Internal migration in Northern Ireland: Are people becoming more stuck in place?

The analysis uses the Northern Ireland Longitudinal Study to explore migration trends from 2001 to 2015. Its contributions are to extend this analysis to the period after the 2011 Census, to use administrative data to measure migration not as a transition between places but as (potentially) multiple events over the full 2001 – 2015 period over different distance bands, to add to the UK evidence base on internal migration trends, and to show how long-term trends appear to override economic cyclical effects. The results show that internal migration rates over all distance bands fell from 2001 onward, continuing a decline in migration between Super Output Areas recorded since 1981, the first date for which there are data. This indicates that Northern Ireland, despite its unique political and social history, has not been immune to the long-term migration decline observed elsewhere in Great Britain and other countries such as the United States. The only caveat to this is that when moves from Northern Ireland to other places are included in the long-distance band of 50 km or more then long-distance migration returns to near its 2001 – 2003 level by 2013. The analysis also shows that the same personal characteristics as elsewhere are related to different frequencies and distances of moving.

trajectory within an increasingly devolved United Kingdom, which has seen the decentralisation of some powers.However, at the same time, it shares many features in common with other advanced economies such as counterurbanisation, deindustrialisation, occupational shifts, ageing, and growing levels of education, all of which influence migration rates (Green, 2018) over longer-term periods of 30 or 40 years.This combination of uniqueness and shared common factors makes NI an interesting case study to examine how general the decline in internal migration might be.There is also considerable interest in understanding how mobile the country's population is against the background of changing levels of residential segregation and in meeting the aspirations for a society that is more mixed in religious as well as on other social and economic dimensions than it is currently (Northern Ireland Executive, 2013).
The paper builds on previous work on migration rates in NI, which examined trends in people's propensity to move between decennial censuses over the period 1981-2011 (Campbell, 2018).It traces migration events taking place between April 2001 and October 2015, using administrative data from the healthcare system that provides details of people's addresses at six-monthly intervals, which are linked to data on people's characteristics drawn from the 2011 Census to see if the migration decline noted after 1981 continued after 2011 and whether economic and political shocks after 2001 influenced migration levels.It explores the correlates of moves over different distance bands and with different frequencies, and finally, it offers a direct comparison to the Scottish analysis of McCollum et al. (2020), which employs similar data and methods.Before the results are presented, they are contextualised by the next sections, which introduce the NI background, present the expectations from previous literature, describe the NILS data, and outline the methods used.The paper concludes with a discussion of its findings and their implications.

| NORTHERN IRELAND: UNIQUE OR NOT?
As just mentioned, it is tempting to view NI as being unique, at least in postwar Europe, because of its troubled political history, social divisions, and legacy of religion-based residential segregation (Shirlow & Murtagh, 2006;Shuttleworth & Lloyd, 2009).As its foundation during the partition of Ireland was based in large part on demographic considerations to ensure a Protestant majority, it is no surprise that population has become a highly political issue whether in terms of total numbers or geographical distribution (Anderson & Shuttleworth, 1998).Religion is a marker of British and Irish national identities in a geographical arena where neither British nor Irish state building has been wholly complete or successful.Three decades of civil conflict in the so-called Troubles from 1969 also set NI apart, as also do its current governmental arrangements based on the assumption of two voting blocks, British Unionists and Irish Nationalists.This background, including the transition since the 1990s from explicit violence, cannot fail to have influenced internal migration rates over the long term, given the role of political violence and intimidation in forcing people to change address as The Troubles developed in the early 1970s (Darby & Morris, 1974).It sets the scene for the decrease in migration observed between 1981 and 2011 presumably from higher levels of movement in the politically and economically turbulent 1970s, which in turn is the context for our post-2001 analysis.Other available indirect evidence suggests that migration rates were high in the later 1960s and early 1970s, with substantial local demographic changes occurring as a result (CRC 1971;Gregory, Cunningham, Ell, Lloyd, & Shuttleworth, 2013).A similar inference can be made from considering residential segregation rates, which, for example, increased markedly between 1971 and 1991 as the two communities drew apart but then remained steady between 1991 and 2001, with some evidence of a fall between 2001 and 2011 (Shuttleworth & Lloyd, 2009, 2013).The apparent fall in migration in more recent years could be read perhaps as a result of the start of the Peace Process or as a consequence of the same factors causing the migration decrease seen in other societies.
