Pelegrin
(2002,2003), Rodríguez (2005) and Rodríguez and Pallas (2008). However, as
regional economic realities in Spain are far from being homogeneous and FDI is
highly unevenly distributed across regions, an analysis of the different degrees of the
region’s attractiveness for FDI seems to be appropriate. This paper tries to add to this
literature by investigating the regional distribution of FDI in Spain and its main
determinants. The paper contributes in four ways: first and to get more insight about
the distribution than in previous papers, it employs different indicators of inward
FDI; secondly, it uses factor analysis to reduce the number of variables explaining
regional distribution of FDI to four uncorrelated composite variables or factors;
1 Although the theoretical impact of FDI on economic growth goes through various ways, the most relevant tend to be the increase in financial and physical capital and the improvement in technology and know-how. However, empirical evidence documenting these effects is not conclusive. Lim (2001) offers an interesting survey of the literature on the relations between FDI and growth.
3
thirdly, it carries out estimates of the FDI equation to derive the main drivers
(factors) of FDI not only at aggregate but also at sectoral level; and fourthly, it
explores the existence of negative spatial spillovers in the attracting factors to assess
whether regions compete for FDI.
The remainder of the paper is organised as follows. In section 2 a thorough review of
the regional distribution of FDI in Spain is carried out. Afterwards, section 3 offers a
very brief review of the literature on FDI determinants, mainly related to the Spanish
case. Drawing from this, section 4 proceeds in two steps: first, it performs an
exploratory factor analysis, and second, it specifies the model employed to uncover
FDI determinants, estimates it and presents the main results. Section 5 enlarges the
previous model by including spatial lags of FDI attraction factors and offers
estimates from this spatial model. Finally, some concluding remarks and policy
implications are offered in section 6.
2. The regional distribution of FDI in Spain
In this section the regional distribution of FDI in Spain is analysed. The sample
comprises 17 regions or autonomous communities. The data about inward FDI,
provided by the Department of Trade and Investment of the Spanish Ministry of
Industry, Tourism and Trade, refer to gross FDI. As for the rest of the variables
employed in the analysis, the main data source is Cereijo et al. (2007), although some
of them have been directly drawn from the Spanish Statistical Institute (INE)
website. For statistical reasons (availability of data) the period of analysis goes from
1995 to, depending on the specific issue at hand, 2005 or 2008. Finally, it is
important to note that all monetary variables are expressed in real terms (constant
euros of 2000).
As shown in Figure 1, FDI in Spain has increased over time although following three
clearly differentiated paths: a rapidly increasing one between 1995 and 2000; a
declining path from 2000 to 2005; and, finally, an increasing one from 2005 to 2008.
This same time paths can be appreciated when FDI is scaled with either GDP or
population.
4
Different features typify the regional distribution of FDI in Spain. In this section the
main focus is on its accumulated (absolute) value; however, it also pays some
attention to its origin and sectoral distribution. As for the accumulated value of FDI,
Table 1 unveils two main features:
- Firstly, that it is highly concentrated in just a few regions. Madrid and
Cataluña received, on average, 79% of total FDI, although being the volume of the
first four and a half times bigger than that of the second. Should we add the FDI
received by the País Vasco and Comunidad Valenciana (4.5 and 4.3%, respectively)
the total received by these four regions would amount to nearly 90% of total inward
FDI in Spain.
- And secondly, that there is a remarkable persistence in the ranking of FDI
regional distribution. Madrid, Cataluña, País Vasco and Comunidad Valenciana
nearly always qualified as the firsts in the ranking,2 while Cantabria, Extremadura
and La Rioja are always the last regions in the ranking.
Although the information given in Table 1 is very interesting, it does not actually
offer any clue about how well a region is performing relative to the others. The
UNCTAD’s World Investment Report for 2002 introduced two indices to benchmark
success in attracting FDI: the Inward FDI Performance Index and the Inward FDI
Potential Index. The Performance Index, or location index, simply consists in scaling
FDI inflows by GDP;3 according to the information displayed in the first three
columns of Table 2 for the period 1995-2005, only Madrid has a Performance Index
consistently greater than one. Generally speaking, the results provided by this table
are in line with those previously offered: Madrid gets a share of Spanish FDI much
greater (three and a half times) than its share of GDP, while Cataluña and País Vasco
(the second and third in the previous ranking) get a bit less than expected according
2 The third position in the ranking for Aragon in the period 2001-2005 is due to the high value obtained in 2003 in the automotive sector. 3 The index (IFDIPI) is given by the expression
17
1
17
1
)(
)(i
i ii
i
i ii
GDPGDP
FDIFDIIFDIPI
5
to their relative GDP, but much more than the rest of regions.4 In addition, it is also
convenient to note that there is no apparent geographical pattern to the regions with
high/low Performance Index. To confirm this, we compute the two most commonly
used spatial dependence indicators: Moran’s I and Geary’s C statistics (Moran 1948;
Geary 1954). The results obtained for both statistics, using the inverse of the
standardized distance as a distance matrix, clearly reveal there is no positive spatial
dependence in attracting FDI between Spanish regions; to be precise, the Moran’s I
statistic indicates the presence of negative spatial dependence whereas Geary’s C
shows no dependence at all.
While the Performance Index is appealing and highly used in applied research for its
simplicity and easiness of computation, UNCTAD also proposed to use the Potential
Index, which is constructed by taking into consideration more scaling variables than
the mere size (GDP) of the economy under analysis. Drawing from UNCTAD (2001)
we have constructed our own Inward FDI Potential Index5 for the Spanish regions by
using the following variables: the growth rate of GDP, per capita GDP, R&D
expenditures as percentage of GDP, exports plus imports as percentage of GDP,
average number of years in school, internet users as percentage of total population,
unit labour costs and public capital over total population. The results obtained,
displayed in the last three columns of Table 2, offer a somewhat different view to
that shown by the Performance Index. In particular two results stand out: first,
although Madrid continues to be a privileged location for FDI, Navarra and the País
Vasco slightly but consistently outweigh Madrid; and second, the regions with the
poorest Performance and Potential indices do not match completely, with the cases
of Cantabria and Baleares as perhaps the most prominent cases.
The comparison of the regional ranking on the two inward FDI indices yields a 4*4
matrix showing whether the regions can be considered as front-runners, above
4 The case of Aragón should be considered, once again, as an exception. 5 The index for a region i is computed as the simple average of the scores on the chosen variables for that region. The score for each variable is given by the expression: Score= (Vi-Vmin)/(Vmax-Vmin) where Vi refers to the value of the variable for region i and Vmin and Vmax refer to the lowest and highest values of the variable among the regions.
