Does the order of claims to assets on the balance sheet reflect equity risk?

Abstract

Purpose

This paper empirically examines whether the order of liability and preferred stock accounts presented on the balance sheet is consistent with how the stock market values their riskiness.

Design/methodology/approach

This paper measures a firm’s riskiness with idiosyncratic risk and employs the first-difference design to test the relation between idiosyncratic risk and the order of current liabilities, noncurrent liabilities and preferred stock, respectively. Further, the paper tests whether operating liabilities are viewed as riskier than financial liabilities. Finally, the authors partition their sample based on the degree of financial distress and investigate whether the results differ between the two subsamples.

Findings

The paper finds that current liabilities are viewed as riskier than noncurrent liabilities and preferred stock is viewed as less risky than current and noncurrent liabilities, consistent with the ordering on the balance sheet. Further, the paper finds that operating liabilities are viewed as riskier than financial liabilities. Finally, the authors find that total liabilities and preferred stock (redeemable and convertible classes) are viewed as riskier for distressed firms than for nondistressed firms.

Originality/value

The authors thoroughly investigate the riskiness of several classes of claims and document that the classification of liabilities and preferred stock classes is relevant to common stockholders for assessing their associated risk.

Citation

Liu, C.Z., Hu, X.S. and Reichelt, K.J. (2022), "Does the order of claims to assets on the balance sheet reflect equity risk?", China Accounting and Finance Review, Vol. 24 No. 3, pp. 290-322. https://doi.org/10.1108/CAFR-05-2022-0062

1. Introduction

Our study examines whether the stock market prices the claims on the right-hand side of the balance sheet, and whether the pricing is consistent with their required order on the balance sheet. Modigliani and Miller (1958, 1963) posit that financial leverage increases financial risk to common stockholders in a frictionless market (Appendix 1), which prior studies empirically support (Botosan & Plumlee, 2005; Dhaliwal, Heitzman, & Li, 2006). The balance sheet typically lists current liabilities before noncurrent liabilities and preferred stock. The United States (U.S.) Securities and Exchange Commission (SEC) Regulation S-X (CFR (code of federal regulations) 17 210.5-02) lists these claims in order of current liabilities (e.g. accounts payable, income taxes payable, short-term debt), long-term debt (e.g. bonds, mortgages and other liabilities) and preferred stock (redeemable and nonredeemable classes), before common stock and other shareholders’ equity (additional paid-in capital and retained earnings). The Financial Accounting Standards Board (FASB) conceptual framework (Concept Statement No. 8, OB2 and 3) implies that financial reporting should provide information to investors to assess the riskiness of future cash inflows (Financial Accounting Standards Board, 2021). Our study examines whether the order of these claims is relevant to investors by how the stock market assesses their riskiness.

The claims on the right-hand side are characterized by several dimensions which include current and noncurrent liabilities, operating and financial liabilities and equity-like and liability-like claims. Beginning with the first dimension, the FASB (ASC (accounting standards codification) 210-10-45-8, 9) states that current liabilities are typically due within one year, implying that noncurrent liabilities are typically due after one year [1]. Diamond (1991) proposes that liabilities with a shorter maturity-term have greater liquidity risk. Liquidity risk arises when firms do not have sufficient cash to settle, cannot refinance and consequently sell their assets at distressed prices (Diamond, 1991). Current liabilities are also riskier because of the operating risk from suppliers and employees who threaten continuing operations if unpaid (Cunat, 2007). We posit that current liabilities are riskier than noncurrent liabilities due to higher liquidity risk and operating risk.

Liabilities are also characterized by their cash flow activities from operating and financing activities (ASC 230-10-20, Nissim & Penman, 2003). Operating activities incur liabilities owed to suppliers, employees, customers and government (i.e. operating creditors). Financing activities incur liabilities owed to financial institutions and holders of debt securities (i.e. financial creditors). We posit that operating liabilities are associated with greater operating risk than financial liabilities because operating creditors can disrupt and even halt operations if not paid (Cunat, 2007). On the other hand, financial institutions and other creditors, who have liquidation rights, more often restructure debt payments to preserve the firm’s going concern value rather than liquidating assets (Casey, 2020; Warren & Westbrook, 2009).

Within these categories, we further examine individual accounts, some of which are controversial among standard setters. Within current liabilities, we separate trade payables and accrued liabilities (e.g. accrued payroll and warranties) from income taxes payable to compare the critical nature of suppliers, employees and customers to the strong collection powers of tax authorities (Internal Revenue Code, 26 U.S.C. § 6331). Within noncurrent liabilities, we separate two controversial accounts: deferred taxes and capital leases. Deferred taxes are not legally enforceable claims but it is unclear whether they are priced by the market (Graham, Raedy, & Shackelford, 2012). On the other hand, capital leases are a legally enforceable substitute for debt (Yan, 2006).

In addition to liabilities, the right-hand side reports preferred stock which has both liability-like and equity-like features. Preferred stock is liability-like in that it pays dividends on a fixed or determinable basis. However, it is also equity-like because the dividends are paid at the discretion of the board of directors and non-payment cannot force the firm into bankruptcy. Further, specific classes of preferred stock may be redeemable and convertible, that are liability-like and equity-like, respectively. Redeemable preferred stock has a liability-like feature permitting the holder to redeem it for cash. Convertible preferred stock has an equity-like feature permitting the holder to convert it into common shares. Because redemption and conversion features are unique features, we examine them separately. Since 1986, the FASB has grappled with developing standards that distinguish liabilities from equity, including preferred stock (FASB, 2016), and they have updated standards in recent years (FASB, 2020).

Our analysis starts with 102,928 observations from 1986 to 2016. The dependent variable, idiosyncratic risk, is measured in three ways: the current year, the current and past two years and the current and past four years. To control for time-invariant unobservable firm effects, we use a first-difference design that regresses the change in idiosyncratic risk on the change in liabilities and preferred stock. We use the first-difference design because it is more efficient than firm fixed effects when serial correlation of the residuals is high (Wooldridge, 2016). We control for known risks (operating, agency, information and bankruptcy risk), and control for potential feedback effects by including lagged independent variable terms. We standardize regression coefficients to permit comparisons.

We find that current liabilities are more positively associated with idiosyncratic risk than noncurrent liabilities, consistent with the theoretical prediction by Diamond (1991) and the greater operating risk. This result suggests that the riskiness of current liabilities agrees with the balance sheet order, and the separate classification of current liabilities from noncurrent liabilities is relevant to common stockholders. We also find that operating liabilities are more positively associated with idiosyncratic risk than financial liabilities, suggesting that operating liabilities are riskier claims than financial liabilities.

