The spatial distribution of population in Spain: an ... · Documentos de Trabajo N.º 2028. Author:...

THE SPATIAL DISTRIBUTION

OF POPULATION IN SPAIN:

AN ANOMALY IN EUROPEAN

PERSPECTIVE

2020

Eduardo Gutiérrez, Enrique Moral-Benito,

Daniel Oto-Peralías and Roberto Ramos

Documentos de Trabajo

N.º 2028

THE SPATIAL DISTRIBUTION OF POPULATION IN SPAIN: AN ANOMALY

IN EUROPEAN PERSPECTIVE

Documentos de Trabajo. N.º 2028

2020

(*) We thank Mario Alloza and Carlos Sanz for sharing the municipality data. We also thank David Cuberes, JavierPérez, Diego Puga, Jacopo Timini and seminar participants at Banco de España for useful comments. The opinions and analyses are the responsibility of the authors and, therefore, do not necessarily coincide with those of the Banco de España or the Eurosystem.

Eduardo Gutiérrez, Enrique Moral-Benito and Roberto Ramos

BANCO DE ESPAÑA

Daniel Oto-Peralías

UNIVERSIDAD PABLO DE OLAVIDE

THE SPATIAL DISTRIBUTION OF POPULATION IN SPAIN:

AN ANOMALY IN EUROPEAN PERSPECTIVE (*)

The Working Paper Series seeks to disseminate original research in economics and fi nance. All papers have been anonymously refereed. By publishing these papers, the Banco de España aims to contribute to economic analysis and, in particular, to knowledge of the Spanish economy and its international environment.

The opinions and analyses in the Working Paper Series are the responsibility of the authors and, therefore, do not necessarily coincide with those of the Banco de España or the Eurosystem.

The Banco de España disseminates its main reports and most of its publications via the Internet at the following website: http://www.bde.es.

Reproduction for educational and non-commercial purposes is permitted provided that the source is acknowledged.

© BANCO DE ESPAÑA, Madrid, 2020

ISSN: 1579-8666 (on line)

Abstract

We exploit the GEOSTAT 2011 population grid with a very high 1-km2 resolution to

document that Spain presents the lowest density of settlements among European

countries. We uncover that this anomaly cannot be accounted for by adverse geographic

and climatic conditions. Using techniques from spatial econometrics, we identify the

clusters that exhibit the lowest densities within Spain after controlling for geo-climatic

factors: these areas mainly belong to Teruel, Zaragoza, Ciudad Real, Albacete, Sevilla

and Asturias. We also explore the attributes that characterize the municipalities located

in these low-density areas: larger population losses during the 1950-1991 rural exodus,

higher shares of local-born inhabitants, longer distances to the province capital, higher

shares of population employed in agriculture, and larger increases in regionalist vote

after the Great Recession.

Keywords: economic geography, Spain.

JEL classifi cation: R10.

Resumen

Utilizando información sobre la distribución de la población por cada kilómetro cuadrado

de Europa (GEOSTAT 2011), este artículo documenta que España es el país con un

mayor porcentaje de territorio deshabitado y una mayor concentración de la población

en determinadas zonas geográfi cas. Esta anomalía en la distribución de la población a lo

largo del territorio no puede explicarse por condiciones geográfi cas y climáticas adversas.

Asimismo, utilizando técnicas de la econometría espacial, se identifi can las zonas que

exhiben las densidades más bajas dentro de España al tener en cuenta los factores

geoclimáticos: estas áreas pertenecen principalmente a Teruel, Zaragoza, Ciudad Real,

Albacete, Sevilla y Asturias. Finalmente, se exploran las características de los municipios

ubicados en estas áreas de baja densidad: mayores pérdidas de población durante el

éxodo rural de 1950-1991, mayor proporción de habitantes nacidos en el municipio,

mayores distancias a la capital de la provincia, mayor proporción de población empleada

en la agricultura y mayores aumentos en el porcentaje de voto regionalista después de la

crisis fi nanciera global.

Palabras clave: geografía económica, España.

Códigos JEL: R10.

BANCO DE ESPAÑA 7 DOCUMENTO DE TRABAJO N.º 2028

1Allen and Arkolakis (2014) show that geographic location alone can explain at least 20% of the spatial variationin income across the United States.

2According to Duranton and Turner (2018), most daily activities take place within an area of 10-km2 that doesnot necessarily coincide with administrative areas. In addition, other dimensions beyond administrative units suchas the number of centers within an area or the compactness of development might be relevant in determining urbanoutcomes (Harari 2020).

1 Introduction

How human settlements are distributed across space shapes economic interactions and influences

economic development (see e.g. Krugman 1991).1 This article explores the spatial distribution of

population in Spain compared to that of other European countries. Using the GEOSTAT 2011

population grid, we uncover an anomaly in Spanish settlement patterns. We find that Spain presents

a much lower share of inhabited areas than its European neighbors. Moreover, this pattern cannot

be accounted for by geographic and climatic factors. The remaining of this introduction provides a

detailed advance of the different parts and findings of the paper, which should enable a reader to

grasp in advance the essence of our contribution.

In Section 2, we describe the different datasets we exploit throughout the paper. The GEOSTAT

2011 population grid is the main data source that allows us to analyze the distribution of population

and settlements at a very granular level. In particular, it provides the figures of population living

in each squared kilometer within Europe. This information is crucial to properly measure the dis-

tribution of population across space as opposed to the traditional density indicators (based on the

ratio of population to surface of administrative areas) that do not appropriately reflect the density

actually faced by individuals and firms (see Duranton and Puga, 2020).2 We thus consider three

alternative indicators to compare the actual density experienced across European countries: (i) the

ratio of population to surface within each 250-km2 grid cell; (ii) the so-called settlement density

given by the percentage of 10-km2 cells that are inhabited within each 250-km2 grid cell; (iii) an

indicator of population concentration given by the percentage of the population living in the most

populated one percent of the territory within each 250-km2 grid cell. Section 2 also describes the very

granular information considered in our analysis on geo-climatic factors such as temperature, rainfall,

altitude, ruggedness or soil quality for the different European countries at the grid-cell level. The

municipality-level database exploited in the last section of the paper to characterize underpopulated

areas within Spain is also presented in Section 2.

Section 3 uncovers an anomaly of Spanish settlement patterns in European perspective that

cannot be accounted for by geo-climatic conditions. In particular, the regions with the lowest amount

of settlements relative to its surface within Europe are located in Spain, Iceland, Finland, Norway

and Sweden. However, while the low settlement density is fully explained by adverse geographic and

climatic factors in Scandinavian countries, this is not the case in Spain, which remains the only outlier

in terms of low settlement density after accounting for geo-climatic conditions. Using the alternative

measures of spatial distribution of the population, we conclude that Spain is characterized by a

very low settlement density and a very high concentration of the population in a small share of the


3On the contrary, Madrid, Malaga and Barcelona stand out as the provinces with higher settlement density withinSpain.

4The 35 European countries included are Albania (AL), Austria (AT), Belgium (BE), Bosnia and Herzegovina(BI), Bulgaria (BG), Croatia (HR), Czechia (CZ), Denmark (DK), Estonia (EE), Finland (FI), France (FR), Germany

(DE), Greece (EL), Hungary (HU), Iceland (IS), Ireland (IE), Italy (IT), Kosovo (KO), Latvia (LV), Liechtenstein(LI), Lithuania (LT), Luxembourg (LU), Montenegro (ME), Netherlands (NL), Norway (NO), Poland (PL), Portugal(PT), Romania (RO), Serbia (SR), Slovakia (SK), Slovenia (SI), Spain (ES), Sweden (SE), Switzerland (CH) and theUnited Kingdom (UK).

territory. These two characteristics render traditional population density indicators somehow in line

with other European countries, albeit relatively low. According to Oto-Peralıas (2020), insecurity

characterizing the Reconquest period may explain the anomalous spatial distribution of Spanish

population. Continuous warfare and insecurity heavily conditioned the nature of the colonization

process, characterized by the leading role of the military orders as colonizer agents, scarcity of

population, and a livestock-oriented economy (Gonzalez-Jimenez 1992). This determined a spatial

distribution of the population characterized by a low density of settlements and an economy based

on ranching, as mobile assets were favored by military conditions (see Bishko, 1975).

In Section 4, we characterize underpopulated areas within Spain along two dimensions. First,

we use spatial econometric techniques to identify clusters of underpopulated areas once we account

for geography and climate. Our results indicate that these areas are mainly located in the provinces

of Teruel, Zaragoza, Ciudad Real, Albacete, Sevilla and Asturias.3 Second, we exploit density

measures at the municipality level together with socio-economic characteristics of each municipality

in order to characterize the underpopulated areas. Our findings suggest that low-density areas

belong to municipalities characterized by higher shares of local-born inhabitants that experienced

large population losses during the 1950-1991 rural exodus, longer distances to the province capital,

higher shares of population employed in agriculture, and larger increases in regionalist vote after the

Great Recession. Finally, underpopulated municipalities tend to present lower income per capita in

nominal terms but, interestingly enough, higher income per capita if we use a PPP-adjustment based

on municipality-level housing prices taken from Registro de la Propiedad.

2 Data

2.1 GEOSTAT 2011 population grid

The GEOSTAT population grid provides data on population distribution in space at a very high

1-km2 resolution (Eurostat 2016). The sample consists of the territory covered by GEOSTAT 2011

after excluding the overseas regions (see Figure 1), an area slightly larger than 5 million km2 with

approximately 2.08 million populated 1-km2 grid cells.4

We construct three indicators measuring different dimensions of the spatial distribution of popu-

lation in the different countries. First, a settlement density indicator that measures the distribution

of settlements along the territory. More specifically, it refers to the percentage of 10-km2 grid cells


that are inhabited in each 250-km2 cell. A 10-km2 grid cell is considered to be populated if it contains

at least one 1-km2 populated cell within it. We choose 10-km2 as cell area because it is a meaningful

size from an economic point of view. For instance, Duranton and Turner (2018) conclude that most

daily activities take place within an area of 10-km2. Also, the average size of a commune in France

is 15-km2 and, typically, each commune has more than one settlement. It is worth noting that,

according to this indicator, settlements are identified through the presence of populated 1-km2 cells.

An important advantage of this way of identifying settlements is its homogeneity across countries.

Other alternatives such as data on municipalities or on local administrative units cannot be used for

comparative purposes since they are heterogeneous across countries.

Second, we compute an indicator of population concentration that measures the percentage of

the population living in the most populated one percent of the territory within each 250-km2 grid

cell. It is worth stressing that the level of spatial aggregation used is important to this indicator.

Population concentration may be high at the country level but moderate or low at the sub-national

level. However, there is in practice a strong correlation (0.86) between the country level value and

the average of grid-cell level values.

Third, we also consider a traditional indicator of population density given by the ratio of pop-

ulation to surface in each 250-km2 grid cell. Regarding the correlation among these indicators,

settlement density is negatively correlated with population concentration (-0.75) and positively with

the logarithm of population density (0.77), while population concentration and population density

are negatively correlated (-0.58). Table A.1 provides some descriptive statistics of the three indicators

at the grid-cell level.

2.2 Geographic and climatic factors

We construct several geographic and climatic variables at the grid cell level, including temperature,

rainfall, altitude, ruggedness, distance from the coast and soil quality (see Table A.1 for descriptive

statistics).

Temperature and rainfall refer to annual average temperature and annual precipitation in hun-

dred of milliliters, respectively. They are both sourced from the WorldClim database (Hijmans et al.,

2005). Altitude refers to average altitude of the surface area of the observation unit while ruggedness

corresponds to the standard deviation of the altitude of the territory corresponding to the observa-

tion unit, both are constructed using data from GTOPO30 (Data available from the U.S. Geological

Survey). Distance to the coast refers to the geodesic distance between the centroid of the observa-

tion units and the nearest point of the coast (in km). Soil quality, taken from from Fischer et al.

(2008), corresponds to the average of seven key soil dimensions important for crop production: nu-

trient availability, nutrient retention capacity, rooting conditions, oxygen availability to roots, excess

salts, toxicities, and workability (the average value for each component is calculated for the surface

area corresponding to the observation unit). Finally, we also consider latitude and longitude, the

geographic coordinates of the grid cell centroids in decimal degrees.


5Collantes and Pinilla (2011) provide a detailed description of rural-urban migrations over this period.6Note that the fiscal variables only refer to the city budget and include items such as spending on garbage collection

or street lighting and revenues from property taxes. However, they do not include public spending on e.g. health andeducation, and public revenues from income or value added taxes.

2.3 Municipality-level variables for Spain

We also conduct part of the analysis at the municipality level within Spain. We thus collect several

demographic, political, fiscal and economic variables.

Demographic variables include the dependency ratio and the share of citizens born in the mu-

nicipality, which are extracted from the 2011 Census. The former refers to the share of citizens

older than 65 over the population aged 15-64. We also consider the share of citizens born in the

municipality as an indicator of the attraction of the area with an average of almost 50% of citizens

living where they were born. We also explore how migration flows during the so-called rural exodus

from 1950 to 1991 can explain current disparities in population density across municipalities.5

With respect to political variables, we use the share of discontent vote in 2019 general elections

and the increase of regionalist votes between 2007 and 2019. These municipal variables derive from

the Spanish Ministry of Home Office. While extreme political parties surged after the crisis reaching

18% of the vote on average in 2019, the share of regionalist votes increased by around 3pp over the

same period.

Lastly, we measure fiscal characteristics of municipalities with public spending and public revenues

per capita;6 and socio-economic conditions with the share population employed in agriculture from

the 2011 Census and the income per capita in 2015 published by INE. Average public spending

is around 1,222 euros per capita, slightly below average revenue per capita. The average share of

population employed in agriculture is around 7% and income per capita in the (unweighted) average

municipality is around 10,000 euros. Descriptive statistics at the municipality level are available in

Table A.2.

3 The Spanish anomaly in settlement patterns

This section documents that the spatial distribution of the population in Spain represents an anomaly

with respect to other European countries. In particular, it presents an abnormally low settlement

density with the highest prevalence of uninhabited areas even after accounting for geographic and

climatic conditions. The section also provides some heuristic discussion on the potential causes of

this pattern, which seem to be rooted in historical patterns of warfare.

3.1 A first glimpse at the data

We first describe the main features of the population location in Spain in comparison with other

European countries. Compared to other European Union countries, population in Spain stands out

for two characteristics. First, total population is low. And second, it is highly concentrated. The


first two columns of Table 1 show that, despite being the second country by land area, Spain only

ranks 5th in terms of total population. For example, the number of inhabitants is 25% lower than

in the United Kingdom, despite the land area being twice as much.

The low number of people living in Spain stems from the fact that there is a very high amount of

uninhabited areas. Figure 1, which plots in red (white) the inhabited (uninhabited) 1 km2 grid cells

in Europe, shows that this feature stands out in sharp contrast with other countries. Indeed, Spain

appears comparable only to areas where geographical and climatological conditions deter population

settlements, such as the Scandinavian Peninsula or the Alps. Column 4 of Table 1 shows that only

13% of the Spanish land area is inhabited, which is the lowest value in the European Union.