6
potential economies, below potential economies or under performers.6 As can be
seen in Table 3, in 1995-2005 there are 4 front-runners among which Madrid
distinctively stands out. The group of above-potential economies also includes four
regions, among which the two archipelagos. The group of below-potential economies
comprises four regions as well, with Navarra as having the worst performance
relative to its potential. Finally, some of the Spanish poorest regions (Andalucía,
Castilla-León, Castilla-La Mancha, Extremadura and Galicia) are clearly under-
performers. When the two subperiods (1995-2000 and 2001-2005) are considered, it
is shown that, for good or worse, some regions consistently stay in their cells while
some others, among which Murcia is the most relevant example (from being initially
an under-performer it became a front-runner), have experienced important changes.
Another important feature of the FDI flows to the Spanish regions is that they mostly
have their origin in the EU15. This implies that the Spanish regions, as for the
Italians (Iammarino and Santangelo, 2000), have been more attractive for EU
investors than for non-EU investors, this not only suggesting the presence of an “EU
integration effect” but also that in this regional scheme arrangement FDI and trade
have been more complements than substitutes.7 With an average close to 89% of the
total between 1995 and 2008 (Table 4), this EU integration effect, however, has not
been equally distributed across regions as in some of them (namely Asturias, La
Rioja and Castilla-La Mancha) the EU share is above 95% while in others
(Extremadura and Andalucía) does not reach 60%. Despite this, no clear geographic
and/or economic pattern seems to exist with relation to the regional weight of FDI by
area of origin. Additionally it is worth noticing that for some (mostly small) regions
the share of European FDI is rather unstable.
Spanish regions do not only differ in the amount of FDI they attract but also in the
type of FDI they attract. Therefore, an analysis of the sectoral distribution of FDI in
the Spanish regions also offers some new interesting insights (Table 5). For this, we
consider just 6 sectors: primary sector, mining and quarry, manufacturing, energy,
6 The dividing value is always the mid-point of the ranking, which, for easy of representation we include in the low group. 7 Blonigen (2001) points out that regional trade arrangements can increase FDI when FDI and trade are complements, but decrease FDI when they are substitutes.
7
construction, and services; additionally, following OECD taxonomy based on
technology intensity we divide manufacturing into 3 branches and also split services
into 7 branches. For the period 1995-2008 four main conclusions arise:
- First of all, FDI is mainly concentrated in manufacturing and services, as
these two sectors represent, on average, 83.8% of total FDI.
- Although the coefficient of variation shows that regional disparities in
these two sectors are much lower than in the others, it is convenient to note that they
are remarkable. For instance, for manufacturing it happens that five regions have
shares over 70% while three regions are below 20%, with Baleares standing out with
even less than 1%. On the contrary, in the services sector Baleares keeps the highest
share (close to 70%), while in the opposite side Asturias and Extremadura have a
share well below 10%.
- Regarding the manufacturing sector, most of FDI (75.8%) is concentrated
in medium-technology industries, while low-technology and high-technology
industries account for, respectively, 20.2 and 4% of the sector’s total. From a
regional perspective differences are rather important in medium-technology
industries although it happens that both low and high-technology industries
distributions are much less evenly distributed (see the last row of the table).
Specifically, FDI in high-technology industries is quite important in Cantabria,
Castilla y León and Navarra but is completely irrelevant in regions such as Canarias,
Asturias, Extremadura, Galicia and Murcia.
- With respect to the services sector, wholesale and retail trade (with a
weight of 39.3%) and communications (with a share of 24.1%) are the most relevant
branches for FDI. Once again, the regional distributions diverge markedly for the
different branches, although to a lesser extent than in manufacturing; according to
the value of the coefficient of variation, the more homogeneous distribution occurs in
communications while the less homogeneous happens in hotels and restaurants.
8
A complementary, yet synthetic perspective on the sectoral distribution of FDI in
Spanish regions can be obtained by computing an indicator of specialisation. For a
region r, the so-called Entropy index of specialisation (EIS) is defined as:
riri rir XXXXEIS ln
where i refer to sector/branch8 and X is the FDI value. The index ranges between 0,
when just one sector/branch concentrates all FDI, and ln(n) when FDI is equally
distributed in all n sectors/branches; in this case ln(14)=2.69. According to the results
shown in Table 6 for the whole sample period, Aragón, Asturias and Murcia are the
regions in which the sectoral distribution of FDI is more specialised, while regions
such as Andalucía, Canarias, Cataluña and Madrid, with EIS over 2, are the less
specialised; put it in a different way, these last four regions are the ones in which the
sectoral distribution of FDI is more symmetrical. As for the change in the degree of
specialisation, ten out of seventeen regions have decreased the value of the indicator,
which implies the FDI sectoral distribution became less symmetrical in the final sub-
period of the sample than in the first. In particular, the major changes have taken
place in Canarias and Cantabria in a more specialised direction and in Asturias and
La Rioja in a less specialised direction.
3. Determinants of the FDI regional distribution in Spain: a brief review
Following the tremendous growth of FDI in the world in the last decades, the
analysis of FDI determinants turned out to be one of the topics which has attracted
more and even increasing attention in the economic literature (a comprehensive
review on this issue can be found in Blonigen, 2005). Yet, since our paper only
examines the regional location of FDI in Spain, it does not need to pay any heed to
classical determinants such as the exchange rate, taxes, institutions, trade effects,
political risks, etc., as the values of all of them are considered to be the same across
the Spanish regions. On the contrary, the paper focuses on factors helping to
understand why the FDI is located in a specific region (say Madrid) but not in other
region (say Extremadura): that is, why a particular region within Spain is chosen for
8 As shown in Table 5, we are considering 4 sectors (primary sector, mining and quarry, energy and construction) and 3 branches for manufacturing and 7 for services.
9
the location of FDI. The best summary of the literature on this issue lies in what has
been dubbed as the “eclectic paradigm”, which, in one of its three “legs”, 9 is a
mixture of the agglomeration and comparative advantage approaches to FDI:
according to the first one, agglomeration economies -mainly driven by knowledge
and pecuniary spillovers- are the key determinants of FDI, whereas from the point of
view of the neoclassical, comparative advantage approach, FDI is driven by
differences in input (labour, transport, energy, etc.) costs and inputs availability.
As mention in the Introduction, a great number of studies have empirically examined
the issue of the determinants of the location of FDI at international and national
levels, but the empirical evidence on the FDI location decisions within a country is
much less abundant (Mullen and Williams, 2005). As for Spain, the situation is
roughly the same; there are quite a few papers analysing the issue at the national
level (see, again, the references in the Introduction), but just a small number devoted
to the study of FDI at regional level.
