Comparing Internal Migration Intensities around the Globe

Comparing Internal Migration Intensities around the Globe Paper prepared for the IUSSP to be held in Busan Korea, 26-31 August 2013 Martin Bell, University of Queensland, martin.bell@uq.edu.au; Elin Charles-Edwards, University of Queensland, e.charles-edwards@uq.edu.au; Philipp Ueffing, University of Queensland; Marek Kupiszewski, Polish Academy of Sciences, m.kupisz@twarda.pan.pl ; Dorota Kupiszewska, Central European Forum for Migration and Population Research, d.kupisz@twarda.pan.pl; John Stillwell, University of Leeds, j.c.h.stillwell@leeds.ac.uk; Yu Zhu, Fujian Normal University, zhu300@fjnu.edu.cn Introduction This paper reports results from the IMAGE project, a four year international collaborative research program designed to provide a framework for systematic comparisons of internal migration, the ultimate goal being to develop and apply a robust set of measures that can be used to advance understanding of the way in which Internal Migration varies Around the GlobE. The stimulus to this work derives from the fact that, compared with fertility and mortality, surprisingly little attention has been given to understanding the way internal migration varies between nations. With the progressive convergence in fertility and life expectancy as countries complete the demographic transition, internal migration, along with international mobility, is becoming the dominant force affecting the composition and distribution of populations around the globe. While the significance of migration is increasingly being recognised (see e.g. The World Bank 2009; United Nations 2009), comparative indicators of internal migration are conspicuous by their absence from international statistical collections, such as the UN Demographic Yearbook and there is no comprehensive national league table of internal mobility akin to those ranking countries according to rates of birth, death and international migrant stocks. The deficit in robust comparison of internal migration can be traced in part to the absence of statistical standards in the collection and dissemination of migration data, including a lack of standard definitions, discrepancies in whether moves or movers are captured, and differences in the time period and spatial framework over which migration is measured. Such comparisons are further complicated by the multifaceted nature of internal migration. Bell et al.(2002) identified four discrete dimensions of migration capturing particular facets of mobility the overall prevalence of movement, the distances over which people move, the effects on population redistribution, and the way migration acts as a mechanism connecting localities and regions each of which call for specialised metrics and pose thorny analytical challenges. The scholarly literature contains valuable contributions to understanding on all of these dimensions of mobility (see e.g. Rogers and Castro 1981; Nam, Serow et al. 1990; Long 1991; Rees and Kupiszewski 1999; Bell and Muhidin 2009). To date, however, no concerted attempt has been made to assemble a comprehensive assessment on any of these dimensions for countries around the globe. The aim of the current paper is to utilise data from the IMAGE Repository and IMAGE Studio to generate the first comprehensive league table ranking countries on a single

dimension of internal migration - the overall level of migration intensity, that is the proportion of people changing their place of usual residence in a given interval - matching similar indices long available in international statistical collections for other demographic processes. The level of migration intensity measured in a country depends fundamentally on the number of zones into which that country is divided. Since countries differ widely in size, statistical geography and patterns of settlement, simple cross-national comparisons of migration intensities reported by national statistical agencies cannot be compared directly. To circumvent this problem, we compare a measure of aggregate migration intensity which encompass all permanent changes of address, and aim to provide estimates for one and five year periods for a large sample of the 193 UN Member States. Only a small minority of countries collect these data directly, via censuses, surveys and registers, so we employ a range of statistical methods to harmonise the data as far as possible for differences in temporal and spatial frameworks and hence generate comparable estimates of aggregate intensities. By way of background we first draw on the IMAGE Inventory to summarise the types of migration data collected, the intervals over which migration is measured and the spatial frameworks employed, and show how these vary between countries around the world. We then summarise the data held in the IMAGE Repository which brings together migration matrices and counts for a subset of 193 UN member states. We also describe the IMAGE Studio which was developed to facilitate comparative analysis of internal migration data across countries in our sample. We describe the four sub-systems that comprise this system, including a spatial aggregation routine which enables migration intensity and other metrics to be calculated for different spatial configuration within countries. Following this, we identify the key challenges in making broad-based comparisons of aggregate migration intensities and propose a solution based around the index k, originally outlined by Courgeau (1973) and recently modified by Courgeau et al. (2012 ). Supplemented with observed data on aggregate migration intensity for those countries which collect data on all moves, we present a league table of internal migration intensities for some 71 countries around the world. We seek explanation for differences in overall intensity using a range of social, economic and demographic variables. Why compare population mobility? Population