Predicting future firm performance is important but also challenging for academics, financial analysts, and the wider public. From an accounting perspective, it is an interesting question if current financial statements provide information on future firm success. This can be helpful for analysts as well as answer in part, the question of whether specific accounting information is forward-looking or not. This study focuses on specific accounting information on tax loss carryforwards that has often been criticized due to its complexity and compliance burden. In addition, the study accounts for empirical evidence that profits are more persistent than losses. From this perspective, current losses might be less helpful for performance predictions when compared to current profits.
Recent studies have explored whether accounting information regarding deferred taxes from tax loss forward and negative firm performance can improve predictions of future firm performance. Existing research provides evidence of a significant association between financial reporting information on deferred tax assets and future firm performance that increases the explanatory power or forecasting regressions. In addition, broad empirical evidence also suggests a higher persistence of positive than negative performance.
Models can perform very differently in in-sample and out-of-sample testing. Even the “true” explanatory model can underperform in predictions, since the minimization of expected prediction errors is not equivalent to the minimization of explanatory bias.Sebastian Eichfelder
Is accounting information on losses helpful for predictions?
Corresponding to International Accounting Standard (IAS) 12.34, deferred tax assets from tax loss carryforwards are only recognized to the extent that the tax benefit’s realization is likely. This is only the case if the firms generate sufficient future (taxable) profits to offset tax loss carryforwards. It has been argued that firms can use additional deferred tax assets from tax loss forward to signal positive future firm performance.
Considering the lower persistence of losses compared to profits, it seems more likely that negative firm performance is a weaker signal. Transitory losses are a primary reason for that. Due to business cycles, economic shocks, restructurings, and similar issues, firms might be able to transform their current losses into future profits. Thus, positive performance outcomes are more persistent and have higher predictive validity than losses.
Information on deferred taxes and tax loss carryforwards not helpful
This study uses a unique hand-collected panel of 835 observations of International Financial Reporting Standards (IFRS) accounts of firms listed on the German stock market. Unlike the US Generally Accepted Accounting Principles (GAAP), the tax footnote of IFRS accounts contains mandatory details on the amount of unrecognized (i.e., the nonvaluable component of) tax loss carryforwards. Since this information should be based on a firm’s internal estimate of future taxable earnings, it could be a helpful predictor of future pre-tax earnings, post-tax earnings, and cash flows.
This study also analyzes the usefulness of voluntarily disclosed accounting information on tax loss carryforwards (i.e., the total amount of tax loss carryforwards, the book value, and changes in valuation allowances for deferred tax assets from tax loss carryforwards) for performance predictions. The analysis confirms previous findings suggesting a negative association between unrecognized (deferred taxes from) tax loss carryforwards and future firm performance in in-sample tests. However, out-of-sample tests, including a battery of robustness checks, reveal that such items typically reduce predictive validity.
A theoretical explanation for the findings of this study relates to model overfitting, where the model overfits its training data, capturing unstable relations. This problem is especially relevant for noisy predictors with potential measurement errors. Regarding unrecognized tax loss carryforwards (ULCFs), there are three main reasons for measurement error. First, ULCFs result from the internal estimates of managers that, by themselves, can be subject to forecasting error. Second, previous empirical research already suggests that ULCFs are used for earnings management, which reduces the accuracy of that information. Third, differences in tax and financial accounting induce additional measurement errors if ULCFs are used for the prediction of consolidated earnings and cash flows.
How to enhance performance predictions by loss information
This study presents robust empirical evidence that common forecasting approaches that treat positive and negative performance similarly overestimate the persistence of current negative firm performance. This holds especially true for long-run prediction horizons, increasing the likelihood of loss reversal. Considering differences in the persistence of negative and positive current performance in regressions significantly increases the explanatory power and predictive validity. In additional out-of-sample tests, the analysis provides mixed evidence for standard proxies of persistent and transitory losses. Overall, prediction models can easily be enhanced by adding an interaction term of current performance and an indicator variable for negative or positive performance.
This research has three main takeaways. First, while in-sample tests typically find a significant association between deferred taxes from tax loss forward with future tax payments and firm performance, out-of-sample tests show that such items typically worsen predictions. This holds even for predictions of after-tax cash flows, suggesting the limited usefulness of deferred tax components in predicting cash taxes.
The findings align with recent research indicating a notable correlation between stock prices and deferred tax items, but not of stock prices and deferred tax assets from tax loss carryforwards. The overall evidence suggests that IFRS information on tax loss carryforwards is not helpful for performance predictions. Thus, for auditors and analysts, these items might be of limited relevance, and standard setters might think of oversimplifying corresponding standards (e.g., IAS 12) to reduce financial reporting costs.
Second, there is strong and robust evidence that considering the asymmetry in the information content of negative and positive performance with an interaction term of a dummy for negative performance and current performance enhances predictions. Thus, this study is a simple way to improve predictions that can be helpful for analysts, auditors, investors, managers, and researchers. As a more general implication, this research documents that negative performance contains less information than positive performance as it is less persistent.
Third, the analysis documents that models can perform very differently in in-sample and out-of-sample testing. Thus, researchers, as well as practitioners, might need different types of models for explanatory analysis (i.e., models that want to explain the causes of a phenomenon) and predictive analysis (i.e., models that want to predict future developments). Thus, before choosing a specific model, researchers and practitioners should clearly define their targets. While in-sample tests are an appropriate statistic for explanatory models, predictive models clearly need out-of-sample testing to identify the best-performing model.
Dreher, S., Eichfelder, S., & Noth, F. (2023). Does IFRS information on tax loss carryforwards and negative performance improve predictions of earnings and cash flows?. Journal of Business Economics, 1-39. https://doi.org/10.1007/s11573-023-01147-7