Within the separate categories of liabilities, we find some differences in their association with risk. Most categories of liabilities are associated with idiosyncratic risk: current operating liabilities (excluding income taxes payable), other noncurrent liabilities, short-term debt, long-term debt and capital leases. However, income taxes payable and deferred tax liabilities are not associated with idiosyncratic risk, suggesting that investors do not consider them risky claims.

We also find that preferred stock is positively associated with idiosyncratic risk, but the association is less than current liabilities and noncurrent liabilities. The result is consistent with the order of presentation on the balance sheet. Within preferred stock, we find that only three categories (redeemable, nonredeemable and convertible) are positively associated with idiosyncratic risk, suggesting they are liability-like. Nonconvertible preferred stock is not associated with idiosyncratic risk.

We also partition our analysis between financially distressed firms and nondistressed firms. Financially distressed firms have a different capital structure than nondistressed firms because distressed firms restructure their debt in response to increased creditor bargaining power, which may include issuing redeemable and convertible classes of preferred stock. Further, financially distressed firms bear greater bankruptcy, financial and liquidity risk. We find that total liabilities have a stronger association with idiosyncratic risk for distressed firms than nondistressed firms, supporting the view that creditors gain bargaining power and the liabilities are riskier. Current operating liabilities, short-term debt and preferred stock have a stronger association with idiosyncratic risk for distressed firms than nondistressed firms, suggesting they are a riskier and important source of credit for distressed firms. Among preferred stock classes, redeemable and convertible preferred stock have a stronger association with idiosyncratic risk for distressed firms than nondistressed firms, suggesting that the two classes are a riskier but important source of financing for distressed firms.

In short, our study shows that the ordering of liabilities and preferred stock is consistent with how common stockholders view their riskiness. Our paper contributes by showing that the riskiness of liabilities and preferred stock on the right-hand side of the balance sheet is unequal. Debt is a major part of economic activity, accounting for approximately 75% of U.S. gross domestic product (Federal Reserve Bank of St. Louis, 2021). Prior studies have examined overall financial leverage (Christie, 1982; Rosenberg & McKibben, 1973) and specific right-hand side claims for their association with risk. For instance, prior studies have examined capital leases (Yan, 2006), deferred taxes (Chandra & Ro, 1997; Graham et al., 2012) and specific preferred stock classes (Cheng, Frischmann, & Warfield, 2003; Kimmel & Warfield, 1995). However, to the best of our knowledge no other studies have compared multiple classes of claims. The implication of our study is that the classification of liabilities and preferred stock classes is relevant to common stockholders for assessing risk.

Our study extends prior studies of financial distress (Hotchkiss, John, Mooradian, & Thorburn, 2008), by showing that liabilities and preferred stock for distressed firms have a greater positive association with idiosyncratic risk than nondistressed firms. These results suggest that creditors and preferred stockholders gain bargaining power for distressed firms and the claims are riskier. Our analysis shows that investors not only rely on legal contract terms to assess the riskiness of leverage claims, but they also refer to the financial health to evaluate the riskiness of liabilities and specific classes of preferred stock.

Our study shows that operating liabilities have a greater positive association with idiosyncratic risk than financial liabilities, which prior studies have not examined (Cunat, 2007; Nissim & Penman, 2003). Nissim and Penman (2003) find that the positive association of future profitability and operating liability leverage is greater than the association between future profitability and financial liability leverage. They find the same for the association of market to book ratio and operating liability leverage vs. financial liability leverage. We extend their study by examining the association of idiosyncratic risk with operating liability leverage and financial liability leverage. Cunat (2007) find that trade credit is negatively associated with customer collateral levels, suggesting that trade credit is used by customers with fewer alternative financing sources. We extend their study by using a stronger measure of risk (idiosyncratic risk) and show that operating liabilities are riskier than financial liabilities, particularly for financially distressed firms. Further, we extend prior studies of preferred stock (Cheng et al., 2003; Kimmel & Warfield, 1995) by showing that specific features of preferred stock, redemption and convertibility, are positively associated with idiosyncratic risk. As well, our study uses a longer sample period, a stronger measure of risk and a first-difference model to control for omitted variables.

The remainder of the paper starts with discussing the background to risk and individual liability and preferred stock components, followed by data measurement and sample, empirical analysis and results, and discussion and conclusion.

2. Background

2.1 The association of financial leverage with common stockholder risk

Modigliani and Miller’s (1958, 1963) framework (hereinafter “MM”), shows that the financial risk to common stockholders increases with the debt-to-equity ratio (Appendix 1). In a frictionless market, common stockholders earn a residual return (rU−i)(1−τ)DLSL, that increases with the ratio of the market value of debt to common equity [2]. The intuition is that common shareholders earn a residual return after paying a fixed return to creditors. However, paying a fixed return to creditors increases the risk that common shareholders will not earn a positive return. Studies by Botosan and Plumlee (2005) and Dhaliwal et al. (2006) support MM’s prediction by providing evidence that the cost of equity capital is positively associated with the debt-to-equity ratio.

Rosenberg and McKibben (1973) test the MM framework and find that idiosyncratic risk is positively associated with the ratio of senior securities to total assets (LEV). Idiosyncratic risk is specific to a firm’s stock and is uncorrelated with the market return. It is defined as the variance of the residual (n_nt) for firm n in period t, from estimating the following market model:

(a)rnt=α+βntMt+nnt

where r_nt is the logarithm of the stock return for firm n in period t, and M_t is the market return in period t. The authors develop a proxy for idiosyncratic risk and find that it is positively associated with LEV for the period 1954 to 1970.

Christie (1982) theoretically extends MM, by showing that in an MM world with a constant interest rate, no dividends, a single class of riskless debt and constant rate of return volatility:

(b)σS=σV(1+LR)

where σ_S is the standard deviation of the rate of return on equity, σ_V is the standard deviation of the rate of return on the firm (a constant), LR is the market financial leverage ratio (D/S(V)), D is the market value of debt, S is the market value of the firm’s equity and V is the market value of the firm. Following equation (b), the volatility of a firm’s stock return is a positive function of its debt-to-equity ratio. Christie (1982) provides empirical evidence of a positive association between stock return volatility and the debt-to-equity ratio.

However, the assumption that the right-hand side of the balance sheet consists of riskless bonds with a fixed rate of interest may not hold for at least two reasons. First, many liabilities do not have fixed interest rates, rather they may have variable rates, implied rates (e.g. trade payables) or none at all (e.g. deferred taxes and accruals for payroll and warranties).

