International migrations and urbanisation : 1960-2010

International migrations and urbanisation rates have seen a large increase in the last decades. Here we analyse the relationships between migrations and urbanisation by using a panel of ca 200 countries over the period 1960-2010. We describe the main global stylised facts on urbanisation and international migrations focusing on differences in these across world regions. We found that while there was a positive association between immigration and urbanisation, particularly in small and medium-sized cities, the association between emigration and urbanisation in developing countries was inverse. Both associations have become stronger over the few past decades, and our results highlight that international migration is an increasingly relevant and complementary dimension of the traditional rural-urban reallocation of workers which takes place during economic development.


Introduction
In 1960 one third of the world's population lived in cities.In 2012 this figure reached 52.6%.At this speed of growth, which has been particularly faster in the last two decades, in 2050 around two thirds of the world population will be living in cities.The process of urbanisation is related to the process of economic development, with people migrating from lagged rural areas to developing urban ones, which exhibit higher productivity and offer higher wages.However, migration flows currently occur not just within countries but also increasingly between them, with people also looking for opportunities, derived for instance, by international differentials in productivity levels and real wages.Migrations have important consequences on both sending and receiving countries.
In fact, despite the fact that the largest flows have been within countries, the number of persons living outside their country of birth has surpassed 200 million people (almost three per cent of the world's population in 2010).While in OECD countries, the population of foreign origin now represents about 12% of the total population (i.e. are net 'importers' of people), in other parts of the world, such as Central Asia, Central America and the Caribbean, more than 10% of the population has migrated out of their country.Acknowledging these two global trends, the aim of this paper has been to inspect whether international migration can be considered as a relevant factor in the evolution of the internal economic geography of countries.
In this work, we analyse the increasing role of cities all over the world, and the importance of international migrations in the process of urbanisation, using a wide database of 197 countries over the period 1960-2010.We focused on differences across world geographical regions in levels of economic development and rates of urbanisation.Urbanisation was examined both quantitatively (i.e.rates) and qualitatively, i.e. by considering urbanisation in small and-medium sized cities and the process of urban concentration in larger agglomerations.
The paper is structured in five sections.After this introduction we present a brief literature review on the determinants of urbanisation, focusing on a potential role for international migrations, which prepares the ground for inspecting the current trends in these two processes through correlation analyses (section 3).In section 4 we use a parsimonious econometric model for urbanisation.Finally, we discuss the main findings and conclusions.

Determinants of urbanisation: A brief literature review
The central role of urbanisation in the process of economic development is extensively present in the literature (Lewis, 1954;Ranis and Fei, 1961;Harris and Todaro, 1970;Todaro, 1976, to mention a few seminal papers).In most of these models, migration occurs between lagged rural areas and developed urban areas, driven by the higher wages in the latter due to the higher productivity of the urban sector.As migration proceeds, the economy experiences a structural change moving from agricultural-based activities to industrial activities, which are the base for economic development (Gollin et al. 2002).However, urbanisation, and urban concentration in large agglomerations, is also maintained by the higher productivity driven by agglomeration economies (even at the risk of urban underemployment, as expressed in Rauch, 1993), with labour mobility being the human side of the agglomeration story.Urbanisation can be seen as a manifestation of development processes, and "migration is a contributor of development, a corrector of regional imbalances, and a conqueror of the tyranny of space," in the words of Firebaugh (1979, p.199).
More recent literature has focused on the process of urbanisation itself, considering not just urbanisation but also urban concentration.In this line, there are papers providing theoretical modelling (Henderson and Wang, 2005), as well as empirical evidence on the association between the sector shift out of agriculture and urbanisation (Brueckner, 1990;Davis and Henderson, 2003;Castells-Quintana andRoyuela, 2011 andFranch Auladell et al., 2013).Kasarda and Crenshaw (1991) define three proximate determinants of urban growth: the natural increase of urban population; boundary redefinition through annexation of surrounding areas; and migration, both intra-national (rural-urban and urban-urban) and international.However, as the authors highlight, scholars have focused mainly on rural-urban migration rather than on international migration.The relevant role of other factors driving urbanisation has also been highlighted, including institutions (Henderson and Wang, 2007), adverse rural conditions (Firebaugh, 1979) and climate changeespecially in Sub-Saharan Africa (Barrios et al., 2006).Finally, regarding urban concentration, other factors have been found to be relevant.Ades and Glaeser (1995) report as significant determinants, high tariffs, high costs of internal trade and low levels of international trade, as well as political factors, while Davis and Henderson (2003) analyse the role of specific policies like development of interregional infrastructure and fiscal decentralization.
International migrations are an increasingly important factor accompanying traditional internal migrations.However, they are often overlooked in the analyses of urbanisation processes.
International migrations can be understood as an equaliser of the marginal productivity of labour not just between rural and urban spaces, but also between countries.Additionally, in the same way as rural-urban migrations within countries, associations are also expected between international migrations and features of the process of economic development and thus also with changes in the economic geography of countries.Of course, we expect migration processes to affect the economic geography of both sending as well as receiving countries.
Overall, as the World Development Report (World Bank, 2009) stresses, "an important insight of the agglomeration literaturethat human capital earns higher returns where it is plentifulhas been ignored by the literature of labour migration" (WDR, 2009, p. 158).Novel understandings come from considering together agglomeration economies and labour migration, which takes place both at the national and at the international level.

