Revising the Harris-Todaro framework to model labour migration from the Canadian Northern frontier

22
Vol. 24, No. 2, 2007 Journal of Population Research 185 REVISING THEHARRISTODARO FRAMEWORK TO MODEL LABOUR MIGRA T A A ION FROM THE CANADIAN NORT R R HERN FRONTIER Andrey N. Petrov, University of Toronto The Harris–Todaro model of labour migration was developed almost four dec- ades ago, and since has become a classic method of migration analysis in less developed countries. This paper explores the applicability of the Harris–Todaro (HT) framework outside its traditional use, by modelling frontier–metropolis migration in Canada. If appropriate, the framework can potentially be used in other countries with similar regional dichotomies, such as Russia and Australia. The paper argues that the HT model is generally applicable in the context of migration from the resource frontier to large metropolitan areas of the Cana- dian south, although it requires several modifications. The classic HT model is extended to account for northern labour-force heterogeneity (Aboriginal and Non-Aboriginal), the possibility of having or losing jobs in the declining and undiversified frontier economy, and living cost differentials. Further analysis is focused on Aboriginal migration from isolated northern communities. The plausibility of the modified HT model is demonstrated using an empirical test, in which the model is used to assess migration probabilities of Aboriginal la- bour migrants. These new insights into the mechanisms of frontier–metropo- lis migration could provide a better basis for developing planning strategies, aimed to sustain human capital in the Canadian North, and for optimizing wel- fare policies both in the North and in the South. Keywords: labour migration, frontier workers, Aboriginal population, economic models, Canadian North The Harris–Todaro (HT) model initially proposed in Todaro (1969), and restated in Harris and Todaro (1970), became a classic framework for analysing migration and labour-market equilibrium. At the time, the model was very innovative if not revo- lutionary (Allen 2001). It demonstrated that, in a two-sector regional system, migra- tion from underdeveloped (rural) to developed (urban) sectors will persist despite high unemployment in the latter, as long as there are sufficiently high expectations of finding a job and earning more. It was also argued that the creation of new jobs in the urban sector imminently leads to an increase in urban unemployment, a situ- ation known as the ‘Todaro paradox’ (Raimondos 2003). 1 During the last three dec- ades, the model has been significantly modified (e.g. Corden and Findley 1975; Fields 1975; Neary 1981; Gupta 1994; Kritchel and Levine 1999), and was widely applied in Address for correspondence: Andrey N. Petrov, Department of Geography , University of Toronto, Toronto ON M5S 3G3, Canada. Email: andrey[email protected]a

Transcript of Revising the Harris-Todaro framework to model labour migration from the Canadian Northern frontier

Vol. 24, No. 2, 2007 Journal of Population Research

185

REVISING THE HARRIS–TODARO FRAMEWORKTO MODEL LABOUR MIGRATAA ION FROM THECANADIAN NORTRR HERN FRONTIER

Andrey N. Petrov,† University of Toronto

The Harris–Todaro model of labour migration was developed almost four dec-ades ago, and since has become a classic method of migration analysis in lessdeveloped countries. This paper explores the applicability of the Harris–Todaro(HT) framework outside its traditional use, by modelling frontier–metropolismigration in Canada. If appropriate, the framework can potentially be used inother countries with similar regional dichotomies, such as Russia and Australia.The paper argues that the HT model is generally applicable in the context of migration from the resource frontier to large metropolitan areas of the Cana-dian south, although it requires several modifications. The classic HT model isextended to account for northern labour-force heterogeneity (Aboriginal andNon-Aboriginal), the possibility of having or losing jobs in the declining andundiversified frontier economy, and living cost differentials. Further analysisis focused on Aboriginal migration from isolated northern communities. Theplausibility of the modified HT model is demonstrated using an empirical test,in which the model is used to assess migration probabilities of Aboriginal la-bour migrants. These new insights into the mechanisms of frontier–metropo-lis migration could provide a better basis for developing planning strategies,aimed to sustain human capital in the Canadian North, and for optimizing wel-fare policies both in the North and in the South.

Keywords: labour migration, frontier workers, Aboriginal population, economicmodels, Canadian North

The Harris–Todaro (HT) model initially proposed in Todaro (1969), and restated inHarris and Todaro (1970), became a classic framework for analysing migration andlabour-market equilibrium. At the time, the model was very innovative if not revo-lutionary (Allen 2001). It demonstrated that, in a two-sector regional system, migra-tion from underdeveloped (rural) to developed (urban) sectors will persist despitehigh unemployment in the latter, as long as there are sufficiently high expectationsof finding a job and earning more. It was also argued that the creation of new jobs in the urban sector imminently leads to an increase in urban unemployment, a situ-ation known as the ‘Todaro paradox’ (Raimondos 2003).1 During the last three dec-ades, the model has been significantly modified (e.g. Corden and Findley 1975; Fields1975; Neary 1981; Gupta 1994; Kritchel and Levine 1999), and was widely applied in

† Address for correspondence: Andrey N. Petrov, Department of Geography, University of Toronto, Toronto ON M5S 3G3, Canada. Email: [email protected]

186 Andrey N. Petrov

the study of less developed countries (LDCs) (see Allen 2001; Ingene 2001). MichaelTodaro, in numerous publications, emphasized not just the theoretical importance of his findings, but also their implications for migration and development policies.

In the past decades, a number of studies have successfully applied the Harris–Todaro model to developed countries. Some later work used the HT model to studymigration processes in the USA (Suits 1985), within the European Union (Barry 2003)and between EU and developing countries such as Turkey (De Santis 2003). These studies showed that the HT approach may have universal applications, if conditionsand factors of labour migration are similar to those assumed in the model. Eaton andNeher (1975) were, perhaps, the first who asserted that the HT model is generalizableto any dual labour market, which has an institutionally determined higher wage inone sector (region). The current paper argues that the model can be used to analysemigration processes in the underdeveloped and isolated parts of developed coun-tries. The Canadian North is the focus of the paper.

The need to reconceptualize labour migration in the Canadian North is driven byan intensive loss of population in the region over the 1990s, although some regions rebounded in the 2000s. The observed loss in labour force was manifested in a long-term out-migration from the North to the southern parts of Canada (Bone 2003). If in the past a substantial outflow of labour was compensated by an ever-larger in-migra-tion, in the last decade this tendency was not followed (Statistics Canada 2002b). Although the increase in the out-migration from the North has been well recorded,there is still a lack of understanding of its mechanisms. There have been claims thatthe North is the exception in respect to migration ‘laws’, and it was suggested thatfurther research is required to test the applicability of conventional models and pro-vide an empirical framework for migration modelling in the North (e.g. Heleniak 1999; Petrov 2005). Therefore, it remains uncertain whether the out-migration fromthe Circumpolar North is consistent with existing models of labour migration devel-oped and used in other regions.

Given the legacy of the HT approach and its successful application in variousgeographic regions, the HT model may be found useful for the purpose of analys-ing the labour-driven component of the North-to-South migration in Canada. First, it appears reasonable to adopt the HT assumption that northern out-migrants con-sider expected earnings in the South in their moving-out process. Furthermore, theemphasis on the probability of getting a job, as the foundation of the migrant’s deci-sion-making, presents a better alternative to the traditionally used notion of wagedispersion and perfect information about regional wages (see Miron 1978). Second,the two-sector character of the economy, assumed by the HT model, resembles the relationship between the northern resource economy and the diversified service andmanufacturing economy of the Canadian South. This dichotomy is evident not only in the occupational structure, but also in regard to trade, when the resource sectortends to be disadvantaged compared to the industrial core (e.g. Sarkar 1994). Theexistence of the traditional sector in frontier regions makes the analogy with LDCseven more relevant. In addition, the non-metropolitan and rural origins of migra-tion in the North correspond with predominantly urban-oriented destination choices (Statistics Canada 2002b). As in the HT model, the act of migration from the Northfrequently appears to result in the migrants’ unemployment at the destination: it isespecially true for the Native northerners. Therefore, the HT framework is useful forstudying the labour out-migration from the North.