On the other hand, whilst there is a credible case to be made that NI is unique, it should not be pressed too far.NI does not exist in a bubble-it has also experienced many of the labour market, social and demographic changes that have been seen elsewhere in the United Kingdom, and, indeed with variations, in other high-income countries.
It is not the intention to survey these in detail here but it is worthwhile noting that, just as elsewhere, there has been a decline in manufacturing jobs, growth in service employment, an increase in the proportion of the population with secondary and higher educational qualifications, and an increase in the population's average age (Rowland, 2019;Teague, 1993) all of which are compositional drivers of internal migration (Green, 2018).To these can be added geographical shifts in the population, notably a fall in the population of Belfast and particularly of its urban core (Gregory et al., 2013;Power & Shuttleworth, 1997), no doubt partly as a result of violence as noted above but also because of housing clearance/improvement and the move to suburbs and smaller settlements seen in other national contexts (Champion, 1989).In summary, there is enough to set NI apart as a special case, and thus a good test for the generalisability of the migration decline, but enough in common with other parts of the United Kingdom (and other countries too) to make useful analytical comparisons.

| EXPECTATIONS AND EVIDENCE FROM THE EXISTING LITERATURE
It is unnecessary to review the whole literature on declining rates of internal migration ab initio as this is done elsewhere (Cooke, 2011;Champion et al 2018), but it is useful to reflect on the social and demographic correlates of migration as it is through changes in the composition of a population in terms of these that a start can be made on analysing temporal trends in migration rates over the long term.Green (2018) provides a comprehensive review of population changes and their expected impacts on aggregate migration levels, with key features including population ageing, rising educational levels, increasing home ownership, rises in the numbers of single and divorced people, and a variety of changes in the labour market.To these might be added technical factors, such as internet use, and cultural preferences such as Cooke's (2011) "secular rootedness."The longer the analytical time period, the bigger these changes will be and the greater their effect, although compositional changes should have some effect over the 14-year period between 2001 and 2015 of the analysis in the present paper.
As regard ageing, the expected effect would be to decrease migration rates, everything else being equal, because older people tend to move home less frequently.Therefore, a growing proportion of older people will shift the population into less migratory age groups and will tend to force migration rates down.The growth of owner occupation and decline in renting in the housing market will also push migration rates down, because renters tend to be more mobile than owners but the increase in private renting since 2001 might increase migration propensities.
The other two changes suggested above would tend to act to increase migration rates.More educated people tend to move more over longer distances, and hence, the growth of the share of the population with school and postschool qualifications would be expected to boost longer-distance migration rates.The same outcome would also be forecast as a result of changes in relationship statusincreased proportions of single and divorced people, who tend to be more spatially mobile than married people (Cooke 2008), should increase overall address changing and longer moves.
Labour market change might also be expected on balance to act to increase migration rates in the long term and to lead to higher migration in our analytical period.The number of jobs in NI steadily increased from 1991 onwards before being checked by the Great Recession but then resuming growth, and there has been a fall in the number of unemployed people since 2001 who are often less mobile-apart from in the immediate aftermath of the recession (Rowland, 2019).At the same time, there has been a historical sectoral move away from older manufacturing industries such as shipbuilding towards service employment, which has been matched by occupational shifts to service-type jobs.These developments might be expected to increase migration rates as the share of people with characteristics associated with higher rates of migration has grown into the 21st century (Green, 2018).
The factors that are less easy to hypothesise about are those specific to NI.The ending of the Troubles and the transition to peace may have acted to reduce migration as fewer people were compelled to move.Moreover, the net fall in Belfast's population, which was mostly between 1971 and 1991, seems to have ceased in later decades, and segregation grew from 1971, peaked in 1991, and was thereafter stable or falling, reinforcing the impression that NI's population was more dynamic and mobile before the 1990s than since then.On the other hand, the transition to peace in the 1990s, coupled with growing affluence, might have served to release pent-up demand for moving home and thereby increase address changing rates.