Secondly, liabilities are not riskless because some firms default. Consequently, debt covenants and bankruptcy laws restrict how firms can invest, finance and pay creditors and investors. Debt covenants may require the firm to maintain a minimum working capital level and prohibit additional borrowing unless specific financial ratios are met (Barclay & Smith, 1995). If a firm defaults on payment, creditors generally have the right to force the firm into bankruptcy, which by a court-supervised proceeding either reorganizes the debt contracts or liquidates the assets (Barclay & Smith, 1995). When the assets are liquidated, secured creditors (including lessors) are granted settlement priority over unsecured creditors and preferred stockholders, respectively (Barclay & Smith, 1995). In short, all claims on the right-hand side of the balance sheet are not identical in risk. Following prior studies (Diamond, 1991; Kimmel & Warfield, 1995; Nissim & Penman, 2003), we examine three dimensions of claims (current vs. noncurrent liabilities, operating vs. financing liabilities and liability-like and equity-like features of preferred stock). To provide further granularity, we examine individual liabilities including controversial liabilities (deferred income taxes and capital leases).

Further, the MM framework does not consider the costs of bankruptcy that are associated with lower observed leverage levels (Megginson, 1997). We examine whether financially distressed firms allocate risk to the right-hand side of the balance sheet differently from nondistressed firms.

We next discuss the background to all three dimensions of the right-hand side of the balance sheet, including specific classes and the associated risks of each category. We follow with our predictions.

2.2 Current and noncurrent liabilities

We begin our discussion with the classification of current and noncurrent liabilities. The classification is required by the FASB codification (ASC 210-10-05-4) and Regulation S-X (17 CFR § 210.5-02 – Balance Sheets) [3]. The FASB codification requires separated classification of current liabilities and it provides examples such as current operating liabilities and short-term debt (ASC 210-10-45) [4]. Similar to the FASB codification, Regulation S-X lists specific current liabilities on the face of the balance sheet (e.g. amounts payable to trade creditors), while permitting note disclosure for short-term debt (borrowings from financial institutions and commercial paper), and it requires a total for current liabilities. Regulation S-X also lists categories of non-current liabilities reported on the face of the balance sheet or by note disclosure (bonds, mortgages and capitalized leases). Deferred taxes are shown on the face of the balance sheet.

Current liabilities consist of legally enforceable claims from operating creditors and financial creditors. Operating creditors are owed for amounts of trade payables, payroll, sales taxes, income taxes and warranties. Financial creditors are owed for short-term debt including commercial paper, short-term loans and the current portion of long-term debt. In bankruptcy proceedings, secured short-term debt has a higher settlement priority than unsecured short-term debt.

Noncurrent liabilities consist of amounts owing to operating creditors and financial creditors. Operating creditors are owed for deferred income taxes and pension liabilities. Financial creditors are owed for long-term borrowings from financial institutions, bondholders and capital leases. Pension liabilities, long-term borrowings and capital leases are legally enforceable while deferred income taxes are not suggesting that deferred income taxes are less associated with financial risk than other noncurrent liabilities. Compared to operating creditors, financial creditors have unique contractual rights. Long-term borrowings typically have debt covenants that restrict investment, payout and financing policies, such as meeting working capital balances and restricting borrowing if specific financial ratios are not met (Barclay & Smith, 1995). Capital leases typically grant repossession rights to the lessor in the event of default, and they have the highest priority of claims among secured creditors under bankruptcy laws, while creditors of secured long-term borrowings have priority over unsecured creditors (Barclay & Smith, 1995).

Current and noncurrent liabilities differ by three risks: liquidity, operating and interest-rate risk. Liquidity risk is “the risk that a solvent but illiquid borrower is unable to obtain refinancing” (Diamond, 1991). Short-term borrowings must be refinanced after they mature and before the investment pays off. If the lender is unwilling to refinance, the borrower must liquidate the assets, to settle the short-term debt, sometimes at less than their intrinsic value. Because current liabilities mature before noncurrent liabilities, we expect that they are associated with greater liquidity risk.

Arguably, operating risk is higher for current liabilities than noncurrent liabilities. Operating risk occurs when creditors have the power to suspend operations if they are not paid or if payments are too slow and uncertain, both of which can increase the risk of bankruptcy. Suppliers hold the threat to stop delivery, leaving few if any practical substitutes for unique materials, brand-name merchandise and utilities (Petersen & Rajan, 1997; Cunat, 2007). Unpaid employees may stop working and litigate, leaving few if any practical substitutes for skilled or unionized labor (Findlaw, 2021). Governing authorities, such as the U.S. federal government, have strong collection powers such as seizure of property without a court ruling when the taxpayer refuses to pay (Internal Revenue Code, 26 U.S.C. § 6331).

A counterargument is that noncurrent liabilities include secured creditors (financial institutions, bondholders and lessors) who in the event of default could stop operations by forcing liquidation or repossessing assets. However, in the event of default, the U.S. Bankruptcy Code constrains their powers to liquidate the firm’s assets and destroy firm value. Chapter 11 of the Code (11 U.S.C. § 1101-1195) puts in place “guardrails” for management to negotiate a plan to restructure the debt contracts with the creditors (Casey, 2020). Under Chapter 11, an automatic stay prevents creditors from foreclosing on their collateral until the firm exits bankruptcy (Gilson, John, & Lang, 1990). Consequently, management more often negotiates a restructuring plan to continue operations, either privately or by court-supervision, rather than liquidating the firm’s assets (Barclay & Smith, 1995; Warren & Westbrook, 2009) [5]. In short, because bankruptcy laws constrain repossession and liquidation rights for secured creditors, we expect that operating creditors of current liabilities impose more operating risk and bankruptcy risk.

Finally, we argue that interest-rate risk is higher for noncurrent liabilities than current liabilities, but the effect may be mitigated. Investors demand a premium for holding securities with a fixed rate of return over a long maturity term (e.g. fixed-rate notes payable and capital leases); thus, noncurrent liabilities potentially bear higher interest rates than current liabilities. However, firms mitigate interest-rate risk with variable interest rate terms, hedging with derivatives and frequent adjustments to discount rate estimates for defined benefit pension plan liabilities.

Liabilities also bear financial risk arising from fixed, variable and implied interest rates, but it is unclear whether current liabilities bear greater risk than noncurrent liabilities. Financial risk is arguably greater for variable rates than fixed rates because the interest charges vary more over time. However, short-term debt and long-term debt potentially use both variable and fixed rates, while accounts payable and capital leases have implied rates.

