Abstract
Fiji, located in the Pacific Ocean, is classified as a Small Island Developing State and is characterised by a high rate of internal migration. This research explores the factors influencing mobility at the individual level in Fiji. Our study aims to uncover patterns of migrant selectivity and variations in migration tendencies among diverse groups within the Fijian population. The analysis considers variables such as age, gender, ethnicity and socio-economic indicators (including education and occupation). We analyse data from national censuses carried out between 1976 and 2007. The analysis reveals that individuals moving from remote or rural areas to urban centres tend to have higher socio-economic status. Youth aged 16–29 are the most mobile sub-group, and women tend to migrate more than men. This is likely explained by social norms around marriage and the greater educational and economic prospects available to women in cities. Additionally, migration patterns differ among ethnic groups: indigenous Fijians generally show a higher mobility rate than others. However, this pattern varies across time periods, and in the most recent data, Indo-Fijians from remote regions display the highest migration rates, which possibly reflect underlying ethnic divisions within the country.
Introduction
Globally, internal migration far surpasses international migration, with this trend being particularly pronounced in the Global South (UN DESA, 2018). In many cases, people move away from rural or remote areas to escape risks or pursue better opportunities. In the Pacific Island Countries (PICs), much attention is given to climate-related migration. However, beyond environmental challenges, individuals and families in small island states also face steep travel expenses between main and peripheral islands, poor infrastructure, limited public facilities and poverty (Cabezon et al., 2019). Consequently, many choose to leave the outer islands for larger, more developed islands where they can access better services, secure higher incomes or pursue education (Black et al., 2011).
To date, emigration from PICs has received the greatest scholarly focus, largely due to the significant role remittances play in these economies (contributing 30%–50% of national income across various island nations) (UN Population Fund Pacific Sub-regional Office, 2014).
However, mobility within and across the PICs is frequently the first phase preceding international migration and represents the most significant flow of people (Naidu & Vaike, 2016). An increasing number of studies has begun to explore if migration from remote outer islands and rural areas to coastal urban centres positively influences economic growth and structural transformation in these countries (Bertram & Watters, 1985; Naidu & Vaike, 2016; Sofer, 1993; Ward, 1961), and while anthropological studies have examined individual drivers of migration in PICs (Bryant-Tokalau, 2014; Mamak, 1974; Williksen-Bakker, 2001), there remains a gap in using demographic data to analyse this phenomenon. Research from other regions shows that internal migration is highly selective, with costs and risks often preventing the poorest and most vulnerable from migrating (Massey et al., 1993; Todaro, 1969). As extreme weather events become more frequent, internal migration in the Pacific is likely to increase, making it essential to expand our understanding of both those who migrate and those who remain behind.
Like many Small Island Developing States (SIDS), Fiji is sea-locked, remote and resource-limited while also being at the forefront of the struggle against climate change (Salem & Rosencranz, 2020). What sets Fiji apart, however, is its unusually high internal migration compared to other SIDS (Bell et al., 2015). Suva, the largest city in Fiji and the second largest in the Pacific, has seen substantial expansion in informal settlement. Additionally, Fiji’s ethnic divisions between iTaukei (indigenous Fijians) and Indo-Fijians have led to governmental instability, resulting in four coup d’états since the late 1980s and related streams of migration.
Against this backdrop, our study contributes to the growing body of research on internal migration within SIDS. We do this by analysing individual-level factors influencing internal migration patterns in Fiji. Specifically, our focus is on examining how various demographic and socio-economic characteristics are associated with migration behaviour among different segments of the Fijian population. These characteristics include gender, age, ethnicity and socio-economic indicators. For our analysis, we draw on four rounds of census data on internal migration in Fiji, collected between 1976 and 2007. 1
The census data contain information on the place of residence 5 years before the census and allow us to assign a migration status to each individual and compare with stayers (non-migrants).
The Fijian Context
Fiji is an archipelago of 330 islands spread in the heart of the Pacific Ocean, with a population of more than 900,000. Only 16% of Fiji’s landmass is suitable for agriculture and is found mainly along coastal plains, river deltas and valleys. Fishing and agriculture continue to be vital for the Fijian economy, yet their contribution to GDP and employment has consistently declined since the 1980s.