mobility is integral to the process of human development (Zelinsky 1971; Skeldon 1997; United Nations 2009): it is the primary demographic process shaping patterns of human settlement; it underpins efficient functioning of the economy; it enables individuals and families to meet their aspirations; it is key to enabling transitions through the life course; it is also an essential mechanism in flight from risk and danger. An understanding of human mobility is therefore central, both to informed social science and to the health of national space economies. Strong arguments can also be made for a rigorous framework that enables robust cross-national comparisons: measures of mobility, like all aspects of social behaviour, become more meaningful when placed in a comparative framework; such comparisons aid theorisation and bring to light the diversity of migration dynamics; they also encourage

greater precision in individual country studies. These developments, in turn, hold the promise of improvements in policy approaches to migration (Bell, Blake et al. 2002). Bell et al. (2002) took steps to develop this framework with proposals for a suite of 15 rigorously defined measures covering four broad dimensions of migration (see also Rees, Bell et al. 2000). Implementation of such measures, however, has been fundamentally constrained by a deficit of basic information as to the nature of the internal migration data that are collected by statistical agencies around the world. If analysts are to undertake rigorous comparisons of the way internal migration differs between countries, a sound understanding of the way it is being measured is indispensable. More broadly, if the study of internal migration is to be placed on the same comparative footing already enjoyed by its demographic sister processes of fertility and mortality, a comprehensive global inventory of data collections was an essential pre-requisite. The IMAGE Inventory While there is a sporadic and fragmented literature comparing aspects of internal migration in selected countries (see Rogers and Castro 1981; Long 1991), data availability has been recognised as a fundamental constraint to comparative research. The most prominent systematic analysis is due to Rees and Kupiszewski (1999) who assembled internal migration data for 28 European countries, but the only attempt to establish a global inventory dates from a survey conducted by the UN Statistical Commission in the 1970s. The Commission concluded that although internal migration was an extremely important phenomenon for most countries the wide diversity of national needs and practices made it difficult to formulate recommendations on migration statistics (United Nations 1978 p. iii). Despite these reservations, the Commission identified 121 countries that collected migration data and reported on a range of features including the type of data collected and the geography of the migration defining regions. A clear understanding of the available data is an indispensable first step towards comparative analysis. The IMAGE inventory was structured to provide comprehensive global coverage of the 193 UN member States and to identify availability of the data needed to implement the comparative measures proposed by Bell et al. (2002). Specific information was sought as to the: instrument used to collect migration data (census, register or survey); type of data collected (transitions, events, duration of residence); interval over which migration flows were measured (e.g. 1 year, 5 years, etc); and the zonal system against which migration was recorded. To date, complete or partial information has been assembled for 178 countries, all but three of which collect internal migration statistics in some form (Bell, Charles-Edwards et al. under review). The remaining 175 employ a mix of data sources but the most common was the census with 159 countries (91%), while 38 countries (21%) drew data from population registers and 16 (9%) employed a survey. Table 1 illustrates the substantial variation between countries and regions in the way migration was measure at the 2000 round of censuses (conducted between 1995 and 2004).

One hundred and thirty-seven countries were identified as collecting some form of internal migration data during this round. Place of birth (within the country) emerges as the most common point of reference (118 countries), and a substantial number of countries measure migration over a fixed interval, a minority employ intervals of one (28) or five years (52). Latest move data, derived by combining information on duration of stay and place of previous residence was collected in 53 countries. While ongoing, evidence to date points to a similar level of diversity in migration data collection at the 2010 census round. Table 1 about here Among the 38 countries drawing migration data from a population register, the vast majority report the number of moves occurring within a fixed interval (e.g. one year). Selected countries (e.g. Sweden, United Kingdom) also release transition data from registers (generated by comparing current addresses with addresses one year prior), but overall is there is less variation than seen in censuses. Latest move data is the most commonly collected by surveys around the world (Bell, Charles-Edwards et al. forthcoming), however, these data tend to lack any detailed spatial breakdown, limiting their utility for cross-national comparisons. Notwithstanding, a number of countries do conduct nationally representative surveys on internal migration, for example the US Current Population Survey, which can be used for comparative purposes. The IMAGE Repository The collection of internal migration data by national statistics agencies does not automatically translate to dissemination of data at an equivalent spatial scale either off the shelf or on request. The format of data available varies widely across countries, ranging