Abstract
This paper surveys both the theoretical and the empirical archival literature on conservatism when accounting information is used for debt contracting. The theoretical literature shows mixed results whether conservative accounting is desirable, which depends on the underlying agency problem, the information available, and the contracting space. The empirical literature takes a more holistic view in measuring the degree of conservatism. It studies a broad array of possible effects of conservatism in debt financing, but also beyond. The results overwhelmingly support the view that conservatism plays a useful role in debt contracting, although there are also some mixed results. We describe key results and empirical designs, and we provide suggestions for future research.
Acknowledgements
We thank the organiser, Edward Lee, the sponsor of the Symposium, Accounting and Business Research, and two anonymous reviewers for helpful comments.
Disclosure statement
No potential conflict of interest was reported by the authors.
ORCID
FERNANDO PENALVA http://orcid.org/0000-0002-5206-3754
ALFRED WAGENHOFER http://orcid.org/0000-0001-5919-942X
Notes
† This paper is based on our contributions to the ‘Accounting and Business Research Special Issue Symposium: The Role of Accounting Information in Debt Markets’ at the Annual Congress 2017 of the European Accounting Association.
1. Dickhaut, Basu, McCabe, and Waymire (Citation2010) even suggest ‘that the ultimate explanation for the Conservatism principle may derive from how gains and losses are differentially processed by neurons within the human brain’ (p. 244).
2. Further examples are discussed in Barker and McGeachin (Citation2013) and Wagenhofer (Citation2014).
3. This is a limitation not imposed by conditional conservatism but by current accounting standards, which prevent the use of fair value for most assets with the exception of financial assets. But even in the case of financial assets, Badia, Duro, Penalva and Ryan (Citation2017) show that firms are conditionally conservative in the measurement of the fair value of Level 2 and Level 3 financial assets, where there is more uncertainty.
4. It is worth noting that the use of accelerated depreciation among U.S. firms has experienced a sharp, monotonic decline over the last decades (Jackson, Liu, and Cecchini, Citation2009). This trend persists until 2018.
5. DeFond et al. (Citation2016) use the proxies of unconditional conservatism developed by Penman and Zhang (Citation2002), which are based on capitalised and amortised research and development and advertising expenditures.
6. For a more detailed survey see Ewert and Wagenhofer (Citation2011).
7. Gigler, Kanodia, Sapra, and Venugopalan (Citation2009) provide a definition for continuous signals in the same spirit. Gao (Citation2013) suggests a more elaborate procedure with two steps: step 1, in which transactions are recorded and differentially verified, and step 2, in which management biases the report by earnings management.
8. Such disclosure models have become known as Bayesian persuasion games. See Kamenica and Gentzkow (Citation2011).
9. Jiang and Yang (Citation2016) show a similar result.
10. Francis and Martin (Citation2010) argue that conservatism induces managers to abandon poorly performing projects earlier. If a manager does not benefit from continuing the project (B = 0) then it is not clear why biasing information for managerial decision making should be beneficial.
11. The reverse result holds as well: If the net present value were negative, then a conservative accounting system would be optimal. But in this case the firm would not invest in the first place.
12. Beyer (Citation2012) considers aggregation of signals and finds that conservatism is beneficial when the amount of debt financing is high because it improves efficient liquidation.
13. In another asset substitution model, Burkhardt and Strausz (Citation2009) find that less information, e.g., pure historical-cost based accounting, is preferable to more information, e.g., lower-of-cost-or market accounting.
14. Neither Gao and Liang (Citation2018) nor Laux (Citation2017) explicitly consider accounting biases.
15. See also Guay and Verrecchia (Citation2006).