Early school leaving among immigrants in Toronto secondary schools.

Early school leaving among immigrants in Toronto secondary schools. IMMIGRANTS ARE WILLING TO BEAR THE costs of moving to a new andunfamiliar country in order to pursue economic opportunities, a betterstandard of living and, what is especially important to most, a brighterfuture for their children. Surveys of immigrants consistently indicatethat nearly all hold postsecondary educational aspirations for theirchildren (Krahn and Taylor 2005). Given the need for skilled labor andan informed citizenry, it is equally important to Canada that thechildren of immigrants are well-educated. Completing high school is, ofcourse, a prerequisite to gaining access to postsecondary studies and isconsequently among the first, and most important, steps immigrant youthtake toward making the most of the opportunities that served to motivatetheir family's move to Canada. Failing to complete high schooljeopardizes the economic prospects of immigrant youth and imposes asocial cost on Canadian society that it can ill afford. For thesereasons it is important to better understand the factors underlying theacademic performance of first- and second-generation immigrant youth.First-generation immigrants are the foreign-born children of immigrantsand the second generation is the Canadian-born children of immigrants.The educational aspirations of both generations tend to be high,reflecting the optimism each feels toward attaining their social andeconomic goals by succeeding in the education system (Glick and White2003; Krahn and Taylor 2005; OECD 2006). However, the ability of immigrant youth to realize their broaderambitions by educational means is not assured given the rapidly changingdemographic composition of Canada's population coupled with theeconomic downturn experienced by recent immigrant cohorts. At issue is,first, the capacity of large metropolitan educational systems like theToronto District School Board (TDSB) to accommodate the diversebackgrounds and needs of immigrant youth and, second, whether immigrantfamilies have the resources to support their children's educationalendeavors. Newcomers to Canada have experienced unprecedented obstacles tosuccess that have translated into lower incomes and higher poverty ratesrelative to previous immigrant waves (Aydemir and Skuterud 2004; Bloom,Grenier, and Gunderson 1995). Research indicates, for instance, that theearnings gap between recent immigrants (i.e., those that have arrivedwithin the previous five years) and the "like" Canadian bornhas been increasing significantly with each successive cohort since the1970s (Frenette and Morissette 2003). Given that employment earningsrepresent the major source of income for most families, it is nosurprise that the decline in relative earnings has corresponded to anincrease in the proportion of new immigrants living below the low-incomecutoff (LICO). While low-income rates have fallen or remained constanteven for the most vulnerable groups among the Canadian born, the ratefor new immigrants rose from 25 to 36 percent between 1980 and 2000(Picot and Hou 2003). Moreover, the low-income situation of manynewcomers has been described as chronic. As many as 65 percent ofimmigrants can expect to experience a low-income spell within theirfirst 10 years of living in Canada and, of these, roughly one-third willremain in low income for three years or more (Picot, Hou, and Coulombe2007). The deteriorating economic position of recent immigrants hascoincided with a period of rapid demographic change in Canada(Statistics Canada 2006a, 2006b). Sustained high levels of immigrationsince the latter half of the 1980s together with the removal ofpreferential access for applicants from European countries in 1967brought about a continuous rise in the proportion of ethnic, racial,religious, and linguistic minorities. Once from the predominantly whitecountries of Northern and Western Europe, the new immigration consistslargely of those from Asian and African countries. Data from the 2006Census of Canada indicate that 58 percent of recent immigrants were fromAsian countries and another 11 percent were born in African countries.To be sure, 19.8 percent of Canada's population is foreign bornand, as of 2006, 16.2 percent of the population identified themselves asvisible minorities. In many Canadian cities, such as Toronto, Vancouver,Mississauga, and Markham, immigrants and visible minorities make upclose to one-half of the population or more (Statistics Canada 2006b,2008). Canada's immigrants represent 220 countries and nearly 150languages; fully 70 percent of Canada's foreign-born population hasa mother tongue that is neither English nor French; and it is expectedthat roughly one out of every five people in Canada will be a visibleminority by 2017, when Canada celebrates its 150th birthday (StatisticsCanada 2005, 2006b). This study employs student administrative data obtained from theTDSB to investigate the extent to which living below the LICO affectsthe likelihood of dropping out, while taking into considerationadditional risk factors associated with leaving school before graduation(e.g., country of origin, age at arrival, family structure). Further tothis, the extent to which the association between LICO and academicperformance varies by generational status is measured. The(longitudinal) research design employed involves analysis of datacollected from the year 2000 grade 9 cohort that were followed to 2006,or two years after the normal year of graduation. In employing TDSB administrative data, the study complements thefindings of previous immigrant youth studies based on nationallyrepresentative data supplied by Statistics Canada (Finnie and Mueller2008; Worswick 2001). Administrative data are particularly advantageousin that they canvas an entire population. As such, the errors that ariseas a result of survey sampling and design are avoided (Warburton andWarburton 2004). While the use of administrative data has its ownchallenges, (1) the present study makes use of a highly accurate set ofdata collected by the largest school board in Canada. In doing so, wetake advantage of a unique opportunity for a detailed investigation of alarge population of students that in many ways represents the futuredemographic profile of the country (Yau and O'Reilly 2007). PATHS TO ASSIMILATION The changing demographic composition of Canada's populationand the deterioration of immigrant labor market outcomes raise importantquestions about the socioeconomic integration of new immigrants,intergenerational mobility, and the way in which the future economicposition of the first and second generation hinges on ethnic background.The traditional theory of immigrant adaptation suggests a"straight-line" or "linear" form of assimilation,whereby time spent in the host country determines the degree ofassimilation both within and between generations (Gans 1992, 1997). Withrespect to academic achievement, the straight-line approach suggeststhat children who immigrate at a younger age will have better outcomesthan those who arrive later. Further, those belonging to the third-plusgeneration should outperform those belonging to the second generation,who in turn will have better outcomes than the first generation. (2)Consistent with this approach, age at migration has been found to be astrong predictor of academic performance among immigrants: the youngerimmigrants are upon arrival, the better their academic outcomes tend tobe (see, e.g., Cahan, Davies, and Staub 2001). Boyd (2002), for example,reported that the 1.5 generation--first-generation immigrants whoarrived in Canada before age 15--are more likely to complete high schooland tend to attain higher overall levels of education than those whoarrived later. Contrary to the straight-line hypothesis, Boyd (2002)further revealed that the educational attainment of the 1.5 generationtends to exceed that of both the second and third generations. Recognizing that the straight-line model ignores the various pathsto assimilation that various immigrant groups might follow, thesegmented assimilation hypothesis first articulated by Portes and Zhou(1993) highlights the potential for horizontal or downward assimilation,particularly in the U.S. context. Segmented assimilation theoryhighlights the different patterns of adaptation that characterizedifferent immigrant groups. Some may indeed follow the classicalstraight-line route to assimilation into