from detailed origin-destination matrices, to counts of in-flows and of out-flows from geographic regions, to overall counts of movers measured at different spatial scales. Internal migration data are sometimes disaggregated by age, sex and other characteristics, but usually only at a national level. This lack of international data dissemination standards further complicates crossnational comparison of internal migration. In assembling the IMAGE Repository, we privileged geographic coverage, targeting data in a variety of formats and at varying spatial scales. We focus on data corresponding to the 2000 and 2010 UN round of censuses. In the first instance, we sought origin-destination migration matrices for UN member states at multiple levels of geography, along with national counts of all movers/all changes of address. Where origin-destination matrices were either unavailable or available at only a course spatial grain, we sought matrix marginal totals, i.e. counts of in-flows and out-flows. In constructing the IMAGE Repository we drew on established data collections, along with holdings from national statistics agencies. The Integrated Public Use Micro-data - International (IPUMs-International) holds census micro-data sample files for 74 countries dating back to the 1960s. Origin-destination matrices and/or counts of migrants have been extracted from sample files for 42 countries. Origin-destination matrices for 22 Latin American countries and Mexico were extracted from the Centro Lationamericano y Caribeno de Demografia (CELADE) for major and minor administrative regions. In contrast to the

IPUM-International holdings, these data are based on complete census counts. Data for some 50 countries were acquired from national statistics agencies through published reports and by request. For a number of countries, data from multiple sources are held. As of mid-2013, the IMAGE repository held data for 103 countries (Table 2), including 97 origin-destination matrices. Table 2 about here For this paper, we analyse a subset of data held in the repository, first assembling aggregate migration intensities capturing all permanent changes of address. These data were available for 24 countries. Where these data were unavailable, we targeted countries with detailed origin-destination matrices (> 200 zones), along with country matrices at multiple levels of geography. From these data we were able to estimate aggregate migration intensities using the method proposed by Courgeau et al.(2012 ). These data were held for 47 countries in the repository, bringing our league table coverage to 71. The IMAGE Studio The IMAGE Studio has been developed to facilitate the analysis and modelling of internal migration in any country, dependent upon the provision of an origin-destination area matrix of flows between Basic Spatial Units (BSU), a vector of area populations and a set of digital boundaries of the corresponding BSUs. The IMAGE Studio is structured as a set of four linked but autonomous systems: (i) data preparation, (ii) spatial aggregation, (iii) internal migration indicators, and (iv) spatial interaction modelling. The data preparation system (i) readies data from the IMAGE Repository, including origin-destination matrices, populations at risk of moving and digital boundary files of BSUs, for input to the IMAGE Studio analytical systems. Outputs include migration flows between origin-destination, vectors of populations and spatial continuity files. The spatial aggregation system (ii) creates novel spatial configurations of migration input data by stepwise aggregation of BSUs into Aggregate Spatial Regions (ASRS) of varying shapes and sizes. Migration is fundamentally dependent on the spatial framework over which it is measured. The construction of multiple levels of geography from a single set of BSUs allows users to explore how the number and shape of zones over which a migration is measured affects a range of migration indices, including migration intensity. This facility was developed purposely to address the Modifiable Areal Unit Problem (MAUP) identified by Openshaw (1984). The internal migration indicators system (ii) computes global internal migration indicators for the spatial configurations generated by the spatial aggregation system and for BSUs, and local (areaspecific) indicators for the set of BSUs. The indicators include those suggested by Bell et al. (2002) as being suitable for comparing migration in different countries including Crude Migration Intensity, Aggregate Net Migration Rate, Migration Effectiveness Index, Migration distance, The Coefficient of Variation and the Gini Index of Connectivity, among others. The fourth system calibrates a doubly constrained spatial interaction model (SIM) either for the migration flows for the initial set of BSUs or for the migration flows for each set of ASRs. Impediments to cross-national comparisons

Data on internal migration collected around the world is the least standardised of any demographic variable. As previously described, there is variation in the migration measure employed (event/transition/latest move), the interval over which a migration is recorded, and the spatial framework. Harmonising statistics for cross-national comparisons over these dimensions presents a range of challenges: some of which are minor and can be safely ignored, others which are intractable, and finally those which can be addressed through statistical means. We review these impediments and our proposed solutions in turn. Transitions, events, and latest move data There are three principle data types on recent mobility collected by countries around the world: transition data, event data and latest move data. These data types differ in how migration is measured as well as in population coverage (Table 3). Transition data are most commonly collected by census and surveys, and compare place of residence at the time of enumeration