2. The vertical information transfer effects of earnings restatements along the supply
2.4. Empirical results
2.4.3. Supplier contagion returns and accounting quality of suppliers
The main objective of this section is to test whether restatement-induced supply chain contagion stock returns are related to proxy for earnings quality of suppliers and customers.
To test accounting quality concern, I use performance-adjusted discretionary accruals to proxy for suppliers’ earnings quality (denoted by DA). Following Hribar and Collins (2002), I use the direct approach to compute the total accruals (TACC). I estimate a modified Jones model (Dechow et al. 1995) on a cross-sectional basis for each Fama and French (1997) industry with 20 or more firms in year t:18
Where
TACC equals to operating income less operating cash flows adjusted for
discontinued operations and extraordinary items; A is total Assets at the end of yeart-1;
ΔSALES is change in sales for firm i in year t; ΔREC is change in accounts receivable for firm i in year t; PPE equals property, plant and equipment for firm i in year t.I compute the performance−adjusted discretionary accruals based on Cahan and Zhang (2006), an alternative approach to control for companies’ performance effect.
That is, for each Fama and French (1997), I divide my sample into deciles based on sample companies’ return on assets (ROA). I then adjust each discretionary accrual estimated from Equation (2) by subtracting the median discretionary accruals for the firm’s industry-ROA deciles.19 I predict that supplier chain effects are related to performance-adjusted discretionary accruals.
In addition to capturing whether earnings restatement trigger investors to concern the accounting quality of other firms in supply chain, I also consider factors that extant literature suggests might lead to cross-sectional variations in the nature and extent of supplier and customer contagion. I first consider the severity of the restatement. The abnormal returns to the restating firm capture the information in the
18 Following prior research (e.g., Kothari et al. 2005; Cahan and Zhang 2006), I winsorize all distributions to the 1st and 99th percentiles in estimating Equation (2).
19 This approach does not impose linearity on the relation between accruals and the performance measure (ROA) (Cahan and Zhang 2006).
restatement (Durnev and Magen 2008). I include the restating firm’s three-day abnormal stock return (ARRE) to control for differences in investor perceptions of the severity and importance of the restatement and related information in the announcement. Given the underlying economics of the relationship between suppliers and the restating firms, one would expect to observe the coefficient on ARRE is positive.
I also consider that the economic bond between the restating firm and its suppliers. The literature suggests a vertical information transfer between two firms increase with their correlation in economic activities (Pandit et al. 2007; Olsen and Dietrich 1985). Applying this notion, I expect that suppliers will suffer more pronounced supplier contagion effects when suppliers’ economic activities are more dependent on the restating firm. To measure the strength of economic bond between the restating firm and its suppliers (DEPENDCENC), I use the percentage of sales made by a supplier to the restating firm to assess the how reliant the supplier is on the restating firms for sales revenues.
Prior studies have viewed alliance as a form of relationship-specific investment by suppliers (e.g., Fee et al. 2006; Raman and Shahrur 2008). When suppliers invest in more relationship-specific investments to doing business with the restating firm, the more implicit/explicit claims held by customers and suppliers depend on the restating firm. Specifically, this variable captures the presence of specific contracts between the restating firm and its suppliers and customers. Thus, my second proxy for economic bond is alliance agreement. I expect that suppliers with alliance agreement with the restating firm will suffer pronounced supplier contagion. To gather information on alliances, I search for whether the restating firm and its suppliers were listed together in the Securities Data Corporation (SDC) strategic alliance database. I define ALLIANCE as a dummy variable that takes a value of one
if the firms in a relationship had a formal alliance agreement with the restating firm over three years before earnings restatement and zero otherwise.
Prior research documented that supplier power have an effect on suppliers by influencing the term of trading contracting. For example, restating firm relies on a larger suppler for its product as alternative source is not available or large enough.
Thus, I expect that larger suppliers suffer less negative stock price effects at the time of earnings restatement announcements. Following prior research, I measure the degree of concentration of the restating firm by the sale-based Herfindahl index (HERFINDAHL), which equals the sum of the squared fraction of industry sales by all firms in the industry.
To mitigate problems of potentially omitted correlated variables, I include several variables into my cross-sectional regression to control for the characteristics of the restating firm, customers, and suppliers that might affect contagion stock returns to customers and suppliers. Other information transfer studies have indicated that the size of the restating firm may have an impact on stock returns (e.g., Gleason et al. 2008). Thus, I add RESIZE, the natural log of total assets, into the regression.
The supplier long-term debt and debt in current liabilities divided by total assets (LEVERAGE) is used to control for the potential impact of financial leverage on abnormal returns to suppliers at the restatement announcement date (Hertzel et al.
2008). I also consider the effect of suppliers’ firm size on the contagion effects (SIZE).
CFS, sales to cash flows ratio, enters the regression is to control the profitability of
suppliers. Finally, Book-to-market ratio (BM) is to control for suppliers’ growth opportunity.In summary, to conduct my main tests of accounting quality concern argument, I estimate Equation (3) for suppliers:
ε
Where CAR is the restatement-induced contagion stock returns to suppliers during the three-day event period (-1, 1) surround the restatement announcements, the independent variables are as described above, and ε is a random disturbance term.