In particular, the papers by Egea and López (1991), Díaz (2002) and Rodríguez
(2005) offer a rather precise description of this regional distribution, with the main
message of FDI being highly concentrated in just a few regions, while those of
Pelegrin (2002, 2003) and Rodríguez and Pallas (2008) add an econometric analysis
of the FDI determinants. Pelegrin (2002), by using different estimation
methodologies (OLS, LS with fixed effects and GLS), estimates for the period 1993-
1998 a FDI equation where the dependent variable is per capita gross effective
foreign investment. She identifies three key determinants of the FDI location, namely
market size, human capital and public incentives; however, infrastructures were not
found significant while labour costs presented a (puzzling?) positive coefficient. In
another paper, Pelegrin (2003) extends the sample period by two years (to 2000) and,
paying particular attention to agglomeration factors, concludes that “manufacturing
agglomeration, concentration of R&D activities and the availability of skilled labour
9 The eclectic paradigm (Dunning, 1980, 1988, 1998, 2001) considers that the pattern of FDI can be explained in terms of three legs: “ownership-specific advantages”, “internalization advantages” and “location advantages”. The location advantages, which are specific to an area (country, region, province), are mainly reflected in agglomeration and comparative advantage variables. For a precise, clear and enlightening summary of this paradigm see Chapter IV of UNCTAD’s World Investment Report (1998).
10
are important determinants of manufacturing foreign direct investment, but
congestion costs sometimes act as a centrifugal force, leading to the rejection of
foreign investment” in the manufacturing sector (Pelegrin, 2003, p. 23).
Considering also gross effective FDI as the dependent variable of the model,
Rodríguez and Pallas (2008) analyse its determinants in Spanish regions for the
period 1993-2002. The model estimated, using both GLS (cross-section weights)
and instrumental variables, concludes, mostly in line with Pelegrin (2002 and 2003),
that differences in regional unit labour costs, the volume of FDI at the start of the
period under analysis, GDP and human capital are key factors in explaining the
regional distribution of FDI in Spain.
4. Determinants of the FDI regional distribution in Spain: an empirical analysis
Although the aforementioned papers are very interesting, empirical evidence on what
explains the various fortunes in the FDI attractiveness across Spanish regions is still
incomplete. For this reason, this section tries to present a more comprehensive
approach to the issue and offer new valuable insights; for statistical reasons we just
confine our analysis to the sample period 1995-2005. Drawing from the literature on
FDI determinants previously surveyed and from data availability, we have initially
chosen a set of 16 variables (see Annex) to explain regional FDI patterns in Spain.
As can be seen in Table 7, descriptive statistics of these variables reflect the fact that,
on average, for some of them (GDP, population, PTK and TTK), there are huge
regional disparities while, for the others, regional differences tend to be rather low;
additionally, it is shown that in most cases (11 out of 16) the distribution is skewed to
the right and presents excess kurtosis (values over 3).
Anyway, as such a large number of variables is likely to cause problems with the
regression analysis due to the presence of some collinearity across them, we remove
redundant variables by performing a standard exploratory factor analysis. However,
before doing that we compute the correlation between each pair of variables (Table
8), this showing that all of them have at least one correlation coefficient over 0.5; as
it also happens that the determinant of the correlation matrix is null, the conclusion is
that using factor analysis in this case is quite appropriate. Using this process of data
11
reduction a new set of orthogonal and non-collinear variables (named factors) is
obtained. In particular, by using principal component analysis and following the
Kaiser’s criterion for factor extraction, the analysis identifies four “unobservable
variables” or factors with eigenvalues greater that 1 (Table 9), explaining 87.3% of
the cumulative variance of the original 16 variables, a proportion that is quite
satisfactory (Hair et al., 2006). The composition of these four factors is shown in
Table 10. As for the economic interpretation of these factors, the first one (F1),
explaining more than 47.9% of the entire variance, is labelled economic potential as
is a combination of labour productivity, total and private technological capital, R&D
investment, human capital, internet users and per capita GDP. The second factor
(F2), comprising variables such as activity, employment and occupation rates plus
the inverse of unit labour costs, is called labour conditions. The third factor (F3) is
made up of GDP and total population so is labelled market size. Finally, the fourth
factor (F4) contains three indicators (roads infrastructure, economic structure and
openness degree) and is clearly related to competitiveness.
The four-factor solution previously mentioned has three additional virtues (see also
Table 10): firstly, almost all original variables are highly correlated with just one
factor and quite weakly with the others; secondly, all variables have at least one
factor loading greater in absolute value than 0.5, which experts regard to be very
significant; and thirdly, the reliability of the extracted factor structure is patent as it
explains, but for the variable “inverse of unit labour costs”, between 80 and 96.5% of
the variance of each original variable.
Considering the four aforementioned extracted factors as our independent variables,
we apply panel data techniques to estimate an equation in which the dependent
variable is gross effective FDI, expressed as percentage of GDP or in per capita
terms; the panel dataset is then made up of 17 regions from 1995 to 2005, which total
187 observations. As in many other cases, a log-linear form was employed to
transform the unknown relationship between the dependent and independent
variables into a linear relationship.
Before estimating our FDI equation, however, we have tested for the existence of
panel unit roots and panel cointegration. To do that, we implement the ADF test, the
12
results showing that all series are I(1) and cointegrated in the long run. Then, for
model selection and to check the existence of individual effects in this equation, we
perform the Hausman test for the presence of fixed versus random effects. The
results reveal the model should be estimated with fixed effects, which is confirmed
by a classical fixed effects Chow test; by including fixed effects, we minimize the
potential burden of omitted variables in the estimates. In consequence, the equation
proposed is as follows:
itititititi
it
FFFFPOPGDP
FDI
4321
)( 4321 (1)
In order to decide between least or generalised least squares to estimate the FDI
equation, we have computed the fairly general test of Breusch-Pagan for
heteroscedasticiy; the results show that the null hypothesis of homoscedasticity can
be rejected at a significance level of .05. Therefore, and according to the results of all
tests computed, the estimation of the model is finally performed by Generalized
Least-Squares (GLS) with fixed effects.
Table 11 (first half) presents the regression results for total FDI.10 As expected, FDI
(scaled either by GDP or population) is positively and significantly correlated to the
economic potential, labour conditions and competitiveness of the regions. As for the
market size variable, economic analysis suggests that its relationship to FDI should
be positive; however, the results show that in our case the relationship is negative but
statistically non-significant, a conclusion that, in our opinion, could be explained by
the fact that, once a multinational corporation has decided to invest in Spain, the size
of the local (regional) market in which the investment effectively takes place is of no
relevance at all. Finally, it is worth mentioning that regional fixed effects are
statistically significant in all cases.
