How do perceptions, fears, and experiences of violence and conflict affect considerations of moving internally and internationally?

Abstract

This article draws on cross-country survey and qualitative data for local areas within Afghanistan, Ethiopia, Nigeria, and Somalia to explore how perceptions, fears, and experiences of violence affect how young adults consider whether or not to move, internally, within their own countries, or internationally. We shed new light on how different forms and intensities of violence and conflict, ranging from tribal violence in Northern Nigeria to the encroachment of the Taliban in Afghanistan, affect these considerations. We show that, in most cases, perceiving insecurity, fearing, or having experienced violence does not result in people considering moving. However, where significant, it is specific fears and experiences that affect considerations, rather than general perceptions of insecurity. We observe distinct patterns across the local areas, related to the duration, intensity, and nature of the conflict, underscoring the salience of sub-national-level data to better understand the interplay of conflict and mobility.

Introduction

If feeling in danger, or having experienced violence, how do people consider the option of leaving, within or beyond the country in which they are? Intuitively, if at risk, escape is an obvious strategy. Some existing evidence indeed shows a linear relationship between conflict and mobility: the more violence experienced, the greater mobility (Adhikari 2013; Davenport et al. 2003; Lubkemann 2008; Moore and Shellman 2004). However, evidence suggests that there is far more nuance, agency, and careful consideration at work in the processes underlying people's moving and staying behaviour in insecure settings (Bohra-Mishra and Massey 2011).

Individuals in conflict settings consider a range of alternatives, including but not limited to a binary between ‘staying or leaving’ (Moore and Shellman 2006). Which type of exposure to conflict people have matters; for instance, directly witnessing or being exposed to violence may result in post-traumatic growth and a delay in deciding to leave (Schon 2019). Having endured living in conflict longer appears to boost individuals’ capacity to cope both with violence and with mobility and associated risks (Ghosn et al. 2021). Despite the growing body of literature on the dynamics of conflict-related mobilities, we know little about people’s considerations about leaving, internally or internationally, from settings affected by violence and insecurity.

Significant attention in the research literature has been paid to movements across national borders as a response to conflict (Braithwaite et al. 2019; Polzer and Hammond 2008). However, the numbers of people moving within a country tend to be much larger than those crossing international borders, including in conflict-affected areas. According to the UNCHR, in 2022, some 53.2 million of the 89.3 million ‘forcibly displaced’ people globally were internally displaced (UNHCR 2022), that is, they had not left their country.

Therefore, a key aim of this article is to contribute to a more nuanced understanding of the varying roles of perceptions, fears, and experiences of violence and conflict in considerations of moving internally or internationally from conflict-affected areas. To this end, we build on other studies that analyse the links between conflict and violence, and different types of mobility. A study on Nepal finds that higher levels of violence lead to larger distance movements (Bohra-Mishra and Massey 2011). Other cross-country studies find that cross-border movements are more likely to involve state-sponsored genocide/politicide compared to other types of civil war or state coercion (Moore and Shellman 2006), or if the main perpetrator of violence is a state actor, or if the state is an ally of the main perpetrator (Steele 2019). Conversely, higher levels of internal displacement are found to be associated with the prevalence of rebel violence (Turkoglu 2022). A study on Syrians finds that cross-border movements are more likely for those who experienced personal persecution, imprisonment, and torture (Müller-Funk 2023), as compared to a generalized perception of growing insecurity (Schon 2019; Ghosn et al. 2021).

Investigating people’s motivations for and considerations about leaving or staying in insecure and conflict-affected settings may by some be seen as ethically questionable, inasmuch as a link is often made between the necessity of leaving, in the face of threat, as well as to the right to protection as a refugee, having crossed an international border, in line with the 1951 Refugee Convention (Schon 2019). We build on the recognition that voluntariness in decision-making is best understood as a continuum, from extreme cases of absolute involuntariness to extreme cases of entirely free choice (Erdal and Oeppen 2018). This means that the scope for human agency in the face of different structural constraints and opportunities will vary, both between individuals in similar contexts and between contexts, based on a range of factors (Ghosn et al. 2021; Turkoglu 2022).

Moreover, evidence suggests that even in highly insecure and violent settings, ‘traditional migration determinants’, such as economic opportunities, remain highly relevant (Bohra-Mishra and Massey 2011; Engel and Ibáñez 2007). This is not to underplay the constraints on many people’s agency present in conflict-affected settings, resulting in involuntary immobility in such contexts, likely at a greater human cost than in many non-conflict settings (see also Carling and Schewel 2018; Lubkemann 2008).

We propose that the approach we build on and adopt lends itself to challenging the risk of group-level victimization of refugees and other ‘forced migrants’ per se by instead focusing our attention on unpacking the very real considerations underlying people’s decisions to leave (or stay) when these are made while facing different types and intensities of violence, from crimes to conflict. For this reason and because our underlying survey data uses this term, we use the term migrant for all those moving, rather than pigeonholing sub-groups into specific boxes.

The present understanding of whether it is people’s perceptions of insecurity, actual experiences of violence, or both that matter most in relation to whether they consider leaving or not remains inconclusive (Steele 2019). Schon (2015) draws attention to the idea of fear of conflict actors as an explanation for increases in displacement. Müller-Funk (2023) discusses the idea that all types of violence—whether self-experienced or not—are subjectively perceived and affect migration decisions as such. Reinforcing the idea that it is not just actual experience of violence that leads to a decision to leave, Schon (2016) analyses the dramatized—but isolated—incidents of violence employed at checkpoints by armed state and non-state actors in Somalia and Syria to create uncertainty, which act as a deterrent for leaving. Meanwhile, studies that do take into account subjective experiences and perceptions of violence and conflict stand in contrast to much of the literature which remains focused solely on ‘objectively measured’ data on conflict, often either in specific local contexts or at more aggregate national or cross-national levels (e.g. Moore and Shellman 2006). Furthermore, few studies, to our knowledge, strive to combine analysis of both internal and international migration (considerations) jointly in contexts of violence and insecurity, despite the recognition of various interconnections of internal and international mobilities (Echevarria-Coco and Gardeazabal 2021; King and Skeldon 2010).

In this article, we contribute new knowledge on the roles of violence and conflict in migration decision-making, first by disaggregating violence and conflict into more subjective measures such as perceptions, fears, and experiences of conflict and violence. Second, we focus on the specific issue of whether or not mobility is being considered—not whether it happens or whether it is desired. Third, we distinguish explicitly between considering internal and not only international mobility.

In this article, we answer the research question:

How do different perceptions, fears and experiences of conflict and violence affect considerations about internal and international migration among people living in conflict-affected local areas?

We analyse this relationship using a unique dataset covering more than 5000 respondents across ten local areas in Afghanistan, Ethiopia, Nigeria and Somalia, allowing us to delve into specific dynamics at the sub-national level. The article continues in the next section by briefly describing these areas and the types of violence and conflict seen in each. We then give an overview of the data utilized and the empirical strategy. The next section presents findings, showing the effect of perceptions, fears, and experience of violence on considerations about internal and international migration. We then discuss the implications of these findings while the final section concludes.

