The Relationship Between Inequality and Potential Emigration: Evidence from the Gallup World Poll

Abstract

The existing literature presents conflicting findings regarding the relationship between inequality and potential emigration. This paper utilizes individual-level data from the Gallup World Poll and country-level income inequality measures for 150 countries to contribute new evidence. The findings indicate that increasing inequality is linked to decreased desires and plans for emigration, a consistent global trend observed across various inequality measures and specifications. Notably, this association is more pronounced for women, individuals without overseas networks, and those lacking financial and human capital. Additionally, the study sheds light on how the level of economic development in countries influences the relationship. In low- and middle-income countries, rising inequality is negatively associated with emigration intentions. Conversely, in affluent nations, heightened inequality stimulates greater desires to emigrate, particularly among high-income and highly educated individuals. These insights provide a deeper understanding of how inequality shapes emigration in diverse world regions and across different cohorts, bridging gaps between previous divergent findings.

1. Introduction

The determinants of international migration stocks and flows are well-established in the social science literature (e.g., Mayda 2010; Czaika and Reinprecht 2022). Push and pull factors, such as poverty, unemployment, and formal and informal institutions, are among the most important causes underpinning migration. In recent years, the broadening of the conceptual frameworks for studying migration beyond the neo-classical models, such as in Carling and Schewel (2018) and de Haas (2021), and the availability of high-quality data on emigration aspirations as part of the Gallup World Poll (GWP) have deepened the understanding of the factors influencing potential and actual emigration, both at the country level (e.g., Docquier, Peri and Ruyssen 2014) and at the individual level (e.g., Migali and Scipioni 2019; Auer, Römer and Tjaden 2020; Manchin 2023).

Several studies have explicitly examined the relationship between inequality and emigration (Liebig and Sousa-Poza 2004; Mihi-Ramírez, Kumpikaitė-Valiūnienė and Cuenca-García 2017; Maestri, Migali and Natale 2017). Their findings differ based on methodological choices and empirical rigor. Other papers in the migration literature have studied inequality as one among several migration determinants (Zaiceva and Zimmermann 2008a; Mayda 2010; Otrachshenko and Popova 2014) or have a different focus of analysis but show complementary results featuring inequality (Borjas 1987; Czaika 2013; Cooray and Schneider 2016). Yet, these studies show conflicting results regarding the link between emigration and inequality, with some documenting a positive relationship, others a negative one, and still others — a nil one (see Supplemental Table B1).

Adapting the aspirations/(cap)ability framework from Carling and Schewel (2018) and de Haas (2021), this paper proposes that inequality is a factor that can influence both migration aspirations (i.e., moving abroad as a desired course of action) and the actual ability to migrate (i.e., the capacity to undertake the move). First, concerning the direct effect of inequality on moving, inequality may be both positively and negatively correlated with emigration, depending on the broader meaning of inequality within the context of different societies. On the one hand, inequality can symbolize prospects of upward mobility (POUM) by signaling that society values and rewards skills, talents, and hard work (Benabou and Ok 2001). On the other hand, individuals may perceive inequality as unfair or believe it is a symptom of dysfunction and injustice, which they may wish to escape by emigrating. Inequality may also limit the actual ability to move — and therefore also influence emigration plans — by creating what Carling (2002) refers to as cumulative immobility, that is, factors that make other individuals decide to stay. For example, by increasing the number of poor people in a country, inequality mechanically limits emigration because emigration is costly. Households that can finance migration are typically richer than those not planning to move (Clemens and Mendola 2020). Structural inequality also may entail that the socioeconomic arrangements within society make migration an appealing course of action for some cohorts (e.g., those with many opportunities) while depriving other cohorts of rights and subjecting them to discrimination (de Haas 2021).

This paper provides new evidence on whether and how country-level economic inequality affects potential emigration measured by individual emigration desires (i.e., whether the respondent would like to migrate to another country permanently) and plans (i.e., whether the respondent plans to move permanently to another country in the next 12 months). To this end, I utilize individual-level information from the GWP and country-level income and wealth inequality data from the World Inequality Database (WID).

The paper makes several contributions to the literature. First, to my knowledge, it is the first study on the migration–inequality nexus that leverages aspirations/(cap)ability theory. Second, it is the only study on how within-country inequality levels affect the emigration desires and plans of individuals that draw from as many as 150 origin countries. Third, I distinguish between two aspects of emigration aspirations — hypothetical migration desires and concrete migration plans — which help provide more nuance about the role of inequality at different stages of the migration process. Fourth, the richness of the data allows explorations of how the relationship between inequality and emigration varies with respondents’ characteristics and the level of economic development in the respective countries of residence. This adds a level of nuance that prior studies lack. Fifth, the paper utilizes four income inequality measures in addition to wealth inequality, providing a broader understanding of inequality's effect on emigration.

I find that income inequality is negatively correlated with emigration desires and plans. In other words, as economic inequality in an origin country increases, the probability of formulating emigration desires and plans decreases. Important nuances appear when I split the global sample based on countries’ level of economic development. Respondents living in poor and middle-income countries drive the negative correlation between inequality and emigration desires. In contrast, inequality is positively associated with emigration intentions in rich countries, especially among high-income and well-educated individuals in those countries. These findings are consistent with the explanation that inequality acts as a structural barrier to the opportunities of some individuals while allowing those with existing capabilities to take advantage of migration (de Haas 2021). In additional specifications, I show that as home country inequality increases, migrant networks abroad, education, and income cushion some of the negative influence of inequality on potential emigration.

2. Theoretical Insights

Push–pull and neo-classical models of migration have recently been criticized for being too simplistic and unable to describe empirical patterns related to migration (e.g., de Haas 2021; Clemens 2022). Traditional neo-classical models view emigration as an investment decision associated with monetary and nonmonetary costs and benefits. People move if the expected utility — a function of income — at the destination exceeds that of the origin, net of migration costs. ¹

Another “mainstream” model of migration — the push–pull framework (Lee 1966) — views income differences between countries as a prime emigration driver. For example, an increase in the average wage differences between origin and 14 Organisation for Economic Co-operation and Development (OECD) destination countries of 1,000 USD (at 2,000 PPP) increases immigrant flows by 10–11 percent of their initial levels (Ortega and Peri 2009). Recent work on GDP per capita as a push factor (Clemens 2020) demonstrates that emigration increases until country per capita income levels of $5,000 at PPP, slows between $5,000 and $10,000, and decreases thereafter, suggesting that the relationship between GDP per capita at the origin and emigration is non-monotonic. Additional push and pull factors include unemployment, poverty, tax systems, public goods and amenities, unhappiness, and institutions (Graham and Markowitz 2011; Polgreen and Simpson 2011; Cai et al. 2014; Chindarkar 2014; Dustmann and Okatenko 2014; Otrachshenko and Popova 2014; Colussi 2016; Manchin 2023). ²

The utility maximization and push–pull migration models often cannot explain real-world migration patterns (e.g., de Haas 2021). For example, if income differences between countries were indeed the main driver of migration, then migration levels would be much higher than observed. The aspirations/ability model by Carling (2002) and Carling and Schewel (2018) instead is based on comparing and contrasting the aspirations (i.e., the conviction that moving abroad is better than nonmoving) and ability to migrate (the capability to undertake the move among those with migration aspirations). In this model, generalized by de Haas (2021), ability and aspirations are both determined by micro- and macro-level circumstances.

The framework by Carling (2002) and Carling and Schewel (2018) is appropriate in the context of this study because it helps understand aspirations and whether inequality is a factor that determines aspirations. Within the aspirations/ability model, inequality can be viewed as a macro-level factor that determines the migration environment (the sociopolitical and economic circumstances in which all residents of a country live). Finally, the immigration interface in the aspirations/ability model is about the modes of migration (e.g., labor migration, illegal migration, and family reunification) along with the associated opportunities and hurdles that determine the ability to move.

Empirically, we cannot follow respondents with migration aspirations and understand whether they undertook the move. Therefore, this paper specifically focuses on aspirations and the migration environment. Nevertheless, studying how inequality as a key component of the migration environment influences aspirations and plans to move can potentially provide insights about the ability to move. The fact that I have information on two degrees of emigration intentions — hypothetical desires and concrete plans — helps shed light on the role of inequality at different stages of the migration process. Specifically, inequality may be viewed as both a factor that determines the migration environment (aspirations) and a factor that determines the ability to move (e.g., by serving as a barrier to moving). This may happen through several mechanisms.

