Analysis and diagnosis of income smoothing in Spain

Abstract

This paper examines income smoothing behaviour for a sample of Spanish listed companies over the period 1991–97. Discretionary accruals are used as a measure of managers' accounting discretion. Our results strongly support the income smoothing hypothesis which predicts aggressive (conservative) accounting policies when a firm's current pre-managed performance undershoots (exceeds) target earnings, as proxied by median earnings for the firm's industry. Consistent with DeFond and Park (1997), we also provide some evidence supporting the anticipatory income smoothing hypothesis, whereby current period smoothing behaviour is partly conditioned on expected future performance.

ACKNOWLEDGEMENTS

The authors are very grateful for the helpful comments and suggestions provided by two anonymous reviewers, the Editor and José Moreira in earlier versions of this paper. We also want to thank I/B/E/S database for the information supplied, and Joan Allbraight and Victor M. Sánchez for the language revision.

Notes

¹There is no precise distinction between this concept and that of ‘earnings management’ in the literature. Until the beginning of the 1980s, the term ‘income smoothing’ was dominant but nowadays it has been displaced by the more general term ‘earnings management’, within which the former is viewed as a special case (Schipper, 1989; Apellániz and Labrador, 1995).

²Apart from aggregate discretionary accruals, other approaches used in the literature to measure managerial discretion include accounting changes (e.g. Moses, 1987; Sweeney, 1994), specific components of discretionary accruals (McNichols and Wilson, 1988) and components of discretionary cash flows (Bartov, 1993).

³Extant studies that use discretionary accruals to measure managers' accounting discretion include: Healy (1985), Jones (1991), DeFond and Jiambalvo (1994), Apellániz and Labrador (1995), Gaver et al. (1995), Holthausen et al. (1995), DeFond and Park (1997), Key (1997), Young (1998) and Kasznik (1999).

⁴Subramanyam (1996) and DeFond and Subramanyam (1998) conclude that the cross-sectional version of the Jones model obtains better results than its time-series version. Also DeFond and Jiambalvo (1994), Gaver et al. (1995), Teoh et al. (1998), Beneish (1997) and Peasnell et al. (2000) propose cross-sectional versions of these models.

⁵Our results are robust to different ways of computing discretionary accruals. We repeated the analysis using three other models: the modified Jones model proposed by Dechow et al. (1995), which consists of estimating coefficients in the same way as the Jones model, and then replacing change in revenue in equation (1) with change in revenue less change in receivables, just to compute discretionary accruals; and both the Jones model and the modified Jones model including dummy variables to represent the industries in the sample and estimating just one regression per year instead of one regression per year and industry sector.

⁶Spanish accounting rules do not require firms to disclose cash-flow information. Following extant studies (e.g. Jones, 1991; Dechow et al., 1995), total accruals (TA) are therefore computed as: TA _it  = ΔCA _it  − ΔCash _it  − ΔCL _it  + ΔDCL _it  − Dep _it where ΔCA is the change in current assets; ΔCash is the change in cash and cash equivalents; ΔCL is the change in current liabilities ΔDCL is the change in debt included in current liabilities; Dep is the depreciation and amortization expense; and the subscripts i and t represent firms and years, respectively.

⁷Law 19/1989 of 25 July, relating to Partial Reform and Adaptation of Commercial Legislation to EC Company Law Directives, established the general accounting principles to be followed by Spanish companies while the new General Accounting Plan (1990) gave detailed rules about the content of annual accounts. Companies of a certain size were also obliged to seek the auditing of annual accounts by an independent auditor (see Giner, 1993, for more details about the Spanish accounting framework). Limiting our analysis to the post-1990 period helps to ensure that financial statements were prepared under similar accounting guidelines during the sample period. Apellániz and Labrador (1995) present evidence of a reduction in the amount of earnings management carried out by managers in a group of Spanish listed firms in the period following the implementation of the General Accounting Plan of 1990 relative to the previous period.

⁸DeFond and Park use the earliest between (1) the second median forecast following management's announcement of current year earnings, and (2) the median forecast in the fourth month following the end of the current period. After collecting announcement dates for a subset of firms in our sample from the Sequencer database, we observed that those announcements are usually made between February and May. Therefore, to ensure that our forecast data were issued after managements' announcement of current year earnings, we perform the analysis using the median forecast in the fourth, fifth and sixth month after the end of the current year.

⁹Lim and Lustgarten's analysis was also repeated using analysts' forecasts as proxy for future expected performance. In all three analyses (using different forecast dates) mean and median NDA are significantly negative, at 1%, within all the cells. Again, although it can be observed that mean and median values of NDA are less negative in cell iii than in cell i, and more negative in cell ii than in cell iv, these differences are never significant when comparing cells ii and iv, and they are only significant, at 5%, when testing the difference in means between cells i and iii in two of the three analyses.

¹⁰Another indication that our findings are indeed capturing some of the predicted earnings management activity has already been pointed out by our main results. Specifically, if we assume that observed DA in cells i and iv, where fewer earnings management practices are predicted, represent an approximation of the bias introduced by the ‘backing-out’ problem, then the fact that the mean and median values of DA are significantly higher (lower) in cell iii (ii) than in cell i (iv) suggests that the initial conclusions of significant earnings management in cells ii and iii may be valid.

¹¹Leverage and logarithm of total assets are included as explanatory variables of DA since they are found to be associated with discretionary accruals behaviour in previous studies (e.g. Becker et al., 1998, and Young, 1998). The inclusion of lagged DA is based on the intuition that managers' ability to ‘borrow’ or ‘save’ earnings in one period probably depends on what they did in previous periods (DeFond and Park, 1997: 132).

¹²The analysis was also carried out with the sample using analysts' forecasts as proxy for future expected performance. The results are fairly similar in both cases.