Evidence on longer-term trends is available from Campbell (2019) arising from previous work on the NILS.This showed a progressive fall in the rate of NI's internal migration since the 1980s.Whereas 42% of NILS members aged 25-74 in 1981 and who also had a census record in 1991 had moved between Super Output Areas (SOAs) over the decade, the proportion of the equivalent population in the intercensal decade 1991-2011 was just 36% and it fell further to 34% in the 2001-2011 period.There are 890 SOAs in NI with an average population of around 2,000 people.It was also found that the majority of this 8 percentage point decline between the 1980s and the 2000s was attributable to changes in migration behaviour, with only a very small proportion (less than one-eighth) being explained by changing population composition.In this latter regard, it would seem that population ageing and changes in housing tenure that tend to pull migration down (for example, shifts towards older people and away from renting) have over this longer term tended to offset the effects of the sorts of changes that would tend to push migration rates upwards (such as the increase in single and divorced people).Increased higher educational participation probably had an upwards effect but was not analysed because of difficulties with the education variables in 1981.
This finding-that migration rates have fallen at least at one spatial scale (that of the SOA) -meshes well with circumstantial evidence about what is known about the NI's population from the other perspectives noted earlier.For instance, the large increase in residential segregation between 1971 and 1991 (Shuttleworth & Lloyd, 2009) implied substantial migration as a contributing factor as did the large population losses from inner-city Belfast and indeed from the city as whole (Gregory et al., 2013;Power & Shuttleworth, 1997)

| USING THE NILS TO ANALYSE MIGRATION
NILS is the only source that has sufficient detail to answer the questions set out in the previous section.This dataset is based on the routine linkage of records from NI's healthcard register to census and administrative data such as that from the Valuation and Lands Agency.
The healthcard register is nearly universal and includes anyone registered with the National Health Service at a doctor's practice.It captures address changes, from 2001, as they are reported to health professionals-it is essential to provide either a new or updated home address when registering with a doctor, getting treatment, or receiving a prescription.Moves from NI are captured by deregistrations, moves to there by new or re-registrations.Some demographic groups (for example young men) lag in reporting address changes (Barr & Shuttleworth, 2012) but most update their records after a delay (Foley et al 2017) and, when assessed, migration statistics drawn from administrative sources are closely comparable with those from the census in which they are self-reported (Foley, 2017).This permits migration to be measured as a series of multiple events for any one individual, with a maximum of one move for each 6-month period.
Furthermore, since house locations are georeferenced very accurately, it is possible to calculate the precise distances moved.For the purposes of this paper, the distance variable was categorical since a precise individual distance of move was judged to be disclosive.These categories were aggregated to give those used in the paper.
The NILS comprises an approximately 28% sample of the NI's population (drawn from a random selection of 104 birthdates), which translates to a sample size of between 450,000 to 500,000 individuals at any single time point but cumulatively far more if the total is reckoned as anyone who has ever been a NILS member.The NILS is linked to the 1981, 1991, 2001, and 2011 Censuses and holds vital and migration events after 2011 used for this particular analysis.A full range of individual census variables are available for each NILS member, but there is also census information available for other household members.Since the NILS contains individual-and household-level microdata from multiple censuses and administrative sources, the data can only be accessed by accredited researchers in a secure and supervised data laboratory in the Northern Ireland Statistics and Research Agency (NISRA).
These detailed data on internal migration are conceptualised in two ways in this study; as binary (any address change as opposed to no address change) and then as a multiple category variable (moves of less than 10 km, moves of 10-50 km, and moves of 50 km or more [with or without emigration from NI]).NI is relatively small geographically-some 160 km from east to west-so sample members who left NI were included as the equivalent of longer-distance moves as, for instance, those between regions in England and Wales or as longer moves in Scotland.Around 75% of moves outside NI were to other parts of the United Kingdom at the start of the period, this proportion falling towards its end but remaining at over 50% and including an unknown share of movers within the island of Ireland and who crossed the border.This variable, along with all address changes, is our main focus.Apart from one important exception-see below-the results were the same whether or not moves to locations outside NI were included in the more than 50-km category.
There are some caveats.To preserve confidentiality, a continuous distance measure was not provided; instead, category data were recoded as described above.Administrative datasets are also subject to cleaning and updating.In this case, 2012 saw many more moves reported than other years because of a concerted attempt to revise the healthcard register.This is acknowledged and, in the later quantitative analysis, dealt with by a year dummy for April and October 2012.The sample used was for NILS members aged 16-64 in 2011 and possessing a census record in that year, and whose georeferenced location in 2011 in the healthcard register and census matched at the time of the census or within 6 months of it.On the basis of these criteria, there were approximately 252,000 sample members.Note that this sample would be aged 6-54 in 2001 and would be 20 to 68 years old at the end of our reference period in 2015.
