India ranks among the world's largest consumers of gold, yet existing research has relied predominantly on conventional macroeconomic frameworks centred on inflation, exchange rates, and interest rates to explain gold demand, while systematically overlooking the psychological and behavioural forces that substantially govern Indian investor decision-making. This study addresses that gap by examining gold demand dynamics in India over the period 2015–2024 through an integrated behavioural and machine learning framework grounded in Prospect Theory, Mental Accounting Theory, and Safe Haven Theory. Using secondary data sourced from the World Gold Council, the Reserve Bank of India, and leading financial data repositories, the study incorporates macroeconomic variables alongside behavioural proxies including the CBOE Volatility Index (VIX), financial media sentiment scores, and investor attention indices. The analytical architecture spans descriptive statistics, OLS regression, Granger causality testing, dynamic rolling correlations, event study methodology, and supervised machine learning models specifically Random Forest and Gradient Boosting algorithms. Key findings reveal that inflation (India CPI r = 0.93) and rupee depreciation (USD/INR r = 0.88) are the dominant long-run structural drivers of gold prices in India, with the exchange rate amplifying dollar-denominated returns by approximately 35–40% in rupee terms. Gold's safe haven properties are confirmed as episodic and regime-dependent, intensifying during sustained crisis windows such as the COVID-19 pandemic and Russia–Ukraine conflict. Machine learning models outperform traditional OLS benchmarks in out-of-sample accuracy, with variable importance analysis validating the primacy of inflation and currency variables. The study contributes emerging-market-specific evidence to safe haven and mental accounting theory while offering actionable insights for retail investors, institutional portfolio managers, and policymakers.
Anushka Leni (Mon,) studied this question.