In short, we predict that common stockholders view current liabilities as riskier than non-current liabilities because operating risk and liquidity risk is greater while interest-rate risk is mitigated.

2.3 Operating and financial liabilities

Liabilities are also characterized by their cash flow activity – operating and financing types (Nissim & Penman, 2003). Operating activities “generally involve producing and delivering goods and providing services,” and generally are from “transactions and other events that enter into the determination of net income” (ASC 230-10-20). They are owed to operating creditors: suppliers, employees, customers and government. On the other hand, financing activities involve “borrowing money and repaying amounts borrowed, or otherwise settling the obligation; and obtaining and paying for other resources obtained from creditors on long-term credit” (ASC 230-10-20). They are owed to financial institutions (e.g. banks and leasing companies) and holders of debt securities (e.g. commercial paper and bonds) and they operate in well-functioning capital markets (Nissim & Penman, 2003). Both operating liabilities and financial liabilities have financial risk, but we do not predict which category is greater.

We argue that common stockholders view operating liabilities as riskier than financial liabilities because they bear more operating risk that leads to greater bankruptcy risk. Operating creditors hold the threat to prematurely destroy the going-concern value of the firm, and suppliers have an information advantage over financial creditors (Cunat, 2007; Giannetti, Burkart, & Ellingsen, 2011; Petersen and Rajan, 1997). As discussed, suppliers and employees can halt operations if not paid on time, or if payments are too slow and uncertain. Suppliers also have an information advantage over financial institutions to sooner stop or threaten to stop delivery (Giannetti et al., 2011; Petersen & Rajan, 1997) [6].

On the other hand, as mentioned, financial creditors’ rights to liquidate assets are constrained by the U.S. Bankruptcy Code. Chapter 11 of the Code provides “guardrails” for management to restructure debt contacts, continue operations and minimize parties from holding each other up (Casey, 2020). A counterargument is that suppliers are also parties to debt restructuring plans and their liquidation rights are also constrained. However, suppliers still have the “threat of stopping the supply of intermediate goods to their customers” that financial creditors lack (Cunat, 2007), suggesting that operating liabilities have greater operating and bankruptcy risk than financial liabilities.

In the accounting literature, Nissim and Penman (2003) find that operating leverage has a higher association with future profitability (and the price to book ratio) than financing leverage. However, they do not examine and compare the association of operating leverage and financing leverage with risk. In short, we predict that common stockholders view operating liabilities as riskier than financial liabilities.

2.4 Preferred stock

Preferred stock is a hybrid of liability and equity. Like traditional debt, it requires periodic payments in the form of perpetual fixed-rate dividends. However, like common stock, the dividend payments can be postponed or suspended until liquid resources are available (Heinkel & Zechner, 1990). Further, preferred stockholders cannot force the firm into bankruptcy (Barclay & Smith, 1995). Like equity, some classes of preferred stock have voting rights (Kimmel & Warfield, 1995). In bankruptcy proceedings, preferred stockholders rank lower in settlement priority over secured and unsecured creditors and higher than common stockholders (Barclay & Smith, 1995).

An example is Chicken Soup for the Soul Entertainment, Inc. (CSSE) Series A preferred stock, issued in 2018, with an annual cumulative dividend rate of 9.75% [7]. The shareholders have no voting rights except to elect two members to the board of directors if dividends are in arrears for eighteen or more consecutive or non-consecutive months. In the event of liquidation, the shareholders receive $25 per share and any accumulated and unpaid dividends before common stockholders are paid.

The accounting standards for classifying preferred stock are less straight forward because of several FASB standards (topics 480, 815 and 505) and two balance sheet categories under Regulation S-X (17 CFR § 210.5-02). These requirements define various classes of preferred stock and specify how they are classified on the balance sheet (liabilities, mezzanine equity and stockholders’ equity). The FASB defines equity as “the residual interest in the assets of an entity that remains after deducting its liabilities” (ASC 505-10-05-3), implying that unless it is classified as a liability, it is equity. In the FASB Master Glossary, preferred stock is a “security that has preferential rights compared to common stock.” SEC Regulation S-X, since 1979, classifies two types of preferred stock outside of liabilities. The first type, redeemable preferred stock, is classified as mezzanine equity, between liabilities and stockholders’ equity (17 CFR § 210.5-02.27). The regulation defines redeemable preferred stock as mandatorily redeemable by the holder or redeemable outside the control of the issuer. The second category, which we refer to as nonredeemable preferred stock, is classified as stockholders’ equity (17 CFR § 210.5-02.28). It includes preferred stock that is not redeemable or is only redeemable by the issuer (i.e. callable).

FASB topic 480 and subtopic 815-15 address classification of preferred stock as a liability and the bifurcation into liability and equity components, respectively. ASC 480, Distinguishing Liabilities from Equity, since 2003, requires preferred stock to be classified as a liability if it is certain to be redeemed (mandatorily redeemable preferred stock) or has an unconditional obligation to issue a variable number of equity shares based on a fixed monetary value, variations in something other than the fair value of equity shares or variations inversely related to changes in the fair value of equity shares (ASC 480-10-25) [8]. ASC 815-15, Derivatives and Hedging – Embedded Derivatives, since 1999, bifurcates the embedded derivative of preferred stock into a fair-valued liability and mezzanine equity or stockholders’ equity [9]. Embedded derivatives include redemption rights by the holder or the issuer, rights by the holder to convert to common stock and other features (ASC 815-15-25-17D). The subtopic applies when embedded derivatives in their entirety are not clearly and closely related to the host preferred stock instrument’s economic risks and characteristics (liability-like or equity-like), together with other requirements (ASC 815-15-25-1) [10].

While preferred stock has several features, our analysis is constrained by the four categories that Compustat provides: redeemable, nonredeemable, convertible and non-convertible. Other features (voting rights, callability, embedded derivatives and mandatory redemption features) require hand-collection which is not practicable for our sample size (102,928). One study that examines redeemable preferred stock with voting rights was limited to a sample of 239 observations (Kimmel & Warfield, 1995). Rather, we test the preferred stock classes that are recorded by Compustat, which includes either mezzanine equity (redeemable) or stockholders’ equity (nonredeemable, convertible and non-convertible) [11]. Liability classification of embedded derivatives (ASC 815-15) and mandatorily redeemable preferred stock (ASC 480-10-25) are not examined.