Although Fiji gained independence from the British Empire in 1970, colonial-era core–periphery structure persisted. As a result, geographic disparities in wealth and living conditions remain evident, with outer islands consistently serving as the main origin of internal migrants since the 1960s (Sofer, 1993; Ward, 1961). Even with substantial urban growth (55% of the population was urbanised in 2017; Fiji Bureau of Statistics, 2018), rural poverty has intensified, and poverty in urban areas tripled between 1977 and 2003 (Gounder, 2013).
Fiji is divided into 15 administrative provinces, as illustrated in Figure 1. The provinces of Ba, Naitasiri and Rewa have the highest population concentrations, collectively representing approximately 59% of the national population. Rewa includes the capital city, Suva, while Ba province contains the other major urban centres, Nadi and Lautoka. Suva alone has an estimated population of 77,366 (in 2023), making up about 9% of the country’s total. The remaining outer provinces, Bua, Macuata, Cakaudrove, Lau, Lomaiviti, Kadavu and Rotuma, together account for roughly 21% of Fiji’s population.
Access to the vast majority of land in Fiji (nearly 90%) is tied to clan affiliation through the mataqali system. For the iTaukei people, land is not only a means of subsistence but also holds deep cultural and spiritual value. In traditional Fijian villages, inheritance typically follows the male lineage, and women customarily relocate upon marriage. As urbanisation and inter-island mobility increase, urban areas have seen the development of distinct ethnic communities. In cities like Suva, for example, some groups continue to live under traditional vakavanua arrangements, maintaining strong connections with their home islands through practices such as remittances, ongoing communication and circular migration (Mohanty, 2006). Community events are often held to meet village needs and to help raise children with a sense of belonging to their village origins.
Most Indo-Fijians are descendants of indentured labourers brought by the British colonial authorities from South Asia between 1879 and 1916 to work in the sugar industry. They continue to constitute the core labour force within that sector. Since the early 2000s, however, a rising number of iTaukei landowners have chosen not to renew agricultural leases, leading many Indo-Fijian farmers to abandon farming altogether (Voigt-Graf, 2008). In contrast to the iTaukei, who live in tightly knit, kinship-based village communities, Indo-Fijians tend to settle in a more dispersed fashion, typically leasing land wherever it is available (Nakamura & Kanemasu, 2020). Their communities are less bound by traditional hierarchies or familial structures. Education has historically served as the primary means of achieving upward mobility for Indo-Fijians (Naidu & Vaike, 2016). Nevertheless, ongoing sociopolitical challenges in Fiji have increasingly marginalised them, making migration abroad an attractive alternative for many.
Previous Research
Research on internal migration in lower middle-income countries consistently highlights that mobility is strongly influenced by factors such as age, gender, education and ethnicity (Auwalin, 2020; Lall et al., 2006). Younger individuals are more likely to migrate, partly because they are often landless and unemployed in their home communities and also because they stand to gain more from relocating compared to older age groups. Earlier literature suggests that women are generally less mobile than men, with their migration largely driven by marriage. This trend is linked to cultural expectations that limit women’s participation in the labour market outside the household (White & Lindstrom, 2005). More recent research, however, indicates a shift in gender dynamics, with growing numbers of women engaging in migration (Camlin et al., 2014). Improved infrastructure, wider access to education and technology, and more employment opportunities in smaller towns and urban areas have all contributed to this shift, resulting in what many scholars describe as a feminisation of internal migration (Camlin et al., 2014; Islam, 2013). Although men still dominate temporary migration flows, women often follow once these movements become more established and structured (White & Lindstrom, 2005). Education is another major determinant, as migration tends to be selective based on educational attainment. Both highly educated and low-skilled individuals are drawn to cities by the prospect of employment (Bernard & Bell, 2018; Lall et al., 2006).
Compared to other regions, research on internal migration within the PICs remains limited. Among these nations, Fiji has received the most scholarly attention, serving as the primary focus of empirical studies on internal migration in the region. Prior to independence, Ward (1961) estimated that nearly one-third of the Fijian population resided outside their native villages. The migrants were predominantly men aged 15–29. The primary source of migration was the outer islands, where land was in short supply. 2 Once migrants settled in Suva, for instance, there were not many incentives to go back. Key factors influencing migration included age, job opportunities, access to education, relief from communal responsibilities and the availability of urban services.
