Changes in the Regional Distribution of New Immigrants to Canada

View the most recent version.

Archived Content

Information identified as archived is provided for reference, research or recordkeeping purposes. It is not subject to the Government of Canada Web Standards and has not been altered or updated since it was archived. Please "contact us" to request a format other than those available.

Archived

This page has been archived on the Web.

by Aneta Bonikowska, Feng Hou, and Garnett Picot
Social Analysis and Modelling Division, Statistics Canada

Skip to text

• Abstract
• Executive summary
• 1 Introduction
• 2 Factors associated with changes in regional distribution of immigrants in the United States and Canada
• 3 Data and methods
• 4 Regression decomposition results
• 5 Conclusion and discussion
• 6 Appendix tables
• Notes

Text begins

Abstract

Canada and the United States have recently experienced an increase in the regional dispersion of entering immigrants. American research suggests that a mixture of economic push factors (away from states like California) and pull factors (toward states with growth of low-wage jobs), as well as changing government policies and regulations contributed to the development of the ‘New Gateways.’ Very few studies have been conducted to determine why the regional dispersion of entering immigrants occurred in Canada. This paper assesses the relative importance of immigrant selection programs and immigrant source regions in accounting for changes in the regional dispersion of entering immigrants during the 2000s. Using data from immigrant landing records and income tax files, this study shows that different factors accounted for changes in the share of immigrants settling in different destinations. Changes in immigrant selection programs, notably the Provincial Nominee Programs (PNPs), are the primary reason for increasing numbers going to Winnipeg and the province of Saskatchewan. Shifts in immigrant source regions were found to be an important factor in the decrease in immigration to Toronto and in the increase to Montréal.

Executive summary

Canada and the United States have recently experienced an increased regional dispersion of entering immigrants. American research suggests that a mixture of economic push factors (away from states like California) and pull factors (toward states with growth of low-wage jobs), as well as changing government policies and regulations contributed to the development of the ‘New Gateways.’ Very few Canadian studies have provided comprehensive explanations for the geographical redistribution of entering immigrants, although studies of changes in immigration policy suggest that the Provincial Nominee Programs (PNPs) were, at least in part, responsible for increased numbers having chosen the western provinces. However, these studies did not consider other possibilities, such as changes in immigrant source regions. This paper assesses the relative importance of immigrant selection programs and immigrant source regions in accounting for changes in the regional dispersion of entering immigrants.

The analysis is conducted separately for two different samples. The first focuses on the intended destinations identified on the landing records of all immigrants who arrived in Canada between 1999/2000 and 2009/2010. The second focuses on the actual destinations of immigrants who filed a tax return for the first full year after landing. A comparison of actual and intended destinations shows that there is a discrepancy between the two for many entering immigrants. However, while most regions lose some immigrants who intended to settle there, they gain immigrants from other regions. For the three major destinations—Toronto, Vancouver, and Montréal—the difference between the intended and actual number of immigrants is about 6%. Furthermore, the factors accounting for changing settlement patterns are very similar, whether actual or intended destinations are used.

A regression decomposition method is employed to assess the effect of changes in entry programs and source regions on the destination choice of new immigrants. Destinations were defined as major cities and the remaining provincial areas outside of those cities. Seven intended destinations experienced more than a one-percentage-point change in their share of new immigrants. They were Toronto, Montréal, Winnipeg, Calgary, Edmonton, Saskatchewan and Alberta (outside of Calgary and Edmonton). The regression analysis is restricted to these seven regions.

The results show that changes in the type of programs under which immigrants entered (specifically, the PNPs) accounted for virtually the entire rise in the shares of new immigrants going to Saskatchewan and Winnipeg, and played an important role in Alberta outside of Edmonton and Calgary. Changes in immigrant programs also tended to reduce the share of immigrants to Montréal.

