The possibility to compensate taxable profits with taxable losses from prior years is important in explaining firms’ tax incentives, tax planning, and tax aggressiveness. The tax loss carryforward (TLCF) is the total amount of taxable losses from the past that can be used to offset future taxable income. Data on this TLCF are, however, often missing in Compustat or not available at all for several countries. In our article “Estimating and imputing missing tax loss carryforward data to reduce measurement error” in the European Accounting Review (https://doi.org/10.1080/09638180.2021.1924812), Malte Max, Eelke Wiersma and I propose a method to estimate and impute missing TLCF data for research purposes. In the article, we show that these imputations are good estimates of missing TLCFs, we show the relevance of using these more accurate data to obtain more reliable research results, and we provide access to the algorithm, code, and data to facilitate the use of this imputation by other researchers
Our estimation method estimates the TLCF if it is missing based on earlier available TLCF information, an estimation of taxable income, and the applicable tax loss carryforward and -backward rules. Most research papers assume the TLCF to be zero when it is missing. According to our estimation, this is incorrect in about 63% of the cases with missing information. Indeed, comparing our estimated values with information in 10-K forms, or using our algorithm to estimate values that are available in Compustat, shows that it provides good estimates of the TLCF.
The amount of missing values implies that assuming it to be zero in earlier research leads to measurement error. Using our estimated values reduces this measurement error largely. We illustrate this by re-analyzing parts of the studies by Frank et al. [(2009) Tax reporting aggressiveness and its relation to aggressive financial reporting. The Accounting Review, 84(2), 467–496] and Dyreng et al. [(2017). Changes in corporate effective tax rates over the past 25 years. Journal of Financial Economics, 124(3), 441–463]. A specific risk is that firms with missing information are classified as tax aggressive while lower tax payments are actually a result from the fact that they had a TLCF. Dyreng et al. (2017) document a negative time trend in firms’ effective tax rates over the last 25 years. Because over the last decades relatively more firms have a TLCF, we expect and show that part of this downward trend is explained by measurement error in the TLCF variable. Indeed, 18% of the time trend disappears only because of using our estimated TLCFs instead of assuming missing TLCFs to be zero.
In conclusion, in tax studies, and especially tax aggressiveness studies, it is important to properly control for relevant tax variables of which the TLCF is a very crucial one. Assuming it is be zero when it is missing is inaccurate, leads to measurement error, and may even yield incorrect conclusions. We therefore suggest to try to reduce this measurement error in all future tax studies as much as possible. While measurement error will remain, every step forward is a step in the right direction and our results suggest that using our estimations is a large step in that direction. To facilitate using this approach, we make the approach and data easily accessible. Data for Compustat North America that are missing can be downloaded from https://doi.org/10.34894/N9J1WE. The algorithm and Python code are shared in an online appendix to the article to allow the use of this method also in other countries and settings.