the white middle class, whileothers fall into poverty and join the ranks of the underclass and stillothers will experience varying degrees of upward and downward mobilitybetween generations (Portes 1995; Portes and Rumbaut 2001; Rodriguez2002; Zhou 1997). The declining at-entry labor market position ofCanada's immigrants, no matter the reason, may be taken as anindication that the newer immigrant cohorts are experiencing moredifficultly integrating into their host society. It also raisesquestions about the extent to which future generations will be able toassimilate, which groups they will assimilate into, and the educationaland occupational pathways they will follow. The segmented assimilation hypothesis is therefore useful forguiding investigation into the educational experiences of variousimmigrant groups. Such factors as race, language ability, place ofbirth, socioeconomic status, and age at arrival are said to determinethe segment of society into which immigrants will assimilate (Zhou1997). As new arrivals become increasingly diverse, their paths toassimilation are more varied: children of the newer immigrant cohortscome from a multitude of ethnic, linguistic, religious, andsocioeconomic backgrounds, they are likely to experience some degree ofeconomic hardship while in Canada, and they come from a wide array ofnational origins. Each of these factors has been linked with educationalperformance and subsequent opportunities for upward social mobility. The impact of economic hardship on youth receives considerableattention in the growing literature on poverty, exclusion, andeducation. Educators, community advocates, and policymakers have drawnattention to the fact that student achievement is significantly lower inurban schools with high levels of poverty (Levin 2007:2). Research hasfurther documented that children reared in low-income conditions tend tohave poorer physical and mental health, experience more punitivediscipline styles and abuse, live in poorer neighborhoods, and are morelikely to be delinquent than are children raised in wealthier households(Jones et al. 2002; Luthar 1999). The Canadian Council on SocialDevelopment identified 27 elements important to child development,including family functioning, neighborhood safety, aggression, health,math and vocabulary scores, and participation in sports or clubs. For 80percent of the elements examined, family income played a critical role:among children living in families with annual incomes below $30,000, therisk of negative child outcomes and the likelihood of poor livingconditions were considerably higher than in families with higher incomes(Ross and Roberts 1999). Given that immigrants and visible minoritiesare more likely to experience poverty (Edward Herberg 1998; Kazemipurand Halli 2000, 2001a, 2001b), one would expect that these groups aremore likely to experience more negative child outcomes, included poorerschool performance and a greater incidence of dropout. An important concept to consider in understanding the complexrelationship between poverty and educational outcomes is social capital.Young people that experience sustained periods of poverty often lack thekinds of social capital or networks that are key for achieving success,first in school and, subsequently, in the labor market. There are twotypes of capital that have different benefits and consequences for youngpeople. The first, called bonding capital, consists of close family andfriends who belong to the same social, cultural, and economicbackground, while bridging networks are made up of individuals who varyin terms of their backgrounds (Putnam 2000). The former creates andsustains relationships within groups, while the latter refers to tiesthat developed between groups. It is bridging capital that providesaccess to the mainstream and to information and services that areotherwise unavailable; to this end, it is these heterogeneous bridgingnetworks that are believed to help people "get ahead." It has been argued that the poor and socially excluded are strongin bonding networks, but weak in bridging networks (Kunz and Frank2004:5). Studies, such as Poverty by Postal Code, demonstrate thatToronto has many more concentrated areas of poverty than in the past, aswell as a higher proportion of immigrants living in these poorneighborhoods (United Way of Greater Toronto and the Canadian Council onSocial Development 2004). Since 1981, there has been a 484 percentincrease in the "poor" immigrant family population living inhigh poverty neighborhoods (from 19,700 in 1981 to 115,100 in 2001) and,as of 2001, immigrant families accounted for nearly two-thirds offamilies living in Toronto's higher poverty neighborhoods.Furthermore, Toronto is home to a significant number of ethniccommunities. Given that residential segregation favors bonding overbridging, immigrant youth who live in and attend schools in poorneighborhoods and ethnic enclaves, are more likely to network or bondwith peers of similar social, cultural, and economic backgrounds.Opportunities for immigrant students to accrue bridging capital maytherefore be limited, thereby increasing the probability of pooracademic performance and leaving school early without graduating. Empirical research from Canada and elsewhere has indicated thatimmigrant and visible minority children tend to have more difficulty inschool. For instance, historically disadvantaged groups in the UnitedStates, particularly black and Hispanics, experience severe economic andacademic disadvantage upon arrival that is likely to persist acrossgenerations (see, e.g., Glick and White 2003). In Canada, researchfindings are less conclusive but, perhaps, more optimistic. Worswick(2001), for instance, found that language proficiency and age at arrivalwere particularly salient factors that influence settlement andadaptation to school life. Looking at school children up to age 15, hereported that children of immigrants perform, on average, at least aswell as children of the Canadian born along several dimensions of schoolperformance, including reading, writing, and mathematics. However, thechildren of immigrants whose first language is neither English norFrench tend to have lower reading and writing scores than children ofnative-born parents. Nevertheless, with more years spent in the Canadianschool system, reading and writing test scores tend to converge (seealso Corak 2005). However, looking specifically at students for whoEnglish was not their mother tongue, Gunderson (2007) found starkdifferences in the academic performance of different ethno-culturalgroups. Based on a sample of 5,000 English as a Second Language (ESL)students enrolled in the Vancouver school system between 1991 and 2001,Gunderson (2007) revealed that Mandarin and Cantonese speaking studentsin grades 8 through 12 outperformed English-speaking Canadians in allsubjects with the exception of grade 12 English, while Indian-,Vietnamese-, Tagalog-, and Spanish-speaking students generally performedless well than the Canadian born. Though the paths to assimilation appear to be varied for immigrantsand their offspring in both Canada and the United States, the contextualfactors facing Canadian immigrants are very different, and it has beenargued that there is little evidence of second generation decline inCanada (Boyd 2002; Boyd and Grieco 1998). Boyd (2002) points to the lackof urban ghettos and the smaller black and Hispanic populations inCanada relative to the United States, and suggests that downwardassimilation is unlikely simply because there is no identifiableunderclass. Boyd (2002) also points out that the greater proportion ofimmigrants entering Canada may create and sustain a "criticalmass" that supports education as a tool for upward social mobilityamong immigrants and their children. Yet the unlikelihood of segmentedassimilation occurring en mass in Canada does imply that upward mobilityis inevitable. Relative to their parents, "horizontalmobility" of the second generation may be more likely (Alba and Nee2003:268). Recent successions of immigrant cohorts entering Canada arefacing lower at-entry earnings, higher rates of unemployment, and higherlow-income rates while the native born are enjoying decliningunemployment and poverty rates (Aydemir and Skuterud 2004; Picot andSweetman 2005). Despite the improbability of assimilation into an"underclass," the economic outlook for children of the foreignborn may not be much brighter than that of their parents. IMMIGRANT