Changes in source regions can also influence immigrants’ destination choices. Immigrants from a particular country or ethnic group tend to enter destinations with a pre-existing community of earlier immigrants from the same region or ethnic group, although the establishment of the pre-existing community might originate from certain regional characteristics, such as climate, dominant language, distance to home country, and labour market niches, that are attractive to a given immigrant group. The analysis suggests that the change in source regions was an important factor in the fall in immigration to Toronto and the increase in immigration to Montréal, but played a small role for any other destination.

Changing immigrant programs played a minor role in the remaining three destinations: Toronto, Calgary and Edmonton. Changes in regional economic conditions were likely important factors contributing to the changing shares of new immigrants going to Toronto, Montréal, Calgary and Edmonton.

In summary, although the changing entry program mix was significant, it was not the sole contributor to the geographic dispersion of immigrants who have come to Canada since 2000.

1 Introduction

Immigrants to North America have traditionally clustered in specific locations. In the United States, since the Second World War, cities such as Miami, Fort Lauderdale, Los Angeles, San Diego, Houston, Chicago, Jersey City and New York have been gateways for immigrants (Gozdziak and Martin 2005; Massey and Capoferro 2008; Singer 2004). By 2000, the states of California, Texas, New York, New Jersey, Illinois and Florida still accounted for more than two-thirds of all immigrants in the United States. In Canada, the concentration has been even more pronounced, with the majority of immigrants residing in the three largest cities—Toronto, Montréal, and Vancouver. As late as 2001, three-quarters of immigrants entering Canada went to one of these three metropolitan areas. Directing new immigrants toward other regions has been a policy goal of Citizenship and Immigration Canada (CIC) since the 1990s (CIC 2001).

Both countries have recently witnessed considerable geographical dispersion of entering immigrants—in the 1990s in the United States, and after 2000 in Canada. During the 1990s, U.S. immigration rates were very high, and the foreign-born population rose by 57% over the decade. At the same time, new immigrants became less geographically concentrated. The foreign-born population increased at double the U.S. national average in 13 states, mainly in the west and southeast, which traditionally had not been immigrant receivers^Note 1 (Camarota and Keeley 2001; Singer 2004). Although the majority of immigrants continued to settle in traditional immigrant-receiving states, some destinations were losing their appeal. American research suggests that a mixture of economic push factors (away from states like California) and pull factors (toward states with growth in low-wage jobs), as well as changing government policies and regulations, contributed to the development of the New Gateways.

Since 2000, a declining concentration of new immigrants has also been observed in Canada, with a downturn in immigration to Toronto and gains for non-traditional regions. Between 2000 and 2010, annual immigration to Canada rose from 227,500 to 280,700, but the percentage of immigrants intending to settle in Toronto fell from 48% to 33% (Table 1). At the same time, Alberta’s share of entering immigrants rose from 6.3% to 11.6%, with about half going to Calgary. Manitoba’s share rose from 2.0% to 5.6%. Saskatchewan’s share rose from less than 0.8% to 2.7%. The Atlantic provinces’ share increased from 1.3% to 3.0%. In the early years of the 2010s, the distribution continued to shift away from Toronto and Vancouver towards Alberta, Saskatchewan, and Quebec.

No Canadian research has provided comprehensive explanations for the geographical redistribution of entering immigrants, although studies of changes in immigration policy (CIC 2011; Carter, Morrish and Amoyaw 2008; Pandey and Townsend 2011, 2013) suggest that the Provincial Nominee Programs (PNPs), programs developed by the federal and provincial governments in the late 1990s, were at least partly responsible for the increased numbers of immigrants having chosen the western provinces. However, other possibilities, such as changes in regional economic conditions and changes in immigrant source regions, were not explored in these studies.

The goal of this analysis is to assess not all possible factors associated with the shifts in destinations of new immigrants but, rather, to assess specifically the relative importance of immigrant selection programs and immigrant source regions. Data limitations prevent the direct identification of the influence of changes in regional economic conditions. Using a regression decomposition method, this analysis quantifies the extent to which selection programs, notably the PNPs, and immigration source regions contributed to the geographic dispersion of immigrants who came to Canada over the 2000s. The results suggest that these factors played different roles for specific destinations.