This analysis has also been carried out for the manufacturing and services sectors, as
it was shown that most of the FDI was concentrated in them. Additionally,
estimations for the main branches of these two sectors have been performed. Before
10 We have also checked for the presence of spatial dependence in the residuals of our FDI equation. The results for both Moran’s I and Geary’s C statistics clearly indicate there is no spatial dependence.
13
presenting the results, we want to point out that, considering the similarity of them to
those of the aggregate equation, now we have just estimated the equation by scaling
FDI with GDP; additionally, given the lack of data for some branches in some
specific years, an unbalanced panel has been estimated in these cases. The results
obtained (first half of Table 12) confirm that market size is non-significant whatever
the sector/branch considered. Additionally, these results also show that, for both total
and medium technology manufacturing activities, labour conditions and
competitiveness are the main factors explaining FDI location, while economic
potential now becomes non-significant. As for the services sector on the whole and
its communications and business services branches, economic potential and
competitiveness are positively and significantly related to FDI, whereas labour
conditions turn out to be non-significant. Surprisingly, it is worth noticing that none
of the four extracted factors are statistically significant for the wholesale and retail
trade branch. This result requires further investigation and, thus, opens a clear
avenue for future research based on the study of wholesale and retail trade FDI
location and the role of different factors as FDI attractors and hence regional
boosters.
Finally, in view of the potential endogeneity of some of the independent variables,
we think of interest to address this issue in our estimations. Although it is argued that
the approach of generalized method of moments (GMM) deals consistently and
efficiently with endogeneity problems, this procedure critically hinges on the
assumption of no second-order serial correlation in errors of the level equations; in
this case, the AR(2) tests suggest that, generally speaking, our equations do not
satisfy such a necessary condition and, therefore, we do not employ this approach.
Thus, we have opted to solve the problem by estimating our equations by Two-stage
GLS; in so doing, we use lags of the independent variables as instruments, showing
the Sargan test that the null hypothesis of validity of instrumental variables can not
be rejected at 95%.
The results are shown in the second half of Tables 11 (for the aggregate) and 12 (for
main sectors and branches). Regarding aggregate FDI, the main difference with GLS
estimates refers to the role played by labour conditions, which now seem to be non-
14
significant. As for the main sectors and branches the results tend to confirm those
obtained with GLS. However, there are some minor differences related to the roles of
competitiveness and economic potential in explaining, respectively, FDI in medium
technology intensity manufacturing and in services.
5. Spatial dependence
As stated in Section 2, there is no spatial dependence in the regional distribution of
FDI in Spain. In other words, regions with high (low) shares of FDI do not tend to be
geographically concentrated. This being so, in this section we take another
complementary view. Having uncovered the main factors behind the regional
distribution of FDI, the aim of this section is to detect whether regions compete for
attracting FDI; that is, whether the improvement of an FDI attraction factor in a
given region may bring forth a fall of FDI in neighbouring regions.
To do that, we estimate an enlarged version of equation (1) in which the spatial lags
of the three statistically significant attracting factors -considered as a spatial
weighted average of them at neighbouring locations- have been included as
independent variables (Rey and Montouri, 1999; Lim, 2003). These spatial lags are
computed by multiplying each factor F by a distance matrix W which role is to put
less weight on observations that are further from the region considered; this distance
matrix is the inverse of the standardized geographical distance and, more precisely, is
the inverse of the great-circle distance (the shorter distance between any two points
on the surface of a sphere) between regional capitals.
The estimation results -for the sake of simplicity we only present those obtained by
GLS in the case of aggregate FDI scaled by GDP- are reported in Table 13. Two
main conclusions can be drawn. First of all, regarding the influence of the four
original determining factors on FDI it is important to note that the results are not
substantially different to the previous ones. Second, the inclusion of the factors
spatial lags reveals negative and significant geographical spillovers associated to
economic potential and competitiveness; this reflects that regions do compete for FDI
flows and that an improvement in any one of these two factors in neighbouring
15
regions would decrease the flow of FDI in the considered region. This is not the case,
however, for the factor labelled as labour conditions.
6. Concluding remarks
Although inward FDI has increased tremendously in Spain in the last thirty years,
large differences in FDI patterns across Spanish regions tend to persist. This paper
analyses this issue from 1995 onwards and tries to offer new insights with relation to
the regional distribution of FDI in Spain and its determining factors.
Concerning the regional distribution of FDI, the paper employs different indicators
and concludes that: 1. FDI in Spain has been highly concentrated in just two regions,
namely Madrid and Cataluña. 2. This being the case, the so-called FDI Potential
Index shows that two other regions, Navarra and País Vasco, seem to be even better
positioned to receiving FDI than Madrid and Cataluña. 3. By comparing the
Performance and Potential FDI indices, it is concluded that Madrid, Cataluña,
Aragón and País Vasco are considered as front-runners whereas Andalucia, Castilla
y León, Castilla-La Mancha, Extremadura and Galicia (some of the poorest regions
in Spain) are labelled as under-performers. 4. By area of origin, most inward FDI in
Spain comes from the EU, although with striking regional differences. 5. From a
sectoral perspective, it has been shown that, although manufacturing (mainly medium
technology intensity industries) and services (above all wholesale and retail trade,
and communications) concentrate most of FDI, there are important regional
differences as to the weight of these sectors (branches); in particular, FDI is much
more sectoral concentrated in regions such as Asturias, Murcia and Aragón than in
Canarias, Cataluña and Andalucía.
The second part of the paper is devoted to the analysis of the main determinants of
FDI in the Spanish regions. After briefly reviewing the literature on the issue, and
being conscious about the large number of variables potentially affecting FDI, we
perform an exploratory factor analysis to reduce independent variables in the FDI
equation to a manageable number of extracted factors; these are labelled as economic
potential, labour conditions, market size, and competitiveness. The estimation of the
FDI equation using these four factors as independent variables has shown that, at the
16
aggregate level, economic potential, labour conditions and competitiveness are
important for attracting FDI; on the contrary, it seems that market size is not relevant
at all. As for the main sectors and branches, the estimates tend to confirm the
aggregate results but for three exceptions: economic potential becomes non-
significant for manufacturing; the same occurs with labour conditions in services;
finally, no one of the four extracted factors is significant as FDI determinant for the
wholesale and retail trade. Back to the aggregate level, it is also important to note
that the inclusion of spatial lags in the exogenous variables of the FDI equation
reveals the presence of negative geographical spillovers associated to the economic
potential and competitiveness factors.
