## wpiea2024097-print-pdf - References

## Source details

**Canonical URL:** [wpiea2024097-print-pdf - References](https://www.imf.org/-/media/files/publications/wp/2024/english/wpiea2024097-print-pdf.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2024/english/wpiea2024097-print-pdf.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2024/english/wpiea2024097-print-pdf.pdf.json)

---

### I. Introduction — scope and key contributions
- Scope:
  - Quantitative assessment of drivers of migration within Africa using data for 52 African countries during the period 1990-2020.
- Context and motivation:
  - More than half (53 percent) of Africa’s migrants in 2017 lived in another African country; for sub-Saharan Africa the intra-continental share increased to more than 80 percent.
- Key contributions and findings:
  - Demographic factors (population size and structure of the home country) matter for intra-African migration.
  - Political risk and ethnic tensions in the home country are relevant push factors.
  - Income level of the destination country is an important pull factor.
  - Climate shocks—including natural disasters, precipitation, and temperature—are important drivers.
  - Distance between home and destination countries, and human capital development in the home country, are also important.
- Policy implications (high level):
  - Strengthen institutions to reduce political instability and ethnic tensions.
  - Invest in education and health.
  - Intensify efforts to combat climate change.

### II. Literature — thematic overview
- Push and pull factor categories summarized:
  - Economic drivers: income level, employment, poverty, distance, macroeconomic conditions (Mayda, 2005; Berthelemy et al., 2007).
  - Political factors: conflicts, political instability, terrorism as push factors (Moore and Shellman, 2004; Dreher et al., 2011).
  - Institutional drivers: good institutions attract migrants; corruption encourages emigration (Bertocchi and Strozzi, 2008; Poprawe, 2015).
  - Climate factors: unsettled relationship between climate change and migration; climatic shocks may trigger, mitigate, or have no impact on migration (Abel et al., 2019; Berlemann and Steinhardt, 2017).
- Prior intra-African migration findings:
  - Economic and socio-political factors in host country and migration networks matter (Ruyssen and Rayp, 2014; Flahaux and De Haas, 2016).

### III. Stylized facts on intra-African migration (key statistics)
- About half of African migrants, or 17.8 million migrants, remained in Africa in 2020.
  - Caveat: figure likely underestimated due to incomplete UN coverage and limited national tracking.
- Regional patterns:
  - Bulk of intra-African migration takes place in sub-Saharan Africa; in the 1990s eight out of ten migrants from SSA stayed in another SSA country.
  - SSA continues to retain almost two-thirds of its migrants.
  - Less than two percent of SSA migrants travel to North Africa.
  - Intra-regional migration more intense in West Africa (visa-free ECOWAS travel; spread of ethnic groups).
  - Northern Africa: 98 percent of migrants leave the continent.

### IV. Data, variables, and empirical model
- Data sources and sample:
  - Main bilateral migration data: United Nations Department of Economic and Social Affairs (Global Migration Database, International Migrant Stock 2020).
  - Additional: World Development Indicators (WDI), CEPII GeoDist, UNDP HDI, International Country Risk Guide (ICRG), Emergency Events Database.
  - Time coverage: five-year data periods spanning from 1990 to 2020.
  - Country sample size: 52 African countries.
- Key variables (as specified in model):
  - Dependent variable: M_ijt (stock of migrants, bilateral, by year).
  - Demographic: log(pop_i t) (total population in millions), Rural_i t, Rural_j t.
  - Geography: log(dist_ij) (log distance between most populated cities, km), Contig_ij.
  - Culture: ComLang_ij (common language spoken by ≥9 percent), OffLang_ij.
  - Economic/human development: log(GDP_i t), log(GDP_j t) (current thousands of US$), hdi_i t, hdi_j t, life expectancy, expected years of schooling.
  - Politics: EthnicTension_i t, PoliticalRisk_i t (ICRG indices).
  - Climate: Climate_i t vector (total disasters; share affected; five-year average temperature and precipitation; deviations; anomalies using 1950 long-run stats).
  - Fixed effects: u_i, u_j; time dummies λ_t.
- Estimation approach:
  - Poisson pseudo maximum likelihood estimator (Santos Silva and Tenreyro, 2010).
  - Full sample and regional subsamples (North Africa excluded from regional estimation due to small flows with SSA).

### V. Correlation analysis and baseline econometric results (highlights)
- Unconditional correlations:
  - Negative co-movement: migration with political factors; migration with human development and components.
  - Positive co-movement: migration with destination GDP; migration with number of disasters, incidence of natural disaster, average temperature.
  - Geographic proximity matters: distance negatively correlated; contiguity positively correlated.
  - High co-movement (≥ 0.7) between political risk rating and ethnic tension; between HDI and its components—these pairs not included together to limit multicollinearity.
- Baseline Poisson regression (selected significant coefficients, robust s.e. in parentheses):
  - Population (origin):
    - (1) 0.016*** (0.006)
    - (2) 0.012*** (0.004)
    - (3) 0.015*** (0.006)
    - (4) 0.010** (0.004)
  - Rural population (destination):
    - (1) -0.048*** (0.015)
    - (2) -0.050*** (0.015)
    - (3) -0.041*** (0.014)
    - (4) -0.044*** (0.014)
  - Political risk rating (origin):
    - (1) -0.028*** (0.009)
    - (2) -0.019** (0.008)
  - Ethnic tensions (origin):
    - (3) -0.352*** (0.096)
    - (4) -0.275*** (0.092)
  - Human development index (origin):
    - (1) -6.308*** (2.075)
    - (3) -7.717*** (1.767)
  - Log GDP (destination):
    - (2) 0.192* (0.106)
    - (4) 0.239** (0.107)
  - Log distance (most populated cities):
    - (1) -0.466*** (0.067)
    - (2) -0.480*** (0.071)
    - (3) -0.465*** (0.067)
    - (4) -0.478*** (0.070)
  - Contiguity:
    - (1) 1.877*** (0.094)
    - (2) 1.890*** (0.091)
    - (3) 1.876*** (0.093)
    - (4) 1.884*** (0.091)
  - Common language:
    - (1) 0.757*** (0.115)
    - (2) 0.709*** (0.117)
    - (3) 0.754*** (0.114)
    - (4) 0.705*** (0.116)
- Model fit and sample size:
  - Observations: (1) 3,040; (2) 3,240; (3) 3,040; (4) 3,210.
  - R-squared: (1) 0.81; (2) 0.83; (3) 0.82; (4) 0.84.

