4.2.2. The Value Relevance and Earnings Predictability of Non-GAAP and Equivalent GAAP Earnings: Conditioning on Firm-Level Characteristics
The analyses in the preceding section suggest that there are substantial differences between non-GAAP and equivalent GAAP earnings in terms of value relevance and earnings predictability. In this section, we investigate whether such differences can vary across various firm-level characteristics, such as market value of equity, accruals quality, analyst coverage, and managerial ability of a firm, and report the results in
Table 4,
Table 5,
Table 6 and
Table 7, respectively. Like the format in
Table 3, Panels A and B of each table report results for the value relevance and the earnings predictability tests, respectively. It is worth noting that the total observations within each cross-sectional regression vary due to the availability of the data for each firm-level characteristic.
Table 4 repeats the analysis of
Table 3 but splits the sample into two sub-samples based on whether a firm has a market value of equity above or below the median market value of equity in the sample each quarter. The market value of equity can either positively or negatively affect a firm’s earnings quality. On one hand, firms with a higher market value of equity tend to make income-decreasing adjustments to their earnings when facing greater regulatory scrutiny, but such income-decreasing adjustments can negatively affect the quality of earnings (
Watts & Zimmerman, 1986). On the other hand, the market value of equity can also be a proxy for information uncertainty since firms with a lower market value of equity are less diversified and have less available information for the market (
Zhang, 2006). It is thus plausible that firms with a lower market value of equity can more aggressively inflate their earnings due to higher internal control deficiencies or lower disclosure preparation costs (
Zhang, 2006;
Doyle et al., 2007).
Panel A of
Table 4 reveals that the value of the coefficient on GAAP earnings (
GAAP Earnings per share) increases from 2.473 in the low-market-value-of-equity sub-sample to 18.09 in the high-market-value-of-equity sub-sample. Similarly, the value of the coefficient on non-GAAP earnings (
Non-GAAP Earnings per share) increases from 12.18 in the low-market-value-of-equity sub-sample to 40.90 in the high-market-value-of-equity sub-sample. The result is consistent with the findings from extant literature that firms with a high market value of equity report more value relevant non-GAAP and equivalent GAAP earnings than firms with a low market value of equity (
Zhang, 2006;
Doyle et al., 2007). Interestingly, the magnitude of the difference for the coefficients between non-GAAP and equivalent GAAP earnings is larger in the high-market-value-of-equity sub-sample, although the significance level of the difference remains identical (
p < 0.01) in both sub-samples. Further, the analysis of the adjusted R
2 confirms that the overall explanatory powers of the regression in columns (2) and (4) are higher compared to the corresponding values in columns (1) and (3), thus suggesting that the difference in the value relevance between non-GAAP and equivalent GAAP earnings is not completely mitigated by the market value of equity.
Panel B of
Table 4 also confirms the findings from Panel A that the value of the coefficients on both GAAP earnings (
GAAP Earnings) and non-GAAP earnings (
Non-GAAP Earnings) is higher in the high-market-value-of-equity sub-sample than the value of the corresponding coefficients in the low-market-value-of-equity sub-sample. Panel B, however, documents a slightly different result to the one in Panel A that the magnitude of the difference for the coefficients between non-GAAP and equivalent GAAP earnings is larger in firms with a low market value of equity than in firms with a high market value of equity, although the difference remains highly statistically significant (
p < 0.01) in both sub-samples.
Overall, the empirical evidence from
Table 4 reveals that the market value of equity of a firm can alter but cannot completely mitigate the phenomenon that non-GAAP earnings are more value relevant and have higher predictive power towards future operating earnings of a firm than equivalent GAAP earnings.
Table 5 repeats the analysis in
Table 3 but with separate estimates for firms with above and below median values of accruals quality in a year. When a firm has lower quality accruals, the firm is normally viewed as reporting lower quality GAAP earnings. It is reasonable to assume that if a firm reports lower quality GAAP earnings, investors may rely more heavily on non-GAAP earnings to assess the value of the firm, and the non-GAAP earnings of the firm are therefore more relevant to the equity value. Similarly, lower quality GAAP earnings may contain a lot of non-recurring “noisy” items and thus have lower predictability towards future operating income or cash flow. It is thus plausible to assume that non-GAAP earnings have better predictability towards the future operating earnings under this circumstance. However, if managers of a firm intentionally report low-quality GAAP earnings to manipulate earnings, the same group of managers could have reported low-quality non-GAAP earnings as well. Thus, it is an empirical question to examine the conditioning effect of accruals quality on the value relevance and earnings predictability of the non-GAAP and equivalent GAAP earnings.
