Contributions to Financial Econometrics: Asset Pricing in a DSGE Framework and Volatility Discovery in Cryptocurrency Markets

Abstract

This dissertation presents three distinct research studies in the field of financial econometrics. The first study critiques the performance of DSGE asset pricing models by econometrically assessing the plausibility of one renowned model’s estimated asset pricing parameters. The partial indirect inference (PII) estimation method is adapted and applied for this purpose. The superiority of this method lies in its ability to achieve consistent estimates for the asset pricing parameters of interest, by formally allowing for the calibration of the nuisance parameters, that are an inherent feature of highly structural models. The analysis shows that the model under investigation is capable of resolving the infamous equity premium puzzle, yet the risk-free rate puzzle still poses an unresolved problem. The second study investigates the estimation challenges posed by the inherent misspecification in highly structured models. The parameter estimates obtained from the PII method in the first study are compared with the estimates obtained from two other indirect inference estimation methods. The novel „dark matter“ measure is then adapted and utilized to investigate, beyond statistical inference, the fragility of the different estimation methods to the underlying misspecification in the model. As a result, a modified PII method is established, whereby some nuisance parameters are included in the estimation process. This improves the overall estimation quality of the model, especially with regards to its implied business cycle dynamics. Finally, the third study, departs from the macro-level view of the economy and closely investigates the structural interdependence of volatility in the cryptocurrency markets. Given that the same cryptocurrencies are traded on different exchange markets, the aim is to discover the market driver of volatility. The well-known price discovery methodology is adapted for this purpose, taking into account the long memory property characterizing volatility series. The results indicate that the different cryptocurrencies have different volatility market leaders, and that the market with the highest trading volume is not necessarily also the volatility leader.