Revising the Harris–Todaro Framework 187

Study problem and objectives

First, this paper reviews the foundations and examines the use of the classic behav-ioural HT framework for studying regional labour migration, as it relates to the cur-rent case study. The purpose of this brief review is to inquire if, theoretically, the HTmodel can be applied to the case of labour out-migration in the circumpolar frontier–metropolis context. Secondly, the paper introduces several modifications aiming toaccommodate the unique socio-economic attributes of the region. Thirdly, an empiri-cal example of selected northern communities is used to confirm that the model pro-duces plausible results, that is, to test if predicted out-migration probabilities matchwith an observed out-migration. Because the case study uses the data for Aborigi-nal migrants, this study also works toward developing a comprehensive modelling framework of the Aboriginal migration in Canada.

Study area

For the purposes of this study, the definition of the Canadian northern frontier is lim-ited to 20 northern Census divisions of Canada: it includes the Northwest Territories,

Figure 1 Northern frontier communities and major metropolitan areas inCanada

Earnings data source: Statistics Canada 2003b. AB: Alberta; MB: Manitoba; NB: New Brunswick;NF: Newfoundland and Labrador; NS: Nova Scotia; NT: Northwest Territories; NU: Nunavut;ON: Ontario; PE: Prince Edward Island; QC: Quebec; SK: Saskatchewan; YT: Yukon.

188 Andrey N. Petrov

Nunavut, Yukon and the northern parts of seven provinces. This designation fol-lows the recommendations suggested by Statistics Canada to delineate the CanadianNorth (McNiven and Puderer 2000). Henceforth, these areas are referred in the paperas ‘the North’ (see Figure 1). In contrast, the southern regions of Canada around thelargest metropolitan centres are considered as the core (‘metropolis’) areas. They arereferred to here as ‘the South’.2

A classic Harris–Todaro modelTT

The initial Todaro (1969) model (TM) assumed that a decision to migrate from ruralto urban areas was dependent upon the ‘expected’ income maximization objective shared by all migrants. This thesis was not unique in Todaro’s theorization. Whatwas novel, however, was the second assumption that migrants, making a decisionto migrate, considered not only wage differentials, but also the probability of being employed in the city. Todaro stressed a ‘fundamental role of job opportunities andprobabilities of employment’ as crucial for the migration decision-making process (Todaro 1969: 140). With this notion, he was able to accommodate within the modelan observed paradoxical increase in labour pull into urban settlements with exist-ing high unemployment. Todaro postulates that ‘migration proceeds in response tourban–rural differences in expected rather than actual earnings’ (Todaro 1971: 364, hisemphasis). Here, expected income is a function of wage differential and probabilityof obtaining a job, which, in its turn, is contingent on a probability of being selected from a surplus labour force. According to Todaro, a rural migrant (who is assumed to be perfectly informed and pragmatic) weighs these probabilities against an alterna-tive of staying. It is implicit to the model that the planning horizon is uniform among all migrants and moving costs are identical.

Mathematically, the TM can be presented as the following (Todaro 1969, 1971). If V is the discounted present value of the expected net urban–rural stream over theVmigrant’s time horizon, YuY (t) is the average real income in the urban sector and YrY (t)in the rural sector, n is the number of time periods in the migrant’s planning horizonand r is the discount rate reflecting the degree of time preference, the decision tomigrate V is positive or negative after solving:V

n

V(0) = ∫ [p(t)YuY (t) – Yr(t)]e–rtdt – C(0), (1)

t=0

where C(0) is the cost of migration and p(t) represents the probability that the migrantcan find an urban job providing average income in period t. Further, Todaro defined the probability of being employed in the urban sector (p) as a function of the prob-ability π of ‘having been selected in that or any previous period from a given stock of unemployed or underemployed job seekers’ (Todaro 1971: 411). Hence, Todaro insisted that migrants evaluated their migration prospects using information about potential success in finding a job. Employment probability p(x) within x periods aftermigration can be formalized as

p(x) = p (x – 1) + [1 – p (x – 1)] π (x), (2)

while

Revising the Harris–Todaro Framework 189

π (x) = mN/[S – N] , (3)

so that m is the net rate of urban new job creation, S is total urban labour force andN is number of employed in urban sector. Therefore, under the model’s assumpN -tions, the rural to urban migration will occur until urban real wages, offset by theexpected probability of receiving them within a certain time window, will maintainthe urban–rural real wage differential.

The Harris–Todaro (HT) model is a slight modification of the initial Todaro model that accounts for trade in a two-sector system (urban and rural sectors) in order todetermine real wages and urban–rural wage differential (expressed in terms of man-ufactured goods) as well as welfare and distributional consequences of migration(Harris and Todaro 1970). The model uses traditional neoclassic mechanisms andbehavioural rules of factor use, output in each sector, terms of trade and so forth. Italso assumes the existence of a fixed minimum urban wage WuW fix and links between migrants and their families in the rural sector, as is frequently the case in LDCs.Since the migration is a disequilibrium phenomenon, Harris and Todaro analyse the conditions of equilibrium to find out under what circumstances migration ceases to continue. In particular, if

WrW = WuW N / S, and WuW > or = WuW fix , (4)

where WrW and WuW are rural and urban wages respectively, N and N S as denoted above (3) is the condition of migration termination; the increase in urban employment Nwill require adequate growth of rural wages WrW to sustain equilibrium. In reality, rural wages are relatively stable, and, thus, migration accelerates and, paradoxically, urban unemployment rises as a result of new manufacturing jobs creation; see the concept of induced migration (Todaro 1976).

Todaro (1969, 1971, 1976, 1980) and Harris and Todaro (1970) used a framework of a new model to draw some important public policy conclusions. They asserted that the cut in unemployment and the overall welfare improvements are possible, whenlimited wage subsidies correspond with migration restriction. In the long run, to reduce labour influx into the cities they suggested investing in the agricultural sectorand rural development.

Evaluating the relevance of the Harris–Todaro model in the Canadian northern TTfrontier

There are several important points to be raised when assessing the applicability of the HT framework to the Canadian context. First of all, similarly to the classic HT model, the out-migration from the North predominantly occurs toward core regions,with no high demand for extra labour, an abundant labour force and significantunemployment. In particular, the north-to-south migration has historically gravi-tated to large metropolitan areas of Edmonton, Calgary, Vancouver, Ottawa–Hull,and Toronto (Statistics Canada 2002b).

Secondly, and more importantly, the nature of the north-to-south migration resem-bles rural-to-urban migration in LDCs, while the character of produced goods andterms of trade are also similar to the two-sector urban–rural dichotomy. Northernproducts, which are predominantly raw materials and various crafts (analogous to

190 Andrey N. Petrov

‘agricultural’ goods in the HT model), tend to be valued below more sophisticatedmanufactured goods produced in the South. Consequently, southern goods in theNorth become extremely expensive both because of their higher comparative valueand because of substantial transport costs.3 Therefore, in addition to the price dif-ferential caused by the ‘nature’ of goods, assumed in the HT model, there is a dif-ferential determined by the geography of production. As a result, the real income of northern workers diminishes, if compared to the real income of their southern coun-terparts. Higher expenses required to support individuals and households in theNorth offset the large nominal wages some workers receive. This fact, coupled with the severe working environment and social well-being issues, forces many to con-sider out-migration. The frontier labour force is very unstable and prone to migrate(Heleniak 1999).

