Knowledge that Transforms

Portfolio selection based on high-dimensional covariance matrices is a key challenge in data-rich environments with the curse of dimensionality severely affecting most of the available covariance models. We challenge several multivariate Dynamic Conditional Correlation (DCC)-type and Stochastic Volatility (SV)-type models to obtain minimum-variance and mean-variance portfolios with up to 1000 assets. We conclude that, in a realistic context in which transaction costs are taken into account, although DCC-type models lead to portfolios with lower variance, modeling the covariance matrices as latent Wishart processes with a shrinkage towards the diagonal covariance matrix delivers more stable optimal portfolios with lower turnover and higher information ratios. Our results reconcile previous findings in the portfolio selection literature as those claiming for equicorrelations, a smooth dynamic evolution of correlations or correlations close to zero.

In this paper, we provide several theoretically relevant and empirically significant improvements to the general affine realized volatility (GARV) model of Christoffersen et al. (2014). We impose hidden volatility components in both the return-based conditional variance and the realized variance and augment their combination with another jump component. This new composition nests within a common framework several empirically well-tested models such as the GARV model mentioned above. To facilitate practical implementations we obtain the closed-form formulas to evaluate VIX and its futures through a variance-dependent kernel. Our empirical studies demonstrate that the volatility-component specification provides a further evident improvement in VIX forecasting and its futures pricing across maturity and volatility levels; more importantly, these hybrid and hidden features turn out to be complements rather than substitutes, and their prominence is further intensified by the jump.

We present a meta-analysis of the impact of higher capital requirements imposed by regulatory reforms on the macroeconomic activity (Basel III). The empirical evidence derived from a unique dataset of 48 primary studies indicates that there is a negative, albeit moderate GDP effect in response to a change in the target capital ratio. Meta-regression results suggest that the estimates reported in the literature tend to be systematically influenced by a selected set of study characteristics, such as econometric specifications, the authors’ affiliations, and the underlying financial system. Finally, we discuss the publication bias.

We examine the impact of corporate board reforms around the world on stock price crash risk. Using a sample of firms in 41 economies that passed major board reforms between 1990 and 2012, we find that board reforms are associated with a significant reduction in crash risk of about 13%. The effect of reforms on crash risk is stronger among firms with more severe ex ante agency problems. Our analysis further suggests that board reforms reduce crash risk by improving financial transparency and enhancing investment efficiency. In sum, our findings are consistent with the notion that board reforms improve board oversight and mitigate agency problems.

Examining a loan-level matched sample of Japanese banks and firms, we study the factors determining the termination of bank–firm relationships. We find that such terminations are mainly driven by bank-side factors and that these bank-driven terminations increase when banks’ capital conditions worsen. Furthermore, a longer relationship duration decreased the probability of termination substantially when the Japanese banking system was stable, whereas the duration effects weakened when the system became fragile.

I study the role of sovereign risk in determining the effects of expected bailouts on banks’ portfolio decisions. Empirically, data on Italian banks show that they decrease lending to firms and increase purchases of government bonds following an increase in the probability of a bailout, if the risk of sovereign default is sufficiently low. Crucially, the portfolio adjustment becomes weaker and eventually reverses sign as sovereign risk increases. To interpret these results, I develop a model in which the relation between the bailout probability and the corresponding payoff to bank owners (“bailout rents”) depends on sovereign risk. The model’s predictions are consistent with the key features of the data.

This paper investigates the effect of media talk on bank stock returns in response to corporate governance news. Using Loughran and McDonald's (2011) dictionary, we create four categories of word lists that define the positive/negative tone and degree of certainty/uncertainty of news. We document three relevant findings. First, negative news significantly affects bank stock returns. Second, media coverage and the degree of certainty of the news are associated with more severe stock market losses. Third, bank capital and risk-adjusted performance mitigate the effect of negative news on stock prices. Overall, our study suggests that media talk on bank corporate governance events is an important determinant of abnormal stock returns.

To capture systemic risk related to network structures, this paper introduces a measure that complements direct exposures with common exposures, as well as compares these to each other. Trying to address the interconnected nature of financial systems, researchers have recently proposed a range of approaches for assessing network structures. Much of the focus is on direct exposures or market-based estimated networks, yet little attention has been given to the multivariate nature of systemic risk, indirect exposures and overlapping portfolios. In this regard, we rely on correlation network models that tap into the multivariate network structure, as a viable means to assess common exposures and complement direct linkages. Using BIS data, we compare correlation networks with direct exposure networks based upon conventional network measures, as well as we provide an approach to aggregate these two components for a more encompassing measure of interconnectedness.

When it comes to individual stock option pricing, most applications consider a univariate framework. From a theoretical point of view this is unsatisfactory as we know that the expected return of any asset is closely related to the exposure to the market risk factors. To address this, we model the evolution of the individual stock returns together with the market index returns in a flexible bivariate model in line with theory. The model parameters are estimated using both historical returns and aggregated option data from the index and the individual stocks. We assess the model performance by pricing a large set of individual stock options on 26 major US stocks over a long time period including the global financial crisis. Our results show that the losses from using a univariate formulation amounts to 11% on average when compared to our preferred bivariate specification.