2 Factors associated with changes in regional distribution of immigrants in the United States and Canada

A number of explanations have been proposed for the pattern of geographic dispersion of immigrants entering the United States in the 1990s (Massey 2008). The new destinations tend to have well-developed and growing, low-skilled service sectors that attract immigrants, particularly those from Mexico (Donato et al. 2008; Leach and Bean 2008). Leach and Bean (2008) argue that industrial restructuring and a shift in the location of economic growth in the United States, particularly associated with low-skilled jobs, were some main contributors. Donato et al. (2008) offer a similar explanation for increased settlement of immigrants in the Southeast and Midwest. They conclude that new immigrants are responding to the growth of low-wage jobs in industries such as manufacturing, construction and services (Johnston, Johnson-Webb and Farrel 1999; Johnson-Webb 2003; Parrado and Kandel 2008).

Other U.S. research focuses specifically on the increased geographical dispersion of Mexican immigrants (Kandel and Parrado 2005; Massey, Durand and Malone 2002; Zuniga and Hernandez-Leon 2005). Between 1990 and 1996, the percentage of Mexican immigrants going to California fell from 58% to 46%, and the percentage going to non-traditional gateway states rose from 10% to 21% (Durand, Massey and Charvet 2000). A conjunction of economic, social and political circumstances during the 1990s contributed to the increased geographic dispersion of Mexican immigrants. First, as a result of increased restrictions and security at the border, these immigrants found it increasingly difficult to enter the United States via the traditional crossings in Arizona, New Mexico and California, and so selected other entry points. Second, the poor economic conditions in California during the 1990s made it a less hospitable place for new immigrants. Third, the economic boom in other parts of the country in the late 1990s, notably the Midwest, Northwest and Southeast, created a sustained demand for unskilled and semi-skilled labour, with rising real wages. Finally, the Immigration Reform and Control Act (IRCA) of the late 1980s legalized 2.3 million Mexican immigrants, allowing them to move more freely in the United States without the fear of exposure in new regions.

In Canada, too, researchers have reported associations between employment opportunities and immigrant destinations. Hou (2007) documented changes in immigrants’ initial destinations and their subsequent distribution from the 1970s through the 1990s. The study suggested that the geographic concentration of immigrants in major metropolitan areas in the 1970s and 1980s reflected the tendency for immigrants to be drawn to large cities because of high demand for workers. Hou (2007) also suggested that immigrant concentration during the 1990s resulted primarily from the shift in source regions. Krahn, Derwing and Abu-Laban (2006) and Haan (2008, 2009) also reported associations between immigrant destinations and economic opportunities during these decades.

Changes in regional economic conditions, together with broader changes in the structure of the Canadian economy, may have contributed to the dispersion of new immigrants in the 2000s in Canada. As a result of these changes, the growth in the manufacturing sector, largely based in southern Ontario and parts of Quebec, was outpaced by commodities, particularly the petroleum sector, based primarily in the west. Hence, among the 20 destinations examined in this analysis, the unemployment rate rose most in Toronto, Hamilton, and "the rest of Ontario" between 1999/2000 and 2009/2010 (Table 2). On the other hand, in provinces west of Ontario, unemployment rates were lower and rose more slowly than in southern Ontario. In short, labour market conditions were better in western Canada than they were in the rest of the country. If regional economic conditions affect immigrants’ destination choices, Ontario’s share of new immigrants (notably, Toronto’s) would have fallen, even if immigrant selection programs had remained stable over time. Indeed, the large decline in the share of economic-class immigrants going to Toronto suggests that changes in relative economic conditions may have played a role. The decline in numbers of economic class immigrants accounted for almost 90% of the total decline in the number of new immigrants going to Ontario between 2000 and 2010 (CIC 2009, 2010b).

