Heterogeneous agent models to determine spillover effects in the context of quantitative easing

Koziol, Tina

Heterogeneous agent models to determine spillover effects in the context of quantitative easing

Doctoral Thesis

2019

Abstract

We develop heterogeneous agent models to investigate financial spillover effects in the context of Quantitative Easing (QE). We consider these spillover effects from two perspectives. The first perspective studies spillovers within a network of financial institutions. The aim is to understand where amplification effects occur in the event of a shock. For this purpose, we calibrate a model of fire-sale contagion to the South African banking sector. We use cross-sectional balance sheet data for 29 South African banking institutions. Fire-sale externalities are pecuniary externalities that operate through prices. They pose a threat to the financial system because they amplify price shocks across assets and thus lead to liquidation spirals. In the first step, we investigate general shock propagation scenarios to an unsecured lending portfolio of a large bank and to a marketable asset held by all banks, i.e. South African government bonds. We rank individual banks according to their contribution to systemic risk and show the importance of cash liquidity buffers in reducing risk of fire-sale occurrences. Further, we find a critical threshold parameter which, if exceeded, makes the banking system highly unstable. In the second step, we build on findings presented by Cecchetti et al. (2017) that determine a relationship between Quantitative Easing and risk-taking behavior of financial institutions in emerging markets. Assuming that QE increases banks’ leverage, we show that the fire-sale contagion channel becomes much more pronounced. The same shock to the government bond asset class leads to higher banking sector instability. The risk to banking sector losses is not linear, but rather increases exponentially with higher leverage ratios. The second perspective of the dissertation considers spillovers between financial markets in the context of QE. We contribute to the literature that investigates the portfolio balance effect associated with QE. In essence, the portfolio balance channel is the consequence of an assumed imperfect substitutability of assets. To account for this, we develop a dynamic agent-based model to study international asset price spillover. Our two-country model features heterogeneity in assets and in investor preferences. Both are crucial for a meaningful model-based impact assessment of QE because preferences for asset maturity, asset class (bonds, equities and currencies) and whether an asset is issued at home or abroad can influence the substitutability of assets, and hence the portfolio balance effect of central bank asset purchases. We implement a novel pricing mechanism that allows us to approach market clearing prices. This allows us to take advantage of the flexibility of the agent-based methodology, while keeping the model comparable to more standard equilibrium-based portfolio balance models. We calibrate the two countries in our model to the Eurozone (EZ) and a representative sample of rest-of-the-world (ROW) countries in order to estimate the international impact of the ECB’s asset purchase program announced in January 2015. For this purpose, we compile data on asset holdings of 15 374 EZ and 25 930 ROW open-end investment funds from the Morning Star Database, as well as data on investment portfolios of EZ and ROW banks from the ECB’s Statistical Warehouse and Bankscope. When simulating our model, we find a negative impact of central bank asset purchases on both domestic and foreign returns. While the effects of QE on domestic bond yields and the exchange rate are rather modest and smaller than commonly assumed in the literature, they can cause domestic stock prices increase substantially. Somewhat surprisingly, however, we find that spillovers from portfolio balancing to the rest of the world are negligible.

Keywords

Economics

Reference:

Collections

PhD / Doctoral

Full item page