Similar to migration from rural hinterlands to major urban centres in the clas-sic HT model, the geography of north-to-south migration flows in Canada showsa bias toward large CMAs, with abundant and skilful labour supply. Expectedly,under strong competition, many migrants remain unemployed, whereas they had ajob in the North; in a comparable case in Russia, the unemployment of former north-ern residents after migrating is 23 per cent compared to less than 10 per cent before (Heleniak 1999). In Canada, the same is likely to be true, especially for the Native migrants. The unemployment rate of Aboriginal residents in some CMAs-migrationdestinations is, in fact, higher than in the North itself, where this rate is often inflated by underreporting employment in the traditional sector (Statistics Canada 2003a). According to the Aboriginal Peoples Survey public use microdata (Statistics Canada2006a), 31 per cent of Aboriginal migrants to CMA cited work as a reason to move. At the same time, 25 per cent of these economic migrants, who moved into metro areas in 1996–2001 and participated in the labour force, were unable to secure a full-timejob at destination. In comparison, for the non-migrants, that is the Indigenous popu-lation continuously living in CMAs, the reported unemployment was 19.5 per cent.

The aforementioned arguments suggest that the general framework of the HT theory could be potentially applicable in the North. However, there are important problems that have to be resolved. First of all, such assumptions as randomnessof job selection process, the utility-maximizing motivation of migration, the equalcompetitiveness of workers at the market, and the homogeneity of the labour force are not realistic in our case. More importantly, in contrast to the core Harris–Todaroassumption, the unemployment exists both in the origin (North) and the destination (South).4

Another limitation of the HT framework in this context is that it was designed to reflect only migration related to job-search. For many North-to-South migrants, espe-cially Aboriginal northerners, job opportunities are not a major reason to migrate to CMAs (Statistics Canada 2003a). Therefore, the model developed below will be limited in its explanatory power to labour migration: it will be explaining about one-third of the total out-migration flow.

Modifications of Harris–Todaro model for North-to-South migration in Canada

The most important issue for the HT model’s applicability is the heterogeneity of the northern labour force. It refers to substantial differences in migration behaviour and decision-making process. To address it, one may suggest developing two different

Revising the Harris–Todaro Framework 191

versions of the model for the two major groups of the northern labour force: one forthe Native labour force and another for the non-Native labour force.

The Canadian Native5 labour force is still a disadvantaged work force (Bone and Green 1984; Hull 2000; Kuhn and Sweetman 2002), in terms of its labour-market com-petitiveness. This labour force in its characteristics is principally similar to the rural population in LDCs, for which the HT model was originally designed. On the other hand, unemployment related to migration to large urban centres (i.e. the Todaroparadox), is also evident (Statistics Canada 2003a). Given these similarities, the HTmodel may be accurate in portraying the labour migration of Indigenous northerners to the South.

Although existing studies demonstrate that wage per se is not the major factor in Native migration to the southern urban centres (Clatworthy 1996; Norris et al. 2004b), more detailed qualitative inquiries suggested that it was a ‘search of bet-ter opportunities’ that motivated most Indigenous people to move into Canadian CMAs (Cooke and Bélanger 2006). Employment and higher wages were among those expected opportunities.

In contrast to the HT assumption about full employment in the rural sector, this is not the case for the northern Aboriginal labour. Although the definition of unem-ployment for this group is problematic because of the involvement in the traditional economy, the official unemployment rates are very high, indicating the low propen-sity to find a modern-sector job at home (Hull 2000).6 Many Native Canadians sufferunemployment and resulting difficult socio-economic conditions that play an impor-tant role in the decision to migrate to urban centres (Trovato, Romaniuc and Addai1994). To make the analysis more accurate, it could be important to include this factin the model.

It has also been argued in the previous section that in a two sector-two regionNorth–South framework, in addition to the price differential caused by the uneventerms of trade already assumed in the HT model, there is another cost differential,determined by high transport costs. As a result, the real income of northern workersdiminishes, if compared to the real income of their southern counterparts. Therefore,a coefficient should be introduced to account for this cost differential.

With these adjustments in mind, the wage-driven Aboriginal out-migration from the North may be successfully described using general HT equilibrium conditionswith minor adjustments:

W’n = ϕp’W’sNs / ρ’Ss , (5)

where W’n denotes wage of Indigenous peoples in the North, W’s in the South, NsNand Ss are employment and labour force in the South, ϕ is a coefficient of extra costs associated with living in the North, p’ is an adjustment coefficient for possibility of securing a job for Native workers (for example, it may be derived as a ratio between the Native and average unemployment rates) and ρ’ is a coefficient reflecting pos-sibility of obtaining a modern-sector job in the origin; ρ’ is analogous to π in eq. (3).Therefore, the decision to migrate will depend on solving the following expression.Y’s(t) and Y’n(t) are real incomes for Indigenous northerners, the rest as in (1), and migration will occur if V(Ind) > 0: n

V(Ind) = ∫ [p’(t)Y’s(t) – ρ’(t)–1ϕ (t)–1Y’n(t) ] e–rtdt – C. (6) t=0

192 Andrey N. Petrov

In future it may be useful to include some variables characterizing comfort, remote-ness, development of social infrastructure, size of community and its Native popula-tion. However, this is beyond the limits of the current pilot study.

The first step in modelling the non-Aboriginal north-to-south migration usingthe HT framework is to reject the assumption about the absence of unemploymentat the origin. In other words, the modified model should consider the probabilityof losing a job in the origin (in addition to the probability of obtaining a job in bothorigin and destination) among the factors in migration. The important differencebetween Native and non-Native migrants would be that owing to the existing tra-ditional economy, and community and government support, a job loss for potentialNative migrants is unlikely to be a defining factor in out-migration. In contrast, as has been pointed out in many studies, a push factor related to job loss, or other job circumstances, is the major reason for non-Indigenous residents to leave the Arctic(Heleniak 1999). Hence, in migration decisions made by non-Aboriginal northerners,the propensity to lose a job in the origin will be no less important than the possibility of getting one in the destination. In the World Bank study of Russia, 27 per cent of migrants (86 per cent of whom were not born in the North) cited wage-related issues as a reason to out-migrate, while only six per cent mentioned the opportunity of ajob in the South as such a reason (Heleniak 1999). Probability of losing a job (b) in adeclining economy7 within x periods of time can be determined similarly to prob-ability of obtaining a job:

b(x) = b(x – 1) + [1 – b(x – 1)]μ(x), (7)

while

μ = nN /[S – N ] , (8)

so that n is the net rate of urban job loss, S the total urban labour force and N theNnumber of employed in the urban sector. A higher probability of losing employmentmakes the worker’s wage expectations in the South more modest.

Incorporating the job loss probability will give the following form to the HT equi-librium equation for non-Native migrants from the North:

WnW = ϕb–1WsW Ns/ρSs , (9)

in which WnW denotes the wage of non-Aboriginal workers in the North, WsW in theSouth, NsN and Ss are employment and labour force in the South, ϕ is a coefficient of extra costs associated with living in the North and ρ is a possibility of being selectedfor a new job at the origin. The migration decision expression with the outlined mod-ifications will look thus (see variables’ denotations in the text):

n

V(Non-Ind) = ∫ [p(t)Ys(t) – ρ(t)–1ϕ(t)–1b(t)–1Yn(t)]e–rtdt – C. (10) t=0

The introduced modifications to the classic HT framework produce two distinct mod-els of the frontier–metropolis migration for Aboriginal and non-Aboriginal workers.This suggests that two distinct groups of northerners will demonstrate somewhatdissimilar mechanisms and patterns of labour migration to the South. It is interesting

Revising the Harris–Todaro Framework 193

that additional variables incorporated into both models encourage the out-migra-tion from the North: a probability of losing a job and a living cost differential may bestrong push-factors from the North. The modified HT model implies that the act of migration is even more likely to occur in the North–South direction than in the classic model. Certainly, it is a disturbing observation for those who are concerned with the labour force drain from northern communities.

