Obesity, defined by the WHO as having a body mass index (BMI) of 30 or higher (World Health Organization, 1995), is one of the five leading global risks for mortality (Murray et al., 2020). It is a major determinant of cardiovascular diseases, diabetes and certain types of cancer (Hruby et al., 2016), and it poses a significant international public health concern due to its substantial social, health and economic costs. There has been an unprecedented increase in overweight and obesity rates all over the globe, including in Australia (Australian Bureau of Statistics, 2018b; World Obesity Federation, 2023). For example, the prevalence of overweight and obesity among Australians aged 18 years and over increased from 56% in 1995 to 67% in 2017–18. This increase was mainly driven by the proportion of people in the obese category, which rose from 19% in 1995 to 28% in 2014 and 31% in 2018 (Australian Bureau of Statistics, 2018a, b).

In general, foreign-born (FB) individuals appear to have lower body mass index (BMI) and are less likely to be overweight or obese upon arrival in the host country than native-born (NB) people (for a review, see Goulão et al., 2015; Murphy et al., 2017; Oza-Frank and Cunningham, 2010). However, an increase in obesity levels with increased duration of residence (DoR) in the host country has been noted among various immigrant groups in the USA (e.g. Antecol and Bedard, 2006; Bates et al., 2008; Commodore-Mensah et al., 2016; Goel et al., 2004; Gordon-Larsen et al., 2003; Himmelgreen et al., 2004; Kaplan et al., 2004; Kaushal, 2009; Oza-Frank and Narayan, 2010; Sanchez-Vaznaugh et al., 2008), Canada (e.g. McDonald and Kennedy, 2005; Setia et al., 2009), Australia (e.g. Hauck et al., 2011; Menigoz et al., 2018; Menigoz et al., 2016), Germany (e.g. Sander, 2008) and Norway (e.g. Gele and Mbalilaki, 2013). Past research has also shown that immigrants who arrive at younger ages are at higher risk of being overweight or obese with increasing length of residence than immigrants who arrive at older ages (Kaushal, 2009; Oza-Frank and Narayan, 2010a; Roshania et al., 2008). However, the heterogeneity in weight and obesity rates among immigrant populations upon arrival and over time as a result of actual differences between groups, such as gender, age at the time of migration and duration of residency in the host country, have also been acknowledged and documented (e.g. Antecol and Bedard, 2006; Delavari et al., 2013b; Gordon-Larsen et al., 2003; Himmelgreen et al., 2004; Kaushal, 2009; Murphy et al., 2017; Sanchez-Vaznaugh et al., 2008).

Menigoz et al. (2016, 2018) explored the influence of origin (i.e. region of birth), duration of residence and age at arrival on BMI and obesity patterns, but did not account for variations in body weight due to differing durations of residence and ages at arrival among immigrants from the same region. Hauck et al. (2011) studied the influence of generational status on BMI, yet this approach confounded the effects of acculturation in the different environments (i.e. country of origin for the first generation and Australia for the second generation) experienced by first- and second-generation immigrants. Exposure to different environments will likely result in different habits and views associated with factors influencing body weight. Our paper advances this body of Australian work by examining the influence of duration of stay and age of arrival on obesity patterns among immigrants from different origins and comparing these patterns with those of non-immigrants. This fills a critical gap not addressed by previous studies.

