Consistent risk management: Three models outlined

This article introduces three models on how to understand the demand for consistent risk management. The first model, which accords with traditional risk analysis, is called the Standard Model. In this model, the decisive criterion of whether or not to accept a risk is if the total benefit exceeds the total cost. Since this model cannot protect the individual from unfair risk exposure two more models are outlined. The arguments in the Model of Inviolable Rights and in the Model of Procedural Justice evolve around the separateness of individuals, rights and fair risk taking. It is argued that risk management needs to acknowledge a variety of morally salient factors to avoid exposing people unfairly to risks.     

Derived factors in event studies

We examine the utility of the statistical factor model of the process generating stock returns in the context of event studies. For a variety of estimation procedures and experimental designs we find limited value added relative to the use of a simple market model. We would attribute this finding to misspecification of the statistical factor analysis model, and suspect that there exist more robust procedures for estimating the factor structure of stock returns.     

Levels of significance in event studies

Standard event-study methods compare the returns in the window with forecasted values based on a benchmark period. These methods fail to account for the fact that errors in parameter estimates in the benchmark period give serially correlated errors in the window's abnormal returns. This is true both of mean-adjusted returns and market-model adjusted returns. This oversight leads to overstatement of significance levels, the degree of overstatement varying positively with the ratio of periods in the window to periods in the benchmark period. Sometimes the overstatement is minimal, but it can be substantial if the window is large relative to the benchmark period.     

Sample selection and event study estimation

The anomalies literature suggests that pricing is biased systematically for securities grouped by certain characteristics. If these characteristics are related to selection in an event study sample, imprecise predictions of an event study method may produce erroneous results. This paper performs simulations to compare a battery of short-run event study prediction and testing methods where samples are grouped by market equity, prior returns, book-to-market, and earnings-to-price ratios. Significant statistical errors are reported for both standard and newer methods, including three- and four-factor models. A characteristic-based benchmark model produces the least biased returns with the least rejection errors in all samples.     

Statistical inference in multiperiod event studies

We examine several event-study test statistics that can be used to detect abnormal performance during amultiperiod event window. We demonstrate that one of the most commonly used test statistics does not, under the assumptions made, have the distribution claimed (standard normal), and thus tests using it will be biased. The magnitude of that bias is shown to increase with the length of the event window and can generally be expected to lead to excessive rejection of the null hypothesis. We also compare the relative power of alternative test statistics that are normally distributed and are straightforward to apply.     

Bayesian estimation of financial models

This paper outlines a general methodology for estimating the parameters of financial models commonly employed in the literature. A numerical Bayesian technique is utilised to obtain the posterior density of model parameters and functions thereof. Unlike maximum likelihood estimation, where inference is only justified in large samples, the Bayesian densities are exact for any sample size. A series of simulation studies are conducted to compare the properties of point estimates, the distribution of option and bond prices, and the power of specification tests under maximum likelihood and Bayesian methods. Results suggest that maximum–likelihood–based asymptotic distributions have poor finite–sampleproperties.     

Accounting-based versus market-based cross-sectional models of CDS spreads

Models of financial distress rely primarily on accounting-based information (e.g. [Altman, E., 1968. Financial ratios, discriminant analysis and the prediction of corporate bankruptcy. Journal of Finance 23, 589–609; Ohlson, J., 1980. Financial ratios and the probabilistic prediction of bankruptcy. Journal of Accounting Research 19, 109–131]) or market-based information (e.g. [Merton, R.C., 1974. On the pricing of corporate debt: The risk structure of interest rates. Journal of Finance 29, 449–470]). In this paper, we provide evidence on the relative performance of these two classes of models. Using a sample of 2860 quarterly CDS spreads we find that a model of distress using accounting metrics performs comparably to market-based structural models of default. Moreover, a model using both sources of information performs better than either of the two models. Overall, our results suggest that both sources of information (accounting- and market-based) are complementary in pricing distress.     