First, at the country level, inequality may directly influence the ability to emigrate through a mechanical effect (McKenzie 2017). Holding average income constant, higher inequality entails more poor individuals who lack access to finance and opportunities to borrow to cover emigration costs. This lower ability to emigrate can also translate to fewer emigration intentions (aspirations and plans) at the individual level. Even if a particular individual is not liquidity-constrained, the fact that fewer compatriots are emigrating may also discourage this individual from emigrating. Fewer compatriots migrating entails fewer information channels and fewer opportunities for cost-sharing (e.g., through traveling together). This makes emigration more costly and less likely for the individual, independent of income. Inequality may thus impose a migration cost that acts to discourage potential emigration. Persistent inequality and discriminatory policies may also constrain the emigration of unprivileged cohorts through social structures that promote exploitation and restrictive social norms (de Haas 2021).

Inequality also positively and negatively influences emigration aspirations. First, inequality levels may signal POUM (Benabou and Ok 2001) and high returns to skill. According to the POUM hypothesis, which has been empirically confirmed in Eastern Europe and the United States (Ravalion and Lokshin 2001; Alesina and La Ferrara 2005; Cojocaru 2014), poor people oppose high taxation and redistribution if they believe they or their children can become rich at some point in the future (Benabou and Ok 2001).

This implies that individuals may tolerate inequality as a symbol of the high rewards for hard work and individual effort or if they believe that they can benefit from inequality (Alesina and Giuliano 2011). In this sense, inequality may discourage the emigration of individuals who think that they can get ahead in life and improve their financial circumstances by living in their home country. Societies experiencing economic growth and transformation processes are more tolerant of inequality as they view it as a marker of future success (Hirschman and Rothschild 1973; Senik 2005; Grosfeld and Senik 2010). Thus, inequality may be negatively associated with emigration intentions if it proxies societal-level rewards for hard work and belief in social mobility and opportunity.

Country-level economic inequality and emigration intentions may be positively associated, whereby increases in inequality may encourage emigration. First, high levels of inequality may signal that the system is non-meritocratic and that the concentration of high incomes at the top of the distribution results from luck and connections (Oishi, Kesebir and Diener 2011; Kuhn 2019). Inequality may be frustrating, as Hirschman's tunnel effect — a metaphor for inequality as a symbol of future mobility — illustrates. It refers to a hypothetical traffic jam on a two-lane road. When the other lane starts moving, the individual initially feels optimistic that it will soon be their turn to move on. Nevertheless, as only the other lane is moving, individuals stuck in the traffic jam feel frustrated as their expectations to leave the traffic do not materialize (Hirschman and Rothschild 1973). In such societies, increasing income disparities may trigger calls for redistribution (Ahrens 2022), protests (Schoene and Allaway 2019; Alexeev and Zakharov 2022), or rising populism (Stoetzer, Giesecke and Klüver 2023). Similarly, increases in inequality may also encourage emigration aspirations.

Second, high levels of economic inequality may accompany low trust (Barone and Mocetti 2016), poor institutions (Savoia, Easaw and McKay 2010), and low-quality or low levels of public goods (Anderson, Mellor and Milyo 2008; De la Croix and Doepke 2009; Stiglitz 2015). Moreover, inequality can lower the incentive to cooperate with fellow citizens (Rothstein and Uslaner 2005; Aksoy 2019) and may hinder economic growth (Cingano 2014; Brueckner and Lederman 2015; Cerra, Lama and Loayza 2021) and lead to poor health (Pickett and Wilkinson 2015) and unhappiness (Ferrer-i-Carbonell and Ramos 2014, 2020). In other words, inequality may proxy poor quality of the social fabric and a weak social contract, which individuals may be trying to escape through emigrating.

Finally, the New Economics of Labor Migration (NELM) suggests a positive association between emigration and inequality if country-level inequality is a proxy for relative deprivation (Stark, Byra and Kosiorowski 2020). Income comparisons with “relevant” peers can trigger dissatisfaction and feelings of relative deprivation (Stark and Bloom 1985; Stark and Taylor 1989; Stark 2006; Stark, Byra and Kosiorowski 2020). Migration can therefore be a tool for individuals to change their relative position in the income distribution or alter their reference group altogether (Stark and Bloom 1985). Higher levels of economic inequality may lead to greater feelings of relative deprivation and trigger emigration. Nevertheless, defining and measuring relative deprivation is difficult in practice, as the relevant reference group may change with migration (Gelatt 2013).

In summary, inequality can activate or constrain people's ability to emigrate, which may, in turn, influence their responses to migration aspiration questions. Finally, in addition to the theoretical mechanisms, the relationship between inequality and potential emigration may depend on empirical choices, such as the measure of emigration, the choice of the countries included in the sample, and the period under investigation. Section 3 details the insights from the extant work on the topic.

3. Empirical Findings of Previous Studies

The existing literature on the emigration–inequality nexus offers conflicting results (Supplemental Table B1). Several papers find a positive relationship (Liebig and Sousa-Poza 2004; Zaiceva & Zimmermann, 2008b), others a negative relationship (Borjas 1987; Czaika 2013), and still others — no relationship (Fouarge and Ester 2007; Otrachshenko and Popova 2014) or a nonlinear relationship (Mayda 2010). One study finds a positive relationship among rich countries but not among poor ones (Mihi-Ramírez, Kumpikaitė-Valiūnienė and Cuenca-García 2017). Another one discovers a negative relationship that disappears with the inclusion of additional control variables (Maestri, Migali and Natale 2017). A report finds a marginally statistically significant positive relationship but only for those with middle levels of education (Fouarge and Ester 2007).

Part of the explanation for these divergent findings lies in the methodological choices: using different data and methods and measuring inequality and emigration (intentions). For example, some studies focus on emigration rates, others on migration stocks, and still others on migration intentions.

Data sets relying on immigrant stocks lack information on pre-migration characteristics, including migrants’ earnings and education levels before leaving, which makes addressing the issue of self-selection of migrants into emigration difficult to tackle. For example, analyses that omit information about the emigrants’ sociodemographic characteristics may wrongly produce a statistically insignificant relationship between emigration and inequality. Specifically, emigrants tend to be relatively young, high-skilled, and male, and these cohorts may be relatively uninformed about or insensitive to inequality. As such, approaches that include the pre-migration characteristics of those who leave, such as this paper, can produce more credible results regarding the relationship between inequality and emigration.

Among papers relying on migration intentions data, there are also large differences in the wording of the migration aspirations questions (Carling and Schewel 2018). Some papers use hypothetical migration aspirations (Liebig and Sousa-Poza 2004) and others moving intentions concerning moving to another city, region, or country in the next 5 years (Zaiceva & Zimmermann, 2008a, 2008b).

The extant studies in the literature also utilize different econometric techniques. While most studies employ multivariate regressions, one study uses bivariate correlations (Czaika 2013), and some authors only summarize but do not fully report their empirical results (Maestri, Migali and Natale 2017; Mihi-Ramírez, Kumpikaitė-Valiūnienė and Cuenca-García 2017).

4. Data and Variables

4.1. The Gallup World Poll and Information on Emigration Intentions

The individual-level data are based on the GWP, which surveys individuals living in over 150 countries worldwide, representing 99 percent of the world's adult population aged 15 and older. While the survey started in 2005/2006, I focus on 2009–2019 as income and employment information is only available since 2009. In 2020, there are only very few countries where the emigration intention question is asked, so I dropped it from the analysis. Interviews are conducted via phone in countries and areas where telephone coverage is widespread (Northern America, Western Europe, developed Asia, and Gulf Cooperation Council countries). Face-to-face interviews occur in Central and Eastern Europe, much of Latin America, former Soviet Union states, and nearly all of Asia, the Middle East, and Africa. ³ Different individuals are polled each year, and as such, the data set presents pooled cross-sections rather than a panel tracing the same individuals over time.

Several features make the GWP more advantageous than other data sources with emigration intention questions explored in the literature (Nikolova, 2016; Carling and Schewel 2018). First, while other surveys containing information about emigration intentions have limited geographic coverage, the GWP is a nationally representative survey providing about 1,000 observations per country for a large sample of countries (see Supplemental Tables B2–B3). Second, the GWP elicits information about different aspects of emigration aspirations: desires, plans, and preparations. Finally, it contains rich individual-level information ranging from household and individual sociodemographics to opinions, well-being, and actual and intended behaviors. These variables are important factors for the decision to move and thus feature as control variables in my analysis.

Despite its advantages, the GWP does not provide the reason for desired emigration, which makes it impossible to distinguish between voluntary (e.g., economic vs. family-based migrants) and involuntary migrants (i.e., refugees and asylum seekers). Inequality levels may be irrelevant or relatively unimportant for family migrants and those escaping climate change. If such groups of migrants dominate the analysis sample, I may wrongly conclude that inequality is unassociated with emigration levels. Nevertheless, most international movers are economic migrants (McAuliffe and Triandafyllidou 2022), which alleviates such concerns.