A significant limitation in much of the previously published research concerning immigration and obesity is that, until recently, a substantial portion of both international and national research evidence relied on single or repeated cross-sectional datasets (e.g. Antecol and Bedard, 2006; Gele and Mbalilaki, 2013; Hauck et al., 2011; Menigoz et al., 2016; Sanchez-Vaznaugh et al., 2008). These datasets offer only snapshots in time, showcasing differences in health outcomes between immigrants and non-immigrants. However, processes such as immigration are dynamic, and several years may need to pass before their full impact on health can be observed. Consequently, assessing the effect of immigration on obesity becomes more challenging when relying solely on cross-sectional data. Additionally, estimating the effects of immigration introduces additional complexities because exposure to immigration is not a random occurrence. Immigrants and non-immigrants may differ significantly due to self-selection and other underlying processes. For example, immigration to a new country initiates a labour market adjustment process that non-immigrant workers do not undergo. This adjustment process can introduce confounding factors that bias the estimation of the exposure-outcome relationship (Rothman et al., 2008). Likewise, many factors influencing obesity, such as diet and exercise, can change over time for individuals, and particularly for immigrants as they assimilate to the host culture (Delavari et al., 2013b). Failing to account for these individual-level changes can result in biased estimates. In contrast, Setia et al. (2009) employed longitudinal data and mixed-effects models to compare 12-year changes in BMI among White and non-White immigrants with those among Canadian-born individuals. However, Setia et al. (2009) did not measure the changes in BMI over time for various immigrant groups relative to those for Canadian-born individuals. Without this crucial comparison, it becomes difficult to attribute the distinct trajectories in obesity among immigrants solely to their immigrant status.

Furthermore, it is important to note that many longitudinal studies have employed balanced panels, often overlooking potential biases stemming from panel attrition (Chiswick et al., 2008; De Maio and Kemp, 2010; Kim et al., 2013; Setia et al., 2011). Additionally, the pathways and mechanisms responsible for changes in obesity over time remain poorly understood, limiting the development of effective policies aimed at improving the health of all individuals, including immigrants.

Such methodological and theoretical considerations motivated this paper. Longitudinal data can be used to help identify differences in causal relationships between obesity and potential determinants for FB and NB people. Longitudinal data are preferred over cross-sectional data, even when the latter include information on age and time of arrival, due to several key advantages. First, observing the same individuals at multiple points in time allows for a more detailed investigation of causal relationships. It also accounts for individual differences, clarifying how unique traits influence outcomes, which is crucial for designing effective interventions. Furthermore, using longitudinal data reduces bias and confounding by controlling for these individual differences, isolating effects within participants. Finally, longitudinal data can be used to track changes in individuals’ obesity status over time, providing a dynamic view of how immigration affects health outcomes. This is crucial for understanding the temporal relationship between immigration and obesity, which cross-sectional data cannot capture.

The present study builds upon the methodological framework and dataset of Jatrana et al. (2018) to explore a distinct research question: namely that of the relationship between nativity, duration of residence and obesity. Our study advances the immigrant health literature by (i) using a nationally representative longitudinal dataset, the Household, Income and Labour Dynamics in Australia (HILDA), with 16 years of follow-up data to provide estimates of the nativity/DoR gap in obesity (i.e. nativity/DoR effects are measured relative to native Australians); (ii) providing a baseline for future fixed effects obesity analyses of HILDA data using the same method as the one used in this paper; and (iii) recognising that confounding is not an issue for nativity/DoR exposures, as no factor has a causal influence on an individual’s nativity status since that status was determined at birth. However, nativity status can causally influence a range of socio-economic and cultural factors (see below). Similarly, duration of residence is fixed in this study at the first wave, meaning that it cannot be causally influenced by other variables that occur at a later time. Furthermore, we (iv) reduce bias from loss to follow-up by using the unbalanced panel and (v) explore some possible mechanisms through which obesity changes over time post-immigration.

Using so-called hybrid regression models (explained in the next section) that separate within-person and between-person variations over time, we examine the associations of nativity (NB and immigrants from ES (English-speaking) and NES (non-English-speaking) countries of origin) and DoR with the prevalence of obesity, and whether the association between nativity/DoR and obesity prevalence is modified over time by age at arrival (AA) in Australia. We also examine the mediating roles of English language proficiency, socio-economic status (SES) and health behaviour factors in the association between nativity/DoR and obesity.

