• Selected
• Entire List

Abstract:
One of the implications of the creation of Basel Committee on Banking Supervision was the implementation of Value-at-Risk (VaR) as the standard tool for measuring market risk. Since then, the capital requirements of commercial banks with trading activities are based on VaR estimates. Therefore, appropriately constructed tests for assessing the out-of-sample forecast accuracy of the VaR model (backtesting procedures) have become of crucial practical importance. In this paper we show that the use of the standard unconditional and independence backtesting procedures to assess VaR models in out-of-sample composite environments can be misleading. These tests do not consider the impact of estimation risk and therefore may use wrong critical values to assess market risk. The purpose of this paper is to quantify such estimation risk in a very general class of dynamic parametric VaR models and to correct standard backtesting procedures to provide valid inference in out-of-sample analyses. A Monte Carlo study illustrates our theoretical findings in finite-samples and shows that our corrected unconditional test can provide more accurately sized and more powerful tests than the uncorrected one. Finally, an application to S&P500 Index shows the importance of this correction and its impact on capital requirements as imposed by Basel Accord.

Backtesting; Basel Accord; Conditional Quantile; Estimation Risk; Forecast evaluation; Fixed, rolling and recursive forecasting scheme; Risk management; Value at Risk

Abstract:
A numerical approximation of the critical values of Cramér-vonMises (CvM) tests is proposed for testing the correct specification of general conditional location parametric functionals. These specifications include conditional mean and quantile models. The method is based on the estimation of the eigenelements of the covariance operator associated with the CvM test, and it has the advantage that it requires the practitioner to estimate the model only one time under the null hypothesis. A Monte Carlo experiment shows that the proposed approximation compares favorably with respect to the sub-sampling method in terms of size accuracy, power performance and computational time.

Cramér-von Mises, Principal Components, Eigenvalues, Empirical Processes

Abstract:
Backtesting methods are statistical tests designed to uncover excessive risk-taking from financial institutions. We show in this paper that these methods are subject to the presence of model risk produced by a wrong specification of the conditional VaR model, and derive its effect on the asymptotic distribution of the relevant out-of-sample tests. We also show that in the absence of estimation risk, the unconditional backtest is affected by model misspecification but the independence test is not. Our solution for the general case consists on proposing robust subsampling techniques to approximate the true sampling distribution of these tests. We carry out a Monte Carlo study to see the importance of these effects in finite samples for location-scale models that are wrongly specified but correct on average. An application to Dow-Jones Index shows the impact of correcting for model risk on backtesting procedures for different dynamic VaR models measuring risk exposure.

Backtesting, Basel Accord, Conditional Quantile, Forecast evaluation, Model Risk, Risk management, Value at Risk

Abstract:
The purpose of the present paper is to relate two important concepts of time series analysis, namely, nonlinearity and persistence. Traditional measures of persistence are based on correlations or periodograms, which may be inappropriate under nonlinearity and/or non-Gaussianity. This article proves that nonlinear persistence can be characterized by cumulative measures of dependence. The new cumulative measures are nonparametric, simple to estimate and do not require the use of any smoothing user-chosen parameters. In addition, we propose nonparametric estimates of our measures and establish their limiting properties. Finally, we employ our measures to analyze the nonlinear persistence properties of some international stock market indices, where we find an ubiquitous nonlinear persistence in conditional variance that is not accounted for by popular parametric models or by classical linear measures of persistence. This finding has important economic implications in, e.g., asset pricing and hedging. Conditional variance persistence in bull and bear markets is also analyzed and compared.

Conditional Mean, Nonlinear time series, Non- linear Persistence, Nonlinear correlograms, Persistence in variance, Bull and bear markets.

Abstract:
This article proposes a general class of joint and marginal diagnostic tests for parametric conditional mean and variance models of possibly nonlinear non-Markovian time series sequences. The use of joint and marginal tests is motivated from the fact that marginal tests for the conditional variance may lead misleading conclusions when the conditional mean is misspecified. The new tests are based on a generalized spectral approach and, contrary to existing procedures, they do not need to choose a lag order depending on the sample size or to smooth the data. Moreover, the proposed tests are robust to higher order dependence of unknown form, in particular to conditional skewness and kurtosis. It turns out that the asymptotic null distributions of the new tests depend on the data generating process, so a new bootstrap procedure is proposed and theoretically justified. A simulation study compares the finite sample performance of the proposed and competing tests and shows that our tests can play a valuable role in time series modeling. Finally, an application to the S&P 500 highlights the merits of our approach.

