MAFC Research Papers

First, to identify the most suitable risk models for stress testing, we apply an extensive back testing procedure that focuses on extreme market movements. We consider eight possible risk models including both conditional and unconditional models and four possible return distributions (normal, Student’s t, empirical and normal mixture) applied to three heavily traded currency pairs using a sample of daily data spanning more than 20 years. Finding that risk models accommodating both volatility clustering and heavy tails are the most accurate predictors of extreme returns, we develop a corresponding model-based stress testing methodology. Our results are compared with traditional stress tests and we assess the implications for capital adequacy. On the basis of our results we conclude that the new recommendations for market risk regulatory capital calculation will have little impact on current levels of foreign exchange regulatory capital.

The first method extends Posner and Milevsky (1998) to jump-diffusion models: a Johnson variable is found that matches the first four moments of the arithmetic average; this leads to an analytical approximation for the average rate option price. This method has been found to be extremely accurate as well as instantaneously fast when modelling asset prices as a geometric Brownian motion. In the second method, Fourier transforms are used to obtain closed form pricing for the equivalent geometric average option. This is used as an effective control variate to price the original arithmetic average rate option using Monte Carlo simulation. The Johnson variable approximation is fast and gives accurate prices when the jumps are not too large or skewed. However for parameters consistent with some observed implied volatility smiles, the approximation is less accurate. Our results suggest it is necessary to match six moments to ensure enough of the distribution of the average is captured to obtain a satisfactory approximation. Although slower, the simulation method is still quite fast and consistently gives accurate results.

Estimators of the Sharpe ratio have less helpful distributions than estimators of mean and variance. The error in the estimate of the Sharpe ratio can be simply too large to make useful conclusions. Investors are often overly swayed by historical performance in making investment decisions: why make matters worse by obscuring what information we have by combining past performance characteristics in an unhelpful ratio?

The implication of this view is that it can be difficult to distinguish between alternative strategic asset allocation benchmarks and to determine whether or not rebalancing portfolios or making tactical asset allocation tilts is statistically meaningful. This paper develops more powerful tests to show that small changes in asset allocation weights are often statistically significant, challenging those views.

However, regulatory intervention may also create distorted incentives for banks (refer Gennotte and Pyle (1991)). The provision of safety nets to the banking sector is a clear example. Protection provisions mitigate the cost of risk-taking normally borne by risky firms. In the standard model of bank moral hazard banks can increase shareholder wealth by taking on more risk.
Empirical studies suggest banks shift risk to safety net providers but fail to maximise the value of any (explicit or implicit) subsidy. Other factors act to mitigate the extent to which banks are able, or willing, to transfer wealth from the regulator to themselves.
One factor that may deter risk-taking behaviour is bank charter value. We examine the effectiveness of bank charter value as a deterrent to risk-taking in the Australian context. We find Australian banks exhibit strong, positive charter value over the period reviewed. However, charter value is found to be, generally, an ineffective deterrent to moral hazard induced risk-taking. Support for the hypothesis that capitalisation levels affect the risk/charter value relationship is found.

It also includes a simple real-world trading experiment assessing whether strategies designed to exploit the breakdown of UIP have been profitable.The study reveals a host of intriguing results, based largely on cross exchange rates not previously examined in this context. First, forward rates of industrialized countries are unambiguously biased predictors of future spot rates, as spot rates invariably change less than the amount of the forward spread. This indicates violations of UIP are ubiquitous and not simply restricted to the key currency blocks which have already been studied. Second, contrary to popular belief, the forward ‘puzzle’ anomaly – where high yielding currencies appreciate, not depreciate as UIP predicts - is not a pervasive phenomenon. Third, regression results and the trading experiment demonstrate that deviations from UIP are both size and time dependent. For instance, during points in time where interest differentials have been relatively large, UIP correctly predicts the sign, but not the magnitude, of exchange rate changes.
More recently, the breakdown of UIP has become increasingly severe as interest differentials have converged around the world. Finally, results from the trading experiment demonstrate that in contrast to US dollar based findings presented in earlier studies, strategies exploiting the forward bias have in fact been something of a money tree - deviations from UIP have been large and consistent enough to generate impressive profits and risk-return tradeoffs for simple multi-currency trading strategies.

This research is limited by the relatively short history of reliable data and the reliance on classical statistical tests that rely, in turn, on the assumption of normal returns and linear parameters. This paper develops an alternative test for hedge fund skill that addresses these shortcomings. I set up a multi-factor generalized method of moments (GMM) test and then introduce information to condition this test for the possibility that betas rise in bear market conditions. The test confirms the presence of a significant long-run hedge fund manager skill premium which, for most strategies, is robust to underlying market conditions.