### VI. Regional baseline heterogeneity (selected results)
- West Africa (Observations: 1,301; R-squared: 0.90):
  - Log distance: -1.114*** (0.122)
  - Contiguity: 1.258*** (0.115)
  - Common language: 0.749*** (0.159)
- East Africa (Observations ~737–741; R-squared: 0.64–0.74):
  - Log distance: -0.537*** (0.152)
  - Contiguity: 1.969*** (0.193)
  - Common language: 1.181*** (0.254)
- Central Africa (Observations: 498; R-squared: 0.88):
  - Log distance: -0.181*** (0.070)
  - Contiguity: 2.282*** (0.221)
  - Common language: -0.624* (0.373)

### VII. Climate, disasters, landlockedness, and income-group heterogeneities
- Natural disaster incidence (share of population affected) — positive association with outward migration:
  - Table 4, Panel A incidence (origin): 1.368** (col 5); 0.629 (col 6); 0.988* (col 7); 0.761* (col 8).
  - Table 5, Panel A (landlocked): Incidence (origin) 4.514*** (col 1); 5.423*** (col 2); 4.260*** (col 3); 5.077*** (col 4).
  - Interaction Incidence*Landlock: -4.264** (col 1); -5.650*** (col 2); -4.373*** (col 3); -5.113*** (col 4).
- Number of disasters: generally small or not significant; interaction with landlock often positive and sometimes significant:
  - # of disasters*Landlock: 0.038* (col 1); 0.063*** (col 2); 0.044** (col 3); 0.065*** (col 4).
- Precipitation effects (Table 4, Panel B):
  - Precipitation anomaly positive: 0.238** (col 9); 0.209** (col 10); 0.207** (col 11); 0.135 (col 12).
  - Precipitation level / deviation sometimes positive (e.g., Precipitation level 0.020* in some columns; Precipitation deviation 0.020*).
- Temperature effects and landlock interactions (Table 5, Panel B):
  - Temperature level*Landlock positive and significant in multiple columns: e.g., 1.000** (col 2); 1.323*** (col 4); 0.856** (col 6); 1.179*** (col 8).
  - Temperature deviation*Landlock positive and significant: e.g., 1.002** (col 4); 1.336*** (col 6).
  - Temperature anomaly*Landlock positive and sometimes significant: e.g., 0.384* (col 9); 0.476*** (col 10).
- Income-group heterogeneity:
  - Climate/disaster effects stronger for low-income countries (LICs) versus middle-income countries (MICs).
  - Precipitation variables more robust in LICs; temperature variables have greater impact in MICs.

### VIII. Commodity dependence and other controls
- Commodity dependence of destination (Table A.7) — evidence that energy and mining exporters attract more migrants in some specifications:
  - Energy (destination): 3.096*** (col 1); -0.890 (col 2); 2.627*** (col 3); -1.073 (col 4).
  - Mining (destination): 1.507** (col 1); 4.106*** (col 2); 1.240** (col 3); 3.703*** (col 4).
- Conflict and institutional quality checks:
  - Presence of conflict (dummy) contributes to migration; fatalities less consistently predictive.
  - Institutional quality controls (Political Stability and Absence of Violence, Rule of Law) not statistically significant once risk factors are included—strong co-movement noted between risk factors and institutional quality.

### IX. Endogeneity strategy and robustness checks
- Instrumenting approach:
  - Three instrument sets: (i) lagged five-year average (e.g., 2015-19 for 2020), (ii) alternative lagged five-year average (e.g., 2011-15), (iii) value at beginning of period (e.g., GDP in 2016 for 2020).
  - Appendix Table A.6 results broadly in line with main results.
- Additional robustness checks:
  - Include commodity exporter status, presence of conflicts, quality of institutions, alternative risk measures (state fragility index, World Uncertainty Index).
  - Main results robust to these additions where sample-size reductions permit.

### X. Interpretation, distributional concerns, and policy recommendations
- Interpretation of empirical findings (summary):
  - Geographic proximity (contiguity, shorter distance), common local language, higher destination GDP, and demographic size drive intra-African migration.
  - Political instability and ethnic tensions in origin, and lower human development in origin, push migration.
  - Climate shocks (incidence, precipitation anomalies, temperature deviations) are associated with higher outward migration, especially from landlocked and low-income origins.
- Policy recommendations:
  - Facilitate intra-African migration to capture benefits: better allocation of production factors, higher productivity, deeper trade integration.
  - Strengthen institutions to reduce political instability and ethnic tensions to reduce forced migration.
  - Invest in education and health to reduce incentives to migrate and improve migrant outcomes.
  - Improve regional infrastructure and trade integration to reduce migration costs.
  - Mitigate climate change and manage climate shocks to contain “climate refugees”.
- Note on distributional effects and future research:
  - Migration can cause brain drain in origin countries and pressures on public services in host countries; future research needed to quantify net benefits and costs.

