B. PAPER 2
2. HYPOTHESIS DEVELOPMENT AND RELATED RESEARCH 1 The influence of debt on SDEM
In settings with public equity capital markets earnings management incentives are widely researched, and most research designs investigate earnings management arising from agency conflicts between owners (shareholders) and managers (see review by Dechow et al. 2010).
However, such agency conflicts are practically absent in owner-managed firms, in which agency conflicts arise between the owner-manager and other firm stakeholders, such as lenders; banks being the most significant source of capital (OECD 2017a)2. Thus the presence and magnitude of debt is expected to influence the financial reporting decisions of owner-managed firms.
In our setting, the owner-manager is approximately indifferent between labor income and dividends because taxation of labor income and capital income (at the personal level taxed as corporate income and dividend income) is largely aligned. While dividends are disclosed in the annual report, manager salary, however, is not. To the extent that the owner-manager expects that the bank will not require salary data (private information) the owner-manager has incentive to use SDEM. We expect the owner-manager’s propensity to use SDEM to increase in the magnitude of debt, because the benefits in the form of lower expected interest rates are higher, and because lenders are not expected to unravel the owner-manager’s use of SDEM when their debt investment is not at stake.
However, this effect is expected to reverse when debt is high, because financially risky firms (i.e. firms with high magnitudes of debt) are subject to lender scrutiny due to increased agency costs between firm owners and lenders (Haw et al. 2014), which is expected to increase borrower firms’ perceived risk of managing earnings and mitigate their propensity to manage earnings. Further, the capacity for using SDEM is limited when debt is high because equity is
2 This relation is even more pronounced in European firms compared to US firms.
96
low, and hence paying dividends is associated with litigation risk3 and risk of breaching capital based debt covenants.
H1: The relationship between the magnitude of debt and the propensity to use salary-dividend earnings management has an inverted u-shape.
2.2 Consequences of SDEM for the cost of debt
Financial statements and their quality are important factors in the lending decision. For example, Agarwal and Hauswald (2010) use a dataset on loan applications and outcomes from private SME firms, provided by a major US small-business lender, and find that 70-80 percent of the bank’s score of (potential) borrowers is based on hard information. Donelson et al. (2017) survey 492 US lending officers and provide similar insights: they find that their survey respondents make credit decisions “more on the basis of financial statements than on the soft information provided by relationship lending” (p 2053). Further, prior research has found that attributes of private firms’ financial statements, such as audit status (audit vs. non-audited), reporting format (accrual-based vs. cash flow based), earnings smoothness, and earnings quality, influence firms’ credit access and cost of debt (Minnis 2011; Allee and Yohn 2009; Gassen and Fülbier 2015; Vander Bauwhede et al. 2015). Income and cash flow statement items (i.e. items that are influenced by SDEM) are considered important for lenders because they function as debt covenant trip wires (Dyreng et al. 2017; Christensen and Nikolaev 2012) and feed into banks’ credit scoring models4.
Although banks have the ability and an obvious interest in monitoring financial statement quality, the costs associated with scrutiny of financial reports limit banks’ capacity of carefully looking into each borrower firm’s financials. In our setting (explained in detail in section 3) all limited liability firms are mandated to publish financial reports. With easy access to financial statement data lenders’ processing costs of financial statement data are very low (Kaya and
3 Legally, dividends cannot be paid if it leaves the company without adequate financial resources (see:
https://www.ret-raad.dk/blog/hvornaar-maa-man-udlodde-udbytte-i-et-selskab). Further, dividend payments may lead to debt covenant violations, or attract lenders’ attention.
4 We are not aware of any research that specifically aims to uncover banks’ credit scoring models. However, from our interviews (discussed later) we learn that such data feed into the credit scoring models of all interviewee banks.
Further providing indirect evidence, (1) Kraft (2015) shows that Moody’s use and adjust both profitability and cash flow measures in their credit rating process, and (2) profitability/cash flow measures are standard variables in probability of default models (Beaver et al. 2005; Shumway 2001).