We predict that preferred stock is positively associated with risk because it bears financial risk. Perpetual fixed-rate dividends are like interest payments on debt and are senior in payment priority to common stock dividends, which increase return volatility to common shareholders. Some preferred stock securities have cumulative rights to pay amounts in arrears, similar to accrued and unpaid interest. Further, preferred stock may have voting rights if dividends are sufficiently in arrears, permitting preferred stockholder to influence payment (e.g. OME and CSSE) [12]. Preferred stock also enhances debt capacity so that the firm can borrow more (Heinkel & Zechner, 1990), which also bears financial risk. However, preferred stock has little or no liquidity or bankruptcy risk because postponement or suspension of dividends cannot force liquidation or trigger bankruptcy. A counterargument is that if a firm suspends dividends indefinitely, there is no financial risk associated with preferred stock. However, such firms are likely insolvent and few. For solvent firms, when dividends are declared, preferred stockholders are paid before common stockholders.

Further, since we are interested in whether the ordering of preferred stock after total liabilities matters, we predict that preferred stock is less associated with risk than total liabilities. Namely, preferred stock bears less liquidity risk and bankruptcy risk than liabilities because the dividends can be deferred and deferral cannot force liquidation or trigger bankruptcy.

Prior empirical studies (Cheng et al., 2003; Kimmel & Warfield, 1995; Linsmeier, Partridge, & Shakespeare, 2020) find mixed results from examining the association of preferred stock with systematic risk and common stock prices. Kimmel and Warfield (1995) find that redeemable preferred stock with voting rights or conversion rights is negatively associated with systematic risk, incremental to operating risk (1979–1989). Cheng et al. (2003) find that nonredeemable preferred stock is positively associated with systematic risk (1993–1997), incremental to operating risk. Our study differs from these two studies by using a stronger measure of risk specific to the firm (idiosyncratic risk) [13], a longer sample period (1986–2016), and we use a first difference model to control for time-invariant omitted variables. In a related working paper, Linsmeier et al. (2020) examine whether investors view preferred stock as equity-like (liability-like) by whether it is positively (negatively) associated with common stock price. They find that preferred stock is positively (negatively) associated with common stock price for financially distressed (non-distressed) firms. Our paper differs in that we compare preferred stock and other claims to idiosyncratic risk for distressed and nondistressed firms, and we use a first-difference design. In short, while prior empirical findings are mixed, preferred stock arguably bears financial risk, thus we predict that it has a positive association with risk. Further, we predict preferred stock is less associated with risk than total liabilities (current and noncurrent liabilities, and operating and financing liabilities) because it bears less liquidity risk and bankruptcy risk.

2.4.1 Redeemable and nonredeemable preferred stock

Regarding redeemable preferred stock, we predict that it is positively associated with risk. Redeemable preferred stock bears financial risk because it grants the holder the option to a one-time payment. The one-time payment is useful for early-stage companies that require more cash to invest. For instance, OME, an early-stage company in 2010, had series G redeemable preferred stock that grants holders the option to redeem it after a specific date for an amount that increases over time [14]. The one-time payment bears financial risk because the redemption value increases over time, similar to a zero-coupon note accruing interest. In turn, the return volatility to common stockholders increases over time.

Regarding nonredeemable preferred stock, we predict that it is positively associated with risk, following our prediction for preferred stock in general. Namely, it bears financial risk by paying a perpetual fixed-rate dividend and it is senior in dividend and liquidation rights over common stockholders. Further, as stated, the dividend deferral feature bears financial risk by permitting additional borrowing.

2.4.2 Convertible and non-convertible preferred stock

We predict that convertible preferred stock is positively associated with risk because it bears financial risk and dilution risk. As stated, convertible preferred stock grants the holder the right to convert preferred stock into common stock at a fixed or determinable amount. It bears financial risk because it pays perpetual fixed-rate dividends but at a reduced rate (Lee & Figlewicz, 1999) [15]. Further, it bears dilution risk to existing common shareholders because the number of common shares outstanding increases if preferred stockholders convert their shares.

Regarding nonconvertible preferred stock, like nonredeemable preferred stock, we predict that it is positively associated with risk. Namely, it bears financial risk from paying perpetual fixed-rate dividends.

3. Data measurement and sample

3.1 Data measurement

Our dependent variable is the firm-specific risk borne by common stockholders, measured as idiosyncratic risk. Idiosyncratic risk (IRISK_t) is the portion of stock return variability which is not attributed to overall market return variability. We measure IRISK_t as the standard deviation of the residual from regressing daily stock returns on Center for Research in Security Prices (CRSP) value-weighted market returns for the one-year period starting from the beginning of the fourth month of fiscal year t until the end of the third month of fiscal year t+1. We measure idiosyncratic risk for one year (t) (IRISK_t), three years (t−2 to t) (IRISK_t3) and five years (t−4 to t) (IRISK_t5) (e.g. Ashbaugh-Skaife, Collins, Kinney, & Lafond, 2009) [16]. The latter two measures capture the longer-term effects for noncurrent liabilities and preferred stock. We retrieve return data from CRSP and derive our risk measures based on daily returns starting at the beginning of the fourth month of fiscal year t, t-2 and t-4 through to the end of the third month of fiscal year t +1. The three-month lag permits us to measure the stock return during the approximate period that financial statement information is available to the stock market. IRISK_t, IRISK_t3 and IRISK_t5 are multiplied by 10 to better compare the variable coefficients in our regression analysis.

The independent variables in our analysis include the first two dimensions: current liabilities (CL_t) and noncurrent liabilities (NCL_t), and total operating liabilities (TOL_t) and total financial liabilities (TFL_t). We also examine the intersection of the two dimensions as current operating liabilities (COL_t), noncurrent operating liabilities (NCOL_t), current financial liabilities (STD_t) and noncurrent financial liabilities (NCFL_t).

We then decompose these categories as follows. Current operating liabilities decompose into current operating liabilities excluding income taxes (COLx_t) and income taxes payable (IP_t). Current financial liabilities do not decompose and only include short-term debt (STD_t). Noncurrent operating liabilities decompose into deferred taxes (DTX_t) and other noncurrent operating liabilities (ONCL_t). Noncurrent financial liabilities decompose into long-term debt (LTD_t) and capital leases (CapL_t).

To examine the third dimension, equity-like and liability-like claims, we add preferred stock (PS_t) to our regression analysis. We decompose preferred stock into two groups of features: redeemable preferred stock (PSR_t) and nonredeemable preferred stock (PSNR_t), and convertible preferred stock (PSC_t) and non-convertible preferred stock (PSNC_t). We scale all liability and preferred stock variables by the market value of common stock (CE_t) (Christie, 1982).

