Portal de Periódicos da CAPES

Balance sheet versus earnings conservatism in Europe

2004; Routledge; Volume: 13; Issue: 2 Linguagem: Inglês

10.1080/0963818042000203347

1468-4497

Juan Manuel García Lara, Araceli Mora Enguídanos,

Fiscal Policies and Political Economy

Abstract In this study we extend prior research on the international analysis of accounting conservatism (Joos and Lang, 1994; Ball et al., 2000; Giner and Rees, 2001), by examining the level of accounting conservatism across eight European countries (United Kingdom, Germany, France, Switzerland, the Netherlands, Italy, Spain and Belgium), and assessing the statistical significance of the differences among them. The definitions of conservatism that we use are, on the one hand, the Feltham and Ohlson (1995) definition, which implies a persistent understatement of book value of shareholders' equity (balance sheet conservatism). On the other hand, we use the one proposed by Basu (1997), that is, a timelier recognition of bad news in earnings relative to good news (earnings conservatism). We also address the possible scale problems of the models used to measure balance sheet conservatism. Finally, we check whether our comparative results could be influenced by a different sample composition in each country. Our results show that there are both balance sheet and earnings conservatism practices in all countries under study. In addition, while continental countries show larger balance sheet conservatism, differences in earnings conservative practices between countries are not that pronounced, although they tend to be larger in the UK. We also find that the existence of balance sheet conservative practices is associated with reduced levels of earnings conservatism, which is consistent with the results in Pope and Walker (2003). This paper is part of the Harmonia programme of research into accounting harmonization and standardization in Europe, and we acknowledge the financial contribution of the European Commission through the Human Potential Programme, contract HPRN-CT-2000-00062. It is also a result of the Marie Curie programme (University of Manchester), funded by the European Commission, contract HPMT-CT-2000-00067, as well as of the research project 'La convergencia contable como factor de competitividad empresarial en el mercado bursátil globalizado: el papel de los aspectos institucionales y empresariales en los países de la Unión Europea' funded by Ministerio de Ciencia y Tecnología (Spanish Ministry of Science and Technology), SEC2002-04608-C02-02. The paper has benefited from presentations at the first Harmonia conference, held in Madrid in June 2001, at the 25th European Accounting Association annual congress, held in Copenhagen, in April 2002, at the Harmonia workshop held in Lancaster, in September 2002, at a CentER seminar at Tilburg University, in January 2003, and at the 5th Aidea Giovani conference, held in Milan in July 2003. We acknowledge the comments and suggestions by Philip Brown, Willem Buijink, Begoña Giner, Stuart McLeay, Ken Peasnell, Peter Pope, Maarten Pronk, Bill Rees and two anonymous referees, as well as those by all the members of the Harmonia programme. Notes Facultad de Economía, Departamento de Contabilidad, Avenida de los Naranjos s/n, 46022, Valencia, Spain. E-mail: Araceli.Mora@uv.es 1As Basu (Citation1997) points out, Interpretation 14 (FASB, Citation1976) seems to be an exception to this rule. 2It would be possible to understate earnings consistently if the firm is growing. See Zhang (Citation2000, p. 133). 3Basu (Citation2001) identifies two types of earnings conservatism. The first type responds to the firm's commitment to allocate the original acquisition cost over the life of an asset in a certain pattern, without regard to concurrent information about changes in asset value (very similar to the Feltham and Ohlson, Citation1996, definition of conservatism). The second one focuses on the asymmetric recognition of news in earnings, which is the definition used by Basu (Citation1997), Ball et al. (Citation2000) or Giner and Rees (Citation2001). When we talk in the paper about earnings conservatism we focus also on the asymmetry in news recognition in earnings. 4Note that in the literature we find also a third type of conservatism, the Feltham and Ohlson (Citation1996) conservatism type, resulting from the rapid expensing of operating assets. However, since the differences between market and accounting numbers driven by this type of conservatism are merely temporary, we would include it just as a type of earnings conservatism. For empirical evidence on this type of conservatism see Ahmed et al. (Citation2000). 5We use 'code-law-based' or 'continental' interchangeably throughout the paper. We refer to an accounting regime where financial institutions are the main providers of capital funds. In our sample, we include Germany, France, Italy, Belgium, Spain and Switzerland as code-law or continental countries. 6See also Zhang (Citation2000) for an analytical development of the relationship between balance sheet and earnings conservatism. 7Basu (Citation1997) also argues that the increase in the timeliness of earnings over cash flows is greater for bad than for good news, that negative earnings changes have greater tendency to revert in the following period than positive earnings changes, and that 'good earnings news' (positive earnings changes) leads to larger announcement period abnormal returns than negative earnings changes. 8Gigler and Hemmer (Citation2001) argue that this asymmetry can also be driven by voluntary information provided by the firm. 9See Kothari et al. (Citation1988) and Skinner (Citation1994). 10Results of comparative studies on earnings conservatism can be affected by differences in income smoothing (see Leuz et al., Citation2003). 