Migration Anticipation and Preparedness

Executive summary

Why predicting migration is hard – and yet necessary

Migration is one of the most complex social phenomena to anticipate. People move for deeply interconnected reasons: economic opportunity, political instability, safety concerns, education, family reasons, and personal aspirations. These drivers interact with policy changes, geopolitical crises, environmental shocks, and shifting labour markets. Sometimes flows change gradually; at other times, they surge abruptly in response to conflict or political events. The refugee movements following the Syrian conflict in 2015‑2016 and Russia’s large‑scale invasion of Ukraine in 2022 revealed how difficult it is for governments to anticipate large‑scale displacement.

Yet despite this inherent uncertainty, governments must attempt to forecast migration. Without forward-looking estimates, it becomes nearly impossible to properly allocate budgets, plan housing and integration services, manage asylum systems, or adjust labour market policies. Forecasting supports contingency planning, strengthens crisis preparedness, and enables more transparent public communication. It also facilitates international co‑operation and burden-sharing by identifying potential pressure points before they escalate.

The OECD’s Migration Anticipation and Preparedness (MAP) Task Force developed this Handbook to guide policymakers, civil servants, analysts, and modelers in building more robust forecasting systems. The document emphasises that forecasting is not about achieving perfect prediction. Rather, it is about giving the right tools to reduce uncertainty, prepare for multiple plausible scenarios, and embed anticipation into migration governance.

One size does not fit all: Tailoring forecasts to migration categories

A central argument of the Handbook is that migration cannot be forecast as a single aggregate number. While demographic projections often rely on net migration figures, such aggregates are insufficient for operational policymaking. Effective governance requires understanding who is migrating, under which category, and why.

Different migration streams respond to different drivers and exhibit distinct statistical characteristics. So-called forced migration flows, such as asylum applications and irregular border crossings, are highly volatile and often non-stationary. They can shift rapidly in response to conflict, disasters, or policy changes. In contrast, labour, student, and family migration may follow more stable and structured patterns, often influenced by economic conditions or institutional frameworks.

Because of these differences, forecasting must be category specific. Time‑series models such as ARIMA may work well for relatively stable flows. Econometric models can help analyse structured labour or student migration patterns, although they often struggle to anticipate sudden breaks. Bayesian methods help to improve forecasting results by incorporating prior knowledge (including expert opinion) and uncertainty. More recently, machine learning techniques – including neural networks models – have shown promise for short-term forecasts of high-frequency data, particularly for asylum and irregular migration.

Across OECD countries, forecasting practices vary widely. Many countries regularly forecast asylum applications, but far fewer predict irregular border crossings or regulated migration streams such as labour or student migration. Most systems combine quantitative modelling with expert judgement, acknowledging that statistical models alone cannot capture sudden policy shifts or geopolitical developments. The appropriate level of model complexity depends on the volatility of the flow, the time horizon, data availability, and institutional capacity. In many cases, simpler models (whether augmented by expert elicitation or not) may perform nearly as well as more complex ones while remaining easier to maintain and communicate.

From models to preparedness: Data, capacity, and continuous updating

Reliable data are the foundation of migration forecasting, yet they present significant challenges. Migration statistics often suffer from delays in publication, inconsistent definitions, limited frequency, and discontinuities over time. The Handbook proposes evaluating data sources according to criteria such as accuracy, timeliness, coverage, granularity, continuity, and clarity of definitions.

A key trade‑off emerges between timeliness and accuracy. Highly validated data may arrive too late for operational use, while real-time or high-frequency indicators may lack full reliability. Increasingly, forecasters complement administrative data with alternative “digital trace” sources such as online search data, social media signals, mobile phone data, and event databases. These innovative sources can enhance short-term forecasting, particularly in more volatile contexts.

Beyond methodology and data, institutional capacity is critical. Effective forecasting requires secure IT infrastructure, computing resources, skilled analysts, and cross-ministerial co‑ordination. Forecasting systems must be embedded within governance structures, supported by clear mandates and long-term political commitment. Data-sharing agreements and interagency collaboration are essential to ensure sustainability.

Forecasts must also be regularly updated. Migration systems are dynamic, and models must adapt to new events. Medium-term forecasts for a few years often require annual recalibration, while early warning systems based on high-frequency data and forecasting for a few months may need monthly revisions. Major geopolitical or policy changes may require immediate adjustments outside regular update cycles.

Ultimately, the Handbook frames migration forecasting as a structured approach to managing uncertainty. Perfect prediction is neither possible nor the objective. Interacting regularly with policymakers to acknowledge that matter of fact, alongside disseminating forecasting results to the right network of decision makers on the ground, is key to make sure forecasting, despite its uncertainty, is relevant. By tailoring models to specific migration categories, investing in high-quality data and institutional capacity, combining quantitative analysis with expert insight, and maintaining continuous updates, governments can significantly strengthen their preparedness. Forecasting thus becomes not merely a technical exercise, but a central pillar of evidence‑based and resilient migration governance.