Rice University

A Panel Cointegration Test Using Testing Instruments

Email address: yoosoon@rice.edu

Keywords: Nonstationary Panel, Time Series Econometrics

Abstract:
We construct a panel cointegration test using testing instruments. The testing instruments employed are simulated random walk processes that are transformed under bounded integrable functions. The test is simple to implement: To test for the null of panel cointegration, i.e., the presence of a cointegration in all cross-sectional units, we just need to include the testing instruments as additional regressors in individual cross-sectional regressions and jointly test for insignificance of those superfluously added testing instruments. The null hypothesis is supported if the test detects the superfluousness of the added regressors. Our test relies on the fact that the panel regression becomes authentic only under the null of cointegration. In the absence of cointegration, the panel regression becomes spurious and the test cannot detect the superfluousness of the added regressors. The test has many desirable properties and can be used for very general nonstationary panels. The test statistics has a chi-square limit distribution with degrees of freedom equal to the number of testing instruments, as long as the panels have large individual time series observations and are balanced in a very weak sense. The dimension of cross-sectional units are not restricted; it may be arbitrarily small or large, and the number of observations may differ across cross-sectional units, thereby permitting unbalanced panels. Individual cross-sectional units are allowed to be dependent through general dependency among the errors, and also allowed to have heterogeneous cointegrating relationships defined with different cointegrating vectors and variables involved. Permitted also are the regressors-errors cross correlations and the serial correlations in the errors.

PDF file of paper: Not available.

Session: Panel Data Models

Time: Saturday, 7 July, 2:15pm - 3:45pm

Room: C