Harris–Todaro model for north-to-south migration in Canada: an application to TTAboriginal labour migration

This section offers an empirical exercise, which attempts to estimate the likelihood of Indigenous workers migrating from selected northern communities into Canadian metropolitan areas, using the modified HT model, (6) redefined in equilibrium terms.The example concerns only Aboriginal migrants in selected communities (explanation of the selection process below) and, as always when using the HT model, describesonly the wage-driven labour migration. While the definition of indigeneity is itself problematic, and the data on Indigenous movers are limited, the Native people are the most easily ‘traceable’ movers in Canadian statistics, because they are surveyedapart from the rest of the population and in more detail. Although most Native peo-ple are mostly local movers, their propensity to out-migrate to other areas, includingCMAs, is slightly higher than for the non-Aboriginal population (Frideres, Kalbachand Kalbach 2003). Only the Inuit are less mobile than both other Indigenous andnon-Native Canadians. Table 1 presents evidence that most Aboriginal residents inmajor CMAs have in-migrated there.8

To build the model, it has to be assumed that Indigenous northerners follow the migration pattern set for all Indigenous Canadians to migrate to large urban centres.

Table 1TT Selected socio-economic characteristics of Aboriginal population in major metropolitan areas (2001)

CensusMetropolitanAreas (CMAs)

Unemploymentrate, %

λ

In-migrated,a%

Work—major reasonto move-in,

%

Average incomefor employedfull-time, $C

W’s

Calgary 10.10 82.00 40.00 35067.00Edmonton 13.10 77.00 32.00 34227.00Montreal 12.20 64.00 26.00 31928.00Ottawa 7.30 75.00 28.00 39899.00Saskatoon 22.30 77.00 25.00 30949.00Toronto 8.60 71.00 41.00 41321.00Vancouver 15.30 76.00 22.00 37162.00Winnipeg 14.30 62.00 26.00 29239.00

a In-migrants: per cent of adults who lived in another city, town or community at any time before 2001.

Sources: Statistics Canada 2002a, 2003a.

194 Andrey N. Petrov

Such an argument is plausible, because the Census indicates that the main CanadianCMAs served as primary destinations for migrants from the North and for Indig-enous migrants across the country (Norris et al. 2004b). Since the case in point dealswith small and remote northern communities, it is important to note that Aboriginalcommunities furthest from urban centres tend to have higher out-migration ratesthan more proximate ones, largely because cities provide, or are perceived to pro-vide, much better access to services and employment than remote towns and vil-lages.9 At the same time, the return migration to distant communities is inhibitedby high moving costs (Norris et al. 2004a). Remote communities with small popula-tions also experience higher out-migration rates than do larger ones. Offsetting this,however, communities with high participation in traditional economic and culturalactivities are less prone to out-migration (Norris et al. 2004a).

Various research (Statistics Canada 2003a; Norris et al. 2004a, b) shows thatemployment opportunities play a limited role in Native out-migration to large urbancentres (see Table 1), yielding, first of all, to family-related reasons. In the AboriginalPeoples survey microdata sample, 31.3 per cent of Aboriginal in-migrants to CMA cited work as a reason to move (Statistics Canada 2006a). Thus, from the outset, it isnecessary to emphasize first that labour-force considerations were the main reason to migrate only for 20 to 40 per cent of Native movers, and second that the role of employment factors in migration decision-making was more complex than in theclassic HT framework. Thus, one would expect the explanatory power of the HT model to be within a 20–40 per cent margin. In addition, employment propensitieshave to be considered for both origins and destinations; see equations (5) and (6).

Developing the dataset

An empirical component of this paper is based on labour-market indicators andmigration patterns in northern regions of Canada, obtained from the 2001 CanadianCensus and Aboriginal Peoples Survey (Statistics Canada 1998, 2002a, 2003a, b). Another source includes territorial statistical bureau data on population change and its components (NT Bureau of Statistics 2005a, b). These datasets provide information on the number of residents, their mobility (net migration and in-migration: both arenot directly provided, but are computable from the data), unemployment rates, andfull-time employment wages for both Aboriginal and non-Aboriginal populations.Ancillary data include the ‘isolated posts’ database (Treasury Board 2006), whichcontains cost differentials, and GIS files with geocoded community locations.

The empirical exercise requires selecting appropriate sets of origins and destina-tions, that is, core and frontier communities. There have been multiple attempts todelineate core and peripheral areas in Canada (Bourne 2000; Simmons and McCann2001). Most sources contrast southern densely urbanized parts of Canada with north-ern small town and rural peripheries. However, not all of the largest metropolitan areas are equally important destinations for the inbound migration from the Cana-dian North. Interprovincial migration data show that just a few southern urban cen-tres served as dominating hubs. In fact, seven metropolitan areas concentrated 31.9 per cent of all out-migrants from the Canadian Territories between 2000 and 2003(Statistics Canada 2006c). These urban centres were, in descending order of impor-tance, Edmonton, Vancouver, Calgary, Toronto, Winnipeg, Saskatoon and Montreal.They concentrated 86.4 per cent of territorial migrants coming to large CMAs. Indeed, most of these cities are also traditionally considered as the cornerstones of the Cana-

Revising the Harris–Todaro Framework 195

dian metropolis (Bourne 2000). Saskatoon was also a major destination of northern labour migration, including Indigenous (e.g. Statistics Canada 2002b; Norris et al.2004b; Cooke and Bélanger 2006). Therefore, based on available empirical indicationsand literature recommendations, selected migration destinations, exemplifying theCanadian urban core, included metropolitan areas of Toronto, Montreal, Vancouver,Ottawa, Calgary, Edmonton, Winnipeg and Saskatoon (Figure 1).

To create a sample of northern frontier communities (origins), the study selected populated places, designated by the Canadian government as ‘isolated posts’ (Treas-ury Board 2006). The ‘isolated posts’ database includes remotely located settlements,in which government employees are eligible for additional allowances and otherentitlements, based on local living and transport costs, remoteness, accessibility, pop-ulation size and environmental (climatic) conditions: all ‘isolated posts’ by definition are communities with distinct frontier characteristics. Thus, the list of ‘isolated posts’ constitutes a convenient starting point for creating a sample of ‘frontier’ settlements. However, in the current exercise, two other considerations are crucial: selected set-tlements should have substantial Indigenous populations to be representative forAboriginal migration, and communities in the final sample should be located in theCanadian North.

A number of database manipulations and GIS procedures were necessary to pre-pare the dataset satisfying these criteria. The final set consisted of 57 frontier com-munities (see Figure 1 and Table 2), which satisfied three fundamental criteria: they were isolated, had substantial Indigenous population (over 200) and had northernlocation, that is, they belonged to 20 northern census divisions. In 52 of them, the Indigenous population was over half of the total population.

By design, the HT model compares expected wages at destination with existingreal wages at origin. Therefore, to operationalize the model, it is necessary to iden-tify pairs of migration origins and destinations. Theoretically, migration can occurbetween any frontier community and any metropolitan area in our samples. How-ever, scores of migration studies suggest that moves are generally more likely tooccur between closer locations. The pattern of North-to-South migration in Canada, indeed, is more complex than this simplistic assumption. At the same time, within the limits of the current test, focused on demonstrating the applicability of the HT model, there is no need to create additional complexities, and it appears plausible to use the factor of distance in matching pairs of origins and destinations. In the GISenvironment, each frontier community (origin) from the sample was matched withthe closest metropolitan area (destination). Thus, the closest metropolitan area wasdesignated to represent characteristics of the Canadian ‘core’ regions in each particu-lar case.

