Understanding access to higher education amongst humanitarian migrants: an analysis of Australian longitudinal survey data

Abstract

Humanitarian migrants are amongst the most marginalised population groups in countries within the Global North, including Australia. An important channel for these migrants to successfully settle into the host society and improve their socio-economic outcomes is participation in the local education system, particularly in higher-education options. However, we know surprisingly little about the socio-demographic factors that structure inequalities in humanitarian migrants’ access to (higher) education, with evidence from robust quantitative studies being particularly scarce. The present study fills this important gap in knowledge by analysing Australian longitudinal survey data (Building a New Life in Australia; n = 2109 migrants and 8668 person-year observations) by means of random-effect panel regression models. Key results indicated that higher English-language proficiency and pre-arrival education levels are core factors fostering greater engagement with the Australian higher-education system amongst humanitarian migrants. Humanitarian-migrant women in our sample exhibited a greater adjusted likelihood of being a student than humanitarian-migrant men. Altogether, our findings confirmed inequalities in accessing the Australian higher-education system amongst humanitarian migrants, and that policy attention is required to redress this situation. However, they also stress that a ‘one size fits all’ policy strategy may be neither sufficient nor appropriate to boost their education prospects.

Appendices

Appendix 1: Additional tables

Tables 3, 4, 5, 6, and 7

Table 3 Sample means and standard deviations for analytic variables

Table 4 Sample means for analytic variables using different subsamples

Table 5 Odds ratios from random-effect logistic regression models, alternative age ranges

Table 6 Odds ratios from a random-effect logistic regression model of studying towards a university degree, assigning the value zero to non-students

Table 7 Odds ratios from a random-effect logistic regression models with gender interactions

Appendix 2: Formal model specification

Formally, the random-effect logistic models that we fit take the following form:

$$log\left(\frac{pr({E}_{it}=1)}{1-pr({E}_{it}=1)}\right)=\alpha +{\varvec{\beta}}{{\varvec{S}}}_{it}+{u}_{i}$$

(1)

where subscripts i and t stand for individual and time period, respectively; E is a binary measure capturing being enrolled in an education or HE course in Australia measured at time t; S represents the set of socio-demographic variables described before and measured at time t; α is a model intercept; β is the vector of coefficients of interest to be estimated; and u is a person-specific random intercept (i.e., a random effect). While an array of longitudinal approaches are available to analyse panel data of the sort available in BNLA, this random-effect approach is fit-for-purpose to achieve our research aim of identifying the socio-demographic correlates of HE access amongst humanitarian migrants. These random-effect models are more efficient and less data-demanding than other potential approaches. For example, random-effect models using lagged explanatory variables would reduce the number of waves available for analysis, substantially diminishing the sample size. Similarly, within-group fixed-effect models would preclude consideration of time-constant predictors (e.g. pre-arrival education, gender, and country of origin), and would be highly inefficient in estimating the coefficients on rarely changing variables (e.g. English-language proficiency, marital status, or general health).