Intra-industry information transfer associated with earnings releases by global airlines : a cross-country analysis

Abstract

This thesis examines the presence, magnitude, and determinants of earnings information transfer across multiple countries with distinct financial market characteristics. As information transfer is most likely to occur with respect to earnings announcements for which there is a significant market impact for the announcer itself, and because the existence of information transfers is likely to reduce the information content of earnings releases, this study integrates the study of information transfer with the study of information content. This thesis first investigates whether, for a given degree of unexpected earnings, there is differential market reaction for announcers belonging to three different accounting clusters, i.e. those following an Anglo-Saxon tradition, those following a Continental European tradition, and those coming from the Emerging Markets. Univariate analyses reveal no significant differential information content among these accounting clusters, and this result also holds in subsequent multivariate regression analysis. For the whole sample, there appears to be no clear association between the extent of unexpected earnings and average cumulative abnormal returns surrounding the earnings announcement by the releasing firm. However, a statistically significant association emerges for a sub-sample of the small announcing firms, but not for the sub-sample of large firms. Additional tests using Beaver's U indicate that while, on average, the whole sample of earnings announcements does possess information content, this result is driven by a small subset of firms, mostly small ones. Such evidence is consistent with Bamber et al. (2000) and cautions against over-reliance on tests of information content based on cross-sectional average reactions. To maximize the chance of detecting transnational information transfer, this thesis focuses on the global airline industry, which is characterized by its transnational nature of operations and high level of firm interdependence. The empirical evidence suggests that transnational information transfers, while modest in magnitude for the population of firms, are potentially more significant in sub-samples of firms with special characteristics. In addition to confirming the effects reported in Firth (1996a) of announcer's firm size, correlation of returns and cross-listing on transnational information transfer, this thesis documents that competitive rivalry between the announcer and the non-announcer is a key determinant of information transfer, with closer competition giving rise to more significant information transfers. The majority of the results are robust to alternative model specifications and measures of information content/transfer. Several implications may be suggested in relation to the evidence on transnational information transfers. Firstly, the existence of transnational information transfers (apart from regular releases of traffic statistics by global airlines) might have been responsible for the lack of significant information content associated with the whole sample of announcing airlines in general, and the sub-sample of large airlines in particular. Secondly, the existence of transnational information transfers provides evidence to suggest that current national differences in financial reporting standards are not so insurmountable as to inhibit investors from extrapolating foreign earnings information for use in the continuous and timely revaluation of domestic firms. Such use is also consistent with cross-country linkages at the industry and corporate level, at least for certain global industries. Thirdly, the generally weak evidence of transnational information transfer for the population of firms (especially when signed cumulative abnormal returns are used), vis-a-vis much stronger evidence of information transfer for close competitors, suggests that future studies must use caution in the sample selection and matching process. This also serves to illustrate the recommendation of Schipper (1990) that, if the goal is to detect a hypothesized phenomenon of interest, it may be advantageous to use a sample that is intentionally biased towards maximizing the chance of detecting the hypothesized phenomenon.