STATUS AND THE SCHOOL EXPERIENCE IN ONTARIO The school experience of first- and second-generation youth is aparticularly important issue in Ontario, where one-half of allimmigrants to Canada settle and 28 percent of the population is foreignborn. With respect to youth, 19 percent of Ontario's populationaged 15 to 24 was foreign born in 2006; in Toronto, this figure was 40percent (Statistics Canada 2007). Its large foreign-born population,particularly in urban areas, means that Ontario can not afford tooverlook the disadvantaged socioeconomic position of its immigrants andthe long-term consequences this might have for their children. Data from Statistics Canada's Youth in Transition Survey(2000) indicate that, in Ontario, 11.2 percent of 20-year-old males and7.8 percent of 20-year-old females were not in school and had yet tocomplete the requirements for a high school diploma (Bowlby and McMullen2002). Within Ontario, high school dropout rates vary substantially byfamily income. Data from 2003 show that, among those in the lowestincome quartile, the high school dropout rate at age 19 was 8.3 percent.This is three times higher than the dropout rate of 2.6 percent amongthose in the highest income quartile. Furthermore, the postsecondaryparticipation rate at age 19 was 40 percent higher for those in thehighest income quartile relative to those in the lowest quartile (Zeman2007). When asked about their main reason for dropping out, school factorsare most commonly cited by early school leavers. A 2002 survey of17-year-old Canadians indicated that nearly 45 percent of those who haddropped out of school attributed their departure to the schoolenvironment. School-related factors include boredom or lack of interestin classes, difficulties with school work and with teachers, expulsion,and missing credits (Bushnik, Barr-Telford, and Bussiere 2004). Acomparison of school leavers and school continuers revealed that thereading proficiency of dropouts were one full level below the average,as defined by the Program for International Student Assessment. (3)Early school leavers also reported much lower grades; among those whohad dropped out by age 17, 32 percent reported an overall grade of lessthan 59 at age 15, compared with 8 percent of other students (Bushnik etal. 2004). A study commissioned by the Ontario Ministry of Education andTraining (Hospital for Sick Children 2005) revealed that first- andsecond-generation youth in Toronto and Kitchener-Waterloo experienceunique challenges in secondary school. In-depth qualitative interviewswere conducted with 57 first- and second-generation youth who had leftschool early or were at risk of doing so. Respondents cited the need tolearn a new language, language barriers, unfamiliarity with the Canadianschool system, and inappropriate linguistic assessment and gradeplacement as important risk factors for school disengagement. Stressesassociated with resettlement, loneliness, isolation, and a lack offriends were also reported. The study further demonstrated that age atthe time of migration was especially critical, whereby youth whoimmigrated during the latter years of high school were most at risk ofdropping out. DATA, VARIABLES, AND METHODS Access to student level data was provided by the External ResearchReview Committee of the TDSB. A single grade 9 cohort that began highschool in 2000 was tracked over a six-year period. Seventy-nine percentof this cohort had completed elementary school within the TDSB. Theremaining 21 percent arrived from other school boards, both within andoutside of Canada. The original grade 9 cohort was made up of 18,798students. By October 2006, the official end of year 6 of the study,2,220 students had transferred out of the TDSB to other secondaryinstitutions and could no longer be tracked. Hence, they were omittedfrom the analysis. Another 329 students were removed due to codingerrors, leaving a sample of 16,249 students. By the end of their sixth year of secondary study, 72 percent ofstudents in the sample had graduated with an Ontario Secondary SchoolDiploma (OSSD) or successfully completed 30 or more credits. Two percenthad not graduated but remained in a secondary school for a seventh yearand another 26 percent had dropped out. Dropouts are classified as thosestudents who had left secondary school without having graduated ortransferred to another secondary institution. The administrative data set contains a series of variables thatmeasure a variety of sociodemographic characteristics, including gender,region of birth, language, family status, and the age at which eachstudent entered high school. Region of birth distinguishes among sevenregions, including Canada, Europe, English-speaking Caribbean, Africa,South Asia, West Asia, and Eastern Asia. Respondents born in Canada arefurther divided into two groups: those who speak English at home andthose who do not. First-generation immigrant status was thus defined asbeing foreign born, second generation as being born in Canada but notspeaking English in the home, and third generation as being born inCanada and speaking English in the home. Family status measures thefamily situation of students in their third year of high school andcategorized into two groups: those who live with both parents and thosewho do not. Finally, a variable based on age was included as anindication of whether students began high school at the expected time orif they began late. As noted above, a variety of studies have documented the negativeimpact of poverty on student achievement (Ornstein 2000; United Way ofGreater Toronto and the Canadian Council on Social Development 2004). Tocapture poverty, a variable that measures the proportion of people inthe respondent's immediate neighborhood that fall below the LICO isincluded. This variable was derived from student postal codes that werematched with their dissemination area (DA): the proportion of thepopulation living below the LICO, as reported by the 2001 Census, wasassigned to each student based on the DA in which they lived. Thevariable is coded in deciles by the TDSB, such that value 1 indicatesthe highest incidence (proportion) of residents living below the povertyline, whereas 10 indicates the lowest incidence of residents livingbelow the poverty line. Hence, a higher score means that a respondentlives in a more affluent neighborhood. Finally, three independent variables that provide information onvarious aspects of schooling at the student level were included. Thefirst variable reflects streaming within secondary school. Streamingrefers to the majority of courses taken in grades 9 and 10, and isemployed to classify the student's program of study as academic,applied, or essentials. Under the Ontario secondary school curriculumintroduced by the Ministry of Education in the Fall of 1999, studentsare to choose a program of study that includes grades 9 and 10 coursesthat are classified as academic (university-directed), applied(college-directed), or locally developed essentials(workplace-directed). As with previous studies, the present analysiscategorizes a student's program of study as academic, applied, oressentials based on the majority of courses taken in grades 9 and 10. The second independent variable indicates whether or not a studentis considered to be "at risk." A student is classified as"at risk" if he or she had completed fewer than seven coursesby the end of grade 9. Last, the third variable distinguishes betweenstudents who have taken ESL courses and those who have not. (4) Thisvariable also represents a proxy for language proficiency. The descriptive statistics for the variables used in this analysisare displayed in Table 1. Frequencies are provided for categoricalvariables and means are provided for quantitative variables. Thedescriptive statistics are provided separately for each region oforigin. With the exception of gender, there are statisticallysignificant differences across region of origin for all variables in theanalysis. The most noteworthy findings are discussed below. In terms of dropout levels, students from the Caribbean had thehighest dropout rates (40 percent), whereas students of Eastern Asiawere least likely to dropout of high school (10 percent).English-speaking Canadian students were in between, approximately 20percent dropped out of high school. With regard to age at entering highschool, English-speaking Canadian-born students were most likely toenter on time (97 and 98 percent, respectively), whereas