Figure 3.7: Length of time for which data was collected following the 50 largest displacements in 2016 and the ten largest each year 2008 to 2015 Source: IDMC Figure 3.8: The magnitude of the estimated stock figure for displacement by disasters, based on the length of time for which data was collected Source: IDMC For more than half of the sample, the “last” data point came within a month of the disaster striking, and for several it was more a matter of days. Forty events, less than a third of the total, yielded data recorded three or more months later. Only for around 18 per cent was displacement still reported on after a year. Given that many of these displacement events would have been accompanied by the widespread damage or destruction of homes, livelihoods and basic infrastructure and services, we would expect more of them to have involved displacement continuing for over three months. Underscoring this point, when we examine timeseries data for individual disasters we see that data collection almost always ends before the number of people displaced returns to zero (see Philippines box below). Nor does available data account for multiple displacements of the same people within the same year, if they return only to face ongoing risks, or are confronted with similar or new risks that cause their secondary or onward displacement from their place of refuge. 80 GRID 2017 The lack of monitoring over time and the limited availability of longitudinal data has significant consequences in estimating the total number of people displaced at a given point in time. Compared with data collected several months after a disaster, that collected immediately after it strikes tends to relate more closely to the peak number of people displaced, particularly when this data relates to the number of people evacuated. This obscures the fact that a significant number of people may have been displaced for days or weeks rather than months or years. This is the case when voluntary returns occur and when the homes of evacuated people are not destroyed, rendered uninhabitable or remain inaccessible. If we were to include all events from our sample for which only peak displacement data is available, it would result in a figure of around 89 million people (see figure 3.8). That is an overestimate because it does not account for everyone who returned home or established a new home elsewhere after data collection stopped.