The local areas

This article draws on the MIGNEX survey data for ten local areas (Figure 1), which we selected from the larger dataset of twenty-six local areas because they had all seen violence and fighting in recent months, years, and decades. The types and intensity of violence differ, however, as well as the time elapsed. This gives us an opportunity to understand how different forms of violence may affect mobility considerations.

Figure 1

Location of the ten local areas. Source: Jørgen Carling for MIGNEX.

We cover three local areas in Afghanistan: Sharake Jabrael, Behsud, and Sharake Mahdia. In all three areas, the encroachment of the Taliban as well as decades of conflict and fighting were a central part of people’s lives at the time of data collection (Akakhil et al. 2022; Alizada and Murray 2022; Majidi et al. 2022). More than two-thirds of young people in these areas had feared violence in the past 5 years, and more than a third personally experienced some form of violence (Figure 3). While the fighting between the Taliban government and the Islamic Republic of Afghanistan was present in all areas, they also differed in some ways. In Sharake Jabrael, fear of the Taliban was intertwined with ethnicity and religion, as an area with high numbers of Hazara. In Sharake Mahdia, insecurity also resulted from crime, which is linked to growing poverty—nine in ten young people did not think it was safe to walk the streets at night (Figure 2).

Figure 2

Thinks it’s not safe to walk the streets at night (%). Source: MIGNEX Survey (mxs-restricted-v1.dta). N=13,172. Data are weighted to reflect the survey design.

We draw on two areas in Ethiopia: Batu and Moyale.¹

1

Data was also collected in Kombolcha, but as the 2021 conflict prevented data collection from being completed, the data is excluded from the analysis here.

The civil war between the Tigray region in the North and the government of Ethiopia was the main conflict dominating the country at the time of data collection. Batu has also seen ethnic tensions, while Moyale has seen security challenges due to renewed conflict on the Oromo and Somali border (Adhanom and Murray 2022; Kasavan et al. 2022a). The latter local area stands out because every second young adult has personally experienced violence in the past year, while security and safety concerns are also high (Figure 3).

Figure 3

Experience and fear of violence in past 5 years (%). Source: MIGNEX Survey (mxs-restricted-v1.dta). N=13,172. Data are weighted to reflect the survey design.

We have three areas in Nigeria: Down Quarters, Awe, and Ekpoma. Down Quarters is an area within Kaduna city in the insecure Kaduna state experiencing religious and ethnic violence (Adamu et al. 2022). While Down Quarters is itself seen as a peaceful haven, more than 80 per cent of young people do not consider it safe to walk the streets at night (Figure 2). Heightened tensions between farmers and herders over land resources have led to violent clashes in the area surrounding Awe (Genyi et al. 2022). Ekpoma, a university town, is considered much safer by Nigerian standards, and security concerns are lower (Figure 3); see also Aghedo et al. (2022).

The final two areas are in Somalia: Erigavo and Baidoa. The Somali civil war of the 1990s and 2000s and the ongoing al-Shabaab insurgency are the key forms of conflict. Relative to other locations in Somalia/Somaliland,²

2

For brevity, Somalia will be used throughout the rest of this article.

Erigavo is considered safe by residents, and safety and security concerns are generally low compared to the other areas (see Figures 2 and 4); see also Ahmed et al. (2022). Baidoa saw a lot of fighting during the civil war, but the city itself has become much safer in recent years, while the surrounding countryside is still held by Al-Shabbab (Kasavan et al. 2022b). Hence, large numbers of young adults have feared violence in the past five years (Figure 3).

Figure 4

Concerns by local area (%). Source: MIGNEX Survey (mxs-restricted-v1.dta). N=13,172. Data are weighted to reflect the survey design.

Having established that security and safety concerns are present to some degree in all ten areas, we now compare this to other concerns for the future that young adults have. Figure 4 shows the concerns for the next 5 years that respondents have, including insufficient income, disease and health, conflict and violence, climate change, and losing traditions. It shows that besides concerns about conflict and violence, young adults have other concerns, and for eight of the areas, conflict and violence are not the primary concerns. This varies greatly by area—as to be expected, it is the primary concern in Sharake Jabrael and Sharake Mahdia (both in Afghanistan). Meanwhile, insufficient income is the primary concern in five areas, and concerns about disease and health are also prominent, some 18 months into the COVID pandemic when data was collected. It is also worth noting that concerns about climate change and losing traditions in some areas are of similar proportions to concerns about, for instance, insufficient income or indeed conflict and violence. As such, these descriptive statistics already suggest that attentiveness to a broader range of factors affecting considerations about migration is worthwhile.

Before moving into the main analysis, we present an overview of migration considerations across the ten areas. We use the variable ‘has seriously considered internal/international migration in the past year’. This variable is more factual than attitudinal, as migration aspiration questions often are. In the way it has been operationalized in the question above, we can assume it reflects a mix of respondents’ awareness of migration as an option to be considered and their assessment of this option in terms of both feasibility and desirability (Carling 2019: 9). Since the question asks about ‘seriously’ considering—and has a timeframe: ‘in the past year’, responses offer a view of whether or not leaving (internally or internationally) is something respondents have invested any substantive thought into recently or not. We suggest that this, in contexts of insecurity, violence, or conflict, is a worthwhile and contextually appropriate approach to better understanding conflict and (im)mobility dynamics.

In seven of the ten local areas, 60 per cent or more of respondents had not seriously considered internal or international migration in the past year. In two areas (Moyale and Baidoa), this proportion was 80 per cent. This already points to the need to understand people’s immobility better, as even in conflict settings, the vast majority of young adults do not seem to consider leaving as a relevant option for themselves.

Meanwhile, in Batu, Down Quarters, and Ekpoma, higher proportions have seriously considered migration. As Figure 5 reveals, there is quite a lot of variation between the local areas as to whether those who had considered migration seriously were considering only internal—only international—or both types of migration. In the subsequent analyses, this variation is something we explore further.

Figure 5

Considerations of internal and international migration (%). Source: MIGNEX Survey (mxs-restricted-v1.dta). N=13,172. Data are weighted to reflect the survey design.

These initial descriptions of the local areas and patterns show that there is large variation in both perceptions, fears, and experience of conflict and violence, as well as considerations of internal and international migration. We now move on to describing our dataset and methodology.

Empirical strategy

Survey data

This article draws primarily on the MIGNEX survey data (version mxs-restricted-v1.dta). The focus of this survey was to collect comparative data on migration and development experiences, perceptions, and processes across individuals in twenty-six local areas. These local areas had been previously selected on the basis of nine different ‘developments’. While none of the local areas had been selected on the basis of a deterioration of security or the acute presence of conflict, by the time the data was collected, several of the local areas saw high levels of insecurity.

