Ensaios em gestão de risco e regulação bancária
Título da Revista
ISSN da Revista
Título de Volume
This dissertation consists of three essays. The first essay analyses the public information about the risk of Brazilian banks' loan portfolios and is split in two chapters. The first one compares public data with banks' managerial information and shows the limited scope of accounting data. One conclusion is that is possible to improve disclosure, a fact that has been happening gradually in Brazil through new rules related to Basel II’s Third Pillar and by more detailed reports by the Brazilian Central Bank. The second part of this essay shows the discrepancy between non-performing loans ratio (NPL) and probability of default (PD), it also discusses the relationship between loss allowances and expected loss. The tools used are empirical data (migration matrices) and a simulation based on overlapping vintages whose transition matrices are modeled as a second order Markov process. The second essay makes a linkage between risk management and price discrimination. Is developed a model that consists in a Cournot duopoly in a retail credit market, in which banks may perform third-degree price discrimination. In this model, potential borrowers can be of two types, low or high risk, wherein low risk borrowers are characterized by more elastic demand. Initially, banks cannot observe borrowers’ type. For this to happen the bank needs to invest in risk management and thus becomes able to price discriminate, which in this model coincide with risk-based pricing. According to the model, if costs to observe clients type are high, the banks' strategy is not to discriminate (pooling equilibrium). However, if these costs are sufficiently low, the optimal strategy for banks is to charge different rates for each group. It is argued that Basel II functioned as an exogenous shock that shifted the equilibrium where there is greater price discrimination. The third essay is divided into two chapters. The first one discusses the application of the concepts of subjective probability and Knightian uncertainty to VaR models and the importance of model risk assessment, which is comprised by estimation risk, specification risk and identification risk. The essay proposes that 'four elements' methodology from operating risk (internal data, external data, business environment and scenarios) can be extended to the measurement of other risks (market and credit risk, for example). The second part of this last essay is an application of scenario analysis to measure the conditional volatility on dates of relevant economic releases, specifically on COPOM (Monetary Policy Committee) meeting days.