A Life Course Perspective on the Wartime Migrations of Northern Vietnamese War Survivors

Abstract

Research addressing conflict and migration has made great strides in explaining the relationship between violence and migration. However, it commonly lacks individual-level data on exposure to war. We use survey data from the 2018 Vietnam Health and Aging Study to examine the associations between war-related violence exposure during the American War and the wartime migrations of northern Vietnamese war survivors. Using multilevel mixed-effects count models, we investigate three groups of factors influencing migration—war-related events, economic circumstances, and demographic and life course factors—to explore the relationship between war exposure and migration, inclusive of deployments, economic moves, and displacements. Our findings indicate that the effects of war exposure, socioeconomic status, and demographic characteristics diverge for different types of migration. These findings, framed within the life course and historical context, suggest the need to thoughtfully delineate both war exposures and traditional causes of migration to understand the diverse types of mobility occurring during periods of armed conflict.

Introduction

Unrelenting armed conflict and organised violence plague many regions of the globe, frequently triggering waves of international and internal migration (Bohra-Mishra & Massey, 2011; Davenport et al., 2003; Williams et al., 2012). Conflict-induced displacement damages the livelihoods, security, and health of the displaced, generating ‘disproportionately high levels of malnourishment, disease affliction, physical stunting,’ poor mental health, political disenfranchisement, and lack of access to social resources (Bhattarai-Ghimire & Upreti, 2008, p. 118; Haque et al., 2020). Displacement can also tax the support systems of receiving communities, increase informal economies, and increase social and economic inequality (Ruiz & Vargas-Silva, 2013; Verme & Schuettler, 2021).

Though scholars of displacement have predominantly focused on international migrants displaced by conflict, the “forced” migration¹ conversation has shifted to address internal migration and alternate causes (Betts, 2010; Castles, 2003). Documented drivers of internal migration range from national economic development and accompanying urbanisation (Charles-Edwards et al., 2019; Massey et al., 1993; Zelinsky, 1971) to policies redistributing populations (Charles-Edwards et al., 2019; Dang, 1999; Weisner, 1988) to natural disasters, (Betts, 2010; Castles, 2003; Haque et al., 2020) and finally conflict-induced displacement (Bohra-Mishra & Massey, 2011; Castles, 2003; Davenport et al., 2003; Melander & Öberg, 2007). In Asia, demographers document these same drivers but note that the dominant drivers vary across regions within Asia (Charles-Edwards et al., 2019).

When considering war and migration, little research considers the extent to which traditional forms of migration (e.g., for work, education, and marriage) persist during wartime and whether the drivers change in the context of war (Bohra-Mishra & Massey, 2011; Williams, 2015). Those studies that address traditional migration and the root causes find that these continue to operate during war in some geographic contexts (Ibid.). With its history of successive wars and widespread displacements, the Vietnamese case presents a unique opportunity to examine the tangled relationships between war and the myriad forms of migration.

Prior studies of war-related migration and Vietnam primarily focused on those who left Vietnam following the American War (e.g., Catanzaro & Moser 1982; Gold 1992). Further, while the migration literature is replete with studies examining the drivers of international migration, the immediate consequences of migration, and post-displacement integration processes (see, for example, Bloemraad et al., 2008; Bustamante et al., 2017; Castles et al., 2002; Gold, 1992; Ruiz & Vargas-Silva, 2013; Verme & Schuettler, Kirsten, 2021), studies of forced displacement are more limited (Abbasi-Shavazi & Kraly, 2018; Hugo et al., 2018) as are studies of internal migration in Asia (Abbasi-Shavazi & Kraly, 2018; Charles-Edwards et al., 2019). Moreover, few studies comprehensively delineate both individual- and community-level predictors of migration during wartime (Adhikari, 2013; Bohra-Mishra & Massey, 2011). This study takes a step toward filling that gap by investigating wartime migrations of older adults presently residing in four northern Vietnamese districts that experienced varying levels of bombing and violence during the Vietnam War. Though residents of a given district experienced similar district-level exposure to bombing, individual exposures and responses to wartime violence within a given district varied greatly, e.g., depending upon one’s gender, life course stage, and socioeconomic resources. Shared community exposures, such as the destruction resulting from bombing, varied from village to village (thus varying within districts), and individual exposure to those killed and injured due to war violence further varied within those villages.

Though prior studies document war and armed conflict, in the aggregate, as a proximate cause of migration, most lacked the data necessary to analyse variations in individual exposure to war-related stressors and how these exposures inform migration decisions (Bohra-Mishra & Massey, 2011). Additionally, most prior studies of migration in war contexts overlook migration’s totality, particularly the persistent undercurrent of mobility related to employment, education, or family formation (ibid.). We draw upon survey data collected in the Vietnam Health and Aging Study (VHAS) to investigate how wartime violence exposure influenced diverse migration types. We expand upon previous research in two ways. First, we follow prior researchers in examining geographically aggregated war exposures, but we expand upon this with respondents’ self-reported exposure to specific war-related events and conditions. This enables us to examine the relative influence of direct experiences versus wartime milieu. Second, for all migrations occurring during the Vietnam War, VHAS respondents gave a reason for their migration, allowing us to disaggregate displacements, deployments, economic migrations, and other moves. Side-by-side comparison of predictors for these distinct forms of migration allows investigation of the extent to which traditional motivations for migration persisted during the Vietnam War. Finally, we contribute a novel discussion of military deployments as an important form of migration that shares characteristics with both displacement and economic migrations, but to our knowledge, has not been researched as a form of migration. These contributions generate a nuanced understanding of the diverse forms of wartime mobility met by soldiers, militias, and noncombatants in Vietnam.