The survey consists of more than 13,000 in-person interviews with young adults (aged 18–39) across twenty-six areas in ten across Africa, Asia, and the Middle East. Data was collected by MIGNEX team members or sub-contractors, with extensive training, guidelines (on all aspects of the survey including ethics), and quality monitoring in place for all part of the data collection team (Hagen-Zanker et al. 2023a). The survey sample of 500 per area is roughly representative of young adults at the local area level, drawing on a three-stage, probability-proportional-to-size cluster sampling strategy with random walks (Hagen-Zanker et al. 2023a). For this article, we focus on ten areas, where data collection took place in the time frame of June to August 2021.

The data was collected among residents in the local areas, which means that contrary to the challenge faced in much research on migration and displacement, which surveys those who have moved and therefore suffer from a mobility bias, our data could conversely be challenged as having an ‘immobility bias’. This could be seen as a challenge in relation to analyses of considerations of migration, as we cannot account for the proportion or traits of those who have currently left the local area (though we do control for past mobility; see below), while we can anticipate that demographics, perceptions, and some outcomes might be affected by those who left. Furthermore, we might assume that those staying may either be poorer (and thus unable to leave even if they wanted to) or richer (and thus able to stay despite, e.g. conflict, violence, or other adversity, which they can afford to manage). In our regression analysis, we attempt to control for any observable factors including past mobility; however, we are aware that nevertheless, our findings will retain some bias from the composition of our sample.

Our data was collected among respondents aged 18–39 years, which allows a particular age-based perspective on considerations about migration. The specific age group aids our analysis, as given the 21-year bracket, we are not including vastly different life stages, but instead focusing on young adults, who, according to other studies, are known to have higher levels of migration aspirations than older age groups.

Regressions

To address our research question, we estimate the following equation for each of the ten local areas separately and for the pooled data:

$$\left(1\mathrm{ }or\mathrm{ }2\right)\mathrm{ }{\mathit{MigCons}}_i=\mathrm{ }{\alpha }_i+\beta \left({\mathit{violence}}_i\right)+\gamma ({\mathit{X})}_{\mathit{i}}+{\phi }_c+{\epsilon }_i$$

where the dependent variable ${\mathit{MigCons}}_i$ of young adults corresponds to (1) internal migration considerations, a dummy variable equal to 1 if a young adult has seriously considered migrating internally in the past year, 0 otherwise or (2) international migration considerations, a dummy variable equal to 1 if a young adult has seriously considered migrating internationally in the past year, 0 otherwise. For our independent variable of interest, our article seeks to investigate how violence and conflict, and specifically, which dimensions of them influence considerations to migrate internally or internationally. ${\mathit{Violence}}_i$ refers to nine dummy variable measures of perceptions, fears, concerns, and experience of violence, detailed in Table 1.

Table 1

Dependent and independent variables of interest.

Variable type	Dimension	Measures (all binary variables)
Dependent	Internal migration consideration	Respondent has seriously considered internal migration in the past year
	International migration consideration	Respondent has seriously considered international migration in the past year
Independent	Perceptions of safety/insecurity	Thinks it is not safe to walk the streets at night
		Thinks RA has become more dangerous, compared to 5 years ago
	Concerns/fears of violence	Concerned about conflict and violence
		Has feared violence at political rally/protest (past 5 years)
		Has feared violence by armed forces/militias (past 5 years)
		Has feared other violence in local area (past 5 years)
	Experiences of violence/conflict	Has personally experienced violence (past 5 years)
		Respondent or HH member has experienced physical violence (past 5 years)
		Respondent or HH member has experienced theft (past 5 years)

We assess the effect of each independent variable on migration considerations separately. The vector of covariates ${\mathit{X}}_{\mathit{i}}$ represents a set of controls including respondent’s age, gender, number of household members aged 18–39, respondent’s educational attainment, religion, marital status, whether respondent is a parent, whether the respondent is working, perception of possibility of finding a good job in the local area, perceived household relative living standards, household wealth index, measure of risk aversion, having migrant family/relatives/friends living outside of the local area or abroad, and having lived in a high-income country for at least 1 year.³

3

This last measure allows us to control throughout the analysis for previous migration experiences and to account, to a certain degree, for the “immobility bias” of our sample.

Finally, ${\phi }_c$ captures cluster fixed effects to control for time-invariant unobserved characteristics at the cluster level that could impact both our measure of violence/conflict and migration considerations. Standard errors are also clustered at the cluster level to allow for arbitrary correlation within local area clusters. For the pooled data analysis, combining all ten local areas, we also control for local area fixed effects to account for time-invariant local area unobserved characteristics. We do not pool at the country level, as the local areas within a country are not representative of countries.

The descriptive statistics of our dependent and independent variables of interest are detailed in Table A.1.

We employ a linear probability model (LPM) to estimate Equation 1/2 as our main specification, given that the two outcomes of interest are binary measures. We prefer a LPM over logit and probit estimators, as it has the advantage of allowing the application of fixed effects (Freedman 2008). Given the nested nature of our data, we apply fixed effects to remove time-invariant heterogeneity that introduces bias into our coefficients of interest. Moreover, an LPM allows the direct interpretation of coefficients as probabilities, whereas logit and probit estimators require additional methods to estimate marginal effects or predictions at the means (Angrist and Pischke 2009). We do not intend to claim causality in this study, but employing a more parsimonious model such as the LPM while controlling for the time-invariant characteristics of local areas allows us to isolate the effects of our covariates of interest more carefully.

As a robustness check, we also estimate logit and probit models, for which we compute the marginal effects of the covariates evaluated at their mean value and compare findings across the three models. The effects obtained with logit and probit estimators are consistent with those obtained with the LPM. The differences in the magnitudes of the coefficients across models are small, and the coefficients remain significant across all three models.⁴

4

The full LPM, Logit and Probit regression results are available from the authors upon request.

Qualitative data

The discussion and interpretation of the results in this article are supplemented with insights from MIGNEX qualitative data collected in each of the ten local areas. This consisted of fieldwork, usually with a team of researchers, conducting key informant interviews, focus group discussions, and observation for 1–2 weeks, focused on migration and development dynamics (Erdal et al. 2022).

Results

In this section, we discuss the main regression results, supplemented with qualitative data. For each dimension of independent variables, we analyse the effects on internal and international migration considerations. We discuss the statistical significance and magnitude of the coefficients of each variable, across the ten local areas and at the aggregate pooled data level.

Perceptions of safety and insecurity

We first look at the respondents’ general perceptions of the safety in the local area at the pooled data level and across all local areas (Tables 2 and 3), in terms of the variable Thinks it’s not safe to walk the streets at night. There is only one statistically significant association between this general perception of safety and international migration considerations for Down Quarters, Nigeria. There is a somewhat clearer trend in terms of perceptions of changes in safety, with this subjective assessment of Whether the local area has become more dangerous compared to five years ago affecting international migration considerations in Behsud, Afghanistan, Baidoa, Somalia and at the aggregate level. With respect to internal migration, we observe a statistically significant association with this subjective assessment of perceptions of changes in safety in Down Quarters, Nigeria and Erigavo, Somalia. (Table 3).