I utilize the following questions to capture emigration aspirations (see Table 1): ⁴

- Emigration desires (2009–2019): Ideally, if you had the opportunity, would you like to move permanently to another country, or would you prefer to continue living in this country?

- Emigration plans (asked of respondents with emigration desires and available 2009–2015): Are you planning to move permanently to another country in the next 12 months or not? ⁵

Table 1.

Variable Definitions.

Variable	Definition
Individual variables
Emigration desire	A binary variable based on the question (WP1325) “Ideally, if you had the opportunity, would you like to move PERMANENTLY to another country, or would you prefer to continue living in this country?”; 1 = yes, 0 = no
Emigration plan	A binary variable based on the question (WP10252) “Are you planning to move permanently to another country in the next 12 months, or not?” (Asked only of those who would like to move to another country); 1 = yes, 0 = no; therefore, the 0 = no designates respondents with emigration intentions who do not have emigration plans. When the variable had missing information, I checked whether there were valid answers given to question WP6880 asked in 2008/2009 “Are you planning to move permanently to that country in the next 12 months, or not?” (Asked only of those who specified a country to which they would like to move)
Female	Respondent's biological sex; 0 = male, 1 = female
Age	Respondent's age in years
Immigrant	An indicator of whether the respondent was born in the country of interview; 1 = yes, 2 = no, 3 = missing information
Rural location	An indicator capturing whether the respondent's location is rural or not. 1 = rural, 2 = small town, large city, and suburb, 3 = no information
Married	A binary indicator capturing the respondent's marital status; 1 = married/domestic partnership, 0 = single/widowed/divorced
Tertiary education	A binary indicator capturing the respondent's educational level; 1 = completed 4 years of education beyond high school and/or received a 4-year college degree, 0 = completed elementary education or completed secondary education
Children in the household	A binary indicator capturing whether the respondent has children living in the household; 1 = yes, 0 = no
Income tertile	An indicator variable indicating the within-country per capita annual household income in International USD; 1 = bottom income tertile, 2 = middle-income tertile, 3 = top third tertile, 4 = missing information
Unemployed	Whether the respondent is unemployed or not. 1 = unemployed, 2 = working or out of the workforce
Key independent variables (country-level)
Top 1% income share (lag)	Top 1% share of pre-tax national income for adults, including elderly (20+), equal-split adults (i.e., income divided equally among spouses), lagged one time period, based on the WID
Top 10% income share (lag)	Top 10% share of pre-tax national income for adults, including elderly (20+), equal-split adults (i.e., income divided equally among spouses), lagged one time period, based on the WID
Top 20% income share (lag)	Top 20% share of pre-tax national income for adults, including elderly (20+), equal-split adults (i.e., income divided equally among spouses), lagged one time period, based on the WID
Income Gini coefficient (lag)	The Gini index ranges from 0 (perfect quality) to 100 (perfect inequality). It is a measure of how much the income distribution deviates from a perfectly equal distribution, based on the WID
Alternative independent variables (country-level)
Top 1% wealth share (lag)	Top 1% share of pre-tax national income for adults, including elderly (20+), equal-split adults (i.e., wealth divided equally among spouses), lagged one time period, based on the WID
Top 10% wealth share (lag)	Top 10% share of pre-tax national income for adults, including elderly (20+), equal-split adults (i.e., wealth divided equally among spouses), lagged one time period, based on the WID
Top 20% wealth share (lag)	Top 20% share of pre-tax national income for adults, including elderly (20+), equal-split adults (i.e., wealth divided equally among spouses), lagged one time period, based on the WID
Wealth Gini coefficient (lag)	The Gini index ranges from 0 (perfect quality) to 100 (perfect inequality). It is a measure of how much the wealth distribution deviates from a perfectly equal distribution, based on the WID
Country-level controls
Life evaluations (lag)	Average country-level variable based on the responses to the question asking respondents to position their current lives on an 11-step ladder, where 0 denotes the worst possible life they can imagine for themselves and 10 denotes the best possible life they can imagine for themselves from the Statistical Appendix to the World Happiness Report 2021. The variable’s original source is the Gallup World Poll. The variable is lagged one time period.
Log GDP per capita (lag)	Log-transformed GDP per capita (variable name GDP) in purchasing power parity (PPP) at constant 2017 international dollar prices from the Statistical Appendix to the World Happiness Report 2021. The original data source is the World Development Indicators (WDI), and data from Taiwan, Syria, Palestine, Venezuela, Djibouti, and Yemen are from the Penn World Table. Before log-transforming this variable, we added 1 to 0 observations. The variable is lagged one time period.
Social support (lag)	Share of respondents in the country of interview answering “Yes” to the binary question “If you were in trouble, do you have relatives or friends you can count on to help you whenever you need them, or not?” from the Statistical Appendix to the World Happiness Report 2021. The variable’s original source is the Gallup World Poll. The variable is lagged one time period.
Healthy life expectancy (lag)	Healthy life expectancies at birth are from the World Health Organization’s (WHO) Global Health Observatory data repository (last updated: September 28, 2020). The original data are for 2000, 2005, 2010, 2015, and 2016. The World Happiness Report researchers interpolated and extrapolated missing data for the analysis period 2005–2020. The data are derived from the Statistical Appendix to the World Happiness Report 2021. The variable is lagged one time period.
Freedom (lag)	Share of respondents in the country of interview answering “Yes” to the question “Are you satisfied or dissatisfied with your freedom to choose what you do with your life?” from the Statistical Appendix to the World Happiness Report 2021. The variable’s original source is the Gallup World Poll. The variable is lagged one time period.
Generosity (lag)	The residual of regressing national average of response to the GWP question “Have you donated money to a charity in the past month?” on GDP per capita from the Statistical Appendix to the World Happiness Report 2021. The variable’s original source is the Gallup World Poll and the World Development Indicators. The variable is lagged one time period.
Corruption perceptions (lag)	Average country-level variable based on the responses to the questions “Is corruption widespread throughout the government or not” and “Is corruption widespread within businesses or not?” The variables are coded as 0 = no and 1 = yes and then added up and averaged. The data are from the Statistical Appendix to the World Happiness Report 2021. The variable’s original source is the Gallup World Poll. The variable is lagged one time period.

The GWP question on emigration desires has two key limitations (Carling and Schewel 2018). First, respondents may find it cognitively difficult to distinguish between a hypothetical ideal situation and their attitudes toward migration. Second, it is unclear how respondents who want to emigrate temporarily and not permanently answer this question.

Naturally, emigration intentions reported in surveys are not about actual but rather about intended behavior, and some of those expressing such intentions may never move. Nevertheless, emigration intentions correlate well with actual migration behavior (Simmons 1985; Creighton 2013; Van Dalen and Henkens 2013; Docquier, Peri and Ruyssen 2014; Bertoli and Ruyssen 2018; Tjaden, Auer and Laczko 2019; Adema, Aksoy and Poutvaara 2021). Using the GWP, Bertoli and Ruyssen (2018) show that emigration desires correlate highly with migration to OECD destinations, with correlations ranging from 0.4 to 0.8. ⁶ Docquier, Peri, and Ruyssen (2014) demonstrate that the correlation between emigration desires and actual migration from 138 origin countries to 30 destinations is 0.93 for the college-educated and 0.24 for the non–college educated. ⁷

These studies suggest that emigration desires and plans may overstate actual emigration but are nevertheless meaningful predictors of potential emigration. Furthermore, analyses of emigration intention data offer insight into the prospective emigration flows, thus providing policy input for targeted, proactive migration policies.

4.2. Inequality Data and Measures

Studying the relationship between income inequality and emigration intentions assumes that (i) individuals are aware of and sensitive to the level of their home country's inequality and (ii) the researcher knows which inequality measure is relevant for individuals (Clark and d'Ambrosio 2015; Ferrer-i-Carbonell and Ramos 2020). Regarding the first assumption, country-level inequality levels negatively correlate with people's subjective well-being, suggesting that inequality bothers individuals (Bjørnskov et al. 2013; Clark and d'Ambrosio 2015; Ferrer-i-Carbonell and Ramos 2020).

In addition, extant studies have utilized several measures of within-country income inequality (see Supplemental Table B1), with the majority focusing on the Gini coefficient. As no single measure of inequality can provide a complete picture of a country's income distribution (Alvaredo et al. 2020), I utilize several indicators from the WID. These measures include the top 1 percent share, the top 10 percent share, the top 20 percent share, and the Gini coefficient. Furthermore, while the main focus is on income inequality, I also include complementary analyses with wealth inequality.