To conclude with, and assuming that FDI enhances economic growth, the results
obtained in this paper point out to policies promoting the economic potential and
competitiveness of the regions. Accordingly, focusing on the use of public and
private technological capital, R&D investments, strategic roads infrastructure
projects and trade openness degree may be some of the most efficient ways to attract
FDI. Finally, although labour conditions also emerge as an FDI booster factor, and
the need of a labour reform in Spain is quite evident, it is considered that this reform
would affect in a similar way to all Spanish regions and, as such, would not have a
significant impact on the regional location of FDI.
References
Bajo, O., Díaz-Mora, C. and Díaz-Roldán, C. (2007) Foreign direct investment and regional growth: an analysis of the Spanish case. IEF, Papeles de Trabajo, No. 17.
Bajo, O. and López, C. (2002), “Foreign direct investment in a process of economic integration: the case of Spanish manufacturing, 1986-1992”, Journal of Economic Integration, 17, 85-103.
Bajo, O. and Sosvilla, S. (1994), An econometric analysis of foreign direct investment in Spain, 1964-1989, Southern Economic Journal, 61 (4), 104-120.
Barrios, S. and Strobl, E. (2002), Foreign direct investment and productivity spillovers: Evidence from the Spanish experience, Review of World Economics, 138 (3), 459-481.
17
Barrios, S., Dimelis, S., Louri, H. and Strobl, E. (2004) Efficiency spillovers from foreign direct investment in the EU periphery: A comparative study of Greece, Ireland, and Spain, Review of World Economics, 140 (4), 688-705.
Blonigen, B.A. (2001), In search of substitution between foreign production and exports, Journal of International Economics, 53 (1), 81-104.
Blonigen, B.A. (2005), A review of the empirical literature on FDI determinants, Atlantic Economic Journal, 33, pp. 383-403.
Cereijo, E., Turrión, J. and Velázquez. F. (2007), Indicadores de convergencia real para las regiones españolas. Estudios de la Fundación FUNCAS, No. 23.
Díaz, R. (2002) Un estudio descriptivo de la inversión extranjera directa en España y su distribución territorial, Cuadernos de Economía, 25(70), pp. 277-301
Dunning, J.H. (1980), Toward an eclectic theory of internal production: some empirical tests, Journal of International Studies, 11, pp.9-31.
Dunning, J.H. (1988), The eclectic paradigm of international production: a restatement and possible explanations, Journal of International Business Studies, 19 (1), pp. 1-31.
Dunning, J.H. (1998), Location and multinational enterprise: a neglected factor?, Journal of International Business Studies, 39 (1), pp. 45-65.
Dunning, J.H. (2001), The eclectic (OLI) paradigm of international production: past, present and future, International Journal of Economics and Business, 8(2), pp. 173-190.
Egea, M.P. and López, C. (1991), Un estudio sobre la distribución geográfica de la inversión extranjera directa en España (1986-1989), Información Comercial Española, 696/697, 105-118
Fernández-Otheo, C.M. (2000), Concentración y especialización regional de la inversión directa extranjera en España, Economía Industrial, 335/336, 67-82.
Geary, R.C. (1954), The contiguity ratio and statistical mapping, Incorporated Statistician, 5, 115-145.
Hair, J.F., Black, B., Babin, B., Anderson, R.E. and Tatham, R.L. (2006) Multivariate Data Analysis, Prentice-Hall International.
Herce, J.A., Jimeno, J.F. and Sosvilla, S. (1998), Flujos de capital e integración financiera: El caso de España, Fundación de Estudios de Economía Aplicada, Madrid.
Hu, A.G. and Owen, R.F. (2005), Gravitation at home and abroad: Regional distribution of FDI in China, mimeo.
18
Iammarino, S. and Santangelo, G.D. (2000), Foreign direct investment and regional attractiveness in the EU integration process: Some evidence for the Italian regions, European Urban and Regional Studies, 7 (1), 5-18.
Lim, E. (2001), Determinants of, and the relation between foreign direct investment and growth: A summary of the recent literature. IMF Working Paper WP/01/175.
Lim, U (2003) A spatial analysis of regional income convergence. Planning Forum, 9, 66-80.
Martín, C. and Velázquez, F.J. (1997), The determining factors of foreign direct investment in Spain and the rest of the OCDE: Lessons for the CEECS, Discussion Paper q637, CEPR.
Moran, P.A.P. (1948), The interpretation of statistical maps, Journal of the Royal Statistical Society Ser. B, 10, 243-251.
Mullen, J.K. and Williams, M. (2005), Foreign direct investment and regional economic performance, KYKLOS, 58 (2), 265-282.
Muñoz, M. (1999), La inversion extranjera directa en España: Factores determinantes, Civitas, Madrid.
Pelegrín, A. (2002), Inversión extranjera directa. Factores determinantes de la localización regional, Papeles de Economía Española, 93, 122-133.
Pelegrín, A. (2003), Regional distribution of foreign manufacturing investment in Spain. Do agglomeration economies matter?, Documento de Trabajo 2003/6, Institut d’Economia de Barcelona.
Rey, S. and Montouri, B. (1999) U.S. regional income convergence: a spatial econometric perspective. Regional Studies, 33, 143-156.
Rodríguez, C. (2005), Un análisis comparado de la inversión extranjera directa efectiva y potencial de las regiones españolas, Revista de Estudios Regionales, II(73), pp. 13-41.
Rodríguez, X.A. and Pallas, J. (2008), Determinants of foreign direct investment in Spain, Applied Economics, 40 (19) pp. 2443-2450.
UNCTAD (1998), World Investment Report, 1998. United Nations, New York and Geneva
UNCTAD (2001), World Investment Report, 2001. United Nations, New York and Geneva
UNCTAD (2002), World Investment Report, 2002. United Nations, New York and Geneva
19
Figure 1. Inward FDI in Spain: 1995-2008
0
1
2
3
4
5
6
7
8
9
10
1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008
%
0
5000
10000
15000
20000
25000
30000
35000
40000
Eu
ros
2000
(mil
lion
s)
FDI/GDP FDI/POP FDI
Note: FDI is measured on the right hand side; FDI/GDP and FDI/POP are measured on the left hand side.
