Conservatism in residual income models: theory and supporting evidence

Abstract

In this paper, we develop a framework for evaluating the impact of conservative accounting on the structure of residual income models of equity valuation. We explore specific examples of both unconditional and conditional conservatism and observe a common mathematical structure. We proceed to generalise our model and identify the joint dependency of conservatism and the persistence of abnormal earnings on the weights attached to book values, earnings and dividends. We are able to show theoretically the likely numerical impact of conservatism on price-earnings ratios and under-valuations produced by residual income models. We investigate empirically the interaction between conservatism and persistence and find they accord well with the theory developed. We briefly discuss the implications of testing the effect of conservatism on valuation and linear information dynamics.

Keywords:

• accounting conservatism
• equity valuation
• residual income models
• linear information dynamics

JEL classification:

• M41
• G12

Acknowledgements

The authors acknowledge many helpful comments received from participants from AAA, EAA and BAA annual meetings, Manchester University, Bristol University, Strathclyde University, Loughborough University and Imperial College London for the early versions of the paper. We would also like to thank James Ohlson, Ken Peasnell, Martin Walker, Kenton Yee and Bharat Sarath (AAA discussant) for their constructive suggestions. This paper has also benefitted from the comments of two anonymous referees as well as from an associate editor and Vivien Beattie (the editor).

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes

1. This does not undermine the use of the classification of conservatism in Beaver and Ryan (2005) as a conceptual framework for discussing accounting policy or the empirical classification for event studies such as carried out by Ball and Shivakumar (2005) and Basu (1997).

2. We do not consider an ‘other information’ variable here. Accounting conservatism is a long-run property and ‘other information’ in the standard residual income-based valuation model is generally regarded as behaving as a stationary auto-regressive process and, therefore, it is not expected to be critical to valuation in the long run.

3. Alternatively, we could interpret this part of our analysis as comparing two identical firms save for differences in accounting policies, with references to cash flows referring to only the equity portion of these.

4. This structure is more general than it may appear at first sight. We can regard as an average or composite rate of depreciation over different classes of assets, both tangible and intangible. For example, may include a contribution from the expensing of intangibles investments such as R&D or advertising. Note that here is equivalent to (a policy parameter which determines the depreciation rate) in Feltham and Ohlson (1996, p. 215). They state that ‘the overdepreciation case (the cash receipts persistence parameter) is of interest because it induces conservative accounting'. In our analysis, we do not assume the cash inflow dynamics. Therefore, we do not have parameter,. Instead, we directly assume two policy parameters, and , in both biased and unbiased accounting systems.

5. Value is determined purely by investment policy and the model is thus consistent with dividend displacement (Miller and Modigliani 1961). We also assume, as do most residual income models, that risk is held constant.

6. Hughes et al. (2004) examine the value relevance of accounting variables and characterise the impact of inflation on the weights that attach to the accounting items. They are primarily concerned with the question as to whether one can adjust depreciation policy in an inflationary environment to produce an unbiased valuation model. Ashton et al. (2011) examine a special case where and g = 0.

7. It may lead to an improved contracting, decreased litigation risk and reduced information asymmetry (Watts 2003).

8. While lagged book value may capture the size effect of a firm in , Equation (17) in fact describes the stochastic nature of return on equity.

9. In the case of being distributed as a zero mean normal with standard deviation , .

10. Strictly should carry a time subscript, since both forms of conservatism are asset or income specific and hence time dependent. However, provided that the firm's policy is applied consistently across asset classes and the long-run asset mix remains on average constant, we can interpet as summarising the long-run policy of the firm.

11. Yee (2005) assumes a mean-reverting process of abnormal earnings. The conservative adjustment term in his model is a non-zero constant rather than time-varying as in Equation (30).

12. Ashton and Wang (2013) get round this issue by using a variant of Lintner's (1956) dividend smoothing model to forecast .

13. This measurement is also consistent with that given in the P D Leake Lecture to the ICAEW (Barker 2014).

14. We note that the ‘other information’ variable plays a role of minor adjustment in valuation due to the difficulty of specification in prior empirical studies.

15. Adjusted dividends in merged CRSP and Compustat file are ordinary and return-of-capital dividends, adjusted using the price adjustment factor.

16. Our main results are similar when we use a cost of capital of 9% and 15% in our analysis.

17. This may violate the clean surplus accounting assumption. However, it eliminates potentially confounding effects of one-time items. We use reported earnings as a proxy, not only because dirty surplus accounting is not a first order concern, but empirical evidence also shows that the residual income valuation model is robust to dirty surplus earnings (Isidro et al. 2006, Heinrichs et al. 2013).

18. The results (not reported here) are very similar when we use mean P/B or median P/B as the measure of conservatism. Note that there is no commonly accepted measure of accounting conservatism. For example, with respect to the recently developed C-score in Khan and Watts (2009), Ashton and Wang (2013) argue that C-score ‘is really a measure of propensity to follow a conservative accounting policy.’

19. Our results appear to contradict those of Chen et al. (2014). Although they agree that pricing multiples are affected by both persistence of earnings and conservatism, with conservative firms having a greater degree of persistence in earning, they do not model and test the interaction between the persistence of earnings and conservatism. A glance at Table 2 shows that persistence plays a greater role in determing the earnings weights than conservatism.

20. Although Lo and Lys (2000) argue that size or scale differences across firms may explain the observed ‘anomalous’ valuation of dividends for a sample of US firms, scaling variables in the Ohlson (1995) model by lagged market values does not solve the puzzled sign on dividends for European markets (Goncharov and Veenman 2014).

21. In this exercise we set the cost of capital at 12%, . However, we find that the value of R is not crucial and results based on costs of capital of 9% and 15% are very similar.