This exceptionally low settlement density makes population in Spain be very concentrated. In-

deed, column 5 of Table 1 reveals that the population density in inhabited areas reaches 737 people

per squared km, which is the second largest in Europe (being the largest the tiny island of Malta). If

we compute Lorenz curves estimating the inequality in the spatial location of population along the

Table 1: Population Density in European Union Countries (2011).

Surface (km2) Population (thousands) Density Inhabited area (%) Density inhabited area(1) (2) (3) (4) (5)

France 549,060 62,765 114 67.8 168Spain 498,504 46,816 94 12.7 737Sweden 449,896 9,539 21 25.2 84Germany 358,327 80,213 224 59.9 374Finland 337,547 5,340 16 30.0 53Poland 313,851 38,500 123 62.6 196Italy 301,291 59,429 197 57.2 345United Kingdom 247,763 63,154 255 51.7 493Romania 239,068 20,122 84 29.4 287Greece 131,912 10,634 81 19.7 409Bulgaria 110,995 7,365 66 21.3 312Hungary 93,013 9,938 107 29.8 358Portugal 88,847 10,562 119 46.6 255Austria 83,944 8,402 100 51.4 195Czech Republic 78,874 10,437 132 56.0 236Ireland 70,601 4,575 65 80.2 81Latvia 65,519 2,081 32 50.5 63Lithuania 65,412 3,029 46 54.5 85Croatia 56,539 4,290 76 43.4 175Slovakia 49,035 5,399 110 32.4 340Estonia 45,347 1,294 29 45.6 63Denkmark 43,162 5,535 128 90.4 142Netherlands 37,824 16,651 440 81.0 544Belgium 30,668 10,990 358 83.0 432Slovenia 20,277 2,049 101 66.0 153Cyprus 9,249 840 91 37.3 244Luxembourg 2,595 513 198 65.5 302Malta 315 417 1325 92.7 1430

Source: Eurostat.

Note however that this anomaly is somehow masked when looking at traditional population density

measures at the aggregate level (inhabitants per square km) in which Spain ranks slightly below the

mean (see column 3).7

7Rae (2018) shows inhabited areas density in 39 European countries to reflect that the traditional measure is notcomplete.


8We also consider a third measure of population concentration. Since the findings corresponding to this measureare very similar to those of settlement density, its discussion is relegated to Appendix B.

3.2 Testing for the anomaly and the role of geo-climatic factors

The evidence presented in Table 1 and Figure 1 is suggestive of a Spanish anomaly in population

patterns. However, Spain is characterized by high temperatures and large mountainous lands that

might account for the high prevalence of uninhabited areas. This section explores whether geographic

and climatic factors may explain the differences between Spain and other European countries in the

distribution of population across the territory.

Using the GEOSTAT 2011 population grid, we compute two proxies for the settlement patterns in

the different countries: settlement density and population density.8 Crucially, using the GEOSTAT

2011 population grid allows us to identify population patterns through 1-km2 cells.

inhabited 1 km2 grid cells, one can see that Spain is the most spatially concentrated country among

the selected ones. For example, while in Germany, Poland and Portugal 15% of the most populated

cells account for 80% of total population, in Spain they account for 90%. The corresponding figure

for France, Italy and the United Kingdom is 85%.

All in all, Table 1 suggests that population in Spain is very concentrated in the space because

there is an abnormally large number of uninhabited areas in comparison to other European countries.

According to these measures, we are able to identify a Spanish anomaly in settlement patterns.

where yic refers to either settlement density or (log) population density of 250-km2 grid cell i be-

longing to country c. xic is a vector of geographic and climatic variables for each grid cell including

temperature, rainfall, average altitude, ruggedness, soil quality and distance from the coast. In or-

der to allow for non-linear effects of those variables on settlement patterns, quartile dummies are

included with the omitted category being the first quartile. An island dummy and latitude and lon-

gitude coordinates are also included in the vector xic. Table A.3 in Appendix A shows the estimated

coefficients for the geographic and climatic variables included in the regressions.

Finally, our main object of interest in equation (1) is the vector ηc that refers to a full set of

country dummies. In order to test the significance of the Spanish anomaly after accounting for

geography and climate, we analyze the estimated country dummies (ηc) from equation (1) and their

associated standard errors. France is the omitted category in all cases so that the estimated country

dummies reflect the average difference in settlement density between each country and France.

The following regression allows us to better understand the role of geography and climate in

explaining the Spanish anomaly in terms of settlement density and population density:

yic = ηc + x′icβ + εic (1)


Source: Eurostat.Notes:This figure depicts in red (white) the inhabited (uninhabited) 1 km2 grid cells in Europe.

Figure 1: Inhabited Grid Cells in Europe (2011).

Figure 2: The Spanish anomaly in settlement and population density.

AT

BE

DE

FI

IE

IT

LU

NL

PT

UK

ES

−50 0 50 100 −50 0 50 100

Baseline Geo−climatic controls

AT

BE

DE

FI

IE

IT

LU

NL

PT

UK

ES

−5 0 5 10 −5 0 5 10


Figure 2 presents the estimated country fixed effects without controls (baseline) and with ge-

ographic and climatic controls (geo-climatic controls) for both settlement density (left panel) and

population density (right panel). In the baseline panels without controls, Spain is the country with

the lowest share of inhabited 10km2 areas within each 250-km2 grid cell as well as the second lowest


9Figure A.1 in Appendix A shows that these findings also hold in a larger sample of European countries. However,for the sake of clarity we present here the results for the Eurozone sample and the UK.

population density only above Finland. For instance, the average 250-km2 grid cell in Spain presents

a settlement density 50 percentage points lower than the corresponding average cell in France, while

this figure is around 30 pp. in Finland. Interestingly enough, this difference becomes positive and

statistically significant in the case of Finland after accounting for geographic and climatic factors,

while it remains negative and significant in Spain, albeit smaller in magnitude.

The left panel of Figure 2 clearly illustrates that geography and climate cannot fully explain

the Spanish low settlement density, which is the lowest within Europe.9 In sharp contrast, the

Finnish low settlement density is fully explained by geographic and climatic factors. Indeed, after

accounting for those factors, settlement density in Finland is the highest in Europe. This pattern

remains unaltered if we use population concentration instead of settlement density as the dependent

variable in our regressions as shown in Figure B.6 in Appendix B. Population concentration is thus

the highest in Spain even after accounting for geography and climate. Indeed, both variables present

a very high and negative correlation of -0.75. Turning to population density in the right panel of

Figure 2, the difference between the average Spanish and French 250-km2 grid cell is not statistically

significant once we account for geo-climatic conditions.

A potential concern is that we are comparing countries that are very heterogeneous in surface

area. One could suspect that the singularity of Spain is perhaps conditional on the specific territorial

division used in our analysis. In order to address this concern, we compare Spain to a set of virtual

countries with similar surface areas. For that purpose, we randomly choose 1,000 grid cells and

then create virtual countries by selecting up to 2,500 neighbors falling within 700 km from the cell’s

centroid (similar to Spanish land area). Then we run 1,000 regressions of settlement density on the

full set of geo-climatic controls, the Spain dummy, and the virtual countries dummies (included one

by one). Remarkably, in every case the coefficient on Spain for settlement density is lower than

on the virtual region (see Figure A.2 in Appendix A). This figure is slightly above 70% in the case

of population density, which somehow confirms the lack of statistical significance of the Spanish

coefficient in the right panel of Figure 2.

In order to further explore the Spanish anomaly, we now turn to the analysis at the regional

level. In particular, we estimate models analogous to equation (1) but instead of including country

dummies, we now include region dummies at the NUTS3 level corresponding to Spanish provinces.

The omitted category is now the Paris area (FR101) so that all estimated dummies capture the

difference with respect to Paris.

The left panel of Figure 3 confirms the Spanish anomaly in settlement density by region. The

y-axis plots the estimated region dummies without geographic and climatic controls while the x-

axis refers to the estimated dummies with those controls. Interestingly enough, region dummies of

nordic countries such as Iceland (IS), Sweden (SE), Finland (FI) and Norway (NO) present highly

negative coefficients without geographic controls (much lower settlement densities than Paris) that


Figure 3: The Spanish anomaly by regions.

ALB0

ALB1

ALB2

ALB3ALB4

ALB5ALB6

ALB7

ALB8ALB9

ALB99

AT111

AT112

AT113AT121AT122

AT123AT124

AT125AT126AT127AT130

AT211

AT212

AT213

AT221

AT222

AT223

AT224AT225

AT226

AT311AT312AT313

AT314AT315

AT321AT322

AT323

AT331

AT332

AT333AT334

AT335

AT341

AT342

BE100BE211BE212BE213BE221BE222BE223BE231BE232BE233BE234BE235BE236BE241BE242BE251BE252BE253BE254BE255BE256BE257BE258BE310BE321BE322BE323BE324BE325BE326BE327BE331BE332BE334