Table 2

Association between perceptions of safety/insecurity and international migration aspirations.

	AFG1: Shahrake Jabrael	AFG2: Behsud	AFG3: Shahrake Mahdia	ETH2: Batu	ETH3: Moyale	NGA1: Down Quarters	NGA2: Awe	NGA3: Ekpoma	SOM1: Erigavo	SOM2: Baidoa	Pooled data
Panel A
Thinks it is NOT safe to walk streets at night in research area	0.0621	−0.0401	0.0153	−0.0254	0.00807	0.164*	−0.0323	−0.0152	−0.0115	−0.0311	0.00381
	(0.0467)	(0.0457)	(0.0541)	(0.0485)	(0.0368)	(0.0854)	(0.0297)	(0.0590)	(0.0438)	(0.0480)	(0.0165)
Obs.	515	534	531	520	520	479	468	473	462	522	5020
Adjusted R2	0.165	0.274	0.202	0.178	0.178	0.278	0.286	0.245	0.192	0.166	0.211
Panel B
Thinks RA has become more dangerous, compared to 5 years ago	0.0703	0.100**	−0.00805	0.0324	0.0286	0.0610	0.00820	−0.0522	0.0644	0.292*	0.0429**
	(0.0613)	(0.0426)	(0.0562)	(0.0444)	(0.0574)	(0.0470)	(0.0292)	(0.0517)	(0.0670)	(0.152)	(0.0180)
Obs.	513	530	527	518	520	481	460	460	461	522	4988
Adjusted R2	0.171	0.278	0.201	0.180	0.180	0.271	0.271	0.244	0.194	0.185	0.212

Note: Cluster robust standard errors in parentheses. Using MIGNEX Survey (mxs-prep-2023–01-11.dta). N = 13,172. Data are weighted to reflect the survey design.

*

p < 0.1.

**

p < 0.05.

Table 3

Association between perceptions of safety/insecurity and internal migration aspirations.

	AFG1: Shahrake Jabrael	AFG2: Behsud	AFG3: Shahrake Mahdia	ETH2: Batu	ETH3: Moyale	NGA1: Down Quarters	NGA2: Awe	NGA3: Ekpoma	SOM1: Erigavo	SOM2: Baidoa	Pooled data
Panel A
Thinks it is NOT safe to walk streets at night in research area	−0.0286	0.0385	0.0260	0.0298	−0.0231	−0.0427	−0.0189	0.109	0.0243	−0.0235	0.00787
	(0.0500)	(0.0532)	(0.0528)	(0.0445)	(0.0491)	(0.0855)	(0.0392)	(0.0753)	(0.0398)	(0.0547)	(0.0189)
Obs.	507	509	527	519	519	481	464	469	435	475	4,901
Adjusted R2	0.157	0.170	0.077	0.163	0.174	0.241	0.180	0.274	0.175	0.186	0.216
Panel B
Thinks RA has become more dangerous, compared to 5 years ago	−0.00531	−0.000347	0.0508	−0.0846	0.0292	0.0886*	0.00121	0.0226	0.133*	0.305	0.0246
	(0.0565)	(0.0482)	(0.0711)	(0.0546)	(0.0435)	(0.0493)	(0.0382)	(0.0588)	(0.0775)	(0.184)	(0.0190)
Obs.	506	508	524	517	519	482	458	455	434	475	4,874
Adjusted R2	0.158	0.171	0.077	0.178	0.176	0.244	0.180	0.282	0.183	0.203	0.217

Note: Cluster robust standard errors in parentheses. Using MIGNEX Survey (mxs-prep-2023–01-11.dta). N = 13,172. Data are weighted to reflect the survey design.

*

p < 0.1.

As such, general perceptions of safety and insecurity do not seem to play a large role in affecting migration considerations, however, the findings do show that in some instances, more negative perceptions result in a greater likelihood of considering international migration (and to a lesser degree, consideration of internal migration).

Concerns and fears of violence and conflict

We also analyse the effects of concerns and fears of violence and conflict (Tables 4 and 5). They appear to have an effect on both internal and international migration considerations, with different trends across local areas. For instance, we see that in Erigavo, Somalia, Fear of violence by armed groups/militia and being Concerned about conflict and violence in the next five years have a statistically significant, positive, association with internal migration considerations. This finding should perhaps be placed in the context of this local area’s history of out-migration during Somalia’s protracted civil war and, more recently, where half the respondents to our survey said they knew someone who had migrated internally from Erigavo in the past five years (Ahmed et al. 2022). Thus, when it comes to possible responses to fear of violence, it seems plausible that at least considering internal migration as an option might be expected. Moreover, the magnitude of the effect for Erigavo is higher than for other local areas, for example those who have Feared violence by armed groups/militia in the past five years are 24 per cent more likely to have seriously considered internal migration in the past year, than those who have not feared this type of violence.

Table 4

Association between concerns/fears of violence and international migration aspirations.

	AFG1: Shahrake Jabrael	AFG2: Behsud	AFG3: Shahrake Mahdia	ETH2: Batu	ETH3: Moyale	NGA1: Down Quarters	NGA2: Awe	NGA3: Ekpoma	SOM1: Erigavo	SOM2: Baidoa	Pooled data
Panel A
Concerned about conflict and violence (next 5 years)	0.179**	0.0400	−0.0247	−0.0243	−0.0134	0.0529	−0.0260	−0.0621	0.00922	0.0633	0.0236
	(0.0785)	(0.0750)	(0.0588)	(0.0511)	(0.0526)	(0.0482)	(0.0260)	(0.0530)	(0.0493)	(0.0565)	(0.0190)
Obs.	506	516	527	518	518	467	470	473	461	520	4,972
R2	0.185	0.290	0.201	0.177	0.175	0.269	0.270	0.250	0.189	0.162	0.214
Panel B
Has feared violence at political rally/protest (past 5 years)	0.112**	0.0571	0.106***	0.00210	−0.00936	0.119**	0.0109	−0.00106	0.0270	−0.0171	0.0332*
	(0.0532)	(0.0523)	(0.0374)	(0.0528)	(0.0530)	(0.0520)	(0.0414)	(0.0389)	(0.107)	(0.0365)	(0.0170)
Obs.	515	533	531	526	521	483	471	477	462	522	5,037
R2	0.179	0.275	0.213	0.175	0.178	0.278	0.282	0.245	0.190	0.165	0.211
Panel C
Has feared violence by armed forces/militias (past 5 years)	0.195***	0.0244	0.110***	0.0711	0.132**	0.155***	−0.0243	−0.0686	−0.0355	0.00690	0.0420**
	(0.0689)	(0.0386)	(0.0357)	(0.0563)	(0.0589)	(0.0496)	(0.0414)	(0.0485)	(0.0751)	(0.0354)	(0.0168)
Obs.	516	532	531	519	521	483	466	478	463	522	5,028
R2	0.192	0.273	0.215	0.180	0.192	0.287	0.283	0.248	0.191	0.165	0.212
Panel D
Has feared other violence in research area (past 5 years)	0.0918	0.0601	0.0609	0.0810	0.00713	0.121***	−0.0323	−0.0842	0.115	0.0475	0.0194
	(0.0554)	(0.0397)	(0.0516)	(0.0567)	(0.0453)	(0.0431)	(0.0350)	(0.0609)	(0.127)	(0.0689)	(0.0168)
Obs.	515	534	528	526	521	481	468	476	461	522	5,028
R2	0.174	0.275	0.206	0.180	0.178	0.283	0.282	0.251	0.194	0.167	0.211

Note: Cluster robust standard errors in parentheses. Using MIGNEX Survey (mxs-prep-2023–01-11.dta). N = 13,172. Data are weighted to reflect the survey design.