### XI. References — thematic clusters (selected)
- Climate/environment and migration: Abel et al., 2019; Berlemann and Steinhardt, 2017; Cattaneo and Peri, 2016; Kaczan and Orgill-Meyer, 2020; El-Hinnawi, 1985.
- Institutions, conflict, and migration: Bertocchi and Strozzi, 2008; Cooray and Schneider, 2016; Dimant et al., 2013; Dreher et al., 2011; Davies et al., 2023.
- Migration theory, selection, and labor: Borjas, 1987; Borjas, 1999; Lee, 1966; De Haas, 2010; Bodvarsson et al., 2015.
- Empirical methods and datasets: Santos Silva and Tenreyro, 2010; Mayer and Zignago, 2011 (CEPII GeoDist); UNDESA International Migrant Stock 2020; Solt, 2020 (SWIID).

*Source: Content unit "wpiea2024097-print-pdf - References" (text excerpts and tables/figures listing provided).*

### References .............................................................................................................

### wpiea2024097-print-pdf - References

### I. Introduction
- Focus: quantitative assessment of the drivers of migration within Africa using data for 52 African countries during the period 1990-2020.
- Context and motivation:
  - Existing studies focus predominantly on south-north migration and within-country (rural-to-urban) migration.
  - South-south migration, particularly intra-African migration, is extensive: more than half (53 percent) of Africa’s migrants in 2017 lived in another African country (Lucas, 2015; UNCTAD, 2018).
  - For sub-Saharan Africa, the intra-continental share increased substantially to more than 80 percent.
  - Understanding drivers is important for: (i) realizing benefits from deeper trade integration and labor mobility (African Union Protocol on free movement and AfCFTA), and (ii) boosting growth and structural transformation in destination countries through higher labor productivity (UNCTAD, 2018).
- Key contributions and findings summarized:
  - Demographic factors (population size and structure of the home country) matter for intra-African migration.
  - Political risk and ethnic tensions in the home country are relevant push factors.
  - Income level of the destination country is an important pull factor.
  - Climate shocks—including natural disasters, precipitation, and temperature—are important drivers.
  - Distance between home and destination countries, and human capital development in the home country, are also important.
- Policy implications highlighted:
  - Strengthen institutions to reduce political instability and ethnic tensions to reduce “forced” migration.
  - Invest in education and health to reduce incentives to migrate and improve outcomes for migrants.
  - Intensify efforts to combat climate change to contain “climate refugees”.
- Paper structure (as provided):
  - Section II: literature review
  - Section III: migration data and stylized facts
  - Section IV: key drivers based on principal component analysis, data and empirical model
  - Section V: econometric results
  - Section VI: conclusion

### II. The literature
- Overview of push and pull factor categories:
  - Economic drivers:
    - Income level, employment level, poverty, distance, and macroeconomic conditions.
    - Mayda (2005): mean income in host countries has robust pull effects.
    - Berthelemy et al. (2007): increase in income per capita above a threshold in departure country pushes migration; foreign assistance from destination countries to departure countries does not dampen immigration.
  - Political factors:
    - Conflicts and political instability can lead to voluntary and “forced” migration (Moore and Shellman, 2004).
    - Violence has a larger impact on migration than political institutions or average size of the economy.
    - Terrorism is a push factor (Dreher et al., 2011); nationalism and political factors affect migration when people’s well-being is affected (Radnitz, 2006).
  - Institutional drivers:
    - Good institutions attract more migrants (Bertocchi and Strozzi, 2008).
    - Dimant et al. (2013): institutions boost migration alongside population, common language, and common border.
    - Corruption encourages emigration and discourages immigration (Poprawe, 2015); effect stronger for non-high skilled workers (Cooray and Schneider, 2016).
  - Climate factors:
    - Relationship between climate change and migration is unsettled (Abel et al., 2019).
    - Climatic shocks may trigger, mitigate, or have no impact on migration depending on shock nature and region.
    - Climate-related migration is often thought to lead to within-country migration rather than international migration.
    - Evidence of environmental drivers: migration to access new land (Geist and Lambdin, 2001); concept of environmental refugees from environmental deterioration and demographics (El-Hinnawi, 1985).
- Prior intra-African migration studies:
  - Ruyssen and Rayp (2014): economic and socio-political factors in host country are main drivers; migration infrastructure and networks matter.
  - Flahaux and De Haas (2016): marginal, poorer, or landlocked countries have lower extra-continental migration and direct migration primarily toward other African countries; countries with higher extra-continental migration have higher GDP per capita and are further along demographic transition; intra-African migration increasing in absolute terms though declining in intensity.

### III. The extent of intra-African migration: stylized facts
- Key statistics and observations:
  - About half of African migrants, or 17.8 million migrants, remained in Africa in 2020.
    - Note: figure likely underestimated because not all African countries are covered by the United Nations dataset and many African countries do not track migration flows.
  - Bulk of intra-African migration takes place in sub-Saharan Africa.
  - In the 1990s, eight out of ten migrants from sub-Saharan Africa (SSA) stayed in another SSA country.
  - Although this proportion has been trending down over the last three decades, SSA countries continue to retain almost two-thirds of its migrants.
  - Less than two percent of SSA migrants travel to North Africa.
  - Regional patterns:
    - Intra-regional migration is more intense in West Africa, attributed to visa-free travel between ECOWAS member countries and spread of ethnic groups across West-African countries.
    - Northern Africa: 98 percent of migrants leave the continent—attributed to proximity to Europe, colonial/post-colonial ties to France, and labor recruitment agreements with European countries since 1960.
- Data source noted for figures:
  - Raw data from the United Nations Department of Economic and Social Affairs, Population Division (2020), International Migrant Stock 2020, authors’ calculations.
  - Regional breakdown: central, eastern, northern, southern, western (Appendix Table A.1). ROW stands for rest of the world.