BE335

BE336BE341BE342BE343

BE344

BE345

BE351

BE352BE353

BG311

BG312BG313

BG314

BG315BG321

BG322

BG323

BG324BG325

BG331

BG332BG333

BG334

BG341

BG342

BG343

BG344

BG411

BG412

BG413

BG414

BG415

BG421

BG422

BG423

BG424

BG425

BIH10

BIH11

BIH12

BIH13

BIH14

BIH15BIH16

BIH17BIH18BIH19

BIH20

BIH21BIH22 BIH23

BIH24

BIH97

BIH98BIH99

CH011

CH012

CH013

CH021

CH022

CH023CH024CH025CH032CH033CH040

CH051

CH052

CH053

CH054

CH055

CH056

CH057

CH061

CH062

CH063CH064

CH066

CH070

CZ010CZ020CZ031CZ032

CZ041

CZ042CZ051


CZ071CZ072CZ080

DE111DE112

DE113DE114DE115DE116DE117DE118DE119DE11ADE11BDE11CDE11DDE121DE122DE123DE124DE126DE127DE128DE129

DE12A

DE12B

DE12C

DE131DE132DE133DE134 DE135DE136DE137DE138DE139DE13ADE141DE142DE143DE145DE146

DE147

DE148DE149DE212DE213DE214

DE215DE216

DE217DE218DE219DE21ADE21BDE21C

DE21D

DE21E

DE21F

DE21GDE21HDE21IDE21J

DE21KDE21L

DE21MDE21NDE222DE223DE224

DE225

DE226DE227DE228

DE229

DE22ADE22BDE22CDE232DE234DE235DE236

DE237DE238DE239DE23A

DE241DE245DE246DE247DE248DE249DE24ADE24BDE24CDE24DDE251

DE254DE255DE256DE257DE258DE259DE25ADE25BDE25CDE262

DE264DE265

DE266DE267DE268DE269

DE26ADE26B

DE26CDE271DE272

DE274DE275DE276

DE277DE278DE279

DE27A

DE27B

DE27CDE27D

DE27E

DE300DE401

DE402DE403 DE404DE405

DE406DE407DE408

DE409

DE40ADE40BDE40C

DE40DDE40E

DE40F

DE40GDE40H

DE40IDE501DE502DE600

DE711

DE712DE715

DE716DE717 DE718DE719

DE71ADE71BDE71CDE71DDE71EDE721DE722DE723DE724

DE725DE731

DE732DE733DE734


DE803

DE804

DE807

DE808DE809DE80ADE80B

DE80CDE80D

DE80E

DE80F

DE80GDE80H

DE80I

DE911DE912

DE913

DE914DE915

DE916DE917

DE918

DE919

DE91ADE91BDE922DE923DE925

DE926

DE927DE928DE929

DE931

DE932DE933DE934DE935DE936DE937

DE938

DE939

DE93A

DE93BDE944DE945DE946DE947

DE948DE949

DE94A

DE94BDE94C

DE94DDE94EDE94FDE94GDE94HDEA11DEA12DEA13DEA15DEA1BDEA1CDEA1DDEA1EDEA1FDEA22DEA23

DEA26DEA27 DEA28DEA29 DEA2ADEA2BDEA2CDEA2DDEA32DEA33DEA34DEA35DEA36DEA37DEA38DEA41DEA42DEA43DEA44

DEA45DEA46

DEA47DEA52DEA53DEA54DEA55 DEA56

DEA57DEA58DEA59DEA5A

DEA5B

DEA5C DEB12DEB13DEB14

DEB15

DEB16DEB17DEB18 DEB19DEB1ADEB1BDEB21DEB22DEB23

DEB24DEB25DEB32

DEB35DEB36

DEB3ADEB3B

DEB3C

DEB3D

DEB3E

DEB3F

DEB3GDEB3HDEB3J

DEB3K

DEC01DEC02DEC03DEC04DEC05DEC06DED21DED2C

DED2D

DED2EDED2FDED41

DED42DED43DED44DED45DED51DED52DED53

DEE01DEE03DEE04DEE05

DEE06DEE07

DEE08

DEE09DEE0A

DEE0BDEE0CDEE0D

DEE0E

DEF03DEF05DEF06DEF07

DEF08DEF09DEF0ADEF0B

DEF0C

DEF0DDEF0EDEF0FDEG01DEG02DEG03

DEG04DEG05

DEG06

DEG07DEG09

DEG0ADEG0B

DEG0CDEG0D

DEG0EDEG0F

DEG0G

DEG0H

DEG0IDEG0JDEG0KDEG0LDEG0M

DEG0PDK011

DK012DK013DK014DK021DK022DK031DK032

DK041DK042DK050

EE001EE004

EE006

EE007

EE008

EL111

EL112EL113

EL114

EL115EL121EL122EL123

EL124

EL125EL126EL127

EL131EL132EL133

EL134

EL141

EL142

EL143 EL144

EL211

EL212

EL213

EL214EL221

EL222

EL223EL224

EL231

EL232

EL233

EL241

EL242

EL243

EL244EL245EL251 EL252

EL253

EL254

EL255EL300

EL411

EL412

EL413

EL421

EL422

EL431

EL432

EL433

EL434

FI193FI194FI195

FI196FI197FI1B1FI1C1FI1C2

FI1C3FI1C4FI1C5FI1D1FI1D2

FI1D3

FI1D4

FI1D5FI1D6

FI1D7

FI200

FR103FR104FR105FR106FR107FR108

FR211FR212FR213FR214

FR221FR222FR223FR231FR232FR241FR242FR243FR244FR245FR246FR251FR252FR253

FR261FR262FR263

FR264FR301FR302

FR411

FR412

FR413FR414

FR421FR422FR431FR432FR433FR434FR511FR512FR513FR514FR515FR521FR522FR523FR524FR531FR532FR533FR534FR611

FR612FR613

FR614

FR615

FR621

FR622FR623

FR624FR625

FR626

FR627FR628 FR631FR632FR633FR711FR712FR713

FR714

FR715FR716

FR717

FR718

FR721FR722FR723FR724FR811FR812FR813

FR814

FR815FR821

FR822

FR823

FR824

FR825FR826

FR831FR832 HR031HR032

HR033

HR034HR035

HR036

HR037

HR041HR042

HR043HR044HR045

HR046HR047

HR048

HR049

HR04A

HR04BHR04C

HR04DHR04E

HU101

HU102

HU211

HU212

HU213HU221HU222

HU223HU231

HU232HU233

HU311

HU312

HU313

HU321

HU322

HU323

HU331

HU332

HU333

IE011IE012

IE013

IE021IE022IE023IE024IE025

IS001

IS002

ITC11

ITC12ITC13

ITC14

ITC15

ITC16

ITC17ITC18

ITC20

ITC31

ITC32ITC33ITC34ITC41

ITC42ITC43

ITC44

ITC46ITC47

ITC48ITC49ITC4AITC4BITC4CITC4D

ITF11

ITF12ITF13

ITF14

ITF21

ITF22ITF31ITF32ITF33

ITF34ITF35

ITF43ITF44ITF45

ITF46

ITF47

ITF48

ITF51ITF52ITF61ITF62ITF63

ITF64ITF65

ITG11

ITG12

ITG13ITG14

ITG15ITG16

ITG17

ITG18ITG19

ITG25

ITG26

ITG27ITG28

ITG29

ITG2A

ITG2BITG2C

ITH10ITH20

ITH31ITH32

ITH33

ITH34ITH35ITH36ITH37

ITH41 ITH42

ITH43ITH44ITH51ITH52ITH53ITH54ITH55

ITH56

ITH57ITH58

ITH59

ITI11

ITI12ITI13ITI14ITI15

ITI16

ITI17 ITI18ITI19

ITI1AITI21ITI22ITI31ITI32

ITI33ITI34

ITI35

ITI41ITI42

ITI43ITI44

ITI45

KOS25

KOS26

KOS27

KOS28KOS29

KOS30

KOS31

KOS33KOS34

KOS35

KOS36KOS37

KOS39

KOS40KOS42

KOS43

KOS44KOS45

KOS46

KOS47

KOS48

KOS49KOS50KOS51

KOS52

KOS53LI000LT001

LT002LT003

LT004LT005LT006LT007

LT008LT009LT00A

LU000

LV003LV005

LV006

LV007 LV008LV009

ME00

NL111NL112

NL113

NL121

NL122NL123NL131NL132NL133NL211NL212NL213NL221NL224NL225NL226NL230NL310NL321NL322

NL323NL325NL326NL327NL332NL333NL337

NL338

NL339

NL33ANL341NL342NL411NL412NL413NL414NL421NL422NL423NO011

NO012

NO021

NO022

NO031

NO032

NO033

NO034NO041

NO042NO043

NO051

NO052

NO053

NO061

NO062 NO071

NO072

NO073

PL113PL114PL115PL116PL117PL121PL122PL127PL128PL129PL12APL213PL214

PL215

PL216PL217PL224PL225

PL227PL228

PL229

PL22APL22BPL22C


PL323

PL324PL325PL326PL331PL332



PL422PL423PL424

PL425

PL431

PL432

PL514

PL515

PL516 PL517PL518PL521PL522PL613

PL614PL615PL621PL622PL623PL631

PL633PL634

PL635

PT111

PT112PT113PT114PT115PT116

PT117

PT118

PT150PT161PT162PT163

PT164PT165PT166PT167

PT168

PT169

PT16A

PT16B

PT16CPT171

PT172

PT181

PT182

PT183

PT184

PT185RO111

RO112

RO113

RO114

RO115

RO116

RO121

RO122

RO123RO124

RO125

RO126

RO211

RO212

RO213

RO214

RO215

RO216

RO221

RO222

RO223

RO224

RO225

RO226RO311

RO312

RO313

RO314

RO315

RO316

RO317

RO322

RO411

RO412RO413RO414 RO415

RO421

RO422

RO423

RO424

SE110

SE121

SE122SE123

SE124SE125

SE211SE212SE213SE214

SE221SE224SE231

SE232

SE311

SE312

SE313SE321

SE322

SE331

SE332

SI011 SI012SI013

SI014SI016

SI017

SI018

SI021

SI022SI023

SI024

SK010

SK021 SK022SK023

SK031

SK032SK041SK042

SRB54

SRB55

SRB56

SRB57SRB58SRB59

SRB60SRB61SRB62

SRB63SRB64SRB65SRB66SRB67

SRB68

SRB90SRB91

SRB92SRB93

SRB94

SRB95

SRB96SRB97

SRB98

SRB99

UKC11UKC12

UKC14UKC21

UKC22UKC23

UKD11UKD12

UKD31

UKD32

UKD41UKD43

UKD61UKD62UKD63UKD71UKD72UKD73UKD74UKE12UKE13UKE21

UKE22

UKE31

UKE32

UKE41UKE42UKE44

UKE45UKF12UKF13

UKF15UKF16UKF22UKF24UKF25UKF30UKG11UKG12UKG13UKG21UKG22UKG24UKG31UKG32UKG37UKH11UKH12UKH13UKH14UKH23UKH24UKH25UKH32UKH33UKI12UKI21UKI22UKI23UKJ11UKJ12UKJ13UKJ14UKJ21UKJ22UKJ23UKJ24UKJ31UKJ33UKJ34UKJ42UKK11UKK12UKK13UKK14UKK15

UKK21UKK22UKK23UKK30UKK42UKK43

UKL11

UKL12

UKL13UKL14

UKL15UKL16UKL17

UKL18UKL21UKL22UKL23

UKL24

UKM21

UKM22

UKM23

UKM24

UKM26

UKM27

UKM28

UKM31

UKM32

UKM33

UKM34UKM35UKM36

UKM37

UKM38

UKM50

UKM61

UKM62

UKM63

UKM64

UKM65

UKM66

UKN02

UKN03UKN04

UKN05ES111

ES112

ES113

ES114

ES120ES130ES211

ES212

ES213

ES220ES230

ES241

ES242ES243

ES300

ES411

ES412ES413

ES414ES415ES416

ES417

ES418

ES419

ES421ES422

ES423

ES424

ES425

ES431ES432

ES511ES512

ES513

ES514

ES521

ES522

ES523

ES531

ES532

ES533

ES611

ES612

ES613

ES614

ES615ES616

ES617

ES618

ES620

−100

−80

−60

−40

−20

0Se

ttlem

ent d

ensi

ty

−50 0 50 100Settlement density (geo−climatic controls)