Additional research has also identified variations in migration patterns based on ethnicity. Using more detailed census data, Nair (1980) added that the probability of migration had increased for both ethnic groups everywhere, with the provinces of Rewa (where Suva is located) and Ba (an area of sugar cane plantations and mines) being the largest recipients. While 94% of Indo-Fijian households consider themselves permanent migrants, only 60% of indigenous Fijians report the same. This difference is largely due to land ownership, as iTaukei Fijians have customary rights to land through their village chiefs, whereas Indo-Fijians do not hold such entitlements. Nayacakalou (1975) similarly highlights how traditional leadership and social structures among indigenous Fijians maintain strong ties to land and village identity, which could moderate migration decisions.
According to Bedford (1989), the unsuccessful 1985 land reform aimed at granting land access to Indo-Fijians contributed significantly to the large-scale emigration of skilled workers.
The two coup d’états in 1987 led to significant out-migration, with unofficial sources estimating that over 100,000 people left Fiji, almost 90% of whom were Indo-Fijians. Lal (2013) argued that these political upheavals disproportionately affected Indo-Fijian communities, as institutional discrimination and heightened insecurity became powerful drivers of both internal displacement and international migration. Mohanty (2006) estimates that following the 2000 coup, average annual migration nearly doubled, reaching approximately 6,000 individuals per year between 2000 and 2003. This surge resulted in the departure of over 3,800 professionals, technicians and skilled workers. In effect, more than half of the existing pool of medium- and high-skilled labour, especially teachers, left the country during this period (Voigt-Graf, 2003).
Internal Fijian migration has, on average, been male-dominated, but the gender gap between male and female migrants has narrowed over time as urban centres became places of opportunity for employment and education (Naidu & Vaike, 2016). Marriage traditions in Fiji also contribute to female migration. In customary practice, women move to their husband’s village after marriage. Since marriages between people from different regions or islands are fairly common, this often leads to women relocating (Nabobo-Baba, 2006). Ravuvu (1987, 1988) emphasises how iTaukei customs and kinship structures reinforce the relocation of women upon marriage, reflecting broader communal obligations.
Migrant women have emerged as significant economic contributors. Both skilled and unskilled women are often the family’s primary migrant, facilitating the migration of the entire household or, in other cases, leaving the family behind (Bastia & Piper, 2019). These migration patterns can be seen in studies of Fijian transnational migration, encompassing both skilled and unskilled occupations (Scott, 2003; Voigt-Graf, 2008). Rokoduru (2006) further documents intra-Pacific labour migration from Fiji, particularly among nurses and teachers, illustrating the growing role of women in regional mobility.
Theoretical Foundations
According to Todaro (1969), individuals base their migration decisions on anticipated economic gains. Assuming free choice and complete information, individuals typically migrate to areas offering the highest wages. These regions usually have infrastructure that lowers migration costs, which can lessen the motivation to migrate as wage differences between rural and urban areas narrow. In a similar vein, Sjaastad (1962) argues that individuals with higher levels of education are more inclined to migrate, as they expect greater benefits, provided there is sufficient demand for their skills at the destination. However, empirical evidence is mixed; both high-skilled and low-skilled individuals migrate for different reasons (Lall et al., 2006). Highly educated individuals with specific skills are rewarded with better jobs and higher earnings in urban areas, while low-skilled individuals migrate to find manual jobs and escape unemployment in rural areas (Keung Wong et al., 2007). Generally, younger individuals are more likely to migrate than older individuals, as a longer life expectancy correlates with expected positive migration returns. In traditional male-breadwinner societies, male labour migration prevails, with men providing for the family and women handling household tasks (Ullah, 2017). Migration decisions can also be influenced by unobservable personal characteristics (Constant & Massey, 2003; Heckman, 1979). Individual traits and capabilities, beyond measurable factors, may affect the probability of migrating, resulting in varied migration behaviour even among people with comparable age, gender and educational backgrounds (Nakosteen & Zimmer, 1980).
The neoclassical cost-benefit approach to migration has been criticised for its unrealistic assumptions of perfect information, rational decision-making and full market access (Massey et al., 1993). It also overlooks broader external factors such as conflict, the impacts of colonial history and global disparities (Baines, 1990). Economic and institutional factors significantly shape rural-to-urban migration, affecting both willingness and opportunity. Urban job availability encourages migration, while issues such as land ownership regulations can push rural populations to relocate (Lall et al., 2006).
