This project provides an in-depth econometric panel data analysis to investigate the factors influencing the deaths and disappearances of migrant children worldwide. Using a decade of data, we explore the temporal and spatial patterns of these tragic events and build a statistical model to identify key predictors, with the ultimate goal of informing policy and humanitarian interventions.
The primary research question driving this project is:
What are the factors that influence the total number of deaths and missing children in migrant incidents across continents and time?
To answer this, our objectives were to:
- Analyze the temporal and spatial distribution of migrant incidents.
- Specify and estimate an appropriate panel data regression model.
- Identify significant predictors of child deaths and disappearances.
- Provide data-driven insights to support policy interventions aimed at protecting vulnerable populations.
This project was developed for the Statistics for Data Science course in the Master's in Data Science and Advanced Analytics program at NOVA IMS. The work was completed during the 1st Semester of the 2024/2025 academic year.
The data was sourced from the Missing Migrants Project, an initiative by the International Organization for Migration (IOM).
- Dataset: Contains 5,462 entries of migrant incidents from 2014 to 2024.
- Scope: The dataset includes detailed information on incident dates, locations, and the number of deceased, missing, and surviving individuals, including gender breakdowns.
The analysis and visualization for this project were conducted entirely within the R ecosystem.
Our approach followed a structured econometric analysis process:
-
Data Preprocessing:
- Filtered the dataset for high-quality source information.
- Transformed variables using logarithmic transformations to handle non-normal distributions and created dummy variables for categorical features like
Cause of Death.
-
Initial Analysis & Visualization:
- Illustrated temporal trends using polar charts and geospatial patterns with an animated global map, identifying the Mediterranean Sea and Central America as critical high-risk zones.
-
Econometric Modeling:
- Model Selection: We employed Panel Data Regression to control for both time and continent-specific effects. We compared Pooled OLS, Fixed Effects (FE), and Random Effects (RE) models.
- Specification Tests:
- Breusch-Pagan Test: Confirmed the presence of heteroskedasticity (p-value < 0.05), necessitating the use of robust standard errors.
- Robust Hausman Test: Yielded a p-value of 0.929, indicating that both Fixed and Random Effects models are consistent, but the Random Effects model is more efficient. This model was chosen for the final analysis.
The Robust Random Effects model revealed several significant factors influencing the number of child deaths and disappearances (D&M):
- Impact of Female Fatalities: Both
log(Number of Females)andlog(Number of Males)showed positive and significant coefficients. However, the coefficient for female D&M (0.275) is substantially larger than for male D&M (0.069). This suggests a 1% increase in female deaths is associated with a 0.275% increase in child deaths, highlighting the critical role of women as primary caregivers. - Risk of Drowning: The
Drowningdummy variable was positive and significant, indicating that drowning incidents are associated with an approximately 4.92% increase in the expected number of child D&M compared to other causes of death.
Our analysis confirms that migrant deaths and disappearances are concentrated in high-risk regions and that the vulnerability of women is directly linked to an increased risk for children.
Future Research Directions:
- Incorporate data on smuggling networks and conflict levels in countries of origin.
- Conduct a more granular analysis of specific migration routes to guide localized interventions.
- Advocate for improved data standardization across regions to enable more accurate and comprehensive crisis assessment.
- André Silvestre (20240502)
- Filipa Pereira (20240509)
- Umeima Mahomed (20240543)