Caribbean andAfrican students were least likely to enter on time (88 and 89 percent,respectively). With respect to family status, European students weremost likely to live with both parents (74 percent), in contrast toCaribbean students who are most likely to live in another familystructure. Just 26 percent of students from the Caribbean lived in twoparent families. Among English-speaking Canadian born students, exactlyone-half lived with both parents. Turning to the academic variables, descriptive results reveal thatstudents from East Asia were predominantly enrolled in the academicstream (90 percent), followed closely by students from Europe (85percent), English-speaking Canada (78 percent), and students from SouthAsia (78 percent). In contrast, Caribbean immigrants were least likelyto be in the academic track, at just 39 percent. Similar patternsemerged for the "at-risk" variable, whereby students from EastAsia (7 percent) and Europe (10 percent) were least likely to be labeled"at risk," and students from the Caribbean (33 percent) weremost likely to be considered "at risk." Canadian-bornEnglish-speaking students were somewhere in the middle, as 14 percenthad not completed seven or more credits by the end of grade 9. A slightly different pattern emerges with respect to having takenan ESL course. As expected, English-speaking Canadian-born students wereleast likely to have taken an ESL course (< 1 percent). (5) Incontrast, approximately 28 percent of East Asian immigrants had taken anESL course, followed by West Asian (26 percent) and South Asianimmigrants (22 percent). In comparison, 10 percent of Caribbeanimmigrants have taken an ESL course. Finally, using the LICO as an indicator of economicallydisadvantaged neighborhoods, it appears that African immigrants lived inthe most disadvantaged areas, followed by South Asian immigrants andthen by West Asian immigrants. In contrast, English-speaking studentsborn in Canada tended to reside in neighborhoods with the lowestpercentage of families living below the poverty line. Overall, the most consistent pattern revealed by the descriptivestatistics is that East Asian and European immigrants are generally inthe most favorable positions in terms of sociodemographic andschool-related characteristics, whereas Caribbean immigrants are in themost disadvantaged positions. English-speaking students born in Canadatend to fall in between the two extremes, though it is safe to say thatthey are most certainly at an advantage in terms of theirsociodemographic and academic profiles. Regression Results The response variable in our analysis is an indicator of whetherthe respondent had dropped out of the school system. Respondents areconsidered dropouts if they had not graduated by 2006. (6,7) For theregression analysis, we estimate a multilevel model in which individuals(level 1) are nested within neighborhoods (level 2), where neighborhoodis defined according to the DAs in which the students live. Theneighborhood-level variable used in this study is the proportion of thepopulation living below the LICO, as defined by Statistics Canada. Theresponse variable is a binary variable that distinguishes between thosewho dropped out of high school and those who did not. To regress thelevel 1 outcome (dropout) on both level 1 and level 2 predictors, weemploy a mixed logit model. The Bernoulli distribution is specified forthe response variable and a logit link is used to map the mean of theresponse variable to the linear predictor. Then logit link is defined as [[eta].sub.ij] = log ([[PHI].sub.ij]/1 - [[PHI].sub.ij]) where [[phi].sub.ij] is the predicted probability of dropping outfor the ith observation in neighborhood j, and [[eta].sub.ij] is the logodds of dropping out. To estimate the magnitude of variation between neighborhoods indropout levels, we first estimate an unconditional model without anypredictors at either level (Model 1). Since the level 1 variance isheteroskedastic, the intraclass correlation is not as intuitive as it isin the standard hierarchical linear model. Nevertheless, it is still auseful index because it represents the ratio of the level 2(neighborhood) variance to the total variation. In models with binaryoutcomes, the intraclass correlation is best considered in relation tothe latent variable approach, where the level 1 random effect is assumedto have a standard logistic distribution with a mean of 0 and varianceequal to [[pi].sup.2]/3. (8) Using conventional notation the level 1model is specified as [[eta].sub.ij] = [[beta].sub.0j] and the level 2 model is [[beta].sub.0j] = [[gamma].sub.00] + [[mu].sub.0j] where [[mu].sub.0j] ~ N (0, [[tau].sub.00]) In the second equation, [[gamma].sub.00] represents the averagelog-odds of dropping out across the neighborhoods, and [[mu].sub.0j] isthe random effect at level 2. The last term indicates that we areadopting the usual assumption that the error term at level 2,[[mu].sub.0j], is normally and identically distributed with an expectedvalue of 0 and a constant variance, [[tau].sub.00]. This assumption isapplied to all models estimated in this paper. The estimates from Model 1 are provided in the first column ofTable 2. The key estimate in this model is the intraclass correlation,[rho], which indicates that approximately 13 percent of the variation inthe outcome can be attributed to neighborhood characteristics(p<.001). Since it is highly statistically significant, we proceed toinclude a random effect at level 2 in Model 2. (9) The region of origin variable is the only variable included inModel 2, in which the level 1 structural model is specified as [[eta].sub.ij] = [[beta].sub.0j] + [[beta].sub.1j][X.sub.1ij] + ...+ [[beta].sub.kj][X.sub.kij] where [[beta].sub.1j] through [[beta].sub.kj] are the parametersrepresenting the six dummy coded variables for the region of originvariable. The level 2 model is [[beta].sub.0j] = [[gamma].sub.00] + [[mu].sub.0j] The parameters for the dummy coded variables are treated as fixed(i.e., [[beta].sub.pj] = [[gamma].sub.p0] for p>0). The likelihoodratio chi-square test for the region of origin variable is statisticallysignificant (p<.001) and the parameter estimates in Model 2 areinterpreted as the log-odds of dropping out of high school relative tothe reference category, English-speaking Canadian-born respondents. Incomparison with English-speaking Canadian students, only immigrants ofthe Caribbean are more likely to dropout of high school (p<.001). Incontrast, immigrants that are less likely to dropout than the referencegroup are students from Europe (p<.01), South Asia (p<.01), andEastern Asia (p<.001). Perhaps most interestingly, second generationCanadians are no more or less likely to dropout of high school than arefirst-generation Canadians. In Model 2, the estimated variance at level2 remains statistically significant (p<.001), as approximately 11percent of the total variation in dropout levels is attributable toneighborhood characteristics after controlling for country of origin. Model 3 includes the remaining level 1 variables and the level 2variable LICO. (10) The specification of the level 1 structural model is [[eta].sub.ij] = [[beta].sub.0j] + [[beta].sub.1j][X.sub.1ij] + ...