The selection of independent variables for inclusion in the analyses was informed by pragmatic considerations of what was available in the census and also by reference to the literature and past precedent (see Shuttleworth et al 2019;Cooke, 2011).They comprise age, gender, religion, socio-economic status, economic activity, health, education, and SOA social deprivation.There is a problem that is common to all analyses of this kind, which arise from the capture of the independent variables from a census.Many of the explanatory variables (e.g., education, health, housing tenure, and economic activity) naturally vary through time as people fall ill, become healthy, or move house.However, we only know the status of sample members every 10 years when a census occurs so these have to be treated as time  2004-2006, 2007-2009, 2010-2011, 2012, and 2013-15 are used with the exception of 2012, which was, as noted above, a year with administrative updates.

| RESULTS
The results of the first stage of the analysis is set out in Figure 1 and Table 1, which show, respectively, the annual rate (aggregated from the six-monthly data) of address changes by distance band between 2001 and 2015 for all NILS members, and the demographic characteristics of the analytical sample with selected different migratory profiles.On this basis the total address changing rate starts at about 4.2% for the first half-year period and then fluctuates between 6% and 10% for whole years-remember this may also record two moves by the same person in a single year-peaking in 2006/2007 at a time when the NI housing market was buoyant but falling afterwards despite a small increase in 2011.Barring this peak, the overall trend for migration within NI appears stable or perhaps slightly declining but with the important caveat that, when migrants who leave NI are counted in the greater than 50-km category, this rate increases towards the end of the period.However, there are limits to the conclusions that can be drawn from visual inspection of this figure; aggregate annual migration rates may alter, for example, because of economic and political events/shocks or else the continuation of longer-term trends.
Two important initial points to note, before turning to Table 1, are that few people have high numbers of multiple address changes; most people do not move at all or only change address once or twice (Foley, 2017) and those who do move seldom move over great distances.Bearing in mind the outcome variable that is the focus for later quantitative analysis (no moves, moves of less than 10 km, moves of 10 to 50 km, and moves of more than 50 km excluding or including people who leave NI), it was possible to create a typology of migration trajectories when a sample member changed address.There were a vast range of possible migration sequences, and it was simply impossible to tabulate them all so Table 1 shows the sample profile, that of members who did not move, those who moved more than once over 50 km (including leaving NI), and those who had two or more moves of less than 10 km.These were chosen on the basis of illustrating differences between different types of sample members' migratory profiles.The percentages sum to 100% vertically.
Table 1 shows that there are considerable differences in the demographic characteristics of the subgroups from the composition of the whole sample.For age group, nonmovers tend to be concentrated more in the older groups than the sample, whereas multiple movers-whether over long or short distances-tend to be younger.This accords well with what might be expected, and it is also worth noting the high proportion of long-distance migrants among those aged 16-25 in 2011 perhaps because of higher-education enrolment.There are interesting differentials by marital status with singles dominating the long-distance category of 50 km or more and the divorced and separated being overrepresented in the multiple short-distance move category of 10 km or less.There are major differences by housing tenure; there are a high proportion of owner occupiers in the nonmovers group, but renters are overrepresented in the short-distance category and private renters in the long-distance group.In terms of religion, the demographic subgroup that really stands out is the one with no religion-this group forms a greater proportion of the long-and short-distance categories than proportionally in the sample, which is partly a result of its younger age structure.With respect to educational attainment, there seem to be few differences between no movers and the entire sample, but long-distance movers are more highly educated than the other groups and they also tend to be healthier.There are some geographical differences by migration trajectory-those who are frequent short-distance migrants tend to be concentrated in socially deprived places and towns, whilst longdistance migrants tended to be disproportionately drawn from the least deprived quartile of SOAs (in terms of their place of residence in 2011) as defined by the 2010 Measure Deprivation Measure.Some differences by NS-SEC are also apparent; professionals and students are concentrated tend to move longer distances, those in routine occupations more over short distances.
There are clear differences in the demographic profile of nonmigrants, migrants of different types, and the full NILS analytical sample.
