Adapted World Food Insecurity Data (1990-2024)

Description:
The Global Food Insecurity dataset used in this study was constructed by harmonizing data from multiple public sources, in order to obtain a consistent and continuous time series covering the period 1990–2024 at the global (World) level. As no single international source provides complete and uninterrupted food insecurity data for this entire time range, the dataset was compiled by carefully integrating information from authoritative reports and databases, as follows:

1. 1990–2014 period:

For this period, data were manually extracted from The State of Food Insecurity in the World (SOFI) reports published annually by the Food and Agriculture Organization (FAO).
The indicator used was the Prevalence of Undernourishment (PoU), defined as the estimated percentage of the population whose habitual food consumption is insufficient to provide the dietary energy levels required to maintain a normal, active, and healthy life.
Although the PoU focuses primarily on chronic caloric undernourishment, it has been widely used in the literature as a key indicator of global food insecurity prior to the development of experiential-based measures.

Relevant data points from each annual SOFI report were manually compiled, cross-checked for consistency across editions, and transformed into an annual time series from 1990 to 2014. In cases where FAO reported multi-year moving averages (e.g., 3-year averages), the values were assigned to the central year of the corresponding period.

1. 2015–2022 period:

From 2015 onwards, the dataset incorporates data from the Our World in Data platform, specifically the indicator “Share of population with severe food insecurity” derived from the FAO’s Food Insecurity Experience Scale (FIES).

FIES is a direct, experiential measure of food insecurity based on household surveys, and it corresponds to SDG indicator 2.1.2. This metric complements and updates the previous PoU measure by capturing the severity and experience of food access problems.
The OWID dataset, in turn, is based on FAO’s FIES data and was used for the period 2015–2022 as reported by FAO and OWID.

1. 2023–2024 period:

Since full-year validated data for 2023–2024 are not yet available in FAO or OWID publications at the time of writing, values for these years were extrapolated using trend-based estimation from the most recent available observations (e.g., using linear trend extrapolation or smoothing across the latest 5 years), in line with common practice in time series construction for policy analyses.

1. Harmonization:
   As the underlying indicators (PoU vs. FIES severe insecurity) are conceptually different but related (both representing serious forms of food insecurity), the entire series was harmonized into a common scale of “% of population food insecure” to allow for temporal comparisons. This was done by cross-validating overlapping years and adjusting for indicator definitions based on FAO and SDG metadata documentation.

Limitations:
The constructed series represents a best-effort harmonization of available public data. While PoU and FIES severe insecurity do not capture identical dimensions of food insecurity, their trends and values are broadly comparable and widely used in global monitoring. The adapted dataset is suitable for trend analysis and causal modeling as conducted in this study, but caution should be exercised when comparing absolute levels across time segments.

Table 1. Food Insecurity Data Sources (1990–2024, adapted dataset)