Non-Immigrant Mobility Responses to Immigrant Inflows in Canada: a Panel-Data Analysis

Transcription

1 Non-Immigant Mobility Responses to Immigant Inflows in Canada: a Panel-Data Analysis (DRAFT Pepaed fo the 44 th Annual Confeence of the CEA) Athe H. Akbai & Yigit Aydede 1 Abstact: Population plannes in smalle povinces of Canada have embaked on the mission of inceasing thei povinces shaes in national immigation intake in ode to at least patially offset thei povinces population declines. Within each povince, maintaining population balance though immigation acoss smalle communities is also a desied objective. Yet, ising immigation may also lead to out-migation of the native-bon population fom a community in the shot-un. If thee is a stong link between immigant inflows and native outflows, instead of mitigating demogaphic poblems, immigation may cause a demogaphic balkanization acoss Canada. This pape expands the cuent Canadian liteatue on immigation in thee new diections: Fist, we question whethe a linkage exists between immigation and native-bon outflows as spatial scales get fine. Second, we employ a multiegional famewok in which net intenal migation is analyzed on national scale. And lastly, to contol fixed egional and time-specific effects, we apply panel estimations using fou population censuses that cove the peiod fom 1991 to 2006 at two geogaphical levels: metopolitan aeas and census divisions. Ou esults indicate that locations with high levels of immigation ae less likely to etain o eceive non-immigant esidents in Canada. Acknowledgment: We thank D. E. Falais (Univesity of Delawae) fo his valuable comments on an initial daft. Responsibility fo any shotcomings emains with us. Keywods: Displacement, Immigation, Native-bon Mobility JEL Classification: J15, R31 1 Both authos ae membes of the Economics Depatment at Saint May s Univesity and associated with Atlantic Metopolis Cente. athe.akbai@smu.ca, yigit.aydede@smu.ca. 1

2 Based on its medium gowth scenaio, Statistic Canada pojects that the natual gowth ate of population will be negative in 2028 (Statistic Canada 2005). In othe wods, without immigation, not only will the Canadian population shink, but it will also be aging much faste than today. Smalle povinces and ual aeas have aleady begun to face population declines as they also face out-migation of the local population. As a esult, they ae now adopting initiatives to incease thei shaes in annual Canadian immigant inflows and also to etain the new aivals (Akbai, 2009). Yet, ising immigation levels in an aea may esult in out-migation of nativebon. This is likely to happen if immigants displace native-bon wokes in employment, bid down wages o cause house pices to ise though inceased demand fo shelte. Besides these economic easons, the native-bon may also have degee of social avoidance against immigants. The economic and social impacts of immigation have been the stongest in Monteal, Toonto, and Vancouve whee 80 pecent of immigants land each yea. The impact of ising immigant inflows on the depatues of native-bon and established immigant populations has been studied fo these cities by Hou and Boune (2004) and Ley (2007). Whethe these intenal migation effects of immigation also apply in smalle aeas has not yet been studied in Canada. Thee has been a discussion in the liteatue in the United States that the cowding-out effect of immigation is not only a phenomenon of lage metopolitan aeas but can also be obseved in smalle aeas (Cad 1997, Wight et al., 1997). If this is also tue in Canada, then instead of mitigating the demogaphic poblems faced by smalle povinces and thei communities, ecent policies towads egionalization of immigation may augment the cuent demogaphic tend with a ising sepaation of immigants fom the est of the population acoss egions. The pesent study investigates the mobility esponses of native-bon Canadians to immigant inflows in thee diffeent ways. Fist, we question the existence of spatial egulaities between native-bon mobility esponses and immigation as spatial scales get fine. As the gavity model of migation hypothesizes (Zipf, 1946), the volume of migation between two places is invesely popotional to the distance between the two. This implies that migation is pimaily a phenomenon of neighboing localities and it is highly likely that we lose obsevations about population flows between two locations as thei size gets lage. Hence, using census divisions that can be made consistent acoss censuses is a unique aspect of this study. Second, even though about 80 pecent of immigants aive and live in majo cities, lage agglomeations ae distinctive in thei fom and ecent evolution. Hence, the native-bon may be leaving lage cities because of paticulaities in 2

3 them, athe than because of immigation to them (Wight et al. 1997). This pape employs a multiegional famewok in which net migation is analyzed acoss Canada by using data on smalle aeas. Thid, people may also move intenally not only due to spatial diffeences in amenities, climate, cime levels, and cultue but also due to time-specific easons such as a tansition in national economy that might make some local populations moe susceptible than othes due to specifics of the local economic stuctue (Jackman and Savoui, 1992). The panel data that we have pooled ove fou population censuses coveing the peiod fom 1991 to 2006 enable us to emove those fixed effects so that estimations can identify the impact of immigation on the native-bon mobility. Ou esults indicate that locations with high levels of immigation ae less likely to etain o eceive non-immigant esidents in Canada. The est of the pape is oganized as follows: we summaize the elated liteatue in Section 1 and the data in Section 2. Section 3 shows the estimation esults of a model fist applied by Bojas et al. (1997). A spatial equilibium famewok is developed and estimated by using population gowth ates in Section 4 and by using mobility measues in Section 5. Ou intepetation of the esults is given at the end. 1. Liteatue eview Although the displacement effect of immigants in gateway cities may be lifestyle diven, thee is a gowing ageement among eseaches that low-income, less-educated native-bon population ae sensitive to immigant inflows because they ae most likely to be in diect competition with new immigants fo low-skill, low-paying jobs. Evey yea, Canada eceives about 225,000 immigants. Ove the last two decades moe than thee million immigant wokes have enteed into local labou makets. Despite the fact that Canada is one of the majo immigant eceiving counties in the wold, studies investigating the effects of immigation on local makets ae scace. Labou maket studies (Akbai and Aydede 2010, Islam 2009, Akbai and DeVoetz 1992, Roy 1997, 1987) epot impefect substitutions between immigant and native bon wokes implying that new immigants ae not in diect competition with native-bon wokfoce. 2 In thei ecent study, howeve, Aydemi and Bojas (2007) find a stong negative impact of immigant inflows on labou maket outcomes in Canada. Unlike othes, they question how much of the dispaity in the outcomes of diffeent skill goups of native-bon wokes can be attibuted to immigation that shifts the elative 2 Thee ae also two ecent studies on housing makets (Akbai and Aydede, 2009; Ley and Tutchene, 2001) indicating a weak negative linkage between immigation and local housing pices. 3

