Canadian Rural Partnership - Rural youth: stayers, leavers and return migrants

Rural Youth: Stayers, Leavers
and Return Migrants

Acrobat Portable Document Format (.PDF)

Report submitted to:
the Rural Secretariat of Agriculture and Agri-Food Canada
and to the Atlantic Canada Opportunities Agency

by: Richard Dupuy¹
Francine Mayer¹
and René Morissette¹

1 : Business and Labour Market Analysis Division
Statistics Canada

Richard Dupuy : (613) 951-3611
Internet: dupuy@statcan.ca
Francine Mayer : (613) 951-4536
Internet: mayefra@statcan.ca
René Morissette : (613) 951-3608
Internet: moriren@statcan.ca
FAX: (613) 951-5403

March 31, 2000

According to our analysis, in virtually all provinces young people 15 to 19 years of age are leaving rural areas in greater proportions than urban areas -- in part to pursue post-secondary education. While there are more complex migration patterns affecting the 20-29 age group, the net result of all migration is that the Atlantic provinces - as well as Manitoba and Saskatchewan - are net losers of their rural population aged 15-29. The problem is particularly acute in Newfoundland. In the Atlantic provinces, rural areas which fare worse than the national average - in terms of net gains of youth population - do so not because they have a higher than average percentage of leavers but rather because they are unable to attract a sufficiently high proportion of individuals into their communities.

Of all individuals who move out of their rural community, at most 25% return to this community ten years later. The implication of this result is clear : one cannot count on return migration as a means of preserving the population size of a given cohort. Rather, rural areas must rely on inflows from other (urban) areas to achieve this goal. Some rural communities achieve this; that is, they register positive net in-migration of persons aged 25-29 or older, even though they incur a net loss of younger people.

Individuals who move out of rural areas generally experience higher earnings growth than their counterparts who stay. However, it remains an open question in which direction the causality works: is the higher earnings growth the result of the migration process itself or does it reflect the possibility that people with higher earnings growth potential are more likely to become movers?

There has been for some time substantial concern regarding the loss of young people in rural communities. There is a sense that most rural communities offer few opportunities for their younger people, requiring them to leave for urban communities, most likely not to return. While there is a considerable body of research on interprovincial migration (Courchene, 1974; Grant and Vanderkamp, 1976; Finnie, 1998a, 1998b), very little is currently known about migration patterns between rural and urban areas in Canada. To our knowledge, no Canadian study has answered yet questions such as : 1) what percentage of individuals move out of rural and urban areas ? 2) what percentage move into rural and urban areas ? 3) where do movers go to ? 4) where do entrants come from ? 5) what are the characteristics of the movers and stayers ? 6) is return migration as frequent in rural areas as it is in urban areas ? 7) do movers experience higher earnings growth than stayers ?

The goal of this report is to provide answers to these questions. To do so, we assemble data from various sources : 1) the Censuses of 1991 and 1996, 2) administrative data based on T1 tax records which cover the 1986-1997 period and, 3) the Survey of Labour and Income Dynamics of 1993-1997. We apply the rural/urban distinction, not only at the province level, but also within each economic region of a given province. For instance, because Ontario has 11 economic regions, this allows us to construct 11 rural areas and 11 urban areas for Ontario. Similarly, we use 16 economic regions for Quebec and 15 for the Atlantic provinces.

The aforementioned questions are important as they help policy makers establish the basic facts about the migration patterns of the rural youth population. Doing so is a necessary step to understand what policy intervention, if any, is needed to help some rural areas stop the decline of their youth population and favour economic growth.

The plan of the study is the following. Section 2 contains a brief overview of the literature on rural/urban migration. In section 3, we define the concepts used in the report. In section 4, we sketch a profile of youth in rural and urban areas. Next, we document migration flows out of/into rural and urban areas (Section 5). Among other things, we check whether the propensity to leave a geographical unit is higher in rural areas than in urban areas. Some of those who migrate eventually return to their area of origin. We investigate the magnitude of this phenomenon, called return migration, in Section 6. In section 7, we compare the characteristics of movers to those of stayers and investigate whether movers enjoy higher earnings growth than stayers. Concluding remarks follow in section 8.

Until recently, research on urban/rural migration has been hampered by a lack of appropriate data. In Quebec, a team of researchers led by Gauthier (1997) has recently published a collection of papers which examine issues related to rural/urban migration. While some of these papers (Côté, 1997; Roy, 1997) examine data for some specific rural areas, there is no analysis provided for all administrative areas in Quebec. This gap should be filled in the near future since this team of researchers has recently conducted a Quebec-wide survey on rural migration, whose data has yet to be analyzed.

One of the arguments often made in the aforementioned papers is that migration out of rural areas is driven only partly by economic factors. For instance, Roy (1997, p. 95) finds that if young individuals could hold the job they desire in their community, four out of ten would still be willing to move out to an urban center. This is evidence that other factors, such as one's desire to experiment with different life experiences or to fulfill one's aspirations, play a role in explaining migration out of rural community. Obviously, the challenge of any researcher is to disentangle the relative importance of economic, sociological and psychological factors. We do not attempt such an exercise in this paper. Rather, our goal is to provide Canada-wide information about the basic facts regarding migration out of /into rural and urban communities.

The data used in this paper come from three separate data sets : 1) the Censuses of 1991 and 1996, 2) T1 tax records and, 3) the Survey of Labour and Income Dynamics (SLID) of 1993-1997. The data sets constructed from Census data and T1 tax records have extraordinary sample sizes. Those used for Census data cover 20% of the population and thus allow a very detailed examination of migration patterns in narrowly defined geographical units. Such a detailed analysis can also be performed using T1 tax records since the files used from this data source cover the whole population of tax filers. Only SLID has a more limited sample size : the analysis which can be conducted with it is limited at the national level.

Census data allow analysts to examine migration flows because it contains the following question : "Where did [you] live 5 years ago?". Census data has two advantages over T1 tax records. First, it contains information on labour market conditions (employment rate, full year full-time employment rate, wages of full year full-time workers) in each geographical unit. Second, it contains information on workers' education level, occupation and industry of employment at the time of data collection (e.g. May 1996 for Census 1996). None of this information is available in the tax data. However, contrary to tax data and SLID, it cannot be used to examine whether earnings of movers grow faster than those of stayers since it does not include information on workers' earnings five years ago. Furthermore, Census data cannot be used to document the characteristics of workers before the move because it contains no information on workers' attributes 5 years ago.

SLID is the only Canadian data set which allows an examination of workers' education level, occupation and industry of employment both before and after migration, i.e. both in the area of origin and in the destination area. Another advantage of SLID is that - like T1 tax records - it allows an analysis of the earnings growth of movers and stayers. Contrary to T1 tax records, it contains several covariates (education level, industry, occupation, union status, firm size) which make possible a multivariate analysis of the earnings growth of movers and stayers. However, as pointed out above, its sample size precludes an analysis of migration patterns at the economic region level.

Because it covers a long time interval, data from T1 tax records permit an analysis of return migration, which is not possible with Census data and is possible only to a limited extent with SLID data.

In this paper, we define urban areas as geographical units belonging either to a census metropolitan area (CMA) or a census agglomeration (CA). Rural areas and small towns are defined residually as geographical units which are neither in a CMA nor a CA. A CMA consists of an urbanised core having a population of at least 100,000 people while a CA consists of an urbanised core which contain between 10,000 and 100,000 individuals. We use the term "rural areas" to refer to rural and small town communities (see Appendix A for details).

While part of the analysis conducted in this paper is done at the Canada level or the province level, a substantial portion is also performed at the economic region level. The economic regions used in the study are those of the Labour Force Survey. In Canada, there are 74 economic regions, 62 of which have both a rural and an urban component. The size of the population aged 15 and over varies markedly across economic regions. It amounts to 2,129 in the urban component of Gaspésie-Iles-de-la-Madeleine and reaches a maximum of 3,510,138 in urban Toronto (Appendix Table 2). We use the term geographical unit to refer to a given economic region/rural-urban status. For instance, in this paper, Cape Breton rural is one geographical unit.

We use three age categories to define young individuals : 1) those aged 15-19, 2) those aged 20-24 and, 3) those aged 25-29. We use the term youth to refer to individuals aged 15 to 29.

Young individuals live in regions which differ widely in their degree of ruralness. For instance, in Gaspésie-Iles-de-la-Madeleine, 97% of youth live in rural areas (Appendix Table 1). Conversely, only 2% of youth living in the Halifax economic region stay in rural areas. A less extreme scenario is observed in Vancouver Island and Coast where 15% of youth live in rural areas.

The share of youth in the population aged 15 and over is generally smaller in rural areas than in urban areas : this is true for roughly three quarters of economic regions which have both a rural and urban component (Appendix Table 2).⁽¹⁾ In Canada, youth represent 24.6% of the population in rural areas and 26.3% of the population in urban areas.

Young individuals living in rural areas are less educated than those living in urban areas. For instance, of all individuals aged 25-29 living in rural areas, 31% had a postsecondary education in 1996, compared to 46% for those living in urban areas (Appendix Table 3). This pattern generally holds when we consider the percentage of individuals who have a university degree. It is widespread, i.e. it is observed in most economic regions both for individuals aged 20-24 and those aged 25-29.⁽²⁾ Of course, there are some exceptions. In Cape Breton, individuals aged 20-24 living in rural areas have a post-secondary education more often (32%) than those living in urban areas (27%) (Appendix Table 4). The same scenario is observed in Annapolis Valley, Gaspésie-Iles-de-la-Madeleine, Toronto, Stratford - Bruce Peninsula, North Central Manitoba, Saskatoon - Biggar, Prince Albert, Red Deer - Rocky Mountain House, Wood Buffalo - Camrose and Northeast British Columbia.