with place of previous residence at an early point in time, most commonly one or five years ago. Transition data provide a measure of the number of movers within a population over a given interval, but do not capture repeat or return movers. Transition data also do not measure movements by people who enter or exit the population (through birth, death and international migration) during the interval. By contrast, event data are typically collected by population register and count all movements that occur over a given interval. Event data capture multiple moves by individuals over a given period including return and repeat migration, as well as internal movements by people who enter or exit the population during the measurement interval. These two measures of migration diverge as the interval lengthens due to repeat movers in the population, with the number of moves continuing to increase arithmetically as the number of movers stabilises. Table 3 about here Latest move data are the third type of internal migration data widely collected by countries around the globe. These data measure migration by combining information on respondents duration of residence with information on their place of previous residence. Unlike transition data, latest move data enumerate all movers over a defined interval, including return migrants who are missed by fixed interval questions. Repeat movers are also enumerated in latest move data, however, the origin of these movers will differ from the origin captured by a fixed transition question. The population coverage also differs. Unlike transition data, latest move data include movements by people entering the population (through births and international migration) during the interval. While individuals not alive at the beginning of the interval are readily excluded, international migrants who make a subsequent internal move during the interval will be counted as movers. Latest move data arguably provide a more complete coverage of migration than transition data, capturing all members of the end of interval population who make a move, but unlike event data they do not capture all moves. These differences in the migration measurement and population coverage between the three data types are an important limitation for cross-national comparisons over extended intervals. Over short intervals (e.g. less than one year) the impact, however, is likely to be relative small reflecting the low likelihood of individuals making multiple movements in a single

year. These three data types can therefore be considered to be broadly comparable provided migration intensities are measured over a single year interval. Temporal harmonisation Harmonisation of transition data measured over different length intervals is a major impediment to cross-national comparisons of internal migration. Over short intervals (e.g. one year) the number of moves and movers are roughly equivalent (Long and Boertlein 1990). As the interval increases, however, the number of moves and movers diverge due to the presence of repeat movers within the population. The upshot is that five year transition rates rate not equivalent to five times the one year transition rate, and the relationship between the two will vary both over time and between countries, depending on the level of repeat movement within the population. Despite persistent attempts, no straightforward analytical solution has been found to the problem of comparing migration intensities measured over different length intervals (see e.g. Rogerson 1990) so cross-national comparisons need to be confined to countries collecting data over intervals of the same length.. Spatial harmonisation Even within groups of countries collecting the same type of data over the same time intervals, comparisons are prejudiced by differences in the number of spatial units into which they are divided: the finer the spatial mesh, the greater the number of migrants/migrations observed. Since countries differ widely in size, statistical geography and patterns of settlement, simple cross-national comparisons of migration intensities reported by national statistical agencies are not viable. To circumvent this problem, we propose a measure of aggregate migration intensity which encompasses all permanent changes of address. The problem is that very few countries around the world collect or disseminate statistics that capture all residential moves. Globally, combining information on census transitions with data on duration of stay, together with estimates from population registers and surveys, delivers direct estimates of aggregate migration intensities for 24 countries. To generate estimates for those countries which do not collect such data, we build on the approach originally developed by Courgeau (1973) and subsequently adapted by Courgeau et al. (2012 ), which generates an estimate of aggregate migration intensity for each country by fitting a regression line to intensities measured at a range of geographic scales. The underpinning logic is that, as the number of zones into which a territory is divided increases, so the number of inter-zonal migrants rises. Courgeau et al. (2012 ) deduce a linear relationship between the crude migration intensity (CMI) observed at a given level of disaggregation, j, and the logarithm of the average number of households per zone, H, at that spatial level. In equation 1, the parameter k scales this relationship and w is a constant. For countries which provide migration data at more than one level of spatial scale (e.g. states, provinces, counties, etc) it is therefore possible to estimate equation 1. Substituting Hj = 1 corresponds to a hypothetical level of spatial resolution at which there is just one household per zone and therefore captures all migrations. Since ln(1) = 0, the corresponding CMI can be read directly from the y intercept on a graph or computed from equation 1 as the constant w.