4 demand and supply of diffeent skills at the national level. In the immigation liteatue, this appoach is justified by the fact that if thee ae substantial nativebon outflows in esponse to immigant inflows, a naïve spatial empiical study may even find a positive impact of immigation on local maket outcomes. One of the poposed solutions to this bias in the estimates of immigants impact is to test diectly eactions of native-bon populations to immigation. Most studies on native-bons esponses to immigation have been conducted fo the United States, whee evidence is mixed. 3 Fey (1994, 1995, 1996, 2002), found stong native-bon mobility esponses leading to demogaphic balkanization of U.S. cities. Bojas, Feeman, and Katz (1997) epoted consistent evidence confiming substantial outmigation of native-bon in esponse to immigant inflows on national scale. Fey s displacement hypothesis has been challenged by White and Imai (1994), Wight, Ellis, and Reibel (1997) and Haison (2002) who found that net in-migation of the native-bon is eithe positively elated o unelated to immigation in metopolitan aeas. In fact, thei esults indicate that the net loss of unskilled native-bon wokes fom metopolitan aeas is pobably a function of those cities population size and industial estuctuing athe than immigant inflows to them. Moeove, Cad and DiNado (2000) estimated the net impact of immigation inflows on the elative skill distibution of diffeent cities in the U.S. and found that inceases in immigant population in specific skill goups led to small inceases in the population of native-bon individuals of the same skill goup. Hatton and Tani (2005) eviewed migation pattens acoss 11 egions of the United Kingdom using annual data fo the peiod fom 1981 to They found a stong negative link between immigation flows and native-bon mobility esponses. Moe specifically, fo all 11 egions, thei esults showed that a 1 pe cent incease in immigation educes net in-migation of native-bon by pe cent, implying that immigation induces native-bon esidents to elocate to othe cities. No Canadian study has eviewed native-bon mobility esponses on national scale. Two ecent studies, Hou and Boune (2004) and Ley (2007), have found that the gowth in ecent immigation co-vaies with out-migation ates among the less-educated native-bon in Toonto and Vancouve. While Ley compaes Sydney and Toonto by using time-seies data between 1977 and 2002, Hou and Boune calculate in and out-migation ates by using multivaiate logistic egession techniques on a sample of mico-data dawn fom five censuses fom 1981 to 2001 fo the woking population aged between 25 and 64 living in thee CMAs (Toonto, Monteal, and Vancouve). Hou and Boune compaed the 3 An excellent liteatue eview can be found in the Hou and Boune (2004) pape. 4

5 effects of economic estuctuing, housing maket conditions, and immigant ininflows in a CMA on the tends in intenal migation fom and to that CMA. They found a significant coelation between gowth in the ecent immigant population and the inceased out-migation ate among the low-skilled Canadian-bon in Toonto and Vancouve. Howeve, this association becomes insignificant acoss CMAs implying that immigation may not be the majo souce of out-migation of the native-bon in gateway cities. 2. Data We use the 1991, 1996, 2001, and 2006 population censuses to build panel data at two geogaphical levels: census metopolitan aeas (CMA) and census divisions (CD). 4 One poblem with pooling censuses is that the geogaphical coveage of some CDs and CMAs changes though time. Theefoe, we used the concodance tables, povided by Statistics Canada, to dop the egions whose boundaies have changed significantly (moe discussion in this egad is povided in Appendix). Although we pooled only fou censuses with 31 CMAs and 238 CDs to limit the numbe of egions dopped fom the panel, the peiod this study coves is peculia because of two easons: (1) between 1991 and 2000, Canada eceived aound 2.2 million new immigants the lagest decadal inflow in the past 100 yeas; (2) duing 1990s the majo souce of immigant sending egions shifted futhe away fom Euope to Asia and othe thid wold counties. Immigants coming fom these egions ae distinguished as visible minoities in the censuses afte Ou analysis investigates local population flows including all people, not just labou flows. This is mainly because data on the mobility of non-immigant wokes by skill goups o by labou foce status ae not available to us at the CD level and the cost of obtaining such data is too high. Although labou maket effects on elocation decisions will be diffused in this setting, these data do have the advantage of including people who ae not in the labou foce but move to othe localities due to housing maket outcomes and social avoidance. 5 Lastly, we use data on net population gowth ates and mobility of the local population. Hence, ou esults ae not able to distinguish out- and in-migation esponses to immigation but eveal (1) spatial egulaities between net gowth 4 To conduct a population census, Statistics Canada divided the county into 289 povincially legislated census divisions (CDs), o smalle communities that ae intemediate geogaphic aeas between the povince/teitoy level and the municipality (census subdivision). A CMA efes to the main labou maket aea of an ubanized coe having 100,000 o moe population. CMAs ae ceated by Statistic Canada and ae usually known by the name of the uban aea foming thei ubanized coe. 5 This is simila to the Fey s models which cove all people. 5