To create the proxy for accruals quality, we adopt the method from
Dechow and Dichev (
2002) by considering accruals as a function of past, present, and future cash flows in a regression model. Additionally, we incorporate the thoughts of
McNichols (
2002) by including in the regression model the change in sales and the level of property, plant, and equipment. Specifically, we estimate the following cross-sectional regression by each industry based on the 2-digit SIC code and year.
where Δ
WC is the change in working capital from year t − 1 to year t,
CFO is the cash flow from operations, Δ
REV is the change in sales from year t − 1 to year t, and
PPE presents the level of property, plant, and equipment at year t. All variables are scaled by the average total assets between the year t − 1 and t. To include each industry-year into the regression model, we require the industry-year to have at least ten observations. After estimating the above regression, we generate the residuals and calculate the standard deviation of the residuals over a rolling window of three years as the measure of accruals quality. The higher value for the variable indicates the lower quality of earnings.
Panel A of
Table 5 reports a similar pattern to the one in Panel A of
Table 4 that the value of the coefficient on GAAP earnings (
GAAP Earnings per share) increases from the low-quality sub-sample to the high-quality sub-sample, whereas the value of the coefficient on non-GAAP earnings (
Non-GAAP Earnings per share) decreases from the low-quality sub-sample to the high-quality sub-sample. Therefore, the magnitude of the difference for the coefficients between non-GAAP and equivalent GAAP earnings is larger in the low-quality sub-sample but the significance level of the difference remains identical (
p < 0.01) in both sub-samples. The analysis of the adjusted R
2 also confirms that the overall explanatory powers of the regression in columns (2) and (4) are higher compared to the corresponding value in columns (1) and (3), thus suggesting that the difference in the value relevance between non-GAAP and equivalent GAAP earnings is not completely mitigated by the accruals quality of a firm. The empirical evidence partially corroborates the previous conjecture that investors rely more heavily on the non-GAAP earnings to estimate the equity value of a firm if the firm reports low-quality GAAP earnings.
Panel B of
Table 5 reports that both values of the coefficients on GAAP earnings (
GAAP Earnings) and on non-GAAP earnings (
Non-GAAP Earnings) increase from the low-quality sub-sample to the high-quality sub-sample, except that the ones between columns (8) and (6) are nearly indifferent from each other. Further, the magnitude of the difference for the coefficients between non-GAAP and equivalent GAAP earnings is larger in the low-quality sub-sample while the significance level of the difference remains identical (
p < 0.01) in both sub-samples.
Overall, the empirical evidence from
Table 5 unveils that the difference in the value relevance and earnings predictability between non-GAAP and equivalent GAAP earnings is more evident when a firm reports lower-quality GAAP earnings, however, the phenomenon persists that non-GAAP earnings are more value relevant and have better predictability towards future operating earnings of a firm than equivalent GAAP earnings, even when accruals quality of a firm is high. It is worth noting that the result (un-tabulated) remains identical when we use alternative measures for accruals quality, such as the one estimated from the performance matched modified
Jones (
1991) model (
Dechow et al., 1995;
Kothari et al., 2005).
Table 6 repeats the analysis of
Table 3 but splits the sample into two sub-samples based on whether the number of analysts following a firm is above or below the median value in the sample each year. The extant literature demonstrates that greater analyst coverage serves as an external governance mechanism to monitor a firm’s financial reporting practice, including non-GAAP reporting (
Healy & Palepu, 2001;
Yu, 2008;
Bradshaw et al., 2018;
Christensen et al., 2021). Similarly, fundamental corporate theories (
Jensen & Meckling, 1976;
Fama, 1980) suggest that stronger governance of a firm can better monitor CEOs’ behaviors, thus reducing their rent-seeking behaviors. It is thus reasonable to conjecture that stronger analyst coverage can improve the quality of earnings and potentially increase the value relevance and the earnings predictability of both non-GAAP and equivalent GAAP earnings. However, other factors, such as whether investors care about a firm’s quality of governance, can also affect the value relevance of earnings. Hence, it is an empirical question to examine whether analyst coverage can moderate the value relevance and earnings predictability of the non-GAAP and equivalent GAAP earnings. To further examine the topic, we obtain the analyst coverage information from the IBES dataset and apply it to the empirical analysis.