20
Table 1. Accumulated Inward FDI in Spanish regions
1995-2000 2001-2005 2006-2008 1995-2008 1995-2000
2001-2005
2006-2008
1995-2008
Andalucía 1775622 1676298 1181896 4633815 2.1 1.6 1.9 1.9 Aragón 635179 5727866 418353 6781398 0.8 5.5 0.7 2.7 Asturias 1083630 2160990 75748 3320368 1.3 2.1 0.1 1.3 Baleares 980052 877198 588858 2446108 1.2 0.8 1.0 1.0 Canarias 3344542 2839570 721985 6906097 4.0 2.7 1.2 2.8 Cantabria 40940 39728 98933 179601 0.0 0.0 0.2 0.1 Castilla y León 239838 457613 98096 795546 0.3 0.4 0.2 0.3 Castilla-La Mancha 136424 431679 279540 847642 0.2 0.4 0.5 0.3 Cataluña 13885630 16331440 6826953 37044023 16.5 15.7 11.1 14.8 C. Valenciana 1979831 5253765 3554126 10787722 2.3 5.0 5.8 4.3 Extremadura 102280 135375 9170 246825 0.1 0.1 0.0 0.1 Galicia 265345 1448782 127016 1841143 0.3 1.4 0.2 0.7 Madrid 55158628 61419866 43890388 160468882 65.5 58.9 71.3 64.2 Murcia 207221 1090581 251901 1549703 0.2 1.0 0.4 0.6 Navarra 461914 132594 198986 793494 0.5 0.1 0.3 0.3 País Vasco 3882154 4112360 3224183 11218697 4.6 3.9 5.2 4.5 Rioja (La) 94935 106525 2739 204198 0.1 0.1 0.0 0.1 Spain 84274165 104242230 61548869 250065264 100.0 100.0 100.0 100.0
21
Table 2. Inward FDI Performance and Potential Indices, 1995-2005
FDI Performance Index FDI Potential Index 1995-2000 2001-2005 1995-2005 1995-2000 2001-2005 1995-2005 Andalucía 0.157 0.118 0.135 0.336 0.338 0.337 Aragón 0.234 1.771 1.069 0.543 0.588 0.563 Asturias 0.559 0.955 0.770 0.388 0.405 0.396 Baleares 0.477 0.333 0.397 0.284 0.236 0.262 Canarias 1.009 0.667 0.818 0.365 0.336 0.352 Cantabria 0.039 0.030 0.034 0.411 0.413 0.412 Castilla y León 0.049 0.081 0.066 0.372 0.447 0.406 Castilla-La Mancha 0.046 0.122 0.088 0.337 0.387 0.360 Cataluña 0.867 0.830 0.846 0.545 0.590 0.566 C. Valenciana 0.244 0.515 0.396 0.409 0.392 0.401 Extremadura 0.071 0.078 0.075 0.320 0.347 0.332 Galicia 0.058 0.271 0.172 0.364 0.404 0.382 Madrid 3.795 3.316 3.532 0.615 0.625 0.620 Murcia 0.104 0.416 0.282 0.406 0.441 0.422 Navarra 0.317 0.074 0.183 0.622 0.684 0.651 País Vasco 0.729 0.636 0.678 0.638 0.658 0.647 Rioja (La) 0.147 0.137 0.141 0.439 0.442 0.441
22
Table 3. Region Classification by Inward FDI Performance and Potential Indices
1995-2005
Performance High Low
Potential
High
Front-runners Aragón Cataluña Madrid País Vasco
Below potential Cantabria Murcia Navarra Rioja (La)
Low
Above potential Asturias Baleares Canarias C. Valenciana
Under-performers Andalucía Castilla y León Castilla-La Mancha Extremadura Galicia
1995-2000
Performance High Low
Potential
High
Front-runners Cataluña C. Valenciana Madrid Navarra País Vasco
Below potential Aragón Cantabria Rioja (La)
Low
Above potential Asturias Baleares Canarias
Under-performers Andalucía Castilla y León Castilla-La Mancha Extremadura Galicia Murcia
2001-2005
Performance High Low
Potential
High
Front-runners Aragón Cataluña Madrid Murcia País Vasco
Below potential Castilla y León Navarra Rioja (La)
Low
Above potential Asturias Canarias C. Valenciana
Under-performers Andalucía Baleares Cantabria Castilla-La Mancha Extremadura Galicia
23
Table 4. Inward FDI from UE15 (%)
1995-00 2001-05 2006-08 1995-08 Andalucía 62.3 68.0 66.3 65.5 Aragón 74.8 92.6 95.1 85.4 Asturias 98.8 98.9 84.6 98.7 Baleares 72.1 73.4 75.1 73.5 Canarias 89.5 66.7 83.0 79.6 Cantabria 57.8 40.8 84.6 71.0 Castilla y León 89.4 98.8 60.3 91.2 Castilla-La Mancha 91.1 95.6 96.2 95.1 Cataluña 78.9 79.7 82.8 80.2 Comunidad Valenciana 89.9 96.8 96.1 95.1 Extremadura 87.2 33.6 26.8 52.6 Galicia 71.2 95.5 36.8 87.6 Madrid 88.7 87.0 96.2 91.6 Murcia 75.8 96.1 94.8 93.7 Navarra 61.8 82.8 83.9 71.1 País Vasco 89.0 90.3 88.4 89.1 Rioja (La) 95.1 98.5 83.8 96.7 Spain 85.7 86.2 93.5 88.7 Coef. Variation 0.15 0.23 0.22 0.15
24
Table 5. Sectoral distribution of Inward FDI across regions (1995-2008 average)
Primar
y sector
Mining
and
quarry
Manufacturing
Energy
Construction
Services
Low-
technology
intensity
Medium-
technology
intensity
High-
technology
intensity
Total
Whole
sale
and
retail
Trade
Transpo
rt
Hotels and
restaurant
s
Commun
i-cations
Financial
intermediatio
n
Business
services
Other
service
s
Total
Andalucía 3.40 5.41 50.47 44.21 5.32 15.81 0.85 20.04 32.94 0.11 6.07 3.31 16.16 37.93 3.48 54.48 Aragón 0.21 0.00 3.04 95.68 1.27 68.37 1.93 0.19 82.51 0.10 0.14 1.13 5.18 2.14 8.80 29.30 Asturias 0.00 0.00 0.62 99.24 0.15 37.31 59.09 0.57 13.41 0.66 1.61 4.02 33.19 46.54 0.58 3.02 Baleares 0.38 0.00 65.88 32.95 1.17 0.81 0.18 28.67 3.54 3.67 30.80 1.36 21.10 32.83 6.70 69.96 Canarias 2.56 4.77 22.49 77.51 0.00 30.72 7.79 2.61 19.55 2.22 9.83 35.40 6.73 12.31 13.95 51.55 Cantabria 2.60 0.00 4.56 41.75 53.69 49.12 0.00 2.10 1.58 13.27 1.75 8.06 67.45 7.02 0.87 46.17 Castilla y León 2.68 5.53 12.66 68.12 19.23 18.58 0.28 12.23 63.49 0.33 0.08 1.70 0.32 31.43 2.64 60.70 Castilla-La Mancha 1.30 0.51 31.78 57.68 10.54 28.86 