The study distinguishes itself from its predecessors by paying particular attention to the mechanism driving the contemporaneous relationship between intervention and trading rule returns, in a market which has received virtually no attention in this context. In addition, by assessing the profitability of trading rule returns in several clearly defined phases of RBA intervention policy, a more comprehensive analysis of this relationship is possible relative to that reported in studies of other central banks.
The results demonstrate that whilst a simple momentum trading rule is largely unprofitable in the absence of RBA intervention, it generates very attractive returns when taking positions in the opposite direction to the RBA during periods of intervention. The most profitable opportunities occur when the RBA attempts to smooth or correct a strong trend in the $A (particularly an aggressive depreciation) over successive days. This may suggest that on occasions where the RBA ‘leans against the wind’, public resources are lost to trend-following speculators. However it is not immediately obvious that the RBA would be particularly concerned with taking a loss on its short-term intervention activities.

fixed income arbitrage and some equity hedge (technology) funds, although prior estimates of ‘instant history’ bias however, are greatly exaggerated.We extend on prior research by examining the impact of survivorship on higher moments of the distribution of hedge fund returns (volatility, skew and kurtosis) and across different hedge fund style groups. We also consider probable causes of death after trawling the extensive TASS ‘notes’ fields maintained in the TASS database for over 1000 defunct funds, referred to affectionately by TASS as “the graveyard”. We argue that many funds actually close themselves down when their net asset value drifts well below their previous high watermark for incentive fees. We also find sudden short-term losses are more likely to lead to termination than poor returns or high volatility per se. There is little evidence that hedge fund death is related to leverage, the extent of trading discretion, or whether or not managers are personally invested. We do however, find a significantly higher incidence of death among funds that use technical/trend-following processes.

Firstly, the Monte Carlo procedure demonstrates that the unconditional distribution of high frequency foreign exchange returns cannot be approximated by the unconditional distribution of returns simulated by autoregressive conditional heteroskedastic (ARCH) processes. Secondly, a range of standard ARCH models are shown to produce accurate forecasts of realized daily variance when frequently sampled data is used to generate and appraise variance forecasts. This appears to be an artefact of the noise inherent in using the daily squared return as an estimator of realized daily variance. In short, this paper demonstrates that whilst standard econometric models do not capture the intraday foreign exchange return generating process, this should not immediately preclude these models from high frequency applications. Instead, the forecasting exercise demonstrates practical benefits are easily attainable from using high frequency data to develop and evaluate existing asset pricing models.

The analysis finds very little difference in risk management practice between Hong Kong and Singapore. When these two Asian centres are compared with the United States, the survey highlights that a higher proportion of firms in Hong Kong and Singapore use derivatives and they do so with greater intensity. The reason for the popularity of derivatives may be that firms in both countries have a high proportion of revenues and costs denominated in foreign currency and they often have significant foreign currency borrowings. Firms in Hong Kong and Singapore are more likely than their US counterparts to use selective or active hedging strategies based on market views. Therefore it is worrying to note some significant gaps in the oversight of risk management in these firms, especially in relation to reporting and valuation of derivatives. This tendency to mix risk management with "speculation" runs counter to the efficient markets literature. The survey finds that only a minority of firms in Hong Kong and Singapore have implemented risk-adjusted performance measurement for treasury so it is unclear whether or not their activities add value for owners.

In the absence of adequate risk measurement techniques, the paper recommends that financial institutions should instead focus their attention on managing operational risk using an agency framework. Agency risk can be seen as the underlying cause of many operational losses. The paper examines several recent disasters to illustrate this point eg Barings, Orange County, Long-Term Capital Management and Procter & Gamble/Bankers Trust. Agency theory helps to identify situations where such losses are most likely to occur and also suggests improvements to organisational structure to minimise the possibility of losses. The paper also briefly considers the importance of character in operational risk management.

Banking regulators and governments, as the ultimate payee for any resulting contingent liability, are concerned about the extent to which banks take advantage of mis-priced deposit insurance provisions, and whether outside discipline is sufficient to offset risk-shifting incentives. In this paper, we use share price data to estimate the value and volatiltity of assets of ten Australian banks and any arising contingent claim over a recent time period. A potential bias in risk-shifting results is eliminated by testing for risk-shifting with respect to the two components of banking risk, asset risk and leverage risk. We find that while forces do act to mitigate risk-taking incentives, they are insufficent to prevent individual banks extracting a benefit over the period considered. Results also indicate the regional bank group in the sample extracted the greater benefit from the subsidy.

Researchers rely on large simulation methods to handle the complexity induced by the massive path dependence inherent in GARCH models. In this paper, we develop an efficient lattice algorithm to price European and American options under discrete time GARCH processes. We show that this algorithm is easily extended to price options under generalized GARCH processes, with many of the existing stochastic volatility bivariate diffusion models appearing as limiting cases. We establish one unifying algorithm that can price options under almost all existing GARCH specifications as well as under a large family of bivariate diffusions in which volatility follows its own, perhaps correlated, process.

Researchers rely on large simulation methods to handle the complexity induced by the massive When a limit is reached trading stops and the equilibrium price is not observed. The statistical problem is to maximize the likelihood function by replacing the unobservable squared error terms with their expected values. We evaluate the performance of the algorithm by extensive simulation and apply it to Treasury bill futures data from a period of high volatility and frequent limit moves.

The study also highlights the pitfalls of using the traditional Fixed Window method to estimate variance/covariance for asset allocation decisions. The implications for asset allocation decisions, hedging decisions and international diversification strategies are examined.

Corporate bond markets in the region are generally of minor importance, but a widely held view is that they have potential to contribute more to economic development. The analysis here suggests that efficient bond markets improve the performance of the financial sector and the economy in several ways.

The risk measures considered include the traditional fixed window method, exponential smoothing and GARCH (in increasing order of complexity). Our findings confirm that the choice of risk measure can make a significant difference to the efficiency of asset allocation decisions and therefore to investment outcomes and fund ranking. We find that the traditional fixed window method is rarely optimal and that exponential smoothing is the most attractive risk estimation method in most cases. However beyond this, added complexity does not necessarily improve investment efficiency.