Given a risk measure, the main problem is that of risk management. Its twin
facets of the accurate estimation of input parameters and the correct development
of hedging strategies are both addressed in the other two papers in this issue.
In the first article by Jokivuolle and Peura, we are reminded that the global
financial crisis has demonstrated that a bank’s financial distress can start well before
an actual default: at a fairly high although deteriorated rating level. To avoid the
adverse consequences of such a distress, the bank should hold enough capital not
only against default – that is, economic capital – but also to support, at least, a
minimum non-distressed rating at all times. This requires much higher capital buffers
than standard economic capital models typically imply. In their article, Jokivuolle
and Peura argue that such minimum rating targeting may have explained many
banks’ actual capital levels; at least in relation to the risks banks, regulators and
rating agencies were able to anticipate and measure in the years preceding the crisis.
The authors provide a two-stage simulation-based framework in the context of a
corporate credit portfolio, in which the amount of capital needed to support a desired
minimum target rating can be measured.
The second article, by Yu et al, deals with risk measurement. The VaR concept
has been widely adopted by financial regulators all over the world for designing
capital adequacy standards for banks and financial institutions. In addition, financial
firms have adopted VaR for internal risk management and the allocation of resources.
However, its potential failure at correctly measuring diversification has led to an
increase in an alternative, namely ES, or expected loss beyond VaR. In their article,
Yu et al propose a non-parametric, kernel-based approach that mitigates the occurrence
of bias in the estimation of tail distribution. They exploit the representation
of ES as an integral of the quantile function to develop a one-step kernel estimator.
In a comparison with existing kernel estimators, they conduct a Monte Carlo study
that supports the substantial improvement in the accuracy and efficiency provided by
their technique.
The calculation of accurate futures hedge ratios is important for the practice
of risk management. While the calculation of such ratios using naïve one-to-one
or simple regression approaches are often preferred due to their simplicity, the
more involved generalized autoregressive conditional heteroskedasticity (GARCH)
approach is often reported to be more accurate. The third paper, by McMillan and
Garcia, examines the performance of hedge ratios constructed using the realized
volatility approach, which is simple in construction but grounded in a rigorous
theoretical base. Through constructing hedged portfolios, the paper examines the
performance of the simple regression, rolling simple regression, GARCH and
realized hedge ratio measures. The results support the view that, based on minimizing
the portfolio variance, the static regression, rolling regression and GARCH-based
methods perform well; however, this is often at the expense of negative mean
values. Portfolio performance that takes into account both mean and variance using
the Sharpe ratio almost unanimously supports the portfolios constructed using the
realized hedge ratios. A final issue that remains, however, is that the realized hedge
ratio itself is volatile such that any benefits in portfolio construction may be negated
by the transaction costs required. Therefore, future research could examine ways,
such as smoothing, to allow the gains to be realized.
The fourth and final article, by Maller et al deals with parameter estimation
in the context of portfolio diversification. The Sharpe ratio is one of the most
common and also most important measures of the return–risk ratio of a portfolio. In
practice the ratio must be estimated from returns data, and it is well known that the
corresponding sampling error will transmit to the Sharpe ratio itself. Generalizing
earlier analyses, Maller et al obtain in their paper for the first time, under very
general conditions and in a definitive and highly usable form, the large-sample
distribution of the estimated maximal Sharpe ratio. This distribution represents the
spectrum of possible optimal return–risk tradeoffs that can be constructed from the
data ex ante. There are many possible uses of it; a particular example, focusing on
the question of whether it is better to select, ex ante, a suboptimal portfolio from
a large class of assets or to perform a Markowitz optimal procedure on a subset of
the assets, is discussed. The authors illustrate applications of the theory by analyzing
a large sample of US companies, giving a comparison of constant correlation and
momentum strategies with the optimal strategy. Simulations based on this data are
also given for illustration. |
