Following previous studies (Aboody et al., 2005; Francis et al., 2004; Francis et al., 2003; Myers et al., 2003; Van der Meulen et al., 2007),this study measured the quality of earnings using the accrual quality model introduced by Dechow and Dichev (2002) with some modifications as suggested by McNichols (2002). The original Dechow and Dichev (2002) model is based on the assumption that accruals maps into cash realisations, thus any measurement errors in accruals not realised in cash flows indicate low quality of accruals. The modifications suggested by McNichols (2002) adjust for changes in firms economic environment, resulting in a better proxy for earnings quality (Jaggi et al., 2009).
According to Dechow (1994) and Dechow and Dichev (2002), the accrual-based earnings solve the potential timing and matching problems associated with the use of cash flows as a short-term performance measure. In other words, the function of accruals is to adjust the recognition of cash flows over time, so that it better reflects firm performance. However, as accrual-based earnings include measures that are subject to estimations, the earnings figure could be manipulated. This will results in lower quality of earnings.
The Dechow and Dichev (2002) model relates total current accrual (TCA), to lagged, current and future cash flows from operations. It is based on the assumption that
direct or indirect control interest. This also includes their investment through nominees' institutions and others means.
current accruals are estimates of future year‟s cash flow realisation. The accruals are temporary adjustments that delay or anticipate the recognition of realised cash flows. Thus, the quality of accrual depends on the precision of these estimates. In this model, total current accrual is measured by changes in working capital, since related cash-flow realisations generally occur within one year, which is as follows:
TCAi,t = α0 + α1,iCFOi,t-1 + α2,iCFOi,t + α3,iCFOi,t+1 + i,t (4.1)
where TCAi,t is the total current accrual21 of firm i in year t scaled by average assets,
CFOi,t is the cash flow from operations of firm i in year t scaled by average assets and
i,t is the residual of firm i in year t.
The Dechow and Dichev (2002) model captures both intentional and unintentional accrual estimation error by management, which is the inverse measure of earnings quality (Hermanns, 2006). To distinguish the unintentional accrual estimation error from the intentional, McNichols (2002) suggests future research to incorporate some variables from the Jones (1991) model, the change in revenues (REV) and property, plant and equipment (PPE), as additional explanatory variables in the Dechow and Dichev (2002) model. McNichols (2002) highlights that the combination of these models provide a more powerful approach to estimation of accrual quality in the presence of management discretion. This is because the inclusion of REV and PPE
21
Total current accrual is measured as changes in working capital. It is also equivalent to the changes in current assets minus changes in current liabilities, minus changes in cash and plus with changes in short term debt.
would control for unintentional estimation errors, which are related to firm and industry characteristics. The accrual quality model proposed by McNichols is as follows:
TCAi,t = α0 + α1,iCFOi,t-1 + α2,iCFOi,t + α3,iCFOi,t+1 + α4,iREVi,t + α5,iPPEi,t + i,t (4.2)
where REVi,t is the change in revenue of firm i in year t scaled by average assets,
PPEi,t is the gross property, plant and equipment of firm i in year t scaled by average
assets, and all other variables are as previously defined.
According to Dechow and Dichev (2002), accrual quality is an inverse function of the standard deviation of firm i‟s estimated residual. This measure is based on the time-
series expectation of accruals behaviour, in which a larger standard deviation of the residuals implies lower accrual quality. Based on this measure, a firm that has consistently large residuals and thus small standard deviation of residuals is considered to have relatively good accrual quality because the accrual behavior is predictable. However, time-series estimation of residuals requires firms to have sufficient time-series data points. For example, Dechow and Dichev (2002) require firms to have at least eight years of annual data to be included in the sample. Since the current study uses sample from an emerging capital market, this requirement would result in selecting only few established firms that are not representative of the population of firms in the capital market. In addition to survivorship bias, there are other problems related to time-series estimation. As noted by Saleh et al. (2005), the
serially correlated residuals (Peasnell, Pope, and Young, 2000); the coefficient estimates on the change in revenue (ΔREV), and the property, plant and equipment (PPE) variables in the modified model are unlikely to be stationary over time; and there may be confounding effects in the estimation period that are not related to intentional estimation error (Dechow et al., 1995; Guay, Kothari, and Watts, 1996).
To avoid the problems related to time-series estimation of accrual quality, the current study uses a cross-sectional version of Dechow and Dichev (2002). Accrual quality is measured as the absolute value of residuals from industry-specific yearly regression of the modified Dechow and Dichev (2002) model, rather than the standard deviation of residuals22. In other words, this study focuses on the magnitude rather than the volatility of measurement errors. This is based on the intuition that larger residuals from the cross-sectional estimations of the model represent lower accrual quality.
Using the model in equation (4.2), the regressions are run cross-sectionally for each industry and year based on the Bursa Malaysia industry classification, to estimate the accrual quality (AQUALITY) values for each firm from the year 2003 to 2008. The regression is performed in each industry portfolio to minimise the effect of variation of accrual behavior in different industries, as well as the different impact that the economy may have on different industries. Following Francis et al. (2005), this study requires a minimum of 20 firms in each industry group to produce good estimates of
22 Consistent with Baxter and Cotter (2009) and Srinidhi and Gul (2007), I do not use the standard deviation of the
residuals from my cross-sectional industry model, as this would provide a measure of earnings quality across all companies in the industry group rather than just the company of interest.
residuals from the regressions. These regressions generate six yearly residuals for each firms, i,t, which represents measurement errors in total current accruals
(Dechow and Dichev, 2002)
For an easier interpretation of the results of this study, the absolute value of the residuals estimated from the modified Dechow and Dichev (2002) model are multiplied by negative 1 to represent earnings quality (AQUALITY), so that higher value reflects better quality of earnings.