Changes in population structure play a significant role in driving internal migration. During the demographic transition, mortality declines before fertility, particularly in urban areas. Higher fertility in rural regions creates a surplus of labour, which can fuel growth in urban sectors (Lewis, 1954). The urban sector, dependent on industry and services, typically has higher productivity than agriculture, the primary occupation in rural areas. This productivity gap incentivises rural-to-urban migration, which is vital for structural transformation and poverty reduction (Timmer, 1988).
Migration decisions in societies like the Fijian, which operate on a collective principle, extend beyond individual cost-benefit analyses. Instead, they are deeply intertwined with families and communities. Within these collective societies, individuals often make migration choices collectively, with the overarching goal of maximising income while minimising risks. The decision to send a family member in search of urban employment, for instance, is typically made within the family unit (Massey et al., 1993). In many cases, extended families choose younger members to migrate, with the expectation that they will send financial support back to their households (Azam & Gubert, 2006).
Existing social networks in destination areas help reduce both financial and social risks for incoming migrants (Skop et al., 2006). As information and experience build within these networks, the migration process becomes self-sustaining, increasing the flow of individuals along specific routes (Massey et al., 1999).
This collective perspective is known as the New Economics of Labour Migration (NELM), which focuses on how migration affects the communities migrants leave behind, particularly rural areas reliant on agriculture (Taylor, 1999). When migration is analysed through the lens of NELM, conventional push factors tend to have a stronger influence on young people and men.
Internal migration also impacts the well-being of those left behind. While there is a short-term loss of income when family members migrate, this is often offset by remittances. If labour is the primary motive, families tend to gain financially (de Brauw & Harigaya, 2007), but if education is the reason, it may incur costs. Migration costs vary by household, with wealthier families better able to afford migration or to stay in place during tough times. Wealthier individuals also have more resources to cover the often high short-term costs of migration.
Hypothesis
Our objective is to identify the defining traits of a typical internal migrant in Fiji. Drawing from theoretical frameworks and earlier research, we recognise the significance of individual characteristics and the selective nature of migration. Based on this, we hypothesise the following:
Individuals who expect greater benefits from migrating, such as those with higher educational attainment, are more likely to migrate; Migration from both remote and rural areas is predominantly undertaken by younger age groups, who generally gain more from education and face better employment opportunities; Migration displays a gendered pattern, with an increasing proportion of female participation over time; Indo-Fijians are more likely to migrate than iTaukei due to educational selectivity and limited land access; and Migrants typically have stronger employment prospects and higher skill levels compared to those who remain in their places of origin.
Data
Fijian Census Data
To evaluate the hypotheses outlined above, we utilise four rounds of census data from Fiji, collected in the years 1976, 1986, 1996 and 2007. The data were obtained through systematic sampling, selecting every 10th dwelling following a random starting point. A dwelling is defined as a standalone residential structure that is neither attached to another unit nor used for commercial or industrial purposes. A household, in turn, is defined as individuals who regularly share meals and contribute jointly to food expenses. The final working sample consists of approximately 50,000 individuals aged 16 and over, drawn from urban, rural and remote areas. Each respondent was asked where they had lived 5 years prior to the census. By comparing that response with their current province, we identify their migration status.
While the census data serve as a valuable resource for examining internal migration patterns in Fiji, it is important to note its limitations. Specifically, our data set lacks crucial information concerning migrants’ motivations, the role of family and community decisions, and comprehensive insights into their migratory behaviours.
Thus, certain factors highlighted in the theoretical framework, such as the influence of social networks and family decision-making processes, remain beyond the scope of this article’s investigation due to the constraints of the available data.
Variables
Most internal migrants in Fiji come from remote or rural areas and move to larger towns and cities. As a result, our analysis focuses specifically on these types of migrants. To investigate the characteristics of remote-to-urban and rural-to-urban migration, we conduct separate analyses for each group and treat migration status as a binary variable. Remote migrants are defined as individuals who originate from locations, mainly islands, that require expensive air travel or long boat journeys to access major urban centres. Rural migrants, on the other hand, come from provinces located closer to urban areas, where the distance and cost of migration are lower. For remote migrants, individuals who remain in remote areas are used as the reference category. Similarly, rural stayers serve as the comparison group for rural migrants. This distinction allows us to compare the two types of migrants, since the nature of remote migration involves greater distance and cost, while rural migration is considered shorter and less demanding. All individuals with migration status in our data live in one of the urban provinces (which includes the four largest cities in Fiji: Suva [Rewa], Lautoka and Nadi [Ba] and Labasa [Macuata]) at the time of the census and originate from the rural provinces (provinces without any more prominent cities located in the two main islands of Fiji: Viti Levu and Vanua Levu) or from the remaining provinces defined as remote. Table 1 shows the share of individuals in the data who live in each province at the time of the census.
