Abstract
Persistent food price inflation has intensified global interest in how price changes are transmitted through different stages of the food supply chains, known as vertical price transmission (PT). This study combines two interrelated meta-analytical approaches to assess the determinants, patterns, and predictive modeling of PT outcomes.The first chapter introduces the topic and explains why advancing PT research is both timely and necessary. Building on a comprehensive dataset over 140 studies and 411 observations, the second chapter investigates the presence of asymmetric price transmission (APT) and the direction of price causality using semi-nonparametric maximum likelihood and multinomial logit models. Results indicate that nonlinear cointegration methods are more likely to detect asymmetry, while the use of scanner data and studies based in Europe or in markets with high retail concentration are associated with lower likelihoods of detecting asymmetry. Retail trade regulations also significantly influence price transmission dynamics. Price causality is most frequently observed from upstream to downstream, although downstream to upstream and bidirectional causality are more common under strict price controls and greater import dependence. Additionally, higher retail concentration is linked to a greater likelihood of detecting bidirectional causality.
Expanding this analytical framework, the third chapter applies machine learning (ML) tools to improve both the efficiency and predictive accuracy of meta-analysis using a comprehensive dataset from 94 studies. Leveraging ASReview, literature screening time was reduced by over 60%, while Random Forest Regression (RFR) significantly outperformed traditional weighted least squares (WLS) models with improved R2 by up to 67 percentage points. Through resampling (SMOGN) and feature selection (RFE), we identify key predictors driving PT outcomes, including adjustment speed, asymmetry in speed, and asymmetry in magnitude of price responses. The findings underscore the importance of product characteristics such as perishability and processing levels, and demonstrate that ML methods capture non-linear relationships
and market heterogeneity more effectively than conventional approaches. The final chapter
concludes the study’s insights and highlights avenues for future research.
Taken together, these chapters deliver both methodological advances and empirical insights that enhance our understanding of PT across diverse food systems. The study highlights how market structures, policy environments, and data characteristics interact to shape PT behavior, providing actionable evidence for more effective policy interventions aimed at improving market efficiency and safeguarding consumer welfare in times of persistent global food price volatility.