Data for forecast and foresight

Concerns have been raised about the ability of governments to address sudden shocks and urgent societal pressures that are harder to anticipate such as climate change, global epidemics, accelerating digitalisation and shifting global power dynamics. The traditional way of governing migration is often more reactive than proactive, collaborative and experimental (Demos Helsinki, 2023). Therefore, there is growing interest in methods for forecasting the future of migration. 

There are two types of foresight activities, resilient activities try to maintain the functioning of a system despite external shocks or pressure, such as establishing an early warning system for migration trends, while transformative activities aim to change the way in which the system operates, such as introducing the use of digital nomad visas (Demos Helsinki, 2023). Knowledge is essential to support anticipatory policymaking and can be conceptualized as predictive (what is likely to happen?), exploratory (what could happen if), or visionary (what should happen?). Predictive knowledge is more about forecasting the future while exploratory and visionary knowledge are more about foresight, e.g., what would happen if the conditions were different or if other alternatives were pursued. 

Examples of methods for generating predictive knowledge include scanning for trends, data analysis, artificial intelligence methods, predictive models and early warning systems.  Examples of methods for generating exploratory knowledge include horizon scanning, influence maps, Futures Wheel, Scenario building, Delphi and Wild cards. Finally, examples of methods for generating visionary knowledge include Backcasting, Dotmocracy, Behaviour-based experiments, Humble Governance model and Participatory visioning process. The figure below provides examples of policy questions that could be answered using predictive, exploratory and visionary anticipatory knowledge.

Figure 3: examples of questions answered using predictive, exploratory and visionary knowledge

(Source: Demos Helsinki, 2023)

Forecasts produce a quantitative estimate of future migration and refer to forecasts, predictors and projections (IOM, 2020a). Forecasts rely on numerical information from the past to derive a trend for the future. Some forecasting (particularly demographic projections) calculates the past effect of key drivers of migration (i.e., employment) and model how migration may change in relation to a change of that particular factor. For example, how much do migration levels drop if unemployment in destination countries goes up? However, past migration data are often limited, of poor quality or not comparable, making forecasting difficult (IOM, 2020a).

Foresight is different from forecasts in that it is more about storytelling than about numbers. Foresight (or often called scenarios) are qualitative narratives about the future of migration that emphasize possible changes and their consequences for migration—a “what if…” approach. Scenarios draw on a practitioner-driven, strategic, iterative (repeating steps and improving the scenario over time) and discursive (i.e. experts talking to each other) methodology (Sohst et al., 2020). Migration scenarios are an increasingly popular method of generating foresight. However,  the limitation of this approach is that experts tend to hold their own biases and that scenarios about the future are not as useful to  policymakers looking action-oriented knowledge (IOM, 2020a).

Finally, early warning systems monitor migration trends or potential drivers of migration and forced migration in real time, with the goal of signalling surges in migration to policymakers with as much advance notice as possible. Early warning systems select a set of “indicators” (e.g. how many people cross the border every week) and establish a number of thresholds (e.g. how many people need to cross the border per month before taking action) to automatically set in motion a pre-determined protocol of action. However, it is challenging for experts and policymakers to know which indicators are meaningful and which thresholds are most appropriate. Furthermore, data collection has to be consistent and comparable over time to ensure that observed changes are not caused by changes in the way the data are collected (IOM, 2020).