Accounting for good news and accounting for bad news: Some empirical evidence from the Czech Republic

Abstract

This paper is motivated by the links that continue to be forged between security pricing and accounting, building on recent findings that firms tend to be asymmetrically conservative in the timeliness of earnings recognition. The evidence is that firms in the European Union tend to recognise unrealised losses more quickly in their earnings than unrealised gains (Giner and Rees, 2001; Raonic et al., forthcoming), and there is evidence of even greater accounting conservatism in the USA (Basu, 1997; Ball et al., 2000; Givoly and Hayn, 2000). This paper investigates whether the Czech market exhibits conformity with the behaviour that has been documented elsewhere by examining the earnings/returns relationship, focusing to begin with on the impact of losses on earnings response coefficients and then considering the asymmetric timeliness of income recognition in the Czech market. The findings indicate that the Czech market is similar to more developed markets, at least in one respect: there is statistically significant evidence of different market effects of profits and losses, in that profits are more persistent than losses. However, contrary to the findings in more developed markets, there is no statistically significant evidence of earnings conservatism in the Czech market. These results are most probably due to the continuing influence of restrictive tax regulations that mitigate any tendency towards conservatism, as well as the transitional nature of the economy. A further reason is likely to be that the regulatory environment in the Czech Republic is close to the kind of stakeholder corporatism that is described by Ball et al. (2000), who show that conservatism tends to be less pronounced in such regimes where there are fewer managerial incentives to bias current earnings. In conclusion, if changes in market prices signal good news and bad news about future risky outcomes, there is no evidence of asymmetry in the Czech market in accounting for such risks.

Notes

1. Conservatism in accounting as described by Basu (1997) is partially explained by the risk adverse behaviour of corporate accountants. The speed with which bad news is recognised in earnings differs from the speed of good news recognition. As previously recognised by Hayn (1995), the slope of the earnings response regression changes with negative earnings which is consistent with the transitory nature of corporate losses.

2. Timeliness reflects the strength of the link between earnings and returns, as the speedy recognition of the bad news that is appreciated by the market implies cautiousness on the part of corporate accountants. Greater conservatism implies risk aversion and risk-responsible accountability.

3. The greatest change in Czech accounting took place in 1991 when the new Act on Accounting came into effect. Since then there have been several frequent amendments (1994, 1997, 2000, 2001 and 2002) as well as by-laws that address issues of detail, but no further fundamental changes. A new Act on Accounting comes into effect in 2004. This research does not make ex post adjustments with regard to these legislative changes.

4. At this time, the situation is not entirely uniform. Companies that are listed have to prepare IAS accounts and submit them to the Securities Commission. Other companies are not obliged to do so, although all companies have been obliged to report according to Czech standards. This does not have an impact on the data-set used in the present study.

5. Subsequently, about 700 new companies were listed following the second wave of voucher privatisation in 1995.

6. The efficiency of the Czech capital market has been empirically tested (e.g. Sommer, 1996; Filacek et al., 1998). Such studies have concentrated on different aspects of market efficiency and are not congruent. However, the results mostly imply that the market is not efficient, with low turnover and a lack of transparency. Although this may have an impact on the results of our study, since we use only end-of-the-year data, low turnover will not affect the results.

7. Administrative changes positively affect data availability, but not the results. Indeed, it is worth noting that such changes did not stem the outflow of international capital investment (Fact Book – Prague Stock Exchange, 2002, pp. 32–41).

8. The univariate model has the form R _t  = α + β X _t /P _t − 1 + ϵ _t , where R _t is the return over a one-year period, X _t is the earnings per share (EPS) for the year t, and ϵ _t is a normally distributed error term with zero mean and constant variance.

9. In a previous study on the Czech market (Jindrichovska, 2001), the models did not distinguish between profits and losses.

10. Note that asymmetry in the timeliness of incorporating economic events in reported earnings may be measured in different ways, not only the incremental response to bad news relative to good news but also the relative sensitivity of earnings to bad and good news and the relative explanatory power of negative and positive return regressions.

11. To improve the explanatory power of this model to explain more of the earnings variance, the introduction of lagged returns on the right-hand side of the model would be necessary.

12. The ratio (β ₁ + β ₂)/β ₁is also considered as a measure of the relative sensitivity of earnings to bad news compared with their sensitivity to good news (Givoly and Hayn, 2000). This ratio is expected to be greater than one under conservative reporting, but its magnitude should be interpreted with care given that zero is the expected value of β ₁ when accounts are not timely.

13. This sample represents data on all industrial companies that were available simultaneously in both databases on the Czech market.

14. During testing, we also excluded these 12 observations from our sample, but the impact was not material.

15. This result is not in line with other markets. During the period under investigation, negative returns in the market were far more frequent than negative earnings yields (175 versus 44) – see Tables 1 and 2. Therefore, much of the bad news (negative returns) corresponds contemporaneously with reported profits rather than reported losses. This causes a negative sign on the slope coefficient.

16. We control for influential values of the observed variables by weighting out of the regression those observations with high Cook statistics (>2.25 for this sample). As a result, five unduly influential observations (three loss makers and two profit makers) were omitted from the regression. The ERCs remain consistent with the full data, increasing from 0.219 (t = 2.43) to 0.380 (t = 3.46) overall, from − 0.341 (t = −2.63) to − 0.334 (t = −1.84), in the case of losses, and from 0.855 (t = 4.89) to 0.997 (t = 3.21) in the case of profits, with a change in significance level only in the case of losses, where the degree of confidence decreases from 95 to 90%, i.e. still 10% significant. Note, however, that the coefficient of determination is still the largest in the case of the loss-making sub-sample.

17. Controlling again for observations that have a significant influence on the estimation (see n. 16 above), the timeliness coefficient β ₂ in Model (2) of 0.084 (t = 2.43) for the full data is relatively unchanged at 0.081 (t = 3.14) when unduly influential observations are weighted out. The estimations from Model (3) are also robust in this respect, with the conservatism measure β ₂ remaining insignificant: β ₂ is 0.045 (t = 0.37) for the full data and 0.051 (t = 0.53) for the reduced data.

18. We also excluded all negative earnings yield observations from the sample and tested the timeliness model (2) separately on the sample of positive earnings yield (totalling 173 observations) partitioned into good and bad news. The estimated slope coefficients are positive and statistically significant in both cases. However, the coefficient of determination shows that the model is now less well specified.