Unlike other inequality data sources that primarily rely on household surveys, the WID uses various data sets to construct its measures, such as tax data, national accounts, surveys, and wealth rankings (WID 2022). ⁸

4.3 Other Data Sources

I utilize country-level controls, including life satisfaction, GDP per capita, social support, generosity, healthy life expectancy, freedom to make life choices, and corruption perceptions from the Statistical Appendix of the 2021 World Happiness Report (Helliwell et al. 2021). I impute missing information from the nearest neighboring observation for each country or, in a few cases, from the average values for neighboring countries. The World Happiness Report is based on the GWP and provides the most complete coverage of country-level information for the countries in the GWP.

GDP per capita is originally sourced from the World Development Indicators and the Penn World Tables, and the healthy life expectancy comes from the World Health Organization's (WHO) Global Health Observatory data repository. All other country-level controls are based on country-level averages of variables from the GWP (see Table 1 for further clarifications and definitions).

5. Methods

5.1. Baseline Empirical Specification

The emigration intention M of individual i in time period t living in country j is: $$M_{ijt}=\alpha +\gamma Inequalit{y}_{jt-1}+X_{ijt}^{\mathrm{\prime }}\beta +C_{jt-1}^{\mathrm{\prime }}\phi +{\pi }_r+{\tau }_t+{\pi }_r\times d+u_{ijt,}$$ (1)where Inequality_ijt−1 is the within-country inequality measured as the top 1 percent income share of pre-tax national income, the top 10 percent income share of pre-tax national income, the top 20 percent income share of pre-tax national income, or the Gini index, lagged one time period, X is a vector of individual-level control variables (age, gender, immigrant status, marital status, education level, income, the presence of children in the household, urban or rural location, and employment status), C captures country-level variables (life satisfaction, log GDP per capita, social support, healthy life expectancy, freedom perceptions, generosity, and corruption perceptions), π_r is the geographic region of residence dummies, τ_t is the time dummy, π_r × d is the interaction between the region of residence and a linear time trend, and u_ijt is the stochastic error term.

All estimations use the Gallup-provided survey weight and standard errors clustered at the country × year level. I estimate Equation (1) using a logit estimator. Ordinary least squares (OLS) estimations, available upon request, provide qualitatively similar results.

In separate analyses, I also use wealth inequality, measured by the net personal wealth share held by the top 1 percent, top 10 percent, top 20 percent, and the wealth Gini index. Net personal wealth captures the total value of nonfinancial and financial assets (such as housing, land, deposits, bonds, and equities) held by households, net of their debts.

Inequality and other country-level variables are lagged one time period to account for the fact that the link between inequality and emigration intentions is not instantaneous or that the country-level data may be released after the Gallup interview date. The individual controls — standard sociodemographic controls that are correlated with the emigration decision — are based on the related literature (e.g., Cai et al. 2014; Dustmann and Okatenko 2014; Adema, Aksoy and Poutvaara 2021; Graham and Nikolova, 2018; Manchin 2023). Including these variables mitigates issues related to the self-selection of individuals into potential emigration and allows for comparing individuals with similar sociodemographic characteristics, living in comparable country circumstances, and subject to observed and unobserved shocks, such as the economic crisis of 2007–2009. Importantly, because income is included as part of the control variables, the analyses compare how inequality affects the emigration decisions above and beyond the influence of individual income.

Equation (1) does not include country-fixed effects because within-country inequality does not change much over time. For example, the within-country standard deviation of the lagged inequality measures ranges between 0.012 and 0.014. For comparison, the within-country standard deviation of lagged life satisfaction is 0.337. Therefore, adding country-fixed effects essentially absorbs all the within-country variation in inequality and wrongly produces statistically nonsignificant results. Therefore, like other papers that study inequality, I use region-fixed effects instead of country-fixed effects and a rich set of country-level variables capturing the socioeconomic and institutional conditions. Furthermore, the π_r × d fixed effects account for economic or political development trends within geographic regions. Such trends include climate or economic crises that differentially affect world regions.

5.2. Econometric Challenges and Causality

Studying the causal effect of inequality on emigration faces the primary challenge of reverse causality. This is because inequality may influence migration, but migration can also impact inequality through, among other things, remittances (Alpaslan et al. 2021). For example, if individuals from richer households are more likely to emigrate, their families left behind may become even richer due to remittances (McKenzie and Rapoport 2007). In some countries, such as El Salvador, Liberia, and Nepal, remittances account for over 20 percent of GDP (Alpaslan et al. 2021). However, if migration costs are low or liquidity constraints are nonbinding, poorer individuals are more likely to migrate, possibly reducing inequality in the origin country (McKenzie and Rapoport 2007). In addition, migration can affect wealth inequality if wealthy individuals move across international borders and take their assets with them.

My analysis is based on intended rather than actual emigration, which somewhat mitigates the issue. Nevertheless, to the extent that emigration decisions are correlated with actual migration behavior, some endogeneity concerns remain.

To deal with this endogeneity issue, I report results based on an instrumental variable technique in Appendix C. I instrument current inequality levels with information on traditional inheritance practices from Giuliano and Nunn (2018). These additional results align with my baseline findings.

Furthermore, self-selection into migration is important in migration economics (Borjas 1987; Chiswick 1999; Nikolova 2015). I control for sociodemographic characteristics to rule out the possibility that self-selection drives the results. Yet, this does not fully resolve the issue, and results should be interpreted with caution.

6. Descriptive Statistics

Figures 1 and 2 map the distribution of the top 1 percent income share and the Gini coefficient for the 2009–2019 period. Darker colors indicate higher income inequality. The top 1 percent share ranges from 0.07 (i.e., the richest 1 percent of individuals earn 7 percent of the national income) in the Netherlands, North Macedonia, and Slovenia to 0.30 in Mozambique and 0.31 in the Central African Republic.

Figure 1.

Top 1 percent Income Share in the Analysis Sample, by Country (2009–2019).

Figure 2.

Gini Coefficient (Income) in the Analysis Sample, by Country (2009–2019).

The Gini coefficient patterns for 2009–2019 are similar to those of the top 1 percent share (Figure 2). The countries with the lowest income inequality are the Czech Republic, Iceland, Slovakia, Sweden, Norway, and the Netherlands, while the Central African Republic, Mexico, and Namibia are among the countries with the highest Gini index inequality.

I next turn to the measures of potential emigration for 2009–2019. Figure 3 details that about a fifth of respondents reported emigration intentions, which aligns with the findings of other studies using the GWP (Cai et al. 2014; Graham and Nikolova, 2018; Adema, Aksoy and Poutvaara 2021). The average share of those with emigration desires ranges from 0.02 in Indonesia and 0.03 in Thailand to slightly over 0.6 in Liberia and Sierra Leone.

Figure 3.

Average Country-Level Share of Respondents Reporting Emigration Desires in the Analysis Sample (2009–2019).

Furthermore, about 15 percent of those with emigration intentions plan to emigrate in the next year, which comprises 3 percent of the overall analysis sample (see Supplemental Figure B1 and Table B3). The share of those with emigration plans ranges from 0.01 in Japan to 0.51 in Libya.

Table 2 and Supplemental Table B4 report the summary statistics for the analysis sample concerning emigration intention and plan analysis samples, respectively. Respondents with and without emigration intentions differ along key sociodemographic variables, such as age, rural/urban location, marital status, employment status, and the presence of children in the household.

Table 2.

Summary Statistics: Emigration Desire Sample 2009–2019.