However, relative changes in regional economic performance are unlikely the only reason for the dispersion of immigrant destinations, given the significant changes in Canada’s immigration selection programs over the 2000s. More recent Canadian studies have emphasized the role of the PNPs (CIC 2011; Carter, Morrish and Amoyaw 2008). The PNPs, which allow provincial governments to actively select and nominate immigrants to meet their needs for population growth and economic development, could be viewed as a federal–provincial joint effort to achieve a more even distribution of immigrants across the country. Under the provisions of the program, provincial nominees are supposed to go to the province that arranged for their selection and entry, although no mechanism is established to enforce this. Over the 2000s, the PNPs considerably increased the number of immigrants going to destinations that previously received few immigrants. Empirical studies show that PNPs accounted for significant increases in immigration flows to Manitoba, Prince Edward Island, as well as New Brunswick, and provincial nominee immigrants were more likely than other economic immigrants to stay in the province of their initial destination (Pandey and Townsend 2011, 2013).

The percentage of immigrants arriving in Canada as provincial nominees rose from less than 1%, in 2000, to 13%, in 2010. The western provinces felt this increase most strongly. For example, in 2000, fewer than 15% of immigrants whose intended destination was Saskatchewan or Winnipeg entered under the PNPs; by the end of the decade, the figure was around 70% (Chart 1). In contrast, virtually none of the immigrants intending to settle in Ontario or Quebec during the 2000-to-2010 period entered under the PNPs. However, these two provinces were indirectly affected, because annual immigration to Canada remained more or less constant over the decade, and the rise in the PNPs was accompanied by a decline in the share of immigrants entering as skilled workers. The rise in the PNPs, and the decline in the share of skilled workers, will affect the locations in which new immigrants settle.

Description for Chart 1

The regional distribution of new immigrants could also be affected by changes in source countries. Immigrants from particular source countries (ethnic groups) tend to be drawn to destinations with large numbers of previous immigrants from the same country. Over the 2000s, a shift in leading source countries occurred (CIC 2010a). One possible reason for the shift is the modifications to the points system of selecting immigrants, under the Federal Skilled Worker (FSW) program. The Immigrant and Refugee Protection Act (IRPA), introduced in 2002, put in more stringent language requirements that might disproportionately affect certain source countries, particularly China. The percentage of the FSWs from China dropped from 28% before IRPA to 16% under IRPA (CIC 2010a). In the late 1990s, China (includes Taiwan and Hong Kong in this analysis) and India were the leading immigrant source countries, and these immigrants tended to settle in Toronto and Vancouver. By 2010, the Philippines had become the largest, single-source country, with the number of Filipino immigrants tripling through the 2000s. These immigrants had more diverse destinations, partly because their pre-existing communities were more dispersed across Canada. Such changes in source countries tend to increase the geographic dispersion of entering immigrants, even if the program mix remains constant. Changes in immigrant source countries would likely have a particularly strong effect on Toronto and Montréal. As shown in Table 3, the decline of immigrants from China, Southeastern Europe, and "other South Asia," would tend to reduce the share of immigrants settling in Toronto, which previously attracted a disproportionate share of immigrants from these source regions. In comparison, the increase of immigrants from Africa, South America, Central America and the Caribbean would increase the share of immigrants settling in Montréal, which was the main destination for immigrants from these regions.

It is difficult to separate the effects of changes in regional economic conditions, immigrant selection programs, and immigrant source regions. Their effects are likely overlapping, since these factors would all tend to direct new immigrants from Ontario towards western provinces. Moreover, changes in immigrant selection programs might be, at least partly, a response to changes in regional economic conditions. Indeed, a main rationale for provinces increasing their use of the PNPs was to meet perceived demands in local labour markets. Shifts in immigrant source regions might also be a result of changes in immigrant selection programs because immigrants admitted through PNPs tend to have different source-region composition than skilled workers. For instance, the large increase in the number of immigrants from the Philippines was related, at least in part, to the rapid expansion of the PNP in Manitoba, which has a relatively high concentration of immigrants from that country.

