Background
Heavy rains have been affecting Haiti since 5 November 2016, resulting in flash floods, water surges and landslides in 4 departments: Grand Anse, Northwest, North and Northeast ( ECHO , PAHO ). Ten deaths were reported, seven of which in Cap Haïtien. In this supplementary report, we present estimates of population movements for the Department of Nord.
Call detail records (CDRs) are registered by mobile operators for billing purposes. They include information on the cell tower used by subscribers when sending and receiving text messages and making calls. In de-identified formats, they can be used to estimate mobility patterns of a population. The CDR analysis described here was undertaken in compliance with the GSMA privacy guidelines developed in the context of the Ebola outbreak ( GSMA , 2014).
Methodology
Population flow estimation
Haiti has an estimated population of 11 million people. Digicel is the largest operator, having approximately 4.7 million subscribers (85% market share).
We estimate population flows based on movements of de-identified Digicel subscribers who lived pre-hurricane in the affected Departments of Sud, Grande Anse and Nippes. We show estimated flows of people within and out of the affected departments. We show results for Communes and Communal Sections. Population movement estimates are calculated by combining de-identified data on SIM card movements with available population data.
We assume in these analyses that movements of mobile phone subscribers are representative of movements of the population in the three Departments. Mobile phone use is relatively lower in several groups including women, children, the elderly, and the poorest. If these groups have substantially different movement patterns than groups with high mobile phone use, results will be biased. In general the relative distributions of flows across the country are more reliable than absolute numbers given per area. Our previous research in Haiti (post-earthquake) and Kenya (stable conditions) show that overall estimates of mobility corresponded well to population-level data (Bengtsson et al., 2011; Wesolowski et al., 2013). However the estimates provided here should be interpreted with the above mentioned caveats in mind.
Specifically, we show estimated absolute and relative flows as well as flows above and below normal levels.