Model specification

The model (5) was operationalized using ‘threshold expected wages’ (TEW) and‘threshold origin wages’ (TOW). An assessment of TEW and TOW was used forestimating the likelihood of north-to-south labour migration. TEW is an expectedsouthern wage level at which the HT model reaches the equilibrium (i.e. the startingpoint for migration), whereas TOW represents the minimal level of northern wagerequired to prevent migration:

W’n = ϕp’W’s(1 – λ s)/ρ’; W’s = TEW,W W’n = TOW. (5’)

196 Andrey N. Petrov

Table 2TT Selected socio-economic characteristics of Aboriginal population in sampled Northern settlements

Northerncommunitya

Unem-ploymentrate, %b

λn

Average income for employed full-time, $Cc

W’n

Living costs

differentiald

φ

Nearestlarge

metropolitanarea

Aklavik, NT 27 39327.00 1.57 VancouverArctic Bay, NU 22 34827.00 1.72 WinnipegArviat, NU 21 34721.00 1.67 WinnipegBaker Lake, NU 29 34461.00 1.77 WinnipegBeauval, SK 21 35596.00 1.17 SaskatoonBuffalo Narrows, SK 23 32203.00 1.22 SaskatoonCambridge Bay, NU 20 43485.00 1.57 SaskatoonCanoe Lake, SK 27 26845.00 1.17 SaskatoonCape Dorset, NU 27 26918.00 1.62 OttawaChemawawin, MB 39 24972.00 1.27 WinnipegChesterfield Inlet, NU 18 31412.00 1.72 WinnipegChisasibi, QC 16 32985.00 1.22 OttawaClyde River, NU 28 34506.00 1.67 MontrealCoral Harbour, NU 16 35160.00 1.77 WinnipegFort Good Hope, NT 18 37925.00 1.62 VancouverFort Hope, ON 36 27300.00 1.57 WinnipegFort Liard, NT 22 31970.00 1.32 VancouverFort McPherson, NT 23 38090.00 1.52 VancouverFort Nelson, BC 25 48886.00 1.22 VancouverFort Providence, NT 23 30732.00 1.32 EdmontonFort Resolution, NT 21 35910.00 1.37 EdmontonFort Simpson, NT 19 41672.00 1.37 EdmontonFort Smith, NT 17 44398.00 1.27 EdmontonGjoa Haven, NU 28 37271.00 1.62 WinnipegHall Beach, NU 31 31377.00 1.62 OttawaHay River, NT 17 44719.00 1.37 EdmontonIgloolik, NU 31 35672.00 1.62 OttawaInukjuak, QC 16 26137.00 1.52 OttawaInuvik, NT 11 43710.00 1.47 VancouverIqaluit, NU 18 46116.00 1.57 OttawaIvujivik, QC 20 25425.00 1.62 MontrealJohn d’Or, AB 37 26460.00 1.22 EdmontonKimmirut, NU 21 29086.00 1.62 MontrealKugaaruk, NU 13 35336.00 1.82 WinnipegKugluktuk, NU 26 39400.00 1.57 EdmontonKuujjuaq, QC 9 39279.00 1.57 Montreal

Revising the Harris–Todaro Framework 197

This model is a slight modification of the conceptual model (5), where λs is a south-ern unemployment rate; for other variables see equation (5). To further specify themodel, one can represent ρ’, a coefficient reflecting possibility of obtaining a modern-sector job in the origin that equals one in the traditional HT model (full employment),as a function of the unemployment rate at the origin (λn). This allows replacing ρ’with (1 –λn) to reflect the local labour-market conditions: p’ in (5’) stands for the prob-ability of Indigenous workers being employed at destination as opposed to all other groups of jobseekers. As discussed above, Aboriginal workers in Canada are typically in a disadvantaged position in the labour market. In essence, p’ can be calculated as a ratio between Aboriginal and non-Aboriginal unemployment. However, this would

Table 2TT (continued)

Northerncommunitya

Unem-ploymentrate, %b

λn

Average income for employed full-time, $Cc

W’n

Living costs

differentiald

φ

Nearestlarge

metropolitanarea

Kuujjuarapik, QC 16 26154.00 1.42 OttawaLa Loche, SK 51 26462.00 1.22 EdmontonMakkovik, NF 23 31143.00 1.27 MontrealNorway House, MB 28 34082.00 1.22 WinnipegOxford House, MB 29 28071.00 1.47 WinnipegPangnirtung, NU 26 38344.00 1.62 MontrealPinehouse, SK 36 27479.00 1.22 SaskatoonPond Inlet, NU 26 38733.00 1.67 MontrealPort Hope Simpson, NF 50 30106.00 1.17 MontrealPuvirnituq, QC 15 30843.00 1.52 OttawaQikiqtarjuaq, NU 27 34924.00 1.67 MontrealRankin Inlet, NU 19 41486.00 1.72 WinnipegRepulse Bay, NU 30 29674.00 1.77 WinnipegSandy Bay, SK 26 24006.00 1.22 WinnipegSandy Lake, ON 21 25467.00 1.62 WinnipegSanikiluaq, NU 19 34655.00 1.62 OttawaTaloyoak, NU 29 30111.00 1.67 WinnipegTuktoyaktuk, NT 29 45545.00 1.62 VancouverWhale Cove, NU 19 34268.00 1.82 WinnipegWhitehorse, YT 26 34767.00 1.22 VancouverYellowknife, NT 12 49463.00 1.22 Edmonton

a AB: Alberta; MB: Manitoba; NF: Newfoundland and Labrador; NU: Nunavut; NT: NorthwestTerritories; ON: Ontario; QC: Quebec; SK: Saskatchewan; YT: Yukon.Sources: b Aboriginal Peoples Survey 2001 (Statistics Canada 2003a); c Census Aboriginal Pop-ulation Profile 2001 (Statistics Canada 2003b); d Isolated Posts and Government Housing Direc-tive (Treasury Board 2006).

198 Andrey N. Petrov

be necessary only if λs is derived from the overall unemployment in the South, assuggested in (5’). If, instead, one uses the Aboriginal-specific unemployment rate,there will be no need to introduce p’, because now λs would already reflect the actualpropensity of Indigenous jobseekers to be employed. Finally, to be consistent withthe HT framework, the model uses average wages for persons employed full-timetwelve months a year as the entries for W’s and W’n. The latter is adjusted by ϕ (liv-ing cost differential) to approximate a real wage compared to earnings in the South.All variables are presented in Tables 1 and 2. The unemployment figures are takenfrom the Aboriginal Population Survey of 2001, wage figures are from the AboriginalCensus of 2001 and living costs differentials are adopted from the latest edition of theIsolated Posts and Governmental Housing Directive (effective 1 April 2003). The finalequation looks as follows:

W’n = ϕW’s (1 – λs)/(1 – λn); W’s = TEW,W W’n = TOW (5’’)

where λs and λn are Aboriginal unemployment rates at origin (North) and destination (South).

Table 3 offers a resulting matrix that includes the ratio (K) between real wagesin the North (W’n) and TOW for Native migrants in respect to selected destinations.According to the HT hypothesis, migration is likely to occur in cases where this ratio is less than one,10 that is, when

K = W’n /TOW and 0 < W K < 1. (11)

As is easily seen from Table 3, the outflow from the North is expected to exist. Note that smaller values of the coefficient K correspond to the greater likelihood of migratK -ing from northern frontier settlements to the nearest major metropolitan area; ori-gin–destination pairs are as specified in Table 2.

The most striking observation from Table 3 is, indeed, that among 57 northerncommunities there was only one, the Northwest Territories’ capital of Yellowknife,in which Indigenous jobseekers were not likely to out-migrate to the nearest CMA. Another highlight of Table 3 is a great diversity in labour migration probabilitiesamong northern frontier communities. This suggests that it is, perhaps, overly sim-plistic to talk about ‘the frontier’, whereas it is more appropriate to identify multiplefrontiers with distinct socio-economic and labour migration patterns.

Before any further discussion, it is necessary to ensure the plausibility of the HTmodel results. For this purpose, it is useful to analyse the association between pre-dicted out-migration probabilities and actual out-migration pools. Unfortunately,municipal-level migration cross-tabulations between origins and destinations speci-fied in this study are not available. Under such circumstances, the origin-based out-migration data must be used as the best proxy to measure existing migration flows from frontier communities.