0.42 0.53 61.03 0.80 1.09 0.38 28.53 7.27 0.89 68.37 Cataluña 0.32 0.43 32.53 61.77 5.70 42.93 3.17 3.62 24.65 6.48 6.73 19.39 14.25 23.90 4.60 49.53 C. Valenciana 0.19 0.49 2.69 96.10 1.21 72.54 0.03 1.85 9.65 0.88 1.27 3.13 28.53 55.00 1.55 24.89 Extremadura 5.28 14.60 64.20 35.37 0.43 71.41 0.08 0.47 33.81 0.69 4.65 0.04 43.92 14.90 1.97 8.15 Galicia 1.30 0.43 4.67 94.84 0.49 52.24 1.07 6.36 6.55 3.75 1.44 16.32 64.30 7.14 0.50 38.60 Madrid 0.27 0.76 22.08 73.43 4.48 22.90 15.46 2.04 43.60 2.11 1.50 27.25 12.00 11.33 2.22 58.58 Murcia 1.11 0.00 3.87 96.04 0.10 79.03 1.09 5.44 8.53 5.25 3.09 4.27 55.24 23.54 0.08 13.32 Navarra 1.41 0.00 43.65 41.20 15.15 70.73 0.17 2.91 6.93 0.57 16.05 14.88 36.06 24.77 0.74 24.78 País Vasco 0.10 0.09 14.40 82.51 3.08 58.72 2.60 12.42 46.30 4.56 0.89 19.98 6.71 19.01 2.56 26.07 Rioja (La) 0.49 0.00 78.32 19.55 2.13 72.03 0.12 0.07 1.17 0.00 0.10 0.01 71.14 27.57 0.01 27.28 Spain 0.41 0.84 20.18 75.85 3.96 31.77 11.56 3.35 39.32 2.71 2.89 24.11 12.99 14.98 3.01 52.06 Coef. Variation 3.56 4.55 1.25 0.35 3.33 0.76 1.24 2.39 0.63 1.26 2.72 0.45 1.78 0.99 1.23 0.40
25
Table 6. Entropy index of specialisation
95-00 01-05 06-08 95-08 Change 95-00/06-08 (%)
Andalucía 2.25 2.04 1.80 2.19 -19.84 Aragón 2.02 0.69 1.66 1.08 -17.90 Asturias 0.29 0.52 1.47 0.88 414.55 Baleares 1.84 1.45 1.56 1.76 -14.94 Canarias 1.93 2.20 1.02 2.29 -47.37 Cantabria 2.05 0.94 1.22 1.91 -40.73 Castilla y León 1.93 1.09 1.74 1.82 -9.99 Castilla-La Mancha 1.58 1.21 2.07 1.68 31.43 Cataluña 2.13 2.02 2.30 2.21 7.97 Comunidad Valenciana 1.67 0.79 1.01 1.14 -39.42 Extremadura 1.03 1.29 1.43 1.46 38.60 Galicia 2.04 1.31 1.68 1.60 -17.45 Madrid 1.78 1.94 1.78 2.08 -0.21 Murcia 1.63 0.34 1.66 1.00 2.19 Navarra 1.90 1.49 1.41 1.88 -25.91 País Vasco 1.31 1.27 1.77 1.69 35.50 Rioja (La) 0.79 0.88 1.44 1.22 81.53 Spain 2.01 1.96 2.05 2.15 1.71
26
Table 7. Descriptive statistics of regional FDI determinants
Mean Median Max Min SD CV Skew Kurtosis GDP 36576.3 21385.0 117793.8 4698.9 35298.7 0.97 1.32 3.41 POP 2416.6 1752.6 7426.7 275.0 2159.9 0.89 1.18 3.15 GDPpc 18046.6 17215.6 24373.1 11665.3 3778.6 0.21 0.14 1.85 LP 40342.0 41576.7 47997.2 32877.6 5096.6 0.13 -0.04 1.60 RII 90.4 84.9 126.7 67.9 18.5 0.20 0.46 1.97 PTK 1007.2 768.5 2597.1 244.1 683.9 0.68 1.23 3.38 TTK 1278.4 980.1 4154.0 331.6 915.3 0.72 2.00 6.83 H 9.8 9.8 10.6 9.0 0.4 0.04 0.03 2.77 R&D 0.7 0.7 1.6 0.2 0.3 0.48 1.18 3.82 NET 15.3 14.2 22.9 8.7 4.3 0.28 0.24 2.03 ER 37.8 38.5 45.3 30.2 4.4 0.12 -0.10 1.92 AR 66.3 66.4 72.1 59.0 3.3 0.05 -0.18 3.04 OR 83.4 84.4 91.2 73.1 5.0 0.06 -0.38 2.32 1/ULC 1.7 1.7 1.8 1.6 0.1 0.04 0.57 2.39 OP 36.1 39.7 67.6 10.3 16.5 0.46 0.18 2.17 IND/GDP 17.4 17.2 29.6 5.4 7.4 0.43 -0.09 2.09
27
Table 8. Correlation matrix
GDP POP GDPpc LP RII PTK TTK H R&D NET ER AR OR 1/ULC OP IND/GDP GDP 1.00 POP 0.94 1.00 GDPpc 0.31 0.05 1.00 LP 0.21 -0.02 0.89 1.00 RII -0.53 -0.51 0.00 0.16 1.00 PTK 0.56 0.36 0.70 0.74 -0.03 1.00 TTK 0.62 0.41 0.65 0.67 -0.17 0.95 1.00 H 0.36 0.12 0.86 0.81 -0.06 0.79 0.77 1.00 R&D 0.64 0.46 0.63 0.63 -0.06 0.98 0.96 0.75 1.00 NET 0.50 0.28 0.91 0.75 -0.32 0.74 0.71 0.85 0.70 1.00 ER 0.27 0.06 0.84 0.51 -0.17 0.38 0.37 0.67 0.37 0.82 1.00 AR 0.34 0.16 0.72 0.42 -0.07 0.33 0.27 0.48 0.32 0.65 0.85 1.00 OR 0.07 -0.13 0.79 0.50 -0.07 0.27 0.28 0.60 0.25 0.72 0.95 0.69 1.00 1/ULC 0.11 0.12 0.17 -0.08 -0.33 -0.21 -0.15 0.06 -0.19 0.21 0.66 0.58 0.51 1.00 OP 0.32 0.19 0.55 0.52 0.59 0.65 0.52 0.60 0.66 0.51 0.39 0.51 0.32 -0.22 1.00 IND/GDP 0.01 -0.10 0.47 0.61 0.56 0.50 0.31 0.45 0.44 0.30 0.17 0.25 0.19 -0.28 0.78 1.00
28
Table 9. Principal components analysis. Total variance explained
Factors Eigenvalue % Variance Cumulative % Variance
1 7.669 47.934 47.934 2 2.666 16.663 64.597 3 2.600 16.249 80.847 4 1.035 6.470 87.317 5 0.667 4.170 91.486 6 0.392 2.451 93.938 7 0.320 1.999 95.937 8 0.209 1.306 97.243 9 0.187 1.169 98.412
10 0.101 0.629 99.041 11 0.077 0.483 99.524 12 0.034 0.214 99.738 13 0.020 0.122 99.860 14 0.013 0.081 99.941 15 0.006 0.038 99.979 16 0.003 0.021 100.000
29
Table 10. Principal components analysis. Rotated component matrix
Variables F1
(Economic potential)
F2 (Labour
conditions)
F3 (Market
size)
F4 (Competitiveness)
Communalities
GDP 0.335 0.131 0.912 -0.047 0.962 POP 0.116 -0.017 0.959 -0.078 0.939 GDPpc 0.684 0.674 -0.024 0.206 0.965 LP 0.802 0.292 -0.158 0.271 0.826 RII -0.094 -0.160 -0.507 0.716 0.805 PTK 0.894 0.055 0.299 0.256 0.957 TTK 0.904 0.042 0.340 0.070 0.938 H 0.806 0.401 0.037 0.153 0.835 R&D 0.830 0.019 0.412 0.219 0.907 NET 0.692 0.624 0.230 -0.031 0.922 ER 0.299 0.932 0.057 0.044 0.963 AR 0.104 0.799 0.244 0.303 0.800 OR 0.264 0.895 -0.144 -0.013 0.891 1/ULC -0.199 0.592 0.059 -0.377 0.535 OP 0.386 0.238 0.254 0.776 0.873 IND/GDP 0.353 0.063 -0.102 0.845 0.852
Note: The variables loading on each factor are shown in bold.