Panel A of
Table 6 shows that both values of the coefficients on GAAP earnings (
GAAP Earnings per share) and on non-GAAP earnings (
Non-GAAP Earnings per share) increase from the low analyst coverage sub-sample to the high analyst coverage sub-sample. The empirical evidence thereby confirms that the value relevance of both non-GAAP and equivalent GAAP earnings improves when analyst coverage is higher. Interestingly, the value relevance of non-GAAP earnings increases more than the value relevance of GAAP earnings when more analysts cover a firm, thus leading to an increase in the magnitude of the difference in the coefficients between non-GAAP and equivalent GAAP earnings in the high analyst coverage sub-sample. However, in both sub-samples, the significance level of the difference remains identical (
p < 0.01). The analysis of the adjusted R
2 also confirms that the overall explanatory powers of the regression in columns (2) and (4) are higher compared to the corresponding value in columns (1) and (3), thus suggesting that the difference in the value relevance between non-GAAP and equivalent GAAP earnings remains statistically identical after controlling for the analyst coverage of a firm. The empirical evidence lends support to the previous conjecture that the value relevance of earnings improves when analyst coverage is higher but does not support the notion that the difference in value relevance between non-GAAP and equivalent GAAP earnings is completely mitigated by the analyst coverage.
Panel B of
Table 6 reports a similar pattern to the one in Panel A of
Table 6. Specifically, both values of the coefficients on GAAP earnings (
GAAP Earnings) and on non-GAAP earnings (
Non-GAAP Earnings) increase from the low analyst coverage sub-sample to the high analyst coverage sub-sample. Further, the magnitude of the difference in the coefficients between non-GAAP and equivalent GAAP earnings is larger in the high analyst coverage sub-sample, whereas the significance level of the difference remains identical (
p < 0.01) in both sub-samples.
Overall, the empirical evidence from
Table 6 reveals the difference in the value relevance and earnings predictability between non-GAAP and equivalent GAAP earnings is more pronounced when more analysts cover a firm. Admittedly, the phenomenon persists that non-GAAP earnings are more value relevant and have better predictability towards future operating earnings of a firm than equivalent GAAP earnings, even when analyst coverage of the firm is low.
Table 7 repeats the analysis of
Table 3 but splits the sample into two sub-samples based on whether a firm has above and below median values of the managerial ability in a year.
2 The extant literature reveals that management quality has a positive impact on corporate policies and outcomes, as well as accounting choices (
Bamber et al., 2010;
P. R. Demerjian et al., 2013;
Ge et al., 2011;
B. Francis et al., 2019;
Kim, 2023;
Abdel-Meguid et al., 2021). Specifically,
P. R. Demerjian et al. (
2013) suggest that superior managers have better knowledge about their firms and businesses, which leads to better judgment and estimates on how to conduct business and, eventually, to a better quality of earnings. Their empirical evidence suggests that the presence of more capable managers is associated with better predictability of earnings and accruals on future operating earnings. Additionally,
B. Francis et al. (
2019) document evidence to suggest that managerial ability is a positive intangible asset of a firm that can increase the value relevance of earnings of the firm. Reversely,
Cheng (
2017) reports mixed evidence regarding the impact of managerial ability on the quality of non-GAAP earnings. Based on the preceding discussion, it is an empirical question to examine how managerial ability affects the value relevance and earnings predictability of both non-GAAP and equivalent GAAP earnings. To further examine the topic, we obtain the proxy for managerial ability from
P. Demerjian et al. (
2012) and apply it to the empirical analysis.
Panel A of
Table 7 demonstrates that both values of the coefficients on GAAP earnings (
GAAP Earnings per share) and on non-GAAP earnings (
Non-GAAP Earnings per share) increase from the low managerial ability sub-sample to the high managerial ability sub-sample. The empirical evidence thereby confirms the findings from extant literature that the value relevance of earnings improves when managerial ability is higher (
B. Francis et al., 2019). Additionally, the magnitude of the difference for the coefficients between non-GAAP and equivalent GAAP earnings is larger in the high managerial ability sub-sample while the significance level of the difference remains identical (
p < 0.01) in both sub-samples. The analysis of the adjusted R
2 also confirms that the overall explanatory powers of the regression in columns (2) and (4) are higher compared to the corresponding value in columns (1) and (3), thus suggesting that the difference in the value relevance between non-GAAP and equivalent GAAP earnings remains statistically identical after controlling for the managerial ability of a firm. The empirical evidence supports the previous conjecture that the value relevance of earnings improves when managerial ability is higher, but it does not support the notion that the difference in value relevance between non-GAAP and equivalent GAAP earnings is completely mitigated by managerial ability.
Panel B of
Table 7, however, reports slightly different results from the one in Panel A of
Table 7. Specifically, the values of the coefficients on GAAP earnings (
GAAP Earnings) increase from the low managerial ability sub-sample to the high managerial ability sub-sample, whereas the values of the coefficients on non-GAAP earnings (
Non-GAAP Earnings) decrease. This situation leads to a lower magnitude of the difference for the coefficients between non-GAAP and equivalent GAAP earnings in the high managerial ability sub-sample, although the significance level of the difference remains identical (
p < 0.01) in both sub-samples.
