FIN 48 and Risk-taking

To test the first hypothesis, I set up the following regression model (time and firm subscripts suppressed for brevity):

𝑅𝑖𝑠𝑘𝑠 = 𝛽₀+ 𝛽₁𝑈𝑇𝐵 + 𝛽₂𝑃𝑂𝑆𝑇07 + 𝛽₃ 𝑈𝑇𝐵 ∗ 𝑃𝑂𝑆𝑇07 + 𝑿’𝜁 + 𝑖 + 𝑡 + 𝜀 (1) 𝑿 stands for control variables. I also control for industry (i) and year (t) fixed effects.

I use the 48 industry classifications in Fama and French (1997) for the industry fixed effect. To evaluate the risk-taking of firms, I follow the work of Ljungqvist et al. (2016).

They offer several measures to proxy for the dependent variable Risks. Those measures include ROA volatility, ROIC volatility, Operating Cycle Change Ratio, and Capex. ROA

volatility is the standard deviation of quarterly pretax returns on assets and ROIC

volatility is the standard deviation of quarterly pretax returns on invested capital. Both are

calculated over the period from t to t+2. They both are suitable proxies to reflect firms’

aggregate risk taking. Operating Cycle Change Ratio is the ratio of the difference

between current and one-year ahead operating cycle to current operating cycle. This ratio is expected to detect whether firms reduce their risk-taking by shortening their operations.

If operating cycles are shorten, the operating capital put in risk becomes less. Capex in my definition is one-year ahead net capital expenditure (capital expenditure less sale of property) over the book value of lagged total assets.

UTB is an indicator variable equal to one if an observation has any data in the

following items that are provided in Compustat: beginning balance of unrecognized tax benefits, ending balance of unrecognized tax benefits, decreases to unrecognized tax benefits relating to settlements with taxing authorities, interest and penalties related to uncertain tax positions, and increases to unrecognized tax benefits arising from uncertain tax positions taken in a prior year. POST07 is equal to one if an observation is in 2007 or afterwards, and zero otherwise. 2007 is the year when FIN 48 is validated. My main independent variable is UTB*POST07, the interaction effect of UTB and POST07. As my first hypothesis states, I expect β3 to be negative because I argue that firms reduce their risk-taking in the FIN 48 regime.

Following prior literature, I include the firm characteristics that may affect firms’

investment and financing activities as control variables (e.g. Fama and French 1993 1995 2002; Harris and Raviv 1991; Coles et al. 2006; Jackson et al. 2009; Dechow and Dichev 2002; Liu and Wysocki 2016; Kothari et al. 2005). Specifically, I control for leverage (LEV), the accounting method choice of depreciation (DEP_PUREAC), the log of firm age (LOGAGE), market-to-book value (MTB), the log of market value of equity (LOGMVE), the volatility of sales (STDSALES), the volatility of investments

(STDINVEST), the log of operating cycle (LOGOPCYCLE), and

performance-matched modified-Jones model discretionary accrual (PERFDAROA).

(1) LEV stands for leverage, equal to long-term debt divided by common equity. A large stream of finance literature illustrates the corporate capital structure decisions (e.g. Fama and French 2002; Harris and Raviv 1991). Meanwhile, Coles et al. (2006) find positive relation between leverage and CEO

compensation schemes with higher sensitivity to stock volatility. In light of the literature, I believe the choices of capital structure are made to optimize the firm value. Hence, leverage is a cumulative outcome of corporate decisions and will positively affect firms’ risk-taking.

(2) DEP_PUREAC is an indicator showing whether the firm adopts the accounting method choice of accelerated depreciation or the units of production method.

Jackson et al. (2009) find that the external financial reporting choice of

depreciation method affects managerial decisions on capital investments. Firms using accelerated depreciation make larger investments than firms that use straight-line depreciation. Thus, I argue that the aggressive choice of depreciation method has an positive impact on firms’ risks ex ante.

(3) LOGAGE is the log of firm age. MTB is market-to-book value. LOGMVE is log of market value of equity which proxies for firm size. These variables are common variables to control for firm-level characteristics. First, the nature of firms’ operations changes as they grow so firms’ age can capture their

developments over time. I expect LOGAGE to have negative effects on

risk-taking for firms’ development may become stable as they grow. Secondly, Fama and French (1993) find evidence that size and market-to-book equity proxy for sensitivity to risk factors in stock returns. Furthermore, market-to-book value can be an indicator of the growth opportunities since firms with higher

market-to-book value are typical of higher average returns (growth stocks) and strong earnings ( Fama and French 1995). Therefore, I suppose that MTB and

LOGMVE are positively associated with my proxies for corporate risks.

(4) STDSALES and STDINVEST represent the volatility of sales and investments.

These two variables proxy for firms’ historic operating volatilities.

LOGOPCYCLE is the log of operating cycle. Prior literature (e.g. Dechow and

Dichev 2002; Liu and Wysocki 2016) provides evidence that these operating features affect accruals quality. In light of their work, a portion of operating decisions come along spontaneously and are without managerial discretions such

as accounting principle choices. Hence, I expect these variables to capture some variation of firms’ risk-taking and the effects are positive. Higher operating

volatilities and longer operating cycles are indicators of higher risk-taking.

(5) PERFDAROA is the performance-matched modified-Jones model discretionary

accrual developed by Kothari et al. (2005)². It’s a proxy for financial reporting quality in the sense that larger value of PERFDAROA refers to poorer quality of

financial reporting. Francis et al. (2005) conclude that accruals quality is a risk factor priced by investors. Given this idea, I suppose it’s necessary to control for

the reporting quality. The relation may be positive because firms need to react to capital providers’ higher required rate of return for higher information risk

brought out by poor accrual quality.

The detailed definitions of variables can be referred to in appendix A.