### IV. Data and empirical model
- Data sources and scope:
  - Main bilateral migration data: United Nations Department of Economic and Social Affairs (Global Migration Database).
  - Dependent variable: total number of migrants from African countries to other African countries.
  - Time coverage: five-year data periods spanning from 1990 to 2020.
  - Country sample size: 52 African countries.
  - Additional data sources: World Development Indicators (WDI) for population and GDP; CEPII gravity database for geographic variables; UNDP for Human Development Index and components; International Country Risk Guide (ICRG) for political risk and ethnic tensions; Emergency Events Database for climate data.
- Variable categories and measurement (Table 1 summary):
  - Demographic:
    - Population size (origin), share of rural population (origin and destination).
    - Expectation: migration increases with population size; high rural share in destination may be associated with lower cross-border migration.
  - Geography:
    - Distance: defined as distance between the two most populated cities of origin and destination countries.
    - Contiguity: dummy = 1 if destination and origin share a common border, 0 otherwise.
    - Expectation: migration increases with contiguity and decays with distance.
  - Culture:
    - Common language spoken by at least nine percent of the population.
    - Common official or primary language.
    - Expectation: common language increases immigrant stocks.
  - Economic and human development:
    - Log of GDP (origin and destination).
    - Human Development Index and sub-components (life expectancy at birth, expected years of schooling).
    - Expectation: higher development in destination increases immigration; higher development in origin decreases emigration.
  - Politics:
    - Political risk rating and ethnic tensions indices from ICRG.
    - Political instability, conflicts, and disruption of public goods provision push migration.
  - Climate variables (wide range included):
    - Total number of disasters.
    - Share of the population affected by natural disasters (incidence).
    - Five-year average temperature and precipitation.
    - Temperature (precipitation) variability: difference between five-year average and long-run average.
    - Temperature (precipitation) anomaly: temperature (precipitation) deviation divided by the long-run standard deviation.
    - Source: Emergency Events Database.
- Modeling and empirical approach notes:
  - Focus on push and pull factors related to demographic, geographic, cultural, economic, political, and climate characteristics of origin and destination.
  - Principal component analysis used to discuss key drivers (Section IV indicates PCA will be applied).
  - Concern about measurement error in migration data noted; authors prefer UN harmonized dataset and argue measurement error is less concerning because migration is the dependent variable (errors likely captured in residual).
  - Five-year average macroeconomic explanatory variables (lagged) and other robustness and heterogeneity tests are conducted (implied by table listings in the document: Tables 2–5 and Appendix tables A.1–A.9).
- Table 1 (migration pull and push factors) reproduced as thematic mapping (as presented in source):
  - Pull factors:
    - Demographic: Rural population; Population size.
    - Geography: Distance; Common border.
    - Culture: Common language.
    - Economic and human development: Prospect of higher income; Potential for improved standard of living.
    - Politics: Safety and security; Political freedom.
    - Climate: (not explicitly repeated under pull in table but climate variables are part of model).
  - Push factors:
    - Demographic: Rural population.
    - Economic and human development: Poverty; Unemployment; Lack of basic health and education.
    - Politics: Conflict, security, and violence; Poor governance and corruption.
    - Climate: Total natural disasters; Population affected by natural disasters; Level and deviation of temperature and precipitation.

*Source: Content unit "wpiea2024097-print-pdf - References" (text excerpts and tables/figures listing provided).*

### 4.2 Correlation analysis

### 4.2 Correlation analysis

### Correlation findings (unconditional)
- Negative co-movement between migration and political factors (consistent with literature on conflict and migration).
- Negative co-movement between migration and human development (and its components).
- Positive correlation between migration and economic growth in the destination country (measured by GDP), suggesting economic opportunity in the destination country as a pull factor.
- Negative correlation between distance and migration; positive co-movement between contiguity and migration, indicating geographic proximity matters.
- Positive co-movement between migration and:
  - number of disasters,
  - incidence of natural disaster,
  - average temperature.
- High co-movement (correlation ≥ 0.7) observed between:
  - political risk rating and ethnic tension,
  - human development index and its components.
  - These pairs are not included together in the same regressions to limit multicollinearity.

### Empirical model (specification)
- Dependent variable: M_ijt (stock of migrants, bilateral, by year).
- Full specification (variables listed exactly as in the source):
  - log(pop_i t) = total population (in millions) in the country of origin
  - Rural_i t and Rural_j t = share of the rural population in origin and destination
  - log(dist_ij) = log distance between most populated cities (km)
  - Contig_ij = 1 if share a common border, 0 otherwise
  - ComLang_ij = 1 if both countries share a common language
  - OffLang_ij = 1 if both countries share common official or primary language
  - log(GDP_i t) and log(GDP_j t) = gross domestic products (current thousands of US$) of origin and destination
  - hdi_i t and hdi_j t = human development indices of origin and destination
  - EthnicTension_i t = degree of tension in origin attributable to racial, nationality, or language divisions
  - PoliticalRisk_i t = political risk rating assessing political stability of origin
  - Climate_i t = vector of climate-related factors (long run average (standard deviation of) temperature (precipitation) is the average (standard deviation of) temperature (precipitation) from 1950)
  - u_i and u_j = origin and destination country fixed effects
  - λ_t = time dummy
  - ε_ijt = error term
- Estimator: Poisson pseudo maximum likelihood estimator (Santos Silva and Tenreyro, 2010) to handle count data and zeros.
- Sample: full sample and by region (North Africa excluded from regional estimation due to small flows with sub-Saharan Africa).

