Abstract
Abstract This research establishes a framework that combines econometric modelling and retrieval-augmented generation to explain the price dynamics of agriculture. With the help of the econometric analysis using monthly data of rainfall and the Wholesale Price Index (WPI) of paddy for Tamil Nadu (2012–2018), non-stationarity (ADF, KPSS), long-run cointegration, volatility clustering (GARCH), and structural breaks (PELT) were identified. The findings indicate that roughly 5% of the variance in medium-term prices can be attributed to rainfall. Adjustments lagging behind shocks indicate a rather inefficient market. Econometric models generate statistical signals but lack explanations for decision-oriented purposes. In order to bridge this gap, the paper proposes Retrieval-Augmented Econometrics or RAE, wherein econometric outputs serve as triggers for the retrieval of contextual knowledge from IMD reports, MSP policy documents, and TNAU advisories. They deploy large language models to generate bilingual context-relevant advisory messages. A prototype system generated a coherent explanation (4.6/5), is strongly grounded through retrieval (93%) and translates correctly (BERTScore 0.94) to offer an explainable and decision-oriented forecast for agriculture.