	Overall Sample, N = 1,455,295		Emigration Desires = yes, N = 316,512		Emigration Desires = no, N = 1,138,783
Variables	Mean	Std. dev.	Mean	Std. dev.	Mean	Std. dev.
Emigration intention	0.223	0.417
Biological sex
Male	0.490	0.500	0.534	0.499	0.477	0.499
Female	0.510	0.500	0.466	0.499	0.523	0.499
Age	39.257	17.395	32.345	14.083	41.246	17.745
Immigrant status
Native	0.921	0.270	0.916	0.277	0.922	0.268
Immigrant	0.052	0.222	0.059	0.236	0.050	0.217
No information	0.028	0.164	0.025	0.155	0.028	0.166
Location
Rural location	0.704	0.456	0.748	0.434	0.691	0.462
Urban location	0.278	0.448	0.235	0.424	0.291	0.454
No information	0.018	0.131	0.017	0.130	0.018	0.132
Marital status
Married	0.580	0.494	0.459	0.498	0.615	0.487
Not married/divorced/widowed	0.420	0.494	0.541	0.498	0.385	0.487
Education
Primary or secondary education	0.883	0.321	0.872	0.334	0.886	0.318
Tertiary education	0.117	0.321	0.128	0.334	0.114	0.318
Children in the household
Yes	0.562	0.496	0.603	0.489	0.550	0.498
No	0.438	0.496	0.397	0.489	0.450	0.498
Within-country income tertile
Poorest third	0.395	0.489	0.383	0.486	0.399	0.490
Middle third	0.323	0.468	0.314	0.464	0.326	0.469
Richest third	0.251	0.434	0.263	0.440	0.248	0.432
No information	0.030	0.170	0.039	0.194	0.027	0.163
Unemployment status
Not unemployed	0.910	0.286	0.875	0.330	0.921	0.270
Unemployed	0.066	0.248	0.105	0.307	0.054	0.227
No information	0.024	0.152	0.019	0.138	0.025	0.156
Key independent variables (country-level)
Top 1% income share (lag)	0.162	0.050	0.161	0.051	0.163	0.050
Top 10% income share (lag)	0.456	0.088	0.460	0.088	0.455	0.088
Top 20% income share (lag)	0.602	0.082	0.606	0.081	0.601	0.082
Gini (lag)	0.565	0.082	0.569	0.082	0.564	0.082
Country-level controls
Life evaluations	5.391	1.072	5.223	1.039	5.439	1.076
Log GDP per capita	9.331	1.114	9.158	1.119	9.381	1.108
Social support	0.805	0.120	0.792	0.119	0.809	0.119
Healthy life expectancy	63.138	7.119	62.056	7.625	63.449	6.935
Freedom	0.732	0.142	0.711	0.139	0.738	0.142
Generosity	−0.009	−0.163	−0.022	−0.142	−0.006	−0.168
Corruption perceptions	0.748	0.181	0.778	0.157	0.740	0.187

Note: See Table 1 for variable definitions. The values are calculated using the Gallup-provided survey weight.

7. Results

7.1. Results Concerning the Global Sample

Table 3 presents the baseline results from estimating Equation (1) for all four inequality measures. For ease of interpretation, average marginal effects are presented. The dependent variable is emigration desires in Models (1)–(4) and plans in Models (5)–(8). All regressions include controls for individual-level sociodemographic factors (biological sex, age, immigrant status, children, marital status, rural/urban location, education, income group, and unemployment status), country-level factors (life satisfaction, corruption, generosity, social support, GDP per capita, life expectancy, and freedom perceptions), and year, region, and time trend×region fixed effects and report standard errors clustered at the country×year level. For brevity, I only report the coefficient estimates of the key independent variables. Full econometric output is available upon request.

Table 3.

The Relationship Between Income Inequality Levels and Emigration Desires and Plans.

	Emigration Desires, 2009–2019						Emigration Plans, 2009–2015
	(1)	(2)		(3)	(4)		(5)	(6)	(7)	(8)
Top 1% share (lag)	−0.490***						−0.239***
	(0.065)						(0.086)
Top 10% share (lag)		−0.340***						−0.168**
		(0.048)						(0.065)
Top 20% share (lag)			−0.347***						−0.146**
			(0.051)						(0.070)
Gini index (lag)						−0.314***				−0.102
						(0.048)				(0.065)
Pseudo R2	0.093	0.093	0.092			0.092	0.056	0.056	0.056	0.056
Observations	1,455,295	1,455,295	1,455,295			1,455,295	184,295	184,295	184,295	184,295
Year FE	Y	Y	Y			Y	Y	Y	Y	Y
Individual controls	Y	Y	Y			Y	Y	Y	Y	Y
Country-level controls	Y	Y	Y			Y	Y	Y	Y	Y
Region FE	Y	Y	Y			Y	Y	Y	Y	Y
Region × linear time trend	Y	Y	Y			Y	Y	Y	Y	Y

Note: The table reports the average marginal effects of logit estimations using robust standard errors clustered at the country×year level. The dependent variable in Models (1)–(4) is emigration desires, and in Models (5)–(8), it is emigration plans. All regressions include year fixed effects, individual controls (biological sex, age, immigrant status, children, marital status, rural/urban location, education, income group, and unemployment status), geographic region-fixed effects, region×time trend controls, and country-level controls (life satisfaction, corruption, generosity, social support, GDP per capita, life expectancy, and freedom perceptions). All regressions are estimated using the survey weight. See Table 1 for variable definitions.

*** p < 0.01, ** p < 0.05, * p < 0.1.

First, Models (1)–(4) in Table 3 demonstrate that emigration intentions negatively correlate with all inequality measures. According to Model (1), a one percentage point increase in the top 1 percent income share is associated with a 0.490 percentage point decrease in the probability of reporting emigration desires. Given that the average share of those reporting emigration desires in the sample is 0.22 and the top 1 percent income share is 0.16 (see Table 2), increasing the top 1 percent income share from 0.16 to 0.17 would imply a decrease in the probability of reporting emigration intentions from 0.220 to 0.215 (or from 22% to 21.5%). This is a modest, though meaningful, impact in terms of economic significance. These results are qualitatively in line with Czaika (2013) and Borjas (1987), who found that income inequality is associated with lower male immigration to the United States.

These findings differ from those of Liebig and Sousa-Poza (2004), who find that inequality positively correlates with emigration intentions. The differences with Liebig and Sousa-Poza (2004) can be due to several factors, including the different time period, the larger number of countries, and the coverage of countries at lower levels of economic development that I use in my estimations. By contrast, Liebig and Sousa-Poza (2004) used 23 mostly high- or upper-middle-income countries in 1995. ⁹ They also use a different emigration intention variable and different control variables.

The association between inequality and emigration plans is also negative (Models (5)–(8) of Table 3) across all inequality measures. In Model (8), the Gini index is not statistically significantly associated with emigration plans. One interpretation of the smaller magnitude compared with the emigration desire models is that inequality acts as a barrier regarding aspirations but not for making emigration plans. In other words, economic inequality may be less relevant once individuals have made their emigration decisions and are planning the move. This first set of findings implies that inequality is a deterrent and not a push factor for emigration decisions.

7.2. Robustness Checks

To check whether the main results depend on the choice of the control variables, I conduct specification curve analyses (Simonsohn, Simmons and Nelson 2015, 2020). The main idea is to estimate modifications of Equation (1) using alternative control variables, weighting schemes, and subsamples and graphically plot the distribution of the results and their confidence intervals.

Figure 4 provides the results concerning the emigration intention sample for the top 1 percent income share (reporting average marginal effects). The main results from Model (1) in Table 3 are highlighted in dark blue and use all control variables (all sociodemographic variables, country-level controls, and year, region, and year×time fixed effects). ¹⁰ I consequently plot the estimates and confidence intervals from alternative specifications. Specifically, (i) I only control for the year, region, and year×time fixed effects and then (ii) include only exogenous demographic variables and exclude so-called bad controls (Angrist and Pischke 2009), which may be the outcome of inequality themselves. Specifically, the exogenous variables I include are gender, age, and immigrant status. The endogenous demographics I exclude are sociodemographics related to rural/urban location, marital, employment, education status, children's presence in the household, and household income. Next, (iii) I include the lagged country-level controls (life satisfaction, log real GDP per capita, social support, life expectancy, freedom perceptions, generosity, and corruption perceptions). The next set of specifications (iv) exclude the Gallup weight, (v) limit the analysis sample to respondents between ages 18 and 60 to better capture working-age cohorts, and (vi) exclude the foreign-born whose emigration desires may reflect return migration. The last set of specifications excludes one geographic region at a time.

Figure 4.

Specification Curve Analysis, Emigration Desire Sample, Results Related to the Top 1 percent Share Model (1) of Table 3.

The specification curve analysis suggests that the choice of the controls, the weighting scheme, or particular subsamples do not drive the main results, and the conclusions are robust. In addition, Appendix C documents the instrumental variable analysis, whose findings align with the main specification.

I also check whether the results hold when I use wealth inequality measures (Table 4). Supplemental Figures B3 and B4 in the appendix plot the geographic distribution of wealth inequality. The Gini wealth index in our sample is lowest in Malta and highest in South Africa. The top 1 percent wealth share is lowest in Belgium and Slovakia and highest in South Africa.

Table 4.

The Relationship Between Wealth Inequality Levels and Emigration Desires and Plans.