+ [[beta].sub.kj][X.sub.kij] where [[beta].sub.1j] through [[beta].sub.kj] are now used toconveniently denote the parameters for all of the quantitative andcategorical dummy coded explanatory variables in the model. At theneighborhood level, only the intercept [[beta].sub.0j] is a function ofthe level 2 predictor [W.sub.j], which is our measure of LICO: [[beta].sub.0j] = [[gamma].sub.00] + [[gamma].sub.10][W.sub.j] +[[mu].sub.0j] whereas all of the other parameters are treated as fixed. Hence, [[beta].sub.pj] = [[gamma].sub.p0] for p > 0 Most of the variables included in the model are statisticallysignificant (p<.001), while holding constant the value of the randomeffect, [[mu].sub.0j]. (11) The only exception is the variable whichdistinguishes between respondents who have taken ESL classes, which isnot statistically significant. The magnitude of the estimates for the region of origin variable isreduced in Model 3; however, the pattern of estimates is similar toModel 2. It should be noted that the country of origin estimates arereduced when the remaining variables are included in the model. Thus,much of the impact of region of origin appears to be due to othervariables in the analysis. The most noteworthy change occurs amongCaribbean immigrants, as their relative chances of dropping out declinedramatically when the control variables are included. In fact theirdropout levels are no longer significantly different fromEnglish-speaking Canadian-born students. The relative chances ofdropping out also decline for students born in Africa. When the controlsare included in the model, their dropout levels become significantlylower than Canadian-born students (p<.05), as are the dropout levelsof students from South Asia (p<.01) and Eastern Asia (p<.001).Similar to the findings obtained in Model 2, there are no differencesbetween first- and second-generation Canadians in terms of theirlikelihood of dropping out of high school. With respect to the level 2 variable, LICO, respondents residing inneighborhoods with lower proportions of residents living below thepoverty line (i.e., lower poverty) are less likely to dropout of highschool than are respondents residing in neighborhoods with higherproportions of residents living below the poverty line. (12) The estimates for the other variables that are statisticallysignificant are also in the expected direction. For example, males aremore likely than females to dropout. This is consistent with theliterature on school dropout which shows that, despite a recent decreasein the overall dropout rate, gender differences remain. In 1990/1991,just over one-half of dropouts were male (58.3 percent); by 2004/2005,the proportion had increased to 63.7 percent (Bowlby 2005). In terms ofthe academic achievement variables, students in the academic program,the reference category, are least likely to dropout of high school,whereas students in the essentials program have the highest probabilityof dropping out. This finding is generally consistent with past research(Human Resources Development Canada 2000). King et al. (1988) haveobserved that the levels at which courses are taken by secondary schoolstudents is the best predictor of dropping out. The parameter estimatefor the family structure variable reveals that students in two parentfamilies are less likely to dropout of high school than students livingin other family structures. Similarly, students who start high school on time are less likelyto dropout than students who begin a year late. Finally, students whoare classified as "at risk," that is, those who completedfewer than seven credits in grade 9, are more likely to dropout thanstudents who completed seven or more credits in grade 9. Whencontrolling for the explanatory variables, the proportion of varianceattributable to level 2 (LICO) is reduced to approximately 5 percent,but nevertheless remains statistically significant (p < .05). DISCUSSION AND CONCLUSION High school graduation is a prerequisite to advanced education andtraining in Canada. Consequently, the educational and occupationalfutures of those who dropout of high school are severely restricted.Immigrant adolescents generally recognize the importance of furthereducation and invest considerable effort in their high school studies(Krahn and Taylor 2005). However, not all newcomer youth are successfulin school and those who exit before graduation represent a significantcost to their parents. Canada too pays economic and social penaltieswhen immigrant children fail to integrate into the school system,perform well, and subsequently contribute to the broader society. Education, then, is a key factor in the integration process ofimmigrant children. This integration process has been described bylinear and segmented assimilation theories. Explanations for thevariation in school performance among immigrant children and youth thenhave focused primarily on the amount of time that has passed since theirarrival or, alternatively, by socio cultural differences that shape theinteractions between individuals and schools. Where integration isviewed as a linear progression, earlier generations are expected toperform better than newer generations, and those who arrive at a youngage will have better school outcomes than older immigrant youth.Differences in school performance are therefore assumed to be a functionof institutional exposure, as indicated by age at arrival or time spentin school. When considered from the perspective of cultural values andlinguistic differences in proficiency, the school performance ofimmigrant children and youth is expected to vary by source country orregion. This study explored both generational-difference andcultural-difference explanations for immigrant dropouts. We furtherexamined the extent to which barriers to school completion reflected theindividual characteristics, personal situations, and economic resourcesfound among immigrant youth and their families. Results indicate little support for the straight line assimilationmodel as it applies to academic achievement among immigrant youth.Rather, findings supported the segmented assimilation hypothesis. Regionof origin was a significant predictor of dropout when first-generationyouth were compared with the native born. Students from the Caribbeanwere significantly more likely than native-born English students todropout of school, while students from Europe, Eastern Asia, and SouthAsia were less likely to leave school early. Underlying cultural factorsare often associated with regional differences in immigrant outcomes.Cultural differences between their region of origin and the host societypresent newcomers with challenges to adaptation and integration into anew school environment. In the case of Caribbean immigrants, youth oftenfind themselves isolated, falling behind and falling in school, andgrowing increasingly frustrated (Anisef and Kilbride 2003). It isimportant to note that the dropout rate among such groups declines whenother factors are considered, including school conditions and curricularpolicies. Neighborhood effects were found also to be statisticallysignificant. Initial findings revealed that 13 percent of the variationin the odds of dropping out can be attributed to neighborhood levelfactors. Furthermore, close to two-thirds of the neighborhood-levelvariance was explained by the poverty indicator. Specifically, higherdropout rates were found among youth living in neighborhoods with ahigher proportion of residents living below the LICO. It is important topoint out that that this effect was statistically significant despitecontrols for region of origin and the various individuallevel factors.Given that increased numbers of immigrant youth are living below thepoverty line--and therefore in neighborhoods with cheaper housing andpoorer residents--this finding has important implications that requireattention by researchers and policymakers at all levels of government. Individual differences also influenced the dropout rate. Genderdifferences found in this study parallel those found in the literatureon male "underachievement." The effect of family structurealso is consistent with the general literature. Single parent familiesgenerally possess fewer material and social-emotional resources thatpromote student achievement. Several factors that describe the academic potential andperformance of children were also included. "Age of entry"indicates whether the student entered secondary school at the modal ageof native-born children. Late entry may result from additional timeneeded by newcomers to adjust to the TDSB classroom or because of pooracademic performance of those immigrant children who arrived at anearlier age but who struggled with the elementary school program. Thosewho have failed to accumulate the required credits by grade 9 are moreinclined to dropout of high school. Both adjustment and academicachievement require language competence. It is, therefore, interestingthat relatively few immigrant students take an ESL course in highschool. This is consistent with Gunderson's (2007) work, whichfinds many immigrants are reluctant to enroll their children in ESL asit limits the time available to study the core-curriculum courses.Schoolrelated factors reflect district policies and practices oropportunity structures available within the system. The most salient ofthese in the general literature relates to the academic streaming ofchildren. Those who enter the "vocational" stream are morelikely to dropout than those who elect to follow the"university" pathway. Choice of a school pathway is determinedby several factors. However, the effects of streaming are of particularimportance to immigrant children who may take a lengthy period of timebefore