3.2 Sample

Our sample initially consists of 112,842 firm-year observations covering 32 fiscal years from 1985 to 2016 for all firms (except financial and utility firms) and all available data from Compustat and CRSP to compute the variables. Our research design includes first-difference variables, which are computed as the difference between the current period and the prior period amounts. After first-differencing, the sample size is reduced by one year to 102,928 firm-year observations covering 31 years (1986 to 2016). Table 1 reports descriptive statistics. All variables are windsorized at the 1^st and 99^th percentiles. Mean (median) IRISK_t is 0.384 (0.313). Kothari, Li, and Short (2009) report mean (median) return volatility (multiplied by 10) of 0.270 (0.230) for the period from 1996 to 2001, suggesting that our sample has riskier firm-year observations. Mean (median) IRISK_t3 is 0.388 (0.332) and IRISK_t5 is 0.390 (0.342), comparable to IRISK_t. We report first-difference variable amounts in Panel B.

Table 2, Panels A and B, report correlation coefficients (Pearson in the upper right of the diagonal and Spearman in the lower left) for first-difference variables, which we use in our multivariate analysis. Panel A reports correlations between the three risk measures and broader categories of liabilities (current, noncurrent, operating and financing) and preferred stock. Panels B reports correlations between the three risk measures and individual accounts. Panels A and B report that all three first-difference risk measures are positively correlated with virtually all first-difference liabilities and preferred stock variables.

4. Empirical analysis

Our empirical analysis examines how the right-hand side claims by creditors and preferred stockholders are associated with idiosyncratic risk. To ensure that we are measuring the risk of each claim, we control for firm-level time-constant omitted variables by first-differencing dependent and independent variables. Further, first-differencing is more efficient than firm fixed-effects estimation when serial correlation of residuals is high (Wooldridge, 2016).

4.1 Liabilities analysis

We begin by examining whether current liabilities and noncurrent liabilities are positively associated with risk, and whether the positive association of current liabilities is greater than noncurrent liabilities. We estimate the following first-difference model, where i and t are firm and year subscripts, respectively, RISK_it denotes IRISK_it, IRISK_it3 and IRISK_it5, and Δ is the difference between year t and year t−1:

(1)ΔRISKit=β0+β1Δ(CL/CE)it+β2Δ(NCL/CE)it+β3Δ(STD_ROA)it+β4Δ(BID_ASK)it+lagged_termsit+IND_DUMMIESit+SIZE_DUMMIESit+BM_DUMMIESit+YEAR_DUMMIESit+ε.

Lagged level terms for leverage components are added as instruments because of potential feedback effects influencing risk levels. For instance, firms that are financially distressed may substitute current liabilities for noncurrent liabilities because creditors expect a shorter investment horizon.

We control for several risks that may not be adequately controlled by first-differencing. We control for operating risk (Lev, 1974) by including the standard deviation of ROA (net income before interest and taxes scaled by total assets) over years t, t−1 and t−2. We add industry fixed effects, using Fama and French (1997) 48 industries, to control for bankruptcy risk that may vary by industry life-cycle stage [17]. In later analysis, we partition the sample between financially distressed firms and nondistressed firms. Industry fixed effects also control for industry financing practices such as greater debt financing in capital intensive industries (Megginson, 1997). Agency risk is controlled by adding firm-size-decile dummies (Gayle & Miller, 2009). Information risk is controlled by adding book-to-market decile dummies (Rajgopal & Venkatachalam, 2011), firm-size-decile dummies (Lang & Lundholm, 1993) and bid-ask spread. Year dummies are added to control for temporal effects, such as changes in accounting standards, regulations and laws. Standard errors are clustered by firm to adjust for within-firm serial correlation (Petersen, 2009). Standardized coefficients are reported to afford comparability.

The results of estimating equation (1) are presented in Table 3. Operating risk (ΔSTD_ROA)_t and bid-ask spread (ΔBID_ASK)_t are positively associated with all three risk first-difference measures. The maximum variance inflation factor (VIF) for all independent variables (excluding dummy variables and lagged variables) is well below the multicollinearity threshold of 10 (Kennedy, 2008). Our variables of interest, Δ(CL/CE)_t and Δ(NCL/CE)_t, are positively associated with all three idiosyncratic risk measures (ΔIRISK_t, ΔIRISK_t3 and ΔIRISK_t5), although the effect is reduced for the latter two measures. The coefficient F-tests reveal that Δ(CL/CE)_t has a greater standardized coefficient than Δ(NCL/CE)_t, suggesting that current liabilities have a stronger association with idiosyncratic risk than noncurrent liabilities, supporting our prediction.

In equation (2a), we classify liabilities by operating liabilities Δ(TOL/CE)_t and financial liabilities Δ(TFL/CE)_t. Equation (2b) extends (2a) with current and noncurrent classification: current operating liabilities Δ(COL/CE)_t, noncurrent operating liabilities Δ(NCOL/CE)_t, current financial liabilities Δ(STD/CE)_t and noncurrent financial liabilities Δ(NCFL/CE)_t. Equation (2c) expands with individual accounts: current operating liabilities excluding income taxes Δ(COLx/CE)_t, income taxes payable Δ(IP/CE)_t, deferred taxes Δ(DTX/CE)_t, other noncurrent liabilities Δ(ONCL/CE)_t, short-term debt Δ(STD/CE)_t, long-term debt Δ(LTD/CE)_t and capital leases Δ(CapL/CE)_t.

(2a)ΔRISKit=β0+β1Δ(TOL/CE)it+β2Δ(TFL/CE)it+β3Δ(STD_ROA)it+β4Δ(BID_ASK)it+lagged_termsit+IND_DUMMIESit+SIZE_DUMMIESit+BM_DUMMIESit+YEAR_DUMMIESit+ε.

(2b)ΔRISKit=β0+β1Δ(COL/CE)it+β2Δ(NCOL/CE)it+β3Δ(STD_CE)it+β4Δ(NCFL_CE)it+β5Δ(STD_ROA)it+β6Δ(BID_ASK)it+lagged_termsit+IND_DUMMIESit+SIZE_DUMMIESit+BM_DUMMIESit+YEAR_DUMMIESit+ε.

(2c)ΔRISKit=β0+β1Δ(COLx/CE)it+β2Δ(IP/CE)it+β3Δ(DTX_CE)it+β4Δ(ONCL_CE)it+β5Δ(STD_CE)it+β6Δ(LTD_CE)it+β7Δ(CapL_CE)it+β8Δ(STD_ROA)it+β9Δ(BID_ASK)it+lagged_termsit+IND_DUMMIESit+SIZE_DUMMIESit+BM_DUMMIESit+YEAR_DUMMIESit+ε.