6 ates of non-immigants and immigants in the local population; 6 (2) the sensitivity of local population mobility to changes in the density of immigation acoss censuses and egions. The esulting 5-yea inteval between censuses povides a easonable time window fo the individuals to espond to changes in local conditions. Table 1 below gives an idea about the povincial diffeences in population gowth ates between 1971 and It shows that while Bitish Columbia, Ontaio, and Quebec expeienced lage inceases in thei immigant population elative to the incease native-bon population, opposite was tue in Albeta, Saskatchewan, and Manitoba. Manitoba and Saskatchewan expeienced a decline in thei immigant populations duing the same peiod. Table 1: Components of population gowth ( ) pecent Povinces Total Native-Bon Immigants Canada Newfoundland and Labado (NL) Pince Edwad Island (PEI) Nova Scotia (NS) New Bunswick (NB) Quebec (QE) Ontaio (ON) Manitoba (MB) Saskatchewan (SK) Albeta (AB) Bitish Columbia (BC) Souce: Authos calculations based on the data. Table 2 below shows the povincial distibution of immigant and native bon populations befoe and afte This is the yea since when smalle povinces began to take geate inteest in attacting and etaining immigants to offset thei population decline and aging. Except fo Atlantic Povinces, the shaes of immigant and native-bon populations tended to move in opposite diections with an acceleated ate paticulaly afte While Bitish Columbia, Ontaio, and Quebec expeienced a lage-scale immigation afte 1996, we obseve the opposite tend in the Paiie Povinces. We also calculated annualized population gowth ates (not shown hee). An inceasing immigant population afte 1986 coincided with a declining gowth in 6 Bojas et al. (1997), File (1992), Walke et al. (1992), and Fey (1994, 1995) use net netmigation models. 6

7 the native-bon population in Bitish Columbia, Ontaio, and Quebec. Finally, we also eviewed the distibution of native-bon and immigant populations fo 36 CMAs 7 and calculated components of thei population gowth ates fo the peiod 1981 to 2006 (not shown hee). It was inteesting to obseve, fo example, that while Vancouve expeienced a 123-pecent incease in its immigant population compaed to a 40-pecent incease in its native-bon population, all majo cities in Bitish Columbia expeienced an opposite tend. In fact, this was almost tue fo all thee immigants eceiving povinces: all neighboing CMAs to Monteal, Toonto, Calgay, and Ottawa had highe gowth in thei native-bon populations than in thei immigant populations. Table 2: Distibution of population by immigant status, % Canada BC QE ON N M N M N M N M MB SK AB Atlantic N M N M N M N M Souce: Authos calculations based on the data. M and N epesent immigant and native-bon populations. Atlantic Povinces ae NL, PEI, NS, and NB. Although the data shown in tables 1 and 2 ae instuctive and eflect immigation policy changes, using a povincial-level classification can mask possible cowding-out pattens acoss metopolitan aeas and/o smalle neighboing egions within the same povince. 7 Each subsequent census has new additions to thei cuent CMA list, as some census agglomeations (CA) each to 100,000 o moe population. Hence, to constuct a balanced panel, we emoved all new additions afte

8 3. Fist & double diffeence egession models Bojas, Feeman, and Katz (1997) compaed native intestate migation to immigation in the US duing by estimating fist diffeence and double diffeence models. To undestand whethe non-immigant individuals adjust to the impact of immigation in an aea by moving thei labou o capital to othe localities, we begin with fist diffeence model, which takes the following fom: n ( t, t j) m ( t, t j) e, (1) whee e is the stochastic eo and n and m ae the annualized population gowth ate contibutions of the native-bon and immigants in egion, espectively, as computed below: n ( t, t j) N t j jp t N t, and m ( t, t j) M t j jp t M t, whee N and M ae the numbes of native-bon and immigant individuals living in egion in yea t and P is the population. Equation (1) coelates the annual gowth ate of native-bon population in egion i to the gowth ate of immigants in the same egion, elative to the egion s total population. Hence, the coefficient β measues the immediate effect of one additional immigant aiving in egion between t + j and t on the change in the numbe of native-bon individuals living in that egion duing the same peiod. 8 As pointed out by Bojas et al., one poblem with (1) is that the fist-diffeence specification implicitly assumes each egion could have the same gowth ate in native-bon population absent immigation. Howeve, even if thee is no immigation, many CDs (CMAs) pobably would have diffeent gowth paths in any given two subsequent peiods. As seen in Table 2, the suge in immigant inflows afte 1996 has changed the distibution of population, paticulaly in thee majo immigant eceiving povinces. Thus, having contasting two peiods peteatment peiod fom 1991 to 1996 and post-teatment peiod allow us 8 Some bias in the coefficient estimate is expected since ou calculations do not eflect changes in the numbe of the second geneation immigants bon in Canada and the death ates among nativebon individuals. 8