Share (%) of Population in Fijian Provinces.
Period of Study
The four census rounds used in this study span a 40-year period marked by significant political and social changes in Fiji. During this time, the country moved from colonial rule under Britain to periods characterised by political instability and ethnic divisions.
The first census was conducted in 1976, 6 years after Fiji gained independence from Great Britain in 1970. Although the country has experienced considerable political instability throughout its post-colonial period, the political direction up until the late 1980s was generally oriented towards national unity and the development of a multicultural society (Kant, 2019).
In 1987, 1 year after the second census was conducted, national elections led to a coalition government formed by the Indo-Fijian-led National Federation Party and the Labour Party, which had strong backing from trade unions. However, this government was overthrown within weeks through a coup d’état. The primary aim of the coup was to strengthen protections for iTaukei interests, particularly in relation to land ownership. A second coup later that same year led to the introduction of a new constitution, which established political dominance for indigenous Fijians and reinforced legal safeguards for Fijian land rights (Kant, 2019).
In 1999, Mahendra Chaudhry became Fiji’s first Indo-Fijian prime minister, a development that triggered strong opposition from Fijian nationalists. This unrest culminated in a coup d’état in 2000, which involved widespread looting and the destruction of Indian-owned businesses in Suva, eventually leading to a military takeover. An interim administration was subsequently formed, which called for general elections held in 2001. These elections resulted in a win for the nationalist Fiji United Party. The final census used in this study was conducted in 2007, meaning that individuals identified as migrants in that data set would have moved during or shortly after the 2000 coup and the ensuing period of political instability.
Descriptive Patterns
Table 2 presents the group distribution of both migrants and non-migrants across the 4 census years, revealing a decreasing proportion of individuals remaining in remote and rural areas over time. Table 2 also includes urban dwellers, defined as individuals residing in urban areas who did not change their location within the previous 5 years. As expected, the proportion of this urban non-migrant group has grown steadily over the study period.
Group Distribution of Migrants and Non-migrants over Time.
Figure 2 illustrates a notable increase in educational attainment across the census periods. In 1976, fewer than 10% of individuals in any group had completed secondary education, with most groups falling below 5%. At that time, educational differences between the groups were minimal. By 2007, however, the proportion of individuals who had completed secondary education rose significantly among remote migrants (39%) and rural-to-urban migrants (41%), surpassing that of urban non-migrants (33%). Despite some progress, remote and rural stayers continued to have the lowest levels of secondary education compared to the other groups. This indicates that the educational gap between migrants and non-migrants remained over time, even as people moved from less developed areas to urban centres. Overall, those who stayed behind had fewer years of education than those who migrated, suggesting that education plays a central role in motivating internal migration.
Empirical Methods
Census data are cross-sectional, which introduces certain methodological constraints. One limitation is the inability to address endogeneity in migration decisions. Additionally, this type of data does not track individuals over time, making it impossible to capture repeat migration or to distinguish migrants from earlier waves. As a result, some individuals classified as stayers may, in fact, be past migrants. However, given that the dominant migration flows in Fiji move from rural or remote areas towards urban centres, and not the reverse, this issue is likely to be limited in scope. Even so, comparisons between migrant and non-migrant groups should be interpreted with caution. Furthermore, we are unable to account for environmentally driven migration, as such data are currently unavailable.
This study investigates the key individual-level factors influencing remote and rural-to-urban migration. To do so, we model migration status as the dependent variable in a logistic regression, where the outcome is coded as 1 for remote and rural–urban migrants and 0 for stayers. P is defined as the likelihood that Y = 1. Xi is a vector of individual-level variables, Hi captures household-level variables and Ti represents the census year.
We compare remote and rural migrants to urban migrants as there may be differences in migrant selectivity, since remote migration is long-distance while rural migration is defined as short-distance. We include categorical control variables for age (youth: 16–24, adults: 25–65, seniors: 50–64 and retired: 65+) and ethnicity (iTaukei, Indo-Fijian, others). We apply a dummy variable for gender. Given the small number of individuals with university-level education, we also use a simplified education variable, coded as 0 for less than secondary education and 1 for secondary education or higher. Employment consists of two groups: (a) employed and (b) unemployed or inactive as well as skill level, denoted by the values 0 for blue-collar job and 1 for white-collar job. 3
In all models, we control for civil status and the number of children in the household. Table 3 shows the variable means.