	(1)	(2)	(3)	(4)
	Panel A: Emigration Desires, 2009–2019
Top 1% share (lag)	−0.295***
	(0.036)
Top 10% share (lag)		−0.280***
		(0.036)
Top 20% share (lag)			−0.282***
			(0.044)
Gini index (lag)				−0.270***
				(0.043)
Observations	1,449,317	1,449,317	1,449,317	1,449,317
Pseudo R2	0.094	0.093	0.093	0.092
	Panel B: Emigration plans, 2009–2015
Top 1% share (lag)	−0.061
	(0.052)
Top 10% share (lag)		−0.043
		(0.052)
Top 20% share (lag)			−0.053
			(0.063)
Gini index (lag)				−0.038
				(0.062)
Observations	183,400	183,400	183,400	183,400
Pseudo R2	0.055	0.055	0.055	0.055
Year FE	Y	Y	Y	Y
Individual controls	Y	Y	Y	Y
Country-level controls	Y	Y	Y	Y
Region FE	Y	Y	Y	Y
Region × linear time trend	Y	Y	Y	Y

Note: The table reports the average marginal effects of logit estimations using robust standard errors clustered at the country×year level. The dependent variable in Panel A is emigration desires and in Panel B emigration plans. All regressions include year fixed effects, individual controls (biological sex, age, immigrant status, children, marital status, rural/urban location, education, income group, and unemployment status), geographic region-fixed effects, region×time trend controls, and country-level controls (life satisfaction, corruption, generosity, social support, GDP per capita, life expectancy, and freedom perceptions). All regressions are estimated using the survey weight. See Table 1 for variable definitions.

***p < .01. **p < .05. *p < .1.

The results in Table 4 show that wealth inequality is negatively associated with emigration desires but not plans. The magnitude of the coefficient estimates, which are average marginal effects, in Panel A of Table 4 is also lower than those in Models (1)–(4) of Table 3. A potential explanation is that, unlike income inequality, wealth inequality does not immediately produce a larger number of poor people in the country. Thus, it does not constrain their emigration plans to the same extent that income inequality does.

My main finding is that globally, emigration intentions and plans are negatively associated with country-level income inequality. In other words, increases in within-country income inequality correspond to decreases in the probability of expressing emigration desires and plans. In the next two sections, I uncover nuances and patterns in this relationship based on the characteristics of individual respondents (Section 8) or the level of economic development of the countries in which they live (Section 9).

8. Heterogeneity According to Respondents’ Sociodemographic Characteristics

In Table 5, I add interactions with whether the respondent has tertiary education (Panel A), whether the respondent belongs to the richest within-country quartile (Panel B), whether the respondent is female (Panel C), and whether they have networks abroad (Panel D). The coefficient estimates are reported directly following the logit estimations. Nevertheless, the patterns in the associations can be discerned by looking at the raw coefficient estimates.

Table 5.

The Relationship Between Inequality Levels and Emigration Desires: Heterogeneity Analysis.

	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)
	Panel A: Interactions with having Tertiary Education				Panel C: Interactions with being Female
Top 1% share (lag)	−3.241***				−3.028***
	(0.421)				(0.421)
Top 1% share (lag) × variable	1.154***				−0.196
	(0.274)				(0.190)
Top 10% share (lag)		−2.261***				−2.064***
		(0.306)				(0.304)
Top 10% share (lag) × variable		0.866***				−0.219**
		(0.152)				(0.094)
Top 20% share (lag)			−2.323***				−2.117***
			(0.326)				(0.325)
Top 20% share (lag) × variable			0.954***				−0.194*
			(0.169)				(0.100)
Gini index (lag)				−2.103***				−1.962***
				(0.307)				(0.306)
Gini index (lag) × variable				0.775***				−0.083
				(0.170)				(0.096)
Variable	0.020	−0.181***	−0.361***	−0.230**	−0.203***	−0.134***	−0.117**	−0.188***
	(0.044)	(0.066)	(0.098)	(0.095)	(0.030)	(0.042)	(0.059)	(0.055)
Pseudo R2	0.093	0.093	0.092	0.092	1,455,295	1,455,295	1,455,295	1,455,295
Observations	1,455,295	1,455,295	1,455,295	1,455,295	0.093	0.093	0.092	0.092

	Panel B: Interactions with the richest within-country income category				Panel D: Interactions with networks abroad
Top 1% share (lag)	−3.503***				−3.402***
	(0.423)				(0.424)
Top 1% share (lag) × variable	1.457***				1.816***
	(0.172)				(0.317)

	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)
Top 10% share (lag)		−2.407***				−2.256***
		(0.305)				(0.310)
Top 10% share (lag) × variable		0.904***				0.695***
		(0.090)				(0.179)
Top 20% share (lag)			−2.460***				−2.355***
			(0.325)				(0.333)
Top 20% share (lag) × variable			0.950***				0.688***
			(0.096)				(0.192)
Gini index (lag)				−2.226***				−2.267***
				(0.305)				(0.316)
Gini index (lag) × variable				0.846***				0.764***
				(0.095)				(0.188)
Variable	−0.222***	−0.401***	−0.559***	−0.468***	0.308***	0.285***	0.191	0.172
	(0.030)	(0.043)	(0.059)	(0.055)	(0.055)	(0.086)	(0.120)	(0.110)
Pseudo R2	0.093	0.093	0.093	0.092	0.103	0.103	0.103	0.103
Observations	1,455,295	1,455,295	1,455,295	1,455,295	1,352,058	1,352,058	1,352,058	1,352,058
Year FE	Y	Y	Y	Y	Y	Y	Y	Y
Individual controls	Y	Y	Y	Y	Y	Y	Y	Y
Country-level controls	Y	Y	Y	Y	Y	Y	Y	Y
Region FE	Y	Y	Y	Y	Y	Y	Y	Y
Region × linear time trend	Y	Y	Y	Y	Y	Y	Y	Y

Note: The table reports the logit coefficient estimates using robust standard errors clustered at the country×year level. The dependent variable in all models is emigration desires. The table presents the results from regressions with different interaction variables. In Panel A, the interaction variable is whether the respondent has a tertiary education degree or not; in Panel B, it is whether the respondent belongs to the highest income category within their country of residence (i.e., whether the respondent belongs to the top tertile in their country of origin's income distribution); in Panel C, it is whether the respondent is female; and in Panel D, it is whether the respondent has a network of family and friends abroad. All regressions include year fixed effects, individual controls (biological sex, age, immigrant status, children, marital status, rural/urban location, education, income group, and unemployment status), region-fixed effects, region×time trend controls, and country-level controls (life satisfaction, corruption, generosity, social support, GDP per capita, life expectancy, and freedom perceptions). All regressions are estimated using the survey weight. See Table 1 for variable definitions.

***p < .01, **p < .05, *p < .1.

Does the negative relationship between inequality and emigration desires depend on the respondent's education level? The results of Panel A of Table 5 show that, as before, higher income inequality is associated with lower emigration desires and that tertiary-educated individuals are generally less likely to want to move abroad (compared with those with secondary and primary education). Nevertheless, the negative association between inequality and emigration intentions is lower in magnitude for the high-skilled. This is evident from the fact that while the coefficient estimates on all inequality measures are negative, the coefficient estimate on the interaction term between inequality and tertiary education is positive. In other words, while inequality is a barrier to potential emigration, its influence is slightly less negative for the high-skilled. Education, therefore, cushions, though it does not fully offset, the negative effects of inequality for emigration. The results could reflect the fact that inequality makes potential emigration difficult, but being more educated mitigates parts of this burden. Table 7 further details how this finding depends on the level of economic development of the respondent's home country.

Table 6.

The Relationship Between Inequality Levels and Emigration Desires, by Level of Economic Development and Within-Country Income Tertile.

