Accounting standards permit management to use its discretion in determining how certain components of earnings are classified, or where they are reported in the financial statements. Prior to 2001, Australian accounting standards required companies to separately classify and report any unusually large items of income and expense as “abnormal items” to distinguish such items from “normal” earnings. In the U.S. such items are referred to as “special items” or “unusual items”. The term “abnormal items” was subsequently removed from the accounting standards, a move in part driven by the perception of regulators and shareholder interest groups that companies were opportunistically classifying items as abnormal to enhance the reported earnings before abnormal items in order to meet earnings benchmarks. However, there is little empirical evidence on whether earnings were classified as abnormal to manage earnings. The U.S. evidence on special items suggests managers opportunistically classify such items to enhance reported “normal” earnings and that special items are associated with classification shifting for the purpose of meeting or beating earnings benchmarks. McVay (2006) found that U.S. firms opportunistically shift expenses from “core” expenses by reporting them as special items. Some limited evidence of Australian companies engaging in earnings management by opportunistically classifying normal earnings items as abnormal was provided by Cameron (1998) and Cameron and Gallery (2001). This thesis extends prior research to examine whether Australian companies used abnormal items to manage earnings to meet earnings benchmarks. The predictions in this thesis are based on U.S. research findings that special items are asymmetric with a bias to negative special items and that special items are used to: manage the change in earnings from the prior year, reduce the current year’s earnings levels, minimise earnings or “take a bath”, and avoid reporting losses and earnings forecast errors. A sample of Australian top-500 firms for the seven-year period from 1994 to 2000 is used to test the following predictions. (1) The frequency of companies reporting abnormal items increased over the study period. (2) There is asymmetry in reporting frequency and magnitude of negative and positive abnormal items, with a bias to negative abnormal items. (3) The incidence and magnitude of reported abnormal items are associated with three earnings benchmarks, which are: level of current period earnings, annual change in earnings, and analysts’ earnings forecasts. The assumption that firms align earnings before abnormal items to analysts’ earnings forecasts is also empirically tested. Although the prediction of an increase in the frequency of reporting abnormal items over the study period is not supported, the magnitude of abnormal items increased significantly. The findings support the prediction of asymmetry in both the incidence and magnitude of abnormal items, showing that the frequency of reporting negative abnormal items and their magnitude is consistently greater than positive abnormal items over the study period. Logistic regression results indicate that firms with higher levels of earnings and changes in earnings were more likely to report abnormal items. The results of multiple regression analysis reveal that although there is no association between the amount of abnormal items and earnings levels, the significant negative relationship between earnings changes and abnormal items and the fact that abnormal items are, on average, negative, suggests that for firms that have a decline in earnings over the previous year, the lower the level of earnings before abnormal items, the greater the magnitude of negative abnormal items. This infers “bath-taking” behaviour. Partitioning the sample according to the sign of the earnings level, the sign of the earnings change and the sign of abnormal items yields significant results, suggesting firms have differing incentives to manage earnings depending on whether the firm reports a profit or a loss, or an increase (decrease) in the change in earnings. Findings infer that profit firms that raised positive abnormal items did so to maximise current period profits, whereas loss firms reporting positive abnormal items did so to minimise or smooth reported losses, or to shift from reporting a loss before abnormal items to reporting a profit after abnormal items. The significant positive relationship for the positive earnings change sub-sample reporting positive abnormal items suggests positive abnormal items were used to maximise the earnings change, which has the effect of maximising net earnings. On the other hand, the significant negative relationship for the negative earnings changes subsample that raised positive abnormal items did so to dampen the bottom line negative change in earnings and thereby smooth net earnings. The significant positive association for the negative earnings change sub-sample raising negative abnormal items infers negative abnormal items were used to maximise the negative earnings change which could also reflect “big bath” behaviour. The significant negative relationship for the positive earnings change sub-sample raising negative abnormal items indicates negative abnormal items were used to dampen the earnings change and smooth net earnings. Finally, the results show that firms which met or beat analysts’ forecasts were more likely to report abnormal items than firms which missed the forecast. Moreover, the finding of associations between analysts’ forecast errors and abnormal items, for the full sample and sub-samples, indicates that reporting negative abnormal items had the effect of shifting losses from earnings before abnormal items. Thus, earnings before abnormal items are more closely aligned with analysts’ earnings forecasts. The results from this thesis represent a comprehensive analysis of abnormal items in the context of earnings management, providing fresh insights into the earnings management phenomenon. This study also extends U.S. literature by including income-increasing abnormal items in the analysis. Collectively, the results suggest that firms used the classification of earnings as abnormal items for earnings management purposes. The findings suggest that the standard setters may have been justified in removing the “abnormal” classification from the accounting standard. However, it cannot be assumed that all firms acted opportunistically in the classification of items as abnormal. They may have done so to signal the transitory nature of the item. Removal of discretion to use such a signalling device may limit the scope for firms to effectively communicate information about the nature of items presented in financial reports.