Background

Historical Context

While early analyses framed the Vietnam War principally as an episode in American history, scholars have begun to reveal the Vietnam War as a Vietnamese War and one that profoundly shaped Vietnamese society (Merli, 2000; Miller & Vu, 2009). The protracted conflict, officially waged between the Democratic Republic of Vietnam (DRV, or North Vietnam) and the Republic of Vietnam (RVN, or South Vietnam), is conceived in official Vietnamese accounts as ‘Kháng chiến chống Mỹ,’ or ‘the American War,’ given the Americans’ financial and military support to the South Vietnam regime and the casualties and damage wrought by American bombings and military engagement. The societal impacts of the American War have been multifaceted and numerous. The military success of the DRV was contingent upon society-wide mobilisation, inclusive of women and men across diverse ages and social class backgrounds (Teerawichitchainan, 2009). The toll of the American War, estimated widely as causing one to two million Vietnamese deaths, an excess of mortality concentrated especially among men in their young adulthood, imposed unfathomable familial and societal costs (Clodfelter, 1995; Hirschman & Nguyen, 2002; Merli, 2000). War’s violent legacy lives on among those combatants and noncombatants suffering from long-term illnesses and injuries linked to a wide range of violent, traumatic stressors (Korinek et al., 2020; Rydstrøm, 2012), and it is deeply embodied and embedded in the landscape through dioxin and UXO contamination which continue to cause death and injury and challenge healthy livelihoods (Stellman & Stellman, 2018). Finally, as this paper highlights, war’s social and demographic consequences extend beyond mortality and morbidities to encompass diverse migrations during war and post-conflict.

The American War in Vietnam was not the only war to affect development and migration processes in 20^th century Vietnam. This war came on the heels of the First Indochina War and was followed by conflicts with Cambodia and China. While the Vietnamese populace was no stranger to the adverse effects of armed conflict, the bombings and violence of the American War were, by far, the most intense (Miguel & Roland, 2011) and created extensive displacement.

‘It is estimated that in South Viet-Nam close to 12,000,000 civilians, over half the population, fled from one side or the other from armed conflict, or were evacuated or relocated, some of them more than once… In North Viet-Nam the dislocations were also enormous, though the figures are hard to come by’ (Weisner, 1988, p. xvii).

Though the war exerted considerable pressures to migrate, other aspects of Vietnam’s historical and social context were also consequential for migration, specifically, economic development and the population distribution policies of the DRV (Anh et al., 2012; Dang, 1999).

Migration in the DRV during the 1960s and 1970s unfolded amidst war, a state-controlled industrial structure, and collectivisation of production (Porter, 1993). After the departure of the French, DRV leadership aimed to establish Soviet-style urban industrial centres, but these attempts largely fell short, and Vietnam remained ‘under urbanised’ due to scarcities and weak investment in infrastructure (Thrift & Forbes, 1985, p. 288). Furthermore, collectivisation of the economy, along with extensive bombing damage, labour shortages, and redirection of resources for the war effort, stalled industrialisation in the DRV during the 1960s and 1970s (Dang, 1999; Stubbs, 1999). Weakly developed transportation infrastructure, worsened by wartime damage, also constrained internal migration at this time (Porter, 1993).

The household registration, or ho khau, system, further constrained mobility. Ho khau functioned as a local ledger, recording official community members and providing registered members with access to services and communal land. In addition to surveilling the populace, ho khau restricted labour mobility, as potential migrants needed to obtain a ‘moving certificate’ from the community they wished to leave. Household registration policies were tools of the State’s regional planning and population redistribution efforts (Desbarats, 1987), aiming to resettle residents of overcrowded areas and quell urbanisation since it was ‘deemed detrimental to economic progress’ (Hardy, 2001, p. 201). While the DRV’s ho khau system allowed for migration toward rural and upland areas in advancement of socialist development goals, it generally acted to slow urbanisation and migration (Desbarats, 1987; Hardy, 2001; Thrift & Forbes, 1985).

Supplementing these societal influences on migration, individual characteristics, such as gender, positioned people within society, structuring opportunities, social responsibilities, and social pressures. Though Vietnamese military organisations were overwhelmingly male (Goldstein, 2003), ‘the war…blurred the rigid roles of women and men whenever women replaced men: not only men but women too would enlist in the army; women would take over production tasks and even act as heads of household in the absence of a husband’ (Rydstrøm, 2012, p. 277). While the vast majority of deployments in the People’s Army of Vietnam were filled by young men, young women were heavily represented in the Thanh Niên Xung Phong (TNXP) production units that repaired damaged roads, bridges and otherwise supported the war efforts. Women in the TNXP, many in their teens, were often sent on distant missions, such as in the mountainous terrain which sheltered the Ho Chi Minh Trail, for several years at a time (Van Dyke, 1972).

In the absence of war, the predominance of patrilocality meant that women tended to join their husbands’ families after marriage, a custom that often entailed women’s migration (albeit across short distances) in early adulthood (Hirschman & Nguyen, 2002; Thao & Agergaard, 2012). Wartime workforce shortages may have encouraged women’s geographic mobility in a structure that otherwise favoured men for educational and professional opportunities (Werner, 1981). Since gender conditions family roles, employment, and military participation, we expect overall migration rates to be lower for women than for men. Moreover, since gender influences modes of military participation, we anticipate that gender will interact with military participation in predicting migration.

