K. Rao Kadiyala

Numerical Methods for Estimation and Inference in Bayesian VAR-Models

In Bayesian analysis of vector autoregressive models, and especially in forecasting applications, the Minnesota prior of Litterman is frequently used. In many cases other prior distributions provided better forecasts and are preferable from a theoretical standpoint. Several of these priors require numerical methods in order to evaluate the posterior distribution. Different ways of implementing Monte Carlo integration are considered. It is found that Gibbs sampling performs as well as, or better, then importance sampling and that the Gibbs sampling algorithms are less adversely affected by model size. We also report on the forecasting performance of the different prior distributions

Risk Measurement for Event-Dependent Security Returns

Financial economists often test security pricing models for the destabilizing effects of informational uncertainty. In particular, Bar-Yosef and Brown (1977) examined the market model and concluded that systematic risk is concave during stock-split event periods. Unsystematic risk, however, was assumed constant. Further steps are taken in this article. A generalization of the random coefficients model of Hildreth and Houck (1968) is conducted, and case-by-case testing for event-period concavity of both systematic and unsystematic risk is conducted. In contrast to past findings, the results support concavity for unsystematic risk, not systematic risk.

Measuring investment performance with a stochastic parameter regression model

Abstract This paper develops a generalization of the Hildreth and Houck (1968) random coefficient model for use in evaluating the macroforecasting ability of portfolio managers. Most studies have assumed that astute managers ignore the strength of the market trend and create a binary portfolio beta with one value during up markets and a lower value during down markets. In contrast, this paper examines a model in which the superior manager adjusts beta period by period according to changing market conditions. The findings indicate that portfolio managers are not superior macroforecasters.

Forecasting foreign exchange rates in developing economies

We investigate the ability of a variety of exchange rate models to forecast parallel exchange rates for developing economies. In contrast to earlier studies, which use actual values of the exogenous variables, we employ time series forecasts of the exogenous variables. An error correction version of a monetary model proposed by us, that incorporates the dynamics of both short-run and long-run adjustment processes, outperforms all other models that have been suggested earlier.

Asymptotic probability concentrations and finite sample properties of modified LIML estimators for equations with more than two endogenous variables

This paper investigates the distributional properties of a class of modified limited information maximum-likelihood (LIML) estimators. It is shown that the asymptotic distributions of these estimators are more concentrated than those of the modified LIML estimators suggested by Fuller. Additionally, the results of an extensive Monte Carlo investigation of the finite sample properties of the proposed estimators show that when the equation of interest has more than two endogenous variables, the LIML estimator is often highly inefficient so that substantial gains in precision are realized by using the modified estimators in place of the LIML estimator.

Alternative tests for heteroscedasticity of disturbances:a comparative study

This paper presents three small sample tests for testing the heteroscedasticity among regression disturbances. The power of these tests are compared with two of the leading tests for this hypothesis, one by Goldfeld and Quandt [5] and the other by Theil [17]. We also provide a heuristic method of selecting the number of middle observations to be deleted for the Goldfeld-Quandt type of tests.

The estimation of missing observations in related time series data: further reuslts

Earlier work has provided an efficient method for the prediction of missing data in a dependent variable series using a system of grouped regression equations. This paper extends the previous literature in two ways. First , a test statistic capable of indicating the advantage of the grouped procedure is derived. Second, it is demonstrated through an empirical application that the most prevalent methodology used for examining the impact of financial economic events is a special case of the missing data estimation problem.

The estimation of missing observations in related time series data: Further results

Earlier work has provided an efficient method for the prediction of missing data in a dependent variable series using a system of grouped regression equations. This paper extends the previous literature in two ways. First, a test statistic capable of indicating the advantage of the grouped procedure is derived. Second, it is demonstrated through an empirical application that the most prevalent methodology used for examining the impact of financial economic events is a special case of the missing data estimation problem.