One reason for the lower level of education of rural youth is that the type of jobs available in rural areas may require lower skills than those required by jobs in urban areas. Appendix Table 5 provides some evidence which is consistent with this view. Of all individuals aged 15-29 who were not full-time students and who were employed in May 1996 in rural areas, 22% were employed in professional and managerial occupations. The corresponding percentage for urban youth is 30%. Conversely, almost half (48%) of employed rural youth are blue-collar workers, a much higher percentage than that observed for urban youth (36%). This pattern holds for all provinces. Since it is reasonable to assume that the skill requirements of professional and managerial occupations are higher than those of blue-collar workers, these differences in the distribution of employment by occupation are likely to explain part of the differences in educational attainment documented above.

At least three other factors may explain the differences in educational attainment between individuals living in rural areas and those living in urban areas. First, pecuniary and non-pecuniary costs of pursuing post-secondary education - which is generally not available in rural areas - are likely to be higher in rural areas than in urban areas. If the benefits of going to a post-secondary institution are not sufficiently higher in rural areas, the proportion of individuals who will choose to go to a post-secondary institution will be lower in rural areas. Second, the education level of parents may play a role. As long as the probability of a young individual pursuing post-secondary education is positively correlated with his/her parents' education level, and as long as parents in rural areas are generally less educated than their counterparts in urban areas, young individuals in rural areas will be less likely to pursue post-secondary education. A third reason is that, because post-secondary institutions are generally absent from rural areas, individuals originating from rural areas must have left their community and returned to it to be counted as a rural resident with post-secondary education. The fact that only a fraction of leavers return to their rural community - as we shall see below - implies that the proportion of youth with post-secondary education will be lower in rural areas.

Rural youth and urban youth are employed in different industries. Unsurprisingly, young rural workers are more likely than their urban counterparts to be employed in agriculture and in forestry and mining, where natural resources are predominant (Appendix Table 6). In contrast, the former are less likely than the latter to be employed in services, which are concentrated in urban areas. These patterns are found in all provinces.

One reason which is often cited to explain why young individuals leave rural areas is the fact that labour market conditions are less favorable in rural areas than in urban areas. Appendix Table 7 compares labour market conditions in rural and urban areas. It shows data on : 1) the percentage of individuals who are employed in May 1996 (employment rate) and, 2) on the percentage of workers who were employed full year full-time during the year 1995 (full year full-time employment rate). The numbers are presented for individuals aged 15-19, 20-24, 25-29, 15-29 and 25-54. For each labour market indicator, two samples are considered : 1) individuals who are not full-time students and, 2) all individuals.

Rural/urban differences in labour market conditions are not the same for all age groups. Among non-students aged 15-19, the employment rate is, at the Canada level, the same in rural areas and urban areas (51%). The full year full-time employment rate is also the same (12%). In contrast, for individuals aged 20-24 and those aged 25-29, both the employment rate and the full year full-time employment rate are lower in rural areas than they are in urban areas. Note that the employment rate, considered in isolation, can give a misleading picture of labour market conditions. For instance, among individuals aged 25-29 living in Prince Edward Island, it is very similar in rural and urban areas (between 77% and 79%). However, looking at the full year full-time employment rate, labour market conditions look much worse in rural areas : the percentage of workers employed full year full-time equals 31% in rural areas, compared to 50% in urban areas.⁽³⁾ Hence, the main message conveyed by Appendix Table 7 is that labour market conditions appear to be tougher in rural areas for individuals aged 20-29 but not for teenagers (i.e. individuals aged 15-19). Detailed results by economic region are presented in Appendix Table 8.

Appendix Table 9 presents unemployment rates in rural and urban areas for each province. Among non-students aged 15-19, the unemployment rate is, at the Canada level, virtually the same in rural areas and urban areas (22-23%). However, at the national level, the unemployment rate is higher in rural areas than in urban areas for individuals aged 20-29.

There are drastic differences in rural unemployment rates by province. Among individuals aged 15-29 who are not students, the unemployment rate in rural areas reaches a maximum of 40% in Newfoundland and a minimum of 11% in Alberta. Undoubtedly, these differences in unemployment will have an impact on migration flows in these provinces. We expect migration flows to be much less favourable in Newfoundland than in Alberta. As we shall see below, these expectations are confirmed by the data.

A fourth labour market indicator which can be used to compare labour market conditions in rural and urban areas is the annual wages of full year full-time workers. Appendix Table 10 shows that women aged 20-24 who worked full year full-time and who were living in rural areas earned on average $1,383 less than their urban counterparts in 1995, i.e. the year before 1996 Census data was collected. Since individuals living in rural areas are less educated than those in urban areas and since highly educated people generally earn more than their low-educated counterparts, part of the earnings gap may be due to differences in schooling. This is indeed the case, after controlling for individual differences in educational attainment (and in the province of residence), the earnings gap is reduced to $558. A similar pattern is observed among women aged 25-29 and those aged 30-44. Thus, the evidence brought in Appendix Table 10 suggests that women living in rural areas do earn less than those living in urban areas.

Unexpectedly, the story is different for young men. Men aged 25-29 living in rural areas appear to earn $948 less than their urban counterparts but once controls for education and province of residence are included, they end up earning $690 more. Similarly, after performing the multivariate analysis, men aged 20-24 living in rural areas earn $1,447 more than those living in urban communities.⁽⁴⁾ This pattern is not observed among men aged 30-44, however. For these men, annual earnings in rural areas are at least $5,000 smaller than those in urban areas, in the raw data. The earnings gap is reduced to about $1,900 in the multivariate analysis. Part of this gap could be related to the possibility that men aged 30-44 have lower access to high-paying industries in rural areas and/or that within industries, they receive lower wages in rural areas. In any event, the evidence for men aged 30-44 is consistent with the notion that earnings are lower in rural areas than they are in urban areas. At the very least, our analysis indicates that rural/urban differences in earnings vary significantly, not only by age group, but also by gender.

It should be emphasized that our results regarding rural/urban differences in earnings must be interpreted with caution since potentially relevant factors, such as the number of years of experience, have not been controlled for.

The Atlantic provinces have 15 economic regions, some of which are entirely rural (South Coast - Burin Peninsula in Newfoundland and Southern Nova Scotia) and some of which are almost entirely urban (Halifax). The share of individuals aged 15-29 in the population aged 15 and over varies between 21% and 31%, depending on the geographical unit considered. As is observed for the rest of Canada, individuals living in rural areas are less educated than their urban counterparts. Of all young persons (i.e. persons aged 15-29) living in rural areas in the Atlantic provinces, only 17% have some postsecondary education, compared to 26% for those living in urban areas. This pattern is found in all economic regions but is less pronounced in Cape Breton, North Shore (Nova Scotia), Annapolis Valley and Campbellton-Miramichi.

As expected, key labour market indicators confirm that labour market conditions are less favorable for youth living in the Atlantic provinces than for those living in Canada. The Atlantic-Canada differences are more substantial in rural areas than in urban areas. For instance, among individuals aged 15-29 who are not full-time students and who live in urban areas, 70% are employed in the Atlantic provinces, compared to 74% in Canada (Appendix Table 7). In rural areas, the corresponding percentages are 56% and 66% respectively. A similar conclusion applies when we examine the full year full-time employment rate, i.e. the percentage of workers who are employed full year full-time.

Of all the Atlantic provinces, Newfoundland has the lowest full year full-time employment rate of rural youth (non-students), i.e. 22% (Appendix Table 7). Nova Scotia fares the best, with a full year full-time employment rate of 33%. Within the Atlantic provinces, labour market conditions of rural youth vary markedly across economic regions (Appendix Table 8). The full year full-time employment rate of rural youth (non-students) is about 25% or less in all economic regions of Newfoundland (Avalon Peninsula, South Coast - Burin Peninsula, West Coast - Northern Peninsula and Notre Dame - Central Bonavista Bay), in Prince Edward Island, Cape Breton and in Campbellton-Miramichi. In contrast, it equals or exceeds the national average in Annapolis Valley and Fredericton-Oromocto. The national average for the full year full-time employment of rural youth who are non-students is 39%.

In this section, we examine how many people stay in a given geographical unit during a given time period (stayers), how many leave (outflows) and how many move into a given geographical unit (inflows). We focus on 5-year migration flows but we also compare the magnitude of 1-year flows and 10-year flows. Recall that a geographical unit is an area defined both in terms of the economic region it belongs to and in terms of its rural/urban status. Thus, Cape Breton rural and Cape Breton urban are two distinct geographical units.

We sometimes present data on both the 1986-1991 period and the 1991-1996 period but focus on the latter. The data used in Section 5.1 come from T1 tax records and the Census. The population analyzed consists of individuals who were present in the T1 tax records (in Canada for Census data) both in year t and year t+5.⁽⁵⁾

In Table 1, we show migration flows out of rural and urban areas. The numbers are presented for individuals of all ages. A person is counted as a leaver (or an outflow) if he/she was in a given geographical unit in year t but not in year t+5. For instance, an individual who lived in Prince Albert rural in year t and outside Prince Albert rural in year t+5 is counted as a leaver, whether the destination area is Prince Albert urban or any other geographical unit. However, if that person moved to a different location within Prince Albert rural, he/she will not be counted as a leaver. The numbers presented in the table are outflow rates, i.e. equal the number of leavers divided by the population in year t.

Table 1 conveys two messages. First, as is well known, mobility falls with age, i.e. the percentage of leavers declines as individuals get older, both in rural and urban areas. The relationship is not monotonic for individuals living in urban areas, however. One explanation for the fact that mobility falls with age is that the incentives for moving are greater, the younger you are. This is so simply because the younger you are, the lower the costs of moving (pecuniary as well as non-pecuniary) are likely to be and the higher the expected benefits are likely to be (since you can reap these benefits over a longer period of time). Second, and more important, individuals tend to leave rural areas more often than urban areas. The difference is particularly pronounced for teenagers (i.e. those aged 15-19). Between 1991 and 1996, 28-32% of teenagers left rural areas while only 15-18% did so in urban areas. In contrast, the corresponding percentages are 18-22% and 17-19% for individuals aged 25-29. The fact that outflow rates of youth aged 25-29 in urban areas equal at least 87% of outflow rates of their counterparts in rural areas is important : it reminds us that leaving one's area is a phenomenon which is not limited to rural communities. Among individuals aged 30 to 64, the propensity to leave rural areas is 1-3 percentage points higher than the propensity to leave urban areas. This shows that rural/urban differences in outflow rates fall with age. Put simply, the percentage of leavers in rural areas generally gets closer to the percentage of leavers in urban areas as individuals get older.