Out-migration from the given frontier northern communities was estimated for the five-year period 1996–2001 based on Statistics Canada and territorial statisticalbureau data on total population change, number of births and deaths and mobility(Statistics Canada 1998, 2002a; NT Bureau of Statistics 2005a, b). Using the five-yearperiod was a convenient way to eliminate yearly fluctuations. However, these data did not differentiate between Aboriginal and non-Aboriginal populations; there-

Revising the Harris–Todaro Framework 199

fore, a further task was to adjust the derived out-migration estimates. First, to makea first-cut proxy of the Aboriginal out-migration flow, the total out-migration fig-ures were adjusted to reflect the share of Aboriginal population in each communityand weighted by the average differential propensity of Aboriginal northerners toout-migrate to CMAs, derived from Census 2001 public-use microdata files (Statis-tics Canada 2006b). Secondly, a special coefficient reflecting the overall populationmobility was introduced to eliminate the effect of the differential mobility among communities.

The analysis of associations between the final out-migration estimates and mod-elled migration likelihoods from Table 3 (coefficient K) indicates that the HT modelproduces quite plausible results. The correlation coefficient between these two vari-ables is –0.5 (significant at 0.01 level) suggesting that greater K values are associatedKwith smaller Aboriginal out-migration. The negative relationship here is in complete compliance with the conceptual model, in which the greater values of K correspond with the lower probability of labour migration to the core. Thus, the correlationcoefficient indicates an expected relationship. A bivariate linear regression model of Aboriginal out-migration, as expected, shows that the explanatory power of K is atKthe 25.0 per cent level (r2= 0.25). The latter figure is within the expected explanatoryrange (20–40 per cent) suggested for the HT-type model earlier in this paper. This

Table 3 Threshold origin wage coefficients (K) for selected communitiesa

Name K Name K Name K

Aklavik, NT 0.58 Fort Providence, NT 0.60 Makkovik, NL 0.70Arctic Bay, NU 0.57 Fort Resolution, NT 0.70 Norway House, MB 0.90Arviat, NU 0.72 Fort Simpson, NT 0.83 Oxford House, MB 0.54Baker Lake, NU 0.47 Fort Smith, NT 0.98 Pangnirtung, NU 0.65Beauval, SK 0.97 Gjoa Haven, NU 0.65 Pinehouse, SK 0.68Buffalo Narrows, SK 0.70 Hall Beach, NU 0.36 Pond Inlet, NU 0.63Cambridge Bay, NU 0.91 Hay River, NT 0.65 Port Hope Simpson, NL 0.47Canoe Lake, SK 0.72 Igloolik, NU 0.41 Puvirnituq, QC 0.47Cape Dorset, NU 0.33 Inukjuak, QC 0.39 Qikiqtarjuaq, NU 0.56Chemawawin, MB 0.61 Inuvik, NT 0.84 Rankin Inlet, NU 0.69Chesterfield Inlet, NU 0.58 Iqaluit, NU 0.70 Repulse Bay, NU 0.67Chisasibi, QC 0.61 Ivujivik, QC 0.46 Sandy Bay, SK 0.64Clyde River, NU 0.55 John d’Or, AB 0.46 Sandy Lake, ON 0.53Coral Harbour, NU 0.51 Kimmirut, NU 0.52 Sanikiluaq, NU 0.47Fort Good Hope, NT 0.61 Kugaaruk, NU 0.57 Taloyoak, NU 0.57Fort Hope, ON 0.49 Kugluktuk, NU 0.62 Tuktoyaktuk, NT 0.63Fort Liard, NT 0.60 Kuujjuaq, QC 0.84 Whale Cove, NU 0.62Fort McPherson, NT 0.61 Kuujjuarapik, QC 0.42 Whitehorse, YT 0.67Fort Nelson, BC 0.95 La Loche, SK 0.36 Yellowknife, NT 1.20

a Higher coefficient (K) translates to lower likelihood of labour migration.

200 Andrey N. Petrov

first-cut statistical test appears to confirm the plausibility of the HT model resultsand, therefore, indicated that the HT framework is likely to be useful in the fron-tier–metropolis context in Canada.

To investigate the behaviour of major model components (5’) further, it is possibleto fit an empirical model that is a descriptive supplement to estimating the prob-ability of migration directly from the equation (5’). Such models in conjunction withutility functions were previously found useful in the immigration literature (Han and Ibbot 2005). Similar to the HT framework, this empirical model assumes theimportance of the differential in expected real wage between origins and destina-tions, that is, the expected earning premium, for shaping migration. Typically, thesedescriptive models are operationalized using the log-transformed regression form that allows estimating coefficients associated with each factor of migration. In thepresent case, the empirical model that describes the same behaviour as model (5’)would look like:

lnM = β0+ β

1lnw + β

2lnλ + β

3lnφ + ε (12)

where M is an estimated out-migration coefficient from northern communities, w = WnW / WsW (nominal wage differential) and λ= λn / λs ,β estimated coefficients; othernotations as in (5’’).

Parameters in this model are estimated using available out-migration proxies andinputs of the initial HT model (Table 4). R2 of the model is 0.28, again reflecting themodest role of labour migration in the total migration flow from northern communi-ties. The resulting regression model satisfied the criteria of statistical significance (at0.01 level) and produced coefficients with theoretically expected signs.

The nominal wage differential (here measured as a ratio between the North and southern CMAs, since unadjusted nominal wages in the North are higher than in thesouth) was a ‘pull’ factor mostly constraining the out-migration. Meanwhile, both the unemployment rate and cost differentials served as ‘push’ factors that encour-aged out-migration. However, the unemployment rate differential turned out to be statistically not significant. This empirical result is, overall, well-aligned with the HTlogic. From expression (5’’) one can expect similar signs and slopes.

Table 4 Empirical model coefficientsa

Unstandardized Coefficients

Standardized Coefficients

t Sig. B Std. Error Beta

Constant –0.767 0.284 –2.702 0.009

ln w –0.879 0.403 –0.273 –2.178 0.034

ln φ 2.292 0.563 0.506 4.068 0.000

ln λ 0.264 0.241 0.136 1.095 0.279

a For notations see text.

Revising the Harris–Todaro Framework 201

From the public policy perspective it is notable that, given the model’s coeffi-cients, the cost differential appeared to be the most influential factor among inde-pendent variables, whereas both indicators that estimated the real wage gradient(nominal wage w and unemployment λ differentials) had rather modest slopes. Thismay suggest that the severity of living conditions is still among the leading factors inout-migrating from the North. Even the nominal wage rises and availability of jobscannot guarantee the stability of labour supply in northern regions. This fact marksa substantial departure from the traditional HT framework, where living costs arenot a part of the equation and not considered necessary to describe migration prob-abilities. In contrast, in remote and isolated areas in the North, potential migrantswould be likely to integrate cost differentials when determining possible real wage‘premiums’ resulted from migrating. Nevertheless, the additional cost-of-living vari-able, as has been argued earlier in the paper, is in line with the overall behaviourallogic of the HT model.

According to Table 3, the HT model predicts the strongest migration pull fromless prosperous and more peripheral northern communities to major metropolitanareas. The distinction between the NT capital of Yellowknife and the rest of the Ter-ritory reflects internal inequalities in the North. A similar pattern is noticeable inNunavut, where Iqaluit is somewhat distinct from other isolated posts in that Terri-tory. The same is true for other regional centres in the Territories, such as Inuvik and Fort Smith. Overall, these results clearly indicate that the North is likely to experi-ence the ‘drain’ of Indigenous labour force as far as wage-economy occupations areconcerned, despite the high unemployment in the South. In effect, the most economi-cally active and most qualified Native jobseekers will tend to look for employment inthe core rather than in northern communities, given the real-wage expectations theymay have. However, as in the classic Harris–Todaro case, there is no guarantee thatsouthern jobs are readily obtainable, especially considering a substantial skills gapbetween Aboriginal workers and southerners (Hull 2000). As a result, new migrantsjoin scores of unemployed Native job-seekers and, in turn, increase the dependency and welfare burden in large metropolitan areas in the south, not to mention a widespectrum of social issues.