In light of the ho khau system and the characteristics of Vietnam’s economic development, low migration levels would be anticipated for our sample of older, largely rural origin adults were it not for the war. Our cohort’s prime years of labour force participation concentrated in the years of state-led command economy when ho khau restrictions were difficult to overcome. With limited urbanisation and development in 1960s and 1970s Vietnam, the locational disparities that frequently underlie internal migration and urbanisation were relatively weak. However, military enlistment, which often involved migration, offered an alternate path to economic advancement (Rottman, 2012; Turley, 1972; Van Dyke, 1972). Thus, motivations behind decisions to join the military may have paralleled those for economic migrations. According to Rottman (2012), males age 16–45 were liable for conscription into the North Vietnamese Army (NVA) during the war, and males 18 or older could be drafted. It was also not uncommon for men of all social classes to perform their “sacred” duty to their country by enlisting (Duiker, 1995; Malarney, 2001). Conscripts migrated first for training and later for “mobile combat,” i.e., deployment (Rottman, 2012).

The peak American presence in Vietnam during the War (1965 to 1975) exposed civilians and members of the military alike to unprecedented levels of bombing, toxic chemicals, shortages of food and water, and the effects of widespread violence, such as death and severe injury (Clodfelter, 1995; Miguel & Roland, 2011; Weisner, 1988). Prior research documents the enormous psychological impact of seeing those killed or injured due to war violence (Bustamante et al., 2017; King et al., 1999). These types of exposures likely have a particular immediacy, precipitating perceived threat and fear, which can trigger migration (Czaika & Kis-Katos, 2009; Davenport et al., 2003). In addition, the inhospitable conditions associated with conflict also contribute to the perceived threat and fear that mediate war exposure and migration (Fontana & Rosenheck, 1999). Micro-level event-centred approaches underscore how individual responses to war-related events vary by event type and intensity, recommending that researchers disaggregate war-related event types to permit a more nuanced understanding of decision-making patterns around migration and other major life events (Williams et al., 2012). Thus, we hypothesise that the micro-level experience of war-related events spurred migration despite ho khau restrictions and historically low levels of economic migration, but the effects likely differed across types of exposures.

Theoretical Framework

Demographers have observed an increase in the scale, complexity, and diversity of forced migration in the early 21^st century, with the mobility of internally displaced persons and refugees reflecting a complex interplay of political, economic, and environmental drivers (Hugo et al., 2018). In classical conceptualisations of forced migration (Kunz, 1973), movement is a response, either acute or anticipatory, to cataclysmic events or perceived insecurity and often involves a temporary haven prior to longer-term settlement. Armed conflict is one such event or condition, among many, that threatens security and propels involuntary movement or displacement (Castles, 2003; Hugo et al., 2018).

To inform our analysis of migration in the lives of elder Vietnamese who endured decades of armed conflict and social upheaval, we rely on two theoretical perspectives that integrate micro- and macro-level insights: life-course theory (Elder Jr. & O’Rand, 1995; Jasso, 2003) and the micro-level event centred approach (Williams et al., 2012). We fuse these theories to understand how war-related events combine with individual drivers to shape wartime migration. We attend to three fundamental principles of life course theory: timing, agency, and historical context. Following the micro-level event-centred approach, we disaggregate war-related events experienced at the individual level.

The timing principle of life course theory suggests that there are patterned timings of life events related to age. People commonly migrate for work, education, and first marriage when they are in their early twenties (Findlay et al., 2015; Rogers & Castro, 1981). However, wartime mobilisation simultaneously reduces economically-motivated migration and escalates deployment-related migration, upending traditional migration patterns for those exposed to war in their early adulthood (Hirschman & Nguyen, 2002; Williams et al., 2012). Although the associations between class background and deployment characteristics are not well understood in the DRV setting, several studies indicate that young people’s service was not tightly linked with parental social class (Merli, 2000). Thus, we hypothesise that adult respondents across the spectrums of age and socioeconomic backgrounds were likely to experience deployments during wartime. Though relatively uncommon, we expect economic migrations will remain most prominent among young adults from higher socioeconomic backgrounds.

Historically, migration scholars distinguished voluntary and forced migration based on whether migrants exercised agency in migrating (Davenport et al., 2003). Yet, even forced migrants make decisions about their flight, however coerced or fearful they might be (Korac, 2009, p. 8). They are strategic in their decisions about when and how to flee and actively constitute the social structures that enable their flight by creating networks and relations with various institutional actors (Korac, 2009; Williams et al., 2012). While sociologists often operationalise agency solely intentional action, we follow Marshall in recognising “co-constitution” of structure and agency, such that intention cannot be enacted without sufficient personal resources and an obliging social structure.

Notably, even in times of war, degrees of agency vary (Williams, 2015). Individuals fleeing bombings may have more opportunities to exercise agency in their flight than those ordered to evacuate by government officials (Engel & Ibáñez, 2007; Olsaretti, 1998). Similarly, soldiers deployed to new locations could be considered ‘forced’ migrants, as their opportunity for alternate migration decisions is quite limited (Bartram, 2015; Olsaretti, 1998). Though deployment may share some characteristics of displacement, the deployed are a qualitatively distinct form of ‘forced’ migrant. Moreover, to our knowledge, they are seldom, if ever, discussed as migrants in canons of migration and displacement.

For service-aged men, the distinction between military draftee and volunteer speaks to the exercise of agency related to service. Volunteers to the NVA were quite common during the peak years of the war, encouraged by the rhetoric of political rallies and announcements, or motivated in response to American bombings (Bui, 1999; Rottman, 2012). Members of the military who volunteered for duty exercised their agency, knowing this decision could result in deployment to other locations. Manpower needs of the NVA, which varied significantly over the war years, influenced rates of both volunteer service and draft conscription (ibid.).