Variable Means.
Results
Patterns of Migration
Table 4 presents the main findings from the analysis, based on six separate models. In Models 1, 3 and 5, the dependent variable is coded as 1 for remote migrants and 0 for individuals who remained in remote areas. 4
Migration Status in Fiji, Pooled Census Data (1976, 1986, 1996, 2007) and Logistic Regression Model.
Models control for civil status, number of children in family and year.
In contrast, Models 2, 4 and 6 use a dependent variable coded as 1 for rural migrants and 0 for rural stayers. 5
The empirical analysis was designed to test four key hypotheses. One of our main expectations was that migration in Fiji would be selective based on education. This is supported by the results in Table 4, which show that individuals with at least secondary education are more likely to migrate. The odds ratio for remote migrants with secondary education or higher is 1.178, while for rural migrants it is slightly higher at 1.222. We also find evidence of age-based selectivity, using the 16–29 age group as the reference category. Adults, older individuals and retirees generally exhibit a lower likelihood of migrating. Notably, the pattern of age selectivity differs between remote and rural migration. The effect of age appears more pronounced for rural-to-urban migrants, suggesting that shorter-distance moves are more strongly influenced by age.
We expected a gendered pattern of migration, and the results confirm that women in Fiji are more likely to migrate than men, particularly in the case of long-distance migration, as shown in Model 1 (remote migrants). This suggests that remote migration may be less influenced by typical urban pull factors and more closely linked to traditional marriage practices, where women relocate to their husband’s village, supported by strong social ties within the mataqali. In contrast, rural migrants typically face lower migration costs due to their closer proximity to urban centres. As a result, gender differences in migration are less pronounced among rural migrants.
When comparing ethnic groups, the results indicate that iTaukei individuals are more likely to engage in remote migration than Indo-Fijians, as reflected by an odds ratio of 0.682 for the latter. For rural migration, however, there is no statistically significant difference between the two groups. This finding runs counter to our initial expectation that Indo-Fijians would be more inclined to migrate due to their higher educational attainment, limited access to land and exposure to ethnic tensions. One possible explanation lies in the strong mataqali-based social networks within the iTaukei community. These close-knit tribal ties may facilitate chain migration from remote regions to urban centres, as existing networks support and connect migrants across locations. In contrast, Indo-Fijians tend to live in communities with less-rigid social structures, meaning that tribal affiliation is unlikely to serve as a motivating factor for migration in the same way.
In Models 3 and 4 of Table 4, we add a variable representing employment status, and the results clearly show that remote and rural migrants are less likely to be employed than stayers. One explanation for this could be that migrants lack established networks at their destination, which may delay their entry into employment. While this may partly explain the pattern, a closer look at the Fijian context and the data suggest an alternative interpretation. Table A1 presents the types of occupations for both migrants and stayers and shows that migrants are less likely to work in agriculture. In Fiji, agricultural employment often consists of small-scale fishing or farming on family plots for subsistence purposes. Much of the agricultural work recorded in the census falls into this category. Since access to land is a key factor for participating in agriculture, and because 89.75% of land in Fiji is held under iTaukei customary ownership (as outlined in the section ‘The Fijian Context’), migrants have limited access to land. This lack of land access likely contributes to the lower levels of agricultural employment observed among migrants when compared to stayers.
In this part of the analysis, we focus exclusively on individuals who are employed and compare their types of occupations, as shown in Models 5 and 6. The results indicate that migrants have higher odds of being employed in white-collar occupations compared to stayers. This finding aligns with earlier evidence of educational selectivity among migrants. Additionally, remote migrants are more likely to hold white-collar positions than rural migrants, suggesting a stronger association between long-distance migration and higher-skilled employment.
Table 5 presents the interaction effects of key variables to further explore the patterns identified earlier. In Models 1 and 2, we examine interactions between gender and ethnicity. As shown previously in Table 4, women are more likely than men to migrate both to urban areas and from remote locations, and iTaukei individuals are the most likely group to undertake remote migration. The interaction results reveal that, in comparison to iTaukei men, iTaukei women have higher odds of being remote migrants. In contrast, both Indo-Fijian men and women are less likely to migrate from remote areas. This finding once again points to the influence of traditional Fijian marriage customs, which often lead to relocation for iTaukei women. For rural migration, we observe that both iTaukei and Indo-Fijian women are more likely to migrate than iTaukei men, which may reflect the higher returns to education for women in the labour market in urban settings compared to in rural settings.
