3 Data source and methodology

View the most recent version.

Data for this study were drawn from the Longitudinal Survey of Immigrants to Canada (LSIC). The target population of the survey, which was conducted jointly by Statistics Canada and Citizenship and Immigration Canada (CIC), includes all immigrants who (1) arrived in Canada from October 1, 2000 to September 30, 2001; (2) were aged 15 or older at the time of landing; and, (3) landed from abroad and applied through a Canadian mission abroad.

The sampling frame for LSIC was an administrative database of all landed immigrants to Canada maintained by CIC. The LSIC sample was created using a two-stage stratified sampling method. The first stage involved the selection of immigrating units (IUs) using a probability proportional- to-size method, and the second stage involved the random selection of one member within each IU. Only the selected member was followed throughout the survey. Individuals aged 15 or older in each IU were eligible to be selected as the respondent.

Three LSIC questionnaires were fielded during the course of the survey. Approximately 12,000 immigrants were interviewed from April 2001 to March 2002, about six months after their arrival in Canada; approximately 9,300 of the same immigrants were re-located and interviewed in 2003, about two years after their arrival; and, about 7,700 of the same immigrants were re-located and interviewed a third time, about four years after their arrival. The approximately 7,700 LSIC respondents who were re-located over the three waves are nationally representative of approximately 157,600 new immigrants, of whom 104,400 were economic immigrants, 42,600 were family-class immigrants and 9,700 were refugees.⁴

During the second and third LSIC interviews, respondents were asked

"Since your last interview, have you sent money outside Canada to relatives or friends?"

Respondents who said 'yes' were subsequently asked

"How much money have you sent outside Canada to relatives or friends?"

During the first LSIC interview—six months after arrival—respondents were only asked if, since arriving in Canada, they had sent money outside Canada to relatives or friends, they were not asked about the amount sent.

In the research literature, remittances are predominantly discussed in terms of the money sent to family members rather than to friends abroad. Consequently, we have limited our analysis to those LSIC respondents who have family members living outside Canada. This results in the exclusion of 256 of the 7,716 respondents, representing approximately 4,900 of the 157,600 immigrants in the landing cohort. Their exclusion has virtually no effect on model estimates. In addition, respondents who remitted less than $100 or more than $25,000 were excluded in order to reduce the effects of outliers on model estimates. This caused the exclusion of another 26 cases, resulting in a final sample of 7,434 respondents.

Because remittance questions were asked at the individual, rather than at the family, level (that is, "… have you sent money …?"), LSIC may yield underestimates of remittance activities of immigrant families. This is because it is possible that while a LSIC respondent did not send money abroad, someone else in their family did. This may be most likely when the respondent is not a primary income earner in the family—for example, teens living with their parents or seniors living with their adult children. A variable identifying 'principal applicants' and 'spouses and dependents' in the migrating unit had been initially included in our analysis, but it was dropped because it did not yield significant results. Instead, a variable identifying whether or not the respondent is the 'person most knowledgeable' about the family's financial situation is included. About three quarters of our respondents (73.5%) identify themselves as the person most knowledgeable.

⁴ Individuals who applied and landed from within Canada were excluded from the survey since they may have been in the country for a considerable length of time before 'landing' and may therefore demonstrate different integration characteristics than those of immigrants arriving more recently. Refugees claiming asylum from within Canada were also excluded from the survey. Interviews were conducted in one of 15 languages covering approximately 93% of the new immigrant population in Canada and were conducted face-to-face, or by telephone when a face-to-face interview was not possible. For more information on the Longitudinal Survey of Immigrants to Canada see the microdata user guide available at http://www.statcan.ca/english/sdds/document/4422_D1_T1_V3_E.pdf

Three other points regarding our sample are warranted. First, the question of whether remittance activities diminish over time since arrival in the host country can only be addressed within the four-year window of the survey. Data sources are not available to determine the extent to which immigrants in Canada continue to remit 5, 10 or 15 years after arrival. Second, temporary residents in Canada are not included in LSIC. The objective of the survey is to track the experiences of permanent residents over their first four years here and short-term, temporary residents fall outside the scope of LSIC. Therefore, we are unable to compare the remittance behaviours of temporary and permanent residents. In 2005, there were approximately 160,000 foreign nationals residing in Canada on temporary work permits (CIC 2007). Finally, it is important to note that the characteristics and experiences of the LSIC immigrant cohort may or may not be the same as those of cohorts that arrived in Canada at earlier or later points in time. For example, the downturn of the high-technology sector may have had particular consequences for immigrants arriving from 2000 to 2001, including their financial capacity to remit. While it would be informative to estimate inter-cohort trends or differences in remittance activities, our data do not allow us to do so.