Table 4 reports the results of estimating equations (2a)–(2c). Results from estimating equation (2a) show that Δ(TOL/CE)_t and Δ(TFL/CE)_t are positively associated with all three idiosyncratic risk measures, and the standardized coefficient on Δ(TOL/CE)_t is greater than Δ(TFL/CE)_t supporting our prediction that operating liabilities have a stronger association with idiosyncratic risk than financial liabilities. Results from estimating equation (2b) show that Δ(COL/CE)_t, Δ(NCOL/CE)_t, Δ(STD/CE)_t and Δ(NCFL/CE)_t are positively associated with all three idiosyncratic risk measures. Results from estimating equation (2c) show that Δ(COLx/CE)_t, Δ(ONCL/CE)_t, Δ(STD/CE)_t, Δ(LTD/CE)_t and Δ(CapL/CE)_t are positively associated with all three idiosyncratic risk measures. Notably, income taxes payable and deferred taxes are not associated with idiosyncratic risk.

4.2 Preferred stock analysis

To examine preferred stock (PS/CE)_it, we estimate the following models:

(3a)ΔRISKit=β0+β1Δ(CL/CE)it+β2Δ(NCL/CE)it+β3Δ(PS/CE)it+β4Δ(STD_ROA)it+β5Δ(BID_ASK)it+lagged_termsit+IND_DUMMIESit+SIZE_DUMMIESit+BM_DUMMIESit+YEAR_DUMMIESit+ε.

(3b)ΔRISKit=β0+β1Δ(TOL/CE)it+β2Δ(TFL/CE)it+β3Δ(PS/CE)it+β4Δ(STD_ROA)it+β5Δ(BID_ASK)it+lagged_termsit+IND_DUMMIESit+SIZE_DUMMIESit+BM_DUMMIESit+YEAR_DUMMIESit+ε.

We decompose preferred stock into two features provided by Compustat: redeemable (Δ(PSR)_t) and nonredeemable (Δ(PSNR)_t) (3c), and convertible (Δ(PSC)_t) and nonconvertible (Δ(PSNC)_t) (3d), as follows:

(3c)ΔRISKit=β0+β1Δ(COLx/CE)it+β2Δ(IP/CE)it+β3Δ(DTX/CE)it+β4Δ(ONCL_CE)it+β5Δ(STD/CE)it+β6Δ(LTD/CE)it+β7Δ(CapL/CE)it+β8Δ(PSR/CE)it+β9Δ(PSNR/CE)it+β10Δ(STD_ROA)it+β11Δ(BID_ASK)it+lagged_termsit+IND_DUMMIESit+SIZE_DUMMIESit+BM_DUMMIESit+YEAR_DUMMIESit+ε.

(3d)ΔRISKit=β0+β1Δ(COLx/CE)it+β2Δ(IP/CE)it+β3Δ(DTX/CE)it+β4Δ(ONCL_CE)it+β5Δ(STD/CE)it+β6Δ(LTD/CE)it+β7Δ(CapL/CE)it+β8Δ(PSC/CE)it+β9Δ(PSNC/CE)it+β10Δ(STD_ROA)it+β11Δ(BID_ASK)it+lagged_termsit+IND_DUMMIESit+SIZE_DUMMIESit+BM_DUMMIESit+YEAR_DUMMIESit+ε.

Table 5, Panel A, reports the results of estimating equations (3a) and (3b) and Panel B reports the results of estimating equations (3c) and (3d). Results of estimating equation (3a) reveal that the standardized coefficient on Δ(PS/CE)_t is positively associated with all three idiosyncratic risk measures, supporting our prediction. Other coefficient values are consistent with those reported in Table 3. The coefficient F-tests reveal that Δ(PS/CE)_t has a smaller standardized coefficient than Δ(CL/CE)_t and Δ(NCL/CE)_t. These results suggest that preferred stock has a smaller positive association with idiosyncratic risk than current liabilities and noncurrent liabilities, consistent with our prediction.

Results of estimating equation (3b) show that Δ(PS/CE)_t is positively associated with all three idiosyncratic risk measures, supporting our prediction. Δ(TOL/CE)_t and Δ(TFL/CE)_t are positively associated with all three idiosyncratic risk measures, consistent with Table 4 equation (2a) results. The F-tests reveal that the standardized coefficients on Δ(PS/CE)_t are less than Δ(TOL/CE)_t and Δ(TFL/CE)_t. These results suggest that preferred stock has a smaller positive association with idiosyncratic risk than operating liabilities and financial liabilities, consistent with our prediction.

Results of estimating equation (3c) reveal that Δ(PSR/CE)_t and Δ(PSNR/CE)_t are positively associated with all three idiosyncratic risk measures, supporting our predictions. Standardized coefficient signs on other leverage components are consistent with results reported in Table 4 equation 2(c). Compared to other leverage components, redeemable and nonredeemable preferred stock have a weaker positive association with idiosyncratic risk than many liability categories, except Δ(IP/CE)_t and Δ(DTX/CE)_t.

Results of estimating equation (3d) reveal that Δ(PSC/CE)_t is positively associated with all three idiosyncratic risk measures, supporting our prediction. Δ(PSNC/CE)_t is only positively associated with ΔIRISK_t3 and is not statistically associated with ΔIRISK_t and ΔIRISK_t5, thus we make no inference of an association with idiosyncratic risk, contrary to our prediction. Standardized coefficient signs on other leverage components are similar to results reported in Table 4 equation 2(c). Compared to other leverage components, convertible preferred stock has a weaker positive association with idiosyncratic risk than many liability categories, except Δ(CapL/CE)_t, Δ(IP/CE)_t and Δ(DTX/CE)_t.

4.3 Robustness – Altman bankruptcy prediction z-score

For robustness purposes, we add ALTMAN as an additional variable to equations (1) through (3d) to control for bankruptcy risk that is not captured by industry dummy variables (untabulated and available upon request). We find our inferences are unchanged.

4.4 Financial distress analysis

Financial distress may change the association of idiosyncratic risk with creditor and preferred stockholder claims because the firm’s capital structure changes. When firms are financially distressed, creditors gain bargaining power over shareholders because they can threaten to exercise liquidation rights and halt operations if payments are in default. In turn, the firm may restructure their debt contacts to meet their obligations (Hotchkiss et al., 2008). They may renegotiate terms with creditors such as reducing or deferring obligations and replacing debt with common and preferred stock allowing even greater flexibility to defer payments without forcing bankruptcy while reducing the debt-to-equity ratio (Hotchkiss et al., 2008). They may even issue additional debt or equity to others (Hotchkiss et al., 2008).