Table 2 presents outflow rates for individuals aged 15-19, 20-24 and those aged 25-29 for each province. The results show that in all provinces except New Brunswick, teenagers tend to leave more often rural areas than urban areas. In New Brunswick, outflow rates of teenagers living in urban areas are essentially the same as for those living in rural areas.

The patterns are different for individuals aged 20-24. In the relatively rich provinces of Ontario, Alberta and British Columbia, those in their early 20s move out of rural areas much more often than they move out of urban areas. Quite surprisingly, the opposite pattern is found in all the Atlantic provinces: individuals aged 20-24 living in rural and small town areas leave less frequently than their urban counterparts. The rural/urban differences observed in the Atlantic provinces for this age group are small in the tax data and more pronounced in Census data. Depending on whether we use tax data or Census data, we reach different qualitative conclusions for Quebec, Manitoba and Saskatchewan.

Among individuals aged 25-29, outflow rates are still lower in rural areas than in urban areas for the Atlantic provinces. The same pattern is also observed in Quebec and Saskatchewan. The percentage of leavers remains higher in rural areas than in urban areas for Ontario, Alberta and British Columbia. Outflow rates in rural areas are fairly close to those of urban areas for Manitoba.

The findings presented above highlight the need to disaggregate age categories when analyzing the mobility patterns of youth. They reveal that individuals aged 15-29 are not a homogeneous group and suggest that the reasons for leaving may differ depending on the particular age group one analyzes.

In Ontario, Alberta and British Columbia, outflow rates in rural areas exceed those in urban areas for all age groups. One explanation for this pattern is that urban centers in rich provinces may offer several opportunities for high-quality jobs and career progress, increasing the attractiveness they exert on rural youth in these provinces.

Interestingly, young persons in Saskatchewan, Alberta and British Columbia leave their rural communities in much greater proportions than their counterparts in the Atlantic provinces. At first, this may seem puzzling. Because labour market conditions in the Atlantic provinces are generally worse than those in Western provinces, one would expect proportionately more people to move out of the former than moving out of the latter. However, the distance required to move to a promising labour market may be much greater in the Atlantic provinces than in the Western provinces. This could explain why the percentage of rural leavers is smaller in the Atlantic provinces. An alternative hypothesis is that the attachment to one's community could be stronger in the Atlantic provinces.

Whatever the reasons are, this finding suggests that to model properly outflow rates, one requires labour market indicators in the destination area as well as in the origin area, data on the distance associated with a move and data on the population size of the origin and destination areas (used to capture the variety of employment opportunities) (Grant and Vanderkamp, 1976). When there are n areas to consider, the number of possible origins-destinations equals n times n-1. In our case, we have 136 (rural and urban) areas, which yields a total of 18,360 origins-destinations (136 times 135). We do not attempt this modelling exercise in this report.

Apart from the outflow rates, Tables 3 to 5 show the destination of leavers. The destinations are classified into four categories : 1) moving to a rural area inside the province of origin, 2) moving to a rural area outside the province of origin, 3) moving to an urban area inside the province of origin and, 4) moving to an urban area outside the province of origin. For all three age groups and all provinces except Newfoundland, the main destination of individuals who leave rural areas is an urban area inside the province of origin. In Newfoundland, the main destination is an urban area outside the province of origin.⁽⁶⁾ Thus, except in Newfoundland, young individuals who leave their rural community go mainly to larger cities inside their province of origin.

For individuals leaving urban areas, the main destination varies. It is an urban area outside the province of origin for youth living in the Atlantic provinces and in the Prairies. In contrast, it is an urban area inside the province for the three biggest provinces, i.e. Quebec, Ontario and British Columbia. As expected, moving to an urban area outside the province is an extremely rare event for individuals living in Quebec.

Within the Atlantic provinces, there is wide variation in outflow rates across economic regions. For all three age groups and for both Census and tax data, the percentage of youth who move out of their rural community is almost always twice as high in Fredericton-Oromocto as it is in Southern Nova Scotia (or Edmunston-Woodstock) (Table 6).

Among teenagers, the percentage of movers in rural areas exceeds that in urban areas in all economic regions except : 1) Note Dame - Central Bonavista Bay, 2) Campbellton - Miramichi, 3) Moncton - Richibucto and, 4) Edmunston - Woodstock. Among individuals in their early 20's, we find that: 1) outflow rates in rural areas exceed those in urban areas in 3 economic regions, 2) are smaller than those in urban areas in 5 economic regions, 3) the relationship is unclear (i.e. Census data and tax data reveal two opposite patterns) in 5 economic regions.⁽⁷⁾ Among individuals aged 25-29, outflow rates in rural areas are smaller than those in urban areas in 8 economic regions and the relationship between the two outflow rates is unclear in the remaining 5 economic regions which have both a rural and urban component.

For New Brunswick and Nova Scotia, young individuals of all ages who leave their rural community first choose to go to an urban area inside the province, presumably to pursue post-secondary education (Tables 7 to 9). Their second choice is an urban area outside the Atlantic provinces. In Newfoundland, the order of the two choices is reversed, i.e. young individuals first move to urban areas outside the Atlantic provinces. The main destination of urban leavers is different. Young people who leave urban communities generally go to an urban area outside the Atlantic provinces. Sometimes, their second choice is an urban area outside the province and inside Atlantic, an urban area inside the province or a rural area inside the province. The second destination of individuals aged 25-29 is generally a rural area within the province of origin. Thus, the destination of youth who move out of their community varies depending on whether the community of origin is rural or urban.

A person is counted as a new resident (or an inflow) if he/she was present in a given geographical unit in year t+5 but not in year t. Inflow rates are calculated by dividing the inflows by the population of the geographical unit in year t.⁽⁸⁾

Table 10 shows 5-year migration inflows in rural and urban areas for the 1986-91 and the 1991-96 periods. The most striking finding that emerges from this table is that, at least at the Canada level, inflow rates for rural areas exceed those for urban areas for all individuals aged 20 and over. Among teenagers, inflow rates in rural areas are smaller than those in urban areas. Since we saw in the previous section that outflow rates were higher in rural areas than in urban areas for this group, we can already infer that net flows of teenagers (i.e. inflows minus outflows) will generally be smaller in rural areas than in urban areas. Likewise, the fact that inflow rates for rural areas are greater than those for urban areas is quite interesting : it raises the possibility that, despite high outflow rates, rural areas may experience net gains of individuals for some age groups. We will examine this issue in Section 5.1.3.

Looking at inflow rates by province, we learn that, among individuals aged 20 and over, inflow rates for rural areas do not always exceed those for urban areas (Table 11). Among individuals aged 20-29, inflow rates in rural areas are actually smaller than those in urban areas for Newfoundland, Prince Edward Island and New Brunswick. In contrast, inflow rates for rural areas are higher than those in urban areas in Ontario, Manitoba, Alberta and British Columbia. This pattern is observed despite the fact that labour market conditions (as measured by the employment rate and the full year full-time employment rate) in these latter provinces are generally worse in rural areas than they are in urban areas. This suggests that non-economic factors play a role in shaping inflow rates.

Among individuals aged 15-19, inflow rates in rural areas are smaller than those in urban areas in all provinces except Ontario, Alberta and British Columbia.

Once again, some interesting interprovincial differences appear. Inflow rates in rural areas of Alberta and British Columbia are twice as high as those in the Atlantic provinces. Newfoundland is the province which received the lowest inflow of new residents in rural areas during the 1991-96 period.

Where do new residents of rural areas come from? Generally, they come from an urban area inside the province in which they live in year t+5 (Tables 12 to 14). Some exceptions are worth noting. In Newfoundland and New Brunswick, the main area of origin is an urban area outside the province for individuals aged 25-29 in year t. The same is also true for teenagers living in Prince Edward Island.

Where do new residents of urban areas come from ? The answer is more complex here. New residents of urban areas sometimes come mainly from rural areas inside the province, urban areas inside the province or urban areas outside the province. Among teenagers, the main area of origin is generally : 1) a rural area inside the province for the Atlantic provinces, Manitoba and Saskatchewan, 2) an urban area inside the province for Quebec and Ontario and, 3) an urban area outside the province for Alberta and British Columbia. These last two patterns also hold for individuals aged 20-29. While teenagers who become new residents in the Atlantic provinces, Manitoba and Saskatchewan come mainly from a rural area inside the province, the same is not true for their counterparts in their late 20's. In this case, the main area of origin becomes generally an urban area outside the province.⁽⁹⁾

For all age groups, individuals coming from rural areas outside the province almost never represent a sizeable portion of new residents. The proportion of new residents who come from rural areas outside the province equals at most 26%.

To sum up, new residents of rural areas differ from those of urban areas in their area of origin. The former generally come from an urban area inside the province of destination while the latter come from a wider variety of areas.

As was the case for outflow rates, there is a great diversity of inflow rates across economic regions. For instance, among individuals aged 25-29 in 1991, inflow rates in rural areas are as low as 5-7% in South Coast - Burin Peninsula and as high as 35-36% in Annapolis Valley (Table 15).

Among teenagers, inflow rates of rural areas are smaller than those of urban areas in most economic regions. Among individuals aged 20-24 and those aged 25-29, both Census and tax data show the opposite pattern in 4 and 5 economic regions, respectively (out of 13 economic regions which have both a rural and an urban component).

For all three age groups, new residents of rural areas come mainly from urban areas inside the province of residence (in year t+5) (Tables 16 to 18). This finding is important since it shows that rural areas in the Atlantic provinces draw their new young residents mainly from the relatively narrow pool of potential residents represented by urban areas within their own province rather than from the much larger pool of potential residents represented by urban areas outside the Atlantic provinces. This contrasts with urban areas in Nova Scotia. These draw their new residents aged 25-34 (in 1996) mainly from urban areas outside the Atlantic provinces.