	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)	(9)	(10)	(11)	(12)
	Top 1% Share (lag)			Top 10% Share (lag)			Top 20% Share (lag)			Gini Index (lag)
Variables	Low-Income	Middle-Income	High-Income	Low-Income	Middle-Income	High-Income	Low-Income	Middle-Income	High-Income	Low-Income	Middle-Income	High-Income
Panel A: Estimates by country income level
Inequality	−3.610***	−2.014***	0.835	−3.123***	−1.527***	2.566***	−3.352***	−1.621***	3.025***	−3.569***	−1.309***	1.885***
	(0.739)	(0.546)	(0.903)	(0.631)	(0.384)	(0.626)	(0.724)	(0.415)	(0.591)	(0.678)	(0.403)	(0.537)
Observations	469,156	547,326	438,813	469,156	547,326	438,813	469,156	547,326	438,813	469,156	547,326	438,813
Pseudo R2	0.103	0.112	0.084	0.103	0.112	0.085	0.103	0.112	0.085	0.103	0.111	0.085
Panel B: Estimates by country income level and the respondent's country-specific income position
Inequality	−3.481***	−2.850***	−0.300	−3.029***	−2.087***	1.997***	−3.209***	−2.239***	2.495***	−3.378***	−1.878***	1.391**
	(0.768)	(0.565)	(0.950)	(0.645)	(0.395)	(0.639)	(0.740)	(0.430)	(0.603)	(0.686)	(0.418)	(0.550)
Second income quartile	0.072	−0.186***	−0.308***	0.045	−0.310***	−0.412***	0.027	−0.445***	−0.524***	−0.004	−0.424***	−0.419***
	(0.058)	(0.041)	(0.051)	(0.123)	(0.067)	(0.078)	(0.178)	(0.096)	(0.112)	(0.161)	(0.088)	(0.103)
Third income tertile	0.137**	−0.339***	−0.459***	0.336**	−0.623***	−0.635***	0.501**	−0.898***	−0.827***	0.459**	−0.760***	−0.703***
	(0.068)	(0.055)	(0.054)	(0.153)	(0.085)	(0.079)	(0.224)	(0.120)	(0.112)	(0.203)	(0.109)	(0.105)
Income tertile missing	0.443	4.719***	0.046	0.084	4.542**	0.212	0.249	6.346*	0.396	1.233	5.673**	0.119
	(0.365)	(1.465)	(0.313)	(1.037)	(2.169)	(0.429)	(1.624)	(3.301)	(0.553)	(1.373)	(2.587)	(0.485)
Inequality X Second income tertile	−0.108	1.072***	1.588***	0.016	0.653***	0.803***	0.039	0.718***	0.786***	0.094	0.718***	0.637***
	(0.324)	(0.226)	(0.385)	(0.245)	(0.139)	(0.205)	(0.277)	(0.154)	(0.212)	(0.266)	(0.148)	(0.208)
Inequality X Third income tertile	0.004	2.020***	2.640***	−0.402	1.335***	1.340***	−0.570	1.465***	1.322***	−0.536	1.305***	1.181***
	(0.382)	(0.300)	(0.394)	(0.306)	(0.174)	(0.204)	(0.349)	(0.191)	(0.209)	(0.336)	(0.181)	(0.210)
Inequality X Income tertile missing	−2.109	−25.320***	−0.405	0.104	−9.283**	−0.554	−0.192	−10.011*	−0.730	−1.962	−9.342**	−0.276
	(2.385)	(8.030)	(1.735)	(2.266)	(4.298)	(0.947)	(2.688)	(5.143)	(0.928)	(2.433)	(4.109)	(0.872)
Observations	469,156	547,326	438,813	469,156	547,326	438,813	469,156	547,326	438,813	469,156	547,326	438,813
Pseudo R2	0.103	0.112	0.085	0.103	0.112	0.085	0.103	0.112	0.086	0.103	0.112	0.085
Year FE	Y	Y	Y	Y	Y	Y	Y	Y	Y	Y	Y	Y
Individual controls	Y	Y	Y	Y	Y	Y	Y	Y	Y	Y	Y	Y
Country-level controls	Y	Y	Y	Y	Y	Y	Y	Y	Y	Y	Y	Y
Region FE	Y	Y	Y	Y	Y	Y	Y	Y	Y	Y	Y	Y
Region X linear trend	Y	Y	Y	Y	Y	Y	Y	Y	Y	Y	Y	Y

Note: The table reports the logit coefficient estimates using robust standard errors clustered at the country×year level. The dependent variable in all models is emigration desires. The key independent variables for inequality are as follows: Models (1)–(3) use the top 1 percent share, Models (4)–(6) use the top 10% share, Models (7)–(9) use the top 20 percent share, and Models (10)–(12) use the Gini coefficient. In Panel A, the results are shown for each inequality variable based on the country income group (low-income, middle-income, or high-income). In Panel B, inequality is interacted with the individual's income position in their own country's income distribution. All regressions include year fixed effects, individual controls (biological sex, age, immigrant status, children, marital status, rural/urban location, education, income group, and unemployment status), region-fixed effects, region×time trend controls, and country-level controls (life satisfaction, corruption, generosity, social support, GDP per capita, life expectancy, and freedom perceptions). All regressions are estimated using the survey weight. See Table 1 for variable definitions.

*** p < 0.01, ** p < 0.05, * p < 0.1.

Table 7.

The Relationship Between Inequality Levels and Emigration Desires, by Level of Economic Development and Education.

	(1)	(2)	(3)
	Low-Income	Middle-Income	High-Income
Top 1% share (lag)	−3.053***	−2.692***	−2.047**
	(0.775)	(0.604)	(1.017)
Secondary education	0.867***	−0.022	−0.234***
	(0.090)	(0.068)	(0.082)
Tertiary education	1.199***	0.017	−0.599***
	(0.134)	(0.083)	(0.099)
Top 1% share (lag) × secondary education	−2.556***	0.715*	2.460***
	(0.499)	(0.416)	(0.573)
Top 1% share (lag) × tertiary education	−4.236***	1.634***	5.990***
	(0.721)	(0.500)	(0.689)
Observations	469,156	547,326	438,813
Pseudo R2	0.108	0.112	0.0852
Year FE	Y	Y	Y
Individual controls	Y	Y	Y
Country-level controls	Y	Y	Y
Region FE	Y	Y	Y
Region × linear time trend	Y	Y	Y

Note: The table reports the logit coefficient estimates using robust standard errors clustered at the country×year level. The dependent variable in all models is emigration desires. The key independent variable is the top 1 percent share in all models. Inequality is interacted with the respondent's level of education, and the results are shown for low-income, middle-income, and high-income countries. All regressions include year fixed effects, individual controls (biological sex, age, immigrant status, children, marital status, rural/urban location, education, income group, and unemployment status), region-fixed effects, region×time trend controls, and country-level controls (life satisfaction, corruption, generosity, social support, GDP per capita, life expectancy, and freedom perceptions). All regressions are estimated using the survey weight. See Table 1 for variable definitions.

*** p < 0.01, ** p < 0.05, * p < 0.1.

In Panel B of Table 5, I interact inequality with an indicator for whether the respondent belongs to the richest third of households in their origin country. The main insights are similar to the findings related to education. Specifically, while those with higher incomes are less likely to want to emigrate in general and inequality is negatively associated with emigration desires, in high-inequality countries, having a high income slightly mitigates the negative consequences of inequality for forming emigration desires. This result is consistent with the explanation of inequality acting as a migration hurdle because belonging to a richer income group (and having richer peers) or being more educated can help cushion.

In addition, I explore whether the main relationship differs by gender (Panel C of Table 5). The results show that women are less likely to want to emigrate than men and high inequality may make women even less likely to have emigration desires. However, the coefficient estimates on the interaction terms are not always statistically significant. This suggests that the constraints that inequality imposes on emigration aspirations are higher for women than men.

Finally, having networks of family and friends abroad can often lower migration costs and encourage more mobility by providing information and job opportunities upon arrival and helping with assimilation (Bertoli and Ruyssen 2018; Massey et al. 1993, 1994). Meanwhile, networks explain over a third of the variation in migration desires in the GWP (Manchin and Orazbayev, 2018). Evidence from the United States shows that annual migrant inflows increase by about five persons if the migrant stock from a particular origin increases by 1,000 people (Clark, Hatton and Williamson 2007). In Panel D of Table 5, I show that having networks abroad can encourage prospective emigration and mitigate the negative consequences of inequality for emigration, which suggests that part of the reason inequality discourages emigration could be through imposing additional migration burdens and structural barriers for individuals. Networks help mitigate, but do not fully offset, these hurdles. All in all, the results in Table 5 strongly imply that inequality acts as a structural barrier that limits the aspirations and potentially the ability of individuals to migrate, providing evidence for the aspirations/(cap)ability hypothesis.

9. Heterogeneity by Level of Economic Development

Tables 6–8 explore the relationship between income inequality and emigration desires by respondents’ characteristics and the level of economic development of their countries as proxied by the country's real GDP per capita. I split the sample into low-income, middle-income, and high-income country groups, as detailed in Supplemental Table B5.

Table 8.

The Relationship Between Inequality Levels and Emigration Desires, by Level of Economic Development, Gender, and Networks.