9 to isolate the impact of immigation on the change in the native-bon population. To addess the poblem with the fist diffeence specification, we apply a diffeence-in-diffeence compaison as specified below: n ( 96,06) n (91,96) [ m (96,06) m (91,96)] e, (2) whee the coefficient β now eveals what would have happened to the egion s native-bon population gowth afte 1996 if the immigant supply shock had not occued between 1996 and Table 3: Estimating the effect of changes in the numbe of immigants on the numbe of the native-bon by fist & double diffeence specifications CD 3(a) 3(b) 3(c) β SE β SE β SE m (91,06) m (96,06)-m (91,96) n (91,96) m (96,06) m (91,96) # of obsevations R 2 CMA m (91,06) m (96,06)-m (91,96) n (91,96) m (96,06) m (91,96) # of obsevations R Notes: (1) Fo all estimations, the odinay least squae (OLS) method is used. (2) Robust standad eos (SE) ae calculated and adjusted by povincial clustes fo the CD-level and egional clustes (ON, BC, QE, Atlantic, and Paiie) fo the CMA-level. (3) The dependent vaiables ae n(91,06) fo 3(a), n(96,06)- n(91,96) fo 3(b), and n(96,06) fo 3(c) as defined in the text above. The fist column in Table 3 epots the esults of (1) estimated fo CDs and CMAs sepaately. A positive coefficient of β ejects a negative link between the gowth ates of immigant and native-bon populations. Because both n and m ae scaled by the same denominato, β quantifies the effect of an immigant aiving to the CD (CMA) on the change in the native-bon population duing the same peiod. 9 The same method was used by Bojas et al. (1997). 9

10 Fo each aiving immigant, the native-bon population inceases by 0.46 (1.06) pesons in CDs and by 0.29 (0.46) pesons in CMAs on aveage. While the second column in Table 3 above epots the esults of double diffeence estimations specified by Equation (2), the thid column shows unesticted vesion of (2) that elaxes the estictions on the coefficients. Contolling the egion s pe population gowth ate changes the sign of the coefficient β fom positive to negative at the CD-level, implying that one pecent incease in the gowth ate of the immigant population educes the native-bon population s gowth ate one pecent (the coefficients ae not significantly diffeent fom -1). At the CMAlevel, howeve, the esults ae not indicative of a negative link. One eason fo this can be a potential simultaneity poblem that gets wose paticulaly fo CMAs. Since abitay adjustment costs ise as the distance between souce and eceiving egions gets fathe acoss CMAs, the non-immigant population may not adjust contempoaneously in esponse to immigant inflows. In addition, specifications (1) and (2) do not contol egional fixed effects efficiently. This poblem gets wose, as spatial diffeences become moe significant acoss CMAs as opposed to neighboing egions. 4. A Spatial equilibium famewok To identify the key channels by which immigation affects out-migation of nonimmigants, we intoduce a simple spatial equilibium model in which population flows ae explained by egional housing and labou maket conditions, the quality of local amenities, and the pesence of social avoidance and/o self-selected ethnic segegations. 10 To undestand how elocation decisions can be made by nonimmigant individuals, we stat with the following sepaable utility function: U i 1 h 1 1 R h w ( P ) A i M. (3a) The tem in the backets captues the net income effect of housing, whee h is the value eceived fom housing sevices that the non-immigant individual i consumes and R epesents the housing ent in egion so that the optimal housing can be expessed as h i = (α/r ) 1/. The individual eans and consumes the cuent egional wage, w, which is a function of the local population (P ) the sum of non-immigant (N ) and immigant (M ) esidents. Assuming that all individuals ae in the labou foce, the demand fo labou can be expessed by w w N M ), whee ρ measues the impact of the population gowth ( 10 This model builds on the famewok that is used by Saiz (2007) to identify the elationship between immigation and local housing makets in the U.S. 10

11 on local wages and ε eflects the degee of substitutability between immigant and 11 non-immigant wokes with 0 1. The value of egional amenities fo the individual is epesented by the tem A, which is heteogeneous among established non-immigants and new immigants. Hence, Ai T an povides a linea appoximation of the congestion effect, whee T is the total capacity of local amenities and a is the individual shae of amenities that each non-immigant consumes on aveage. Finally, immigation has a diect negative impact on the well-being of the nonimmigants expessed by M, whee captues the degee of self-segegation and/o social avoidance. The native-bon and established immigants may want to live in communities with households who have simila cultual and social values. This topic has been a subject of discussions paticulaly in the U.S. Fey has used tems of demogaphic balkanization and white flight in his ealie aticles investigating the acial segegation acoss cities in the U.S. File (1992) found that while white wages ae affected less by low-skilled immigation than black wages in the U.S., mobility esponses wee stonge among whites, implying that thee is something othe than diect labou maket effects influencing native-bons migation. At the steady state, fo a spatial equilibium to hold, U i must equal to a esevation utility level denoted by U. In othe wods, the maginal nonimmigant will be indiffeent between staying and leaving the egion if U i = U whee we nomalize the utility level outside of the egion to U. 12 Fom this spatial equilibium condition, we can deive the supply of non-immigant esidents in egion as follows: N 1 a ( T T ) ( w w ) ( R 1 R 1 ) a M, (3b) 1 1/ whee ( a ) ( )M N and /(1 ). We assume that immigation to a egion fom aboad is exogenously detemined by conditions in souce counties and pevious immigant inflows. 11 We assume that ρ > To incopoate moving costs, we can discount U by μ, hence the equilibium condition becomes U i = µu whee 0 µ 1 depending on the distance. Although discounting does not change the fundamental esults, it implies that spatial diffeences and immigation have weake effects on elocation decisions as the distance gets fathe. 11

