|dc.description.abstract||In chapter 1, given the financial troubles facing state pension plans in recent years, we examine determinants of the ratio of assets to liabilities, or the funded ratio, based on data for 153 pension plans from 2001 to 2014. The focus is on the relationship between both the actual investment return on pension assets and the assumed return used to discount pension liabilities, or the funded ratio. Importantly, only when appropriate empirical techniques are employed to address potential econometric problems do we find that these two factors have the expected relationship with the funded ratio. Surprisingly, we also find the actual and assumed returns are negatively correlated, even though the correlation is quite low. Furthermore, the assumed return is on average higher than the actual return and has a much larger marginal effect on the funded ratio. We therefore show how a relatively high value can be assigned to the assumed return to make a pension plan appear to far healthier than actually is the case.
In chapter 2, we examine the effects of banks’ client stock ownership structure on their governance mechanism and their risk-taking as well as the effects of such ownership ties in the banking sector on systemic risk in the financial system. Importantly, we apply a dyadic level of analysis to provide new insights on the relevance of such cross-ownership as effective mutual monitoring channel and as a possible source of interconnectedness between and among financial institutions. Our empirical results indicate that bank-client cross-ownership of bank stocks is negatively associated with the riskiness of BHCs. This means that large external equity holders have the potential to perform an effective monitoring role and mitigate agency problems in the banking sector. We also find that bank-client cross-ownership of bank stocks is positively associated with systemic risk and the effects of such cross-ownership on systemic risk is stronger in times of a financial crisis.
In chapter 3, we propose dimension reduction methods and shrinkage methods to forecast tier 1 common capital ratio (T1CR) of the five biggest bank holding companies (BHCs) in the U.S. in a data-rich environment. Specifically, we employ two dimension reduction methods – the principal component regression (PCR) and the partial least squares regression (PLSR), and three shrinkage methods – the ridge regression, the Least Absolute Shrinkage Selection Operator (LASSO) regression, and the elastic net regression. We apply these methods to in-sample and out-of-sample forecasting exercises for T1CR, an extremely important banking variable and the most accurate indicator of the ability of banks to absorb losses. Our results show that factor-type models, PCR and PLSR, dominate the other alternative models over 1- to 10-quarter ahead forecast horizons, while shrinkage methods tend to outperform the factor-type forecasting models over 11- to 12-quarter ahead forecast horizons. In addition, we find that bank and stress test variables help produce the most accurate forecasts for short-term forecast horizons, while macro variables are useful in forecasting long-term horizons. Finally, we find that only six factors account for much of the variance of our 162 quarterly time series in the full dataset and that the most accurate forecasts of T1CR are obtained with just a few factors. One interpretation of such findings is that there may be only a few important sources that are necessary to accurately forecast banks’ capital.||en_US