CMI j = w + k ln (H/j) [Equation 1] We compute Courgeau s constant in two ways. For ten countries with matrices containing more than 200 regions, we use the spatial aggregation sub-system of the IMAGE studio to calculate intensities for a cascading sequence of zones beginning with the basic spatial units and aggregating upwards in increments of ten. This delivers estimates of migration intensity at a series of spatial scales for the same country, not readily available from published statistics. Multiple iterations of the spatial aggregation algorithm ensure that the result reflects movement across a random spatial framework (Stillwell, Daras et al. 2013).We then calculate aggregate migration intensities by fitting a regression line through the mean observation at each of these multiple spatial levels and record the y intercept. For countries lacking matrices at a fine-grained level of spatial resolution, we calculate aggregate migration intensity in the same way, but fit the regression line to just the data points derived directly from the available data that is the record of migrants or migrations at various levels of the administrative hierarchy employed in each national statistical system (Law 1999) e.g. regions, provinces, districts and counties. Data are needed for at least three geographic levels to implement this approach. These estimates are less reliable than those produced using the IMAGE Studio, due to the spatial bias inherent in administrative geographies, but still provide some basis for comparison. A league table of aggregate internal migration intensities Table 4 ranks 71 UN member states by aggregate migration intensities. Countries are grouped separately according to whether five year or one year interval data are used. Five year aggregate migration intensities were calculated for 37 countries. These were based on observed data for 12 countries, generated for seven countries using the spatial aggregation subsystem of the IMAGE Studio, and based on multiple administrative geographies for the remaining 18. Five year intensities range between 10.5 per cent for the Philippines to 56.2 per cent in Fiji. The median five year intensity across the sample was 22.5 per cent. The highest five year intensities are observed in the new world countries of New Zealand, the United States of America, Australia and Canada, along with Fiji, South Korea, Cameroon, Chile and Switzerland. Results for the four new world countries and Switzerland are remarkably consistent with Long s (1991) findings on residential mobility in the early 1980s. The lowest five year intensities are observed in countries in South-East Asia (e.g. The Philippines, Indonesia and Vietnam), Africa (e.g. Mauritius) and Latin America (e.g. Mexico, Honduras, and Nicaragua). Taken together, there appears to be a broad positive association between level of development and five year migration intensities. This is most readily evident within regional groupings. For example, southern European countries (e.g. Malta, Greece, and Portugal) register lower five year intensities than those in Western Europe (e.g. France and Switzerland) while countries in South East Asia (e.g. The Philippines, Indonesia and Vietnam) register lower intensities than their more highly developed East Asian counterparts (e.g. Japan and South Korea). Table 4 about here

One year migration intensities were calculated for 41 countries, with the majority of the sample located in Europe (29) and Asia (6), but only thin coverage of North America (2) Oceania (1), Africa (2), and Latin America (1). Around half of the one year intensities are based directly on observed data (19). Estimates for three countries were computed using the IMAGE studio, while the remaining 20 were generated using data observed at multiple administrative levels. Values for one year intensities range from 1.0 per cent in Macedonia to 19.1 per cent in Iceland. Scandinavian countries display consistently high one year intensities, ranging from 17.0 per cent in Finland to 12.2 per cent in Denmark. More generally, there is a high to low gradient in one year intensities travelling from Northern to Southern Europe, and from Western to Eastern Europe. The picture across the rest of world is more heterogeneous. India records the second lowest one year intensity in the sample (1.3 per cent), while Kenya records the second highest (18.4 per cent). The new world countries (Australia, Canada and USA) again all record high overall intensities echoing the earlier findings of Long (1991). Approaches to explanation There are a range of potential approaches to explaining these differences in migration intensity between countries and regions around the world. In his seminal paper in the early 1990s, Long (1991) sought explanation for high mobility in the four New World countries (USA, Canada, Australia and New Zealand) by reference to peripatetic traditions inherited from immigrant forbears. Also relevant were the associated institutional frameworks, and the relative openness of housing and employment markets. At a broader, conceptual level, Zelinsky (1971) also sought to link the overall level or intensity of migration to progress through the demographic transition, arguing that there were definite patterned regularities in the growth of personal mobility through space-time. On this basis, a close link might be expected between the level of mobility in individual countries and their degree of modernisation, though Zelinsky himself anticipated different trajectories for particular forms of movement, and the overall transition thesis has been subjected to wider criticism (Cadwallader 1993). An alternative approach is to build on the life course model. Which focuses on the role of life course transitions and housing adjustment in triggering migration. For example, recent work by Bernard et al.