adjusting to the Canadian social and educational norms andpractices. Policy Initiatives Recent OECD reviews of school achievement and immigrant adjustmentsuggest several school and community practices designed to facilitatethe integration of immigrant children and youth. These include earlyintervention with preschoolers to develop language skills; programsdesigned to promote the social adjustment of youth; and the removal ofstreaming programs that essentially sort students into vocational anduniversity paths according to ability. Many of these programs take intoconsideration differences in the performance and needs of first- andsecond-generation students (OECD 2007). While this analysis suggests that our proxy for generational statusbears little relationship to educational outcomes among immigrant youthin the TDSB, region of origin nevertheless exerts a significantinfluence on school completion that is substantially reduced whenvarious demographic characteristics and measures of achievement arecontrolled. This finding is important because it provides markers fordevising strategies that may lower dropout rates among specificimmigrant groups from diverse countries of origin. For instance,students from the Caribbean are significantly more likely to enterschool one year late, live in alternate family structures, findthemselves placed in nonacademic streams and be at risk of notcompleting their course of study. Many of these risk factors areresponsive to change by working effectively with schools and family. Forinstance, special transition programs might be considered for studentsentering school late as a means to facilitate adaptation to the socialand academic life of Canadian schools. In the following section, we consider a number of initiatives thateither have been implemented or could be implemented at the federal,provincial, and local levels. Federal Citizenship and Immigration Canada offers adult newcomers a Hostprogram which facilitates their integration by matching them withCanadian volunteers who help with language barriers, with gettingcontacts in their field of work, and with everyday interactions, such asbanking, grocery shopping, enrolling in school, and using the transitsystem. Anisef et al. (2007) recommend that Host be extended to newcomeryouth entering the Canadian school system. Newcomer youth would beprovided a "buddy" or "mentor" to help them bettermanage the difficulties associated with resettlement. Such a programmight be further complemented by calling upon school counselors to workalongside buddies or mentors to address issues associated withadaptation and school performance. In addition, school staff might workclosely with the families of those youth most at risk of leaving schoolearly in order to encourage both academic and social engagement withinthe school environment. Provincial An important initiative launched by the Ontario Ministry ofEducation in 2003 is the Student Success/Learning to 18 Strategy. Thestrategy was designed to ensure that all students successfully completetheir secondary schooling with the knowledge and dispositions requiredto make effective transitions to postsecondary education or employment(Ungerleider 2007). The origins and motivations of the strategy can betraced in part to reactions to a double-cohort longitudinal study byAlan King (King 2002, 2003; King et al. 2005), which cited lowgraduation rates within the province and identified credit accumulationin grade 9 and 10 as a key predictor of graduation. This researchmotivated the development of specific programs to help all studentsacquire the required number of secondary school credits and subsequentlygraduate from secondary school. While school based programs are a keyelement of the Student Success initiative, there is recognition thatprograms must extend beyond the school. Thus, programs have also beendeveloped with the community with parents, employers, communityagencies, and organizations to help inform decision making, and createopportunities for experiential learning. Although the StudentSuccess/Learning to 18 Strategy is in its early stages ofimplementation, a formal evaluation by the Canadian Council on Learningindicates that it is working, providing students with a more respectfuland responsive environment and offering more choices for students notbound for university (Ungerleider 2007:64). Neighborhood Interventions As noted above, the proportion of low-income families living instudent neighborhoods was found to be a significant predictor ofdropping out. Linking student postal codes with low-income data from theCensus measured at the DA level essentially assigns each student aprobability of living in a lowincome family. As such, it remains unclearas to whether this variable is capturing poverty at the neighborhoodlevel (as implied by the model), the family level, or both. To be sure,both family- and community-level socioeconomic status have been linkedto student achievement (see, e.g., Pong and Hao 2007; Webber and Butler2007). The importance of incorporating the neighborhood into schoolinterventions is at the heart of Pathways to Success, a program aimed athighly at-risk students. The highly successful program piloted in RegentPark in downtown Toronto was organized with local community leadership,and since then each additional program has been organized around andinvolving a specific urban community. In many cases (as in Regent Park)these communities contain diverse immigrant populations. Secondary and Postsecondary Interventions Interventions can take place at different parts of the transitionsprocess; in fact, there is a number of ongoing school, government, andprivate programs for immigrant students who are at risk of poorachievement. The Settlement Workers in Schools program was initiallypiloted in the TDSB and has since been implemented in several Ontarioschool boards with large immigrant populations (Ontario Council ofAgencies Serving Immigrants 2009). Among its main goals is to provideassistance to students and parents navigating the Ontario school systemfor the first time. A second Ontario initiative (part of the StudentSuccess/Learning to 18 Strategy) examines the direct transition into theworkplace as well as the process by which students move from the schoolsand the workplace into community colleges. For example, dual creditopportunities leading directly to college and apprenticeships areavailable to TDSB and Toronto Catholic District School Board students inpartnership with all community colleges in the Greater Toronto Areathrough the School/College/Work/Initiative (SCWI). The SCWI is astrategy of Student Success/Learning to 18 initiatives introduced by theMinistry of Education to address the number of students in Ontario whoat risk of not graduating on time. Only students in approved SCWI dualcredit programs may count dual credits toward the OSSD. All programs arebased on collaboration and partnership between school boards andcolleges. While these projects do not specifically target immigrants,the importance and size of the immigrant population within centralOntario means that they are an essential piece of the puzzle. An obviousnext step would involve an examination of workplace and postsecondarytransitions in terms of similarities and differences of subgroups withinthis very diverse population, similar to what has been done here lookingat dropouts and graduation. Future Research Directions The current study contributes to increasing interest in and use ofadministrative data to better specify the conditions that shape theschool performance of immigrant youth. At present, there are a number ofstudies that investigate adolescent and early adult schooling outcomesamong immigrants to Canada and elsewhere. Typically, these employnationally representative samples that do not allow a focused,contextualized examination of immigrant youth (e.g., Finnie and Mueller2008; Worswick 2001). By using detailed school administrative data thisstudy was able to examine dropouts among a grade 9 student cohort in asingle metropolitan city with an ethnically diverse school populationconsisting primarily of first- and second-generation immigrants (Yau andO'Reilly 2007). To the extent that the student body of the TDSB ispredominantly immigrant, it anticipates the future demographic profileof the province and the country. The value of administrative data in studying this emerging culturaldiversity among the school-age population is seen in ongoing