While recognizing these limitations, LSIC does offer other advantages. It includes a large, representative sample of new immigrants from many source countries, thus making it possible to draw cross-national comparisons of remittance behaviours using consistent definitions and methodologies. This provides a breadth of perspective not available from studies focused on immigrants from one or two countries of origin. Furthermore, the longitudinal design of LSIC allows us to build in a time-lag between our independent and dependent variables, thereby minimizing problems of endogeneity. More specifically, two observations are available for each respondent: remittances made (or not made) two years and four years after arrival. For the observation two years after arrival, independent variables are measured using the six-month data file, and for the observation four years after arrival, independent variables are measured using the two-year file. In short, the independent variables are measured at the start of the reference period and thus precede the decision to remit.⁵

3.1 Independent variables

Independent variables within each of the five conceptual areas outlined above are included in the model.⁶ Financial capacity is measured in terms of family income, employment status of the respondent and the value of savings abroad. A flag identifying whether or not the respondent is the person most knowledgeable about the family's income is also included, as noted above.⁷

Human capital is measured by the level of education upon landing in Canada. Self-assessed language ability in English or French was included in earlier versions of the analysis, but it was subsequently excluded because it did not yield significant results. A place of residence variable identifying immigrants residing in Montréal, Toronto, Vancouver, Calgary/Edmonton and other areas is included.

Detailed information on the location of all the members of each immigrant's family is not available on LSIC. The vast majority of LSIC respondents (98%) say that they have family members living abroad, but the relationships with these members are not specified. Two family- related variables are included in our models: the number of children residing in the respondent's household and whether or not the respondent is sponsoring or intending to sponsor (i) a spouse or one or more children or (ii) parents or grandparents to come to Canada.⁸

In terms of migration characteristics, the category of admission to Canada is included in our models. Individuals are admitted to Canada as permanent residents through three main admission categories: economic immigrants, family-class immigrants and refugees. Principal applicants in the economic category are selected for their skills and ability to contribute to Canada's economy, and they are assessed on the basis of educational attainment, language abilities and other factors. Because the earnings gains of economic immigrants far outpace those of family-class immigrants and refugees in the years after arrival (Chui and Tran 2003), they may be more likely to remit and to remit larger amounts, given their greater financial capacity to do so.⁹

In terms of organizational membership, LSIC respondents were asked if they had become a member or taken part in the activities of various groups and organizations since their last interview. Two dummy variables are included, identifying respondents who had or had not participated in or become a member of a (i) church or other religious group or (ii) an ethnic or immigrant association.¹⁰

The region of origin is included in our analysis in two ways. First, regression models based on a pooled sample of all respondents are presented, with nine world regions identified by a series of dummy variables; second, separate regression models are run on respondents from each of these nine regions. Finally, gross domestic product per capita in the respondent's country of origin is included.

Our multivariate analysis includes a logistic regression on the likelihood of remitting and an ordinary least squares regression on the natural logarithm of the amount remitted. Coefficients from the logistic regressions have been converted into predicted probabilities for ease of interpretation.¹¹ Coefficients from the natural logarithm of the amount remitted approximate percentage differences and are discussed in these terms for ease of presentation. All models are calculated using bootstrap weights to correct variance estimates for survey design: a technique called design-based variance estimation.¹²

⁵ One exception is income, which is estimated at the mid-point of each of the reference periods, using a linear interpolation.

⁶ Our initial multivariate models included more than 20 independent variables, many of which did not yield significant statistical results. They were subsequently excluded from our analysis for the sake of efficiency and parsimony. The variables which were excluded are documented in the footnotes to this section for readers who may be interested.

⁷ In earlier versions of the analysis, housing expenditures as a percent of family income was also included. A significant negative correlation was found between such expenditures and both the likelihood of remitting and the amount remitted. However, the housing choices made by immigrants may be influenced by their decision to remit or by their expectation of remitting in the future. Hence, housing expenditures may be a consequence of remittance decisions rather than a determinant of these decisions. Because of this potential endogeneity, the housing expenditure variable was excluded from our final models.

⁸ Respondents sponsoring or intending to sponsor a spouse or child, as well as a parent or grandparent, were coded in the 'sponsoring spouse/child' category; there were fewer than 50 respondents in this situation. Marital status was included in earlier versions of the analysis, but it did not yield significant results and was subsequently excluded.

⁹ In earlier versions of the multivariate models, a variable was included identifying the respondent's single, most important reason for coming to Canada: specifically, economic reasons, family-related reasons, reasons of war or political freedoms, and other reasons. These variables did not yield significant results and were excluded.

¹⁰ A third variable identifying participation in any other type of group or organization was also initially included, but did not yield significant results.

¹¹ Predicted probabilities for each independent variable were estimated by setting the other independent variables to their mean values.

¹² Some researchers have used the Heckman selection model (1979) to take into account the possibility that the sample of immigrants who remit may be a selective sample of those who could have remitted (Funkhouser 1995. Brown and Piorine 2005). Several Heckman models were run using different specifications to address this issue but evidence of selectivity was not found. Our results are consistent with several studies that also report that selection effects are modest or not statistically significant (Menjivar et al. 1998, Funkhouser 1995).