(under review) reveal a strong association between the age patterns of migration and age structure of various life course transitions across 27 countries. As the authors note, there are other contextual factors that trigger migration independent of age, and it is not year clear to what extent these might drive cross-national differences in aggregate migration intensities. For the analysis presented here, we seek explanation for cross-national differences by searching for associations between mobility and a range of demographic, economic and social variables. We calculate correlation coefficients between 13 national indicators sourced from the United Nations and The World Bank and one year and five year aggregate migration intensities (Table 5). Table 5 about here Correlation coefficients for the selected variables range from -0.44 to 0.67 across the one year and five year estimates of aggregate migration intensity. Across the countries in our sample,

the strongest correlation coefficient is recorded between GDP per capita (2005 PPP$) and one year migration intensity (0.67). GDP per capita is also strongly positively associated with five year migration intensities (0.62). There is also a moderate positive association between five year intensity and the human development index (0.60) providing at least partial support for Zelinsky s (1971) hypothesis of the mobility transition in which intensity increases during initial stages of development, however, the relationship between one year intensities and HDI is somewhat weaker (0.49). There is also a positive association between the proportion of the population with a mobile phone subscription and one and five year migration intensities. One interpretation is that greater connectivity is facilitating, rather than substituting for, internal migration. There are moderate associations between a number of demographic indicators and one and five year migration intensities. The level of urbanisation within a country is positively associated with overall intensity, underlying the importance of urban-ward and inter-urban migration flows in driving internal migration. Net international migration rates are also positively associated with five year migration intensities (0.58), but only weakly associated with one year intensities (0.26). This provides mixed support for the substitution of internal migration with international movements in countries experience net international migration losses, as does the negative association between remittances and five year migration intensities (-0.44). The positive association between net international migration rates and five year migration intensities may also reflect displacement of domestic populations from city regions in countries with net international gains. Median age is also positively associated with five year intensities. As populations age, we would expect an increase then subsequent decline in migration intensities, as the population moves through key migration ages. There is no clear evidence of the anticipated curvilinear relationship in these data. Conclusion This paper reports the most comprehensive league table of overall internal migration intensity ever produced. Cross-national comparisons of internal migration are beset by a range of conceptual and methodological challenges. By focusing on overall intensity and utilising the method proposed by Courgeau et al. (2012) it has been possible to generate broadly comparable estimates for some 71 of 193 UN member states. Results reveal substantial variation in the level of mobility across countries. There are striking regionalisations in these data, with high five year intensities recorded across the new world countries, while countries in South-east Asia record among the lowest intensities. Results for one year intensities in Europe suggest a clear geographic gradient likely reflecting variation in economic development and substitution of internal migration with international flows. Further regional assessment of these patterns is required. For the sample of countries reported in this paper, there is a clear association between five and one year migration intensities and the level of economic development. There is also an association between demographic processes of international migration, urbanisation and ageing. Further investigation of these relationships is warranted and sets a clear agenda for future work.