Migration Status in Fiji, Interaction Effects, Pooled Census Data (1976, 1986, 1996, 2007) and Logistic Regression Model.
Models control for civil status, number of children in family and year.
To explore ethnic patterns in more detail, we include interaction terms between education and ethnicity in Models 3 and 4. The reference group in these models is iTaukei individuals with less than primary education. The findings suggest that education plays a role in driving rural migration for both iTaukei and Indo-Fijians. However, the effect appears stronger among Indo-Fijians, as seen in both remote and rural migration contexts. For iTaukei individuals, educational selectivity is evident only in rural migration. These results imply that remote migration among the iTaukei may be influenced by additional factors beyond education, while for Indo-Fijians, education is an important driver of both remote and rural migration.
Time Patterns
The data used in this study span a period marked by significant political instability and ethnic tension in Fiji. The two coups in 1987 and the government takeover in 2000 contributed to increased mobility among Indo-Fijians. This time frame also includes phases of educational expansion and growing urbanisation. To examine how the effects of key variables may have shifted over time, we break down the multivariate results by census year, allowing us to observe potential changes across different historical and social contexts.
Figure 3 presents differences across age categories for remote migrants, corresponding to those shown in Table 3. The data reveal a consistent pattern of youth being more likely to migrate (Figure 3a), especially in the more recent census years. Gender differences, shown in Figure 3b, remain stable across all four census waves, with women in remote areas consistently more likely to migrate than men.
We observe a noteworthy trend among individuals with secondary education (Figure 3d). In the earliest census, those with secondary education were equally likely as those with only primary education to migrate from remote areas. However, in the later census years, individuals with secondary education show a higher likelihood of migration. This suggests that education has become an increasingly important factor in migration decisions over time, a shift that may be linked to the significant expansion of educational opportunities in Fiji, particularly between the 1976 and 1986 census periods.
In terms of ethnicity, Figure 3c highlights substantial changes over time. Indo-Fijians, who were the least likely ethnic group to migrate from remote areas in 1976, became the most likely by 2007. While this shift may be influenced by the political and ethnic tensions that have characterised post-colonial Fiji, the data alone do not allow us to draw firm conclusions. Another possible explanation is that internal migration serves as a stepping stone towards international migration. Supporting this interpretation, Mohanty (2006) reports that approximately 89% of emigrants between 1987 and 2004 were Indo-Fijians. A similar, though less pronounced, increase in Indo-Fijian migration over time is also visible in Figure 4, which focuses on rural migrants.
Migration Status in Fiji, Remote Migrants Using Census Data (1976, 1986, 1996, 2007), Logistic Regression Model by Age Groups, Gender, Ethnicity and Education.
A closer examination of the remaining factors for rural migrants in Figure 4 reveals patterns similar to those observed for remote migrants. Educational attainment becomes a more influential factor in later census years, indicating its growing role in rural-to-urban migration. The data also show that women play a leading role in rural migration, with this trend appearing even more pronounced in the more recent censuses. Additionally, strong age-based selectivity is evident and remains consistent across all time periods.
Migration Status in Fiji, Rural Migrants, Using Census Data (1976, 1986, 1996, 2007), Logistic Regression Model by Age Groups, Gender, Ethnicity and Education.
Conclusion
Since gaining independence from colonial rule, Fiji has experienced high levels of internal population movement. This study adds to our understanding of internal migration in Fiji by analysing migrant selectivity and variations in migration likelihood across different population subgroups.
Based on theory and previous research, we expected internal migrants in Fiji to be younger and more educated than non-migrants, as these groups tend to gain the most from migration. The findings support this expectation, revealing clear selectivity based on both age and education for remote and rural migrants. Importantly, the role of education as a driving factor becomes more pronounced over time, particularly between the first and second census periods.
While earlier research has often shown migration to be male-dominated, more recent trends point to a feminisation of internal migration. In the Fijian context, our findings show that women are more likely to migrate, particularly in cases of remote migration. This pattern can be linked to Fiji’s cultural context, including traditional marriage practices (Nabobo-Baba, 2006), as well as the growing educational and economic opportunities available to women in urban areas, which serve as strong incentives for migration.