Looking first at the variables associated with all address changes, these are investigated via binary logistic regression in Figures 2a (for the individual-level results) and 2b (for the household and geographical results), which graph the results as odds ratios relative to reference categories.In this, sample members who leave NI are included within the outcomes variable along with those who migrate just within NI so it is all address changes that are the focus-although the proportions who emigrate are very small.Since time is the central focus of the paper, the analysis includes temporal dummies.The reference category is 2001-2003, with the other categories being 2004-2006, 2007-2009, 2010-2011, 2012, and 2013-2015  is needed in the interpretation of these results as they are compared against a reference category for the outcome variable-in this case no move-and also against the reference categories for each of the independent variables.This time, the results are presented as relative risk ratios.As before, there are 5% confidence limits on each bar in the chart.
Figure 3a demonstrates large differentials by age group with younger people being far more likely to move over all distance bands but especially the longest.Divorced and separated NILS members are also likely to move between 10 and 50 km and less than 10 km than those who were single, as were those with no or other religion.Those with Level 4+ qualifications (equivalent to degree level and above) were more likely to move more than 50 km.There are smaller differences by NS-SEC and health, which, though statistically significant, have less substantive importance.This latter analysis shows that there are differences in the demographic and social characteristics not only between address changers and nonaddress changers but also between movers over different distance bands.These indicate that the experiences and motives of residential migrants are very varied and that the reasons and motives for why people move differ between distance bands (Niedomysl, 2011).Multiple short-distance residential moves of less than 10 km are associated with some degree of vulnerability, such as poor health and social deprivation (Tseliou, Maguire, Donnelly, & O'Reilly, 2016), whereas longer distance residential changes are associated with youth, employment, and higher education.This is what is expected from the extant literature; NI is not a place apart.
However, there are two features worthy of more comment.First is the difference by religion in internal migration behaviour; relative to Catholics, Protestants are more mobile for the less than 10 km and 10-to 50-km bands, and those with other/no religion are in turn more mobile than Catholics and Protestants.It is probable that since the other/no religion groups are relatively small, they might self-select and be different from the rest of the population in their demographic and other (unobservable) characteristics.The remaining difference between Catholics and Protestants is intriguing, since in terms of characteristics observed in the 2011 Census, these groups are now very similar to each other in census categories such as housing tenure and educational attainment, (Rowland, 2019).This suggests that there are other factors at work-some perhaps social, others perhaps cultural-making Protestants more mobile than Catholics.Figure 3b tells us that the residents, in 2011, of the most deprived quartile of wards were less likely to be long-distance migrants and more likely to make moves of less than 10 km, as were urban dwellers and social renters.
As regard the temporal variables, there are statistically significant decreases across all distance bands until 2011 with a fall in migratory The third research theme was concerned with the individual and geographical correlates of moving over different distance bands.In this regard, the analysis indicates that the demographic factors associated with internal migration in NI are similar to those noted in the literature elsewhere (Green, 2018).In itself, this remark is unexceptional, but it does serve as an antidote to the assumption that NI is unique and utterly unlike elsewhere.In fact, both in its internal migration decline and in the drivers of its migration noted above, NI is just like many other places.The analysis also shows that there are no differences in address changing between SOAs in terms of social deprivation; residents of the most deprived quartile in 2011 were just as likely to have moved than those of more affluent SOAs.However, there are some differences in the distances that are moved by the residents of neighbourhoods at different rungs on the social deprivation ladder.
Compared with the reference category of the least deprived quartile, residents of the most deprived SOAs are less likely to make mediumdistance moves and longer-distance moves of 50 km or more (including emigration).This suggests that socially deprived people tend to make shorter-distance moves than others and that perhaps they do not avail themselves of the full range of regional opportunities.This indicates some degree of possible entrapment because of a greater localisation of those from more socially deprived areas unless some people choose to stay in place.The importance of migration is apparent in Shuttleworth et al (2013) where longer-distance moves are needed to effect substantial moves up or down the social deprivation and religion hierarchy of places.

| CONCLUSION
The paper builds on the contributions of McCollum et al. (2020) and Campbell (2018) by using administrative data alongside census data, thereby permitting geographies to be used independent of official statistical output units, after the 2011 Census.Furthermore, it allows migration to be described and modelled as a series of events-the multiple moves that any NILS member might makerather than as a single transition between time T and time T + 1.In this, and its updateability on a six-monthly basis, the NILS data offer a fresh perspective on migration.This adds to the cross-UK evidence base, and it also offers an analysis that is similar to that remains important as a correlate of migration and, as indicated earlier, this raises questions about why and how this might be.Some of these questions could be addressed with more years of migration event data, but others will require primary data collection to provide answers and yet others from more in-depth analysis of existing NILS data to deal with issues such as the implications of individual religion and neighbourhood characteristics for migratory behaviour.