References Bell, M., M. Blake, et al. (2002). "Cross-national comparison of internal migration: issues and measures." Journal of the Royal Statistical Society 165(3): 435-464. Bell, M., E. Charles-Edwards, et al. (forthcoming). "Internal migration data around the world: Who collects what?" Population and Devlopment Review. Bell, M., E. Charles-Edwards, et al. (under review). "Internal migration data around the world: Who collects what?". Bell, M. and S. Muhidin (2009). Cross-National Comparisons of Internal Migration. Human Development Research Paper 2009/30. New York, United Nations. Bernard, A., M. Bell, et al. (under review). "Explaining Cross-National Differences in the Age Profile of Internal Migration: The Role of Life-Course Transitions ". Cadwallader, M. (1993). "Classics in human geography revisited: commentary 1." Progress in Human Geography 17(2): 213-219. Courgeau, D. (1973). "Migrants and migrations." Population 28: 95-128. Courgeau, D., S. Muhidin, et al. (2012 ). "Estimating changes of residence for cross-national comparison." Population (English edition) 67(4): 631-651. Law, G. (1999). Administrative Subdivisions of Countries: A Comprehensive World Reference, 1900 through 1998. Jefferson, North Carolina, McFarland and Company. Long, L. (1991). "Residential mobility differences among developed countries." International Regional Science Review 14(2): 133-147. Long, L. and C. Boertlein (1990). Comparing migration measures having different intervals. Current Population Reports series, P-23, no. 166. Washington, DC, U.S. Department of Commerce, Bureau of Census. Nam, C. B., W. J. Serow, et al., Eds. (1990). International Handbook on Internal Migration. New York, Greenwood Press. Openshaw, S. (1984). The Modifiable Areal Unit Problem. Norwich, Geo Books. Rees, P., M. Bell, et al. (2000). "Problems and solutions in the measurement of migration intensities: Australia and Britain compared." Population Studies 54(2): 207-222. Rees, P. and M. Kupiszewski (1999). "Internal migration: what data are available in Europe?" Journal of Official Statistics 15(4): 551-86. Rogers, A. and L. Castro (1981). Model Migration Schedules. Laxenburg, International Institute for Applied Systems Analysis. Rogerson, P. A. (1990). "Migration analysis using data with time intervals of differing widths." Papers of the Regional Science Association 69: 97-106. Skeldon, R. (1997). Migration and Development: A Global Perspective. Longman, Harlow, England. Stillwell, J., K. Daras, et al. (2013). The IMAGE Studio: A Tool for Comparative Internal Migration Analysis and Modelling. International Conference on Computers in Urban Planning and Urban Management. Utrecht, The Netherlands. The World Bank (2009). World Development Report 2009: Reshaping Economic Geography. Washington D.C., The World Bank. United Nations (1978). Statistics of Internal Migration: A Technical Report Studies in Methods, Series F23, ST/ESA/STAT/SER F/23. New York, Department of International Economic and Social Affairs, United Nations. United Nations (2009). Human Development Report 2009, Overcoming barriers: Human Mobility and Development. Basingstoke, United Nations Development Program. Zelinsky, W. (1971). "The hypothesis of the mobility transition." Geographical Review 61(2): 219-249.

Table 1: Countries collecting transition data at the 2000 Census round by continent Region Place of Transition interval Latest move Total Birth Countries Other One Five defined year year interval Africa 28 9 9 8 13 31 Asia 22 0 12 4 16 30 Europe 27 14 4 12 10 32 Latin America and Caribbean 28 2 16 2 12 28 North America 3 1 3 0 0 3 Oceania 10 2 8 2 2 13 Total 118 28 52 28 53 137 Source: IMAGE Database: Global Inventory of Internal Migration Note: many countries collect more than one type of data Table 2: Selected data holdings, IMAGE Repository Region Origindestination matrix National counts of movers All moves Countries with data Countries in regions Africa 16 7 1 17 54 Asia 17 18 4 21 46 Europe 35 33 13 36 44 Latin America and 22 22 13 2 Caribbean 32 North America 3 3 2 3 3 Oceania 4 3 2 4 14 Total 97 77 24 103 193 Source: IMAGE Database: Repository of Internal Migration Note: many countries collect more than one type of data Table 3: Selected data holdings, IMAGE Repository Data Type Migration measure Population coverage Events Captures all moves undertaken within a defined interval All individuals in the population over the observation interval Transitions Captures change in residence between the time of enumeration and some point prior (usually one or five years) Individuals resident in the country at both the beginning and end of the interval. Individuals entering and exiting the population during the interval (e.g. births, deaths and international migration) are excluded. The exception are individuals who were resident in the country at the beginning of the interval, departed internationally and then Latest Move Captures most recent move within a defined interval returned prior to the end of the interval. Individuals resident in the country at the end of the interval. Individuals exiting the population during the interval (e.g. deaths and international migration) are excluded. Individuals entering the population (e.g. births/migration) before making an internal move are counted.

