PLACE:  319 Snedecor

SPEAKER:
Helle Bunzel
Department of Economics, Iowa State University

TITLE:
Testing of Trend Models in the Presence of Serial Correlation and Heteroskedasticity

ABSTRACT:

This paper introduces a test statistic that can be used to test hypotheses
 about the parameters of the deterministic trend function of a univariate time
 series.  The test is robust to serial correlation and conditional
 heteroskedasticity of unknown form and is valid for both I(1) and I(0) errors.
 The test proposed in this paper eliminates the need to estimate the correlation
 structure, and hence removes an important source of size distortions. The
 development of the new test relies upon a data-dependent transformation of the
 ordinary least squares estimates of the parameters.  The asymptotic
 distribution of the transformed estimates depends only on the parameters of the
 deterministic trend function, and therefore a test statistic which is invariant
 to the specific form of the correlation structure can be obtained.  Extensive
 simulation experiments investigate the properties of the new test statistic in
 finite samples.  These demonstrate that the new test compares well both with
 traditional tests employed in the literature and more recently developed test
 statistics which also do not require estimates of the correlation structure.
 

COFFEE: 3:45 p.m., 104 Snedecor