Examining differences in obesity levels between immigrants and Australian-born individuals, and how these differences evolve over time, is a crucial policy concern in Australia, which has one of the world’s highest proportions of immigrants in the population (McAuliffe and Triandafyllidou, 2021). Almost 30% of the total population in Australia were born overseas, and net overseas immigration is a significant driver of population growth (Australian Bureau of Statistics, 2023a, b). Additionally, almost half of Australians have at least one parent born overseas (Australian Bureau of Statistics, 2022). As the number of immigrants in Australia continues to rise, it has become increasingly important to comprehend the variations in health risk factors, such as BMI and obesity, between FB and NB individuals. Understanding how these factors change over time is essential, as it aids in the identification of vulnerable immigrant populations. Immigrant health has gained additional significance due to Australia’s evolving immigration program, which now focuses on meeting the labour market needs of the Australian economy (Parliament of Australia, 2021). It is worth noting that good health plays a pivotal role in fully realising the social and economic potential of immigrants. Moreover, under the Australian point-based immigration system, applicants earn points based on their age and English language proficiency (Mence et al., 2017). Consequently, it is important to gain a more comprehensive understanding of how AA and English language proficiency are connected with obesity prevalence, nativity and DoR.

The purpose of this study was to investigate differences in the levels of obesity among foreign-born people from English-speaking and non-English-speaking countries relative to those among native-born Australians, and how those differences changed with duration of residence and age at arrival. We were also interested in examining the mediating roles of English language proficiency, SES and health behaviour factors in the association between nativity, duration of residence and obesity. Unlike analyses that examined these research questions using cross-sectional data, we used 16 waves of longitudinal data to investigate the nature of the association between immigration and obesity in the Australian setting. Extending previous longitudinal work, including that of Setia et al. (Setia et al., 2009, 2011, 2012), we used hybrid regression models to help reduce estimation bias. With respect to our research questions, we found that: 1.

The odds of being obese were patterned by nativity: foreign-born people from both English-speaking and non-English-speaking countries had lower odds of being obese than those of native-born people.

In response to our first research question, our findings are consistent with prior research showing that foreign nativity is associated with lower BMI, overweight and obesity (Antecol and Bedard, 2006; Lauderdale and Rathouz, 2000; Popkin and Udry, 1998; Sanchez-Vaznaugh et al., 2008). By adopting a longitudinal lens, our study contributes new insights into this well-documented pattern, commonly referred to as the “healthy migrant effect”. This hypothesis states that migrants are positively selected on health, either through direct mechanisms such as medical screening, or indirectly via immigration policies that favour individuals with higher education, skills and financial resources, as well as self-selection processes (Akresh, 2007; Antecol and Bedard, 2006; Biddle et al., 2007; McDonald and Kennedy, 2004). Additionally, individuals with obesity have worse health conditions and a higher prevalence of pre-existing chronic conditions, which can be a barrier to the decision to immigrate, especially among the most vulnerable socio-economic groups. For a more detailed discussion of the underlying mechanisms contributing to the healthy immigrant effect, including selection processes and structural determinants, see Jatrana et al. (2018a).

Additionally, our second research question is answered affirmatively, particularly in the case of FB individuals from NES countries. Notably, the health advantage observed among NES immigrants diminished after two decades in Australia. A key explanation for this pattern is the process of acculturation, through which prolonged residence may lead to the adoption of health-damaging behaviours and the erosion of protective cultural practices (Antecol and Bedard, 2006; Delavari et al., 2013b; Wallace, 2022). Changes in diet and physical activity are often cited as primary mechanisms (Akresh, 2007; Delavari et al., 2013a) While our data did not include dietary information, we analysed physical activity and found no strong evidence that it mediated the association between DoR and obesity. However, prior work (Joshi et al., 2017) showed that FB individuals from NES countries have persistently lower odds of engaging in physical activity, even after 20 years of residence.

Although acculturation offers a useful framework, it has limitations. It often assumes that migrants come from healthier environments and that Western societies induce negative behavioural change. This perspective can overlook the global diffusion of unhealthy behaviours prior to migration. For instance, Martínez (2013) demonstrated that many Latino immigrants had already adopted Westernised diets before migrating to the U.S. Similarly, Australia’s infrastructure and environmental conditions – such as its lower pollution, accessible green spaces and recreational facilities – may promote healthier behaviours than some migrants’ countries of origin. In other cases, similarities between Australia and the country of origin may limit behaviour change post-migration.