ALB0ALB1

ALB2ALB3 ALB4

ALB5ALB6ALB7

ALB8ALB9

ALB99

AT111

AT112

AT113AT121AT122

AT123

AT124AT125

AT126AT127

AT130

AT211

AT212AT213

AT221

AT222AT223AT224

AT225

AT226

AT311

AT312

AT313AT314

AT315

AT321AT322

AT323

AT331

AT332

AT333AT334

AT335

AT341

AT342

BE100

BE211BE212

BE213

BE221

BE222

BE223BE231BE232

BE233

BE234BE235BE236BE241BE242BE251

BE252BE253

BE254BE255

BE256BE257

BE258BE310

BE321

BE322

BE323BE324

BE325

BE326

BE327BE331BE332

BE334BE335

BE336

BE341

BE342BE343

BE344BE345

BE351

BE352

BE353

BG311BG312BG313BG314

BG315BG321 BG322BG323

BG324

BG325

BG331

BG332

BG333BG334

BG341

BG342

BG343

BG344

BG411

BG412

BG413BG414

BG415

BG421

BG422BG423

BG424BG425

BIH10

BIH11BIH12

BIH13

BIH14BIH15BIH16

BIH17

BIH18

BIH19BIH20

BIH21

BIH22BIH23BIH24

BIH97

BIH98

BIH99

CH011

CH012

CH013CH021

CH022CH023

CH024

CH025

CH032CH033 CH040

CH051

CH052

CH053

CH054CH055

CH056

CH057

CH061

CH062

CH063

CH064

CH066

CH070

CZ010

CZ020

CZ031CZ032 CZ041

CZ042CZ051CZ052

CZ053CZ063CZ064CZ071CZ072CZ080

DE111


DE117


DE11CDE11DDE121

DE122

DE123DE124

DE126

DE127

DE128DE129

DE12ADE12B

DE12C

DE131

DE132DE133DE134 DE135DE136DE137DE138DE139

DE13A

DE141DE142


DE212

DE213

DE214

DE215DE216

DE217DE218DE219

DE21ADE21BDE21C

DE21D

DE21E

DE21F

DE21G

DE21H

DE21IDE21J

DE21KDE21L

DE21MDE21N

DE222DE223DE224


DE229

DE22ADE22BDE22C

DE232

DE234DE235DE236DE237DE238DE239DE23A

DE241


DE24ADE24BDE24C

DE24DDE251

DE254

DE255

DE256

DE257DE258

DE259DE25ADE25BDE25C

DE262DE264


DE26ADE26B

DE26C

DE271

DE272DE274DE275DE276DE277DE278DE279

DE27A

DE27B

DE27CDE27D

DE27E

DE300

DE401DE402

DE403DE404

DE405DE406DE407DE408DE409DE40ADE40BDE40CDE40D

DE40EDE40F

DE40GDE40H

DE40I

DE501DE502

DE600

DE711

DE712


DE71A

DE71B

DE71C

DE71DDE71EDE721DE722DE723DE724DE725

DE731


DE801DE802 DE803DE804

DE807DE808DE809DE80ADE80BDE80CDE80D

DE80EDE80FDE80G

DE80HDE80I

DE911DE912DE913


DE918DE919

DE91ADE91B

DE922DE923DE925

DE926DE927

DE928DE929

DE931DE932DE933

DE934

DE935DE936

DE937DE938DE939

DE93A

DE93B

DE944DE945

DE946DE947DE948DE949DE94ADE94B

DE94C

DE94DDE94EDE94FDE94GDE94H

DEA11DEA12DEA13

DEA15

DEA1B

DEA1CDEA1D

DEA1EDEA1F

DEA22DEA23

DEA26

DEA27

DEA28

DEA29 DEA2A

DEA2B

DEA2CDEA2D

DEA32DEA33

DEA34DEA35DEA36

DEA37DEA38

DEA41

DEA42DEA43

DEA44DEA45DEA46 DEA47

DEA52

DEA53DEA54

DEA55DEA56

DEA57

DEA58DEA59DEA5ADEA5B

DEA5C

DEB12DEB13

DEB14DEB15DEB16

DEB17DEB18

DEB19DEB1ADEB1BDEB21

DEB22

DEB23DEB24

DEB25

DEB32

DEB35

DEB36DEB3ADEB3BDEB3CDEB3D

DEB3E

DEB3FDEB3G

DEB3HDEB3J

DEB3K

DEC01

DEC02

DEC03DEC04DEC05DEC06

DED21

DED2CDED2D

DED2EDED2F

DED41

DED42DED43DED44

DED45

DED51

DED52DED53

DEE01

DEE03

DEE04

DEE05DEE06DEE07

DEE08DEE09DEE0A

DEE0BDEE0C

DEE0DDEE0E

DEF03

DEF05DEF06

DEF07DEF08

DEF09DEF0ADEF0BDEF0CDEF0D

DEF0E

DEF0FDEG01DEG02DEG03DEG04DEG05

DEG06DEG07DEG09DEG0ADEG0BDEG0CDEG0DDEG0EDEG0FDEG0GDEG0HDEG0IDEG0JDEG0K

DEG0LDEG0MDEG0P

DK011DK012

DK013

DK014DK021

DK022DK031DK032DK041DK042DK050 EE001

EE004EE006

EE007

EE008EL111

EL112EL113

EL114

EL115EL121EL122

EL123EL124

EL125EL126

EL127EL131EL132

EL133

EL134

EL141EL142

EL143

EL144EL211

EL212EL213

EL214EL221

EL222

EL223


EL241EL242 EL243

EL244EL245

EL251 EL252EL253EL254

EL255EL300

EL411

EL412

EL413EL421EL422

EL431

EL432

EL433

EL434FI193

FI194FI195FI196

FI197

FI1B1

FI1C1FI1C2FI1C3FI1C4

FI1C5FI1D1FI1D2

FI1D3

FI1D4

FI1D5

FI1D6

FI1D7

FI200

FR103FR104

FR105FR106FR107

FR108


FR221FR222

FR223FR231FR232

FR241FR242

FR243FR244FR245

FR246FR251FR252

FR253FR261FR262

FR263FR264

FR301FR302

FR411

FR412

FR413FR414

FR421FR422

FR431FR432FR433

FR434FR511FR512



FR532FR533FR534FR611FR612

FR613

FR614FR615

FR621

FR622FR623

FR624FR625FR626


FR633FR711

FR712FR713

FR714FR715FR716

FR717

FR718FR721


FR812FR813

FR814FR815FR821 FR822FR823

FR824FR825FR826

FR831FR832 HR031

HR032

HR033HR034HR035

HR036

HR037

HR041

HR042

HR043HR044HR045

HR046

HR047HR048HR049

HR04AHR04BHR04CHR04DHR04E

HU101

HU102HU211

HU212HU213HU221HU222HU223HU231HU232HU233

HU311

HU312HU313HU321

HU322HU323HU331HU332

HU333IE011IE012

IE013

IE021

IE022

IE023IE024IE025

IS001

IS002

ITC11ITC12

ITC13

ITC14

ITC15

ITC16

ITC17ITC18

ITC20

ITC31ITC32ITC33ITC34

ITC41ITC42

ITC43

ITC44

ITC46ITC47ITC48

ITC49ITC4AITC4B

ITC4CITC4D

ITF11

ITF12ITF13

ITF14ITF21

ITF22ITF31ITF32

ITF33

ITF34ITF35ITF43

ITF44ITF45

ITF46

ITF47ITF48

ITF51ITF52

ITF61ITF62 ITF63ITF64ITF65

ITG11

ITG12ITG13ITG14ITG15ITG16ITG17

ITG18ITG19

ITG25ITG26ITG27ITG28

ITG29

ITG2A

ITG2BITG2C

ITH10

ITH20

ITH31ITH32

ITH33

ITH34ITH35ITH36

ITH37ITH41

ITH42

ITH43ITH44

ITH51ITH52ITH53ITH54ITH55

ITH56

ITH57

ITH58

ITH59

ITI11ITI12ITI13ITI14ITI15

ITI16

ITI17ITI18

ITI19ITI1A


ITI33ITI34

ITI35

ITI41ITI42

ITI43ITI44

ITI45KOS25

KOS26

KOS27

KOS28

KOS29KOS30

KOS31

KOS33KOS34

KOS35

KOS36KOS37KOS39

KOS40KOS42KOS43

KOS44KOS45

KOS46

KOS47

KOS48 KOS49KOS50KOS51

KOS52

KOS53LI000

LT001

LT002LT003LT004

LT005LT006LT007LT008

LT009LT00A

LU000

LV003 LV005

LV006

LV007LV008LV009ME00

NL111NL112

NL113

NL121

NL122NL123NL131

NL132

NL133NL211NL212NL213

NL221NL224

NL225

NL226

NL230

NL310

NL321


NL327NL332NL333

NL337NL338

NL339

NL33A

NL341NL342

NL411

NL412NL413

NL414NL421NL422

NL423NO011

NO012

NO021NO022

NO031

NO032

NO033

NO034NO041

NO042NO043NO051NO052

NO053

NO061

NO062 NO071NO072

NO073

PL113

PL114PL115PL116PL117PL121PL122

PL127

PL128PL129PL12A

PL213

PL214

PL215

PL216PL217PL224 PL225

PL227PL228PL229

PL22A

PL22BPL22C

PL311PL312

PL314PL315

PL323


PL332PL343PL344

PL345

PL411PL414

PL415

PL416PL417PL418

PL422PL423

PL424

PL425PL431PL432

PL514

PL515PL516 PL517PL518PL521PL522

PL613PL614PL615


PL633

PL634PL635PT111

PT112PT113

PT114

PT115PT116

PT117

PT118

PT150

PT161PT162PT163

PT164PT165

PT166

PT167

PT168PT169

PT16A

PT16B

PT16C

PT171

PT172

PT181PT182PT183

PT184

PT185RO111RO112RO113

RO114

RO115RO116RO121

RO122RO123RO124

RO125

RO126

RO211RO212RO213

RO214RO215RO216

RO221RO222RO223

RO224

RO225

RO226RO311RO312

RO313

RO314RO315

RO316

RO317

RO322

RO411RO412

RO413

RO414RO415RO421RO422RO423

RO424SE110



SE214

SE221SE224SE231

SE232

SE311

SE312

SE313SE321

SE322SE331

SE332

SI011

SI012SI013

SI014SI016

SI017SI018

SI021

SI022

SI023SI024

SK010

SK021SK022

SK023

SK031SK032

SK041SK042SRB54

SRB55SRB56

SRB57SRB58SRB59

SRB60SRB61

SRB62SRB63

SRB64

SRB65SRB66SRB67

SRB68SRB90SRB91

SRB92

SRB93

SRB94SRB95

SRB96SRB97

SRB98

SRB99

UKC11UKC12

UKC14

UKC21

UKC22UKC23

UKD11

UKD12

UKD31UKD32

UKD41UKD43

UKD61

UKD62UKD63

UKD71UKD72

UKD73

UKD74

UKE12

UKE13UKE21

UKE22

UKE31

UKE32

UKE41UKE42UKE44UKE45

UKF12

UKF13UKF15

UKF16UKF22UKF24UKF25

UKF30UKG11

UKG12UKG13UKG21

UKG22

UKG24

UKG31

UKG32

UKG37

UKH11UKH12UKH13UKH14

UKH23UKH24

UKH25UKH32

UKH33

UKI12

UKI21UKI22UKI23

UKJ11UKJ12UKJ13UKJ14

UKJ21

UKJ22 UKJ23UKJ24

UKJ31

UKJ33UKJ34UKJ42

UKK11

UKK12UKK13

UKK14

UKK15

UKK21

UKK22UKK23UKK30

UKK42

UKK43UKL11

UKL12

UKL13

UKL14

UKL15UKL16UKL17

UKL18UKL21

UKL22UKL23

UKL24UKM21

UKM22UKM23

UKM24

UKM26

UKM27

UKM28

UKM31UKM32

UKM33

UKM34

UKM35UKM36

UKM37

UKM38UKM50

UKM61UKM62

UKM63UKM64 UKM65UKM66

UKN02

UKN03

UKN04UKN05

ES111

ES112ES113

ES114

ES120

ES130

ES211

ES212ES213

ES220ES230

ES241ES242ES243

ES300

ES411ES412ES413ES414ES415ES416

ES417

ES418

ES419ES421ES422

ES423ES424

ES425ES431

ES432

ES511

ES512

ES513

ES514

ES521

ES522ES523

ES531

ES532ES533

ES611ES612

ES613

ES614

ES615ES616

ES617

ES618ES620

−10

−50

5Po

pula

tion

dens

ity

−5 0 5 10Population density (geo−climatic controls)

turn positive and large after including geo-climatic controls. However, Spanish regions (red dots)

Figure 4: Settlement patterns across Spanish provinces.

Coruña

LugoOrense

Pontevedra

Asturias

Cantabria

Alava

GuipuzcoaVizcaya

NavarraLa Rioja

Huesca

TeruelZaragoza

Madrid

Avila

BurgosLeon

PalenciaSalamanca

Segovia

Soria

Valladolid

Zamora

Guadalajara

Barcelona

Girona

Lleida

Tarragona

Castellon

Ibiza

Mallorca

Menorca

Albacete

Ciudad Real

Cuenca

Toledo

Badajoz

Caceres

Alicante

Valencia

Almeria

Cadiz

Cordoba

Granada

Huelva

Jaen

Malaga

Sevilla Murcia

12

34

56

Popu

latio

n de

nsity

20 40 60 80 100Settlement density

Southern provinces Rest of Spain

Coruña

LugoOrense

Pontevedra

Asturias

Cantabria

Alava

GuipuzcoaVizcaya

NavarraLa Rioja

Huesca

TeruelZaragoza

Madrid

Avila

BurgosLeon

PalenciaSalamanca

Segovia

Soria

Valladolid

Zamora

Guadalajara

Barcelona

Girona

Lleida

Tarragona

Castellon

Ibiza

Mallorca

Menorca

Albacete

Ciudad Real

Cuenca

Toledo

Badajoz

Caceres

Alicante

Valencia

Almeria

Cadiz

Cordoba

Granada

Huelva

Jaen

Malaga

Sevilla Murcia

−4−3

−2−1

01

Popu

latio

n de

nsity

(geo

−clim

atic

con

trols

)

−80 −60 −40 −20 0Settlement density (geo−climatic controls)

Southern provinces Rest of Spain

present in both cases the most negative coefficient estimates among all European regions. Results

are revealing: the 7 regions with the lowest average settlement density are from Spain. Also, 9 out of

the 10 regions with lowest settlement density (and 16 out of 20) are from Spain. The same pattern

(with opposite sign) emerges if we look at population concentration (see Figure B.7 in Appendix B).

In contrast, the right panel suggests that this anomaly is not so prevalent in the case of population

density.

Figure 4 zooms in the differences across Spanish NUTS3 regions (provinces). In particular, Figure

4 plots the relationship between average population density (y-axis, in logs) and average settlement

density (x-axis) at the province level within Spain. On top of the strong positive association between

both measures, the plot shows that southern provinces (red triangles) present, for a given level of

settlement density, population densities higher than the rest of Spanish provinces (blue dots). This

is so even after accounting for geo-climatic factors. This pattern suggests that population is more

concentrated for a given level of settlement density in the Spanish south, which might have been a

limitation to economic development in the past, when the proximity to the farmland was important.


All in all, the evidence presented so far indicates that Spain has a remarkably low density of

settlements, even lower than Scandinavian countries. Moreover, geographic and climatic factors fail

to account for this anomaly.

3.3 History, warfare and the origins of the Spanish anomaly

Our results so far indicate that geographic and climatic factors do not satisfactorily explain Spain’s

population and settlement patterns, particularly in terms of settlement density and population con-

centration, which suggests that the peculiarities of Spanish history may be behind them together

with agglomeration forces. The Spanish anomaly is not a recent phenomenon. Already in the 17th

Figure 5: Persistence in settlement and population density.

10From the Venetian ambassadors Federico Cornaro (1678-81) and Giovanni Cornaro (1681-82), quoted in Brenan(1950: 128).

11To assess the magnitude of these correlations, it is important to bear in mind that the variable measuring densityof population entities in 1787 is not directly comparable to settlement density in 2011. The former captures thenumber of population entities in 1787 divided by surface area while the latter is the percentage of populated 10-km2cells. Similarly, population density in 1787 is not exactly comparable to the indicator in 2011 as the quality of thedata sources is very different. The implication is that correlations would be actually stronger if the methodology ofthe indicators were more similar. See Oto-Peralıas (2020) for more details.

century, European travelers were impressed by the scarcity of settlements: ”One can travel for days

on end without passing a house or village, and the country is abandoned and uncultivated”; ”Spain

gives the impression of being a desert of Libya, so unpopulated it is”.10

The left panel in Figure 5 shows that there is a high correlation between population density in

1787 and settlement density in 2011 (p = 0.65), while the right panel depicts an also strong correlation

between (log) population density in 1787 and (log) population density in 2011 (p = 0.69).11 This

points to the fact that the main features of Spain’s current spatial distribution of the population were

already present at the end of the 18th century, well before Spain began its (late) industrialization

process. Therefore, historical events taking place before this period are probably the responsible

factors.


On the theoretical side, Allen and Donaldson (2018) show that path dependence is important in

determining the level of economic activity across the space and their findings open up the potential

for historical accidents, such as warfare, to play an outsized role in governing where economic activity

occurs. In the case of Spain, previous research stresses the Reconquest as a key historical event in

this regard (Oto-Peralıas and Romero-Avila, 2016; Oto-Peralıas, 2020). Two factors related to the

Reconquest seem to be important here, namely, the pace or rate of the frontier expansion and military

instability due to frontier warfare. Oto-Peralıas and Romero-Avila (2016) argue that a fast frontier

expansion by the Christian kingdoms led to an imperfect colonization of the territory because the

material and human resources available were insufficient relative to the magnitude of the colonization

effort. The consequence was an occupation of the space characterized by a few settlements with large

jurisdictional areas. Consistent with this, the authors find a positive relationship between the rate

of Reconquest and the size of municipalities’ surface areas.

Oto-Peralıas (2020) emphasizes frontier warfare as a key determinant of the way the territory

was colonized by the Christian kingdoms. Military insecurity favored a colonization characterized

by the concentration of the population in a few well defended settlements, scarcity of population

and a livestock-oriented economy. Exploiting a discontinuity in military insecurity across the River

Tagus during the 11th to 13th centuries, Oto-Peralıas (2020) shows that the area south of the river,

more exposed to frontier warfare, has today much lower levels of settlement and population density

and higher population concentration. Interestingly, the difference in settlement density across the

Tagus is close to the whole difference between northern and southern Spain, suggesting that the

explanatory power of historical frontier warfare is important. The evidence supports the conjecture

by some historians linking the prolonged exposure to frontier warfare to population concentration

and scarcity of dispersed village communities (Bishko, 1975).

4 Narrowing down underpopulated areas within Spain

This section identifies the areas within Spain with the lowest settlement and population density

based on hot and cold spot analysis. It also investigates the characteristics of such areas at the

municipality level.

4.1 Hot and cold spot analysis

In order to identify clusters of low-density grid cells, we apply methods from spatial analysis to the

GEOSTAT 2011 population grid. In particular, we consider the hot and cold spot technique from

Getis and Ord (1992). This approach overcomes the border effect problem, that is, the allocation of

population in spatial units such as districts, municipalities or regions.

The so-called Getis-Ord statistic tests whether a grid cell and its neighboring grid cells form a

spatial cluster. In other words, hot and cold spots are detected as spatial outliers. In particular,


the hot (cold) spot is detected when a cell and its neighboring cells have similar values and higher

(lower) values than the average. More formally, the statistic for variable yi in grid cell i is given by:

G∗i (d) =

∑Nj=1 ωij(d)yj∑N

j=1 yj(2)

where yj refers to settlement or population density of grid cell j, N is the number of grid cells and

ωij denotes the ijth element of the spatial weight matrix as

ωij(d) = 1 if dij ≤ d ∀i, jωij(d) = 0 if dij > d ∀i, j

where d is the threshold distance.

Figure 6 plots the resulting hot and cold spots within Spain for both settlement density and

population density using d = 50km.12 According to the upper-left map in Figure 6, most of the

territory in the provinces of Badajoz, Caceres, Ciudad Real, Albacete, Cordoba and Jaen presents

the lowest settlement density within Spain (below the 10th percentile). In contrast, settlement

densities in Galicia, Paıs Vasco and Barcelona are the highest (above the 90th percentile). In the

upper-right map we control for geo-climatic variables, which alters the picture. Low density areas

of Extremadura and Castilla la Mancha as well as high density in Galicia and Asturias are partly

explained by these geo-climatic conditions, which are better in the North that is closer to the coast

and presents better soil quality and more rainfall (see Figure A.3 in Appendix A). Indeed, once

we account for the role of geo-climatic conditions both the Madrid13 and the Malaga areas appear

as the highest-density zones within Spain together with Barcelona. Contrarily, the areas with the

lowest settlement density belong not only to Ciudad Real and Albacete but also to Sevilla, Zaragoza,

Teruel, and Asturias.

A very similar picture arises when looking at the two bottom maps of Figure 6 based on hot

and cold spot analysis for (log) population density. The clusters of high population density after

accounting for geo-climatic conditions belong to the provinces of Madrid, Malaga and Barcelona.

The low-density areas are mainly located in Ciudad Real, Sevilla, Albacete, Zaragoza and Teruel.

The similar pictures arising from the hotspot analysis of settlement and population density are

not surprising since the correlation between the resulting G∗i (d) statistics for both indicators is 0.8.

We thus conclude that underpopulated areas within Spain identified from hotspot techniques are

somehow the same regardless on how we measure underpopulation, either in terms of settlement or

population density. Needless to say, the pattern is also the same with population concentration that

presents a correlation of -0.91 with the G∗i (d) statistic from settlement density (see Figure B.8 in

Appendix B).

12Figure A.4 in Appendix A shows the analysis based on d = 20km.13In Appendix C, we explore the settlement of density in each Spanish region compared with Madrid.


Figure 6: Hotspot analysis.

4.2 Characterizing underpopulated municipalities

In order to characterize the identified underpopulated areas within Spain, we need to exploit data at

the municipality level because economic, demographic, fiscal and political variables are only available

at this level. See Section 2.3 for more details on the municipality database as well as the descriptive

statistics in Table A.2 and the geographical distribution of the different characteristics in A.5.

In order to have a first glimpse at the data, we show in Figure 7 the simple bivariate correlations

between log population density and each of the considered municipality characteristics. In particular,

we group the municipality characteristics into four categories: demographic variables (dependency

ratio and share of inhabitants born in the municipality), rural exodus (population growth rate be-

tween 1950 and 1991), political variables (share of discontent votes in 2019 general elections and

increase in the share of votes to regionalist parties between 2007 and 2019), fiscal variables (public

spending per capita and public revenues per capita) and economic variables (share of population

employed in agriculture and log income per capita).

The patterns in Figure 7 suggest that low-density municipalities are characterized by longer

distances to the province capital, higher dependency ratios, higher shares of local-born population,

negative population growth over the 1950-1991 period, higher prevalence of regionalist vote but lower

prevalence of discontent vote, higher (lower) public spending (revenues) per capita, higher employ-

ment shares in agriculture and lower income per capita in nominal terms but higher in purchasing

power parity (PPP) terms.

In any case, the bivariate correlations shown in Figure 7 do not take into account partial correla-

tions among the different characteristics and they do not allow assessing either the relative strength


14Alternatively, one can also consider as dependent variables the G∗i (d) measures computed at the 250-km2 grid cell

level using surface-based weights to compute municipality-level measures. However, these measures present strongerspatial correlation by construction and significantly reduce variation in the data. Anyhow, unreported estimatesconfirm that the main findings in this section remain robust to this alternative specification.