Nonparametric rank tests for event studies

Because stock prices are not normally distributed, the power of nonparametric rank tests dominate parametric tests in event study analyses of abnormal returns on a single day. However, problems arise in the application of nonparametric tests to multiple day analyses of cumulative abnormal returns (CARs) that have caused researchers to normally rely upon parametric tests. In an effort to overcome this shortfall, this paper proposes a generalized rank (GRANK) testing procedure that can be used on both single day and cumulative abnormal returns. Asymptotic distributions of the associated test statistics are derived, and their empirical properties are studied with simulations of CRSP returns. The results show that the proposed GRANK procedure outperforms previous rank tests of CARs and is robust to abnormal return serial correlation and event-induced volatility. Moreover, the GRANK procedure exhibits superior empirical power relative to popular parametric tests.     

Detecting earnings management using cross-sectional abnormal accruals models

This paper examines specification and power issues in relation to three models used to estimate abnormal accruals. In contrast to the majority of prior work evaluating models estimated in time-series, we examine the performance of cross-sectionally estimated models. In addition to testing the standard-Jones (Jones, 1991) and modified-Jones (Dechow et al., 1995) models, we also develop and test a new specification, labelled the ‘margin model’. Consistent with prior US research employing time-series specifications of the two Jones models, our findings suggest that each of the three cross-sectional models are well specified when applied to a random sample of firm-years. However, the margin model appears to generate relatively better specified estimates of abnormal accruals when cash flow performance is extreme. Analysis of the models' ability to detect artificially induced earnings management indicates that all three procedures are capable of generating relatively powerful tests for economically plausible levels of accruals management (e.g., less than 10% of lagged total assets). Regarding their relative performance, the standard-Jones and modified-Jones models are found to be more powerful for revenue and bad debt manipulations. In contrast, the margin appears to be more powerful at detecting non-bad debt expense manipulations.     

Endogenous problems in cross-sectional valuation models based on accounting information

The main purpose of this paper is to deal with the analysis of the scale effect in the value-relevance of accounting numbers. We examine the impacts of widely used deflators on the adjustment of scale effect. We find that most of the usual deflators employed in the literature generate endogeneity problems. In this paper we recommend the use of exogenous deflator such as the number of employees in market-based accounting research models. This alternative deflator produces, at least for the USA and Canada data, slightly better statistical results than other (endogenous) deflators such as the market value, the book value of equity, or the total assets.     

Evaluating cross-sectional forecasting models for implied cost of capital

The computation of implied cost of capital (ICC) is constrained by the lack of analyst forecasts for half of all firms. Hou et al. (J Account Econ 53:504–526, 2012, HVZ) present a cross-sectional model to generate forecasts in order to compute ICC. However, the forecasts from the HVZ model perform worse than those from a naïve random walk model and the ICCs show anomalous correlations with risk factors. We present two parsimonious alternatives to the HVZ model: the EP model based on persistence in earnings and the RI model based on the residual income model from Feltham and Ohlson (Contemp Account Res 11:689–732, 1996). Both models outperform the HVZ model in terms of forecast bias, accuracy, earnings response coefficients, and correlations of the ICCs with future returns and risk factors. We recommend that future research use the RI model or the EP model to generate earnings forecasts.     

On the estimation of beta-pricing models

An integrated econometric view of maximum likelihood methodsand more traditional two-pass approaches to estimating beta-pricingmodels is presented. Several aspects of the well-known 'errors-in-variablesproblem' are considered, and an earlier conjecture concerningthe merits of simultaneous estimation of beta and price of riskparameters is evaluated. The traditional inference procedureis found, under standard assumptions, to overstate the precisionof price of risk estimates and an asymptotically valid correctionis derived. Modifications to accommodate serial correlationin market-wide factors are also discussed.     