*

p < 0.1.

**

p < 0.05.

***

p < 0.01.

Table 5

Association between concerns/fears of violence and internal migration aspirations.

	AFG1: Shahrake Jabrael	AFG2: Behsud	AFG3: Shahrake Mahdia	ETH2: Batu	ETH3: Moyale	NGA1: Down Quarters	NGA2: Awe	NGA3: Ekpoma	SOM1: Erigavo	SOM2: Baidoa	Pooled data
Panel A
Concerned about conflict and violence (next 5 years)	0.0934	0.0988**	−0.0244	0.113**	−0.0340	0.114**	0.0971*	0.103	0.106**	0.00447	0.0734***
	(0.0722)	(0.0364)	(0.0865)	(0.0424)	(0.0537)	(0.0544)	(0.0474)	(0.0617)	(0.0491)	(0.0576)	(0.0186)
Obs.	498	494	523	518	517	468	468	468	433	473	4,856
R2	0.164	0.181	0.078	0.168	0.178	0.243	0.186	0.272	0.190	0.175	0.221
Panel B
Has feared violence at political rally/protest (past 5 years)	0.100**	−0.0454	0.101**	0.0539	0.0496	0.101*	0.0381	0.0723	0.179	−0.0178	0.0521***
	(0.0397)	(0.0441)	(0.0376)	(0.0498)	(0.0602)	(0.0553)	(0.0555)	(0.0640)	(0.122)	(0.0803)	(0.0183)
Obs.	507	508	527	526	520	484	469	472	434	475	4,918
R2	0.165	0.173	0.089	0.161	0.176	0.243	0.177	0.268	0.185	0.186	0.217
Panel C
Has feared violence by armed forces/militias (past 5 years)	0.144***	0.0826	0.0646	−0.0557	0.0867	0.0864	−0.0324	0.0905	0.243**	−0.100*	0.0505***
	(0.0425)	(0.0585)	(0.0463)	(0.0683)	(0.0571)	(0.0572)	(0.0597)	(0.0547)	(0.0927)	(0.0545)	(0.0180)
Obs.	508	508	527	519	520	484	464	473	435	475	4,910
R2	0.170	0.177	0.082	0.164	0.179	0.241	0.176	0.270	0.202	0.196	0.217
Panel D
Has feared other violence in research area (past 5 years)	0.0686	0.0481	0.0888**	0.133	0.0441	0.0348	0.0785	0.108	0.200	−0.0177	0.0586***
	(0.0487)	(0.0432)	(0.0397)	(0.0827)	(0.0547)	(0.0498)	(0.0686)	(0.0644)	(0.118)	(0.0683)	(0.0187)
Obs.	507	509	524	526	520	482	466	471	434	475	4,910
R2	0.159	0.168	0.079	0.170	0.175	0.238	0.183	0.274	0.186	0.186	0.218

Note: Cluster robust standard errors in parentheses. Using MIGNEX Survey (mxs-prep-2023–01-11.dta). N = 13,172. Data are weighted to reflect the survey design.

*

p < 0.1.

**

p < 0.05.

***

p < 0.01.

Table 6

Association between experience of conflict/violence and international migration aspirations.

	AFG1: Shahrake Jabrael	AFG2: Behsud	AFG3: Shahrake Mahdia	ETH2: Batu	ETH3: Moyale	NGA1: Down Quarters	NGA2: Awe	NGA3: Ekpoma	SOM1: Erigavo	SOM2: Baidoa	Pooled data
Panel A
Has experienced violence (past 5 years)	0.00525	−0.0349	0.00429	0.0268	0.0147	0.0157	0.0201	0.0653	0.336*	−0.0182	0.0155
	(0.0566)	(0.0689)	(0.0587)	(0.100)	(0.0586)	(0.0782)	(0.0530)	(0.0826)	(0.183)	(0.0594)	(0.0239)
Obs.	424	363	355	298	316	245	219	197	89	262	2,765
Adjusted R2	0.200	0.326	0.207	0.277	0.284	0.412	0.448	0.380	0.562	0.219	0.268
Panel B
Respondent or HH member has experienced theft (past 5 years)	0.0763	0.0387	0.0747**	0.0550	0.0807	0.0589	0.0744	0.0741	0.0514	−0.0351	0.0468***
	(0.0484)	(0.0396)	(0.0323)	(0.0529)	(0.0526)	(0.0435)	(0.0581)	(0.0673)	(0.0700)	(0.0406)	(0.0170)
Obs.	516	536	531	524	521	481	483	478	461	522	5,049
Adjusted R2	0.173	0.275	0.208	0.184	0.183	0.270	0.281	0.248	0.194	0.166	0.213
Panel C
Respondent or HH member has experienced physical violence (past 5 years)	0.0528	0.0336	0.0410	0.128*	0.0850	−0.0143	0.0906	0.0530	−0.0199	0.0661	0.0349
	(0.0837)	(0.0475)	(0.0518)	(0.0739)	(0.0660)	(0.0631)	(0.0540)	(0.0654)	(0.0938)	(0.0911)	(0.0219)
Obs.	516	532	531	527	521	483	475	478	460	522	5,042
Adjusted R2	0.169	0.272	0.203	0.182	0.182	0.268	0.287	0.246	0.193	0.167	0.212

Note: Cluster robust standard errors in parentheses. Using MIGNEX Survey (mxs-prep-2023–01-11.dta). N = 13,172. Data are weighted to reflect the survey design.

*

p < 0.1.

**

p < 0.05.

***

p < 0.01.

Table 7

Association between experience of conflict/violence and internal migration aspirations.

