Fig. 1: Global linear trends over time, 2000–2022.

Coefficients of a linear (least squares) regression of the indicator on time (year) for all indicators with data from more than 1 year. Indicator data (response variable) are normalized by minimum–maximum scaling and multipled by 100 before regressing on the variable year with country fixed effects; the original units are defined in Supplementary Table 1. The results are interpreted as average percentage change in the indicator per year over the period analysed. Note the number of country-years differs per indicator. Change defined as desirable if the results per indicator meet the following criteria: (1) the trend line is statistically significantly different from zero, (2) the coefficient estimate is >0 when rounded to the nearest tenth and (3) the sign agrees with the desirable direction of change (defined in Supplementary Table 1). Conversely, the estimates were labelled as changing in the undesirable direction if they were statistically significant with a non-zero coefficient when rounded to the nearest tenth and a sign opposite the desirable direction of change. Coefficients that round to zero, even if statistically significant, are classified as no change, since the magnitude of the coefficient is not practically meaningful. The regressions are weighted by the weighting variables defined in Supplementary Table 1. The sample size (number of countries and years) per indicator included in the analysis is provided in Supplementary Table 4. Excluded indicators do not have time series data and include All-5, minimum dietary diversity (women and children), soft drink consumption, zero fruits or vegetabls (adult and children), NCD-protect, NCD-risk, functional integrity, fisheries health index, social protection coverage, social protection adequacy, child labour, female landholdings and reduced coping strategies.