, whereas the relative stability in residential segregation in 1991-2001 and small declines since 2001 suggest less intense population movement.The analysis builds on this foundation.Firstly, it seeks to evaluate whether the decline in internal migration noted in NI since the 1980s continued after 2011.Secondly, it assesses whether major economic and political shocks to NI since 2001 such as the Great Recession and the Flags Protest (this, starting in December 2012, centred over the flying of the Union Flag over Belfast City Hall, led to street protests, and a worsening of community relations) had any impact on annual internal migration levels.Thirdly, it considers the individual and geographical determinants of address changes over different distance bands and whether migrants who leave NI significantly alter our understandings.Finally, it adds to the comparative migration literature through an analysis similar to that in Scotland by McCollum et al. (2020) thereby also helping to demonstrate the effectiveness of the UK census longitudinal studies for migration research.
invariant.Taking the 2001 Census as the source for our explanatory variables was considered, as was measuring individual transitions between 2001 and 2011, for example in education.However, with the analytical period running forward to October 2015, the 2001 Census would be very dated as would any 2001-2011 transition, so the pragmatic choice was made to take the independent variables from the 2011 Census.These are therefore taken as fixed in 2011 for each person, so the model results for variables that might be expected to change over time in unpredictable ways, especially for migrants, like housing tenure, relationship status, and of course type of residential area (urban/rural, area deprivation level), must be treated with caution.Finally, time dummy variables are used to measure the allimportant aspect of temporal variation.For comparability with McCollum et al. (2020), the same time dummies of 2001-2003 (reference category), Figures 3a and 3bshow the results from a multinomial logistic analysis of moves by distance band (with emigration included in the >50-km band).Results excluding those who leave NI are presented in Figures S4a and S4b-they are largely similar with the one significant Figure S1-are excluded since in this case moves of more than 50 km remain depressed suggesting a continued decline in internal migration within NI but not of migrants leaving NI.
The fourth and final research theme concerned the comparison between NI and Scotland.There are many similarities in the individual factors associated with migration, but there are differences in the temporal trends observed with NI 's continuous decline (especially when leavers are excluded).It is worth noting, however, that the relationship between neighbourhood social deprivation noted in NI is different from that observed byMcCollum et al. (2020) in Scotland and that there is no clear gradient across the different distance bands along the quartile ranking and that the results are sensitive to whether those who leave NI are excluded or included.The descriptive statistics suggest, however, that frequent short-distance moves of less than 10 km are associated with personal disadvantage (see Table1, final column) in health and housing tenure and being in an urban SOA in 2011 and also in the most deprived quartile.This implies that multiple short-distance address changes can be a mark of social deprivation and disadvanbut the frequency of address changing by individuals and why this might be needs further probing.
undertaken byMcCollum et al. (2020) using the Scottish Longitudinal Study for Scotland.Its findings, and the wider background back to 1981, show that internal migration in NI has fallen over the long term as it has in other countries(Champion et al 2018).In this NI appears not to be immune from the broader social trends seen elsewhere.This is also true when the correlates of address changing and of moving different distances are modelled.The Great Recession, and indeed the various political events noted after 2001, does not seem to have had any impact on the address changing or distances moved; the longer-term trend for decreases in between SOA moves that started in 1981 seem mostly to be also observable for different distances of move and to continue after 2011.Either our data or analyses are not sensitive to detect their effects, or longer term forces have proved more important, or their impacts have been manifested in other ways, possibly in the destinations to which migrants move/do not move when they change address.Despite this insensitivity to political events, religion . These categories were chosen to be comparable with those used in Scotland by McCollum et al. (2020) with the exception of the separate dummy for 2012.As mentioned earlier, the purpose Profile of analytical sample by selected migratory trajectories 2001-2015 a Source: NILS Catholics and Protestants, and after controlling for population composition and location, that Protestants are significantly more likely to change address than Catholics.The causes of this require further investigation as it implies that there are unobserved differences between religious denominations not accounted for by the models.
the legacy of conflict and communal division does make NI different from most other comparable societies, notably in the importance of religion as a driver of migration and in the political interpretations of changing population geographies.In this context, it is worthwhile drawing attention to the result that other religions and those with no religion are more migratory than