Forecasting human development with an improved Theta method based on forecast combination

Abstract Forecasting human development is important for tracking sustainable growth and societal progress. However, this task presents statistical challenges. The primary difficulty is the limited nature of the available data, which is a typical problem encountered in forecasting many social time series. In this paper, we propose a novel approach for forecasting short time series based on the Theta method. The classical Theta method decomposes the time series into trend and short-run components. We propose an improved version of the Theta method, called $$\theta $$ <mml:math xmlns:mml="http://www.w3.org/1998/Math/MathML"> <mml:mi>θ</mml:mi> </mml:math> -comb, based on the combination of alternative forecasts for the short-run component. We apply the proposed method to forecast worldwide human development, measured with the Human Development Index, from 1990 to 2022. The results show that the $$\theta $$ <mml:math xmlns:mml="http://www.w3.org/1998/Math/MathML"> <mml:mi>θ</mml:mi> </mml:math> -comb method significantly improves the out-of-sample accuracy in comparison to existing approaches.