Overall, the empirical evidence from
Table 7 indicates that the difference in the value relevance between non-GAAP earnings and equivalent GAAP earnings is more pronounced when firms are operated by more capable managers, whereas the difference in the earnings predictability between the non-GAAP and equivalent GAAP earnings is more evident in the sub-sample consisting of firms operated by less capable managers. Importantly, the difference in the value relevance and the earnings predictability between non-GAAP and equivalent GAAP earnings remain significantly identical after controlling for the managerial ability of the firm.
To summarize, the empirical evidence from this section reveals that the difference in the value relevance and earnings predictability between non-GAAP and equivalent GAAP earnings can be altered by firm-level characteristics, such as market value of equity, accruals quality, analyst coverage, and managerial ability of a firm, but these factors can not completely mitigate the phenomenon that non-GAAP earnings are more value relevant and have better predictability towards the future operating earnings of a firm than equivalent GAAP earnings. Specifically, the difference in the value relevance between non-GAAP and equivalent GAAP earnings is more pronounced when a firm has a higher market value of equity, lower accruals quality, higher analyst coverage, and higher managerial ability. Additionally, the difference in the earnings predictability between non-GAAP and equivalent GAAP earnings is more evident when a firm has a lower market value of equity, lower accruals quality, higher analyst coverage, and lower managerial ability.
4.2.3. Comparison of the Value Relevance and Earnings Predictability Between Non-GAAP and Equivalent GAAP Earnings for Both Pre- and Post-2010 Periods
As an effort to improve the quality and usefulness of information in reports filed to the SEC, the SEC issued revised Compliance and Disclosure Interpretations (C&DI) in 2010, 2016, and 2022. To some extent, these actions, especially the issuance of C&DI 2010, had loosened the reporting standards on non-GAAP reporting, and in turn, led to the rebound of non-GAAP reporting (
Bentley et al., 2018;
D. E. Black et al., 2012), and the increasing concern of non-GAAP earnings being misused to artificially inflate earnings
and being misleading to investors (
Morgan et al., 2018). Thus, to better understand whether the quality of non-GAAP earnings decreased after the issuance of C&DI in 2010, we decide to conduct a supplementary analysis in this section to compare the value relevance and earnings predictability of non-GAAP and equivalent GAAP earnings for both pre- and post-2010 periods. Specifically, we create an indicator variable that equals one if it is post-2010 and zero if it is pre-2010. We use the indicator variable to split the sample into two sub-samples, conduct the regression analysis in each sub-sample, and compare the coefficients on non-GAAP and equivalent GAAP earnings.
Table 8 reports the results.
Panel A of
Table 8 shows that the coefficients on both GAAP (
GAAP Earnings per share) and non-GAAP earnings (
Non-GAAP Earnings per share) increase in the post-2010 period, thus suggesting that the value relevance of both non-GAAP and equivalent GAAP earnings increase in the post-2010 period. Interestingly, the value relevance of non-GAAP earnings increases more than the value relevance of GAAP earnings in the post-2010 period, thus leading to an increase in the magnitude of the difference in the coefficients between non-GAAP and equivalent GAAP earnings in the post-2010 period. But in both pre- and post-2010 periods, the significance level of the difference remains identical (
p < 0.01), which indicates non-GAAP earnings are more value relevant than GAAP earnings in both pre- and post-2010 periods. Moreover, the analysis of the adjusted R
2 confirms that the overall explanatory powers of the regression in columns (2) and (4) are higher compared to the corresponding value in columns (1) and (3), thus corroborating the hypothesis that non-GAAP earnings are more value relevant than GAAP earnings in both pre- and post-C&DI periods.
Panel B of
Table 8 reports that the coefficients on GAAP earnings (GAAP Earnings) increase from column (1) to (3) and from column (5) to (7), indicating that the earnings predictability of GAAP earnings increase in the post-2010 period. The coefficients on non-GAAP earnings (Non-GAAP Earnings), however, decrease from column (2) to (4) and from column (6) to (8), although the decrease is insignificant from a statistical perspective in an untabulated test. As a result, the magnitude of the difference in the earnings predictability between non-GAAP and equivalent GAAP earnings decreases in the post-2010 period, but the significance level of the difference remains identical (
p < 0.01) in both pre- and post-2010 periods. The results indicate that non-GAAP earnings have better predictability for future operating earnings than GAAP earnings in both pre- and post-2010 periods.
Importantly, the preceding analyses do not lend support to the notion that the quality of non-GAAP earnings decreases significantly after the release of the C&DI in 2010. The non-GAAP earnings are more value relevant and have better predictability towards the future operating earnings of a firm than GAAP earnings in both pre- and post-2010 periods. In summary, the superior “informational” role of non-GAAP earnings to GAAP earnings persists in the post-2010 period.