In Table 19, we show net flows at the Canada level for individuals aged 15 and over. Net flow rates are calculated by substracting outflows from inflows and then dividing the resulting number by the population of the relevant geographical unit in year t. Net flow rates must be interpreted as the percentage change in population of a given age cohort which would occur in the absence of deaths and international migration. Even at this broad level of aggregation, some key results appear. First, both Census data and tax data show that, during the 1986-91 and the 1991-96 periods, rural areas seem to have lost 12-16% of their population aged 15-19 to urban areas. This comes as no surprise since our results in Sections 5.1.1 and 5.1.2 have shown that, for this particular age group, outflow rates in rural areas exceed those in urban areas and that inflow rates in rural areas are smaller than those in urban areas. Second, whether rural areas experience a net loss of individuals in their early 20's is unclear : Census data indicate a net gain of 2-6% while tax data indicate a net loss of -2% to -4%. Whether or not their net flow rates are smaller than those of urban areas is also unclear. Third, rural areas experience a net gain of individuals aged 25 to 64 while urban areas experience a net loss of these individuals. In other terms, rural areas lose their teenagers to urban areas, may lose or gain individuals in their early 20's and gain individuals aged 25-64 from urban areas.

What is the net effect of these three trends on the population aged 15 and over ? For the 1991-96 period, the third effect dominates the first two and, as a result, rural areas have enjoyed net gains for their population aged 15 and over during the period considered. The net gains of rural areas amount to roughly 2% while the net losses of urban areas amount to 0.6% . In an accounting sense, these net gains for rural areas explain partly why rural and small town population is growing in the 1990s (Mendelson and Bollman, 1998). Alternatively, the net losses experienced by urban areas suggest that the growth they have experienced (Mendelson and Bollman, 1998) is due to international immigration.

What is the net effect of these trends on the population aged 15-29 ? For the 1991-96 period, the net effect is almost zero : Census data indicates a very small net gain of about 0.2% for rural areas while tax data shows a corresponding percentage of -0.9%.⁽¹⁰⁾

At the Canada level, rural areas lose their teenagers, may lose or gain individuals in their early 20's and gain individuals aged 25-64. This statement applies to all provinces except Newfoundland, New Brunswick and Saskatchewan (Table 20). Rural areas in Newfoundland are net losers of all age groups except those aged 55-64. It is unclear whether rural areas in New Brunswick are net gainers of individuals aged 25-29. Rural areas in Saskatchewan display small negative net migration rates for individuals aged 30-44 and those aged 55-64. ⁽¹¹⁾ The fact that rural areas in most provinces are net gainers of individuals aged 25-64 is quite interesting : it implies that in all provinces except Alberta and British Columbia, the net migration rates of individuals aged 15 and over is higher than those of individuals aged 15-29. For instance, Census data and tax data show that even though rural areas in Nova Scotia are net losers of their youth population, they maintain the size of their population aged 15 and over, i.e. they have net migration flows which average 0%.

At the Canada level, a net gain of individuals aged 25-64 in rural areas implies a corresponding loss of individuals aged 25-64 in urban areas. At the provincial level, this is no longer true. In the previous paragraph, we have identified provinces whose rural areas are net gainers of individuals aged 25-64, whether or not their urban areas are net losers of individuals aged 25-64. What provinces both experience a net gain of individuals aged 25-64 in their rural areas and a net loss of these individuals in their urban areas ? The answer is Quebec, Ontario, Manitoba and Alberta. Prince Edward Island enjoys a net gain of individuals aged 30 and over in urban areas. Nova Scotia has a small net gain of individuals aged 55-64 and British Columbia experiences a net gain of individuals aged 25-64 in urban areas. In Section 5.1.4, we examine in more detail how the net migration flows of various provinces affect the growth rate of their youth population and of their population aged 15 and over.

Over the 1991-96 period, rural areas lose part of their teenagers' population in all provinces (Table 21). In contrast, net flow rates for individuals aged 20-24 are, depending on the data set used, either positive or negative in most provinces, reflecting the uncertainty documented for this age group at the Canada level. Among individuals aged 25-29, net flow rates in rural areas are greater than in urban areas for the six following provinces : Prince Edward Island, Nova Scotia, Ontario, Manitoba, Alberta and British Columbia. Rural areas are net gainers of their population aged 25-29 in all provinces except Newfoundland, New Brunswick and Saskatchewan. The magnitude of these net flows is particularly high in Ontario and British Columbia, where net flow rates in rural areas vary between 8% and 21%.

The fact that, in several provinces, rural areas are net gainers of individuals in their late 20's is unexpected. It puts into question the myth that rural areas lose young individuals of all age groups.

In Table 22, we present net flows by economic region for individuals aged 15-19, 20-24 and 25-29. As expected, the vast majority of rural areas in economic regions experience net losses of individuals aged 15-19. Even within provinces, rural areas are a very heterogeneous group. For instance, in Quebec, rural Bas-Saint-Laurent has experienced a net loss of individuals aged 20-24 of at least 3% between 1991 and 1996. This is in marked contrast with the net gains experienced by rural Lanaudière, which amount to at least 15% for this age group for the same period. Similarly, between 1991 and 1996, in Ontario, the rural component of Kitchener-Waterloo-Barrie has experienced a net gain of individuals aged 25-29 of at least 16%, while the rural component of Northwest has experienced a net loss of 2%.

Out of the 74 economic regions we have, 62 have both a rural and an urban component. Among individuals aged 20-24, it is unclear whether net flows in rural areas are smaller than those in urban areas for 24 economic regions. Net flows in rural areas are smaller than those in urban areas in only 10 regions and are greater in 28 regions.

Among individuals aged 25-29, most economic regions (38) have net flow rates in rural areas which exceed those in urban areas. The opposite pattern is found in 12 economic regions while uncertain patterns are found in 12 economic regions.

These findings support the notion that the two patterns identified above at the Canada level for teenagers and individuals aged 25-29 - i.e. that among teenagers, net flow rates in rural areas are smaller than those in urban areas while, among individuals aged 25-29, net flow rates in rural areas exceed those in urban areas - do not result from a few economic regions with a large population but rather are widespread, i.e. are observed in a large number of economic regions.

In Tables 23 to 25, we present outflows, inflows and net flows for all economic regions of the Atlantic provinces. Three points are worth noting. First, both Census data and tax data show that, during the 1991-96 period, all rural areas in Newfoundland experienced net losses of about 25% of their teenagers' population, which is twice as high as the net losses experienced at the Canada level. In contrast, both data sets show that all rural areas in New Brunswick had net losses which were either fairly close to or smaller than the national average. The bad performance of all rural areas in Newfoundland, as compared to rural Canada, is also found when we examine net losses for individuals aged 20-24 and 25-29. Second, net losses from Census data and those from tax data often differ substantially for individuals aged 20-24. For instance, Census data indicate that net flows for rural Cape Breton are virtually 0% while tax data show net losses of 17%. This simply reflects the uncertainty that we found previously at the Canada level when comparing net losses between the two data sets among individuals aged 20-24. Third, within Nova Scotia, net flows vary markedly across economic regions. Among individuals aged 25-29, rural North Shore exhibits net flows which equal virtually 0% while rural Annapolis Valley has net gains of 9-19%, depending on the data set used. Once again, this confirms the notion that in a given province, migration patterns may be quite heterogeneous across economic regions. The implication is obvious : not all economic regions in a given province may require, if any, the same type of policies.

Whenever rural communities experience smaller net gains or bigger net losses of individuals, compared to the national average, the difference may, in an accounting sense, result from two factors. Net gains may be smaller or net losses may be bigger either because inflows are below the national average (insufficient inflows) or because outflows are above the national average (excessive outflows). In columns 4 and 8 of Tables 23 to 25, we select the rural component of all economic regions whose net flows are smaller than the national average by at least 1 percentage point, and ask whether the main source of the discrepancy is insufficient inflows (labeled ii) or excessive outflows (eo). The answer is clear. For virtually all economic regions selected, the main source of the discrepancy is insufficient inflows. In other words, rural areas of the Atlantic provinces which fare worse than rural areas in Canada - in terms of net flows - do so not because they lose proportionally too many people but rather because they are unable to attract a sufficiently high proportion of individuals into their communities.

This point can be illustrated clearly by looking at rural areas in Newfoundland. Outflow rates show that all four rural areas in Newfoundland lose proportionally fewer individuals aged 25-29 than all rural areas in Canada. However, inflow rates indicate that they attract proportionally much fewer individuals than rural areas in Canada. As a result, they experience net losses of individuals aged 25-29 while rural areas in Canada enjoy net gains of these individuals.

One way to determine whether economic regions are doing well or not is to examine their net flows for both the youth population (i.e. individuals aged 15-29) and the population aged 15 and over. A region where both populations are declining certainly raises more concern than a region where both populations are increasing.

In Table 26, we examine net flows for individuals aged 15-29 and those aged 15 and over for each province. Both data sources indicate that over the 1991-96 period, all the Atlantic provinces, Manitoba and Saskatchewan are net losers of their rural population aged 15-29. Ontario, Alberta and British Columbia are net gainers. Newfoundland, New Brunswick and Saskatchewan are net losers of their rural population aged 15 and over while Quebec, Ontario, Alberta and British Columbia are net gainers. As a result, Newfoundland, New Brunswick and Saskatchewan lose both their rural youth and their rural population aged 15 and over while Ontario, Alberta and British Columbia are net gainers on both populations. Thus, on the basis of these results and abstracting from international immigration, rural areas in Newfoundland, New Brunswick and Saskatchewan appear to be having problems maintaining the size of their youth population and of their population aged 15 and over. The problem seems to be particularly acute in Newfoundland, where net flows of individuals aged 15-29 amount to -13%.⁽¹²⁾

Table 27 shows net flows for individuals aged 15-29 and those aged 15 and over for each economic region. Net flows vary widely across economic regions. Both Census data and tax data show that rural areas in the following economic regions experienced a net loss of at least 5% of their population aged 15-29 during the 1991-96 period :

In contrast, both data sets indicate that rural areas in the following economic regions enjoyed a net gain of at least 5% of their population aged 15-29 :

As expected, most economic regions experienced net losses of rural youth in Newfoundland and net gains of rural youth in British Columbia.