The analytic strategy of this study is to estimate the overall effects associated with changes in selection programs and immigrant source regions on a specific destination, keeping in mind that these estimated effects may be driven, in part, by regional economic conditions. The effect of changes in regional economic conditions cannot be estimated statistically, since economic conditions are measured at the aggregate level and for a given destination there are only two unique data points; i.e., the early 2000s and late 2000s. In comparison, information on the selection programs and source regions is available at the individual level, and thus allows statistical analysis to quantify the contribution of these two factors to the change in the share of immigrants entering a specific destination. It is reasonable to assume that regional economic conditions, along with other factors, may contribute to the portion of the change in the share of immigrants entering a specific destination that is not accounted for by selection programs and source regions.

3 Data and methods

3.1 Data

Background data on trends in the geographic distribution of new immigrants come from the Facts and Figures Web source developed by CIC. These data refer to all immigrants arriving each year.

Two other data sources are used for this analysis. The first source, the Immigrant Landing File,  is a microdata file based on immigrant landing records maintained by CIC. The landing records contain immigrants’ demographic characteristics, the selection programs through which they were admitted, and their intended destinations. With this file, the intended destination can be used as a proxy of immigrants’ initial destination. The advantage of this approach is that the information on intended destination is available for all immigrants. The disadvantage is that many immigrants do not settle in their intended destination (more on this later). Immigrants are free to settle anywhere in Canada, regardless of what they indicated as their planned destination.

The second data source is the Longitudinal Immigration Database (IMDB). The IMDB combines immigrant landing records and annual tax records. From the tax records, the database derives information on earnings and on other income components, taxes paid, current marital status, and geographic location of residence during the tax year. This file allows the immigrants’ actual initial destination to be identified, as reported on immigrants’ tax returns for the first full year in Canada. The advantage to this is that location information reflects where immigrants actually settled. The disadvantage is that not all immigrants file a tax return for their first full year in Canada. Over the study period, from 1999/2000 to 2009/2010, on average, 15% of immigrants, aged 18 to 54, did not file a tax return for the first full year after landing.^Note 2

3.2 Intended versus actual destinations

To assess the difference between intended and actual destinations, geographic distribution based on each definition is generated for immigrants who landed between 1999 and 2010, who were aged 18 to 54 at landing, and who filed a tax return for their first full year in Canada (approximately 85%) in the IMDB. Twenty provincial or metropolitan areas are used to identify both intended and actual destinations (see Table 1).

With the exception of Prince Edward Island, 57% to 87% (varies by destination) of immigrants who intended to settle in a particular destination actually did so in their first full year in Canada (Appendix Table A). For example, of the immigrants who said they intended to go to the rest of Quebec (excluding Montréal and the city of Québec), 65% went there; 21% went to Montréal, and 4% to the city of Québec. Similarly, 85% of immigrants intending to go to Edmonton actually did so; most of the remainder went elsewhere in western Canada.

Any particular destination also gains some immigrants who intended to settle elsewhere. From 9% to 43% of immigrants, who went to a given destination, initially said they intended to go elsewhere. For example, 9% of those who went to Calgary indicated Toronto as their intended destination, and 4% indicated Vancouver (Appendix Table B).

When gains and losses are taken into account, some destinations received fewer than the intended number of immigrants, while others received more. For Alberta, British Columbia, the rest of Quebec, the rest of Ontario, and Hamilton, actual immigrants exceeded intended immigrants. For the Atlantic provinces, Manitoba and Saskatchewan, Toronto, Montréal, and the city of Québec, the number of actual immigrants fell short of the intended number.^Note 3

To summarize, when focusing on individual immigrants, there is substantial discrepancy between their intended and actual destinations. However, in the aggregate, after all the reshuffling, typically, the difference between the intended and the actual number, for most destinations, is not large. The exceptions were Newfoundland and Labrador, Prince Edward Island and New Brunswick, which received only 39% to 80% of their intended immigrants. In contrast, the rest of Alberta and the rest of Ontario received 25% more than the intended number.

Since neither the intended destination nor the actual initial destination provide the full picture about immigrants’ initial distribution patterns, subsequent analyses are conducted using both definitions (and two data sources) to check whether the results are consistent.