Recognising the varying levels of voluntariness of Vietnamese migration, we follow prior scholars in viewing migration on a continuum of agency and compulsion (Erdal & Oeppen, 2018; Williams, 2015); thus, we focus on the reasons for migrating rather than using a forced/voluntary dichotomisation. We also recognise that individual reasons for migrating reflect structural conditions, individual characteristics such as gender, and resources (Bohra-Mishra & Massey, 2011).

We hypothesise that a respondent’s draft/volunteer status, an indicator of the choice component of agency, will influence their deployment-related migration patterns. We also expect that individual agentic actions will be intertwined with their characteristics and resources. Specifically, we hypothesise that respondents’ early life circumstances influenced how they assessed and responded to push and pull factors, in turn influencing their migration aspirations and choices, such that severe economic disadvantage inhibited migration.

Materials and Methods

Vietnam Health and Aging Study

The data for this paper come from the 2018 Vietnam Health and Aging Study (VHAS). The VHAS was designed to investigate the ‘lingering effects of exposure to the Vietnam War on older adult wellbeing, health and mortality’ (Korinek et al., 2019) and included modules documenting migration histories. VHAS collected data from 2,447 men and women aged 60 and older in four purposively selected districts within three provinces in northern Vietnam—Hanoi, Ninh Binh, and Quang Binh. Investigators chose these locations to capture a spectrum of exposures to war, as indicated by the intensity of bombings across the DRV during the 1960s and 1970s (see Figure 1; ibid.; Miguel & Roland, 2011). Within these four districts, stratified random sampling methods were used to select 204 individuals from each of 12 communes (Korinek et al., 2019). It should be noted that the sample is selective in that it is a probability sample only of people who survived the war, remained in Vietnam, and did not die from other causes before 2018.

Figure 1.

Bombing intensity in northern Vietnam.³

Measures

Dependent Variables—Wartime Mobility

Our analysis uses five dependent variables to understand northern Vietnamese men’s and women’s wartime² migration patterns: total number of interprovince migrations, number of total interprovince migrations not including deployments, interprovince deployments, war-induced displacements (e.g., village evacuations), and economic migrations. The VHAS survey gathered complete interprovince wartime migration histories from 1965–1975, documenting the date, destination, and main reason provided by the respondent for each interprovince move lasting three months or longer. The reasons offered include: military deployment; flee/evacuation; neighbourhood/houses were bombed/ destroyed; marriage; job/work; care for elderly/family members elsewhere; education; other reason (specify); and don’t know/don’t remember. We constructed all of the dependent variables as the count of moves of the particular type. To get a picture of overall migration patterns, we dependent variable model the number of all interprovince migrations from 1965 to 1975. However, since deployments comprised roughly 80 per cent of all wartime migrations, we separately examine non-deployment migrations. Finally, to allow comparisons across major migration types, we perform separate analyses of deployments, economic (or traditional) migration, and war-related displacement. We included deployments among migration types for two reasons. First, deployments constituted the most common type of migration for VHAS respondents. Second, on the spectrum of agency as it relates to migration, deployments are a largely involuntary migration consequence of military operations (Erdal & Oeppen, 2018; Williams, 2015). Though the source of involuntariness is the military rather than generalised war violence or village evacuations, we suspect deployments may share commonalities with war-related displacements. Our measure of economic migrations includes the subset of all migrations for work, education, and, for women, marriage. War-related displacement includes migration due to bombing and village evacuations. For this item, we included displacements of any distance and duration.

Independent Variables

Historical context.

Exposure to different types of wartime violence and related threats to life and livelihood, the types which compel displacement, were geographically uneven during the American War (Miguel & Roland, 2011). The VHAS’s purposive sampling of three provinces representing high, moderate, and low-intensity bombing allows us to use location during the war as a proxy for exposure to bombing and related casualties and upheaval. We supplemented this measure with six variables assessing the respondent’s exposure to stressful wartime experiences and conditions: military participation, nearness to death and severe injury, shortages of food and water, inability to sleep due to noise or inhospitable conditions, and having experienced a fear of being injured or killed. Categories of military participation during the American War include none (i.e., civilian), informal military (e.g., militia, Thanh Niên Xung Phong (TNXP)), and formal military (e.g., NVA). To capture the intensity of exposure to death, we constructed a variable delineating the number of times the respondent saw dead or seriously injured people. Because the distribution of values was right-skewed, we logged the variable in our multivariate analyses. Deprivation (food, water, and sleep) and fear of death are binary indicators of the respondent’s wartime experiences.

Timing.

To assess the effect of event timing on migration, we used the respondent’s age at the start of the war (i.e., in 1965), recoded to capture the life course stages traditionally associated with migration or ‘settledness’: childhood (less than 15 years old), adolescence and young adulthood (aged 15–24), and adulthood (24 and older).

Agency and resources.

Our study includes one indicator of agency—whether the respondent volunteered for military duty or was drafted. This variable contains three categories: neither (e.g., civilian), volunteer, and draftee.

To reflect the conjoined nature of structure and agency, we include several measures related to individual characteristics and resources that enable them to exercise agency. These include childhood hunger, childhood health, and parents’ education. Childhood hunger is a binary indicator drawn from the question, ‘Did you experience severe hunger (e.g., going to bed hungry, having to scavenge for food) for a month or longer due to food shortage during your childhood (i.e., from when you were born, up to and including age 15)?’ Childhood health was assessed in five categories ranging from Very poor to Very good. We recoded childhood health as a binary variable with the categories V. poor/Poor/Fair and Good/V.good. Due to extremely low variability in parental education levels, we measured each parent’s education as a binary variable indicating No education or Some education. Finally, we control for gender.