12 A numbe of obsevations can be made on (3b). Fist, unfavoable spatial diffeences in amenities, income levels, and housing costs expessed in the fist paenthesis will have negative effects on the numbe of non-immigant esidents. 13 Second, the effect of immigation on non-immigant mobility is not independent of the paamete of ρ the impact of immigation on local wages. Fo example, if ρ is lage, in the case of impefect substitution, effects of immigation on intenal migation will be diffused, even if the diect disutility fom immigation is substantially lage than zeo. Thid, in this famewok, the labou maket is clea and thee is no unemployment. If wages do not adjust, the intenal migation will be detemined by the inteegional diffeences in unemployment ates. And lastly, the coefficient on M is biased downwad because the elation between local housing ents and immigation is absent in the above setting. 14 Note that, by using (3b), we can also expess diffeences in the supply of nonimmigant esidents between egion j and as a function of spatial diffeences and the divegence in the size of immigant populations as follows: N j N a a M j M, (3c) whee Ω epesents the tem in the fist paenthesis of (3b). 15 Based on (3c), ou estimating famewok takes the following fom: ln( N t ) ln( M t ) kxt 1 Z 1 t t u e, (4) whee the dependent vaiable is the gowth ate of non-immigant esidents measued by the change in the log of N. Since we expect that the adjustment of the non-immigant population may not be contempoaneous in esponse to 13 The list of spatial diffeences that may affect the supply of non-immigant esidents should be longe than what we have in (3b). Fo example, as Ley (2007) points out, global cities have been expeiencing significant shifts in thei economic stuctue with new types of capital gowth and polaized labou demand in the last two decades. This ongoing economic estuctuing has ceated new demand fo manageial and pofessional occupations with deceasing impotance of pimay jobs in some disappeaing manufactuing and sevice industies. As a esult, blue-colla nativebon wokes had to move diffeent places o upgade thei skills while new immigant woke ae willing to wok fo low wages (Sassen 1995). 14 Fom (3b), it is obvious that an incease in R educes N. It can also be shown that, at any given level H (= h x P) and N, an incease in M aises R in the shot-un. 15 See Hatton and Tani (2005) fo a non-immigant labou supply function in tems of gowth ates. Moeove, (3c) puts a constaint on the coefficient of (M j M ) by assuming that the degee of substitutability (ε) and the disutility fom immigation ae identical acoss egions, which may not be an unealistic assumption fo neighboing egions. When we elax it, (3c) becomes 1 1 N N ( a ) ( ) M ( ) M ( a ) j j j j 12

13 immigant inflows and changes in local conditions, we use lagged values fo the gowth ate of immigant esidents, Δln(M t 1 ), and othe explanatoy vaiables, X t-1, which is a 1 x k vecto of vaiables that vay ove egion and time. Lastly, Z is a vecto of vaiables that vay only ove egions, u and τ t ae the unobseved egion specific and time specific effects, espectively, and e t is the idiosyncatic distubance tem which satisfies e t ~ i.i.d. (0, ) fo all t and. 16 Following the liteatue, fist, we pedict that, all things being equal, people tend to move to egions with lowe housing costs. As poposed by Poteba (1991), in equilibium, the expected cost of owning a house should equal the cost of enting. 17 Hence, we will use two altenative vaiables to contol egional housing costs: goss aveage monthly ents fo esidential popeties and aveage housing pices. Second, we contol the linkage between labou maket outcomes and population mobility by changes in the egion s unemployment ate. Ou pediction is that a ising unemployment will be associated with a declining gowth in the non-immigant population. Finally, to emove unobseved spatial diffeences, we fist apply the following fixed effects model that includes a full set of yea dummies: lnn t lnn ( ln M t ln M ) k ( ln X t 1 ln X ) ( e 1 t t whee vaiables with ba epesent panel-level aveages fo thei espective peiods. By its constuction, (5) has explanatoy powe only if the vaiations in ΔlnN aound its mean ae significantly coelated with the vaiations in ΔlnM. The estimation esults of (5) with and without lagged explanatoy vaiables ae povided in Table 4 below. The estimations in the fist column, 4(a), which use non-lagged explanatoy vaiables, show counteintuitive esults and, when compaed with the esults in 4(b), which use lagged vaiables, the estimations appea to be sensitive to the use of lagged covaiates implying that the contempoaneous link between native-bon and immigant population gowth ates might be subject to a simultaneity poblem. Moeove, the test esults confim that unobseved egional fixed-effects ae not andom and, theefoe, e (5) ), 16 As expected, the test esults stongly eject the spatial independence so that the measued gowth in the native-bon population in one location may be coelated those in neighboing locations. We addess this poblem in ou estimations by using clusteed obust standad eos. 17 Use cost = R = P( + τ + λ π) whee R is the imputed goss ent, P is the house pice, is the cost of foegone inteest that the homeowne could have eaned on an altenative investment, τ is the popety tax ate, λ is the ecuing holding cost consisting of depeciation, maintenance and the isk pemium on esidential popety, and π is the expected capital gains (o loss). 13

