Forecasting and Correlation Dynamics in Regional and Global Currency Trios [before doctoral defense]

Kivonat, rövid leírás

This dissertation investigates the forecasting and microstructural determinants of exchange-rate correlations within foreign exchange (FX) currency triangles. Because exchange rates linked through triangular arbitrage cannot evolve independently, correlations are constrained by no-arbitrage relationships. Nevertheless, substantial differences are observed between the correlation structures of major global currencies and those of smaller regional currencies. The thesis aims to explain these differences by combining correlation forecasting, ultra-high-frequency econometric analysis, and market microstructure modeling. Using daily FX and option market data between 2007 and 2025, the dissertation compares option-implied correlation forecasts with traditional time-series approaches, including Historical, EWMA, GARCH, and HAR models. The results show that implied measures outperform historical methods in major currency markets but fail to do so in regional markets. In contrast, EWMA-based models consistently deliver the most accurate correlation forecasts for regional currency trios. The study further demonstrates that superior volatility forecasts do not automatically translate into superior correlation forecasts, despite the theoretical relationship between volatility and correlation under triangular arbitrage. To understand the origins of these differences, the thesis analyzes ultra-high-frequency exchange-rate data at 100- and 200-millisecond intervals using Vector Autoregressive (VAR) models, Local Projections, impulse-response functions, and spillover networks. The findings reveal a clear distinction between major and regional currency triangles. Major currency markets exhibit relatively symmetric and bidirectional information transmission, whereas regional markets display strongly asymmetric and hierarchical spillover structures. These results indicate that, in many regional markets, one leg of the currency triangle is not independently discovered but is instead quoted synthetically from the other two exchange rates. Finally, the dissertation develops a modified agent-based market microstructure model capable of reproducing the correlation regimes observed in practice. The simulations show that quotation conventions and triangular arbitrage mechanisms are sufficient to generate the characteristic correlation patterns of both major and regional currency trios. Overall, the thesis demonstrates that exchange-rate correlation structures arise the organization of price discovery and quoting practices within FX markets, providing a unified explanation for the divergent correlation regimes observed across global and regional currencies.