Data and trends
In this section we describe the main trends in urbanisation and international migrations all over the world by exploring variables on population and urbanisation gleaned from the World Bank World Development Indicators at the country level. 1 Regarding migration, our data sources were the World Bank Bilateral Migration Database 1960-2000 and the World Bank Bilateral Migration Matrix 2010.
Data are from 197 countries for the years 1960, 1970, 1980, 1990, 2000 and 2010. 2 These datasets provide information on stocks rather than on flows.However, we selected these data because immigration stocks data are based on national censuses, and therefore are probably of higher quality than data reporting annual immigrant flows.Censuses deal with unambiguous, net permanent moves.In our paper we have followed Ortega and Peri (2009), who argue that theoretical models of migration can be interpreted as determining the relationships for stocks of migrants, or the analogous flows.As they do, we will consider the stock of migrants as a proxy of long-term migration flows.
Table 1 presents the main urbanisation trends by continent and world subregion. 3The urban world population has increased from 33% in 1960 to 51% in 2010.In Africa, America and Asia urbanisation has increased by at least 20 points, in Europe by 16, and in Oceania by 4 (already having a high rate in 1960).In 2010, 15 subregions had more than half of their population living in cities, while in the remaining eight regions the figure was still below 50%.Urban concentration (defined as population living in cities of more than one million inhabitants) has globally risen by 6% in the last 50 years, being especially important in America, Oceania, and in several other subregions, such as Southern Africa and Western Asia.However, the global urbanisation trend has a stronger pace in small and medium-sized cities (i.e. less than one million inhabitants).These cities have grown two times faster than larger cities over the past 50 years (smaller cities: 20% in 1960 to 32% in 2010 vs. larger cities: 13% in 1960 to 19% in 2010).In two regions, Central Asia and Northern Europe, large cities actually lost weight, while small and medium-sized cities were responsible for the entire increase in urbanisation rates.Moreover, in Europe we can see that more 1 Urban population refers to people living in urban areas as defined by National Statistical Offices.It has been estimated from World Bank population estimates and urban ratios from the United Nations World Urbanisation Prospects. 2The list of countries is reported in Annex 1.These databases can be respectively accessed at http://data.worldbank.org/data-catalog/global-bilateral-migration-database and http://go.worldbank.org/JITC7NYTT0.We used the version collected and revised by Ramos (2013).Over one thousand census and population register records have been combined to construct decennial matrices. 3Annex 1 classifies the countries by continent and geographical region.than 80% of the increase in urbanisation rates was due to the growth of small and medium-sized cities, rather than due to population increases in the large cities.Finally, the growth in urbanisation has been faster in recent decades, and this is true for both small/medium-sized (<1 million inhabitants) and large cities.
Table 2 shows the main demographics trends, also by continent and world subregion.Population growth has slowed down over the last 40 years, although several world subregions still had annual growth rates in excess of 2% in 2010, mainly African regions, Western Asia and Melanesia.Eastern Europe is the only subregion where population has decreased in the past two decades.Regarding migration, although the number of migrants has increased over time, its share of total world population was 3% in 1960 and slightly decreased in 1990 (2.6%), followed by resurgence since then (2.8% in 2010).
Immigration has been particularly pronounced in rich regions, like North America, most of Europe and Australia and New Zealand, compared to other regions.In North America, Western Europe, Australia and New Zealand, immigration largely exceeds emigration.On the other hand, emigration is more important in the Caribbean, Central America, Central Asia, Eastern Europe, Micronesia and Polynesia, with insular regions (such as the Caribbean and the Polynesia) showing the most dramatic increases in emigration rates from 1960 to 2010 and where emigration currently largely exceeds immigration.Interestingly, African regions, Western Asia and Melanesia, with the highest population growth, do not show the highest rates of emigration.
Overall, although urban population has increased in both small and medium-sized cities and in larger cities, among all urban people in the world the proportion of population living in large cities has decreased slightly over the last 50 years (39% in 1960 to 37% in 2010).We found important differences in this across world regions; while in certain regions urban concentration in large cities is substantial (e.g.North and South America, Southern Africa, Australia and New Zealand), in other regions it is the small and medium-sized cities that drives overall urban growth (similarly to European regions and Central Asia).Regarding international migrations, developed regions, which are usually the more urbanised ones, have the highest levels of net immigration (defined as the difference between immigration and emigration), while poorer regions show higher levels of net emigration.1970 1970-1980 1980-1990 1990-2000 2000-2010 1960 1970 1980 1990 2000 2010 1960 1970 1980 1990