Our analysis also reveals distinct ethnic differences in migration patterns. Although we initially expected Indo-Fijians to have a higher likelihood of migrating due to limited access to land and experiences of ethnic tension, the results show that indigenous Fijians are more likely to migrate, particularly in the case of remote migration. This outcome appears to be influenced by strong social networks maintained through mataqali ties among the iTaukei. These tribal connections often link individuals from remote islands to those already settled in urban areas, with remittances serving as one form of continued interaction. Remote migration in this context tends to follow established pathways, where new migrants follow in the footsteps of earlier movers (Massey et al., 1999). In contrast, Indo-Fijians generally lack village-based affiliation, and their inter-island social connections may be weaker, reducing the influence of such networks on their migration decisions.
However, this pattern has shifted over time. In the most recent census, Indo-Fijians from remote areas exhibit the highest likelihood of migration. This change is likely connected to the ethnic tensions that followed the 2000 government takeover. While a significant number of Indo-Fijians emigrated internationally in response to political instability, many also moved within Fiji. Our findings suggest that the two major ethnic groups in Fiji follow distinct migration pathways influenced by different factors. For Indo-Fijians, migration is largely driven by access to education and responses to political unrest. In contrast, iTaukei remote migration appears to be shaped by traditional village structures and social networks maintained through the mataqali system. Meanwhile, rural iTaukei migration tends to respond more to conventional urban pull factors, such as employment and services.
Our study provides strong evidence that internal migrants, particularly those from remote islands, are generally better off than non-migrant groups. These individuals tend to have higher levels of educational attainment and are more frequently employed in white-collar occupations. These findings are consistent with theoretical expectations, which propose that individuals with higher education and specialised skills are more likely to benefit from urban migration through improved employment and income opportunities. Additionally, migration theory suggests that individuals may self-select based on unobservable characteristics (Constant & Massey, 2003; Heckman, 1979). Traits such as innate abilities or specific talents can increase both the likelihood of migrating and the chances of achieving higher educational and occupational outcomes (Nakosteen & Zimmer, 1980).
The chiefly hierarchical structure of Fijian society creates unequal opportunities, benefitting certain individuals and groups while restricting access for others. Closely linked to this is land ownership, which is determined by ethnicity and tribal affiliation. In addition, the financial costs associated with migration play an important role. Households with greater economic resources are more likely to afford the expenses involved in relocating, which can increase their chances of migrating.
The positive selection of migrants across several dimensions suggests underlying inequalities of opportunity in Fiji. Younger, more educated and economically better-off individuals are more likely to migrate, often leaving behind more vulnerable populations. This finding has important implications for discussions on climate-related migration, which is projected to rise due to sea-level rise and extreme weather events. Existing disparities in who is able to migrate may further increase the vulnerability of communities most exposed to climate change. Those who have the skills and resources to contribute to local resilience and recovery efforts may instead relocate to urban areas in search of better opportunities.
Footnotes
Data Availability
The data used for this study are publicly available via IPUMS International.
Declaration of Conflicting Interests
The authors declared no potential conflicts of interest with respect to the research, authorship and/or publication of this article.
Ethical Statement
The use of census data for research must prioritise the privacy and confidentiality of respondents. Researchers should obtain informed consent and be mindful of the sociocultural contexts influencing the data. Data were collected by the Fijian Bureau of Statistics, and the method of data collection was not developed by the authors of this study.
Funding
The authors disclosed receipt of the following financial support for the research, authorship and/or publication of this article: this study was conducted within the project Sustainable Development of Small Island States funded by the Swedish Research Council (Project Number VR 2019-04117).
Appendix
Construction of Employment-level Variables.
| ISCO | Outcome Variable |
| Legislators, senior officials and managers | 1 = White collar |
| Professionals | 1 = White collar |
| Technicians and associate professionals | 1 = White collar |
| Clerks | 1 = White collar |
| Service workers and shop and market sale | 1 = White collar |
| Skilled agricultural and fishery worker | 0 = Blue collar |
| Crafts and related trade workers | 0 = Blue collar |
| Plant and machine operators and assemblers | 0 = Blue collar |
| Elementary occupations | 0 = Blue collar |
| Armed forces | 0 = Blue collar |
| Other occupations, unspecified | 0 = Blue collar |