30
Table 11. Regression results (all sectors)
Dependent variable: FDI/GDP
Dependent variable: FDI/POB
Coefficient t-student Coefficient t-student GLS Economic potential 11.70* 2.68 12.38* 2.82 Labour conditions 3.54* 2.13 4.87* 2.94 Market size -4.96 -0.82 -2.85 -0.33 Competitiveness 8.77* 3.71 9.35* 3.91 Adjusted R square 0.74 0.80 Two-stage GLS Economic potential 33.90* 3.54 34.05* 3.58 Labour conditions 3.05 1.10 4.40 1.60 Market size -17.73 -1.28 -5.35 -0.39 Competitiveness 24.36* 4.15 24.98* 4.33 Adjusted R square 0.70 0.75 Note: All equations include fixed effects: (*) significant at 95%
31
Table 12. Regression results (main sectors and branches)
Manufacturing Services
Medium technology
intensity Total
Wholesale and retail
trade Communications Business services Total
Coefficients t-student Coefficients t-
student Coefficients t-student Coefficients
t-
student Coefficients
t-
student Coefficients
t-
student
GLS
Economic potential -2.54 -0.48 -3.52 -0.61 7.77 1.28 99.71* 3.89 25.53* 6.75 21.78* 3.08
Labour conditions 6.98* 2.45 4.80* 2.27 0.45 0.22 5.20 0.88 1.16 1.15 2.12 1.05
Market size -2.94 -0.39 -1.61 -0.20 -12.97 -1.44 -47.60 -1.23 -3.96 -0.73 -3.45 -0.33
Competitiveness 6.03** 1.77 7.23* 2.30 2.46 0.66 54.64* 3.76 16.53* 7.17 14.82* 3.64
Adjusted R square 0.69 0.73 0.76 0.53 0.87 0.79
Two-stage GLS
Economic potential 7.08 0.70 11.05 1.12 -7.97 -0.80 85.63* 2.56 41.05* 5.77 22.66** 1.76
Labour conditions 14.14* 4.01 9.60* 3.68 -3.76 -0.95 -6.31 -0.76 -1.61 -0.71 -0.15 -0.04
Market size -22.03 -1.60 -13.74 -1.10 10.71 0.73 -43.69 -0.92 -6.28 -0.61 -3.65 -0.20
Competitiveness 17.91* 3.36 23.17* 3.84 -10.21 -1.45 55.82* 2.44 30.44* 6.99 17.56* 2.32
Adjusted R square 0.66 0.69 0.76 0.48 0.77 0.76
Note: All equations include fixed effects; (*) significant at 95%; (**) significant at 90%
32
Table 13. Spatial regression results (all sectors)
Dependen variable: FDI/GDP
Coefficient t-student Coefficient t-student
Coefficient t-student
Economic potential 8.63* 2.07 11.60* 2.90 10.47* 2.85 Labour conditions 9.47* 2.89 2.83 0.51 9.72* 4.51 Market size -5.02 -0.79 -7.53 -1.26 -4.98 -0.91 Competitiveness 6.46* 3.04 8.87* 3.00 7.00* 2.58 W_Economic potential
-3.02** -1.91 - - - -
W_Labour conditions
- - 1.00 0.14 - -
W_Competitiveness - - - - -9.58* -3.23 Adjusted R square 0.75 0.74 0.77 Note: All equations include fixed effects; (*) significant at 95%; (**) significant at 90%.
33
Annex. Regional variables used
Code Description Units Source
GDP Gross Domestic Product Constant euros 2000 Cereijo et. al (2007)
POP Population Thousands Spanish Statistical Institute (INE)
GDPpc Per capita GDP GDP (constant euros 2000) per person Cereijo et. al (2007)
LP Labor productivity GDP (constant euros 2000) per employee Cereijo et. al (2007)
RII Roads infrastructure Synthetic index=100 Cereijo et. al (2007)
PTK Private technological capital Constant euros 2000 per employee Cereijo et. al (2007)
TTK Total technological capital Constant euros 2000 per employee Cereijo et. al (2007)
H Human capital Average years of schooling Cereijo et. al (2007)
R&D R&D investment Percentage on GDP Cereijo et. al (2007)
NET Internet users Internet users over population (%) Cereijo et. al (2007)
ER Employment rate Employment over population (%) Cereijo et. al (2007)
AR Activity rate Active population over population (%) Cereijo et. al (2007)
OR Occupation rate Employment over active population (%) Cereijo et. al (2007)
1/ULC
Inverse of unit labor cost Inverse of the ratio between wages and productivity
Spanish Statistical Institute (INE)
OP Openness degree Exports + Imports over GDP (%) Ministry of Industry, Tourism and
Trade (Datacomex)
IND/GDP Share of industry GDP in total GDP Industry GDP over total GDP (%) Spanish Statistical Institute (INE)