Correlations between international migration and urbanisation processes
The relationships between the processes of urbanisation and international migration were examined by conducting correlation analyses between these variables at the international level on our panel database.Table 3 displays the correlation coefficients between migration (emigration rate at origin and immigration rate at destination) and urbanisation rate variables calculated from the raw data, or from the data after removal of time and/or country effects,4 for all sized cities and separately for small/medium-sized and large cities.Population growth was positively correlated with immigration rates and negatively correlated with emigration rates.The sign and significance of this relationship persists when the time effect is removed, but disappears when country effects are not present.Consequently, the observed correlation is a country-effect issue, i.e. countries with higher population growth are also those with less emigration and more immigration.On the other hand, urbanisation rates are positively correlated with both emigration and immigration rates, and again the country effect dominates, so the more urbanised countries are the ones with a higher propensity to international migration.However, and interestingly enough, we find conflicting signs for the different urbanisation rates with respect to emigration; it is negatively correlated with urbanisation in cities of more than one million, while it is positively correlated with urbanisation in small and medium-sized cities.Similar conflicting signs are found with respect to immigration when we remove time and country effects.
In other words, apparently international immigration is being directed more to smaller cities than to larger ones.Finally, correlations between growth in urbanisation rates and migration rates are generally not significant.The only exception is a significantly negative correlation between urbanisation growth in large cities and immigration: countries with higher growth of large cities (often, developing countries) seem to be those experiencing smaller international immigration rates.Because the analysis of world trends suggest that there are differences across regions according to the level of economic development, we have divided the sample into developed countries and developing countries and have computed the correlation coefficients separately for each5 (Table 4).
When we look at the raw data the basic figures are overall similar to the global ones.The picture is different when we focus on correlations once country and time effects have been removed.We found important differences between developed and developing regions.While in developed countries population growth is significantly correlated with immigration rates, in developing countries this is not the case.On the other hand, urbanisation is negatively correlated with immigration rates in developed regions but still positively correlated in developing countries.The main driver of these differences is the urbanisation rate in small and medium-sized cities, which is negatively correlated with immigration in developed regions but positively correlated in developing countries.

