 |
4/2007 |
The very idea of systemic risk taking is at the same time extremely interesting and disturbing. It is interesting, because it refers to the possibility where individuals interact in a financial environment that induces them to coordinate in undertaking insolvency risk. Such coordination in risk taking could be the outcome of financial liberalization, for example. The notion of systemic risk taking is, on the other hand, disturbing as it involves the possbility of an economy following a risky growth path which, in turn, leads to aggregate financial fragility and occasional crises. Although much of the research on systemic crises and growth takes the experience of middle income or emerging economies as the key motivating factor of the analysis , which is understandable given the perceived weak institutions to manage the informational frictions present in financial sector of these economies, it is conceivable that similar mechanism is present in more advanced economies. It may be the case that the more advanced economies do not suffer from similar contract enforceability problems as the emerging economies do, but expectational biases in the form of optimism of news shocks could well place an economy on a risky growth path, which involves well above the long-term average growth of the GDP and very high and even accelerating credit growth. This question is of extreme importance not only to central banks but also to financial supervisors and by itself clearly merits further research. What is of equal importance in the present context is how resilient do we think the financial system of the more advanced economies is in coping with the possibility of extreme, adverse macroeconomic shocks that bear the seeds of a crisis. Do we have tools to analyze the implications on the financial system of such extreme macroeconomic scenarios? Do we have room for methodological innovations in the relevant research field? |
| |
| So what can we say about the state-of-the-art of macro stress-testing methodologies. One reassuring observation is that substantial progress has been made both in the econometric analysis of financial soundness indicators and in the simulation of value-at-risk measures to assess system-wide vulnerabilities. However, a number of methodological challenges still remain concerning the correlation of market and credit risks over time and across institutions. Further research in this area should also focus more on how to think about and implement macro stress-testing in the context of dynamic stochastic general equilibrium models, since DSGE modelling is the approach followed by an increasing number of central banks to macromodelling and monetary policy analysis and, clearly, dynamic general equilibrium analysis is needed to fully understand the interaction between growth, financial instability and, ultimately also crises. In their forthcoming Bank of Finland discussion paper "GDP at Risk in a DSGE Model: An Application to Banking Sector Stress Testing" Esa Jokivuolle, Juha Kilponen and Tero Kuusi provide the first set of simulation results on generating highly adverse macroeconomic scenarios from a dynamic stochastic general equilibrium model of the Finnish economy to be used in stress tests for the parts of the financial system, for example the banking sector. As they rightly note a DSGE model provides a systematic approach to generting coherent macro scenarios which, in principle, can be given a rigorous economic interpretation. |
| |
| The approach the authors follow is to implement stochastic simulations of a DSGE model, which is done by selecting a set of key exogenous variables of the models, running Monte Carlo simulations on their joint stochastic processes and simultaneously solving for the new equilibrium of the underlying DSGE model to give the paths of the endogenous macrovariables of interest. Parameters of the distribution of the vector of exogenous variables are partly estimated and partly calibrated using empirical moments in the calibration exercise. In the second step of their approach, the authors insert the simulated paths of the relevant macrovariables into banks' loan loss model to generate paths for aggregate loan losses. The exercise also keeps track of the simulated paths of those macrovariables not appearing in the banks' loan loss model, since this information is needed in interpreting the results in economic terms. One limitation of this exercise is that the DSGE model and the bank loan loss model are separate in that no feedback mechanism is considered. Naturally, the ultimate goal in this context should be to incorporate a nontrivial role for money and financial intermediation in a DSGE model, but as the authors rightly point out, this is a very challenging task, primarily for future research and, cosequently, their approach should perhaps best seen as a pragmatic first step in organising the use of a DSGE model in financial stability analysis. A second possible limitation of the approach emerges from simulating a linearized DSGE model together with a linear loan loss model to construct the distribution of loan losses. As the authors discuss in their paper, this probably shows up most clearly the shape of the simulated loan loss distribution, which turns out to be much more symmetric, even normal, than what the stylized facts about the distribution of banks' loan losses would suggest. Also, the simulated loan loss distribution is not very dispersed, with the worst case (five-year average) loan loss provisions amounting to approximately 1.5 % of total loans. The simulation approach followed by Jokivuolle, Kilponen and Kuusi is, limitations notwithstanding, an extremely interesting and important one, both from the point of view of methodology and substance. For a central bank using a DSGE model in its economic analysis and forecasting it is particularly natural to utilize it also in financial stability analysis. This helps bringing the two central areas of central bank analysis closer together and facilitating their mutual communication. Consequently, further research work in this area should be encouraged, particularly the type of research that helps incorporating banks and financial intermediation in dynamic general equilibrium models. |