### Baseline econometric results (summary of Table 2)
- General significant determinants (baseline regressions; dependent variable: stock of migrants; robust standard errors in parentheses; significance: *** p<0.01, ** p<0.05, * p<0.1):
  - Population (total, origin): positive and significant across specifications:
    - (1) 0.016*** (0.006)
    - (2) 0.012*** (0.004)
    - (3) 0.015*** (0.006)
    - (4) 0.010** (0.004)
  - Rural population (share, origin): mixed/weak negative:
    - (1) -0.028* (0.016)
    - (2) -0.007 (0.019)
    - (3) -0.016 (0.016)
    - (4) -0.003 (0.017)
  - Rural population (share, destination): consistently negative and significant:
    - (1) -0.048*** (0.015)
    - (2) -0.050*** (0.015)
    - (3) -0.041*** (0.014)
    - (4) -0.044*** (0.014)
  - Political risk rating (origin): negative and significant where included:
    - (1) -0.028*** (0.009)
    - (2) -0.019** (0.008)
  - Ethnic tensions (origin): negative and significant where included:
    - (3) -0.352*** (0.096)
    - (4) -0.275*** (0.092)
  - Human development index (origin): negative and significant where included:
    - (1) -6.308*** (2.075)
    - (3) -7.717*** (1.767)
  - Log GDP (origin): negative and weakly significant where included:
    - (2) -0.239* (0.139)
    - (4) -0.212* (0.128)
  - Log GDP (destination): positive and weakly/significantly positive:
    - (2) 0.192* (0.106)
    - (4) 0.239** (0.107)
  - Log distance (most populated cities): negative and highly significant:
    - (1) -0.466*** (0.067)
    - (2) -0.480*** (0.071)
    - (3) -0.465*** (0.067)
    - (4) -0.478*** (0.070)
  - Contiguity: large positive and highly significant:
    - (1) 1.877*** (0.094)
    - (2) 1.890*** (0.091)
    - (3) 1.876*** (0.093)
    - (4) 1.884*** (0.091)
  - Common language: positive and highly significant:
    - (1) 0.757*** (0.115)
    - (2) 0.709*** (0.117)
    - (3) 0.754*** (0.114)
    - (4) 0.705*** (0.116)
  - Common official or primary language: not significant at conventional levels:
    - (1) -0.017 (0.117)
    - (2) 0.040 (0.120)
    - (3) -0.014 (0.118)
    - (4) 0.038 (0.120)
- Model fit and sample:
  - Observations: (1) 3,040; (2) 3,240; (3) 3,040; (4) 3,210
  - R-squared: (1) 0.81; (2) 0.83; (3) 0.82; (4) 0.84

### Regional baseline results (summary of Table 3)
- West Africa (columns (5), (6)):
  - Population (total, origin): 0.011* (0.006) and 0.011* (0.006)
  - Rural population (share, origin): -0.038* (0.021) and -0.037* (0.022)
  - Rural population (share, destination): -0.052** (0.026) and -0.053** (0.026)
  - Log distance: -1.114*** (0.122) and -1.118*** (0.126)
  - Contiguity: 1.258*** (0.115) and 1.260*** (0.115)
  - Common language: 0.749*** (0.159) and 0.754*** (0.155)
  - Common official or primary language: 0.293* (0.150) and 0.286* (0.147)
  - Observations: 1,301; R-squared: 0.90
- East Africa (columns (7), (8)):
  - Population (total, origin): 0.031* (0.017) and 0.023 (0.015)
  - Rural population (share, origin): 0.065* (0.037) and 0.079** (0.036)
  - Rural population (share, destination): -0.073** (0.029) and -0.058*** (0.022)
  - Log distance: -0.537*** (0.152) and -0.548*** (0.153)
  - Contiguity: 1.969*** (0.193) and 1.965*** (0.192)
  - Common language: 1.181*** (0.254) and 1.174*** (0.254)
  - Observations: 737 and 741; R-squared: 0.64 and 0.74
- Central Africa (columns (9), (10)):
  - Population (total, origin): 0.016 (0.014) and 0.011 (0.012)
  - Rural population (share, origin): 0.096 (0.087) and 0.111 (0.086)
  - Rural population (share, destination): -0.010 (0.021) and -0.010 (0.021)
  - Log distance: -0.181*** (0.070) and -0.181** (0.071)
  - Contiguity: 2.282*** (0.221) and 2.280*** (0.221)
  - Common language: -0.624* (0.373) and -0.624* (0.375)
  - Observations: 498; R-squared: 0.88

### Interpretation of baseline results (key points)
- Geographic proximity: migration decreases with distance and increases with contiguity; migration mostly occurs between neighboring countries.
- Language: a common language (native/local) significantly increases migrant stock; common official language is not significant.
- Economic pull: migrants move toward destinations with higher GDP (positive relationship between migration and GDP in destination).
- Socio-demographic structure:
  - Rural population (destination) negatively correlated with migration (less diversified economy and fewer opportunities).
  - Rural population (origin) negatively associated with migration (difficulty to migrate directly from rural areas).
- Political factors: greater stability in the home country (higher political risk rating implies lower emigration) and lower ethnic tensions are associated with lower emigration.
- Human development: higher human development in origin is negatively associated with emigration; life expectancy (health care quality proxy) is particularly relevant.

### Density, inequality, and migration
- Population density:
  - Migrants move from less populated to more densely populated countries.
  - Negative effects of rural population and density in departure countries reinforce lack of economic opportunities.
  - High population density in destination associated with concentration of economic activity and pull of migrants.
- Inequality:
  - Gini index not found to have a significant effect on the number of migrants (use of Gini reduces sample size).
  - Human development index and sub-components capture mechanisms associated with underdevelopment and skilled emigration.