One-year estimation uncertainty in some claim development models

ABSTRACT

The paper studies the one-year estimation uncertainty associated with using credibility-based loss reserving methods, when claim development can be described by the models of Bühlmann-Straub or Hesselager-Witting. Having found a formula, it seems natural to minimise the one-year estimation uncertainty in the same way as one can minimise the ultimate uncertainty, i.e. to minimise the MSEP. It turns out that minimisation of the one-year estimation uncertainty leads to unreasonable and unsightly results. This puts into question the sanity of the concept of one-year estimation uncertainty.     

Volatility estimation via hidden Markov models

We propose a stochastic volatility model where the conditional variance of asset returns switches across a potentially large number of discrete levels, and the dynamics of the switches are driven by a latent Markov chain. A simple parameterization overcomes the commonly encountered problem in Markov-switching models that the number of parameters becomes unmanageable when the number of states in the Markov chain increases. This framework presents some interesting features in modelling the persistence of volatility, and that, far from being constraining in data fitting, it performs comparably well as other popular approaches in forecasting short-term volatility.     

Bayesian range-based estimation of stochastic volatility models

Alizadeh, Brandt, and Diebold [2002. Journal of Finance 57, 1047–1091] propose estimating stochastic volatility models by quasi-maximum likelihood using data on the daily range of the log asset price process. We suggest a related Bayesian procedure that delivers exact likelihood based inferences. Our approach also incorporates data on the daily return and accommodates a nonzero drift. We illustrate through a Monte Carlo experiment that quasi-maximum likelihood using range data alone is remarkably close to exact likelihood based inferences using both range and return data.     

Efficient estimation of drift parameters in stochastic volatility models

We study the parametric problem of estimating the drift coefficient in a stochastic volatility model , where Y is a log price process and V the volatility process. Assuming that one can recover the volatility, precisely enough, from the observation of the price process, we construct an efficient estimator for the drift parameter of the diffusion V. As an application we present the efficient estimation based on the discrete sampling with δ _n →0 and n δ _n →∞. We show that our setup is general enough to cover the case of ‘microstructure noise’ for the price process as well.      

Maximum likelihood estimation of stochastic volatility models

We develop and implement a method for maximum likelihood estimation in closed-form of stochastic volatility models. Using Monte Carlo simulations, we compare a full likelihood procedure, where an option price is inverted into the unobservable volatility state, to an approximate likelihood procedure where the volatility state is replaced by proxies based on the implied volatility of a short-dated at-the-money option. The approximation results in a small loss of accuracy relative to the standard errors due to sampling noise. We apply this method to market prices of index options for several stochastic volatility models, and compare the characteristics of the estimated models. The evidence for a general CEV model, which nests both the affine Heston model and a GARCH model, suggests that the elasticity of variance of volatility lies between that assumed by the two nested models.     

The impact of variance estimation in option valuation models

This paper examines some implications of using an estimate of the variance in option valuation models. This procedure produces biased option values. It is shown that the magnitude of this bias is not large. The dispersion induced in the option price is more significant particularly for parameter values of practical interest. The nature and extent of this dispersion is examined by numerical examples. The paper suggests how a Bayesian approach could be used to cope with the estimation error.     

Conducting event studies with Asia-Pacific security market data

We investigate the effectiveness of several well-known parametric and non-parametric event study test statistics with security price data from the major Asia-Pacific security markets. Extensive Monte Carlo simulation experiments with actual daily security returns data reveal that the parametric test statistics are prone to misspecification with Asia-Pacific returns data. Two non-parametric tests, a rank test [Corrado and Zivney (Corrado, C.J., Zivney, T.L., 1992, The specification and power of the sign test in event study hypothesis tests using daily stock returns, Journal of Financial and Quantitative Analysis 27(3), 465-478)] and a sign test [Cowan (Cowan, A.R., 1992, Non-parametric event study tests, Review of Quantitative Finance and Accounting 1(4), 343–358)] were the best performers overall with market model excess returns computed using an equal weight index.