I also test the effect of SEOs on the debt covenant hypothesis using accruals management and classification shifting measures of earnings management. I employ working capital discretionary accruals (A_WCA) as the accruals management measure for regression (5).37 This is because studies show that the former captures more subtle instances
37
I use modified Jones (1991) model to estimate AM by taking into account firm growth and operating performance similar to Collins, Pungaliya, and Vijh (2014). Kothari, Leone, and Wasley (2005) indicate that extreme operating performance should be controlled for to avoid inaccurate estimation of abnormal accruals. Thus, following Collins et al. (2014), I run the following regression to estimate normal working capital accruals: 𝑊𝐶𝐴𝑖,𝑡 𝐴𝑇𝑖,𝑡−1= 𝛼0+ 𝛽1 1 𝐴𝑇𝑖,𝑡−1+ 𝛽2 ∆𝐶𝑅𝑖,𝑡 𝐴𝑇𝑖,𝑡−1+ ∑ 𝛽𝑘 3,𝑘𝑅𝑂𝐴_𝐷𝑈𝑀𝑘,𝑖,𝑡−1+ ∑ 𝛽𝑘 4,𝑘𝑆𝐺_𝐷𝑈𝑀𝑘,𝑖,𝑡−1 to 𝑡+ ∑ 𝛽𝑘 5,𝑘𝑀𝑉_𝐷𝑈𝑀𝑘,𝑖,𝑡−1+ ∑ 𝛽𝑘 6,𝑘𝑀𝐵_𝐷𝑈𝑀𝑘,𝑖,𝑡−1+ ∑ 𝛽𝑘 7,𝑘𝐸𝑃_𝐷𝑈𝑀𝑘,𝑖,𝑡−1+ 𝑒𝑖,𝑡 (6)
of accruals management than total discretionary accruals in the UK (e.g. Peasnell, Pope, &
Young, 2000).38The results are presented in Table 3.6, columns (1). They show that firms do
not employ accruals management to remain within interest coverage covenant limits and this behaviour does not change following SEOs. This is probably because accruals management is not pervasively used and it has no direct cash flow implications (Cohen & Zarowin, 2010; Zang, 2012).
[Table 3.6 around here]
Regarding classification shifting, following Athanasakou et al. (2011) I use a dependent dummy variable, CS, that is equal to 1 for firms that have positive unexpected core earnings39 and higher I/B/E/S earnings40 than net income per share and 0 otherwise for regression (5). This captures those firms that are likely to engage in classification shifting by
where WCAi,t is working capital accruals for firm i in year t, calculated as the change in total current assets
minus the change in cash minus the change in current liabilities minus the change in the current portion of long term debt; ATi,t-1 is total assets for firm i in year t-1; ∆CRi,tis the change in sales for firm i from year t-1 to year t minus the change in accounts receivable for firm i from year t-1 to year t; 𝑘 takes the values of 1, 2, 4, and 5; 𝑅𝑂𝐴_𝐷𝑈𝑀𝑘,𝑖,𝑡−1 is quintile dummies for the return on assets, defined as earnings before extraordinary items
and discontinued operations scaled by lagged total assets. 𝑆𝐺_𝐷𝑈𝑀𝑘,𝑖,𝑡−1 to 𝑡 is quintile dummies for the sales
growth, defined as the change in sales from year t-1 to t divided by sales during year t-1. 𝑀𝑉_𝐷𝑈𝑀𝑘,𝑖,𝑡−1 is
quintile dummies for the market value of equity as of last year t-1; 𝑀𝐵_𝐷𝑈𝑀𝑘,𝑖,𝑡−1 is quintile dummies for the
market to book equity as of year t-1; 𝐸𝑃_𝐷𝑈𝑀𝑘,𝑖,𝑡−1 is quintile dummies for earnings to price, calculated as net
income for year t-1 divided by ending stock price as of year t-1; Each quintile dummy takes the value of 1 if the corresponding firm characteristic belongs to that 𝑘’th quintile, and zero otherwise. Regression (6) is estimated cross-sectionally within industry-years and normal working capital accruals are estimated using the estimated coefficients from regression (6). The difference between actual and normal working capital accruals gives working capital discretionary accruals (A_WCA).
38 The other reason why I employ working capital discretionary accruals rather than total discretionary accruals
is because I use EBITDA-based covenant (interest coverage).
39
Unexpected core earnings (UCE) are estimated following McVay (2006). UCE is the residual from the following regression estimated cross sectionally within industry-years:
𝐶𝐸𝑖,𝑡= 𝛼0+ 𝛽1𝐶𝐸𝑖,𝑡−1+ 𝛽2𝐴𝑇𝑂𝑖,𝑡+ 𝛽3𝐴𝐶𝐶𝑅𝑖,𝑡−1+ 𝛽4𝐴𝐶𝐶𝑅𝑖,𝑡+ 𝛽5∆𝑆𝐴𝐿𝐸𝑆𝑖,𝑡+𝛽6𝑁𝐸𝐺_∆𝑆𝐴𝐿𝐸𝑆𝑖,𝑡+ 𝑒𝑖,𝑡 (7)
where CEi,t is core earnings for firm i in year t scaled by sales where the former is defined as sales minus cost
of goods sold minus selling, general and administrative expenses; ATOi,t is asset turnover ratio, calculated as
sales over average net operating assets; ACCRi,t is accruals, defined as the difference between net income
before extraordinary items and cash from operationsdivided by sales; ∆SALESi,tis percentage change in sales; NEG_∆SALESi,t is percentage change in sales if it is less than 0, and 0 otherwise.
40
reclassifying core expenses as non-recurring items. The results for this logit regression are presented in Table 3.6 column (2). It shows that firms do not use classification shifting to remain within interest coverage covenant limits and this behaviour does not change following SEOs. This is perhaps because classification shifting does not have cash flow implications and thus firms may not employ it to avoid EBITDA-based interest coverage covenant.