	AFG1: Shahrake Jabrael	AFG2: Behsud	AFG3: Shahrake Mahdia	ETH2: Batu	ETH3: Moyale	NGA1: Down Quarters	NGA2: Awe	NGA3: Ekpoma	SOM1: Erigavo	SOM2: Baidoa	Pooled data
Panel A
Has experienced violence (past 5 years)	0.0704	0.149**	−0.0644	0.220***	0.0990	−0.0525	0.201*	0.261***	0.234	−0.00136	0.0828***
	(0.0656)	(0.0605)	(0.0713)	(0.0780)	(0.0944)	(0.112)	(0.0995)	(0.0680)	(0.184)	(0.102)	(0.0303)
Obs.	419	351	354	298	316	244	217	196	86	222	2,700
Adjusted R2	0.180	0.229	0.124	0.224	0.256	0.310	0.403	0.497	0.590	0.434	0.263
Panel B
Respondent or HH member has experienced theft (past 5 years)	0.0458	0.0632	0.0881*	−0.0278	−0.0335	0.0920**	0.248***	0.0994*	0.109*	−0.0142	0.0716***
	(0.0547)	(0.0401)	(0.0490)	(0.0715)	(0.0631)	(0.0440)	(0.0678)	(0.0578)	(0.0610)	(0.0885)	(0.0189)
Obs.	508	511	527	523	520	482	481	473	434	475	4,930
Adjusted R2	0.157	0.174	0.087	0.159	0.174	0.244	0.219	0.271	0.189	0.186	0.219
Panel C
Respondent or HH member has experienced physical violence (past 5 years)	0.0399	0.102**	0.0799	0.0496	0.00105	0.0652	0.223***	0.160***	0.210**	0.0880	0.101***
	(0.0556)	(0.0470)	(0.0534)	(0.0707)	(0.0623)	(0.0654)	(0.0764)	(0.0502)	(0.101)	(0.0992)	(0.0209)
Obs.	508	507	527	526	520	484	473	473	432	475	4,922
Adjusted R2	0.155	0.181	0.082	0.160	0.174	0.238	0.206	0.278	0.196	0.188	0.220

Note: Cluster robust standard errors in parentheses. Using MIGNEX Survey (mxs-prep-2023–01-11.dta). N = 13,172. Data are weighted to reflect the survey design.

*

p < 0.1.

**

p < 0.05.

***

p < 0.01.

For Baidoa, on the other hand, the sign of the statistical coefficient for Fear of violence by armed groups/militia is unexpectedly negative, meaning those with these fears are less likely to have seriously considered internal migration. As mentioned earlier, this can be understood in relation to the context of Baidoa being an island of relative peace and security, within a larger region where territories outside the city are held by Al-Shabbab. The local area is also shaped by significant proportions of internally displaced people who have sought safety in Baidoa in the past decade (Kasavan et al. 2022b). Given the context, a logical response to fear of violence would be to plan to stay and thus not even consider leaving Baidoa.

For Shahrake Jabrael in Afghanistan, Fear of violence at a political rally/protest and Fear of violence by armed groups/militia is positively associated with both internal and international migration considerations. For Shahrake Mahdia, both of these variables are also statistically significant for international migration considerations, and Fear of violence at political rally/protest and Fear of other violence are statistically significant for internal migration considerations. On the whole, the magnitude of the effect for the Afghani local areas is somewhat stronger for international migration considerations.

For instance, taking the example of Shahrake Jabrael, those who have Feared violence by armed groups/militia in the past five years are 20 per cent more likely to have seriously considered international migration in the past year, than those who have not feared this type of violence, and 14 per cent more likely to have considered internal migration. Given the timing of data collection in June–July 2021, where the violence feared by Afghans is likely to have been associated with both the Taliban and the Afghan army, that is, current and imminent state actors, and given the country’s history of protracted conflict and displacement not just internally (see, for instance, Alizada and Murray 2022), but also across its borders, it is perhaps not surprising that considerations of leaving the country were more prominent.

For Behsud, the final Afghani area, only Being concerned about conflict and violence in the next five years has a statistical association with internal migration considerations. This could be related to the demographic composition of the sample in this local area and their experience of conflict: a fifth of the young people sampled in the representative survey were Pakistan-born returnees and a quarter were IDPs who have moved to Behsud (Akakhil et al. 2022). Having potentially been more ‘sensitized’ to conflict and violence, this group may respond differently to fears of violence.

For most of the Ethiopian and Nigerian local areas, concerns and fears of violence and conflict appear to be less relevant in shaping migration considerations. Apart from Down Quarters, the other two local areas only have one statistically significant finding, with no clear patterns. For Down Quarters, Nigeria, being Concerned about conflict and violence in the next five years and Has feared violence at political rally/protest are associated with internal migration considerations. For international migration considerations, three measures of fear of violence are statistically significant, except Concerned about conflict and violence in the next five years. This local area is in Kaduna state and city, both of which have seen high levels of ethno-religious fighting, including frequent acts of violence committed by the Boko Haram group. It is a destination for internal migrants, many of whom have fled violence in other parts of the state and country, with one-third of the sample having been born elsewhere in Nigeria (Adamu et al. 2022). Thus, with a sizeable share of respondents having been internally displaced by violence already, a consideration of international migration in the face of fearing further violence makes sense.

At the aggregate level, all four measures of concerns and fears of violence and conflict exhibit a strong statistically significant association with internal migration considerations. More specifically, being concerned or fearing some type of violence is associated with a 5–7 per cent higher likelihood of considering migrating internally. In the case of international migration considerations, only Has feared violence at political rally/protest in the past 5 years and Has feared violence by armed forces/militias in the past 5 years show a statistical association. Results are stronger in statistical significance and magnitude of coefficient for internal migration considerations, signalling that fearing and being concerned about conflict/violence is a strong determinant of internal migration considerations.

Experiences of violence and conflict

Finally, we look at the association between actual experiences of violence and conflict and migration considerations (Tables 6 and 7). Here we see some strong differences between internal and international migration considerations: on the whole, actual experiences of violence and conflict are much more likely to result in internal migration considerations than international migration considerations. This is not surprising, actually experiencing violence likely increases the urgency of potentially leaving the local area, and internal migration tends to be a more viable and affordable option.

For Behsud, Afghanistan, the Respondent having experienced violence in the past five years or the Respondent or household member having experienced physical violence in the past five years is associated with internal migration considerations. Shahrake Madia, Afghanistan, had seen a rise in criminality (including theft) in the months leading up to the survey (Majidi et al. 2022) and, indeed, the variable Respondent or a household member has experienced theft (past 5 years) is associated with both internal and international migration considerations.

For Erigavo, Somalia, the Respondent or a household member has experienced theft in the past five years or physical violence is also associated with a greater likelihood of internal migration considerations, whereas Has experienced violence in the past 5 years is positively associated with international migration considerations. For the Ethiopian local areas, there is only one statistically significant finding for each measure of migration considerations. In Batu, whether the Respondent has experienced violence in the past five years is positively linked with greater internal migration considerations, while the Respondent or household member has experienced physical violence in the past five years is positively associated with international migration considerations.

However, the clearest trend is seen in the local areas of Nigeria, where there is a positive association between all three measures of conflict/violence experience and internal migration considerations for both Awe and Ekpoma. For Down Quarters, only whether the Respondent or a household member has experienced theft (past 5 years) is positively and statistically associated with internal migration considerations. For Nigerians, internal migration is a common strategy, including as a response to the ethnic tensions that the north of the country (Down Quarters and Awe) has. It is perhaps surprising that Down Quarters, which is officially classified as the least safe of the three areas,⁵

5

The UK Foreign Office classifies Kaduna city and state as red (“Advise against all travel”), the other two local areas are classified green (“See our travel advice before travelling”).

only has one statistically significant finding. One possible explanation is that young people there are more ‘sensitized’ to experiences of violence.

