South-South refugee movements: do pull factors play a role?

Introduction

At least since the Syrian refugee crisis, the topic of forced migration is high on the international political agenda. This is not fully reflected in the academic migration literature. In particular when it comes to analysing the drivers of migration, a majority of the existing studies investigates voluntary migrant flows (e.g. Dao et al. 2018; Mayda 2010; Lanati and Thiele 2018; Ortega and Peri 2013). Underlying these empirical studies is typically a model of international migration in which individuals base their decision on whether and where to migrate on a comparison of the utility associated with each possible destination (e.g. Beine and Parsons 2015). Individuals are assumed to maximize their utility across a pool of destinations, including the home country. Actual migration decisions are then determined by pull factors that make particular destinations more attractive as well as push factors that drive the decision to leave, loosely following the push-pull model of migration developed by Lee (1966).

Pull factors are largely absent from much of the existing literature on forced migration, which has centred on one specific set of push factors, namely the various forms of violent conflict that occur in countries of origin (Brück et al. 2018). There is indeed ample evidence showing that migration by refugees and asylum seekers depends strongly on political factors such as conflict and civil war in their home countries (e.g. Schmeidl 1997; Davenport et al. 2003; Moore and Shellman 2007; Hatton 2009). While most of the forced migrants flee to neighbouring countries (see Figure 1), this often does not restrict their choice to one single destination. Syrians, for example, fled in large numbers to three different countries with which Syria shares a border: Lebanon, Jordan, and Turkey. In addition, a non-negligible share (around 20%) of refugees is moving to non-neighbouring countries. Little is known about the factors driving emigration decisions when refugees have multiple re-location choices.

A small number of papers (Dreher et al. 2019; Murat 2020; Neumayer 2005) addresses the question of whether prospective countries of asylum can at least partly offset the push factors prevailing at origin and dampen the outflow of refugees from developing countries by providing foreign aid. From this literature, no clear picture emerges. According to Dreher et al. (2019), the impact of foreign aid on refugee flows worldwide and to OECD countries in particular is positive in the short run but turns negative after some years, which the authors attribute to lagged positive effects of aid on economic growth. Murat (2020) finds that asylum applications from poor countries are negatively and significantly associated with bilateral aid disbursements in the short run, with mixed evidence of more lasting effects, while applications from medium-income developing economies are positively but weakly related to aid. In both income groups, demand for asylum decreases when donors provide more humanitarian assistance. Neumayer (2005) detects no significant association between the aid disbursed to developing countries and asylum applications in Western Europe during the period 1982–1999.

Several other studies do consider the impact of pull factors on forced migration but focus on OECD destinations where refugees seek asylum (e.g. Brück et al. 2018; Hatton 2016; Kang 2021; Neumayer 2005). Brück et al. (2018) distinguish a “localist” and a “rationalist“ model of refugee migration. The localist model suggests that violence is the predominant push factor and the safety and proximity of the destination the predominant pull factor. Under the rationalist model, additional factors, such as economic opportunities in the destination country, might enter the individual’s migration decision, bringing it closer to the cost-benefit considerations of voluntary migrants predicted by the standard push-pull model. Brück et al. (2018) point to network effects and the employment rate in the destination country as key pull factors that reach beyond the localist model. Hatton (2016), Kang (2021) and Neumayer (2005) also report pull factors including incomes and diasporas at destination to be significant determinants of asylum seeking in OECD and EU countries, respectively.

In a somewhat different within-country setting, Beine et al. (2021) analyse the determinants of the internal mobility of refugees in Turkey using mobile phone data. They find, among other things, that refugees tend to leave relatively poorer locations and are attracted by relatively wealthier ones. To the best of our knowledge, there is so far no study investigating the drivers of south-south refugee movements, even though a large majority of forced migrants ends up in other developing countries.

In this paper, we take the fairly well researched and broadly agreed upon factors that force people to emigrate as given and instead perform a detailed empirical analysis of their destination choices. In contrast to previous studies, our focus is on movements to non-OECD countries, which account for the vast majority of world-wide refugee inflows. This does not only allow us to fill a remaining research gap, but also renders it possible to address the important policy question of whether through their aid allocation decisions OECD countries can have an impact on where refugees actually go. We employ a gravity model of international migration, which by the nature of its dyadic structure is particularly suited to capture key potential drivers of bilateral refugee flows. The drivers we consider include standard dyadic variables – for instance the physical distance between origin and destination country or the existence of migrant networks – as well as pull factors at destination – for instance low unemployment rates or the provision of foreign assistance just mentioned. Inclusion of the latter is important because incoming refugees are in particular need of public services such as health care, the supply of which in turn strongly depends on foreign aid projects in most of the developing country destinations (e.g. Verme et al. 2016).

Our findings suggest that refugees tend to move to safe neighbouring countries but also positively respond to local pull factors such as relatively high per-capita income levels and the availability of education and health services when choosing their country of destination. Donors have the ability to affect the direction of south-south refugee movements by investing in the social infrastructure of potential destination countries.

The remainder of the paper is structured as follows. Section 2 introduces our econometric approach. In Section 3, we describe the data used in the empirical analysis and provide some descriptive evidence on south-south versus south-north refugee movements. In section 4, we present the regression results. Section 5 concludes.

Read the full working paper.