Analytical Approach

This study models potential explanations for respondents’ total number of wartime interprovince migrations as well as their frequency of deployments, economic migrations, and displacements. Because our dependent variables are counts, we used survey-adjusted multilevel mixed-effects Poisson and negative binomial models.

We chose a multilevel mixed-effects model structure because the Vietnamese provinces, districts, and communes are naturally hierarchically stratified. The error structure of hierarchical data often houses correlations between the errors of individuals nested in the area, and single-level models are incapable of compensating for these correlations. Additionally, these strata share characteristics in common that we theorise are relevant to migration; for example, within the areas (and possibly adjacent areas), there are common socioeconomic, developmental, environmental, and wartime exposure characteristics. These potential categories of within-area correlation are not inconsequential to understanding migration. Thus, we use respondent’s province at the start of the war as random intercepts, separating geographically clustered variance and error from the random error.

Our analysis begins by examining descriptive statistics and simple bivariate associations. Next, we apply graphical descriptive techniques adopted from social network analysis and spatial analysis to elucidate migration patterns. Finally, we use the multilevel mixed-effects models to evaluate the relationship of various forms of migration to warzone conditions, exposure to death, socioeconomic wellbeing, and demographic factors.

Results

Descriptive Statistics

The migration histories of the older Vietnamese men and women in this sample reveal that half the sample never experienced a move between provinces despite living through periods of extensive armed conflict and substantial displacement. Among those who had moved, most have since returned to their province of birth. Overall, 71 per cent of the sample reported no migrations during the war, 15 per cent reported only one inter-province move during the war, and 10 per cent reported two to three moves. Despite this broad pattern of low inter-province mobility, the data were skewed, with 20 respondents moving six or more times and two respondents reporting nine inter-province moves lasting at least three months. Also, while war-related displacements were somewhat more evenly distributed across the sample (see Figure 2), 65 per cent of the respondents reported no displacements compelled by wartime violence.

Figure 2.

Number of migrations.

We assessed our respondents’ propensity to migrate based on exposure to the violent effects of war, life course timing, agency, and the individual characteristics and resources that inform decision-making around migration. Table 1 summarises these predictors. It is important to note that the number of displacements is higher than the number of interprovince moves because the former includes some intraprovince displacements. Among the predictors of migration, it is striking that the average number of experiences of seeing dead and seriously injured people is very high. However, 35 per cent of the sample had no experiences of this sort. Seeing the dead and seriously injured was far more common amongst members of the formal military. Finally, paralleling rates of displacement, 35 per cent of respondents reported fearing injury or death, and this fear, while uncorrelated with overall wartime moves, was significantly, though weakly, correlated with displacements (r = 0.29).

Table 1.

Summary Statistics.

	Mean/Proportion	SD	Max.
Migration
Total # of Interprovince Moves (during the war)	0.55	1.12	9
# Interprovince moves (no deployments)	0.10	0.39	5
# Deployments	0.45	1.01	8
# Moves for work, education, etc.	0.06	0.30	4
# War-induced displacements	1.28	2.34	10
# Other types of moves (family, illness, etc.)	0.004	0.06	1
Historical Context – War-related Events
Military participation‡
None (Civilian)	.34
Informal military (Militia, TNXP)	.26
Formal military	.40
Number times saw dead/seriously injured	8.76	9.58	30
Lack of clean water‡	.171	.377
Food shortages‡	.260	.439
Inability to sleep‡	.451	.498
Fear of death‡	.350	.477
Timing
Age in 1965 (Range: 6 – 52)	17.32	8.56	52
Agency
Drafted/Enlisted‡
Civilian	33.67
Volunteer/Enlistee	31.14
Draftee	35.19
Resources
Severe hunger as a child‡	.460	.498
Health as a child‡
Poor/very poor	.060
Fair	.298
Good	.492
Very good	.150
Father’s Education‡
Illiterate/No education	.280
Some education	.720
Mother’s Education‡
Illiterate/No education	.481
Some education	.519
Sex‡
Female	.512
Male	.488

To evaluate migration patterns, we looked at descriptive summaries of the spaces and spatial relationships in our data. First, we examined migrations between pairs of locations to identify patterns of movement. We adopt social network analysis graphing techniques to display these patterns. Figure 3 depicts the frequency of respondents’ wartime migrations between geographic regions, with thicker lines indicating higher frequencies. The absence of a line between two locations indicates that none of our respondents migrated directly between those locations. The figure shows that our respondents’ most frequent direct migrations occurred between the Red River Delta and the North Central regions (occurring 410 times for our respondents), with 52 per cent of those being migrations from the Red River Delta to the North Central regions. This pattern reflects the demands of the war and the locations of intense conflict, which is unsurprising since 82 per cent of wartime migrations were deployments. In contrast, respondents attributed only 2.5 per cent of interprovince wartime migrations to community evacuations or intense bombing. This pattern accords with the gradual spread of respondents away from the Red River Delta and across the country shown in Figure 4.

Figure 3.

Frequency of migrations between regions.

Figure 4.

Proportion of respondents residing in each region.

In preparation for more detailed analyses, we mapped the reasons respondents migrated away from their various regions of residence at the start of the war. Figure 5 confirms that during the American War, the majority of migrations were tied to military participation. However, the patterns become more varied across regions when considering the other reasons for migration. For example, in the two northernmost regions, migrating for work was the second most common reason for migrating, a likely reflection of the limited urbanisation and available jobs in these areas. However, Vietnam’s industrialisation and urbanisation remained low in that era (Dang, 1999), inhibiting work-related migration among most VHAS study participants. Interprovince migrations resulting from community evacuations and bombings were also notable in the Northeast and North Central regions, with fewer (among our respondents) occurring in the Red River Delta. In the following section, we present analyses investigating the factors theorised to influence migration in general and deployments, economic migration, and displacements in particular.