Table 28 classifies the rural and urban components of economic regions into four categories : 1) regions which are net gainers of both population aged 15-29 and of population aged 15 and over, 2) regions which are net gainers of population aged 15-29 but net losers of population aged 15 and over, 3) regions which are net losers of both populations and, 4) regions which are net gainers of population aged 15 and over but losers of population aged 15-29. During the 1991-96 period, out of 71 economic regions which have a rural component, about 30 are net gainers of both populations and another 30 are net losers of both populations. In other terms, roughly 80% of economic regions which have a rural component are either net losers or net gainers of both populations.

Several points are worth mentioning. First, all rural as well as urban areas in Newfoundland are net losers of both populations. Second, in contrast, almost all rural areas in British Columbia are net gainers of both populations. Third, in the two biggest provinces, rural areas which are net losers of both populations coexist with rural areas which are net gainers of both populations. For instance, in Quebec, rural Bas-Saint-Laurent is a net loser while the rural components of Montérégie, Lanaudière, Laurentides and Outaouais are net gainers.

Annual migration outflows are presented in Table 29. They show that year by year, 9-10% of teenagers leave rural areas. Since outflow rates for this group over a 5-year period are roughly 30%, this implies that 5-year outflow rates are less than five times the annual outflow rates. This is so simply because some individuals who leave their community between year t and year t+1 will have returned by year t+5, a phenomenon which is called return migration. Contrary to what is observed for the 5-year migration outflows, outflow rates of individuals in their early 20's living in rural areas exceed those of teenagers. This could occur if return migration among the former is more frequent than among the latter. We investigate this issue in Section 6.

Another piece of evidence regarding return migration can be obtained by looking at migration outflows over a 10-year period. Table 30 shows that between 1987 and 1997, almost 50% of teenagers left their rural community. Once again, this percentage is smaller than ten times the annual outflow rates, suggesting that return migration plays a role here.

1) At the Canada level, individuals tend to leave rural areas more often than urban areas : the difference is particularly pronounced for individuals aged 15-19;

2) In all provinces except New Brunswick, individuals aged 15-19 tend to leave more often rural areas than urban areas;

3) Among individuals aged 20-24, outflow rates in rural areas are higher than those of urban areas in Ontario, Alberta and British Columbia. Contrary to our expectations, the former are lower than the latter in the Atlantic provinces;

4) Among individuals aged 25-29, outflow rates in rural areas are higher than those of urban areas in Ontario, Alberta and British Columbia but are lower in the Atlantic provinces, Quebec and Saskatchewan;

5) For all three age groups of youth and all provinces except Newfoundland, the main destination of individuals who leave rural areas is an urban area inside the province of origin. In Newfoundland, the main destination is an urban area outside the province of origin;

6) For individuals leaving urban areas, the main destination varies. It is an urban area outside the province for youth living in the Atlantic provinces and in the Prairies. It is an urban area inside the province for Quebec, Ontario and British Columbia;

7) At the Canada level, inflow rates for rural areas exceed those for urban areas for all individuals aged 20 and over (20-24, 25-29, 30-44, etc.). However, this pattern is not observed in all provinces;

8) Among individuals aged 20-29, inflow rates in rural areas are smaller than those in urban areas for Newfoundland, Prince Edward Island and New Brunswick. Inflow rates in rural areas are higher than those in urban areas in Ontario, Manitoba, Alberta and British Columbia;

9) Among individuals aged 15-19, inflow rates in rural areas are smaller than those in urban areas in all provinces except Ontario, Alberta and British Columbia;

10) New residents of rural areas generally come from an urban area inside the province in which they live in year t+5. However, in Newfoundland and New Brunswick, the main area of origin for individuals aged 25-29 in year t is an urban area outside the province (Table 18, colums 6 and 7). The same is also true for teenagers living in Prince Edward Island;

11) New residents of urban areas sometimes come mainly from rural areas inside the province, urban areas inside the province or urban areas outside the province;

12) At the Canada level, rural areas lose 12-16% of their population aged 15-19 to urban areas during a 5-year interval. Whether rural areas experience a net loss of individuals aged 20-24 is unclear. Rural areas experience a net gain of individuals aged 25 to 64 from urban areas. These three trends are observed in Quebec, Ontario, Manitoba and Alberta;

13) The net effect of the three trends mentioned in 12) is that, for the 1991-96 period, rural areas have, at the Canada level, enjoyed net gains of their population aged 15 and over of roughly 2% while urban areas have experienced net losses of 0.5%. For the 1991-96 period, the net effect of these trends on the rural population aged 15-29 is unclear;

14) Rural areas are net losers of their population aged 15-19 in all provinces. Net flow rates for individuals aged 20-24 are, depending on the data set used, either positive or negative in most provinces. For the 1991-96 period, rural areas are net gainers of their population aged 25-29 in all provinces except Newfoundland, New Brunswick and Saskatchewan;

15) Over the 1991-96 period, all the Atlantic provinces, Manitoba and Saskatchewan are net losers of their rural population aged 15-29. Ontario, Alberta and British Columbia are net gainers;

16) Over the 1991-96 period, Newfoundland, New Brunswick and Saskatchewan are net losers of their rural population aged 15 and over while Quebec, Ontario, Alberta and British Columbia are net gainers;

17) As a result of 15) and 16), Newfoundland, New Brunswick and Saskatchewan appear to be having problems maintaining the size of their rural youth population and of their rural population aged 15 and over. The problem seems to be particularly acute in Newfoundland;

18) Even within provinces, rural areas are a very heterogeneous group. For instance, the rural component of Northeast Ontario has experienced a net loss of its population aged 15-29 of at least 5% during the 1991-96 period while the rural component of Kitchener - Waterloo - Barrie has enjoyed a net gain of at least 5%.

While the numbers presented in Section 5 provide interesting information on migration flows in rural and urban areas, they do not consider the possibility that some young individuals who moved out of a given rural community between year t and year t+5 will eventually return to this community. Since the process of migration involves several decisions [ 1) where to pursue post-secondary education, 2) where to find a job, 3) where to establish one's family ], it is likely to be a dynamic process possibly involving several destinations. For instance, an individual may leave his/her rural community for an urban area to pursue post-secondary education and then may come back to his/her rural community if job opportunities are favorable enough.

To assess the magnitude of return migration, one needs longitudinal data which track people over time for a sufficiently long period. Census data cannot be used to study return migration since it allows only a comparison of places of residence between two years.⁽¹³⁾ In principle, SLID allows an examination of return migration but the time interval it currently covers, 1993-1997, is still fairly short. To address the issue of return migration, we use T-1 tax records and examine migration patterns between 1987 and 1997.

In Table 31, we ask the following question : of all individuals aged 15-19 who were present in a given community in 1987, what proportion : 1) stayed in 1992 and in 1997, 2) stayed in 1992 but left between 1992 and 1997, 3) left between 1987 and 1992 and returned by 1997, 4) left between 1987 and 1992 and did not return to their community by 1997. We use the term "returner" to refer to the third category. The first four columns of the table define the aforementioned categories. Column 6 shows the percentage of individuals who left between 1987 and 1992 : this is equivalent to the outflows measured in section 5.1.1. Column 7 shows the percentage of individuals who have left between 1992 and 1997 after staying in their community between 1987 and 1992. Column 8 calculates the percentage of leavers who are returners.

The definition of returners which we use here is a natural extension of the work done in Section 5, where we looked at the percentage of individuals who left their community between year t and year t+5. Here, we ask what happened to these individuals five years later, i.e. in year t+10. As we shall see below, this definition of returners is fairly restrictive since it excludes individuals who left between year t and, say, year t+4 (or any year other than t+5) and returned to their community by year t+10. We will use a broader definition of returners in section 6.2. This broader definition will include all individuals who left between year t and any year t+i, i = 1, … 9, and returned to their community by year t+10.

Only 54% of rural youth, aged 15 to 19, are in their original community ten years later (Table 31, columns 1 and 3). Of all individuals aged 15-19 who were present in a given rural community in 1987, 3-6%, depending on the province considered, left between 1987 and 1992 and returned afterwards, i.e. are returners (column 3). The corresponding percentages for urban areas are 3-5%. In both rural and urban areas, of all individuals who left between 1987 and 1992, at most 20% have returned by 1997 (Column 8). In other words, the probability of returning, conditional on having left initially, is at most 20%. This implies that, out of 5 people who leave between year t and year t+5, only 1 will have returned by year t+10. The implication of this result is clear : policy makers in various economic regions cannot count on return migration as a means of preserving the population size of a given age cohort. Rather, they must rely on inflows from other regions to achieve this goal.

At the Canada level, 33% of teenagers left their rural areas between 1987 and 1992. An additional 18% left between 1992 and 1997. The result is that only half of teenagers stayed in their rural community in 1987, 1992 and 1997. In urban areas, the proportion of leavers is - as we saw in Section 5 - smaller. Eighteen percent of teenagers left their urban areas between 1987 and 1992. An additional 14% left between 1992 and 1997.

Except in Newfoundland, the probability of leaving one's rural community if one has stayed in his/her community between year t and year t+5 is smaller than the probability of leaving between year t and year t+5. At the Canada level, 33% of teenagers left their rural community between 1987 and 1992 (Column 6). However, among those who stayed in their community between these two years, only 27% left afterwards (Column 7). In other words, staying in one's community in year t and year t+5 increases one's chances of staying in year t+10. This pattern is not observed in urban areas : the probability of leaving and the probability of leaving conditional on having stayed in year t and year t+5 are essentially the same (18%). Thus, staying initially in one's community lowers one's chances of leaving in rural areas but not in urban areas.