### Climate-related factors and migration (summary of results)
- Overall finding: significant relationship between climate-related factors and migration; several regressions and landlocked-country focus reported.
- Natural disaster incidence:
  - Greater share of people affected by natural disaster (incidence) in the departure country is positively associated with higher migration (Table 4, Panel A).
  - Alternative using log of total number affected shows similar positive correlation but less robust.
- Precipitation:
  - Precipitation level and deviation from the long-run average are positively associated with migration (Table 4, Panel B).
  - Significant deviations from long-run averages increase likelihood of droughts or floods, pushing migration.
- Landlocked countries (Tables 5, Panels A and B):
  - Africa has 16 of the world’s 44 landlocked countries.
  - Higher occurrence of natural disaster in landlocked countries is positively associated with more migration.
  - Being landlocked alone is positive but not significant at conventional levels; when hit by natural disaster, landlock status becomes a push factor.
  - Temperature variables in landlocked countries: deviations from long-run average and abnormal temperatures are associated with more migration.
- Income-level heterogeneity:
  - Analysis by World Bank income groups (LICs vs MICs) confirms previous results (Appendix Table A.5, Panels A and B).
  - Precipitation variables more robustly associated with migration in LICs.
  - Temperature variables have greater impact on migration in MICs.
  - Explanation: rainfall effects on (rainfed) agriculture matter more in LICs; temperature effects on agricultural productivity and income more relevant in MICs.

### Endogeneity strategy and robustness checks
- Endogeneity addressed by instrumenting key explanatory variables (socio-demographic, economic, risk factors) with three sets of instruments:
  - (i) lagged five-year average (e.g., GDP in 2020 instrumented by average over 2015-19);
  - (ii) alternative lagged five-year average (e.g., GDP in 2020 instrumented by average over 2011-15);
  - (iii) value of the variable at the beginning of the period (e.g., GDP in 2020 instrumented by GDP in 2016).
  - Results reported in Appendix Table A.6 are broadly in line with main results.
- Additional robustness checks:
  - Include commodity exporter status of destination (energy and mining dummies defined as: energy = country exports mineral fuels, lubricants and related materials; mining = country exports minerals, ores and metals).
  - Include presence of conflicts in origin and quality of institutions.
  - Tests of other risk measures:
    - State fragility index and World Uncertainty Index tested; not significant at conventional levels and reduce sample size.
    - Political risk rating and ethnic tensions added at destination; main results hold.
- Additional results (density and Gini) available upon request from authors.

*Source: wpiea2024097-print-pdf - 4.2 Correlation analysis*

### Appendix Table A.7) suggest that energy and mining producers attract more migrants

### Appendix Table A.7) suggest that energy and mining producers attract more migrants

### Key empirical findings: climate, disasters, and migration
- Incidence of natural disasters (share of population affected) is positively associated with outward migration:
  - Table 4, Panel A: Incidence (origin) 1.368** (column 5); 0.629 (column 6); 0.988* (column 7); 0.761* (column 8).
  - Table 5, Panel A (landlocked countries): Incidence (origin) 4.514*** (column 1); 5.423*** (column 2); 4.260*** (column 3); 5.077*** (column 4).
  - Interaction with landlock: Incidence*Landlock -4.264** (column 1); -5.650*** (column 2); -4.373*** (column 3); -5.113*** (column 4).
- Number of disasters effects:
  - Table 4, Panel A: # of disasters (origin) coefficients are small and not statistically significant in most specifications (e.g., 0.013, -0.010, 0.013, -0.005).
  - Table 5, Panel A (landlocked): # of disasters (origin) -0.002 (col 1); -0.031** (col 2); -0.005 (col 3); -0.027** (col 4).
  - Interaction with landlock: # of disasters*Landlock 0.038* (col 1); 0.063*** (col 2); 0.044** (col 3); 0.065*** (col 4).
- Temperature and precipitation effects (Table 4, Panel B):
  - Precipitation anomaly is positively associated with migration: Precipitation anomaly 0.238** (col 9); 0.209** (col 10); 0.207** (col 11); 0.135 (col 12).
  - Temperature anomaly coefficients are positive but smaller and often not significant: Temperature anomaly 0.095 (col 9); -0.001 (col 10); 0.170 (col 11); 0.025 (col 12).
  - Precipitation level and precipitation deviation show positive coefficients at times (e.g., Precipitation level 0.020* in some columns; Precipitation deviation 0.020*).
- Landlocked heterogeneities for temperature/precipitation (Table 5, Panel B):
  - Temperature level negative for non-interacted origin but strong positive interaction with landlock:
    - Temperature level -0.364 (col 1); -0.746* (col 2); Temperature level*Landlock 1.000** (col 2); 1.323*** (col 4); 0.856** (col 6); 1.179*** (col 8).
  - Temperature deviation*Landlock shows similar positive interactions: Temperature deviation*Landlock 1.002** (col 4); 1.336*** (col 6); 0.857** (col 8); 1.190*** (col 10).
  - Temperature anomaly*Landlock positive and sometimes significant: Temperature anomaly*Landlock 0.384* (col 9); 0.476*** (col 10); 0.343** (col 11); 0.437*** (col 12).
  - Precipitation anomaly has positive coefficients for origin: Precipitation anomaly 0.268** (col 9); 0.167 (col 10); 0.221** (col 11); 0.109 (col 12), while Precipitation anomaly*Landlock is negative and not significant in these columns.