researchinitiatives undertaken at the regional and national level. There arecurrently a number of projects throughout Ontario where schools boardsare combining their data in similar ways, encouraged by the Associationof Educational Researchers of Ontario, the professional body of Ontarioboard researchers, and the Ontario Ministry of Education. In addition, apilot project used administrative data on grade 9 student cohorts fromToronto, Vancouver, and Montreal to compare achievement and schoolcompletion (McAndrew et al. 2009). The basis for comparison--languagespoken at home--revealed similarities in the pattern of achievementbased on country-of-origin differences. That project allowed researchersto look at similarities and differences in how different culturesinteract with educational systems in the three largest Canadian cities.Future research that employs administrative data promises to contributeto the challenging task of explaining the progress of immigrant youthgroups (defined by country of origin and home language) through theCanadian school systems and their transitions to postsecondary educationand the workplace. References Alba, R.D. and V. Nee. 2003. Remaking the American Mainstream:Assimilation and Contemporary Immigration. Cambridge, M.A: HarvardUniversity Press. Anisef, P. and K.M. Kilbridge. 2003. Managing Two Worlds: TheExperiences and Concerns of Immigrant Youth in Ontario. Toronto, ON:Canadian Scholars Press. Anisef, P., M. Poteet, D. Anisef, G. Farr, C. Poirier and H. Wang.2007. "Issues Confronting Newcomer Youth in Canada: AlternativeModels for a National Youth Host Program." CERIS Policy Matters No.29. Aydemir, A. and M. Skuterud. 2004. Explaining the DeterioratingEntry Earnings of Canada's Immigrant Cohorts: 1966-2000. Ottawa,ON: Statistics Canada. Bloom, D.E, G. Grenier and M. Gunderson. 1995. "The ChangingLabour Market Positions of Canadian Immigrants." Canadian Journalof Economics 28:987-1005. Bowlby, G. 2005. "Provincial Drop Out Rates--Trends andConsequences." Education Matters, Statistics Canada, Cat. 81-004XIE. Bowlby, G. and K. McMullen. 2002. At a Crossroads: First Resultsfor the 18-20 Year Old Cohort of the Youth in Transition Survey. Ottawa,ON: Human Resources Development Canada. Boyd, M. 2002. "Educational Attainments of ImmigrantOffspring: Success or Segmented Assimilation?" InternationalMigration Review 36:1037-60. Boyd, M. and E.M. Grieco. 1998. "Triumphant Transitions;Socioeconomic Achievements of the Second Generation in Canada."International Migration Review 32:853-76. Bushnik, T., L. Barr-Telford and P. Bussiere. 2004. In and Out ofHigh School: First Results from the Second Cycle of the Youth inTransition Survey. Ottawa, ON: Statistics Canada. Cahan, S., D. Davies and R. Staub. 2001. "Age at Immigrationand Scholastic Achievement in School-Age Children: Is There a VulnerableAge?" International Migration Review 35:587-95. Corak, M. 2005. "Equality of Opportunity and Inequality Acrossthe Generations: Challenges Ahead." Policy Options 26:78-83. Finnie, R. and R. Mueller. 2008. "Access to Post-SecondaryEducation in Canada among First and Second Generation CanadianImmigrants: Raw Differences and Some of the Underlying Factors."Working Paper. Montreal: Millennium Foundation. Frenette, M. and R. Morissette. 2003. Will They Ever Converge?Earnings of Immigrant and Canadian-Born Workers over the Last TwoDecades. Ottawa, ON: Statistics Canada. Gans, H. 1992. "Second Generation Decline: Scenarios for theEconomic and Ethnic Futures of the Post-1965 American Immigrants."Ethnic and Racial Studies 15:173-91. Gans, H. 1997. "Towards a Reconsideration of'Assimilation and Pluralism': The Interplay of Acculturationand Ethnic Relations." International Migration Review 31:875-92. Glick, J.E. and M.J. White. 2003. "The Academic Trajectoriesof Immigrants Youths: Analysis within and Across Cohorts."Demography 40:759-83. Gunderson, L. 2007. English Only Instruction and Immigrant Studentsin Secondary Schools: A Critical Examination. Mahwah, NJ: LawrenceErlbaum Associates. Herberg, E.R. 1998. "The Ethno-Racial Socioeconomic Hierarchyin Canada: Theory and Analysis of the New Vertical Mosaic."International Journal of Comparative Sociology 31:206-21. Hospital for Sick Children. 2005. Early School Leavers:Understanding the Lived Reality of Students Disengagement from SecondarySchool: Final Report. Toronto, ON: Ontario Ministry of Education andTraining, Special Education Branch. Human Resources Development Canada. 2000. Dropping Out of HighSchool: Definitions and Costs. Ottawa, ON: Applied Research Branch,Author. Jones, C., L. Clark, J. Grusec, R. Hart, G. Plickert and L.Tepperman. 2002. Poverty, Social Capital, Parenting and Child Outcomesin Canada. Applied Research Branch, Strategic Policy. Ottawa: HumanResources Development Canada. Kazemipur, A. and S.S. Halli. 2000. The New Poverty in Canada:Ethnic Groups and Ghetto Neighbourhoods. Toronto, ON: ThompsonEducational Publishing. Kazemipur, A. and S.S. Halli. 2001a. "The Changing Colour ofPoverty in Canada." Canadian Review of Sociology and Anthropology38:217-38. Kazemipur, A. and S.S. Halli. 200lb. "Immigrants and the NewPoverty: The Case of Canada." International Migration Review38:1129-56. King, A. 2002. "Double Cohort Study: Phase 2 Report for theOntario Ministry of Education." Retrieved February 24, 2010(http://www.edu.gov.on.ca/eng/document/reports/cohortph2.pdf). King, A. 2003. "Double Cohort Study: Phase 3 Report for theOntario Ministry of Education." Retrieved February 24, 2010(http://www.edu.gov.on.ca/eng/document/reports/phase3/report3.pdf). King, A.J.C., W.K. Warren, J.C. Boyer and P. Chin. 2005."Double Cohort Study: Phase 4." Report submitted to theOntario Ministry of Education. Retrieved February 24, 2010(http://www.edu.gov.on.ca/eng/policyfunding/reports.html). King, A.J.C., W.K. Warren, C. Michalski and M.J. Pearlt. 1988.Improving Student Retention in Ontario Secondary Schools. Toronto:Ministry of Education. Krahn, H. and A. Taylor. 2005. "Resilient Teenagers:Explaining the High Educational Aspirations of Visible-Minority Youth inCanada." Journal of International Migration and Integration6:405-34. Kunz, J.L. and J. Frank. 2004. "Poverty, Thy Name isHydra." Horizons: Policy Research Initiative 7:4-8. Levin, B. 2007 "Enduring Issues in Urban Education."Paper presented to the Canadian Society for the Study of Education, May,Saskatoon, Saskatchewan. Luthar, S. 1999. Poverty and Children's Adjustment. ThousandOaks, CA: Sage Publications. McAndrew, M. et al., 2009. "Educational Pathways and AcademicPerformance of Youth of Immigrant Origin: Comparing Montreal, Torontoand Vancouver." Report submitted to the Canadian Council onLearning/Citizenship and Immigration Canada. OECD. 2006. Where Immigrant Students Succeed--A Comparative Reviewof Performance and Engagement in PISA 2003. Programme for InternationalStudent Assessment. Paris: OECD. OECD. 2007. No More Failures: Ten Steps to Equity in Education.Education & Skills Number 15. Paris: OECD. Ontario Council of Agencies Serving Immigrants. 2009."Settlement Workers in Schools (SWIS): BackgroundInformation." Retrieved March 10, 2009(http://atwork.settlement.org/sys/atwork_library_detail.asp?doc_id=1003365). Ornstein, M. 2000. Ethno-Racial Inequality in Metropolitan Toronto:An Analysis of the 1996 Census. Toronto, ON: Municipality ofMetropolitan Toronto, Access and Equity Centre. Picot, G. and F. Hou. 2003. The Rise in Low-Income Rates amongImmigrants in Canada. Ottawa, ON: Statistics Canada. Picot, G., F. Hou and S. Coulombe. 2007. Chronic Low Income andLow-Income Dynamics among Recent Immigrants. Ottawa, ON: StatisticsCanada. Picot, G. and A. Sweetman. 2005. "The Deteriorating EconomicWelfare of Immigrants and Possible Causes: Update 2005." AnalyticalStudies Branch Research Paper Series, Statistics Canada, Cat. No.11F0019MIE-No. 262. Pong, S. and L. Hao. 2007. "Neighborhood and School Factors inthe School Performance of Immigrants' Children." TheInternational Migration Review 41:206-41. Portes, A. 1995. "Economic Sociology and the Sociology ofImmigration: A Conceptual Overview." Pp. 1-41 in The EconomicSociology of Immigration, edited by A. Portes. New York: Russell SageFoundation. Portes, A. and R.G. Rumbaut. 