Figure 7: Bivariate correlations.

Notes: This figure shows the bivariate correlation of (log) population density and different municipality characteristicsbased on a binned scatter plot. To generate the plot, we group population density into equally-sized bins, and computethe mean of each characteristic in each bin.

of the associations or the statistical significance of the correlations. In order to address these issues,

we consider the following empirical specification:

dmp = x′mpδ + w′

mpγ + θp + υmp (3)

where dmp refers to the (log) population density of municipality m in province p.14 xmp is a vector

of geographic and climatic variables for each municipality including temperature, rainfall, average

altitude, ruggedness, soil quality and distance to the coast. These geo-climatic controls are included

non-linearly as in specification (1) above. wmp is the vector of available municipality characteristics

that might characterize low density areas. Finally, a set of province dummies (θp) is included in some

specifications to exploit variation across municipalities within each province.

The vector γ includes our coefficients of interest associated to the different municipality charac-

teristics, namely, 65+ over 16-64 dependency ratio and share of inhabitants born in the municipality

(demographics), the population growth rate between 1951 and 1991 (rural exodus), the share of dis-

content votes in 2019 general elections and increase in the share of votes to regionalist parties between


2007 and 2019 (political variables), public spending per capita and public revenues per capita (fiscal

variables), and share of employment in agriculture and log income per capita (economic variables).

In addition to these variables, we include in all specifications the distance in km to the capital of the

province.

Table 2 shows the estimated coefficients for the model in equation (3). As expected, remoteness

is associated to lower population densities. In particular, a municipality that is 100 km away from

the province capital presents population densities around 1.0-1.9 log points lower. Note that we

control for geo-climatic variables in all regressions but their coefficients are not reported for the sake

of brevity (see Table A.3).

Table 2: Characteristics of low population density municipalities.

(1) (2) (3) (4) (5) (6) (7) (8)Baseline +Demographics +Rural Exodus +Politics +Fiscal +Economic +Fixed Effects +Prices

Distance to capital -0.0195*** -0.0150*** -0.0123*** -0.0121*** -0.0127*** -0.0120*** -0.0103*** -0.0121***(0.00299) (0.00209) (0.00161) (0.00166) (0.00170) (0.00181) (0.00176) (0.00199)

Dependency ratio -0.510*** -0.0192 0.0158 0.0526 -0.0426 0.00598 -0.0619(0.166) (0.106) (0.0962) (0.104) (0.162) (0.170) (0.194)

Share born in mun. -1.656*** -1.327*** -1.282*** -1.294*** -0.925*** -1.055*** -1.156***(0.426) (0.274) (0.265) (0.259) (0.245) (0.219) (0.224)

Pop. growth 1950-1991 0.755*** 0.749*** 0.741*** 0.731*** 0.530*** 0.513***(0.0452) (0.0491) (0.0556) (0.0506) (0.0344) (0.0361)

Regionalist vote growth 0.00324 0.00518 0.00394 -0.0128*** -0.0113**(0.00504) (0.00473) (0.00493) (0.00331) (0.00447)

Share discontent vote 0.00475 0.00724 0.00845 0.0174*** 0.0177***(0.00631) (0.00600) (0.00623) (0.00351) (0.00445)

Public spending pc -0.0267 -0.139* -0.194*** -0.221**(0.0472) (0.0735) (0.0711) (0.101)

Public revenues pc 0.00607 0.0178 0.0530 0.101(0.0440) (0.0534) (0.0411) (0.0652)

Share agriculture -2.187*** -1.705*** -1.923***(0.606) (0.462) (0.590)

Income pc (log) 0.608** 0.445**(0.236) (0.186)

Income pc (PPP, log) -0.0566(0.0337)

Observations 8,007 8,005 7,997 7,995 7,222 6,042 6,042 4,662R-squared 0.551 0.597 0.664 0.665 0.670 0.664 0.726 0.718Geo controls YES YES YES YES YES YES YES YESProvince FE NO NO NO NO NO NO YES YES

Notes: Dependent variable is (log) population density. Geo-climatic controls are not reported but included in allcolumns, namely, temperature, rainfall, average altitude, ruggedness, soil quality and distance to the coast. Standarderrors clustered at the province level. *** p<0.01, ** p<0.05, * p<0.1

Regarding demographic variables added in column (2) of Table 2, we see that low density areas

are associated to ageing populations in which the dependency ratio is higher, and a larger share of

local-born inhabitants. As expected, these patterns indicate that underpopulated areas have received

less immigrants and/or have experienced larger emigration flows in the recent past.

A relevant question related to past migration flows refers to the so-called rural exodus. While

the total number of rural inhabitants in 1940 remained about the same as it had been in 1900,

during the phase from 1950 to 1991 large migration flows from agricultural areas to the leading

industrial regions took place in Spain (see e.g. Collantes and Pinilla, 2011). In column (3), we

include population growth from 1950 to 1991 as a proxy for the rural exodus at the municipality


level. The positive and strong association with population density indicates that underpopulated

municipalities in 2011 are those that experienced population losses during the rural exodus. Indeed,

once we account for this characteristic, the dependency ratio is not statistically significant, which

suggests that young inhabitants leaving rural municipalities during the 1950-1991 period account for

the current dependency ratio of those rural and underpopulated municipalities.

The inclusion of political variables in column (4) indicates that less densely populated municipal-

ities are not associated to higher prevalence of discontent parties, namely, VOX and Podemos, which

is more prevalent in high-density and urban areas. Note that the regressor is the share of discontent

vote in November 2019 General Elections, which is exactly the same to the increase in discontent

vote since 2007, given that the share of Podemos and VOX was zero in all municipalities in that year.

The increase in the support to regionalist parties between 2007 and 2019 is not associated either to

low density areas. It has increased in low density municipalities within each province (see column

7).

Turning to fiscal variables in column (5) of Table 2, the estimates suggest that low-density areas

are characterized by higher spending per capita in the local budgets. Note however, that the asso-

ciation with public revenue per capita is much weaker. These estimates suggest that, if anything,

underpopulated areas in Spain are net recipients of public funds through local budgets. In any event,

note that this finding must be interpreted with a grain of salt since public expenditures and revenues

here do not include the components from regional and central government transfers.

In terms of economic characteristics included in column (6), low-density areas in Spain belong to

municipalities characterized by higher shares of population employed in agriculture as well as lower

average income per capita (note that the number of observations is lower when including income per

capita since this information from INE is only available for municipalities above 100 inhabitants).

These associations are highly significant and robust to the inclusion of province fixed effects in column

(7) of Table 2.

In column (8), we re-estimate the specification in (7) but substituting nominal income per capita

by PPP-adjusted income per capita. In particular, we normalize nominal income by an index of

housing prices available at the municipality level. In particular, we use the census micro-data on

real estate transactions provided by the Spanish Ownership Registry (Registro de la Propiedad) to

the Banco de Espana since 2004. Interestingly enough, the positive and significant association turns

negative and significant! Despite a note of caution is in place since the purchasing power parity

(PPP) adjustment considered here refers only to housing prices, this finding suggests that in PPP

terms, inhabitants of low-density areas present indeed higher incomes or at least, they do not present

significantly lower incomes. Finally, it is worth noting that the number of observations is lower than

in column (7) because housing prices data is only available for municipalities with more than 25

mortgages in the year 2011 in order to ensure good quality of the price index. Still, if we re-estimate

the specification with nominal income per capita in column (7) for the subsample in column (8) the

estimates remain virtually unaltered.


Column (1) in Table 3 refers to the baseline specification from column (6) in Table 2. Low

density municipalities are associated to lower shares of discontent vote (Podemos and VOX) and

higher increases in the regionalist vote between 2007 and 2019. Column (2) shows that these areas

are also associated to higher prevalence of regionalism voting in levels. Interestingly enough, this

bias towards regionalist parties in low density municipalities is robust to the exclusion of Catalan

15Note that voting information refers to the 2019 General Election and the changes with respect to to the 2007election.

In order to gauge the relative strength of the different statistical associations reported in Table 2,

we also estimate the corresponding partial correlations between population density and each regressor

from column (7) by standardizing the variables. These figures are -0.15 for the distance to the

province capital; -0.001 and -0.11 for the dependency ratio and the share of local-born inhabitants;

0.25 for population growth during the rural exodus (1950-1991); -0.09 and 0.10 for the increase in

regionalist vote and the share of discontent vote; -0.10 and 0.03 for public spending and public revenue

per capita; -0.06 and 0.05 for the share of agriculture employment and nominal income per capita. We

also decomposed the goodness-of-fit (R2) of the model in column (7) into contributions of individual

regressor variables (see e.g. Chevan and Sutherland, 1991). After accounting for geo-climatic factors

and province fixed effects, the highest contribution to the overall R2 comes from 1950-1991 population

growth (32%), distance to the capital (18%), the share of local-born inhabitants (15%), the share of

discontent vote (9%), and the agriculture employment share (8%). The remaining regressors jointly

account for around 15% of the overall R2.

All in all, the strongest associations arise between population density and 1950-1991 population

growth (+), distance to the province capital (-), the share of local-born inhabitants (-), the share of

discontent vote (+), and the share of employment in agriculture (-). Needless to say, these estimates

must be interpreted with caution since they refer to mere correlations. While these correlations are

useful for characterizing low density areas in Spain, we acknowledge that identifying the direction of

causality is beyond the scope of the present paper.

Table 3 presents specifications further exploring the voting patterns in low-density municipali-

ties.15 The process of urbanization and globalization have manifested political discontent in rural

and low-density areas of countries such as the US or the UK, which gives rise to populism reactions

in the so-called revenge of the places that don’t matter (see Rodrıguez-Pose 2018).

municipalities from the sample as shown in columns (3) and (4).

Column (5) indicates that the urban bias of discontent vote in Spain is mainly driven by left-wing

parties (Podemos) while the association between right-wing discontent vote (VOX) and population

density is not statistically significant. Note that both Podemos and VOX were not present in the

2007 General Election so that including their voting shares in levels is equivalent to changes from

2007 to 2019.

Finally, in columns (6) and (7) we include an index of vote fragmentation (as in Sanz, Sole-Olle

and Sorribas-Navarro 2019) both in levels and 2007-2019 changes. On the one hand, there is no


Table 3: Characteristics of low population density municipalities — Political variables.

(1) (2) (3) (4) (5) (6) (7)Baseline Regional Without Without Left-right Fragmentation Fragmentation

specification vote Catalonia Catalonia discontent index increase

Distance to capital -0.0103*** -0.0103*** -0.00948*** -0.00942*** -0.0104*** -0.0103*** -0.0103***(0.00176) (0.00177) (0.00181) (0.00181) (0.00175) (0.00176) (0.00176)

Dependency ratio 0.00598 -0.00170 0.0767 0.0732 0.0267 -0.00416 0.00673(0.170) (0.168) (0.164) (0.167) (0.170) (0.168) (0.170)

Share born in mun. -1.055*** -1.042*** -1.109*** -1.108*** -1.098*** -1.025*** -1.058***(0.219) (0.221) (0.254) (0.257) (0.220) (0.222) (0.220)

Pop. growth 1950-1991 0.530*** 0.535*** 0.539*** 0.538*** 0.530*** 0.533*** 0.533***(0.0344) (0.0370) (0.0390) (0.0408) (0.0343) (0.0350) (0.0351)

Regionalist vote growth -0.0128*** -0.0116** -0.0119*** -0.0131***(0.00331) (0.00445) (0.00347) (0.00351)

Share discontent vote 0.0174*** 0.0173*** 0.0184*** 0.0186*** 0.0152*** 0.0179***(0.00351) (0.00362) (0.00352) (0.00363) (0.00324) (0.00346)

Public spending pc -0.194*** -0.187** -0.237** -0.232** -0.199*** -0.196*** -0.194***(0.0711) (0.0714) (0.0929) (0.0928) (0.0701) (0.0711) (0.0714)

Public revenues pc 0.0530 0.0494 0.0715 0.0677 0.0549 0.0535 0.0523(0.0411) (0.0411) (0.0565) (0.0563) (0.0401) (0.0413) (0.0413)

Share agriculture -1.705*** -1.676*** -1.984*** -1.976*** -1.565*** -1.685*** -1.721***(0.462) (0.470) (0.468) (0.471) (0.455) (0.457) (0.450)

Income pc (log) 0.445** 0.460** 0.318 0.318 0.449** 0.431** 0.445**(0.186) (0.195) (0.197) (0.201) (0.182) (0.190) (0.186)

Share regionalist vote -0.00470 -0.00118(0.00378) (0.00235)

Share VOX 0.00419(0.00463)

Share Podemos 0.0270***(0.00373)

Fragmentation 0.0430(0.0327)

Fragmentation growth -0.0139(0.0319)

Observations 6,042 6,042 5,169 5,169 6,042 6,041 6,041R-squared 0.726 0.725 0.690 0.689 0.728 0.726 0.726Geo controls YES YES YES YES YES YES YESProvince FE YES YES YES YES YES YES YES

Notes: Dependent variable is (log) population density. Geo-climatic controls are not reported but included in allcolumns, namely, temperature, rainfall, average altitude, ruggedness, soil quality and distance to the coast. Standarderrors clustered at the province level. *** p<0.01, ** p<0.05, * p<0.1, + <0.15.

statistical association between population density and political fragmentation in levels. On the other

hand, the increase in political fragmentation is indeed lower in low-density areas, which is related

to the rise of the discontent vote (VOX and Podemos) more concentrated in urban areas. All in

all, there is no clear evidence on any association between low density areas and the rise of political

fragmentation through discontent vote. Still, we find some evidence that the rise in regionalist vote

between 2007 and 2019 might have been stronger in low-density municipalities.

Table 4 includes two additional controls: (log) area in km2 as well as a rural dummy for mu-

nicipalities with less than 10,000 inhabitants, which is a commonly-used threshold in Spain to label

rural areas (see Collantes and Pinilla 2011). The relationship between population density and these

two indicators is somehow tautological, but it may well be that some of the characteristics explored

above are related to these variables rather than to (log) population density per se. For instance,


16Also, Table B.4 in Appendix B.4 shows the estimates for population concentration.

public spending per capita or the share of employment in agriculture may be higher in low-density

municipalities because they present larger areas or they are smaller in terms of population size.

According to the results in Table 4, the sign, magnitude, and statistical significance of all the

estimates in our baseline Table 2 remain robust when we control for land area and population size.

Also, not surprisingly, we observe that rural and larger municipalities in terms of surface generally

belong to low density zones within Spain. In particular, rural municipalities with less than 10,000

inhabitants present 0.4-1.3 log points lower density on average, which represents around one standard

deviation of the dependent variable (see Table A.2); also, since both variables are expressed in logs,

the estimates in Table 4 indicate that a 1 percent increase in land area is associated to a reduction

of around -0.4% in population density.

