On the power and interpretation of panel unit root tests

Karlsson, Sune; Löthgren, Mickael

doi:10.1016/s0165-1765(99)00237-2

Cited by 174 publications

(78 citation statements)

References 9 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Their tests control for cross-sectional dependence using demeaned data. The IPS test has low power in panels with small T (Karlsson and Lö thgren 2000), while the reverse is true for the Hadri test (Barhoumi 2005). As shown in Table 5, the IPS generates similar results to the LLC concerning the order of integration.…”

Section: Panel Unit Root and Cointegration Testsmentioning

confidence: 69%

Aluminum Consumption and Economic Growth: Evidence from Rich Countries

Jaunky

2012

Nat Resour Res

View full text Add to dashboard Cite

The article attempts to test the aluminum consumption-economic growth nexus for 20 rich economies for the period 1970-2009. Various panel data unit root and cointegration tests are applied. The series are found to be integrated of order one and cointegrated, especially after controlling for cross-sectional dependence. Moreover, the Blundell-Bond system generalized methods-of-moments is employed to conduct a panel causality test in a vector errorcorrection mechanism setting. Unidirectional causality running from aluminum consumption to real GDP is uncovered in the short-run, while real GDP is found to Granger-cause aluminum consumption in the long-run. Moreover, a 1% increase in real GDP generates an increase of 0.44% in aluminum consumption in the long-run for the whole panel.

show abstract

Section: Panel Unit Root and Cointegration Testsmentioning

confidence: 69%

Aluminum Consumption and Economic Growth: Evidence from Rich Countries

Jaunky

2012

Nat Resour Res

View full text Add to dashboard Cite

show abstract

“…In particular, it has been pointed out that the null hypothesis may be rejected even if a small fraction or, at the limit, one single series in the panel is, in fact, stationary (e.g. Taylor and Sarno, 1998;Karlsson and Lothgren, 2000). Thus, even when the joint panel unit-root null is rejected, further investigating which cross-section units within the panel display stationary behaviour remains relevant.…”

Section: Separating Stationary From Non-stationary Seriesmentioning

confidence: 99%

Current account sustainability in advanced economies

Lanzafame

2013

The Journal of International Trade & Economic Development

View full text Add to dashboard Cite

This paper investigates the sustainability of current accounts in advanced economies, using a panel of 27 countries and annual data over the 1980-2008 period. We find strong evidence in favour of nonlinear but stationary current-account trajectories for 14 countries, while the remaining 13 appear to be nonstationary and, thus, unsustainable. Our analysis indicates that careful empirical modeling of current-account dynamics, particularly in relation to crosssection dependence and nonlinear behaviour, is crucial for appropriate economic policymaking. JEL classification: C33

show abstract

“…By contrast, the LL2 test tends to over-reject the null hypothesis, and the problem worsens as N increases. Finally, it seems to be affected by a rise in T more than in N (see Im et al, 1997, Maddala and Wu, 1999, Karlsson and Lothgren, 2000. Karlsson and Lothgren (2000) have shown that, in general, the power of panel unit root tests (LL1, LL2, t-bar and LR-bar) depends on N, the number of series in the panel, T, the time series dimension in each individual series, and the proportion of stationary series in the panel.…”

Section: Panel Unit Root Testsmentioning

confidence: 99%

“…Finally, it seems to be affected by a rise in T more than in N (see Im et al, 1997, Maddala and Wu, 1999, Karlsson and Lothgren, 2000. Karlsson and Lothgren (2000) have shown that, in general, the power of panel unit root tests (LL1, LL2, t-bar and LR-bar) depends on N, the number of series in the panel, T, the time series dimension in each individual series, and the proportion of stationary series in the panel. For a given proportion of stationary series in the panel, the power increase due to a rise in T is larger than that resulting from a corresponding increase in N. This means that the probability of rejecting the null hypothesis increases with T. Consequently, for large T we may reject the null even if it is not true.…”

Section: Panel Unit Root Testsmentioning

confidence: 99%