On the whole, the pooled data results confirm local area-level findings showing that actual experience of conflict and violence is strongly associated with particularly with internal migration considerations, where the three measures of experience of conflict and violence are highly statistically significant. Having experienced some type of violence, from theft to physical violence, is associated with a 7–10 per cent higher likelihood of having internal migration considerations. Having international migration considerations is also statistically linked with Respondent or Household Member has experienced theft, where this association is lower at 5 per cent.

Control variables

We also explore whether other characteristics including Gender, Household economic situation and Perceived inequality, Employment status, Risk attitudes and Migration networks drive migration considerations; see Supplementary Tables A.1 and A.2 for the pooled data results.⁶

6

The full regression tables at the local area level are available from the authors upon request.

We find that Gender is an important determinant in Afghani and Nigerian local areas, where young women are less likely to consider migrating both internally and internationally compared to young men. In the Ethiopian local areas, Gender is only a determinant of international migration considerations. These findings show that migration considerations are gendered in most local areas.

Other factors like Household economic situation, Perceived inequality and Risk attitudes do not show as important determinants of either international or internal migration considerations, they are statistically significant in very few instances.

Having Migrant networks abroad or internally is the one factor that is consistently a significant driver of both international and internal migration considerations across most local areas in all four countries. This reinforces a large body of literature which shows the importance of migrant networks for migration aspirations and out-migration (see, for example, Manchin and Orazbayev 2018).

Lastly, having had a previous migration experience, specifically Having lived in a high-income country for at least one year is another factor that is associated with higher internal and international migration considerations for many of the regressions. This relationship is mostly positive, but for some local areas and measures of conflict/violence this link is negative or not statistically significant. In turn, the degree to which previous migration experiences influence future migration considerations is largely area specific. However, this measure allows us to reduce to a certain degree the immobility bias of our sample. By controlling for previous migration experiences, we can account for the heterogeneity of our sample in terms of their mobility choices, reflecting that deciding to stay also requires agency (Schewel 2020).

Discussion

Having looked at the detailed regression findings and specific local patterns, we now discuss five key patterns that have emerged from the analysis.

First, the role of violence and conflict appears to play a more important role in shaping migration considerations for the Afghan areas, where we find the highest number of statistically significant findings. This is not surprising, given that the country has seen decades of civil war and high levels of violence, which escalated in the run-up to the survey, compared to the other areas in our analysis which have generally seen more localized, time-bound, or targeted forms of violence.⁷

7

While Somalia has also experienced decades of civil war, the local areas covered in our survey have had comparatively less conflict in recent years.

While our data does not give a measure of intensity of fighting or levels of conflict exposure, we can safely assume that young adults in the Afghan areas have generally been exposed to greater levels of conflict.⁸

8

For instance, more than two thirds of the young adults in the Afghan local areas have feared violence at political protest/rally, by army or militias or other types of violence, see Figure 1.

Such patterns are in line with existing studies, which show that the more threats or violence experienced generally result in more migration (Adhikari 2013; Davenport et al. 2003; Moore and Shellman 2004).

Second, even in local areas where the population feels unsafe and reports high levels of experiences of conflict/violence, these perceptions, concerns, fears, and experiences do not always—and in fact mostly—do not result in migration considerations, bearing in mind that our sample is biased towards those who have currently not left, though we do control for past mobility. Recent studies have drawn attention to the ‘mobility bias’ in migration research (Schewel 2020). In fact, most people either do not aspire to migrate at all or are “involuntarily immobile”, that is, they do not have the capacity to migrate even if they might want to (Braithwaite et al. 2021; Carling and Schewel 2018).

Third, on the whole, we see a stronger association between conflict and internal migration considerations compared to international migration ones. Our pooled data analysis also provides evidence in support of stronger effects for internal migration considerations than for international migration considerations, where seven of the nine independent variables are statistically significant for internal, but only four for international migration. Moving internally is often a more viable option when potential pressures to leave become more acute. This can be explained in terms of internal migration entailing shorter travel distances, which cost less and (might) involve less risk, but also that the availability of social networks, the chance of speaking the language, and thus also more easily being able to find means of surviving, can all be assumed to be higher with internal migration (Bernard et al. 2022; Cirillo et al. 2022).

Fourth, related to the third finding, there is somewhat an exception for the Afghanistan areas. While there is no clear pattern with regard to the number of statistically significant coefficients for internal vs. international migration considerations, the magnitude of the coefficients tends to be larger for international migration considerations in the face of perceptions of insecurity and concerns and fears of violence. The fact that violence and conflict are more likely to result in stronger consideration of international migration in the Afghan local areas is not surprising, given that the insecurity and violence experienced in the summer of 2021—at the time of data collection—were directly related to a civil war between the Taliban and Afghan army. This chimes with existing studies finding that cross-border migration is more likely if the main perpetrator of violence is a state actor (Steele 2019) or a state-sponsored politicide (Moore and Shellman 2006). With a state takeover by the Taliban imminent at the time of data collection, this appears to apply.

The final pattern emerging from the analysis is that there are stronger effects for specific fears and experiences of conflict and violence than general perceptions of insecurity and violence. This also holds for the pooled analysis, where the perceptions of insecurity variables are mostly not statistically significant. This makes intuitive sense; feeling unsafe walking the streets at night on its own is probably not a strong enough reason to consider leaving the place you live in, whereas fearing violence committed by specific groups or having experienced it is a more compelling reason to.

Digging deeper into this, we see an interesting pattern for considerations international migration: For the pooled data and specific local areas, fears of conflict and violence seem to matter more than having experienced them. One explanation could be that those fears of violence are based on observation of these acts of violence actually being committed to others—spillover effects of some form. This could be an explanation for the patterns found in the Afghani local areas, where fears are more important determinants than actual experiences when considering internal/international migration. After decades of conflict, indirect exposure to conflict and having witnessed violence is presumably common across the general population. As such, fears of different forms of violence may be experienced and felt differently than in other areas where conflict is more localized, targeted, or time-bound.

This also resonates with the work of Schon (2016) who finds that the dramatized violence perpetrated at checkpoints results in uncertainty and fear and affects migration patterns in unexpected ways. Müller-Funk (under review), too, finds that both personal threats to life—similar to our experiences of conflict and violence category—and fears of generalized forms of violence both affect displacement trajectories. This also underscores the individual nature of migration aspirations and subjective factors, such as emotions and personality, that can affect migration decisions alongside more tangible factors (see also Hagen-Zanker et al. 2023b).

