
on Econometric Time Series 
By:  Jiti Gao 
Abstract:  In this paper, we consider some identification, estimation and specification problems in a class of semiparametric time series models. Existing studies for the stationary time series case have been reviewed and discussed. We also consider the case where new studies for the integrated nonstationary case are established. In the meantime, we propose some new estimation methods and establish some new results for a new class of semiparametric autoregressive models. In addition, we discuss certain directions for further research. 
Keywords:  Asymptotic theory, departure function, kernel method, nonlinearity, nonstationarity, semiparametric model, stationarity, time series 
JEL:  C13 C14 C22 
Date:  2012–03 
URL:  http://d.repec.org/n?u=RePEc:msh:ebswps:20126&r=ets 
By:  Marian Vavra (Department of Economics, Mathematics & Statistics, Birkbeck) 
Abstract:  The paper examines the robustness of the size and power properties of the standard nonlinearity tests under different conditions such as moment failure and asymmetry of innovations. Our results reveal the following. First, there seems not to be a direct link between moment condition failure and the power variation of nonlinearity tests. Second, the power of the tests is very sensitive to asymmetry of innovations compared to moment condition failure. Third, although we evaluate 9 nonlinear time series models using 8 standard nonlinearity tests, some nonlinear models remain completely undetected. 
Keywords:  nonlinearity testing, Monte Carlo experiments 
JEL:  C15 C22 
Date:  2012–03 
URL:  http://d.repec.org/n?u=RePEc:bbk:bbkefp:1205&r=ets 
By:  Marian Vavra (Department of Economics, Mathematics & Statistics, Birkbeck) 
Abstract:  A new version of the Q test, based on generalized residual correlations (i.e. autocorrelations and crosscorrelations), is developed in this paper. The Q test fixes two main shortcomings of the Mcleod and Li Q (MLQ) test often used in the literature: (i) the test is capable to capture some interesting nonlinear models, for which the original MLQ test completely fails (e.g. a nonlinear moving average model). Additionally, the Q test also significantly improves the power for some other nonlinear models (e.g. a threshold moving average model), for which the original MLQ test does not work very well; (ii) the new Q test can be used for discrimination between simple and more complicated (nonlinear/asymmetric) GARCH models as well. 
Keywords:  nonlinearity testing, portmanteau Q test, autocorrelation, crosscorrelation 
JEL:  C12 C15 C32 C46 
Date:  2012–03 
URL:  http://d.repec.org/n?u=RePEc:bbk:bbkefp:1204&r=ets 
By:  Marian Vavra (Department of Economics, Mathematics & Statistics, Birkbeck) 
Abstract:  In this paper we focus on the finite sample properties of the conditional least squares (CLS) method of threshold autoregressive (TAR) parameters under the following conditions: (a) nonGaussian model innovations; (b) two types of asymmetry (i.e. deepness and steepness) captured by TAR models. It is clearly demonstrated that the finite sample properties of the CLS method of TAR parameters significantly differ depending on the type of asymmetry. The behavior of steepnessbased models is very good compared to that obtained from deepnessbased models. Therefore, extreme caution must be excercised to preliminary modelling steps, such as testing the type of asymmetry before estimating TAR models in practice. A mistake in this phase of modelling can, in turn, give rise to very problematic results. 
Keywords:  threshold autoregressive model, Monte Carlo method, bias, asymmetry 
JEL:  C15 C22 C46 
Date:  2012–03 
URL:  http://d.repec.org/n?u=RePEc:bbk:bbkefp:1206&r=ets 
By:  Pesaran, Hashem (University of Cambridge) 
Abstract:  This paper considers testing the hypothesis that errors in a panel data model are weakly cross sectionally dependent, using the exponent of crosssectional dependence α, introduced recently in Bailey, Kapetanios and Pesaran (2012). It is shown that the implicit null of the CD test depends on the relative expansion rates of N and T. When T=O(N^ϵ), for some 0 < ϵ ≤ 1, then the implicit null of the CD test is given by 0 ≤ α < (2–ϵ)/4, which gives 0 ≤ α < 1/4, when N and T tend to infinity at the same rate such that T/N → к, with к ; with being a finite positive constant. It is argued that in the case of large N panels, the null of weak dependence is more appropriate than the null of independence which could be quite restrictive for large panels. Using Monte Carlo experiments, it is shown that the CD test has the correct size for values of α in the range [0, 1/4], for all combinations of N and T, and irrespective of whether the panel contains lagged values of the dependent variables, so long as there are no major asymmetries in the error distribution. 
Keywords:  exponent of crosssectional dependence, diagnostic tests, panel data models, dynamic heterogenous panels 
JEL:  C12 C13 C33 
Date:  2012–03 
URL:  http://d.repec.org/n?u=RePEc:iza:izadps:dp6432&r=ets 
By:  Jongwook Kim; Teppei Okumura 
Abstract:  We propose a stochastic process driven by the memory effect with novel distributions which include both exponential and leptokurtic heavytailed distributions. A class of the distributions is analytically derived from the continuum limit of the discrete binary process with the renormalized autocorrelation. The moment generating function with a closed form is obtained, thus the cumulants are calculated and shown to be convergent. The other class of the distributions is numerically investigated. The combination of the two stochastic processes of memory with different signs under regime switching mechanism does result in behaviors of powerlaw decay. Therefore we claim that memory is the alternative origin of heavytail. 
Date:  2012–03 
URL:  http://d.repec.org/n?u=RePEc:arx:papers:1203.5581&r=ets 
By:  LudlowWiechers, Jorge 
Abstract:  Some of the most widely used models in economics are based on variables not yet observed, and their specification depends on future observations; the theory that underpins these delivers the backward/ forward solution. We present a newly unified construction, starting with a more general specification of an ARMA model, yet is capable of delivering in closed form, in both the backward and forward cases, leading to an alternative presentation of causal/noncausal and invertible/noninvertible cases. 
Keywords:  Causal models; noncausal models; invertible models; noninvertible models; backward solution; forward solution 
JEL:  C32 C50 C22 C01 
Date:  2012–03–25 
URL:  http://d.repec.org/n?u=RePEc:pra:mprapa:37635&r=ets 