	(1)	(2)	(3)	(7)	(8)	(9)
Variables	Low-Income	Middle-Income	High-Income	Low-Income	Middle-Income	High-Income
	Variable = Female			Variable = Network
Top 1% share (lag)	−3.684***	−1.836***	0.359	−4.243***	−1.929***	0.495
	(0.723)	(0.567)	(0.916)	(0.732)	(0.534)	(0.990)
Variable	−0.334***	−0.164***	−0.298***	0.078	0.536***	0.280***
	(0.078)	(0.048)	(0.043)	(0.104)	(0.075)	(0.079)
Top 1% share (lag) × variable	0.153	−0.364	1.015***	2.649***	0.990**	2.171***
	(0.469)	(0.256)	(0.331)	(0.594)	(0.410)	(0.567)
Observations	469,156	547,326	438,813	455,520	521,891	374,647
Pseudo R2	0.103	0.112	0.084	0.110	0.125	0.094
Year FE	Y	Y	Y	Y	Y	Y
Individual controls	Y	Y	Y	Y	Y	Y
Country-level controls	Y	Y	Y	Y	Y	Y
Region FE	Y	Y	Y	Y	Y	Y
Region × linear time trend	Y	Y	Y	Y	Y	Y

Note: The table reports the logit coefficient estimates of logit estimations using robust standard errors clustered at the country×year level. The dependent variable in all models is emigration desires. The key independent variable in all models is the top 1 percent share. The results are shown for each inequality variable based on the country income group (low-income, middle-income, or high-income). Inequality is interacted with gender in Models (1)–(3) and with having a network of family and friends abroad in Models (4)–(6). All regressions include year fixed effects, individual controls (biological sex, age, immigrant status, children, marital status, rural/urban location, education, income group, and unemployment status), region-fixed effects, region×time trend controls, and country-level controls (life satisfaction, corruption, generosity, social support, GDP per capita, life expectancy, and freedom perceptions). All regressions are estimated using the survey weight. See Table 1 for variable definitions.

The findings in Panel A of Table 6 indicate that the negative relationship between inequality and emigration desires is strongest in low-income countries. The relationship is negative and significant for middle-income countries as well. In these two country groups, inequality acts as a migration deterrent and imposes a barrier to potential emigration. In high-income countries, however, individuals are more likely to want to emigrate as inequality increases (based on estimates with all inequality measures except the top 1% share), suggesting that there are important differences based on a country's economic development level.

In Panel B of Table 6, I also interact the inequality levels with the respondent's own position in the income distribution and examine how that relationship differs by economic level of development. To my knowledge, such explorations by own economic position and the country's level of development have not been done in the literature on potential emigration and inequality.

In low-income countries (Models 1, 4, 7, and 10), inequality acts as a deterrent for emigration aspirations regardless of the individual's income position, which is evident from the nonstatistically significant coefficient estimates for the interactions between inequality and the respondent's income tertile. In middle-income countries, income has a cushioning effect where it partially offsets the negative influence of inequality on emigration intentions that I documented in Table 5, Panel B. In high-income countries, it is the richest individuals, that is, those in the second and third income tertiles, who are also disproportionately concerned with inequality and express emigration intentions.

These findings highlight that the results in the global sample, related to the cushioning effects of income, are driven by respondents wanting to leave middle-income countries as inequality increase. These results are informative, given that emigrant stocks peak as countries reach a middle-income status (Clemens 2020).

In Tables 7–8, I provide additional heterogeneity results based on respondents’ characteristics and the economic level of development of their countries of residence. Only the results for the top 1 percent share are displayed, as the results with the additional inequality measures reveal similar patterns.

Table 7 reveals that in low-income countries, as inequality increases, the most educated respondents are less likely to want to emigrate, likely because they are most likely to benefit from the high rewards for skill and talent in their home countries that inequality brings. The opposite is true in high-income countries, where, as inequality increases, the high-educated individuals are likely bothered by the inegalitarian societies in which they reside and wish to leave them. As inequality increases in rich countries, educated individuals are more likely to want to emigrate.

Finally, the results in Table 8 show that women in high-income countries are most bothered by rising inequality and would like to move as inequality rises (Model 3). Moreover, as inequality increases, networks matter the most for the emigration intentions of respondents in low-income countries (Model 7).

10. Conclusion

This paper investigates the relationship between economic inequality and two emigration aspirations variables — hypothetical desires and concrete plans, based on respondents living in 150 countries worldwide. The findings suggest that inequality imposes a burden, that is, a structural cost, on potential emigration. The barrier is the largest for respondents living in the poorest countries. For those in rich countries, inequality is, in fact, a motivation to leave, especially among the highly educated. There are varying theoretical explanations that are consistent with these results.

First, the fact that inequality discourages emigration destaires in the global sample, and for respondents living in the poorest countries in particular, is consistent with the aspirations/(cap)ability framework (Carling 2002). For individuals living in low-income countries, who potentially have the most to gain by moving, inequality is a structural barrier to emigration, limiting aspirations and the ability to move. For these respondents, economic inequality may entail unequal opportunities, discrimination, and institutions and social norms that favor access to legal emigration and opportunities for some social groups while excluding others of “rights or compelling them into exploitative situations” (De Haas 2021, p. 6).

Second, holding average income constant, higher inequality entails more poor individuals. Such individuals often lack access to finance to cover the high emigration costs. This lower ability to emigrate can also translate to fewer emigration intentions (aspirations and plans) at the individual level. Even if a particular individual is not liquidity-constrained, the fact that fewer compatriots are emigrating may also discourage this individual from emigrating. Fewer compatriots migrating entails fewer information channels and fewer opportunities for cost-sharing (e.g., through traveling together). This makes emigration more costly and less likely for the individual, independent of income, which can result in “cumulative immobility” (Carling 2002). Inequality may thus impose a migration cost that discourages potential (and actual) emigration. This explanation is consistent with my results that the discouraging effect imposed by inequality is smaller for richer and more educated respondents. It is also in line with the fact that networks mitigate the negative consequences of inequality for migration, especially in low-income countries.

From an economic psychology viewpoint, inequality may also negatively influence emigration aspirations if it symbolizes POUM (Benabou and Ok 2001) and high returns to skill. In other words, individuals may view inequality as being the result of high rewards for talent and effort. Thus, inequality may be negatively associated with emigration intentions if it proxies societal-level rewards for hard work and belief in social mobility and opportunity. This explanation is consistent with the finding in this paper that in low-income countries, as inequality increases, the most educated are least likely to want to move, likely because they believe that the returns to their talent are high in their own country. If inequality reflects returns to skills, high-skilled individuals will have few incentives to migrate to another country, while low-skilled individuals will have higher incentives to migrate (Borjas 1987). This is because less-skilled individuals gain from moving to countries with less income inequality than their own: they can benefit from redistribution and higher wages abroad than in their home countries. At the same time, high-skilled people prefer moving to countries with higher income inequalities than their own because they can earn more abroad.

In rich countries, inequality encourages prospective emigration, especially among women and tertiary-educated and high-income individuals. This suggests that inequality in high-income countries and among these sociodemographic groups symbolizes undesirable societal developments, which respondents wish to escape. Another interpretation of this finding is that structural inequality and the socioeconomic structures that come with it make emigration a desired course of action for those with high education and income while depriving other cohorts of such opportunities (de Haas 2021).

While this paper contributes novel evidence of the relationship between inequality and migration, it has limitations and leaves several opportune avenues for future research. Future work should prioritize understanding whether the patterns identified in this paper hold across time and space. Furthermore, the paper deals with inequality's short-run implications on emigration. Therefore, it does not investigate the long-term consequences of inequality for changing societal, economic, and institutional features and, as such, indirectly affecting emigration, which is another opportune avenue of further inquiry. Moreover, the study only focuses on income and wealth inequality but does not consider inequality of opportunity and other types of inequality, such as inequality in well-being. Of course, inequality is one among several factors influencing potential emigration. In this study, I take economic development, institutions, health and life satisfaction, and social cohesion into account in the analyses, but I specifically focus on inequality. Future studies can expand the analyses to explore whether and how inequality interacts with these other determinants. This study only focuses on the push factors of migration because I do not have “destination-level” information for those who do not wish to migrate. Future research can attempt to integrate the push and pull factors of migration into a single framework. Finally, I do not consider temporary vs. permanent migration nor distinguish specifically between economic migrants and other types of migrants. Further data collection efforts and data sets can help shed light on these important distinctions.

Supplemental Material

sj-docx-1-mrx-10.1177_01979183231202991 - Supplemental material for The Relationship Between Inequality and Potential Emigration: Evidence from the Gallup World Poll

Supplemental material, sj-docx-1-mrx-10.1177_01979183231202991 for The Relationship Between Inequality and Potential Emigration: Evidence from the Gallup World Poll by Milena Nikolova in International Migration Review