11Basu et al. (Citation2001) show that those firms audited by the Big Five (Six, Eight) auditing firms are more earnings conservative. Pope and Walker (Citation2003) analyse how the understatement of shareholders' equity, as reflected in the market-to-book ratio, is affecting the measures of earnings conservatism. They find, for the US, that when the market-to-book ratio is small, earnings conservatism is more pronounced. Ryan and Zarowin (Citation2003) analyse the relationship between earnings and returns, studying both the existence of earnings conservatism and of the prices leading-earnings phenomenon. Helbok and Walker (Citation2001) analyse the implication of earnings conservatism on analysts' earnings forecasts. Beekes et al. (Citation2003) analyse whether earnings conservatism measures vary with the composition of the board of directors, and similarly, Huijgen and Lubberink (Citation2001a) link earnings conservatism with managerial incentives. Garrod and Valentincic (Citation2001) argue that the asymmetry can be explained by shareholders' liquidation option. Giner and Grambovas (Citation2001) analyse the existence of earnings conservatism in ten European countries showing that the degree of earnings conservatism is larger in case of monetary crisis. Raonic et al. (Citation2003), focusing on European interlisted firms, argue that differences between European countries can be explained by several institutional factors. Chandra et al. (Citation2001) examine conservatism in US high-tech firms. Finally, Huijgen and Lubberink (Citation2001b) compare Dutch firms cross-listing in the US with those that do not, finding the former more conservative. 12Although in a completely different setting (the United States, early twentieth century and the railroads industry), Sivakumar and Waymire (Citation2003) find evidence inconsistent with this hypothesis. 13We also replicate all our analyses eliminating outliers (first and last percentile, on a country basis) from the residuals of the regressions we undertake. It is likely that through this robustness test we obtain coefficients of the parameters closer to their real values. However, our results are not sensitive to the exclusion of these observations. 14We include in the continental group Germany, France, Italy, Spain and Belgium. We do not include the UK and the Netherlands, typically classified as common-law-based countries (see Nobes, Citation1983). For balance sheet conservatism we also exclude Switzerland from the continental group. If we focus on the descriptive statistics in Table 1, we can see that there is a larger variability of all variables in Switzerland, to the extent that if we include Switzerland in the continental group, Swiss observations would behave as outliers, and drive the results for the whole group. 15Using bottom-line earnings we are closer to maintaining the clean surplus relation, which Feltham and Ohlson (Citation1995) models assume. Additionally, we are also consistent with the definition of earnings that we find more appropriate to measure earnings conservatism. 16As in Feltham and Ohlson (Citation1995), we assume a linear relationship between price and book value and earnings. Although empirical analysis of US data, i.e. Burgstahler and Dichev (Citation1997), suggests that this relation is non-linear, this is beyond the scope of our study. Nevertheless, we believe that further research targeting this particular misspecification of the model should be encouraged. 17Our results obtained from applying the Fama and McBeth (Citation1973) analysis should be interpreted with caution. The analysis by Fama and McBeth (Citation1973) assumes that the relations between variables are stationary over time. However, in our theoretical analysis we assume that Western European countries are converging under the EU directives. Also, the aggregate market-to-book ratio increases systematically over time in most countries. Basu (Citation2001) shows as well serious concerns about the results of the annual regressions of Fama and McBeth, when used in capital-markets-based accounting research. 18To avoid the creation of artificial scale problems that could even lead to the non-linearity of the model, we do not include, following Kothari and Zimmerman (Citation1995), those observations where the deflator (price at the beginning of the period) is smaller than one euro. In Kothari and Zimmerman (Citation1995) they exclude all observations where the deflator is smaller than three dollars. 19We use, as Pope and Walker (Citation1999) or Giner and Rees (Citation2001), the rate of return exclusive of dividends. However, we replicate all our analyses including dividends, obtaining fairly similar results. 20Giner and Rees (Citation2001) find that earnings conservatism decreases as firm size increases. This is consistent with the results in Basu et al. (Citation2001). See Basu (Citation2001) for several explanations for this result. 21Greater persistence of earnings is one of the possible explanations as to why the earnings multiple is smaller in continental countries compared to the UK. Other explanations include, for instance, differences in growth (although this is not supported by our empirical evidence) and risk. Results in other papers, like Ball et al. (Citation2000) or Leuz et al. (Citation2003), suggesting greater income smoothing in code-law-based countries are not necessarily contrary to our results. Income smoothing leads to greater earnings persistence, but this does not necessarily imply that earnings persistence in continental countries will be greater than in the UK. 22We also check whether the total bad news effect (β2 + β3) is statistically significant. For earnings conservatism to exist, it is not only necessarily a significant incremental sensitivity of bad news with respect to good news, but also that the total effect of bad news is statistically significant. We analyse this extent through a Wald test, finding that the total bad news response coefficient is significantly positive in all countries. 23In the case of the pooled regression, if we compare the UK with the group of continental countries and exclude Spain and Switzerland from the analysis, the significance of the difference between the UK and the continental group would increase. But still, we could not strongly argue that there are significant differences, particularly taking into account that when we use Fama and McBeth regressions the differential coefficient of bad news in continental countries with respect to the UK is close to zero. 24This result is not reported, but is available from the authors upon request. 25This is contrary to our hypothesis (and also to what is hypothesized in previous studies), but not exactly contrary to the results in previous literature. Ball et al. (Citation2000) choose the US as reference country, and find the US significantly more conservative than France and Germany (and also more conservative than the UK), but they do not test the significance of the differences between the UK, France and Germany. Also, Giner and Rees (Citation2001) do not find significant differences among the countries in their sample (i.e. UK, France and Germany) as they explicitly point out in page 1327. Additional informationNotes on contributorsAraceli MoraFootnote Facultad de Economía, Departamento de Contabilidad, Avenida de los Naranjos s/n, 46022, Valencia, Spain. E-mail: Araceli.Mora@uv.es