Table 4: League Table of One Year and Five Year Aggregate Migration Intensity Data type Country Region Year Event Latest move 1Y transition 5Y transition Estimation Method 1 Cameroon Africa 2005 41.4 COURGEAU 2 Senegal Africa 2002 29.9 COURGEAU 3 South Africa Africa 2001 21.2 COURGEAU 4 Morocco Africa 2004 21.0 COURGEAU 5 Ghana Africa 2000 19.0 IMAGE 6 Mauritius Africa 2000 12.0 OBSERVED 7 Burkina Faso Africa 2006 4.2 COURGEAU 8 Kenya Africa 1999 18.4 COURGEAU 9 South Korea Asia 2000 52.8 OBSERVED 10 Israel Asia 2000/ 1995 7.0 28.2 OBSERVED 11 Japan Asia 2000 11.9 7.9 27.6 COURGEAU/ OBSERVED 12 Nepal Asia 2001 21.5 COURGEAU 13 Malaysia Asia 2000 17.1 OBSERVED 14 China Asia 2000 15.9 COURGEAU 15 Vietnam Asia 1999 13.5 COURGEAU 16 Indonesia Asia 2000 12.6 COURGEAU 17 Philippines Asia 2000 10.5 IMAGE 18 Cambodia Asia 1998 5.0 COURGEAU 19 India Asia 2001 1.3 COURGEAU 20 Kyrgyzstan Asia 1999 11.7 COURGEAU 21 Turkey Asia 2009 12.4 COURGEAU 22 Switzerland Europe 2000 10.7 36.1 OBSERVED 23 France Europe 2006 34.0 OBSERVED 24 Portugal Europe 2001 25.3 COURGEAU 25 Greece Europe 2001 22.5 COURGEAU 26 Malta Europe 1995 4.7 19.8 OBSERVED 27 Macedonia Europe 2003 1.0 COURGEAU 28 Austria Europe 2002 8.1 OBSERVED 29 Belarus Europe 2004 5.7 COURGEAU 30 Belgium Europe 2005 11.1 IMAGE 31 Bulgaria Europe 2003 4.5 COURGEAU 32 Croatia Europe 2001 2.7 COURGEAU 33 Cyprus Europe 2001 3.8 OBSERVED Czech Republic Europe 2000 3.4 COURGEAU 34 35 Denmark Europe 2006 12.2 OBSERVED 36 Estonia Europe 2000 2.8 COURGEAU 37 Finland Europe 2000 17.0 OBSERVED 38 Germany Europe 2009 8.9 IMAGE 39 Hungary Europe 2001 7.4 COURGEAU 40 Iceland Europe 2000 19.1 OBSERVED 41 Ireland Europe 2002 6.6 OBSERVED 42 Italy Europe 2000/ 2001 8.6 5.1 COURGEAU/ OBSERVED 43 Latvia Europe 2005 5.1 COURGEAU 44 Lithuania Europe 2004 5.5 COURGEAU 45 Poland Europe 2000 2.8 COURGEAU 46 Netherlands Europe 2000 10.1 OBSERVED 47 Norway Europe 2000 12.8 OBSERVED 48 Romania Europe 2000 2.1 COURGEAU 49 Slovenia Europe 2002 2.5 COURGEAU 50 Slovakia Europe 2001 2.7 COURGEAU 51 Spain Europe 2000 4.6 COURGEAU 52 Sweden Europe 2000 12.7 OBSERVED 53 UK Europe 2001 11.5 IMAGE 54 Chile Latin America 2002 41.1 IMAGE 55 Costa Rica Latin America 2000 34.5 COURGEAU 56 Bolivia Latin America 2001 31.5 COURGEAU 57 Paraguay Latin America 2002 29.9 COURGEAU 58 Peru Latin America 2007 25.1 COURGEAU 59 Brazil Latin America 2000 19.0 IMAGE 60 Argentina Latin America 2001 18.6 COURGEAU 61 Ecuador Latin America 2001 18.2 IMAGE

62 Dominican Republic Latin America 2002 17.4 COURGEAU 63 Nicaragua Latin America 1995 14.7 COURGEAU 64 Honduras Latin America 2001 14.3 IMAGE 65 Mexico Latin America 2000 12.9 IMAGE 66 Colombia Latin America 2005 8.6 OBSERVED 67 USA North America 2000 12.5 44.3 OBSERVED (S) 68 Canada North America 2006 13.3 38.5 OBSERVED 69 Fiji Oceania 2007 56.2 COURGEAU 70 New Zealand Oceania 2006 54.7 OBSERVED 71 Australia Oceania 2001 17.6 42.4 OBSERVED Table 5: Correlation Coefficients, One and Five Aggregate Migration Intensities and Selected Indicators Economic Social Demographic Variables 1 year aggregate migration intensity 5 year aggregate migration intensity GDP per Capita ( 2005 PPP$) 0.67 0.62 Gini Coefficient (Income Inequality 2000,2005) -0.03-0.18 Foreign Direct Investment as a proportion of GDP (2000) 0.02-0.13 Female Labour Force Participation (2000) 0.26 0.10 Labour Force Participation (2000) 0.18 0.25 Human development index (2000) 0.49 0.60 Mobile Phone Subscribers (2000) 0.65 0.62 Literacy (2000) 0.09-0.03 E0 (2000-2005) 0.50 0.41 TFR (2000-2005) -0.02-0.29 Growth rate (2000-2005) 0.16-0.28 Median Age 0.21 0.49 Urbanization (2000) 0.60 0.53 Net International Migration rate (2000-2005) 0.26 0.58 Remittances as % of GDP (2000) -0.32-0.44 Nb. Outliers have been removed from the analysis of one year intensities (Kenya and Kyrgyzstan) and five year intensities (Fiji and Cameroon)