97 Pronobis 2016)5, and thus the relative cost of obtaining and analyzing soft and other private information is high, which might lead to fixation on reported numbers or information disclosed in the annual report. The cost of scrutiny is particularly pronounced in small (for the bank) loans, typical for owner-managed firms. For example, Donelson et al. (2017) use a survey design of commercial lenders and find that financial statement quality is viewed as significantly less important when loan officers are dealing with small loans compared to those dealing with large loans.
In a broader context, prior research provides evidence on the variation in banks’ demand for information: Banks are less likely to (1) request financial statements after loan origination when borrower credit risk is very low or very high (Minnis and Sutherland 2017), (2) demand high-quality (audited) financial reports in regions and industries in which the bank has more loan-exposure because concentration fosters lending expertise (Berger et al. 2017), and (3) collect financial statements during periods of economic growth (Lisowsky et al. 2017).
We infer from this literature stream that, in a small business loan context, the demand for financial statement information and applied scrutiny varies between borrowers, and expect that this lack of consistent scrutiny6, bundled with lenders’ reliance on published financial statements (and the information in those) allows opportunistic borrower firm managers to on average extract rents from lenders. That is, to the extent that lenders rely on reported financials we expect firms using SDEM to obtain cost of debt benefits.
H2: Firms engaging in salary-dividend earnings management obtain lower future cost of debt
2.3 Moderating effects: meeting or beating benchmarks
An extensive amount of evidence documents discontinuities in earnings distributions (Burgstahler and Chuk 2017; Dechow et al. 2010; Burgstahler and Dichev 1997). These earnings discontinuities are observed around certain earnings benchmarks, such as zero earnings, last year’s earnings, and expected earnings.
5 Currently, financial statement data are available in XBRL format easily available at cvr.dk. Further, from our interviews (as discussed later) we learn that several banks indeed extract borrower firms’ financial statement data from central databases.
6 From informal interviews with lending officers we learn that banks systematically gather “soft” information that they use in estimating credit scoring models and in their lending decision (for example management quality).
Manager salary, however, is not a piece of information that is collected systematically by any of the interviewees.
Further, banks rely on internally generated credit scoring models rather than external credit ratings, at least for smaller entities.
98
Both analytical (Dye 2002; Guttman et al. 2006) and empirical research (Barth et al. 1999;
Bartov et al. 2002) provide evidence on the benefits of meeting or beating earnings benchmarks.
Exploring non-equity related benchmark beating, Coppens and Peek (2005) plot earnings distributions for private firms (where equity incentives are less pronounced) and find evidence for a discontinuity around zero earnings (loss avoidance) but not around zero earnings changes (decrease avoidance), and Jiang (2008) finds that US public firms meeting or beating earnings benchmarks – zero earnings, last year’s earnings, and expected earnings – obtain higher credit ratings and lower initial bond spreads (i.e. cost of debt benefits)7, supporting the view that lenders use heuristic benchmarks in their credit evaluation. Jiang finds that the effect is strongest when firms beat the zero earnings benchmark because lenders care about downside risk rather than upside potential.
For those reasons, we expect the H1 and H2 hypothesized relationships to be moderated around earnings benchmarks. First, we expect firms with pre-managed earnings just below a benchmark to have a high propensity to use SDEM. Second, we expect firms transforming pre-managed earnings below a benchmark to reported earnings above a benchmark to incrementally benefit from earnings management. As lenders care about downside risk rather than upside potential (Jiang 2008) we expect the effects to be stronger for the zero earnings benchmark than for the last year’s earnings benchmark.
H3: The propensity to use salary-dividend earnings management increases when pre-managed earnings are just below earnings benchmarks
H4: Firms using salary-dividend earnings management to meet or beat earnings benchmarks obtain lower future cost of debt than other firms using salary-dividend earnings management
H5: The H3 and H4 hypothesized effects are stronger for the zero earnings benchmark than the last year’s earnings benchmark
7 Chin et al. (2018) provide evidence on the private loan term benefits (both price and non-price) obtained by firms meeting or beating analyst forecasts.
99