Abstract: | We examine the usefulness of tax allocation accounting (deferred tax) for predicting future tax paid and future tax expense. Deferred taxes increase the explanatory power (R2) of regression models where future taxes paid or future tax expense is the dependent variable. However, the mean out‐of‐sample forecast errors for tax paid (future tax expense) is 30 (45.5) percent. Deferred tax increases predictive ability on pooled data, but is inconsistent on a year‐by‐year basis. We examine three explanations for poor predictive ability: losses, tax changes and asset growth. We discuss the policy and practical implications of our findings. |