### Political, socioeconomic, demographic, and geographic determinants
- Demographic and geographic factors:
  - Population (total, origin) consistently positive and significant: e.g., 0.016*** (Table 4, Panel A col 1); 0.016*** (Table 4, Panel B col 1); 0.017*** (Table 5, Panel A col 1).
  - Rural population (share, destination) negatively associated with migration across tables: e.g., -0.049*** (Table 4, Panel A col 1) and similar magnitudes across specifications.
  - Log distance (most populated cities) strongly negative and highly significant across specifications: examples include -0.463***, -0.480***, -0.467***, -0.481*** in multiple columns.
  - Contiguity strongly positive and highly significant: e.g., 1.879***, 1.892***, 1.875*** across Table 4 columns.
  - Common language (spoken by at least 9% of population) positive and significant: e.g., 0.757***, 0.709***, 0.756*** across Table 4.
- Political risk and ethnic tensions:
  - Political risk rating (origin) negative and significant in many specifications (higher rating = more stability): e.g., -0.029*** (Table 4, Panel A col 1); -0.026*** (Table 4, Panel B col 1).
  - Ethnic tensions (origin) large negative coefficients where included: e.g., -0.362*** (Table 4, Panel A col 3); -0.337*** (Table 4, Panel B col 3).
- Human development and life expectancy:
  - Human development index (origin) strongly negative and significant in many specifications: e.g., -6.628*** (Table 4, Panel A col 1); -6.871*** (Table 4, Panel B col 1).
  - Life expectancy at birth (origin) negative and sometimes significant (e.g., -0.037** in some columns).

### Commodity dependence and migration (Table A.7)
- Commodity dependence of destination matters, with energy and mining exporters attracting more migrants in some specifications:
  - Commodity Dependence: Energy (destination) 3.096*** (column 1); -0.890 (column 2); 2.627*** (column 3); -1.073 (column 4).
  - Commodity Dependence: Mining (destination) 1.507** (column 1); 4.106*** (column 2); 1.240** (column 3); 3.703*** (column 4).
- Model fit and sample:
  - Observations across columns: 3,040; 3,237; 3,040; 3,208.
  - R-squared values: 0.810; 0.829; 0.821; 0.842.

### Robustness and additional checks
- Conflict indicators and institutional quality:
  - Using a dummy “Conflict” (state-based conflict over the last five years) and average number of fatalities (data from Davies et al. (2023) and Sundberg et al. (2013)) shows that the presence of conflict leads to migration and does not invalidate main risk-factor results. The authors note that the presence of conflict seems to play a more important role in migration decisions than the number of fatalities.
  - Institutional quality controls (Political Stability and Absence of Violence, Rule of Law) have coefficients that are not statistically significant in baseline regressions (see Table A.10), implying institutional quality variables do not matter in the migration decision once risk factors are accounted for. The authors note strong co-movement between risk factors and institutional quality and therefore avoid including them together in the same regression (see Table A.9 and Table A.10).
- Heterogeneity by income group and landlockedness:
  - Climate and disaster effects are stronger for low-income countries and for migration originating from landlocked countries (see Table 5 and heterogeneity panels in Tables A.5 and A.6).
  - Precipitation anomaly and precipitation deviation effects are particularly robust in low-income country specifications (e.g., Panel B, Table A.5 shows Precipitation anomaly 0.471*** for LICs in some columns).

### Policy implications and recommendations (conclusions from the paper)
- Facilitate intra-African migration to capture potential benefits:
  - Better allocation of production factors, higher productivity, deeper trade integration, and higher economic growth are potential benefits from intra-African migration.
- Reduce forced migration by addressing political and ethnic risks:
  - Strengthening institutions to curtail political instability and ethnic tensions to reduce forced migration.
- Invest in human capital:
  - Investing in education and health may reduce incentives for migration but improves migrants’ chances to find skilled jobs in destination countries, boosting productivity and remittances.
- Improve regional infrastructure and trade integration:
  - Better regional infrastructure would reduce migration costs and strengthen regional trade integration.
- Mitigate climate change and manage climate shocks:
  - Fighting climate change and managing the incidence, level, and variability of precipitation and temperature can help contain the increasing number of so-called “climate refugees”.
- Note on distributional effects and future research:
  - Migration can harm home countries through brain drain and disrupt social networks; host countries may face pressures on public services and informal sector growth. The authors identify the need for future research to better delineate and quantify net benefits and costs of regional migration for origin and destination countries.

*Source: Appendix excerpts and tables from the IMF working paper (sample of 52 sub-Saharan African countries, 1990-2020) — content unit: wpiea2024097-print-pdf (Appendix Table A.7 and related tables and conclusion).*

### References

### Intra-African Migration: Exploring the role of Human Development, Institutions, and Climate Shocks — References (WP/2024/097)

### Primary themes evident in the references
- Climate and environment as migration drivers
  - Abel, G., M. Brottrager, J. Cuaresma and R. Muttarak, 2019, “Climate, conflict and Forced Migration”, Global environmental change, 54 (January 2019), 239-249.
  - Berlemann, M. and M. Steinhardt, 2017, “Climate Change, Natural Disasters, and Migration—a Survey of the Empirical Evidence”, CESifo Economic Studies, 63(4), 353-85.
  - Cattaneo, C. and G. Peri, 2016, “The migration response to increasing temperatures”, Journal of Development Economics, 122, 127-146.
  - Kaczan, D. J. and J. Orgill-Meyer, 2020, “The impact of climate change on migration: a synthesis of recent empirical insights”, Climatic Change, 158(3), 281-300.
  - El-Hinnawi, E., 1985, “Environmental Refugees”, Nairobi: United Nations Environmental Programme.

- Institutional, governance, and conflict influences on migration
  - Bertocchi, G. and C. Strozzi, 2008, “International Migration and the Role of Institutions”, Public Choice, 137, 81-102.
  - Cooray, A. and F. Schneider, 2016, “Does Corruption Promote Emigration? An Empirical Examination”, Journal of Population Economics, 29, 293-310.
  - Dimant, E., K. Tim and D. Meirrieks, 2013, “The Effect of Corruption on Migration”, Applied Economics Letters, 20(13), 1270-1274.
  - Dreher, A., T. Krieger and D. Meierreks, 2011, “Hit and (They Will) Run: The Impact of Terrorism on Migration”, Economics Letters, 113(1), 42-46.
  - Davies, S., T. Pettersson and M. Öberg, 2023, “Organized violence 1989-2022 and the return of conflicts between states?”, Journal of Peace Research, 60(4), 691-708.
  - Marshall, M. G., and G. Elzinga-Marshall, 2017, “Global report 2017: Conflict, governance and state fragility,” Vienna, VA: Center for Systemic Peace.