3.3 Methods

The change in the regional distribution of immigrants at landing is compared between 1999/2000 and 2009/2010. To increase the sample size and, thereby, examine a larger number of destinations, two years of data, at both the beginning and end of the period, were pooled (i.e., 1999/2000 and 2009/2010). For analysis on intended destinations, the total sample size is 282,000 for 1999/2000 and 365,760 for 2009/2010. For analysis on actual destinations one year after landing, the total sample size is 245,170 for 1999/2000 and 307,590 for 2009/2010.

A regression decomposition method is employed to assess the effect of entry-program changes and source regions on the destination choice of new immigrants. The analysis is restricted to regions that experienced a significant change in their share of entering immigrants. Seven intended destinations experienced more than a one-percentage-point change in their share of new immigrants. For each of these seven regions, three linear probability models were run using pooled data for 1999/2000 and 2009/2010. The dependent variable in all three models takes the value 1, when an immigrant had intended to settle in a particular destination, and takes 0 otherwise. That is, it is the probability that an immigrant had intended to go to a particular destination.

Model 1 has a dummy variable for the landing years 2009/2010 (i.e., 1999/2000 is the reference), but no other control variables. The coefficient on the landing-year dummy variable indicates the observed change between 1999/2000 and 2009/2011 in the probability of immigrants intending to settle in a particular destination. Model 2 adds controls for immigrant entry programs: skilled workers, family class, business class, provincial nominee, refugee, and others (reference group).^Note 4 The difference between Models 1 and 2 in the coefficient on the 2009/2010 landing-year dummy indicates the effect of program changes (the probability of entering under a particular program) on the choice of a given destination.

Model 3 includes 12 source-region dummy variables. These dummy variables are based on 13 source countries or regions of immigration: China; India; the Philippines; the United States; South America, Central America and the Caribbean; other East Asia; other South Asia; other Southeast Asia; West Central Asia and the Middle East; Northwestern Europe (reference group); Southeastern Europe; Africa; and Australia and Oceania. The change in the coefficient on the 2009/2010 landing-year variable between Models 2 and 3 indicates the additional effect of the changing distribution of source regions on choice of landing destination. The coefficients for Models 1 to 3 are shown in Appendix Table C.

Note that the order in which variables are entered into the above-mentioned, sequential, regression models may affect the estimates in terms of their effects on the dependent variable (the probability of an immigrant intending to settle in a particular destination). Selection programs are intentionally entered first, because changes in the composition of source countries or regions could be a result of the alterations in the selection programs. In this way, the estimated effects of selection programs are maximized. Empirically, the results of this study are similar if the order is reversed, or if the decomposition exercise is based on a model that includes both sets of variables simultaneously (Hou 2014).

The above procedures were repeated for the sample of immigrants who filed taxes in the first full year after landing and who indicated their actual initial destinations. The dependent variable, now, is the probability of actually living in a given destination during the first full year since landing. One difference in the models for this sample is the addition of a variable indicating whether an immigrant filed taxes in Canada before the year when he or she became a landed immigrant. This variable is not available in the landing file for the analysis on intended destinations. It is included here to capture, specifically, the increased incidence of transitions from temporary residence status to permanent residence status over the 2000s. This is done, primarily, through the PNPs, as well as through the Canadian Experience Class, which was implemented in the late 2000s. In both Alberta and British Columbia in particular, most provincial nominees were selected from among temporary foreign workers as required by most streams within the programs. It is also possible that immigrants, who were temporary residents in Canada prior to landing, were strongly attached to their place of residence before landing, wherever in Canada it may have been. Therefore, they were more likely to remain there after becoming landed immigrants, regardless of the program through which they landed. The inclusion of this variable would help parse out whether it is changes in selection programs, or increases in transitions from temporary residence status to permanent residence status within each program, that contribute to the dispersion of immigrant destinations.