Figure 5.

Reasons for wartime interprovince migration (by region).

Multivariate Analyses

Our analyses include five models: 1) total number of wartime interprovince moves; 2) total wartime interprovince moves, minus deployments; 3) deployment-related moves; 4) moves undertaken to facilitate socioeconomic advancement; and 5) war-related displacements. The models evaluate the effects of historical context, timing, and agency on each form of migration.

Historical Context and Exposure to War

Key predictors of migration related to historical events and context include military participation, exposure to death, and exposure to inhospitable warzone conditions (see Table 2). First, our proxy for an aggregate measure of exposure, the respondent’s location in 1965, was significant only for non-deployment migrations. For example, residing in one of the northernmost provinces was associated with more moves for socioeconomic mobility. This accords with the region’s typically mountainous and rural character and its disproportionate receipt of ‘new economic zone’ migrants resettled by the government (Hardy 2001). On the other hand, residing within the Central region, Quang Binh province, or another country (Cambodia or Laos), i.e., areas more heavily impacted by war, was associated with more war-related displacements. In contrast, location at the start of the war was unrelated to overall wartime migration and deployment.

Table 2.

Multilevel Mixed Effects Models of Migration (coefficients exponentiated)

	All Wartime Moves	All Without Deploy	Deployment	Economic Moves	Displaced
Fixed Effects
Historical Context – War-related Events
Military participation (Ref = civilian)
Informal Military (Militia, TNXP)	2.15 *	0.69	—	0.73	0.91 **
TNXP Only (Ref = Militia)‡			21.68 ***
Formal Military	4.34 *	0.40 ***	30.52 ***	0.73	0.86
Number times saw dead/seriously injured	1.33 **	1.21 *	1.27 *	1.06	1.17
Lack of clean water	1.31 *	1.21	1.22 *	0.85	0.94
Food shortages	1.14	0.88	1.23+	1.01	1.54+
Inability to sleep	1.06	1.04	1.00	0.89	1.37
Fear of death	0.90	1.11	0.89	1.10	1.37 *
Timing
Life course stage (Ref = Child; < 15 in 1965)
Adolescent/young adult (15–24)	1.71 **	2.03+	1.59	2.38 *	1.15
Adults (25+)	0.92	1.20	0.96	1.21	1.12
Agency
Drafted into military (Yes = 1)			0.82 *	0.42 **
Resources
Mother’s Education (Ref = None)
Some education				0.96
Severe hunger as a child (Yes = 1)	0.71 **	0.70 *	0.75 *	0.67 **	1.04
Health as a child (Good/V. good = 1)	0.96	1.09	0.95		1.04
Sex (male = 1)	1.87 **	2.42 **	1.44+	3.21 ***	1.05
Constant	0.09 ***	0.09 **	0.01 ***	0.06 **	0.58
Random Effects
War Location (variance)	1.26	1.55 *	1.09	1.95 ***	1.36 *
N	2,267	2,433	1,538	2,315	2,297

⁺p<0.10

^*p<0.05

^**p<0.01

^***p<0.001

^‡The deployments model excludes civilians and separates the militia and TNXP.

As expected, military participation had a significant and positive association with the total number of wartime migrations, and the effect was larger for members of the formal military than for militia and TNXP participants. However, when deployments were removed from total moves, the effect was reversed. Military service similarly reduced the experience of displacements; informal military participation was associated with a reduction in displacements, while formal military participation was not significantly associated with displacement.

We observe several significant associations between wartime migration and individual-level exposure to wartime violence and inhospitable living conditions. Exposure to violent death and injury is a significant, positive predictor of every migration type except for employment- and education-related migration. Adverse living conditions exhibited varying effects across the five outcomes. Wartime migrations overall and especially deployments correlate positively with clean water inaccessibility. However, adverse living conditions were unassociated with wartime migrations once deployments were removed. This null result may reflect the nationwide pervasiveness of shortages and difficult circumstances during wartime (and the shorter, intraprovince distances of many migrations precipitated by wartime hardships). Notably, experiencing fear of death or injury was not significantly associated with general wartime moves, whether total moves or moves minus deployments. It was also unassociated with deployments and economic moves.

Adverse wartime conditions demonstrate a significant, positive association with displacement migrations, which encompass both inter- and intra-province migrations. Consistent with displaced persons’ greater odds of witnessing violent death or injury, fear of death was positively associated with displacement. The odds of displacement were also significantly greater among those reporting experiences with food shortages and the inability to sleep during wartime. Models estimated with military service interaction terms (Table 3) demonstrate that, relative to other respondents, members of the formal military who experienced a fear of death were less likely to experience displacement. This pattern suggests that due to their duties and expectations to remain in places of great danger and exigent threat, military participants were less likely to flee from home or village settings when encountering wartime violence and life threat.

Table 3.