14 although we use gowth ates in (4), the use of specification (5), instead of a andom-effect model, is justified. Table 4: Fixed-effect estimatos with gowth ates CD 4(a) 4(b) 4(c) Coefficent SE Coefficent SE Coefficent SE lnm ln(un ) ln(hpice ) L. lnm L. ln(un ) L. ln(hpice ) L. ln(ent ) No of obsevations (t, ) R2 (within) ho CMA (3, 234) (2, 234) (2, 234) lnm ln(un ) ln(hpice ) L. lnm L. ln(un ) L. ln(hpice ) L. ln(ent ) No of obsevations (t, ) R2 (within) ho (3, 31) (2, 31) (2, 31) Notes: (1) The dependent vaiable is lnn. (2) Standad eos (SE) ae obust and adjusted by povincial (fo CDs) and egional clustes (fo CMAs). (3) ho indicates the faction of the unexplained vaiance due to diffeences acoss egions. (4) All egessions have a set of dummy vaiables to contol yea effects (not shown hee). (5) L in font of the vaiable indicates the lagged value of the vaiable Expect fo 4(a), the fixed-effect estimatos consistently show statistically significant and negative elationships between the gowth ates of non-immigant and immigant esidents. Although, the unemployment ate has a pedicted sign, its significance at the CD-level is sensitive to the type of housing cost used in estimations. This could be explained by the fact that the dependent vaiable is the gowth in the non-immigant population not in the numbe of wokes. The esults also imply that, even if the sign on the coefficient agees on what the model pedicts, ising housing pices o ents have no statistically significant effect on the numbe of local non-immigant esidents. By definition, (3a) abstacts fom income effects in housing consumption. Hence, a moe sensible appoach would be to conside income net of housing costs, not housing costs as measued by 14

15 aveage ents o housing pices. Fo example, conside two neighboing egions with identical amenities, one of which has annual wages of $40,000 with annual housing costs of $10,000, while the othe has annual wages of $60,000 and annual housing costs of $30,000. As seen in this example, even though the housing cost is highe in the second egion, highe wages offset the diffeence and in both egions people ean the same income ($30,000, net of housing cost). This poblem leads to a downwad bias in coefficients of housing vaiables in ou estimations Sensitivity of inteegional migation to intenational migation Instead of using population gowth ates to undestand whethe immigation induces non-immigant esidents to elocate to othe egions, we now use inteegional and intenational mobility measues fo local esidents. We estimate the following vesion of (5): lni t lni (ln Et ln E ) k (ln X t 1 ln X ) ( e 1 t t e ), (6) whee I t is the numbe of local esidents who lived in a diffeent egion in Canada five yeas ago and E t is the numbe of esidents who lived in a diffeent county five yeas ago. Hence, the specification (6) eveals the sensitivity of inteegional mobility to immigation when the egional housing and labou maket conditions ae contolled. If the displacement effect of immigation is significant, we expect that egions eceiving moe esidents fom aboad will attact less intenal migations. Table 5 above shows the estimation esults of (6) with and without lagged explanatoy vaiables (5(a) and 5(b), espectively). As befoe, counteintuitive esults in the fist column imply the pesence of a simultaneity poblem. The estimations in the last two columns show statistically significant β coefficient with expected signs both at the CD and CMA levels. Moeove, ising unemployment ates now have significant and negative effects on the numbe of local esidents moved fom diffeent egion in the last five yeas. This is pehaps because the dependent vaiable may now include moe people who ae in the labou foce, assuming that people move mostly when younge fo bette labou maket oppotunities. Finally, housing vaiables have no obust explanatoy powe on the numbe of intenal migants due to the bias as explained befoe. 18 See Glaese (2008) fo moe details. 15

16 Table 5: Fixed-effect estimatos with mobility measues CD 5(a) 5(b) 5(c) Coefficient SE Coefficient SE Coefficient SE ln(e ) ln(un ) ln(hpice ) L.ln(E ) L.ln(un ) L.ln(hpice ) L.ln(ent ) No of obsevations (t, ) R2 (within) ho CMA (4, 234) (3, 234) (3, 234) ln(e ) ln(un ) ln(hpice ) L.ln(E ) L.ln(un ) L.ln(hpice ) L.ln(ent ) No of obsevations (t, ) R2 (within) ho (4, 31) (3, 31) (3, 31) Notes: (1) The dependent vaiable is lni. (2) Standad eos (SE) ae obust and adjusted by povincial (fo CDs) and egional clustes (fo CMAs). (3) ho indicates the faction of the unexplained vaiance due to diffeences acoss egions. (4) All egessions have a set of dummy vaiables to contol yea effects (not shown hee). (5) L in font of the vaiable indicates the lagged value of the vaiable Concluding emaks This study investigates non-immigant mobility esponses to immigant inflows in Canada on national scale fo the fist time in liteatue. It uses data fom fou population censuses coveing the peiod 1991 to 2006, which is a unique peiod in the ecent Canadian immigation histoy. We built a panel ove fou population censuses by using concodance tables to obtain stable bodes at the CMA and CD-levels though time. To ou knowledge, the CD-level spatial scale has not been used to investigate population flows in Canada befoe. Migation is a phenomenon mostly in the neighboing egions. Theefoe, using the smallest available geogaphical classification in a balanced panel is impotant as lage egional sizes mask local population flows. 16