- Human development, labor economics, and selection in migration
  - Borjas, G. J., 1987, “Self Selection and the Earnings of Immigrants”, American Economic Review, 77(4), 531-553.
  - Borjas, G.J., 1999, “The Economic Analysis of Immigration”, in Handbook of Labor Economics, O.C. Ashenfelter and D. Card (eds).
  - Hatton, T.J. and J.G. Williamson, 2005, “What Fundamentals Drive World Migration?” in Poverty, International Migration and Asylum, Borjas, G.J., Crisp, J. (eds), Palgrave Macmillan, London.
  - Lee, E. S., 1966, “A Theory of Migration”, Demography, 3(1), 47-57.
  - Kandemir, O, 2012. “Human development and international migration”, Procedia-Social and Behavioral Sciences, 62, 446-451.
  - Lucas, R. E. B., 2015, “African migration” in Handbook of the Economics of International Migration, Barry R. Chiswick, Paul W. Miller (eds).

- Migration costs, barriers, and networks
  - Adsera, A. and M. Pytlikova, 2012, “The Role of Language in Shaping International Migration”, Economic Journal, vol. 125(586), pp. 49-81.
  - McKenzie, D., 2007, “Paper Walls are Easier to Tear Down: Passport Costs and Legal Barriers to Emigration,” World Development, vol. 35(11), pp. 2026-39.
  - Bertoli, S. and J.F.H., Moraga, 2013, “Multilateral resistance to migration”, Journal of Development Economics, 102(May 2013), 79-100.
  - Manzoor, W., Safdar, N. and H.Z., Mahmood, 2021. “A gravity model analysis of international migration from BRIC to OECD countries using Poisson Pseudo-maximum likelihood Approach”, Heliyon, 7(6), e07357.

### Empirical and methodological sources referenced
- Gravity models, Poisson approaches, and distance measures
  - Ramos, R. and J. Suriñach, 2017, “A gravity model of migration between the ENC and the EU”, Tijdschrift voor economische en sociale geografie, 108(1), 21-35.
  - Santos Silva, J.M.C. and S. Tenreyro, 2010, “On the existence of the maximum likelihood estimates in Poisson regression”, Economics Letters, 107(2), 310-312.
  - Mayer, T. and S. Zignago, 2011, “Notes on CEPII’s distances measures: The GeoDist database”, CEPII.
  - Schwartz, A., 1973, “Interpreting the effect of distance on migration”, Journal of Political Economy, 81(5), 1153-1169.

- Datasets, indexes, and measurement
  - Ahir, H., N. Bloom and D. Furceri, 2022, “The world uncertainty index”, Working Paper w29763, National Bureau of Economic Research.
  - Sundberg, R. and E. Melander, 2013, “Introducing the UCDP Georeferenced Event Dataset”, Journal of Peace Research, 50(4), 523-532.
  - Solt, F. (2020). Measuring income inequality across countries and over time: The standardized world income inequality database. Social Science Quarterly, 101(3), 1183-1199.
  - UNDESA (United Nations Department of Economic and Social Affairs) Population Division, 2020, “International Migrant Stock 2020”.
  - UNCTAD, 2018, “Economic Development in Africa Report 2018: Migration for Structural Transformation”, available at https://unctad.org/system/files/official-document/aldcafrica2018_en.pdf.

### Regional and topical studies focused on Africa and intra-regional flows
- Intra-African migration and sub-Saharan Africa evidence
  - Flahaux, M.-L. and H. De Haas, 2016, “African Migration: Trends, Patterns, Drivers”, Comparative Migration Studies, vol. 4(1), 1-25.
  - Cirillo, M., A. Cattaneo, M. Miller and A. Sadiddin, 2022, “Establishing the link between internal and international migration: Evidence from Sub-Saharan Africa”, World Development, 157, 105943.
  - Ruyssen, I. and G. Rayp, 2014, “Determinants of Intraregional Migration in Sub Saharan Africa 1980-2000”, Journal of Development Studies, 50(3), 426-443.
  - Li, Q. and C. Samimi, 2022, “Sub-Saharan Africa’s international migration constrains its sustainable development under climate change”, Sustainability Science, 17(5), 1873-1897.
  - Udelsmann Rodrigues, C. and J. Bjarnesen, 2020, “Intra-African Migration”, European Parliament. Directorate-General for External Policies of the Union.

### Conceptual and review foundations
- Migration theory, selection, and social dimensions
  - Bodvarsson, O. B., N. Simpson and C. Sparber, 2015, “Migration Theory” in Handbook of the Economics of International Migration, Barry R. Chiswick, Paul W. Miller (eds).
  - De Haas, H., 2010, “Migration Transitions: A Theoretical and Empirical Inquiry into the Developmental Drivers of International Migration”, Working Papers series, 24, Oxford: International Migration Institute.
  - Faist, T., 2016, “Cross-border migration and social inequalities”, Annual Review of Sociology, 42, 323-346.
  - Hartmann, B., 2010, “Rethinking Climate Refugees and Climate Conflict: Rhetoric, Reality and the Politics of Policy Discourse”, Journal of International Development, 22(2), 233-246.
  - Ghatak, S. and P. Levine, 1996, “Migration Theories and Evidence: An Assessment”, Journal of Economic Surveys, vol. 10(2), 159-198.

*References list excerpted from the PDF chapter/section: wpiea2024097-print-pdf - References; Intra-African Migration: Exploring the role of Human Development, Institutions, and Climate Shocks — Working Paper No. WP/2024/097*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2024/english/wpiea2024097-print-pdf.pdf_