It should be realized that the migration probabilities predicted in this exerciseby the HT-type model, despite passing the plausibility test, are based on substantialapproximations and tested using aggregate data. In addition, the heterogeneity of the Aboriginal workforce, whether related to age, occupation, marital status, educa-tion or any other quality, implies a great variability of wage expectations within the considered sample. Results in Table 3 reflect only the general tendency of an ‘aver-age’ person to migrate in a limited set of destination options, but, of course, does notexplain individual behaviour.

Discussion and conclusions

Originally, the HT model was developed as a new approach to policies of welfare-maximization. It showed that new urban jobs eventually induce urban unemploy-ment: the Todaro paradox. As a solution to this problem, Todaro (1969, 1976, 1980, 1995) promoted development of rural areas toward closing the urban–rural wagegap. In the context of the Canadian northern frontier, this approach can be useful aswell. Only keeping and creating jobs in the region can stabilize population dynamics

202 Andrey N. Petrov

and, simultaneously, decrease in-migrants’ unemployment, and thus a welfare bur-den, in large cities. On the other hand, the demand for labour in the North is unstableboth in time and in space. The job-creating capacity of northern economy is limitedand has been uncertain over the last decades. Northern labour loss, to some extent, isan inevitable result of failed economic development practices overlapping with cycli-cal socio-economic crises. However, if we consider any welfare policies, this analysisdemonstrates, as did Todaro’s original work, that such policies should concentrateon investments into sustainable economic activities that may provide long-term localemployment in the frontier both for Aboriginal and non-Aboriginal northerners.

Although there is no doubt that the HT method has been successfully used inmany studies devoted to developing countries, its crucial assumptions are problem-atically adoptable for developed economies.11 Among the most difficult issues couldbe assumptions of an abundant rural labour supply, capital immobility, a simplis-tic way of assessing the expected urban wage, equality of employment opportuni-ties and their unskilled nature. However, the second section argued that the frontierregions such as the Canadian North in many respects demonstrate apparent similar-ity to LDCs.

Overall, the HT model was found to be fundamentally applicable in the caseof North-to-South migration in Canada. However, as it is purely wage-driven, in the present form, it is unable to accommodate the variation of wages in differentsectors of the northern economy, high moving costs and severe climate conditions.These properties may have a significant impact on migration decision-making and may have to be included in the model. Besides, the relevance of the model can bequestioned considering the non-economic bias in Aboriginal mobility. However, theHarris–Todaro model is attractive in its useful approach to modelling income expec-tations as a migration factor, and succeeds in explaining about a quarter of the Abo-riginal migration flow to the South.

Cooke and Bélanger (2006) recently suggested that most statistical models havelittle success in predicting migration for Aboriginal Canadians as a function of humancapital and community-based factors, because these models tend to treat each movein isolation. Although the HT model partly shares this problem by isolating a specificcomponent of migration, it still provides a more advantageous framework of analy-sis by modelling migration behaviour rather than relying on formal population andcommunity characteristics. This makes the HT approach a desirable component indeveloping a complex model of Indigenous population migration in Canada (and,perhaps, in other countries with extensive frontiers), which is yet to emerge.

The welfare and planning component of the HT hypothesis was not in the focusof this paper. However, its results seem to have important public policy implica-tions. First, it is evident that clearer understanding of migration mechanisms in theNorth could provide a better basis for developing planning strategies aiming to sustain human capital in the Canadian North, and for optimizing welfare policiesboth in the North and in the South. Secondly, the results of this analysis suggest, like Todaro’s recommendations for LDCs, that investing in new economic opportunitiesin the frontier may prove to be more efficient than financing costly welfare provi-sions such as social assistance and housing to Indigenous migrants in southern cities.This aspect requires further study, especially in view of Ottawa’s renewed interestin developing the northern frontier and its populations. This research also provides

Revising the Harris–Todaro Framework 203

a potentially useful insight for countries with similar development issues, including Australia and Russia.

Acknowledgments

The author is grateful to John R. Miron for his advice and interesting discussions, andto the anonymous reviewers for their comments on the earlier version of this paper.

Notes

1 It has been pointed out that the universal validity of the ‘paradox’ is not firmly established,since there have been studies that questioned its existence (see Allen 2001).

2 This study focuses on the metropolitan South, which concentrates no less than one-third of migration flows from the North (Statistics Canada 2002; Norris et al. 2004; Cooke andBélanger 2006). The metropolitan South (the Canadian core) is the most natural match forthe assumptions of the HT model.

3 Note that transport of northern goods to the South is much less costly owing to economies of scale available for transporting predominantly bulk cargo (raw materials).

4 Unemployment is usually higher for Native northerners (Hull 2000; Bone 2003). At thesame time, their frequent involvement in traditional economies challenges the very defini-tion of being jobless. Trapping and other traditional activities provide Native people withsome necessities, but yield little income. Another important source of income for this groupis government transfers (see Bone 2003).

5 In Canada, the terms ‘Native’, ‘Aboriginal’ and ‘Indigenous’ are used interchangeably todescribe people, whose ancestors lived in North America before European colonization.Statistics Canada’s data on this group of Canadians, cited in this paper, refer to ‘peoplewith Aboriginal identity’, which includes all individuals who reported identifying with atleast one Aboriginal group, being Treaty Indians or Registered Indians, or being membersof an Indian Band or First Nation.

6 At the same time, not all Indigenous Canadians are engaged in traditional economic activ-ities. According to the Aboriginal Peoples Survey of 2001 (Statistics Canada 2003a), thepercentage of adults who hunted in the preceding 12 months was 58 in Nunavut, 43 inNorthwest Territories and 42 in Yukon. These figures were even lower in the provinces,except for Nunavik in northern Quebec.

7 We consider a case of continuously declining economy, when the propensity to lose job increases in time, as it is the relevant approximation of the real situation in many parts of the North during resource busts. Here, it is assumed that if job demand is growing, con-siderations of losing a job will overall be insufficient in migration decisions. Job turnover(without the net job loss) is understood to be a function of factors other than wage differ-ential.

8 Obviously, the bulk of these in-migrants arrived from nearby reserves or small Native com-munities as well as from other CMAs (Hull 2000).

9 However, in terms of volume, most Indigenous migrants to CMAs come from nearby or moderately distant reserves.

10 Conversely, migration would be likely to occur in cases where existing southern wages exceed TEW.

11 According to Blanchflower and Osvald (1994), these economies are more likely to be char-acterized by the ‘wage curve’. See discussion in Partridge and Rickman (1997).

204 Andrey N. Petrov

References

Allen, J. 2001. The state of the art in modeling migration in the LDCs: a comment. Journal of Regional Science 41(3): 521–528.

Barry, F. 2003. Economic integration and convergence processes in the EU cohesion countries. Journal of Common Market Studies 41(5): 897–921.

Blanchflower, D. and A. Osvald. 1994. The Wage Curve. Cambridge MA: MIT Press.Bone, R.M. 2003. The Geography of the Canadian North: Issues and Challenges. Second edition.

Toronto: Oxford University Press.Bone, R. M. and M. Green. 1984 The northern Native labour force: a disadvantaged work force.

Operational Geographer 3: 12–14.Bourne, L.S. 2000. Living on the edge: conditions of marginality in the Canadian urban system.

Pp. 77–97 in H. Lithwick and Y. Gradus (eds.), Developing Frontier Cities. Global Perspective – Regional Contexts. Boston MA: Kluwer Academic Publishers.

Clatworthy, S.J. 1996. The Migration and Mobility Patterns of Canada’s Aboriginal Population.Ottawa: Canada Mortgage and Housing Corporation, and the Royal Commission on Abo-riginal Peoples.