Tables 32 and 33 replicate Table 31 for individuals aged 20-24 and those aged 25-29, respectively. Roughly 60% (70%) of rural youth aged 20-24 (25-29) are in their original community ten years later (columns 1 and 3). Once again, of all individuals who are present in their community in 1987, very few (2% to 5% depending on the age group and the province selected) leave and return afterwards. Second, no more than 16% of those who leave between year t and year t+5 will have returned by year t+10 (Column 8). Third, for both individuals aged 20-24 and those aged 25-29 and for both rural and urban areas, the probability of leaving conditional on having stayed initially (Column 7) is generally twice as small as the probability of leaving initially (Column 6).

For all three age groups, rural youth in the western provinces are less likely to stay in their rural and small town community for ten years.

In Tables 31 to 33, a returner was a person who had left a geographical unit (defined jointly in terms of economic region and rural/urban status) and had returned to it. Do the qualitative conclusions presented in the previous paragraphs change when we define a returner as a person who leaves an economic region and returns to it later ?

The answer is no. Appendix Tables 11 to 13 replicate Tables 31 to 33 using economic regions (instead of geographical units) as the unit of analysis. As expected, the percentage of stayers (column 1) increases when we move from geographical units to economic regions. However, three qualitative conclusions remain. First, of all individuals present in an economic region in 1987, still very few (1% to 6%) leave and return afterwards, i.e. are returners. Second, no more than 21% of those rural residents in 1987 who left their economic region between year t and year t+5 will have returned to it by year t+10 (column 8). Third, once again, for both individuals aged 20-24 and those aged 25-29 and for both rural and urban areas, the probability of leaving conditional on having stayed initially (Column 7) is generally twice as small as the probability of leaving initially (Column 6).

The main ideas which emerge from Tables 31 to 33 and Appendix Tables 11 to 13 are that, whether we use geographical units or economic regions : 1) returners represent a very small fraction of all individuals present initially in a community, 2) at most 1 leaver out of 5 will return to his/her community ten years later and 3) for individuals aged 20-29, the probability of leaving one's community generally falls by at least 50% if one has stayed in his/her community in year t and year t+5.

Tables 34 to 36 show the location of 1997 residents in 1992 and in 1987. Table 34 considers the group of individuals aged 25-29 in 1997 (i.e. aged 15-19 in 1987). Of all individuals aged 25-29 in 1997 who were living in a rural community in Newfoundland in 1997, 80% were present in this community both in 1992 and 1987. We know, from the first column of Table 31, that, of all individuals aged 15-19 in 1987 who were present in a rural community in Newfoundland in 1987, 51% were present both in 1992 and 1997. Since the number of individuals involved is exactly the same in both cases, it follows that the first percentage (80%) is greater than the second (51%) simply because the size of the cohort (i.e. the denominator) has fallen between 1987 and 1997. The same pattern is found for all other Atlantic provinces.

As was found in Tables 31 to 33, returners (column 3 of Tables 34 to 36) represent a very small fraction of the population of 1997 residents. Alberta and British Columbia appear to have a very "turbulent" population, with at least 35% of their 1997 rural residents coming from outside (column 4 of Table 34).

One way to interpret Table 34 is to abstract from deaths and international migration and imagine that mayors of rural communities hold a meeting every five years since 1987 for a given cohort. If so, mayors of rural communities in Newfoundland would observe that : 1) the size of the cohort aged 15-19 in 1987 is declining, 2) the vast majority (80%) of the individuals remaining in 1997 were present at the first two meetings, 3) very few individuals (16%) present in 1997 were not present at the first meeting held in 1987 and, 4) even fewer individuals are returners (4%). In contrast, mayors of rural communities in British Columbia would find that : 1) the size of the cohort aged 15-19 in 1987 is growing, 2) a little bit more than one third (39%) of the individuals present in 1997 were present at the first two meetings, 3) more than half (56%) of individuals were not present at the first meeting in 1987 and, 4) very few are returners (6%). This simple exercise shows that a rural community may be growing and also experience a large turnover from its workforce. This last point can also be made for individuals aged 30-34 in 1997 and those aged 35-39 in 1997. For these two cohorts, at least 50% of the 1997 residents of rural communities in British Columbia were not present in 1987 (Tables 35 and 36).

So far, we have examined individuals who were present in the tax files in the three following years : 1987, 1992 and 1997. We have defined returners as individuals who left their community between 1987 and 1992 but returned to it by 1997. An alternative strategy is to use more severe selection criteria and restrict our attention to individuals who were present in the tax files for all eleven years of the 1987-1997 period. In this framework, we will define a permanent stayer as a person who stays in the same geographical unit throughout the 1987-1997 period. A returner will be a person who has changed geographical unit at some point during the period but was in the same geographical unit in 1997 as he/she was in 1987. A non-returner will be a person who has changed geographical unit at some point during the period and was in a different geographical unit in 1997, compared to 1987. Recall that a geographical unit is defined jointly in terms of economic region and rural/urban status. Tables 37 to 39 show the results of this alternative strategy for individuals aged 15-19, 20-24 and 25-29 in 1987, respectively.

Because we now require that an individual spends all eleven years of the 1987-1997 period - instead of year t, year t+5 and year t+10 - in his community to be defined as a (permanent) stayer, the percentage of permanent stayers in Tables 37 to 39 is smaller than the percentage of stayers identified in column 1 of Tables 31 to 33. In contrast, our definition of a returner is now broader than it was in Tables 31 to 33. In these tables, a person was defined as a returner if the 1987 location was different from the 1992 location but equal to the 1997 location. In Tables 37 to 39, a person is defined as a returner if the 1987 location is different from any location for the years 1988 through 1996 (i.e. including 1992) but equal to the 1997 location. This new definition includes the first one as a special case. For this reason, the proportion of returners in Tables 37 to 39 should be greater than it is in Tables 31 to 33.

This is indeed the case. While the percentage of returners varied between 1% and 6% in the first set of tables, it now varies between 4% and 14% under the new definition. In other words, the new definition more than doubles the percentage of individuals who are returners. Still, the percentage of non-permanent stayers (or leavers) who return to their original community is fairly low : no more than 1 leaver out of 4 returns to his/her community. Recall that the corresponding ratio was 1 to 5 for Tables 31 to 33. As was the case for the first set of tables, the degree of mobility of rural youth remains fairly high. At the Canada level, only 56% of rural youth aged 15-19 will be in their original community ten years later (Table 37, columns 1 and 2). The corresponding percentages for individuals aged 20-24 and those aged 25-29 are 64% and 74%, respectively.

For policy purposes, it is also of interest to examine how many individuals return to a rural community, whichever it is, in the same province (rather than return to the same rural community). For instance, it is conceivable that, even though a fairly small percentage of individuals who leave rural Avalon Peninsula return to it later, a greater proportion may return to a rural community in Newfoundland, whether it is rural Avalon Peninsula or another one.

Tables 40 to 42 address this issue. At the Canada level, the percentage of individuals who left their rural community and returned to a rural community within their province of origin averages 17%, 15% and 11% for individuals aged 15-19, 20-24 and 25-29, respectively (Tables 40 to 42, column 2). Of all individuals who left their rural community between 1987 and 1996, at most 39% had returned to a rural community within their province of origin by 1997 (Table 41, column 5). The corresponding maximum percentage is much higher for individuals who left their urban community : it equals 75% in Quebec, among teenagers (Table 40, column 5). Hence, for all three age groups and all provinces, return migration remains limited : at least 61 % (i.e. 100% - 39%) of individuals who left their rural community had not returned to a rural community within the province of origin ten years later. The corresponding percentage is 68% (i.e. 100% - 32%) for the Atlantic provinces (Table 42, column 5), suggesting that return migration is more limited in the Atlantic provinces than in the rest of Canada.

In section 6, we have used two definitions of returners. While the first one is a natural extension of the work done in section 5 when looking at 5-year migration flows, the second one should be preferred for analytical purposes since it does not restrict returners to those individuals who were in a different location in a particular year, i.e. in year t+5.

Tables 43 to 45 replicate Tables 31 to 33 for the Atlantic provinces. Data are presented for each economic region of these provinces. Looking across economic regions, the patterns found are generally similar to those obtained at the provincial level. As found above, a very small fraction of individuals (present in the tax file in 1987, 1992 and 1997) both leave their rural community and come back afterwards (1-6%). For those who left their rural community between 1987 and 1992, the chances of being back in 1997 are below 20%. Among individuals aged 20-24 in 1987 who left their rural community between 1987 and 1992, the chances of being back in 1997 are below 10% in Avalon Peninsula (9%), South Coast - Burin Peninsula (8%) and Cape Breton (9%) and equal 14% in Prince Edward Island and Annapolis Valley. Among individuals aged 25-29, the corresponding percentages are 8% or less in Avalon Peninsula (8%) and Saint John - St. Stephen (7%) and equal 14% or more in Prince Edward Island (14%) and Moncton - Richibucto (15%). Consistent with previous findings at the provincial level, the chances of leaving one's rural community between 1992 and 1997 after staying in this community in 1987 and 1992 are much smaller than the chances of leaving one's rural community initially.

Tables 46 to 48 replicate Tables 37 to 39. For all three age groups, the percentage of permanent stayers in rural areas is the lowest in Fredericton - Oromocto and the highest in Southern Nova Scotia.

Among teenagers who had left their rural community, very few returned to Avalon Peninsula (17%), compared to Prince Edward Island (24%). Of all rural communities in the Atlantic provinces, Fredericton - Oromocto has the highest percentage of returners. This is true for all three age groups. The reason for this pattern is not that leavers in this community have a high probability of returning (column 5 of Tables 46 to 48) : it is simply that there is a lot of leavers in this community to start with.

Taken together, Tables 43 to 48 show that whatever definition of returner we use and whatever age group we consider, the percentage of individuals who are returners never exceeds 15%. Likewise, the percentage of leavers who return to their rural community rarely exceeds 25%. As a result, rural communities in the Atlantic provinces, like those in the rest of Canada, cannot rely on return migration as a means of maintaining the size of a given cohort : they must count on inflows from other areas to achieve this goal.