2001. Legacies: The Story of theImmigrant Second Generation. New York: Russell Sage Foundation. Portes, A. and M. Zhou. 1993. "The New Second Generation:Segmented Assimilation and It's Variants." Annals of theAmerican Academy of Political and Social Science 530:74-96. Putnam, R. 2000. Bowling Alone: The Collapse and Revival ofAmerican Community. New York: Simon and Schuster. Rodriguez, T.D. 2002. "Oppositional Culture and AcademicPerformance among Children of Immigrants in the USA." Race,Ethnicity and Education 5:199-215. Ross, D.P. and P. Roberts. 1999. "Income and Child Well-being:A New Perspective on the Poverty Debate." Canadian Council forSocial Development, Report No. 552. Statistics Canada. 2005. The Daily. Ottawa, ON: Statistics Canada. Statistics Canada. 2006a. The Daily. Ottawa, ON: Statistics Canada. Statistics Canada. 2006b. Census of Population. Ottawa, ON:Statistics Canada. Statistics Canada. 2007. The Daily. Ottawa, ON: Statistics Canada. Statistics Canada. 2008. "Community Profiles." RetrievedSeptember 30, 2009 (http://www.statcan.ca). Ungerleider, C. 2007. "Evaluation of the Ontario Ministry ofEducation's Student Sucess/ Learning to 18 Strategy, Stage 1Report." Canadian Council on Learning, Vancouver. United Way of Greater Toronto and the Canadian Council on SocialDevelopment. 2004. "Poverty by Postal Code: The Geography ofNeighbourhood Poverty 1981-2001." United Way of Greater Toronto.Retrieved February 24, 2010(http://www.unitedwaytoronto.com/who_we_help/pdfs/PovertybyPostalCodeFinal .pdf). Warburton, R.N. and P. Warburton. 2004. "Canada Needs BetterData for Evidence-Based Policy: Inconsistencies between Administrativeand Survey Data on Welfare Dependence and Education." CanadianPublic Policy 30:242-55. Webber, R. and T. Butler. 2007. "Classifying Pupils by WhereThey Live: How Well Does This Predict Variations in their GCSEResults?" Urban Studies 44(7): 1229-53. Worswick, C. 2001. School Performance of the Children of Immigrantsin Canada, 1994-98. Ottawa, ON: Statistics Canada. Yau, M. and J. O'Reilly. (2007). 2006 Student Census, Grades7-12: System Overview. Toronto, ON: Toronto District School Board. Zeman, K. 2007. "A First Look at Provincial Differences inEducation Pathways from High School to College and University."Education Matters. Ottawa: Statistics Canada, Catalogue Number81-004-XIE - June 2007, volume 4 number 2. Zhou, M. 1997. "Segmented Assimilation: Issues, Controversies,and Recent Research on the New Second Generation." InternationalMigration Review 31:975-1008. PAUL ANISEF York University ROBERT S. BROWN Toronto District School Board KELLI PHYTHIAN University of Western Ontario ROBERT SWEET Lakehead University DAVID WALTERS Guelph University (1.) Often, administrative data are poorly maintained, subject toadministrative changes, or ill suited to answer research questions.Moreover, access to administrative data can be difficult to obtain. (2.) The first generation are the foreign born, the secondgeneration includes those who were born in Canada to immigrant parents,and the third-plus generation consists of the offspring the Canadianborn. The third generation is often grouped with later generations(referred to as the third-plus generation) for theoretical and empiricalsimplification. (3.) This difference of one proficiency level can be consideredcomparatively large. See Bushnik et al. (2004). (4.) When the cohort study started in Fall 2000, the Ontariocurriculum provided ESL-ESD courses (English as a SecondLanguage/English as a Second Dialect). Since then, "ESD" hasbeen changed to "ELL," English Language Literacy. The vastmajority of courses were ESL, and we will refer to all ESL-ESD-ELLcourses as "ESL" in this paper. (5.) It is probable that many of these English-speaking,Canadian-born students were taking ESD or ELL courses. (6.) Approximately 10 percent of the cohort left the TDSB foranother school board. Since we were unable to track the educationrecords of these students after leaving the TDSB, we removed them fromthe analysis. (7.) The number of students in the full sample was 16,249.List-wise deletion was used for missing data and this resulted in afinal sample of 12,138. The majority of deleted observations are removedas a result of our selection process. (8.) Thus, for mixed models with a Bernoulli sampling distribution,the intraclass correlation is calculated as [rho] =[[tau].sub.00]/([[tau].sub.00] + [[pi].sup.2]/3), where [[tau].sup.00]is variance at level 2. (9.) If this estimate was not statistically significant we wouldhave proceeded to estimate a simple logistic regression model. (10.) The LICO variable is centered at its grand mean. (11.) When not otherwise stated, all of the interpretations forModel 3 are made controlling for the other predictors in the model, andholding constant the value of the random effect, [[mu].sub.0j]. (12.) Since LICO is reverse coded, the negative coefficientindicates that students residing in the lowest incidence neighborhoodsare least likely to drop out of high school. Authors are listed in alphabetical order and we acknowledge theassistance of Etta Baichman-Anisef in editing this paper. Paul Anisef, York University, 335 York Lanes, 4700 Keele Street,Toronto, ON, Canada M3J 1P3. E-mail: anisef@yorku.caTable 1Descriptive Results for Variables in the Study, Separated byRegion of Origin (n = 12,138) Mean/proportion Canada Canada (non- (English) English) Caribbean AfricaDropout Yes .19 .18 .40 .23 No .81 .82 .60 .77Sex Female .49 .47 .50 .53 Male .51 .53 .50 .47Age of entry On time .97 .98 .88 .89 One year late .03 .02 .12 .11Living situation Both parents .50 .72 .26 .42 Alternative .50 .28 .74 .58 structureStreaming level Academic .78 .83 .39 .59 Applied .21 .15 .53 .38 Essentials .02 .02 .08 .03At risk At risk .14 .13 .33 .22 Not at risk .86 .87 .67 .78Taken ESL courses NO .99 .99 .90 .85 Yes .01 .01 .10 .15LICO 6.41 5.48 4.17 3.44n 6,697 1,297 460 338 Mean/proportion Eastern South Western Europe Asia Asia Asia pDropout *** Yes .15 .10 .16 .22 No .85 .90 .84 .78Sex Female .49 .47 .50 .49 Male .51 .53 .50 .51Age of entry *** On time .95 .92 .94 .95 One year late .05 .08 .06 .05 ***Living situation Both parents .74 .59 .49 .66 Alternative .26 .41 .51 .34 structureStreaming level *** Academic .85 .90 .78 .75 Applied .14 .09 .20 .24 Essentials .01 .01 .02 .02At risk *** At risk .10 .07 .11 .16 Not at risk .90 .93 .89 .84Taken ESL courses NO .81 .72 .78 .74 *** Yes .19 .28 .22 .26LICO 4.95 5.37 3.88 4.13 ***n 800 1,027 1,094 425*** p <.001.LICO, low-income cutoff.Table 2Hierarchical Generalized Linear Model Predicting Dropout from theIndependent Variables (n = 12,128) Model l b SE(b) pFixed effects Constant -1.627Country of origin Caribbean Africa Europe Eastern Asia South Asia Western Asia Canada (non-English) Canada (English)Sex Male FemaleAge of entry One year late On timeLiving situation of student Alternative family structure Living with both parentsStreaming level Applied Essentials AcademicAt risk of not completing At risk Not at riskTaken ESL courses Yes NoLevel 2 LICORandom effects Variance of the random .498 intercept Intraclass .131 *** correlation ([rho]) Model 2 b SE(b) pFixed effects Constant -1.544Country of origin *** Caribbean 1.029 (.11) *** Africa 0.152 (.144) Europe -0.298 (.11) ** Eastern Asia -0.808 (.114) *** South Asia -0.305 (.096) ** Western Asia 0.148 (.13) Canada (non-English) -0.153 (.084) Canada (English) (Ref) --Sex Male FemaleAge of entry One year late On timeLiving situation of student Alternative family structure Living with both parentsStreaming level Applied Essentials AcademicAt risk of not completing At risk Not at riskTaken ESL courses Yes NoLevel 2 LICORandom effects Variance of the random .402 intercept Intraclass .109 *** correlation ([rho]) Model 3 b SE(b) pFixed effects Constant -2.168Country of origin *** Caribbean .144 (.129) Africa -.396 (.168) * Europe -.067 (.124) Eastern Asia -.577 (.131) *** South Asia -.295 (.111) ** Western Asia .079 (.148) Canada (non-English) .006 (.095) Canada (English) (Ref) --Sex *** Male .345 (.057) *** Female (Ref) --Age of entry *** One year late .590 (.113) *** On time (Ref) --Living situation of student *** Alternative family .294 (.058) *** structure Living with both parents (Ref) --Streaming level *** Applied 1.014 (.067) *** Essentials 1.336 (.167) *** Academic (Ref) --At risk of not completing *** At risk 2.169 (.076) *** Not at risk (Ref) --Taken ESL courses Yes -.119 (.11) No (Ref) --Level 2 LICO -.055 (.011) ***Random effects Variance of the random .168 intercept Intraclass .049 * correlation ([rho])* p value <.5.** p value <.01.*** p value <.001.Standard errors are in parentheses.LICO, low-income cutoff.Ref, reference category or group.