Regression Analysis
In this section, the correlation analysis between migration rates and urbanisation goes a step further by controlling for other relevant factors in the urbanisation process.Thus, in order to test the influence on international migration of urbanisation rates we followed the parametric strategy proposed in Barrios et al. (2006): where   is the urbanisation rate,   is a vector of time-varying potential determinants, in which we include migration rates,   are time-specific effects common to all countries,   are timeinvariant country-specific effects, and   is the usual error term.We did not aim to perform a strict causality analysis but rather to inspect the correlation between urbanisation and international migration once other factors had been cleared, taking advantage of the panel structure of the database, and performing a set of cross-country regression analyses.
For Urban, our dependent variable, we considered three different rates: urbanisation, urbanisation in larger cities (> 1 million inhabitants), and urbanisation in smaller cities (<1 million inhabitants).
For migration rates we included immigration rate at destination [immigr] and the emigration rate at origin [emigr].As controls, we included a list of classical factors in the applied literature: two economic variables, GDP per capita [GDPpc] and telephones per capita [telph_pc]; three demographic variables, total population [pop_total], the proportion of young [pop_0_14] and older people [pop_m65], and two development variables: life expectancy at birth [life_exp] and infant mortality rates [mort_inf].Even though we assumed that there may be relevant factors that we had excluded from the controls considered, for instance those related to institutions, we deemed that these factors would be mostly captured by development variables or, at a later stage, in the fixed effects specification.Even though these country-specific effects represent a measure of our ignorance, they are a good control for time invariant omitted variables.
The empirical model introduces all variables in logs, except the ones expressed as percentages, and is summarised in the following equation: We calculated the between estimates (BE), which can be interpreted as measuring the long-run effects on urbanisation rates, and the fixed-effects estimates (FE), which capture how time-series changes within a country impinge on the changes in its urbanisation rate over time.In addition, given that these coefficients only reflect within-country time-series variation, they can be interpreted as short-run effects.We also performed a pooled estimation ('Pool'), which can be interpreted as an average of the BE and FE estimates.Table 5 displays the descriptive statistics of the main variables of our analysis.Most of the information on urbanisation rates and immigration rates at destination is cross sectional.Consequently, we expected that the BE estimates captured a substantial part of the variation of the urbanisation variables, while the FE results would explain the variations observed within countries over the 50 years considered in our study.In order to obtain robust estimates of the standard errors, the estimates relaxed the usual requirement that the observations are independent, and consequently we allowed for intragroup correlation in the disturbance term: the observations are independent across groups (clusters) but not necessarily within groups.The considered groups are the world's subregions.The results are displayed in table 6. Immigration rates displayed a positive and significant parameter in all estimates (pooled, between and fixed effects) for the global urbanisation rate, in the between-estimate for larger cities, and in the pooled and fixed effects estimates for the small and medium-sized cities (Table 5).Thus, immigration is clearly linked to within-country increases in urbanisation rates, and this is particularly true for small and medium-sized cities. Quantitatively, the fixed effects estimates report that a 1% increase in the immigration rate is associated to a 0.3% higher urbanisation rate in the small and medium-sized cities The regression of urbanisation on emigration had a negative, although non-significant sign, as was found for bivariate correlation analysis (see section 3.2).In other words: countries with higher (between estimates) or increasing (fixed effects estimates) emigration rates did not experience smaller or decreasing levels of urbanisation.
Emigration was significant and negative in developing regions, but not so in the developed regions.This result suggests, at least from several estimates, that international emigration from developing countries occurs at the expense of local urbanisation: an emigration rate 1% higher is associated to a 0.16% (pool estimate for small and medium-sized cities) to 0.34% (between estimates for the global urbanisation rate) lower urbanisation rate.Our interpretation of this result is that migrants consider as destination both national as well as international urban destinations.The 1960-1980 period displayed substantially less significant results than the 1990-2010 period.
The impact of international migration on urbanisation has been more pronounced in recent decades (i.e. 1990-2010) than earlier .And this is true for both immigration and emigration.We see that the quantitative impact of immigration on urbanisation rates in smaller cities has decreased over time (from 0.31% in the 1960-1980 period to 0.17% in the 1990-2010 period), probably as a consequence of a large increase in urbanisation rates in the smaller cities in developing countries.

Discussion and conclusions
Individual case studies and regional comparisons analysing the impact of international migration on local cities have been common in the literature.Cross-national research, on the other hand, allows for testing general trends, as well as differences across world regions.The large database in this paper (ca 200 countries over a 50-year period) allowed us to capture the global stylised facts of the relationships between international migrations and urbanisation, which have seldom been considered together (WDR, 2009).
We have identified strong and robust empirical relationships.Using panel estimations we have pointed out that immigration is indeed associated with increasing urbanisation, while emigration is only negatively associated with urbanisation in developing countries.Additionally, small and medium-sized cities are apparently more influenced by international migrations than larger cities, and this process has been particularly important in recent decades.This result is in line with recent calls (OECD, 2009) stressing that in many countries it can be the case that medium-sized and small agglomerations enjoy stronger opportunities for economic development6 .
Our results clearly complement the traditional urbanisation studies where the structural change takes place by internal rural-urban migration processes.While there is no doubt that increasing urbanisation worldwide has become a major challenge for sustainable development, especially in developing countries but also in developed ones, the role of increasing international migrations in shaping the economic geography of countries deserves further research.

Table 4 .
Correlation coefficients between countries' migration and urbanisation rates, separately for developed (top table) and developing countries (bottom table).1960-2010.

Table 5 .
Descriptive statistics of main variables in the regression analysis

Table 7 .
Regression results.Parameter estimates.Subsamples according to development and time