Table 5 shows the analogous estimates for settlement density at the municipality level as the

dependent variable. All the specifications are analogous to those of Table 2 but we simply replace

the dependent variable.16

While the statistical association with the share of local-born inhabitants remains strong and

significant, the one between settlement density and the dependency ratio is positive in general.

Table 4: Characteristics of low population density municipalities — Land area and population.


Distance to capital -0.0155*** -0.0121*** -0.0106*** -0.0103*** -0.0107*** -0.0106*** -0.00943*** -0.0110***(0.00225) (0.00172) (0.00139) (0.00142) (0.00144) (0.00151) (0.00165) (0.00189)

Area (log) -0.436*** -0.410*** -0.395*** -0.400*** -0.427*** -0.448*** -0.383*** -0.415***(0.0487) (0.0463) (0.0434) (0.0436) (0.0378) (0.0363) (0.0393) (0.0388)

Rural -1.281*** -1.155*** -0.489*** -0.476*** -0.495*** -0.421*** -0.402*** -0.390***(0.162) (0.128) (0.0847) (0.0888) (0.0870) (0.0800) (0.0734) (0.0733)

Dependency ratio -0.734*** -0.360*** -0.309*** -0.287*** -0.499*** -0.376** -0.525***(0.148) (0.101) (0.0909) (0.0930) (0.144) (0.148) (0.172)

Share born in mun. -0.879** -0.561** -0.519* -0.514** -0.0683 -0.385* -0.357*(0.374) (0.267) (0.272) (0.247) (0.230) (0.198) (0.201)

Pop. growth 1950-1991 0.696*** 0.690*** 0.663*** 0.633*** 0.451*** 0.429***(0.0437) (0.0432) (0.0475) (0.0444) (0.0294) (0.0370)

Regionalist vote growth 0.00278 0.00505 0.00336 -0.0127*** -0.0117***(0.00486) (0.00423) (0.00456) (0.00318) (0.00398)

Share discontent vote 0.00715 0.0100* 0.0100* 0.0184*** 0.0174***(0.00618) (0.00545) (0.00587) (0.00314) (0.00388)

Public spending pc -0.0585 -0.152** -0.184** -0.205**(0.0452) (0.0705) (0.0691) (0.0955)

Public revenues pc -0.00610 0.0176 0.0476 0.103(0.0376) (0.0514) (0.0420) (0.0634)

Share agriculture -1.963*** -1.528*** -1.683***(0.519) (0.448) (0.560)

Income pc (log) 0.443** 0.350**(0.213) (0.173)

Income pc (PPP, log) -0.0875***(0.0315)


Notes: Dependent variable is (log) population density. Geo-climatic controls are not reported but included in allcolumns, namely, temperature, rainfall, average altitude, ruggedness, soil quality and distance to the coast. Standarderrors clustered at the province level. *** p<0.01, ** p<0.05, * p<0.1


Note that the association between the dependency ratio and population density in Table 2 became

insignificant once we accounted for population growth during the 1950-1991 rural exodus in Spain.

Regarding the political indicators, we observe a positive (negative) relationship between settlement

density and the share of discontent vote (increase in regionalist vote) once we include province

dummies in columns (6) and (7).

Turning to the fiscal characteristics of areas with low settlement density, we observe the same

pattern as in Table 2: once we control for socio-economic characteristics and/or province fixed

effects in columns (5) and (6), low settlement density areas belong to municipalities with higher

public spending per capita but not higher public revenues per capita. Indeed, the partial correlation

between settlement density and public spending per capita is -0.15 and statistically significant while

that of public revenues is +0.04 but insignificant.

Finally, looking at columns (6) and (7) of Table 5, we conclude that lower settlement density

is also associated to higher agriculture employment and lower income per capita in nominal terms.

Table 5: Characteristics of low settlement density municipalities.


Distance to capital -0.238*** -0.202*** -0.185*** -0.195*** -0.195*** -0.169*** -0.0821** -0.0935**(0.0364) (0.0342) (0.0333) (0.0341) (0.0347) (0.0384) (0.0351) (0.0389)

Dependency ratio 5.228** 8.175*** 6.585*** 8.182*** 12.64*** 7.269*** 7.111**(2.314) (2.469) (2.427) (2.435) (3.413) (2.543) (3.435)

Share born in mun. -27.60*** -25.71*** -24.48*** -25.50*** -23.16*** -26.33*** -28.64***(6.075) (5.525) (5.042) (4.707) (4.660) (3.624) (3.908)

Pop. growth 1950-1991 4.530*** 4.804*** 4.316*** 4.489*** 1.747** 1.961**(0.812) (0.766) (0.765) (0.746) (0.739) (0.732)

Regionalist vote growth 0.0443 0.0883 0.0304 -0.108* -0.0769(0.128) (0.115) (0.120) (0.0551) (0.0719)

Share discontent vote -0.194 -0.155 -0.142 0.119** 0.131*(0.122) (0.106) (0.120) (0.0539) (0.0693)

Public spending pc -0.520 -3.467*** -2.864*** -2.315**(0.889) (1.188) (0.924) (1.140)

Public revenues pc -0.898 3.29e-05 0.971 0.776(0.640) (0.957) (0.676) (0.748)

Share agriculture -24.12** -11.21 -18.36**(10.54) (6.757) (8.534)

Income pc (log) 14.99*** 6.815**(4.225) (2.899)

Income pc (PPP, log) -0.140(0.423)


Notes: Dependent variable is the settlement density indicator. Geo-climatic controls are not reported but includedin all columns, namely, temperature, rainfall, average altitude, ruggedness, soil quality and distance to the coast.Standard errors clustered at the province level. *** p<0.01, ** p<0.05, * p<0.1

However, if we consider PPP-adjusted income per capita with the only price index available at

the municipality level (i.e. housing prices), the association between settlement density and average

income becomes negative but not significant.


In sum, the main characteristics of low population density municipalities identified from the

analysis in Table 2 also correspond to low settlement density municipalities.

5 Concluding remarks

Using the GEOSTAT population grid with information at the 1-km2 level, this article uncovers an

anomaly in the Spanish distribution of population across the space. Compared to other European

countries, Spain presents the lowest density of settlements (almost 90% of its territory is uninhabited!)

together with the highest population concentration in certain areas. Interestingly enough, these

patterns remain true even after accounting for geographical and climatic conditions. The article also

identifies the areas with the lowest densities within Spain and characterize the municipalities that

17In related projects, Gutierrez et al. (2020) analyze in detail the evolution over the last decades of rural-urbanmigrations in Spain and Budı (2020) explores the macroeconomic consequences of such migration flows.

belong to these virtually uninhabited parcels of land.

Economies of agglomeration rely on increasing returns to scale and lower transportation costs,

and emphasize linkages between firms and suppliers as well as between firms and consumers. The

emptiness of the Spanish territory may thus influence the distribution of economic activity across

firms, sectors and space. The dynamic dimension related to the interaction between internal migra-

tion flows and the process of structural change of the Spanish economy over the last decades may

also shape current economic outcomes.17 Understanding the economic consequences of the Spanish

anomaly in settlement patterns represents an exciting line of open research.


References

[1] Allen, T., and C. Arkolakis (2014) Trade and the Topography of the Spatial Economy. The

Quarterly Journal of Economics, 129(3), 1085-1140.

[2] Allen, T., and D. Donaldson (2018) The geography of path dependence. Mimeo.

[3] Bishko, C. (1975) The Spanish and Portuguese Reconquest 1095-1492, in Kenneth M. Setton

(ed.) A History of the Crusades, Vol. 3: 396-456.

[4] Brenan, G. (1950) The Spanish Labyrinth: An Account of the Social and Political Background

of the Spanish Civil War. Cambridge University Press, Cambridge.

[5] Budı, T. (2020) Rural Depopulation, Urbanization, and Structural Change. CEMFI Master The-

sis.

[6] Chevan, A., and M. Sutherland (1991) Hierarchical partitioning. American Statistician, 45, 90–96.

[7] Duranton, G., and D. Puga (2020) The economics of urban density. NBER WP 27215.

[8] Duranton, G., and M. Turner (2018) Urban form and driving: Evidence from US cities. Journal

of Urban Economics, 108, 170-191.

[9] Collantes, F. and V. Pinilla (2011) Peaceful Surrender: The Depopulation of Rural Spain in the

Twentieth Century. Cambridge Scholars Publishing.

[10] Eurostat (2016) GEOSTAT 2011 population grid. Available at http://ec.europa.eu/eurostat/.

[11] Fischer, G., F. Nachtergaele, S. Prieler, H. van Velthuizen, L. Verelst, and D. Wiberg (2008)

Global agro-ecological zones assessment for agriculture, IIASA, Laxenburg, Austria and FAO,

Rome, Italy.

[12] Getis, A. and J. Ord (1992) The analysis of spatial association by use of distance statistics.

Geographical Analysis, 24, 189–206.

[13] Gonzalez-Jimenez, M. (1992) Frontier and Settlement in the Kingdom of Castile (1085-1350). In

Medieval Frontier Societies, edited by Bartlett Robert, MacKay Angus. Clarendon Press, Oxford,

pp. 49-74.

[14] Gutierrez, E., E. Moral-Benito, and R. Ramos (2020) Tendencias recientes de la poblacion en

areas rurales y urbanas en Espana. Mimeo.

[15] Harari, M. (2016) Cities in bad shape: Urban geometry in India. Processed, Wharton School of

the University of Pennsylvania.


[16] Hijmans, R., S. Cameron, J. Parra, P. Jones, and A. Jarvis (2005) Very high resolution interpo-

lated climate surfaces for global land areas. International Journal of Climatology, 25, 1965-1978.

[17] Krugman, P. (1991) Increasing returns and economic geography. Journal of Political Economy,

99, 483-499.

[18] Oto-Peralıas, D. (2020) Frontiers, warfare and economic geography: The case of Spain. Journal

of Development Economics, 146, 1-19.

[19] Oto-Peralıas, D. and D. Romero-Avila (2016) The Economic Consequences of the Spanish Re-

conquest: The Long-term Effects of Medieval Conquest and Colonization. Journal of Economic

Growth, 21, 409-464

[20] Rae, A. (2018) Think your country is crowded? These maps reveal the truth about population

density across Europe, The Conversation, January 23, 2018.

[21] Rodrıguez-Pose, A. (2018) The revenge of the places that don’t matter (and what to do about

it). Cambridge Journal of Regions, Economy and Society, 11, 189-209.

[22] Sanz, C., A. Sole-Olle and P. Sorribas-Navarro (2019) Betrayed by the Elites: How Corruption

Amplifies the Political Effects of Recessions, mimeo.

[23] Sole-Olle, A. and E. Viladecans-Marsal (2012) Lobbying, political competition, and local land

supply: recent evidence from Spain. Journal of Public economics, 96, 10-19.


A Additional figures and tables

Table A.1: Descriptive statistics at the grid-cell level.

Europe Spain

(1) (2) (3) (4) (5) (6) (7) (8)mean p25 p50 p75 mean p25 p50 p75

Settlement density 72.34 50.00 88.00 100 45.00 24.00 44.00 68.00Pop. concentration 38.00 20.50 31.14 49.27 61.41 44.21 60.58 79.70Pop. density 114.6 5.545 29.05 87.65 133.3 4.920 14.12 51.68Temperature 8.124 5.436 8.505 11.03 13.25 11.41 13.40 15.53Rainfall 7.741 5.842 6.819 8.772 6.222 4.567 5.452 6.953Altitude 371.0 96.75 222.1 516.9 652.7 334.6 651.2 885.0Ruggedness 87.40 17.69 46.76 121.0 109.8 46.63 85.42 145.0Soil quality 8.726 8.429 9 9.429 8.905 8.429 8.934 9.357Dist. coast 134.6 20.02 89.79 205.8 119.4 40.95 109.4 183.6

Table A.2: Descriptive statistics at the municipality level.

(1) (2) (3) (4) (5)VARIABLE N mean p10 p90 sd

Settlement density 8,023 55.20 24 91.67 23.85Pop. concentration 8,024 54.37 32.63 77.72 17.26Pop. density (log) 8,024 3.064 1.204 5.389 1.656Regionalist vote growth 8,108 3.105 -5.051 22.74 12.04Share discontent vote 8,131 18.26 3.823 30.22 9.420Distance to capital 8,113 44.14 15.95 76.41 24.38Share agriculture 7,203 0.0715 0.00786 0.155 0.0604Dependency ratio 8,116 0.505 0.212 0.893 0.312Share born in mun. 8,114 0.466 0.216 0.694 0.177Population growth 1950-1991 8,104 -0.547 -1.371 0.414 0.777Public spending pc 7,338 1.222 0.652 1.938 0.841Public revenues pc 7,338 1.307 0.701 2.101 0.897Income pc (PPP, log) 5,163 10.26 9.655 10.94 0.557Income pc (log) 6,745 9.197 8.924 9.467 0.208


Notes: Standard errors clustered at the country level. *** p<0.01, ** p<0.05,* p<0.1

Table A.3: The role of geography and climate in population patterns.

(1) (2) (3) (4)Sett. density Pop. density Sett. density Pop. density

Temperature pt2 16.27** 0.891*** 18.37*** 0.733(5.445) (0.250) (4.977) (0.519)

Temperature pt3 21.13*** 1.258*** 21.07*** 0.644(5.920) (0.193) (5.666) (0.661)

Temperature pt4 18.47*** 1.667*** 11.18 0.0151(5.647) (0.226) (8.393) (0.896)

Rainfall pt2 5.312*** 0.294 1.840 0.0726(1.666) (0.293) (2.424) (0.160)

Rainfall pt3 10.16*** 0.482 4.346 0.0786(1.987) (0.270) (3.760) (0.296)

Rainfall pt4 9.210** 0.418 2.685 -0.502(2.971) (0.250) (4.094) (0.329)

Elevation pt2 -1.898 -0.443** 3.958* -0.271**(1.448) (0.144) (1.982) (0.130)

Elevation pt3 -8.216*** -1.313*** -3.408 -1.120***(2.592) (0.199) (5.353) (0.406)

Elevation pt4 -11.81*** -1.719*** -19.41** -3.098***(2.665) (0.221) (8.857) (0.825)

Ruggedness pt2 3.419** 0.318*** 5.502*** 0.455***(1.179) (0.0437) (1.289) (0.0647)

Ruggedness pt3 5.047*** 0.427** 6.748** 0.694***(0.714) (0.149) (2.666) (0.206)

Ruggedness pt4 4.423 0.575*** 7.821* 1.366**(2.645) (0.154) (4.443) (0.545)

Soil quality pt2 5.256* 0.208* 7.906*** 0.585***(2.444) (0.0966) (2.080) (0.202)

Soil quality pt3 7.071** 0.288** 9.066*** 0.566**(2.903) (0.120) (2.412) (0.270)

Soil quality pt4 7.562** 0.388** 9.869*** 0.667*(2.862) (0.127) (3.060) (0.341)

Dist coast pt2 -2.475 -0.339* -10.12*** -1.221***(2.811) (0.183) (1.504) (0.193)

Dist coast pt3 -8.852 -0.466 -10.57*** -0.946***(7.091) (0.351) (2.735) (0.175)

Dist coast pt4 -7.531 -0.114 -9.175*** -0.738***(5.606) (0.273) (3.264) (0.226)

Island -1.631 -0.572*** 2.593 -0.289(2.852) (0.140) (2.866) (0.265)

Latitude 26.59*** 1.472*** 18.39*** 0.975***(3.055) (0.169) (4.973) (0.207)

Longitude 1.475 0.111 0.152 0.111(2.071) (0.164) (1.344) (0.109)

Latitude2 -0.278*** -0.0163*** -0.192*** -0.0111***(0.0322) (0.00161) (0.0468) (0.00228)

Longitude2 -0.00970 -0.00493*** -0.0230 -0.00320*(0.0272) (0.00155) (0.0212) (0.00168)

Latitude longitude -0.0143 -0.000385 0.0237 -0.000317(0.0465) (0.00323) (0.0292) (0.00240)

Observations 10,850 10,854 21,172 21,190R-squared 0.697 0.585 0.720 0.681Sample Eurozone Eurozone Europe Europe


Figure A.2: Cumulative distribution of the difference between the coefficient on Spain and on eachvirtual country from 1,000 regressions.