Select Results from Models of Migration with Interactions (coefficients exponentiated)

	Deployment	Economic Moves	Displaced
Fixed Effects
Historical Context – War-related Events
Military participation (Ref = civilian)
Informal Military (Militia, TNXP)		1.05	0.94 **
TNXP Only (Ref = Militia)‡	34.20 ***
Formal Military	67.11 ***	1.40	0.96
Number times saw dead/seriously injured	1.96 *	1.06	1.17 **
Inability to sleep	1.01	0.89	1.36+
Fear of death	0.89	1.13	1.53 *
Timing
Life course stage (Ref = Child; < 15 in 1965)
Adolescent/young adult (15–24)	1.60 ***	2.43 *	1.15
Adults (25+)	0.96	1.29	1.11
Gender
Sex (male = 1)	1.43+	4.99 ***	1.05
Interactions
TNXP X Saw dead/injured	0.74 **
Formal military X Saw dead/injured	0.64 *
Informal military X Male		0.36 *
Formal military X Male		0.37
Informal military X Fear death			0.95
Formal military X Fear death			0.78 **
Constant	0.01 ***	0.05 **	0.46

⁺p<0.10

^*p<0.05

^**p<0.01

^***p<0.001

Event Timing

The life course timing of the American War, as assessed by the respondent’s age in 1965, demonstrates associations that diverge across types of migration. Generally, those in adolescence at the start of the war experienced more wartime migrations than their younger and older counterparts. The singular exception was displacements. Age was unrelated to displacements, suggesting that displacement more so than other migrations involved family units, moving together collectively, irrespective of age.

Agency and Resources

Our sole direct indicator of agency, volunteering for military service, was significant for two types of migration: deployments and economic migration. While only 40 per cent of VHAS respondents were members of the formal military, 74 per cent of those respondents were drafted into service. Compared to military enlistees and civilians, military draftees experienced fewer employment- and education-related migrations, possibly indicating significant disruption to their ‘natural’ migration propensity. Moreover, those who were drafted experienced fewer interprovince deployments than those who report having volunteered for service. These deployment patterns may reflect underlying differences across draftees and volunteers, such as volunteers’ earlier entrance into service and longer durations of service.

Several of the resources that promote respondents’ capacity for agency are important predictors of migration. Among these traditional causes of migration, characteristics associated with economic hardship or difficult early-life conditions, such as experiencing severe hunger as a child, were associated with fewer interprovince moves. As an indicator of socioeconomic status, parents’ education was unassociated with economic migration. This may reflect the limited degree of variation in parental education in this cohort and the unique meaning of social class and status in the post-colonial, communist DRV (Miller and Vu 2009).

Gender is also a critical predictor of various types of wartime migration. Men report a higher number of interprovince migrations across all types of migration, except displacement. As with age, gender is a nonfactor in displacement. However, men report significantly more interprovince military deployments and work- and education-related migrations than women.

Turning to the gender interactions (shown in Table 3), when we disaggregated the types of informal military participation, gender is no longer significant. These forms of military participation, which are deeply gendered, absorb the effect of gender on deployment. Figure 6 demonstrates that when predicting gender differences in migration at different levels of exposure to death and injury, there are significant gender differences in the number of deployments. In sum, gender is a pivotal determinant of most forms of migration. Nevertheless, it is inconsequential for internal displacements, suggesting that displacements are undiscriminating, while military deployments and economically motivated migrations, which were historically more common for men, remain gendered even in the context of war.

Figure 6.

Predicted mean number of deployments by gender and military service type.

Discussion

We began this paper intending to better understand how war affects regular patterns of migration. Zelinsky once described a ‘patterned regularity’ of mobility in response to economic development, the phases of which spread globally ‘like world pandemics’ as nations advance if ‘no major military or ecological disaster will occur’ (1971, p. 241). War and military mobilisations upend structural migration patterns, such as the urbanisation that often accompanies economic development, and disrupt individual migration proclivities prescribed by life course stage and socioeconomic circumstances (Adhikari, 2013; Horst & Grabska, 2015; Raymer & Rogers, 2006). However, as our study demonstrates, war is not a blunt force instrument displacing all in its wake. It exerts differential pressures on distinct types of migration, the effects of which reflect war’s timing, socio-historical context, and the agentic reactions of individuals. Additionally, our analyses suggest that war does not entirely undo deep-rooted causes of migration, nor established gender patterns of migration reflective of engrained cultural beliefs and social order. Rather, it reduces their frequency or postpones them (Bohra-Mishra & Massey, 2011; Hirschman & Nguyen, 2002).

At the start of the American War, the VHAS respondents were distributed in districts experiencing varying levels of bombing. While locations at the start of the war were impactful for wartime displacements and economically motivated migrations, they were unassociated with deployments. In contrast, microlevel exposures to war were strongly but differentially related to all migration types. This pattern of results, indicating divergent impacts of macro- and micro-level violence exposures upon specific migration types, demonstrate a vital need for researchers to consider individual experiences of conflict and violence when modelling wartime migration.

Critically, directly witnessing wartime death and injury was associated with increased migrations with only one exception (employment- and work-related migration). This finding reinforces prior research indicating that such an intense and immediate exposure likely generates a sense of threat and fear for one’s safety (Czaika & Kis-Katos, 2009; King et al., 1999) and supports the notion that threat and fear mediate exposure to war-related events and individual decisions to migrate (Frazier & Caston, 2015). Future studies should explore this mediation to more fully understand individual motivations to migrate in the context of war.

Significant historical events can disrupt the timing of longstanding forms of migration, e.g., migration for education, work, and marriage (Horst & Grabska, 2015). While these migrations were certainly suppressed, comprising only 10 per cent of all reported interprovince wartime moves, migration’s overall life course timing in the DRV seems relatively undisturbed. Nevertheless, the distribution of migration types is substantially altered. In particular, younger adults exhibit more interprovince migrations than their older counterparts during this peak wartime decade. Military participation, a path to socioeconomic mobility typically undertaken in adolescence (for the TNXP) and early adulthood (Merli, 2000), was associated with more frequent moves, deployments, and economic migrations between 1965 and 1975, augmenting migrations during this life course stage.