Empirical Results
Table 2.5 presents cross-sectional analysis of the three-day abnormal returns of the restating firms’ suppliers.20 The results reported here use individual firm observations although the portfolio regressions yield similar conclusions.21 The
t-values are computed with heteroskedasticity consistent standard errors if the tests
reject homoskedasticity at the 10% significance level (White 1980). 22 To conduct my analysis, I winsorize all the dependent and independent variables at the 1st and the 99th percentiles in order to reduce the effect of outliers on my results.[Insert Table 2.5 here]
In Model 1, the results show that the coefficient of ARRE is significantly positive, after including other potentially important variables in the regression.23 This result is anticipated from the results shown in previous section. The supply chain effects are more severe for restating firm that have larger negative abnormal returns around the restatement announcement. The finding supports the notion that the extent of the announcement effects on suppliers are significantly influenced by the extent to which the restatement valuation effects on the restating firms.
20 The results of the analyses for the 21-day abnormal returns are qualitatively similar to the results in this section.
21 Examining individual supplier firm, as opposed to supplier portfolio, allows me to test the importance of firm-specific variables in explaining the cross-sectional variation in abnormal returns to suppliers (as in Gleason et al. 2008 and others).
22 The samples in Table 4 are smaller because of financial data unavailability.
23 Variance inflation factor diagnostic statistics do not indicate multicollinearity as a problem (VIFs are less than 2.0).
As prediction, suppliers with high performance-adjusted discretionary accruals experience a more pronounced contagion stock price decline. This supports my hypothesis H3. The coefficients on DA are negative and significant, indicating that suppliers that have higher performance-adjusted discretionary accruals experience a more pronounced contagion stock price decline than do low-accruals firms. This finding supports the notion that earnings restatements provide useful information that alters investors’ perceptions about the financial reporting credibility of restating firms’ suppliers.
In all models, I add three variables related to types of restatements. As in the univariate results, I find that restatements of accounting fraud are associated with more negative supplier contagion stock returns. Consistent with prior research, my results support the notion that accounting fraud have more negative implications for restating firms’ accounting quality, which in turn increases perceived risk/uncertainty for suppliers and so will induce more severe supply chain contagion effects on suppliers. However, I fail to find significant evidence that REVENUE and
COST have any effect on abnormal returns of suppliers after controlling for ARRE
and FRAUD.24Note that as reported in Model 1, accounting fraud increases perceived risk/uncertainty for suppliers and so is associated with more negative supplier contagion stock returns. Thus, I add an interaction between accounting fraud and performance-adjusted discretionary accruals (FRAUD*DA) into regression model. If the fraud restatements are more likely to prompt investors to question over the suppliers’ accounting quality, then I expect a negative coefficient on the interaction terms. In Model 2, I present the coefficient on the interaction between the proxies for
24 Consistent with prior research (e.g., Palmrose et al. 2004), the result in model 2 suggests a meaningful association between fraud and revenue recognition restatement. I find the correlation between fraud and revenue recognition restatement is positive and significant (p-value<0.01).
earnings quality of suppliers and the accounting fraud. The evidence in Model 2 indicates negative and significant coefficients on the interaction term. This suggests that restatements involving accounting fraud are likely to cause greater concern about the credibility of financial information of restating firms’ suppliers.
To further explore the relation between accounting quality of suppliers and restatement-induced supplier contagion returns, this paper examines that whether contagion stock price effects are more pronounced when restating firm and suppliers use the same external auditor, identify the restating firm’s auditor for the fiscal year before the restatement is announced. Approximately 21 percent of the suppliers in my sample use the restating firms’ auditor. Thus, I add a common-auditor indicator variable (COM_AUDITOR) and an interaction between common-auditor indicator variable and performance-adjusted discretionary accruals into regression model.
In Model 3, I present the coefficients on the interaction between the proxies for earnings quality of suppliers and the common-auditor (COM_AUDITOR*DA). I find that contagion stock returns are negatively related to the incremental effect of DA when suppliers and the restating firms use the same auditor. This suggests that investors seem to impose an incremental contagion penalty on suppliers with high discretionary accruals when the supplier and restating firm employ the same extern auditor.
The coefficient on DEPENDENCE is negative and significant in Model 1, supporting my prediction. This variable measures how reliant the suppliers are on the restating firms for sales revenues and, hence, supplier switching costs. The more dependent the supplier is on a restating firm the more negative will be the supplier’s stock price reaction to the earnings restatement announcement. This evidence suggests that the stock market reaction to earnings restatement announcements takes into account the economic activities that relate suppliers to the restating firms. In
addition, I also find the coefficient on ALLIANCE is negative and significant, suggesting that suppliers have a formal alliance agreement with the restating firm suffer more negative contagion stock price effects.
However, there is no evidence to suggest that the concentration of supplier (HERFINDAHL), leverage (LEVERAGE), restating firm size (RESIZE), book-to-market ratio (BM) and cash flows to sales ratio (CFS) are relevant to determine supplier contagion stock returns.