Essays on the role of memory in financial markets

This thesis investigates the phenomenon of persistence as a key and economically meaningful characteristic of financial markets, one that critically influences forecasting performance, price discovery, and long-term investment outcomes. Despite extensive research on long memory in asset prices, its economic significance and practical importance remain insufficiently understood. This study helps close this gap by demonstrating that persistence captures how information propagates through interconnected financial systems and can be systematically measured and exploited. To operationalise these insights, the thesis develops an integrated framework linking system identification, memory dynamics, and economic application. A spillover-based variable selection approach is introduced to define financial systems through directional shock transmission, ensuring that persistence is estimated within coherent and internally connected market structures. Applied to energy markets, this framework improves forecast accuracy and yields more stable and interpretable memory estimates. Building on this foundation, the thesis shows that persistence shapes price formation through the interaction of historical dependence and forward-looking anticipation. Using fractional cointegration and bidirectional memory measures, it demonstrates that fundamentally driven markets are dominated by backward-looking persistence, while speculative markets exhibit stronger anticipatory dynamics. These distinct memory profiles explain systematic variation in predictability and establish persistence as an information-rich determinant of asset price dynamics. Finally, persistence is embedded into a multi-period asset-pricing framework, transforming memory from a descriptive time series feature into a structural intertemporal risk characteristic. By incorporating fractional memory into closed-form expressions for effective variance and skewness, the thesis develops a memory-augmented Capital Asset Pricing Model that links serial dependence to long-horizon wealth dynamics. Empirical evidence from US and Japanese equity markets shows that memory-aware strategies enhance risk-adjusted portfolio performance and long-horizon efficiency.