The labour demands of waging war also shaped migration patterns among members of the military (Rottman, 2012; Turner & Phan, 1999), both enlistees and draftees, such that heightened deployments disrupted normal migration patterns. Informal military service increased respondents’ exposure to conflict and the effects of violence relative to civilians but likely limited deployments and exposure to death relative to the formal military. Volunteer service in the TNXP and militias likely limited economic migration that may have occurred in the absence of war. Notably, in Vietnamese society, the government encouraged its citizens to support political and military initiatives, and as is typical in many collectivist societies, citizens contributed to those efforts in both major and minor ways (Guillemot, 2009; Parks & Vu, 1994; Rottman, 2012). Thus, cultural pressures also influenced agentic action vis-à-vis military enlistment and TNXP service, a nuanced interaction that we could not disentangle with the VHAS data.

A central contribution of this study surrounds gender’s impact on wartime migration. Our analyses reveal that gender conditions nearly every predictor of migration taking place during the American War. Gender is implicated in military service, the nature and frequency of war exposure, social roles connected to age and family, and economic roles and decision-making processes. Likely reflecting the gender norms of traditional Vietnamese society in which it was more common for men to seek work outside of their village while women remained in the village to maintain the household and agricultural production (Dang, 1999), the predicted number of interprovince migrations for men and women is significantly different, even after controlling for a host of other potential causes. The DRV’s preparations for war in the mid-twentieth century saw thousands of volunteers join the TNXP to aid in constructing and repairing the Ho Chi Minh Trail and other wartime support duties. The call for volunteers for the TNXP spanned gender, creating an avenue for substantial interprovince deployment migrations among women (Guillemot, 2009). However, the feminisation of domestic migration within Vietnam associated with women’s employment in export-oriented manufacturing and service industries would not peak until decades after the war (Bélanger & Pendakis, 2010; Thao, 2013; Thao & Agergaard, 2012). Our data substantiate this pattern, suggesting that during this early stage of economic development, while war waged, men exhibited more moves for employment and education, as well as more deployments and migrations overall. Though gender structures the traditional timing of migration and migration types (Raymer & Rogers, 2006; Werner, 1981), in wartime, gender structured military participation and thereby reshaped gendered migration patterns.

By disaggregating wartime migration into its constituent types, the present analyses reveal how wartime displacements diverge from other forms of migration undertaken during war. Displacements did not discriminate by age or gender. Additionally, displacements do not appear to be shaped by respondents’ social class background or human capital. These results suggest that when evacuation orders or war-related destruction compel populations to move, that compulsion is experienced widely across age, gender, and socioeconomic status. Observers of wartime evacuations in the DRV have commented on the community networks which received and supported displaced families (Bui, 1999). These networks likely lessened barriers to return-migration among evacuees and other displaced persons, including women who may have been migrating alone or with their children. Thus, our study demonstrates that while economic migration can be highly individualised, reflecting age, gender, and class background, war-induced displacements are experienced more uniformly among civilians and members of informal military organisations.

Finally, though our study found distinct sets of life-course predictors are important for each type of migration, our comparison of deployments with war-related displacements and economically motivated migrations illuminates the similarities between deployments and these forms of migration. For example, though the specific predictors differ, the context of war and exposure to the events of war are associated with both displacement and deployment, offering potential support for the notion of deployment as a unique form of “organised” migration with aspects that resemble involuntary migration (Anh et al., 2012). However, our sole measure of agency—draftee status—was associated with fewer deployments, distinguishing it from displacements and aligning it with economic migrations. Deployments further parallel economic migrations in that socioeconomic predictors are associated with both. These findings speak to the shared characteristics of deployments with both displacement and economic migration. Further, these findings offer support for the inclusion of deployments in future studies of wartime migration.

Unfortunately, our study was limited in the amount of information we had regarding individual decision-making contexts. Thus, while we could categorise migration into traditional categories of ‘voluntary’ and ‘forced’ based upon respondents’ broad, primary reasons for migration, we had less information about the voluntariness of military deployments and moves to new economic zones required by the government. We also lacked detailed information on moves of shorter distance and duration, and the retrospective nature of the data collection may have made delineating the timing of migrations and recall of minor moves difficult. Finally, the study data only capture migrations for those in what was then North Vietnam who survived the war. We anticipate that deployment migration numbers may be underestimated, especially among men, due to elevated wartime mortality rates and especially due to the many continuous short-term deployments that characterised wartime service. Despite these limitations, we were able to analyse the more significant moves and disaggregate those moves based on general migration motivations, gaining insights into the migration patterns of older adults who survived the American war and remained in northern Vietnam.

In conclusion, this study employed a life course perspective in its analysis of wartime interprovince migration in Vietnam, a useful tool for interpreting the differing effects of conflict, socioeconomic status, and demographic factors on the manifold types of wartime interprovince migration. Using this perspective, we highlighted the unique historical conditions in Vietnam en route to understanding how macro-level historical, economic, and cultural contexts interacted with micro-level factors, such as personal experiences of war, human capital, economic resources, and other characteristics to influence migration patterns. Notably, this study illuminates how the ramifications of conflict, socioeconomic factors, and demographic characteristics diverge for different types of migration. Moreover, since gendered social relations infuse exposure to conflict, social status, economic opportunity, and family roles, gender shaped both exposure to traditional migration causes and responses to those exposures. This knowledge will further our ability to identify migration patterns in contemporary conflict situations, allowing institutions to respond to public needs in timely and appropriate ways and provide suitable infrastructure and resources to manage migration streams and provide aid to migrants and their families.