17 We fist estimate a model applied by Bojas et al. (1997). At the CD-level, the esults imply that one pecent incease in the gowth ate of the immigant population educes the non-immigant population s gowth ate by one pecent. At the CMA-level, howeve, the esults ae not indicative of a negative link. This may indicate immigation may be causing outmigation of native-bon within a CMA. In the second pat of the pape, we intoduce a spatial equilibium model that identifies the channels by which immigation may impact non-immigant population flows. When we apply panel estimations, the esults consistently show negative and obust impact on location decision of non-immigants. The estimated coefficients on immigation vaiables ae not independent of the elasticity of labou demand, the substitution between immigant and nonimmigant wokes, and the effect of immigants on housing cost, even though we contol housing and labou maket conditions and emove unobseved spatial diffeences. Theefoe, a negative coefficient substantiates its existence but cannot povide an unbiased magnitude of social avoidance and/o selfsegegation. Refeences Akbai, A. (2009) Socioeconomic and Demogaphic Pofiles of Immigants in Atlantic Canada ( ). Repot pepaed fo Atlantic Canada Oppotunities Agency. Akbai, A. and Aydede, Y. (2010) Do the Educational Cedentials of Immigant and Non-Immigant Wokes make them Pefect Substitutes fo Each Othe in Canadian Labou Makets? A Poduction Function Analysis. Woking Pape , Atlantic Metopolis Cente. Akbai, A. and Aydede, Y. (2009) Effects of Immigation on Housing Pices in Canada. Woking Pape , Atlantic Metopolis Cente. Akbai, A. and DeVoetz, D. (1992) The substitutability of foeign bon labou in Canadian poduction: cica Canadian Jounal of Economics, 25. Aydemi, A. and Bojas, G. (2007) Coss-County Vaiation in the Impact of Intenational Migation: Canada, Mexico, and the United States Jounal of Euopean Economic Association 5(4). Bojas, G., Feeman, R. and Katz, L. (1997) How Much Do Immigation and Tade Affect Labo Maket Outcomes? Bookings Papes on Economic Activity 1997(1): Washington, D.C.: Bookings Institution. 17

18 Cad, D. and DiNado, J. (2000) Do Immigant Inflows Lead to Native Outflows? New Issues in Immigation 90(2). Cad, D. (1997) Immigant Inflows, Native Outflows, and the Local Maket Impacts of Highe Immigation. NBER Woking pape File, R. (1992) The Effect of immigant Aivals on Migatoy Pattens of Native Wokes. In Bojas and Feeman (eds), Immigation and the Wokfoce: Economic Consequences fo the United States and Souce Aeas. Chicago: The Univesity of Chicago Pess. Pp Fey, W. (1994) The New White Flight. Ameican Demogaphics. Fey, W. (1995) Immigation and Intenal Migation Flight fom US Metopolitan Aeas: Towads a New Demogaphic Balkanization. Uban Studies, 32(4-5): Fey, W. (1996) Immigation, Domestic Migation, and Demogaphic Balkanization in Ameica: New Evidence fo the 1990s. Population and Development Review, 22(4): Fey, W. (2002) Thee Ameicas: The Rise Significance of Regions. Jounal of the Ameican Planning Association, 68(4): Glaese, L.E. (2008) Cities, Agglomeation, and Spatial Equilibium. Oxfod Univesity Pess Inc. New Yok USA. Haison, R. (2002) Moving Out when Minoities Move In. Ameican Demogaphics, (June): Hatton, J.T. and Tani, M. (2005) Immigation and Inte-Regional Mobility in the UK, The Economic Jounal 115(Novembe): Hou, F. and Boune, L.S. (2004) Population Movement Into and Out of Canada s Immigant Gateways Cities: A Compaative Study of Toonto, Monteal and Vancouve. Statistic Canada, Analytical Studies Banch eseach pape seies, Catalogue no: 11F0019MIE No: 229. Islam, A (2009) The substitutability of labo between immigants and natives in the Canadian labo maket: cica Jounal of Population Economics

19 Jackman, R. and Savoui, S. (1992) Regional Migation in Bitain: An Analysis of Goss Flows Using NHS Cental Registe Data. The Economic Jounal 102(415): Ley, D. and Tutchene, J. (2001) Immigation, Globalisation and Housing Pice Movements in Canada s Gateway Cities. Housing Studies 16: Ley, D. (2007) Countevailing Immigation and Domestic Migation in Gateway Cities: Austalian and Canadian Vaiations on an Ameican Theme. Economic Geogaphy 83(3): Poteba, J.M. (1991) House Pice Dynamics: The Role of Tax Policy and Demogaphy, Bookings Papes on Economic Activity, 2, and Roy, A. (1987) An Analysis of substitutability and complementaity of immigants and Canadian-bon wokfoce. Labou Maket Bulletin 4. Roy, A. (1997) Job displacement effects of Canadian immigants by county of oigin and occupation. Intenational Migation Review 31(1). Statistic Canada (2005) Population Pojections fo Canada: Povinces and Teitoies, Catalog Numbe: XIE. Walke, R., Ellis, M. and Baff, R. (1992) Linked Migation Systems: Immigation and Intenal Labou Flows in the United States. Economic Geogaphy, 68(3): White, M. and Imai, Y. (1994) The Impact of U.S. immigation upon intenal Migation. Population and Envionment, 15(3): Wight, R. A., Ellis, M., and Reibel, M. (1997) The Linkage between Immigation and Intenal Migation in Lage Metopolitan Aeas in the United States. Economic Geogaphy 73(2): Saiz, A. (2007) Immigation and Housing Rents in Ameican Cities. Jounal of Uban Economics, 61: Sassen, S. (1995) Immigation and Local Labo Makets. In The economic sociology of immigation, ed. A. Potes, New Yok: Russel Sage Foundation. Zipf, G. (1946) The [P(1)P(2)/D] hypothesis; on the intecity of pesons. Ameican Sociological Review. 11:

20 Appendix: a note on bounday changes of CDs and CMAs acoss the 1991, 1996, 2001, and 2006 censuses. We used data fom the last fou population censuses, i.e., 1991, 1996, 2001, and 2006, to geneate a panel. In each census, thee ae thee main standad geogaphical classifications (SGC): povinces o teitoies (PR), census divisions (CD), and census subdivisions (CSD). Since CD boundaies tend to be moe stable ove the yeas than CSDs, we fist used CDs to pool all fou censuses. To make each CD consistent in each census, we identified CDs whose coveage was affected by bounday changes by using concodance tables 19 and emoved them fom each census. In total, 51 CDs wee emoved, and thei codes ae 1010, 1011, 1301, 1305, 2419, 2424, 2425, 2426, 2436, 2437, 2442, 2443, 2444, 2449, 2451, 2452, 2455, 2457, 2458, 2459, 2561, 2462, 2477, 2479, 3512, 3514, 3534, 3539, 3548, 3552, 3553, 3557, 4615, 4616, 4803, 4812, 4814, 4815, 4816, 4818, , 5911, 5911, 5913, 5915, 5923, 5925, 5933, 5939, 5947, and Afte this adjustment, the esidual numbe of CDs was We also compaed the land aeas fo each CD in each census and noticed that in the emaining CDs (whee we thought no bounday changes wee made), the land aeas (squae kilometes) have changed acoss censuses. While between 2001 and 2006 diffeences in size wee mino, it was not the case between 1991, 1996, and Statistics Canada povided the following esponse to ou question egading aea changes: Uses should note that even when the boundaies of standad geogaphic aeas did not change between censuses, the land aeas may diffe due to geomety shifts. The shifts ae caused by a change in the undelying database achitectue and by impovements in the absolute positional accuacy of some of the oads. Theefoe, even though we ae able to contol bounday changes in ou estimations, thee can be some measuement eos in pooled data due to geometic shifts within a CD. Statistics Canada uses, in addition to the SGC, a numbe of othe standad geogaphical entities (e.g., census metopolitan aeas and census agglomeations, economic egions, health egions, and counties) fo the povision of statistics. The geneal concept of these standad units is defined as the uban coe, and the 19 Please see 20 We also emoved CDs fom the Nothen Teitoies. 20

21 adjacent uban and ual aeas that have a high degee of social and economic integation with that uban coe, as measued by commuting flows deived fom Census of Population data on place of wok by Statistic Canada. 21 To fom a census metopolitan aea (CMA), the uban coe must have a population of at least 50,000 and the aea (CMA) must have a population of at least 100,000. Once an aea becomes a CMA, it is etained as a CMA even if the population of its uban coe declines below 50,000, o if its total population falls below 100,000. Since thee ae new CMAs in each census, in ode to be able to build a balanced panel, we took the CMA list in 1991 as ou base list and emoved all new additions fom the subsequent censuses. Hence, the esulting list became as follows: Code CMA Code CMA 1 St. John's (N.L.) 559 Windso (Ont.) 205 Halifax (N.S.) 562 Sania (Ont.) 305 Moncton (N.B.) 575 Noth Bay (Ont.) 310 Saint John (N.B.) 580 Geate Sudbuy / Gand Sudbuy (Ont.) 408 Saguenay (Que.) 590 Sault Ste. Maie (Ont.) 421 Québec (Que.) 595 Thunde Bay (Ont.) 433 Shebooke (Que.) 602 Winnipeg (Man.) 442 Tois-Rivièes (Que.) 705 Regina (Sask.) 462 Montéal (Que.) 725 Saskatoon (Sask.) 505 Ottawa - Gatineau (Ont./Que.) 810 Lethbidge (Alta.) 521 Kingston (Ont.) 825 Calgay (Alta.) 529 Peteboough (Ont.) 830 Red Dee (Alta.) 532 Oshawa (Ont.) 835 Edmonton (Alta.) 535 Toonto (Ont.) 915 Kelowna (B.C.) 537 Hamilton (Ont.) 925 Kamloops (B.C.) 539 St. Cathaines - Niagaa (Ont.) 932 Abbotsfod (B.C.) 541 Kitchene (Ont.) 933 Vancouve (B.C.) 543 Bantfod (Ont.) 935 Victoia (B.C.) 550 Guelph (Ont.) 970 Pince Geoge (B.C.) 555 London (Ont.)

22 Moeove, we identified the following 8 CMAs: 205, 521, 529, 543, 575, 602, 810, and 970 with substantial changes in thei land sizes acoss yeas. Afte emoving those 8 CMAs, we had 31 CMAs. 22