It is possible that changes in the composition of immigrants might influence the obesity trend over time. To investigate this, we conducted additional sensitivity analysis (results not shown but available from the lead author upon request) to assess how the composition of immigrants evolved over time. The data did not reveal a significant change in immigrant composition. Given the lack of substantial changes in the makeup of immigrants from regions with notably different obesity rates and dietary habits – such as those from Asia, the Middle East or non-English-speaking countries compared to those from traditional source countries in Europe and North America – it appears unlikely that shifts in immigrant composition are affecting obesity trends.

Our findings regarding the impact of AA differ from those of prior studies (Kaushal, 2009; Oza-Frank and Narayan, 2010a; Roshania et al., 2008) that showed a potential interaction between DoR and the age at which individuals arrived. However, due to wide confidence intervals, our analysis may have lacked sufficient statistical power to detect significant differences across AA groups, and the results should be interpreted with caution pending further validation. As noted in prior research (Kinra, 2004), disentangling the effects of AA from DoR remains a methodological challenge especially in observational studies. In the field of migration studies, the interdependence of age, AA and DoR poses a challenge when attempting to disentangle the individual impacts of age at migration and the overall duration of exposure to a new environment (Kinra, 2004). Consequently, in an individual-level analysis, it becomes difficult to incorporate age, AA and DoR as continuous variables simultaneously. However, this challenge is not insurmountable when these variables are grouped together, even though correlations between the categories may complicate the detection of significant differences. Nonetheless, in our specific study, this issue did not appear to be a significant concern, except for the youngest and the most recently arrived immigrant groups.

Acknowledging the interdependence of age, period and cohort (APC) is essential, as their correlation can introduce multicollinearity problems in longitudinal analyses. However, omitting these factors risks misinterpreting effects, for instance mistaking age effects for period or cohort influences. In this study, we used age at arrival, which is an inherently fixed characteristic and thus enables a clear assessment of its impact on obesity. However, both age and duration of residence increase linearly with each HILDA wave. To address this, we fixed both DoR and age at wave 1 to prevent conflation of effects. Given the overlapping influences of age at arrival, temporal trends and year of arrival, further analyses (see Supplementary material Figures S.2 and S.3) are provided to illustrate respectively the distribution of duration of residence and age at arrival among foreign-born respondents at wave 6, and general trends in obesity by age, period (year), duration of residence and age at arrival. Additionally, Table S.6 provides regression model results for foreign-born people in Australia with obesity as the dependent variable and (i) CoB, DoR and their interaction and (ii) CoB, AA and their interaction as the main exposure variables with DoR and AA as continuous covariates.

We did not find any compelling evidence to suggest that health behaviours played a mediating role in the relationship between nativity, DoR and obesity. However, for immigrants from ES countries, English language proficiency and SES emerged as potential mediators in the relationship between DoR and obesity. After accounting for English language proficiency (Table 3, Model II) and SES factors (Table 3, Model III), previously non-significant findings for immigrants from ES countries relative to NB Australians became statistically significant. While marginal, this significance was evident for English language proficiency (DoR > 20 years) and SES (DoR 10–20 years). We also gained some insight into the role of English language proficiency in mediating the relationship between DoR and obesity when we did a sensitivity analysis by removing the so-called “Western countries” (e.g. France, Germany) from the NES category and adding them to the ES category, i.e. making this a “European” national grouping that ignored language differences. Specifically, the odds ratio estimates for those with a DoR of less than 20 years became statistically significant, while the estimate for those with a DoR exceeding 20 years lost its statistical significance (results not shown but available on request).

There are two plausible mechanisms through which English language proficiency might act as a mediator affecting weight gain among immigrants from ES countries. First, a stronger command of the English language might motivate immigrants to make better use of available information, such as that from mass media, and encourage participation in recreational physical activities (Caperchione et al., 2013). Second, because English proficiency and success in the labour market are closely linked, individuals with higher English proficiency are more likely to belong to higher income quartiles. This finding is consistent with the earlier research evidence suggesting that the ability to speak English influences the obesity levels of immigrants (Gee et al., 2008, 2010). However, the effect of language proficiency on health is complex and may operate through several mechanisms (Gee et al., 2010). The study by Gee et al. (2010) even challenges the presumption that use of the English language represents acculturation, and suggests new avenues of research focused on bilingualism (Schachter et al., 2012).

