Analysis and Forecasting of Youth Employment: A Comparative Study between Shandong Province, China and Klang Valley, Malaysia

Abstract

Youth labor markets in Shandong Province, China, and the Klang Valley, Malaysia, have moved from problems of job quantity to problems of structural friction, yet the two regions are usually studied with incompatible indicators and in separate literatures. This article sets out a quantitative research design for analyzing and forecasting youth employment in both regions, with NEET (not in employment, education or training) status treated as an endogenous determinant of later outcomes rather than only as a residual outcome of weak demand. Four research questions are addressed. The first establishes harmonized baselines and trends since 2014 for labor force participation, unemployment and NEET prevalence among young people aged 15 to 29, together with survey-based measures of long-term disengagement. The second asks whether the duration of disengagement is associated with psychological distress, measured with the Kessler 6 scale, and with downgraded career expectations, measured as a reservation wage gap. The third uses multi-group structural equation modeling, preceded by measurement invariance testing, to compare how perceived involution and skill mismatch shape disengagement trajectories in each region. The fourth evaluates whether adding structural employability variables, including an education and industry matching index and exposure to digital substitution, improves the explanatory power and out-of-sample accuracy of NEET forecasting models relative to a macroeconomic baseline. The design combines official and survey-based secondary series with a harmonized survey of 1,500 school leavers aged 18 to 29, 750 in each region. Decision rules, power calculations, endogeneity checks and ethical safeguards are fixed before data collection. Because the two regions publish very different youth statistics, the article also shows how a comparison can be built when official series are missing or not comparable, and it sets out how the resulting models can be used for ex-ante policy scenarios.