Cooke , M. and D. Bélanger. 2006. Migration theories and First Nations mobility: towards asystems perspective. Canadian Review of Sociology and Anthropology 43(2): 151–164.

Cordon, W.M. and Findlay, R. 1975. Urban employment, intersectoral capital mobility anddevelopment policy. Economica 42: 59–78.

De Santis, R. A. 2003. The impact of a customs union with the European Union on internalmigration in Turkey. Journal of Regional Science 43(2): 349–372.

Eaton, B.C. and P.R.Neher. 1975. Unemployment, undereployment, and optimal job search.Journal of Political Economy 83: 335–375.

Fields, G. 1975. Rural–urban migration, urban unemployment and underdevelopment, andjob search activity in LDC’s. Journal of Development Economics 2(2): 165–188.

Frideres, S., M. Kalbach and W. Kalbach. 2003. Government policy and the spatial redistribu-tion of Canada’s Aboriginal peoples. Pp. 94–114 in J. Taylor and M. Bell (eds), PopulationMobility and Indigenous Peoples in Australasia and North America. New York: Taylor and Fran-cis.

Gupta, M.R. 1994. Foreign capital, income inequality and welfare in a Harris–Todaro model.Journal of Development Economics 45(2): 407–414.

Han, J.D. and P. Ibbot. 2005. Korean migration to North America: some prices that matter.Canadian Studies in Population 32(2): 155–174.

Harris, J. and M.P. Todaro. 1970. Migration, unemployment and development: a two-sectoranalysis. American Economic Review 60: 126–142.

Heleniak, T. 1999. Out-migration and depopulation of the Russian North during the 1990s.Post-Soviet Geography and Economics 40(3): 155–205.

Hull, J. 2000. Aboriginal Post-secondary Education and Labour Market Outcomes. Canada, 1996.Ottawa: Data Strategic Research Directorate, INAC.

Ingene, C.A. 2001. The state of the art in modeling migration in LDCs: a comment. Journal of Regional Science 41(3): 529–543.

Kritchel, T. and P. Levine. 1999. The welfare economics of rural-to-urban migration: the Harris–Todaro model revisited. Journal of Regional Science 39(3): 429–447.

Kuhn, P. and A. Sweetman. 2002. Aboriginals as unwilling immigrants: contact, assimilationand labour market outcomes. Journal of Population Economics 15: 331–355.

McNiven, C. and H. Puderer. 2000. Delineation of Canada’s North: An Examination of the North–South Relationship in Canada. Ottawa: Statistics Canada.

Miron, J.R. 1978. Job-search perspectives on migration behaviour. Environment and Planning A10(5): 519–535.

Neary, J.P. 1981. On the Harris–Todaro model with inersectoral capital mobility. Economica 48:219–234.

Revising the Harris–Todaro Framework 205

Norris, M., M. Cooke, D. Beavon, E. Guimond and S. Clatworthy. 2004a. Registered Indianmobility and migration in Canada: patterns and implications. Pp. 136–160 in J. Taylor andM. Bell (eds), Population Mobility and Indigenous Peoples in Australasia and North America.New York: Taylor and Francis.

Norris, M.J., M. Cooke, D. Beavon, E. Guimond and S.J. Clatworthy. 2004b. Registered IndianMobility and Migration: Analysis of 1996 Census data. Ottawa: Indian and Northern AffairsCanada.

NT Bureau of Statistics. 2005a. Census Populations, Females 15 to 44, Average 5 Year Births, Aver-age Birth and Fertility Rates. Yellowknife: Government of Northwest Territories.

NT Bureau of Statistics. 2005b. Census Populations, Average 5 Year Deaths and Average MortalityRates. Yellowknife: Government of Northwest Territories.

Partridge, M. and D. Rickman. 1997 Has the wage curve nullified the Harris–Todaro model?Further US evidence. Economic Letters 54: 277–282.

Petrov, A.N. 2005. Modelling the effects of remoteness and transport accessibility on migrationprocesses in the Russian and Canadian North [in Russian]. Pp. 138–143 in Y. N. Gladky(ed.), Gumanitarnaya Geografiya v 21 Veke. St Petersburg: Epigraph.

Raimondos, P. 2003. On the Todaro paradox. Economics Letters 42(2-3): 261–267.Sarkar, P. 1994. Terms of trade of the South vis-à-vis the North: a macroeconomic framework.

Pp. 79–90 in D. Sapsford and W. Morgan (eds), The Economics of Primary Commodities. Brook-field VT: Edward Elgar.

Simmons, J. and L. McCann. 2001. The core–periphery structure of Canada’s urban system.Pp. 97–120 in T. Bunting and P. Filion (eds), Canadian Cities in Transition. New York: OxfordUniversity Press.

Statistics Canada. 1998. 1996 Census Community Profiles. Ottawa. <http://www12.statcan.ca/english/Profil/PlaceSearchForm1.cfm> Accessed: 15 February 2007.

Statistics Canada. 2002a. Community profiles 2001 Census. Catalogue # 93F0053XIE. Ottawa.<http://www12.statcan.ca/english/profil01/PlaceSearchForm1.cfm> Accessed: 10 Feb-ruary 2007.

Statistics Canada. 2002b. Profile of the Canadian Population by Mobility Status: Canada, a Nation onthe Move. 2001 Census: Analysis Series. Ottawa.

Statistics Canada. 2003a. 2001 Aboriginal Peoples Survey community profiles. Catalogue # 89–590-XIE. Ottawa. <http://www12.statcan.ca./english/profil01aps/home.cfm> Accessed:15 February 2007.

Statistics Canada. 2003b. 2001 Census Aboriginal Population Profile. Catalogue # 94F0043XIE.Ottawa. <http://www12.statcan.ca./english/profil01ab/PlaceSearchForm1.cfm> Ac-cessed: 15 February 2007.

Statistics Canada. 2006a. Aboriginal Peoples Survey 2001: public use microdata file – adults off reserve. Catalogue # 89M0020XCB. Ottawa.

Statistics Canada. 2006b. Census of Canada 2001: public use microdata file – individuals file.Catalogue #95M0016XCB. Ottawa.

Statistics Canada. 2006c. Provincial and international in-, out- and net-migration estimates, byprovincial regions, annual table #111-0027. Ottawa.

Suits, D.B. 1985. U.S. farm migration: an application of the Harris–Todaro model. EconomicDevelopment and Cultural Change 33(4): 815–828.

Todaro, M.P. 1969. A model of labour migration and urban unemployment in less developedcountries. American Economic Review 59: 138–148.

Todaro, M.P. 1971. Income expectations, rural–urban migration and employment in Africa.International Labour Review 104(5): 387–413.

Todaro, M.P. 1976. Urban job expansion, induced migration and rising unemployment: a for-mulation of simplified empirical test for LDCs. Journal of Development Economics 3: 221–225.

206 Andrey N. Petrov

Todaro, M.P. 1980. Internal migration in developing countries: a survey. Pp. 361–390 in R.A. Easterlin (ed.), Population and Economic Change in Developing Countries. Chicago: Universityof Chicago Press.

Todaro, M.P. 1995. Reflections on Economic Development. The Selected Essays by Michael P. Todaro.Brookfield VT: Edward Elgar.

Treasury Board of Canada. 2006. Isolated Posts and Government Housing Directive. Ottawa: Treas-ury Board of Canada Secretariat and National Joint Council. <http://www.tbs-sct.gc.ca/pubs_pol/hrpubs/ipgh-dpill/ipghd1-1_e.asp> Accessed: 17 January 2007

Trovato, F., A. Romaniuc and I. Addai. 1994. On-and Off-Reserve Migration of Aboriginal Peoples in Canada: A Review of the Literature. Ottawa: Department of Indian Affairs and Northern Development.