Abstract:
This article proposes omnibus specification tests of parametric dynamic quantile regression models. Contrary to the existing procedures, we allow for a flexible and general specification framework where a possibly continuum of quantiles are simultaneously specified. This is the case for many econometric applications for both time series and cross section data which require a global diagnostic tool. We study the asymptotic distribution of the test statistics under fairly weak conditions on the serial dependence in the underlying data generating process. It turns out that the asymptotic null distribution depends on the data generating process and the hypothesized model. We propose a subsampling procedure for approximating the asymptotic critical values of the tests. An appealing property of the proposed tests is that they do not require estimation of the non-parametric (conditional) sparsity function. A Monte Carlo study compares the proposed tests and shows that the asymptotic results provide good approximations for small sample sizes. Finally, an application to some European stock indexes provides evidence that our methodology is a powerful and flexible alternative to standard backtesting procedures in evaluating market risk by using information from a range of quantiles in the lower tail of returns.

Omnibus tests, Conditional quantiles, Nonlinear time series, Empirical processes, Quantile processes, Subsampling, Value-at-Risk, Tail Risk

Abstract:
This note proves the consistency and asymptotic normality of the Quasi-Maximum Likelihood Estimator (QMLE) of the parameters of a GARCH model with martingale difference centered squared innovations. The results are obtained under mild conditions and generalize and improve those in Lee and Hansen (1994) for the local QMLE in semi-strong GARCH(1,1) models. In particular, no restrictions on the conditional mean are imposed. Our proofs closely follow those in Francq and Zakoian (2004) for independent and identically distributed innovations.

GARCH models, Martingales, Semi-strong models

Abstract:
The paper introduces a root-n consistent estimator of the probability density function of the response variable in a nonparametric regression model. The proposed estimator is shown to have a (uniform) asymptotic normal distribution, and it is computationally very simple to calculate. A Monte Carlo experiment confirms our theoretical results, and an empirical application demonstrates its usefulness. The results derived in the paper adapts general U-processes theory to the inclusion of infinite dimensional nuisance parameters.

Density Estimation, Kernel Smoothing, U-processes

Abstract:
This paper proposes an asymptotically optimal specification test of single-index models against alternatives that lead to inconsistent estimates of a covariate's average partial effect. The proposed tests are relevant when a researcher is concerned about a potential violation of the single-index restriction only to the extent that the estimated average partial effects suffer from a nontrivial bias due to the misspecification. Using a pseudo-norm of average partial effects deviation and drawing on the minimax approach, we find a nice characterization of the least favorable local alternatives associated with misspecified average partial effects as a single direction of Pitman local alternatives. Based on this characterization, we define an asymptotic optimal test to be a semiparametrically efficient test that tests the significance of the least favorable direction in an augmented regression formulation, and propose such a one that is asymptotically distribution-free, with asymptotic critical values available from the χ 2/1 table. The testing procedure can be easily modified when one wants to consider average partial effects with respect to binary covariates or multivariate average partial effects.

Average Partial Effects, Omnibus tests, Optimal tests, Semi-parametric Efficiency, Efficient Score

Abstract:
This article investigates model checks for a class of possibly nonlinear heteroskedastic time series models, including but not restricted to ARMA-GARCH models. We propose omnibus tests based on functionals of certain weighted standardized residual empirical processes. The new tests are asymptotically distribution-free, suitable when the conditioning set is in…nite-dimensional, and consistent against a class of Pitman's local alternatives converging at the parametric rate n1=2; with n the sample size. A Monte Carlo study shows that the simulated level of the proposed tests is close to the asymptotic level already for moderate sample sizes and that tests have a satisfactory power performance. Finally, we illustrate our methodology with an application to the well-known S&P 500 daily stock index. The paper also contains an asymptotic uniform expansion for weighted residual empirical processes when initial conditions are considered, a result of independent interest.

Time series models, model speci…cation, ARMA-GARCH mod- els, S&P 500