0.2

.4.6

.81

Cum

ulat

ive

dist

ribut

ion

10 20 30 40 50Difference between the Placebo coefficient and the ’true’ coefficient

Settlement density0

.2.4

.6.8

1C

umul

ativ

e di

strib

utio

n

−1 0 1 2Difference between the Placebo coefficient and the ’true’ coefficient

log Population density

Figure A.1: The Spanish anomaly in settlement density (Europe).

ALATBEBGBI

CHCZDEDKEEELFI

HRHUIEISIT

KOLI

LTLULVMENLNOPLPTROSESI

SKSRUKES

−100 0 100 200−100 0 100 200

Baseline Geo−climatic controlsALATBEBGBI

CHCZDEDKEEELFI

HRHUIEISIT

KOLI


SKSRUKES

−10 −5 0 5 10 −10 −5 0 5 10


Figure A.3: Hotspot analysis for geo-climatic variables.


Figure A.4: Hotspot analysis (20 km).

Figure A.5: Municipality characteristics.


B Population concentration results

Alternatively to settlement density and population density, we also follow Oto-Peralıas (2020) and

consider an indicator of population concentration that measures the percentage of the population

living in the most populated one percent of the territory.

Figure B.6: The Spanish anomaly in population concentration.

AT

BE

DE

FI

IE

IT

LU

NL

PT

UK

ES

−50 0 50 −50 0 50

Baseline Geo−climatic controlsALATBEBGBI

CHCZDEDKEEELFI

HRHUIEISIT

KOLI


SKSRUKES

−100 −50 0 50 −100 −50 0 50


Figure B.7: The Spanish anomaly in population concentration by region.

ALB0

ALB1

ALB2ALB3ALB4

ALB5ALB6

ALB7

ALB8

ALB9

ALB99

AT111

AT112

AT113AT121AT122

AT123AT124

AT125

AT126AT127

AT130

AT211AT212

AT213

AT221

AT222AT223

AT224AT225

AT226

AT311AT312

AT313

AT314

AT315

AT321AT322

AT323

AT331AT332

AT333

AT334

AT335

AT341AT342

BE100

BE211

BE212BE213

BE221

BE222

BE223BE231BE232

BE233

BE234BE235BE236BE241BE242BE251BE252

BE253

BE254

BE255

BE256BE257

BE258BE310

BE321

BE322

BE323

BE324

BE325

BE326

BE327

BE331BE332BE334

BE335BE336

BE341BE342BE343

BE344

BE345BE351

BE352

BE353

BG311BG312

BG313BG314

BG315

BG321

BG322

BG323

BG324

BG325

BG331

BG332

BG333

BG334

BG341BG342BG343

BG344

BG411

BG412

BG413

BG414

BG415BG421BG422

BG423

BG424

BG425

BIH10

BIH11

BIH12

BIH13

BIH14

BIH15BIH16BIH17 BIH18

BIH19

BIH20

BIH21

BIH22BIH23

BIH24

BIH97

BIH98 BIH99

CH011

CH012

CH013

CH021

CH022

CH023

CH024CH025

CH032CH033CH040

CH051

CH052

CH053

CH054CH055

CH056

CH057CH061

CH062

CH063CH064

CH066

CH070

CZ010

CZ020


CZ051CZ052CZ053CZ063CZ064CZ071

CZ072CZ080

DE111DE112DE113

DE114


DE119DE11ADE11B

DE11CDE11D


DE126

DE127

DE128

DE129DE12A

DE12B

DE12C

DE131

DE132

DE133

DE134

DE135


DE13ADE141

DE142DE143DE145DE146DE147DE148DE149

DE212

DE213

DE214

DE215

DE216

DE217


DE21C

DE21D

DE21E

DE21F

DE21G

DE21H

DE21IDE21J

DE21K

DE21L

DE21M

DE21N

DE222

DE223DE224DE225


DE22ADE22B

DE22C

DE232

DE234DE235DE236

DE237

DE238DE239

DE23A

DE241

DE245DE246

DE247DE248

DE249DE24A

DE24B

DE24C

DE24DDE251

DE254DE255

DE256DE257

DE258

DE259

DE25ADE25B

DE25C

DE262DE264


DE269DE26ADE26BDE26C

DE271


DE277DE278DE279

DE27A

DE27BDE27C

DE27D

DE27E

DE300

DE401DE402DE403

DE404

DE405DE406DE407DE408DE409DE40ADE40BDE40CDE40D

DE40E

DE40FDE40G

DE40HDE40I

DE501DE502

DE600

DE711DE712

DE715DE716DE717

DE718

DE719DE71A

DE71B

DE71CDE71DDE71EDE721

DE722DE723

DE724DE725

DE731

DE732DE733DE734

DE735

DE736DE737

DE801DE802

DE803

DE804DE807DE808

DE809

DE80A

DE80B

DE80C

DE80DDE80E

DE80FDE80G

DE80HDE80I

DE911DE912DE913

DE914DE915DE916

DE917DE918

DE919

DE91ADE91B

DE922DE923

DE925

DE926

DE927

DE928DE929

DE931

DE932DE933

DE934

DE935DE936

DE937DE938

DE939DE93ADE93B

DE944DE945DE946

DE947

DE948DE949

DE94A

DE94B

DE94CDE94DDE94EDE94FDE94G

DE94H

DEA11DEA12DEA13

DEA15

DEA1B

DEA1CDEA1D

DEA1EDEA1F

DEA22DEA23

DEA26

DEA27

DEA28DEA29

DEA2ADEA2B

DEA2C

DEA2D

DEA32

DEA33

DEA34DEA35

DEA36DEA37DEA38

DEA41

DEA42DEA43

DEA44

DEA45 DEA46DEA47

DEA52

DEA53DEA54

DEA55DEA56

DEA57

DEA58DEA59DEA5ADEA5B

DEA5CDEB12

DEB13

DEB14DEB15DEB16DEB17

DEB18

DEB19DEB1A

DEB1BDEB21

DEB22

DEB23

DEB24DEB25DEB32

DEB35DEB36

DEB3A

DEB3BDEB3C

DEB3DDEB3EDEB3F

DEB3G

DEB3HDEB3J

DEB3K

DEC01DEC02

DEC03

DEC04

DEC05DEC06DED21

DED2C

DED2DDED2E

DED2F

DED41

DED42DED43

DED44

DED45DED51

DED52DED53

DEE01

DEE03

DEE04DEE05DEE06DEE07

DEE08

DEE09

DEE0ADEE0BDEE0C

DEE0D

DEE0E

DEF03

DEF05DEF06

DEF07

DEF08

DEF09

DEF0A

DEF0BDEF0C

DEF0DDEF0E

DEF0F

DEG01DEG02DEG03DEG04

DEG05DEG06DEG07

DEG09DEG0A

DEG0B

DEG0C

DEG0D

DEG0E

DEG0FDEG0G

DEG0HDEG0IDEG0JDEG0KDEG0LDEG0MDEG0P

DK011

DK012DK013

DK014

DK021

DK022DK031

DK032DK041DK042

DK050

EE001EE004EE006

EE007

EE008

EL111

EL112EL113

EL114

EL115EL121

EL122


EL127EL131EL132

EL133

EL134

EL141

EL142

EL143

EL144

EL211

EL212

EL213EL214

EL221

EL222EL223EL224 EL231

EL232

EL233

EL241

EL242

EL243

EL244

EL245EL251

EL252EL253

EL254

EL255 EL300EL411EL412

EL413

EL421

EL422

EL431

EL432

EL433EL434FI193

FI194FI195

FI196FI197FI1B1FI1C1

FI1C2FI1C3

FI1C4FI1C5FI1D1

FI1D2

FI1D3FI1D4

FI1D5FI1D6

FI1D7

FI200

FR103FR104

FR105FR106FR107

FR108

FR211FR212

FR213FR214


FR241FR242

FR243

FR244

FR245FR246FR251

FR252

FR253FR261

FR262FR263

FR264

FR301FR302

FR411

FR412


FR431FR432FR433

FR434 FR511FR512

FR513FR514FR515FR521FR522FR523FR524FR531FR532FR533FR534

FR611

FR612

FR613

FR614

FR615FR621FR622

FR623FR624

FR625

FR626

FR627FR628


FR713FR714

FR715FR716

FR717

FR718 FR721FR722

FR723FR724

FR811

FR812

FR813FR814


FR824FR825FR826

FR831

FR832HR031

HR032HR033

HR034HR035

HR036HR037

HR041

HR042

HR043

HR044

HR045HR046

HR047

HR048HR049

HR04AHR04BHR04C

HR04DHR04E

HU101

HU102

HU211HU212

HU213HU221HU222HU223

HU231

HU232HU233

HU311HU312HU313

HU321HU322

HU323

HU331HU332HU333

IE011IE012IE013IE021 IE022

IE023IE024IE025

IS001

IS002

ITC11ITC12

ITC13

ITC14

ITC15

ITC16

ITC17

ITC18

ITC20

ITC31

ITC32ITC33ITC34

ITC41

ITC42ITC43

ITC44

ITC46ITC47ITC48

ITC49ITC4AITC4B

ITC4C

ITC4D

ITF11

ITF12ITF13ITF14

ITF21ITF22

ITF31ITF32ITF33

ITF34ITF35

ITF43ITF44

ITF45

ITF46

ITF47ITF48ITF51

ITF52

ITF61

ITF62

ITF63ITF64

ITF65

ITG11ITG12

ITG13

ITG14

ITG15ITG16

ITG17ITG18

ITG19 ITG25

ITG26

ITG27ITG28 ITG29

ITG2A

ITG2BITG2C

ITH10

ITH20

ITH31ITH32

ITH33

ITH34

ITH35

ITH36

ITH37

ITH41ITH42

ITH43

ITH44ITH51ITH52

ITH53ITH54ITH55

ITH56

ITH57ITH58

ITH59

ITI11

ITI12ITI13

ITI14

ITI15

ITI16

ITI17

ITI18

ITI19ITI1A


ITI33

ITI34ITI35

ITI41

ITI42

ITI43ITI44

ITI45

KOS25KOS26

KOS27KOS28

KOS29KOS30

KOS31

KOS33

KOS34

KOS35

KOS36

KOS37

KOS39

KOS40KOS42

KOS43KOS44

KOS45

KOS46

KOS47

KOS48

KOS49KOS50

KOS51

KOS52

KOS53

LI000LT001

LT002LT003

LT004LT005LT006

LT007LT008LT009LT00ALU000

LV003LV005

LV006

LV007LV008 LV009ME00

NL111

NL112

NL113

NL121

NL122

NL123NL131


NL213NL221

NL224

NL225

NL226

NL230

NL310

NL321

NL322NL323NL325NL326NL327

NL332

NL333

NL337NL338

NL339

NL33A

NL341

NL342

NL411NL412

NL413

NL414

NL421NL422NL423

NO011

NO012

NO021NO022

NO031

NO032

NO033

NO034 NO041NO042NO043NO051

NO052

NO053

NO061

NO062NO071NO072

NO073

PL113

PL114


PL127

PL128PL129

PL12A

PL213

PL214

PL215PL216PL217

PL224PL225

PL227PL228

PL229

PL22A

PL22BPL22CPL311PL312PL314PL315

PL323PL324

PL325

PL326PL331PL332

PL343PL344PL345

PL411

PL414

PL415

PL416

PL417PL418

PL422PL423PL424

PL425PL431PL432

PL514

PL515

PL516PL517PL518

PL521

PL522PL613PL614


PL631

PL633

PL634PL635

PT111PT112

PT113

PT114

PT115PT116

PT117

PT118 PT150

PT161PT162PT163

PT164PT165

PT166PT167

PT168

PT169

PT16APT16BPT16C

PT171

PT172

PT181PT182PT183

PT184

PT185RO111

RO112RO113

RO114

RO115RO116

RO121

RO122RO123RO124

RO125RO126

RO211

RO212RO213

RO214RO215 RO216

RO221

RO222

RO223

RO224

RO225

RO226RO311

RO312

RO313

RO314

RO315

RO316RO317

RO322

RO411RO412RO413RO414RO415 RO421

RO422

RO423RO424

SE110

SE121 SE122SE123SE124SE125

SE211SE212

SE213

SE214SE221SE224SE231SE232

SE311

SE312

SE313SE321

SE322

SE331

SE332

SI011

SI012

SI013SI014 SI016

SI017

SI018

SI021

SI022SI023

SI024

SK010SK021

SK022 SK023SK031SK032SK041

SK042

SRB54SRB55 SRB56

SRB57SRB58

SRB59

SRB60

SRB61

SRB62SRB63

SRB64SRB65

SRB66SRB67

SRB68SRB90

SRB91SRB92

SRB93

SRB94

SRB95

SRB96SRB97SRB98SRB99

UKC11UKC12

UKC14

UKC21

UKC22UKC23

UKD11UKD12

UKD31UKD32

UKD41

UKD43

UKD61

UKD62UKD63

UKD71UKD72

UKD73

UKD74

UKE12

UKE13

UKE21

UKE22

UKE31

UKE32

UKE41UKE42UKE44UKE45UKF12

UKF13

UKF15

UKF16

UKF22UKF24UKF25

UKF30UKG11

UKG12UKG13UKG21

UKG22

UKG24

UKG31UKG32UKG37

UKH11UKH12UKH13UKH14

UKH23

UKH24

UKH25UKH32

UKH33

UKI12UKI21

UKI22UKI23

UKJ11UKJ12

UKJ13UKJ14

UKJ21

UKJ22

UKJ23

UKJ24

UKJ31

UKJ33

UKJ34

UKJ42

UKK11

UKK12

UKK13

UKK14

UKK15

UKK21

UKK22UKK23

UKK30

UKK42

UKK43 UKL11

UKL12

UKL13UKL14

UKL15UKL16

UKL17UKL18

UKL21

UKL22UKL23

UKL24

UKM21

UKM22UKM23

UKM24

UKM26

UKM27

UKM28

UKM31UKM32

UKM33

UKM34

UKM35

UKM36

UKM37

UKM38UKM50

UKM61

UKM62 UKM63

UKM64

UKM65

UKM66

UKN02

UKN03

UKN04UKN05 ES111ES112

ES113

ES114

ES120ES130ES211

ES212

ES213

ES220

ES230ES241

ES242ES243

ES300

ES411ES412

ES413

ES414ES415ES416

ES417

ES418

ES419

ES42ES4

ES423

ES424ES425

ES43ES432

ES511

ES512

ES513

ES514

ES521

ES522ES523

ES531ES532

ES533ES611

ES612

ES613

ES614

ES6ES616

ES617

ES618

ES620

−20

020

4060

Popu

latio

n co

ncen

tratio

n

−60 −40 −20 0 20 40Population concentration (geo−climatic controls)

Figure B.8: Hotspot analysis for population concentration.