Our study makes a substantial contribution by addressing certain limitations of prior research by using longitudinal data and a combination of both conventional mixed and hybrid models. Theoretically, our results challenge the oversimplification embedded in traditional acculturation models. Future research should investigate how the immigrant health advantage erodes over time, potentially due to factors like acculturative stress and discrimination, which may influence diet and physical activity through their effects on mental health and identity (Berry et al., 1987; Rudmin, 2009). Evidence from studies on Latino and Iranian immigrants, as well as on Asian Americans, supports the link between stress, unhealthy coping behaviours and obesity (Agne et al., 2012; Delavari et al., 2013a; Gee et al., 2008). Further work should also explore pre- and post-migration dietary patterns and compare immigrants with non-migrant cohorts in their countries of origin (Gibson et al., 2011; McKenzie et al., 2007; Stillman et al., 2009), particularly given the rising obesity rates in developing countries.

From a policy perspective, our findings highlight the need for targeted obesity prevention strategies. Although FB individuals often arrive in better health than NB Australians, this advantage declines over time. As obesity levels among immigrants from NES countries converge with those of NB Australians, the burden of related conditions, such as diabetes, heart disease and cancer, may rise, placing further strain on healthcare systems (Bray, 2004). Tailored interventions are essential, particularly those designed to prevent or delay obesity onset in immigrant communities, recognising the unique challenges they face.

Several limitations in our study warrant attention. First, height and weight data were self-reported, which may introduce bias due to tendencies to underestimate weight and overestimate height (Gorber et al., 2007), though prior studies report strong concordance with measured values (Willett, 2012). We are not aware of evidence suggesting differential reporting by nativity, though response bias remains a possibility. To mitigate this, we controlled for variables commonly associated with reporting bias, including age, gender and education.

Second, our focus on time since immigration meant that we did not distinguish between immigrant generations. First-generation immigrants were categorised as FB, while second-generation migrants were included in the NB group, as they were born in Australia.

Third, analyses by age at arrival were limited to broad categories due to sample size constraints, which also prevented us from exploring the heterogeneity within ES and NES migrant groups. Future studies with larger samples should disaggregate data more finely by nativity, country of birth and age at arrival to uncover important subgroup differences.

Fourth, due to multicollinearity between duration of residence, year of arrival and year of survey, we were unable to separate DoR effects from period or cohort effects. As noted in Jatrana et al. (2018a), this is a common challenge in immigration research, compounded by age-related health changes and external contextual factors over time. Furthermore, the interdependence of age, AA and DoR limited our ability to disentangle their individual effects.

Fifth, while we examined the mediating roles of SES, English proficiency and health behaviours, we could not assess the role of dietary factors in obesity trends. Future research should explore the contribution of diet and nutrition in shaping the obesity trajectories among immigrants compared to those among non-immigrants.

Sixth, we have assumed that the direction of causal effect in Figure 1 goes from ELP to SES. It is plausible that the causal arrow might also go in the other direction at times, e.g. being employed in a predominantly English-speaking environment could improve an immigrant’s proficiency in the English language. This would be an example of reverse causation, an effect that is not amenable to detection with the approach used in this paper (Wooldridge, 2010), and would require more complex methods of causal analysis to identify (Pearl, 2009; Robins and Hernán, 2009). Of course, such methods have their own limitations, one being that rich datasets are required for estimation. Given the limitations of the HILDA dataset, there is still a place for the simpler analyses performed here, provided that the underlying assumptions are borne in mind.

Ours is one of the first studies to investigate the longitudinal association between obesity and nativity, duration of residence and age at arrival, and to do so from a causal perspective. As suggested by the limitations noted above, there is plenty of scope for confirming and strengthening our findings, for example by using more formal causal models, particularly when larger longitudinal studies become available. In addition, these findings are highly relevant for understanding the health trajectories of migrant populations and underscore the need to consider migration-related factors in public health interventions targeting obesity. Future research should explore the role of life course exposures, acculturation pathways and structural determinants in shaping obesity outcomes across different migrant subgroups.