The numbers presented in this section must be interpreted with caution. The reason is that we lose track of some individuals as we attempt to follow them over time. For instance, column 9 of Tables 31 to 33 indicates that of all individuals present in the tax file in 1987 and living in rural areas, 10-14% were lost when we imposed the restriction that they also be present in 1992 and 1997. The degree of attrition is higher when we select the second sample of individuals who are present through all eleven years of the 1987-1997 period. In this case, column 6 of Tables 37 to 39 show that the percentage of missing observations (using the number of individuals present in 1987 as the denominator) rises to 20-30%. If those individuals who are missing are more mobile than the others, the attrition could create selection bias problems. Because of the very limited number of covariates available in the tax file, we do not attempt to correct for potential selection biases. Thus, the reader must keep these limitations in mind when interpreting the numbers.

1) at most 56% of rural youth aged 15-19 are in their original community ten years later. The corresponding percentages for individuals aged 20-24 and those aged 25-29 are 64% and 74%, respectively;

2) at most 1 leaver out of 4 returns to his/her original rural community ten years later;

3) at most 39% of individuals who left their rural community will have returned to a rural community within the province of origin ten years later;

4) depending on whether we use a narrow definition or a broader definition of returners, the percentage of individuals who leave their rural community and return to it later varies between 1% and 6% or between 4% and 14%;

5) the percentage of individuals who leave their rural community and return to a rural community within their province of origin averages 17%, 15% and 11% for individuals aged 15-19, 20-24 and 25-29, respectively;

6) in general, the chances of leaving one's rural community between year t+5 and year t+10 after staying in this community in year t and year t+5 are much smaller than the chances of leaving one's rural community initially.

So far we have documented the magnitude of migration flows over different time periods. However, we have not identified the characteristics of individuals who leave their rural/urban community nor examine how their wage growth compares to that of stayers. Which individuals are the most likely to move ? Which individuals constitute the majority of movers ? How fast do earnings of movers grow relative to those of stayers ? These are the three questions which we examine in this section.

7.1 Which individuals are the most likely to move ? Which individuals constitute the majority of movers ?

These two questions are not easy to answer. The reason is that, ideally, one needs information about the individual characteristics of movers before they move. Tax data contains very few individual characteristics - essentially, sex, age and province of residence - and thus cannot be used to examine these questions. Census data contain individual characteristics after the move and will be used in this section. The only Canadian data set which contain individual characteristics of rural/urban movers before they move is the Survey of Labour and Income Dynamics (SLID). The main disadvantage of this data set is its relatively small sample size which precludes any analysis at the provincial level and thus forces us to conduct our analysis at the national level.

In Table 49, we use SLID to show the individual characteristics of movers in rural areas before they move to a different location. The period considered is 1993-1997 and is dictated by the availability of data. Two samples are selected : 1) all individuals aged 15-29 in 1993 and living in rural areas in 1993 and, 2) individuals aged 15-29 in 1993 who were living in rural areas in 1993 and who were not full-time students in either 1993 or 1997.⁽¹⁴⁾

For both samples, individuals aged 25-29 are less likely to move than their younger counterparts (Table 49, column 1). The propensity to move varies by education level. For both samples, university graduates move out of rural areas more often (40%-42%) than high school graduates (27%-30%). The degree of success on the labour market also matters. Individuals who were employed all year in 1993 were less likely to move than those who were not employed (i.e. either unemployed or out of the labour force) for the whole year. Married individuals are also less likely to move, compared to single individuals. Thus, individuals who are the most likely to move out of rural areas are relatively young, are not married, have a university degree and have limited success on the labour market.

In Table 50, we present individual characteristics of urban movers. As was found for rural movers, the propensity to move is, among non-students, lower for individuals aged 25-29 than among younger individuals (Table 50, column 1). Among non-students, the relationship between education and the propensity to move is different, compared to that observed for rural movers. Specifically, the propensity of university graduates to move out of urban areas is only slightly higher than that of high school graduates.

Which individuals constitute the majority of rural movers ? For both samples, a large share of individuals with post-secondary education (42%-44%) move from rural areas (Table 49, columns 2 and 5). Even though they have the highest propensity to move, university graduates are only a small fraction of rural movers (7%-9%) because of their low demographic weight. The same pattern holds among urban movers. Contrary to university graduates, individuals who are either not employed for the whole year or employed part of the year account for a significant fraction of rural movers (41%-56%). These findings imply that policies which would aim at lowering the percentage of rural movers by focussing only on university graduates would likely have a limited impact on migration statistics.

In Table 51, we use 1996 Census data and examine the characteristics of individuals after they move (during the 1991-96 period). As we found with SLID, individuals aged 25-29 (in 1991 or 30-34 in 1996) are less likely to move out of rural areas. Conversely, university graduates are more likely to move than high school graduates. Individuals whose industry of employment in 1996 was agriculture, fishing or trapping are much less likely to move than other individuals and those whose industry of employment in 1996 was business services are very likely to move. These facts hold in all provinces.

The interpretation of the numbers from Census is more complicated because we look at characteristics of individuals after the move, rather than before the move. For instance, knowing the industry of employment in 1996 may tell us little about the industry of employment in 1991. Agriculture, fishing and trapping could be an exception. The aforementioned finding could reflect the fact that individuals who are initally (i.e. in 1991) employed in agriculture, fishing and trapping have a strong attachment to their rural community and are very unlikely to move.

Table 52 shows the percentage distribution of rural movers by selected 1996 characteristics, for each province. Once again, university graduates represent a small fraction of the movers' population in rural areas.

Two Canadian data sets allow us to examine this question : the T1 tax file and the Survey of Labour and Income Dynamics. Once again, SLID allows only an analysis at the national level. In contrast, the T1 tax file enables us to conduct an analysis at the provincial level.

We first turn to SLID and examine the earnings growth of individuals who had positive wages and salaries both in 1993 and 1997. We also restrict these individuals to have no net income from self-employment in both years. We consider two measures of earnings growth between 1993 and 1997 : 1) the median percentage change in earnings and, 2) the median change in earnings. We choose the median values of these measures because averages can be easily contaminated by extreme values, especially in the case of percentage changes. The results from SLID are presented in the first panel of Table 53.

For all age groups and for both measures, individuals who leave rural areas experience faster earnings growth than those who stay in rural areas. Among individuals living in urban areas, earnings growth measured as the median change in earnings is greater for movers aged 15-19 and those aged 25-29 than for stayers : for individuals aged 20-24, earnings growth is very similar for movers and stayers. When using the median percentage change in earnings, earnings growth is greater among movers aged 20-29.

The second panel of Table 53 uses data from the T1 tax file and presents corresponding results for Canada and all provinces.⁽¹⁵⁾ In all provinces except New Brunswick and Alberta, both measures confirm that individuals who move out of rural areas experience faster earnings growth than those who stay in rural areas. In all provinces and for both measures, teenagers who move out of rural areas enjoy faster earnings growth than their urban counterparts. Among individuals aged 25-29, the opposite pattern is found : individuals who leave urban areas have greater earnings growth - measured as the median change in earnings - in all provinces except Newfoundland, Saskatchewan and British Columbia.

While it would seem reasonable to think that the faster earnings growth of rural movers, compared to rural stayers, results from the process of migration itself, it is not clear whether this is really the case. The faster earnings growth could be related to the possibility that individuals who move out of rural areas have a steeper age-earnings profile, i.e. have greater earnings growth potential, than individuals who stay in rural areas. Thus, it is unclear whether the difference observed in earnings growth is caused by migration itself or by individuals' unobserved heterogeneity. Disentangling these two effects is beyond the scope of this paper.

Individuals who return to their community may be those for which migration was a wrong decision, i.e. they may not have found the type of well paid jobs they were expecting to find (Courchene, 1974). If so, the earnings growth they experience during a given period could be lower than that experienced by individuals who migrated and did not return (i.e. the non-returners). In Table 54, we examine this issue by analyzing the earnings growth of returners, non-returners and permanent stayers over the 1987-1997 period. The sample selected consists of individuals : 1) who were present in the tax file for all eleven years of the 1987-1997 period, 2) who had positive wages and salaries in 1987 and 1997 and, 3) who had income from self-employment neither in 1987 nor in 1997. We use the second definition of returner, i.e. we define a returner as an individual who has changed geographical unit at some point during the period but was in the same geographical unit in 1997 as he/she was in 1987.

The evidence shown in Table 54 is consistent with the view that for some individuals who returned to their rural community, migration was a wrong decision at least in terms of earnings growth. Specifically, for all three age groups (15-19, 20-24, 25-29), the earnings growth of returners - whether measured in percentage change or in absolute change - is, at the Canada level, much smaller than that of non-returners. For instance, individuals who were aged 20-24 in 1987, who migrated and had returned to their rural community by 1997 saw their earnings increase by about $7,700 (in 1992 constant dollars) between 1987 and 1997, compared to roughly $13,400 for their counterparts who migrated but did not return to their rural community. The fact that, for rural areas, earnings growth of returners is smaller than that of non-returners, is observed in all provinces and for all three age groups.

This finding suggests that while return migration may be positive from a community's point of view - it helps maintain the population size of a given cohort - it may have been triggered by a negative labour market experience for those individuals who decide to return to their community. Alternative scenarios can be considered. A young individual may have moved from a given rural area to Toronto, found there a well paid job involving long hours and decided to return to that rural area and be employed in a job with lower wages and shorter hours. Thus, in this case, the return to the rural area would not be motivated by the fact that the person did not find a well paid job in Toronto.

This paper has documented migration patterns in rural and urban areas during the 1990s. Migration patterns in rural areas vary markedly across provinces. At one end of the spectrum, rural areas have been, in demographic terms, booming in British Columbia, showing net gains of individuals aged 15-29 of about 15% during the 1991-96 period. At the other end of the spectrum, rural areas have faced serious problems in Newfoundland : on average, they have experienced net losses of their youth population which are close to 15%. Rural areas in other Atlantic provinces, in Manitoba and Saskatchewan have gone through more moderate losses of their youth population. In Quebec and Ontario, net flows of rural youth population have been close to zero. Finally, Alberta has enjoyed moderate net gains in its population aged 15-29.