Conclusion

This article has contributed to a more nuanced understanding of the relationship between conflict and mobility by disaggregating ‘conflict’ and looking at the specific roles of perceptions, fears, and experiences of violence and conflict in considerations of internal or international migration across a varied range of local areas at the sub-national level. In doing so, we have shown that migration considerations can, in fact, be quantified and measured in very specific ways (see also Carling et al. 2023 on other dimensions of migration aspirations). This allows us to draw out some nuanced insights into the mental processes of people living in conflict contexts.

We have shown that general perceptions of insecurity are less influential in shaping considerations to leave. Instead, it is specific fears and actual experiences of migration that affect considerations of migration. Reinforcing other recent studies analysing migration decision-making in conflict settings (Ghosn 2021; Müller-Funk 2023; Schon 2019), we find that, broadly speaking, specific fears of violence have a stronger effect on international migration considerations, whereas those who have experienced violence are generally more likely to consider internal migration.

Moreover, except for the local areas where the (de facto) state is the main perpetrator of violence, we see much stronger effects of perceptions, fears, and experiences of violence and conflict of insecurity on internal migration considerations, compared to international migration ones. In other words, people affected by conflict mostly consider staying within their country, rather than making a more expensive—and often more dangerous—journey to another country. This is an important finding for humanitarian and migration policy discourses and depictions in the media, which often seem to make the assumption that everyone affected by conflict wants to leave the country.

Meanwhile, while others have noted the prevalence of internal displacement in conflict settings, existing data has severe limitations (Braithwaite et al. 2021; Cantor and Apollo 2020; Turkoglu 2022). Our sub-national dataset of more than 5000 young adults across ten contexts varyingly affected by conflict has thus allowed us to make an important contribution to the conflict and forced displacement/migration literature by showing the differentiated effects of violence on both internal and international mobility considerations.

Strikingly, however, perceptions, fears, and experiences of violence and conflict mostly do not lead to considerations of migration, exposing once again the mobility bias in the forced displacement/migration and conflict fields and broader political discourses. Our unique dataset of young adults who have currently not left their local areas (we do control for past mobility) shows that mobility is only one possible response to conflict and violence. Their immobility can, on the one hand, be understood in the context of broader life aspirations—which may or may not include migration—and, on the other hand, be a result of the lack of ability to migrate: those wanting to leave may not have the financial means, social networks, etc., to actually leave (see also Carling and Schewel 2018; Schon 2019). This question of immobility in conflict-affected places is one that warrants further research.

Based on the above analysis, our contribution adds to the growing body of work moving beyond a binary approach to responses to being conflict-affected. Like Braithwaite et al. (2021), we argue for the continued need for further analysis of internal mobility—both actual and considerations thereof—and taking into account both conflict-related and non-conflict-related factors (see also Schon 2019). Where access to data allows, there are important benefits to being able to analyse (considerations about) movements within and across national borders together, adding to our understanding of how different types of conflict factors, varyingly shape different types of (im)mobility.

Acknowledgements

Many thanks to participants at the 2022 American Association of Geographers conference and the 2022 IMISCOE conference who provided helpful comments on an early draft of this article. The authors are also grateful to two anonymous reviewers who provided helpful feedback. All remaining errors remain our own.

Funding

This publication draws upon research within the project Aligning Migration Management and the Migration-Development Nexus (MIGNEX) which has received funding from the European Union’s Horizon 2020 research and innovation program under grant agreement No 770453.

Supplementary material

Supplementary material is available at Journal of Refugee Studies online.

Table A.1

Descriptive statistics for dependent and independent variables (mean value and standard deviation).

Variable	Nigeria			Ethiopia		Somalia		Afghanistan
	Down Quarters	Awe	Ekpoma	Batu	Moyale	Erigavo	Baidoa	Shahrake Jabrael	Behsud	Shahrake Mahdia
Has seriously considered international migration (past year)	0.35	0.1	0.51	0.25	0.14	0.21	0.11	0.31	0.3	0.27
	(0.48)	(0.3)	(0.50)	(0.44)	(0.35)	(0.41)	(0.31)	(0.46)	(0.46)	(0.44)
Has seriously considered internal migration (past year)	0.47	0.21	0.64	0.36	0.17	0.19	0.15	0.25	0.19	0.2
	(0.5)	(0.41)	(0.48)	(0.48)	(0.38)	(0.39)	(0.36)	(0.43)	(0.39)	(0.4)
Concerned about conflict and violence (next 5 years)	0.69	0.58	0.56	0.62	0.75	0.21	0.25	0.89	0.77	0.96
	(0.46)	(0.49)	(0.5)	(0.49)	(0.43)	(0.41)	(0.43)	(0.31)	(0.42)	(0.2)
Has feared violence at political rally/protest (past 5 years)	0.34	0.25	0.28	0.52	0.34	0.1	0.23	0.5	0.35	0.49
	(0.47)	(0.43)	(0.45)	(0.5)	(0.47)	(0.3)	(0.42)	(0.5)	(0.48)	(0.5)
Has feared violence by armed forces/militias (past 5 years)	0.33	0.23	0.25	0.23	0.34	0.16	0.25	0.74	0.47	0.52
	(0.47)	(0.42)	(0.44)	(0.42)	(0.47)	(0.36)	(0.43)	(0.44)	(0.5)	(0.5)
Has feared other violence in local area (past 5 years)	0.46	0.32	0.24	0.34	0.57	0.07	0.14	0.57	0.48	0.51
	(0.5)	(0.47)	(0.43)	(0.47)	(0.5)	(0.26)	(0.35)	(0.49)	(0.5)	(0.5)
Thinks it is NOT safe to walk streets at night in local area	0.83	0.53	0.68	0.57	0.65	0.33	0.22	0.76	0.61	0.9
	(0.38)	(0.5)	(0.47)	(0.5)	(0.48)	(0.47)	(0.41)	(0.43)	(0.49)	(0.31)
Thinks RA has become more dangerous, compared to five years ago	0.51	0.36	0.34	0.35	0.4	0.14	0.03	0.84	0.56	0.91
	(0.5)	(0.48)	(0.47)	(0.48)	(0.49)	(0.34)	(0.18)	(0.36)	(0.5)	(0.29)
Has experienced violence (past 5 years)	0.32	0.32	0.37	0.13	0.56	0.28	0.21	0.33	0.35	0.36
	(0.47)	(0.47)	(0.48)	(0.33)	(0.5)	(0.45)	(0.41)	(0.47)	(0.48)	(0.48)
Respondent or HH member has experienced theft (past 5 years)	0.36	0.22	0.27	0.25	0.15	0.19	0.11	0.38	0.39	0.44
	(0.48)	(0.42)	(0.44)	(0.43)	(0.36)	(0.39)	(0.31)	(0.48)	(0.49)	(0.5)
Respondent or HH member has experienced physical violence (past 5 years)	0.19	0.14	0.18	0.1	0.11	0.1	0.06	0.15	0.26	0.19
	(0.39)	(0.34)	(0.38)	(0.3)	(0.31)	(0.3)	(0.23)	(0.36)	(0.44)	(0.39)

Note: standard deviation in parenthesis.