C The Madrid anomaly

Left panel of Figure C.9 shows that, within Spain, all regions at the NUTS2 level have a lower

settlement density than Madrid (the omitted category) after accounting for geography and climate.

Similar pattern at the NUTS3-province level with the exception of Barcelona and Girona that are

not significantly different. Right panel of Figure C.9 shows that this excess of settlement density

in the capital region is not observed in France. Contrarily, it holds for other European countries

(omitted category is always the region capital for Germany, Italy, Portugal...), both using NUTS2

Indeed, we can summarize the cross-country evidence on the capital anomaly by simply doing a

scatter plot with zero and different colors for each country to show that Spanish dots are always to

the left of zero but this is not the case for other countries.

and NUTS3 regions (available under request).

Table B.4: Characteristics of high population concentration municipalities.


Distance to capital 0.196*** 0.175*** 0.152*** 0.152*** 0.157*** 0.148*** 0.0778*** 0.0928***(0.0312) (0.0265) (0.0240) (0.0253) (0.0256) (0.0267) (0.0240) (0.0251)

Dependency ratio -2.604 -6.568*** -6.548*** -8.104*** -14.08*** -9.506*** -9.829***(1.960) (2.079) (2.069) (1.981) (2.776) (1.732) (2.308)

Share born in mun. 15.37*** 12.90*** 12.70*** 14.44*** 15.10*** 17.35*** 19.67***(5.417) (4.423) (4.157) (4.100) (4.034) (3.207) (3.486)

Pop. growth 50-91 -6.044*** -6.049*** -5.855*** -6.523*** -4.125*** -3.891***(0.886) (0.895) (0.895) (0.703) (0.409) (0.418)

Regionalist vote growth -0.0122 -0.0197 -0.00777 0.00729 -0.0262(0.0746) (0.0719) (0.0752) (0.0684) (0.0796)

Share discontent vote 0.000941 -0.00516 -0.0583 -0.160*** -0.202***(0.0909) (0.0925) (0.105) (0.0378) (0.0399)

Public spending pc 0.590 1.804 0.751 -0.130(0.683) (1.224) (0.860) (1.156)

Public revenues pc 0.506 0.605 -0.0852 0.708(0.548) (0.858) (0.577) (0.692)

Share agriculture 12.02 6.948 9.974(8.957) (6.906) (7.417)

Income pc (log) -6.560* -3.182(3.545) (2.273)

Income pc (PPP, log) 0.0733(0.454)


Notes: Dependent variable is the population concentration indicator. Geo-climatic controls are not reported butincluded in all columns, namely, temperature, rainfall, average altitude, ruggedness, soil quality and distance to thecoast. Standard errors clustered at the province level. *** p<0.01, ** p<0.05, * p<0.1


Figure C.9: The capital anomaly.

AND

ARA

AST

CANT

CAT

CLM

CVA

CyL

EXT

GAL

MUR

NAV

PV

RIOJ

−50 0 50 −50 0 50

Baseline Geographic controlsFR21FR22FR23FR24FR25FR26FR30FR41FR42FR43FR51FR52FR53FR61FR62FR63FR71FR72FR81FR82FR83

−40 −20 0 20 −40 −20 0 20

Baseline Geographic controls

BANCO DE ESPAÑA PUBLICATIONS

WORKING PAPERS

1920 LUIS J. ÁLVAREZ, MARÍA DOLORES GADEA and ANA GÓMEZ-LOSCOS: Infl ation interdependence in advanced

economies.

1921 DIEGO BODAS, JUAN R. GARCÍA LÓPEZ, JUAN MURILLO ARIAS, MATÍAS J. PACCE, TOMASA RODRIGO LÓPEZ,

JUAN DE DIOS ROMERO PALOP, PEP RUIZ DE AGUIRRE, CAMILO A. ULLOA and HERIBERT VALERO LAPAZ:

Measuring retail trade using card transactional data.

1922 MARIO ALLOZA and CARLOS SANZ: Jobs multipliers: evidence from a large fi scal stimulus in Spain.

1923 KATARZYNA BUDNIK, MASSIMILIANO AFFINITO, GAIA BARBIC, SAIFFEDINE BEN HADJ, ÉDOUARD CHRÉTIEN,

HANS DEWACHTER, CLARA ISABEL GONZÁLEZ, JENNY HU, LAURI JANTUNEN, RAMONA JIMBOREAN,

OTSO MANNINEN, RICARDO MARTINHO, JAVIER MENCÍA, ELENA MOUSARRI, LAURYNAS NARUŠEVIČIUS,

GIULIO NICOLETTI, MICHAEL O’GRADY, SELCUK OZSAHIN, ANA REGINA PEREIRA, JAIRO RIVERA-ROZO,

CONSTANTINOS TRIKOUPIS, FABRIZIO VENDITTI and SOFÍA VELASCO: The benefi ts and costs of adjusting bank

capitalisation: evidence from Euro Area countries.

1924 MIGUEL ALMUNIA and DAVID LÓPEZ-RODRÍGUEZ: The elasticity of taxable income in Spain: 1999-2014.

1925 DANILO LEIVA-LEON and LORENZO DUCTOR: Fluctuations in global macro volatility.

1926 JEF BOECKX, MAARTEN DOSSCHE, ALESSANDRO GALESI, BORIS HOFMANN and GERT PEERSMAN:

Do SVARs with sign restrictions not identify unconventional monetary policy shocks?

1927 DANIEL DEJUÁN and JUAN S. MORA-SANGUINETTI: Quality of enforcement and investment decisions. Firm-level

evidence from Spain.

1928 MARIO IZQUIERDO, ENRIQUE MORAL-BENITO and ELVIRA PRADES: Propagation of sector-specifi c shocks within

Spain and other countries.

1929 MIGUEL CASARES, LUCA DEIDDA and JOSÉ E. GALDÓN-SÁNCHEZ: On fi nancial frictions and fi rm market power.

1930 MICHAEL FUNKE, DANILO LEIVA-LEON and ANDREW TSANG: Mapping China’s time-varying house price landscape.

1931 JORGE E. GALÁN and MATÍAS LAMAS: Beyond the LTV ratio: new macroprudential lessons from Spain.

1932 JACOPO TIMINI: Staying dry on Spanish wine: the rejection of the 1905 Spanish-Italian trade agreement.

1933 TERESA SASTRE and LAURA HERAS RECUERO: Domestic and foreign investment in advanced economies. The role

of industry integration.

1934 DANILO LEIVA-LEON, JAIME MARTÍNEZ-MARTÍN and EVA ORTEGA: Exchange rate shocks and infl ation comovement

in the euro area.

1935 FEDERICO TAGLIATI: Child labor under cash and in-kind transfers: evidence from rural Mexico.

1936 ALBERTO FUERTES: External adjustment with a common currency: the case of the euro area.

1937 LAURA HERAS RECUERO and ROBERTO PASCUAL GONZÁLEZ: Economic growth, institutional quality and fi nancial

development in middle-income countries.

1938 SILVIA ALBRIZIO, SANGYUP CHOI, DAVIDE FURCERI and CHANSIK YOON: International Bank Lending Channel of

Monetary Policy.

1939 MAR DELGADO-TÉLLEZ, ENRIQUE MORAL-BENITO and JAVIER J. PÉREZ: Outsourcing and public expenditure: an

aggregate perspective with regional data.

1940 MYROSLAV PIDKUYKO: Heterogeneous spillovers of housing credit policy.

1941 LAURA ÁLVAREZ ROMÁN and MIGUEL GARCÍA-POSADA GÓMEZ: Modelling regional housing prices in Spain.

1942 STÉPHANE DÉES and ALESSANDRO GALESI: The Global Financial Cycle and US monetary policy

in an interconnected world.

1943 ANDRÉS EROSA and BEATRIZ GONZÁLEZ: Taxation and the life cycle of fi rms.

1944 MARIO ALLOZA, JESÚS GONZALO and CARLOS SANZ: Dynamic effects of persistent shocks.

1945 PABLO DE ANDRÉS, RICARDO GIMENO and RUTH MATEOS DE CABO: The gender gap in bank credit access.

1946 IRMA ALONSO and LUIS MOLINA: The SHERLOC: an EWS-based index of vulnerability for emerging economies.

1947 GERGELY GANICS, BARBARA ROSSI and TATEVIK SEKHPOSYAN: From Fixed-event to Fixed-horizon Density

Forecasts: Obtaining Measures of Multi-horizon Uncertainty from Survey Density Forecasts.

1948 GERGELY GANICS and FLORENS ODENDAHL: Bayesian VAR Forecasts, Survey Information and Structural Change in

the Euro Area.

2001 JAVIER ANDRÉS, PABLO BURRIEL and WENYI SHEN: Debt sustainability and fi scal space in a heterogeneous

Monetary Union: normal times vs the zero lower bound.

2002 JUAN S. MORA-SANGUINETTI and RICARDO PÉREZ-VALLS: ¿Cómo afecta la complejidad de la regulación a la

demografía empresarial? Evidencia para España.

2003 ALEJANDRO BUESA, FRANCISCO JAVIER POBLACIÓN GARCÍA and JAVIER TARANCÓN: Measuring the

procyclicality of impairment accounting regimes: a comparison between IFRS 9 and US GAAP.

2004 HENRIQUE S. BASSO and JUAN F. JIMENO: From secular stagnation to robocalypse? Implications of demographic

and technological changes.

2005 LEONARDO GAMBACORTA, SERGIO MAYORDOMO and JOSÉ MARÍA SERENA: Dollar borrowing, fi rm-characteristics,

and FX-hedged funding opportunities.

2006 IRMA ALONSO ÁLVAREZ, VIRGINIA DI NINO and FABRIZIO VENDITTI: Strategic interactions and price dynamics

in the global oil market.

2007 JORGE E. GALÁN: The benefi ts are at the tail: uncovering the impact of macroprudential policy on growth-at-risk.

2008 SVEN BLANK, MATHIAS HOFFMANN and MORITZ A. ROTH: Foreign direct investment and the equity home

bias puzzle.

2009 AYMAN EL DAHRAWY SÁNCHEZ-ALBORNOZ and JACOPO TIMINI: Trade agreements and Latin American trade

(creation and diversion) and welfare.

2010 ALFREDO GARCÍA-HIERNAUX, MARÍA T. GONZÁLEZ-PÉREZ and DAVID E. GUERRERO: Eurozone prices: a tale of

convergence and divergence.

2011 ÁNGEL IVÁN MORENO BERNAL and CARLOS GONZÁLEZ PEDRAZ: Sentiment analysis of the Spanish Financial

Stability Report. (There is a Spanish version of this edition with the same number).

2012 MARIAM CAMARERO, MARÍA DOLORES GADEA-RIVAS, ANA GÓMEZ-LOSCOS and CECILIO TAMARIT: External

imbalances and recoveries.

2013 JESÚS FERNÁNDEZ-VILLAVERDE, SAMUEL HURTADO and GALO NUÑO: Financial frictions and the wealth distribution.

2014 RODRIGO BARBONE GONZALEZ, DMITRY KHAMETSHIN, JOSÉ-LUIS PEYDRÓ and ANDREA POLO: Hedger of last

resort: evidence from Brazilian FX interventions, local credit, and global fi nancial cycles.

2015 DANILO LEIVA-LEON, GABRIEL PEREZ-QUIROS and EYNO ROTS: Real-time weakness of the global economy: a fi rst

assessment of the coronavirus crisis.

2016 JAVIER ANDRÉS, ÓSCAR ARCE, JESÚS FERNÁNDEZ-VILLAVERDE and SAMUEL HURTADO: Deciphering the

macroeconomic effects of internal devaluations in a monetary union.

2017 FERNANDO LÓPEZ-VICENTE, JACOPO TIMINI and NICOLA CORTINOVIS: Do trade agreements with labor provisions

matter for emerging and developing economies’ exports?

2018 EDDIE GERBA and DANILO LEIVA-LEON: Macro-fi nancial interactions in a changing world.

2019 JAIME MARTÍNEZ-MARTÍN and ELENA RUSTICELLI: Keeping track of global trade in real time.

2020 VICTORIA IVASHINA, LUC LAEVEN and ENRIQUE MORAL-BENITO: Loan types and the bank lending channel.

2021 SERGIO MAYORDOMO, NICOLA PAVANINI and EMANUELE TARANTINO: The impact of alternative forms of bank

consolidation on credit supply and fi nancial stability.

2022 ALEX ARMAND, PEDRO CARNEIRO, FEDERICO TAGLIATI and YIMING XIA: Can subsidized employment tackle

long-term unemployment? Experimental evidence from North Macedonia.

2023 JACOPO TIMINI and FRANCESCA VIANI: A highway across the Atlantic? Trade and welfare effects of the

EU-Mercosur agreement.

2024 CORINNA GHIRELLI, JAVIER J. PÉREZ and ALBERTO URTASUN: Economic policy uncertainty in Latin America:

measurement using Spanish newspapers and economic spillovers.

2025 MAR DELGADO-TÉLLEZ, ESTHER GORDO, IVÁN KATARYNIUK and JAVIER J. PÉREZ: The decline in public

investment: “social dominance” or too-rigid fi scal rules?

2026 ELVIRA PRADES-ILLANES and PATROCINIO TELLO-CASAS: Spanish regions in Global Value Chains: How important?

How different?

2027 PABLO AGUILAR, CORINNA GHIRELLI, MATÍAS PACCE and ALBERTO URTASUN: Can news help measure economic

sentiment? An application in COVID-19 times.

2028 EDUARDO GUTIÉRREZ, ENRIQUE MORAL-BENITO, DANIEL OTO-PERALÍAS and ROBERTO RAMOS: The spatial

distribution of population in Spain: an anomaly in European perspective.

Unidad de Servicios Generales IAlcalá, 48 - 28014 Madrid

E-mail: [email protected]

The spatial distribution of population in Spain: an ... · Documentos de Trabajo N.º 2028. Author:...

Documents

Transcript of The spatial distribution of population in Spain: an ... · Documentos de Trabajo N.º 2028. Author:...