Even within provinces, migration patterns vary substantially across economic regions. In Quebec, rural areas in Lanaudière face much more optimistic prospects than those in Côte Nord or Gaspésie-Îles-de-la-Madeleine. Similarly, rural areas in Kitchener-Waterloo-Barrie have fared much better in the 1990s than rural areas in Northeast Ontario. This diversity of patterns must be kept in mind when thinking about appropriate interventions, if any, to implement.

Moving out of one's community is a phenomenon which is not limited to rural areas : outflow rates in urban areas amount to at least 75% of those in rural areas for individuals aged 20-29. Individuals leave their community for a variety of reasons : to pursue post-secondary education, to find a job, to obtain higher wages, to experiment with different life experiences, to gain independence or to fulfill one's aspirations. The incentives provided by urban areas may be relatively high for women since the rural/urban earnings gap is much more pronounced for them than it is for men.

Contrary to popular perception, rural areas are not net losers of individuals in all age groups. Abstracting from deaths and international migration, rural areas are net gainers of their population aged 25-29 in most provinces. This does not imply that there is no need for concern. As mentioned above, provinces with relatively low incomes (the Atlantic provinces, Manitoba and Saskatchewan) have experienced net losses in their rural population aged 15-29 during the 1991-96 period. Meanwhile, the richest provinces (Quebec, Ontario, Alberta and British Columbia) have been net gainers of their rural population aged 15 and over.

Interprovincial differences in unemployment are likely to be a major factor behind these differences. Among individuals aged 15-29 who are not students, the unemployment rate in rural areas averages 27% in the Atlantic provinces, compared to only 17%, 14%, 11% and 16% in Quebec, Ontario, Alberta and British Columbia, respectively (Appendix Table 9).

Because migration is not a one-step process, it is crucial to examine how many individuals return to their community after having left it. If a substantial portion of leavers were to return to their community, one could count on return migration as a means of maintaining the size of a given cohort in a community. The numbers presented in the paper indicate that such a hope is not justified. At most 25% of leavers return to their rural community ten years later. The implication is that rural areas must rely on inflows from other (urban) areas to maintain the size of a given cohort.

In an accounting sense, rural areas may face problems maintaining or increasing the size of a given cohort either because of insufficient inflows and/or because of excessive outflows. We have examined this issue for the Atlantic provinces. Our results show that in the Atlantic provinces, rural areas which fare worse than the national average - in terms of net gains of youth population - do so not because they have a higher than average percentage of leavers but rather because they are unable to attract a sufficiently high proportion of individuals into their communities.

For policy purposes, it is important to identify both individuals who have a high propensity to move and individuals who represent the majority of movers. We have analyzed this issue and found that even though university graduates have a high propensity to leave rural areas, they represent only a minority of rural leavers. Individuals with post-secondary education, who have a somewhat lower propensity to leave, account for the biggest share of rural leavers.

This paper has provided basic information on the magnitude of migration patterns in rural and urban areas. Because of a lack of appropriate data at the national level, it has not investigated the factors which underlie migration and return migration patterns. Understanding these factors is a necessary step before defining possible appropriate interventions.

Côté, S. (1997) 'Migrer : un choix ou une nécessité. Une enquête à l'échelle d'une région' in M. Gauthier, editor, Pourquoi partir ? La migration des jeunes d'hier et d'aujourd'hui, Sainte- Foy, PUL-IQRC, 315 pages.

Courchene, T. (1974) Migration, Income and Employment, Toronto, C.D. Howe Institute.

Finnie, R. (1998a) 'Interprovincial Mobility in Canada : A Longitudinal Analysis', Working Paper W-98-5E.a, Applied Research Branch, Human Resources Development Canada.

Finnie, R. (1998b) 'Interprovincial Mobility in Canada : Who Moves ? A Panel Logit Model Analysis', Working Paper W-98-5E.b, Applied Research Branch, Human Resources Development Canada.

Gauthier, M. (1997) editor, Pourquoi partir ? La migration des jeunes d'hier et d'aujourd'hui, Sainte- Foy, PUL-IQRC, 315 pages.

Grant, K.E. and J. Vanderkamp (1976) The Economic Causes and Effects of Migration : Canada, 1965-71, Economic Council of Canada (Ottawa : The Queen's Printer).

Mendelson, R. and R.D. Bollman (1998) 'Rural and small town population is growing in the 1990s' Working Paper No. 36, Agriculture Division, Statistics Canada.

Roy, J. (1997) 'La quête d'un espace sociétal' in M. Gauthier, editor, Pourquoi partir ? La migration des jeunes d'hier et d'aujourd'hui, Sainte- Foy, PUL-IQRC, 315 pages.

Two main data sources were used to measure flows of individuals in time: the Census and the Revenue Canada tax filers' files (the T1 files). Both have advantages and disadvantages. The Census covers 1 individual out of 5 and provides researchers with a few covariates. But it only occurs every 5 years, and it is not possible to link individuals across censuses rendering flow investigations spanning more than 5 years impossible. The T1 files on the other hand are annual and cover a substantial portion of the population: those who have filled an income tax statement. However not everybody files an income tax statement. Comparisons with other sources have shown that less than 10% of 15 year old file taxes. This number quickly grows with age such that virtually everybody aged 19 and over are covered in the T1 tax files. The coverage rate for the 15 to 19 age group amounts to 40%. The urban and rural coverage rates are virtually the same for every age categories. The main disadvantage of the T1 is its lack of socio-economic covariates.

These differences have shaped the uses that were made in this study of each set of files. The T1 files were used to generate the bulk of the flows across 1, 5, and 10 year periods while the Census data was used to shed light on the characteristics of both movers and stayers such as education levels and labour force characteristics over 2 five year periods (1991-1996, and 1986-1991). In the cases where the same phenomena could be measured from both sources (such as the 5 year flows of individuals) both sources were used as a data-validation exercise. It turns out that these independent sources of information do not always agree. Where they differ, we show it in the paper.

Flows are the movement of individuals across places, between 2 time periods. To simplify things and make our results comparable with information from other sources we opted to divide the Canadian landscape into 74 economic regions. Inside each economic region individuals living in a census agglomeration or a census metropolitan area were defined as being urban, the rest were defined as being rural. This results in 136 geo/states building blocks (12 economic regions are either all urban or rural).

Individuals were assigned one of these geo/states independently for each year. Flows between regions were registered by measuring the number of people moving from one geo/state to another between two time periods: a base and target year.

As geographies change in time (the urbanisation of our society for example) we always recalibrate the base year geo/states to those of the target year geo/states in order to ensure that the flows we are measuring are not the reflection of changes in municipal boundaries or of geographic typologies. These gains in coherency are at the price of precision regarding the base year urban/rural mix. This observation rests on the fact that during the years covered by this work Canada became more and more urban. Pushing back to a base year the rural/urban mix of a terminal year would thus have the effect of overestimating the size of the base year's urban component at the price of the size of its rural component. In order to minimise these biases - in a way that was practical from an operational point of view - the following choices were made:

Because both municipal boundaries and geographic typologies change over the timeframe covered by this work recalibration is required to ensure consistency in the data.

The most basic levels of geography available for the T1 files and Census are respectively the Postal Code (PC - T1) and the Census Sub Division (CSD - Census). Using postal code conversion files (PCCF) maintained in the Geography Division of Statistics Canada it is possible for a given year to know which CSD a PC spans. It is possible for a PC to span more than one CSD. This multitude of CSDs is recorded in the conversion files but a best choice CSD is also identified for every PC using a weighting scheme based on polulation. Using these files it is thus possible to translate T1 file PC into CSDs - and start on the same footing as Census data. There is a potential chasm here between the two sources as the Census is explicitly coded at the CSD level whereas the T1's CSD are derived. The most notable difference comes from the fact that there are CSD that are entirely comprised of PCs that point to another CSD as their best fit CSD. Although this would be a setback if our geographical unit of analysis was the CSD, the fact that we are working at a much larger level of geography (the economic region) means that this problem is very much contained: indeed, a tiny proportion of individuals is found in PCs that span more than one economic region.

Recalibration of base year CSDs to a set of terminal year compatible CSDs needs to take place because CSDs (municipalities) change in time … they often grow, and sometimes amalgamate. Recalibration is eased by the fact that CSDs do not tend to fracture but rather grow and combine with other CSDs. Because of this, recalibration can be achieved by combining into one CSD the CSDs of a base year that end up being grouped under the umbrella of a terminal year CSD.

Recalibrating allows us to roll back 1997 Economic Region (ER) typologies - ensuring strong geographic consistency in time - while allowing historical urban/rural mix to be respected in the 5 year comparisons.

About 1% of T1 records were excluded every year because of invalid geographic information in the file that precluded us from assigning a geo/state to these records. The main reasons for this were invalid and blank Postal Codes. Although large in numbers (171,391 in 1997) as a percentage of the total (0.81% in 1997) these numbers - although still significant - are judged to be small enough so as not to unduly affect the conclusions that we reach.

Flows are calculated from the T1 files by comparing the whereabouts of individuals in a base year and target year. If a person is in the same geo/state in both periods then that person is defined as a "stayer". If the geo/state of the target year is different than that of the base year then the person is defined as a "mover". Individuals present in a base year that are not found (using the Social Insurance Number - SIN - as the match key) in the target year are defined as "missing". There are many reasons that could explain the existence of this category. People that have died, left the country, or stopped filing an income tax statement correctly fall under this category. More problematic are the "missing" cases associated with individuals who have changed SIN between a base and target year. It has been estimated that about 3% of people change SIN in a given year, a significant number. This situation is all the more significant in the 5 and 10 year comparisons. Because of this, we include the share of "missing" in the tables that are the most affected by this. This problem is limited to the T1 data as the way of matching through time in the Census is different.

The flows measured from the Census do not require merging of separate annual files using SIN - indeed, it would be impossible to do so. Rather, the flows in the Census tables are calculated using a retrospective question that is asked on the long form of the questionnaire (20% of the population are asked to fill a long form rather than the simple basic questionnaire). Having on the same form information on the current and 5 year retrospective geography of an individual ensures that the missing component is nil.