4 Regression decomposition results

4.1 Changes in the distribution of intended destinations

Toronto, the destination with the largest share of new immigrants to Canada, and with the largest change in its annual percentage of new immigrants, illustrates the results of the regression decomposition. The three regression models are presented in Appendix Table C, and the summary results are presented in Table 4. Between 1999/2000 and 2009/2010, the percentage of new immigrants intending to settle in Toronto fell by 14.2 percentage points (Table 4). If immigrant programs were held constant over the period, the share of new immigrants intending to settle in Toronto would have fallen by 11.0 percentage points. Thus, 3.2 points (or 23%), out of the 14.2-percentage-point decline, were associated with changes in immigrant programs at the national level. Holding constant immigrant source regions as well, the share of new immigrants intending to settle in Toronto would have fallen by 8.6 percentage points. Put differently, changes in immigrant source regions account for 2.4 percentage points out of the 14.2-percentage-point decline, net of the change associated with immigrant programs. Together, the two factors accounted for about 40% (5.6 out of 14.2) of the decline in the share of new immigrants intending to settle in Toronto during the 2000s.

The share of new immigrants who intended to settle in Montréal rose by 4.6 percentage points over the 1999/2000 to 2009/2010 period. If there had been no change in immigrant program composition, the share would have increased by 8.1 percentage points. This suggests that the observed overall rise occurred because Montréal received a larger share of immigrants within various programs. However, with further controlling for immigrant source regions, the share of new immigrants who intended to settle in Montréal would have increased by only 4.7 percentage points, again, net of the change associated with immigrant programs. Over the 2000s, the shares of immigrants from Africa and from South America, Central America and the Caribbean increased nationwide, and these immigrants were more likely than others to have chosen Montréal as their intended destination (Table 3). These results suggest that changes in immigrant programs tended to reduce the share of immigrants who intended to settle in Montréal, while changes in the immigrant source regions tended to increase it.

After Montréal, Winnipeg showed the second-largest increase in the share of prospective immigrants. The increase was entirely associated with selection-program changes, particularly the rise in the PNPs. Similarly, in Saskatchewan, almost the entire 1.7-percentage-point increase in its share of prospective immigrants was associated with changing program types.

In Calgary and Edmonton, the two major metropolitan areas in Alberta, about 10% to 30% of the increases in their share of prospective immigrants were associated with changes in entry programs. By comparison, about 60% of the 1.2-percentage-point increase in the share of immigrants intending to settle in Alberta, outside of Edmonton and Calgary,was associated with changes in immigrant programs. In all three cases, changes in immigrant source regions contributed little to the increase in the share of immigrants who intended to settle there.

In sum, the effect of changes in immigrant programs varies across individual destinations. Not surprisingly, the effect is stronger in destinations where PNPs have become the major source of immigrants. In Saskatchewan and Winnipeg, virtually all the gain in prospective immigrant share was associated with changes in entry programs; notably, the rising percentage of immigrants entering under the PNPs. As well, changes in entry programs accounted for most of the increase in the rest of Alberta. Changes in source regions tended to increase the share of immigrants who intended to settle in Montréal, but reduced it for Toronto.

4.2 Changes in the distribution of actual initial destinations

The results on actual destinations among immigrants who filed taxes were generally similar to those reported for the intended destinations. For example, Toronto’s share of new immigrants declined 12.3 percentage points between 1999/2000 and 2009/2010, the largest change for any destination (Table 5, first column). When controls for immigrant program type were introduced, the decline was 9.5 percentage points (Table 5, second column). That is, when immigrant program composition was held constant (Model 2), the change in the probability of an immigrant going to Toronto was 2.8 percentage points less than what was actually observed. Thus, 2.8 percentage points, or 23%, of the 12.3-percentage-point decline can be attributed to the change in the percentage of immigrants entering under various programs, which is the same as the result observed earlier for intended destinations.^Note 5 Prior Canadian residence accounted for little of the decline in the share of new immigrants to Toronto (Model 2a, column 3). Changes in source regions accounted for about 20% of the decline, because the city received an above-average share of its immigrants from source regions that were in decline (China, other South Asia) and a below-average share from regions in ascendancy (Africa and Northwestern Europe) (Table 3). Overall, over half of the decline in the share of new immigrants settling in Toronto was not associated with changes in immigrant programs and source regions.