Trade credit and preferred stock are attractive forms of financing for distressed firms. Trade credit provides liquidity to smaller, younger and higher growth firms and suppliers can better enforce payment (Cunat, 2007). Preferred stock allows dividend deferral without forcing bankruptcy, reduces leverage, and is used by companies with lower tax rates (Houston & Houston, 1990). Convertible preferred stock has a lower dividend rate, as well as the flexibility to defer dividend payments. In turn, cash constrained firms, such as start-ups, plough cash into investment opportunities with the upside potential for preferred stockholders to convert and share in future profits (Lee & Figlewicz, 1999; Jepsen & Wilks, 2018; Johnson, n.d.). Redeemable preferred stock also has the flexibility to defer dividends with a one-time payout.

In short, financially distressed firms likely bear greater bankruptcy, financial and liquidity risk, because they are more likely to fail. In turn, creditors raise interest rates and lend on a secured basis and a shorter term (Tirole, 2006). Preferred stock is likely issued at a higher dividend rate and with a redemption feature. Thus, we expect that liabilities and preferred stock are more positively associated with idiosyncratic risk for financially distressed firms than nondistressed firms.

To examine this prediction, we partition our sample between financially distressed and nondistressed firms. We identify financially distressed firms as those below the Altman (1968) z-score of 1.81. We use the Altman (1968) z-score because it is the dominant method for measuring financial distress and bankruptcy prediction, and thus should better reflect the capital structure of the firm and the underlying risk associated with it [18, 19]. Financially distressed firms comprise 27% of our sample, comparable to Altman (2013) that reports 20% for all U.S. industrial firms in Compustat in 1999.

We first examine total liabilities (TL) by estimating the following:

(4a)ΔRISKit=β0+β1Δ(TL/CE)it+β2Δ(STD_ROA)it+β3Δ(BID_ASK)it+lagged_termsit+IND_DUMMIESit+SIZE_DUMMIESit+BM_DUMMIESit+YEAR_DUMMIESit+ε.

We then examine total liabilities and preferred stock. We estimate the following:

(4b)ΔRISKit=β0+β1Δ(TL/CE)it+β2Δ(PS/CE)it+β3Δ(STD_ROA)it+β4Δ(BID_ASK)it+lagged_termsit+IND_DUMMIESit+SIZE_DUMMIESit+BM_DUMMIESit+YEAR_DUMMIESit+ε.

We then extend equation (4b) by separating liabilities into current and noncurrent categories, and operating and financial categories, as follows:

(4c)ΔRISKit=β0+β1ΔCOL/CEit+β2ΔNCOL/CEit+β3ΔSTD/CEit+β4ΔNCFL/CEit+β5ΔPS/CEit+β6ΔSTD_ROAit+β7ΔBID_ASKit+lagged_termsit+IND_DUMMIESit+SIZE_DUMMIESit+BM_DUMMIESit+YEAR_DUMMIESit+ε.

Table 6, Panels A, B, and C, respectively reports the results of estimating equations (4a), (4b) and (4c), by financially distressed and nondistressed firms and reports the difference in standardized coefficients between the two groups. Panel A reports that the standardized coefficient on Δ(TL/CE)_t is positive and statistically significant for all three risk measures. Standardized coefficients are greater for distressed firms than nondistressed firms, as predicted. The implication is that the positive association of total liabilities and idiosyncratic risk is greater for distressed firms than nondistressed firms, consistent with the notion that creditors gain bargaining power when firms are financially distressed.

Panel B reports that the standardized coefficient on Δ(PS/CE)_t is positive and statistically significant for financially distressed firms only and standardized coefficients are greater for distressed firms than nondistressed firms when the dependent variable is ΔIRISK_t and ΔIRISK_t5, and marginally significant for ΔIRISK_t3 (p < 0.10). The implication is that the positive association of preferred stock and idiosyncratic risk is greater for distressed firms than nondistressed firms, and that the positive association for the entire sample (Table 5, Panel A) appears to be driven by distressed firms.

Panel C reports that standardized coefficients are greater for distressed firms than nondistressed firms for Δ(COL/CE)_t, suggesting that the positive association between current operating liabilities and idiosyncratic risk is greater for distressed firms than nondistressed firms. The standardized coefficients on Δ(STD/CE)_t are greater for distressed firms than nondistressed firms, suggesting that the positive association between short-term debt and idiosyncratic risk is greater for distressed firms than non-distressed firms. Differences in standardized coefficients between distressed and nondistressed firm are not statistically significant across most of the three idiosyncratic risk measures for Δ(NCOL/CE)_t and Δ(NCFL/CE)_t. Panel C also reports that the liability component coefficient signs are consistent with Table 4 equation (2b). Results for Δ(PS/CE)_t are similar to Panel B.

We extend equation (4c) by separating preferred stock into redeemable and nonredeemable categories, and convertible and nonconvertible categories, as follows:

(4d)ΔRISKit=β0+β1ΔCOL/CEit+β2ΔNCOL/CEit+β3ΔSTD/CEit+β4ΔNCFL/CEit+β5ΔPSR/CEit+β6ΔPSNR/CEit+β7ΔSTD_ROAit+β8ΔBID_ASKit+lagged_termsit+IND_DUMMIESit+SIZE_DUMMIESit+BM_DUMMIESit+YEAR_DUMMIESit+ε.

(4e)ΔRISKit=β0+β1ΔCOL/CEit+β2ΔNCOL/CEit+β3ΔSTD/CEit+β4ΔNCFL/CEit+β5ΔPSC/CEit+β6ΔPSNC/CEit+β7ΔSTD_ROAit+β8ΔBID_ASKit+lagged_termsit+IND_DUMMIESit+SIZE_DUMMIESit+BM_DUMMIESit+YEAR_DUMMIESit+ε.

Panel D reports the results of estimating equation (4d). The positive association between redeemable preferred stock and the three idiosyncratic risk measures is only evident for distressed firms. The standardized coefficient is greater for distressed firms than nondistressed firms when the dependent variable is ΔIRISK_t and ΔIRISK_t5, and marginally significant for ΔIRISK_t3 (p < 0.10). The implication is that the positive association observed for the entire sample (Table 5, Panel B) is primarily driven by financially distressed firms, which is consistent with the notion that redeemable preferred stock is used by distressed firms, including cash constrained start-up companies. There is no difference in the association of nonredeemable preferred stock and idiosyncratic risk between distressed and nondistressed firms.

Panel E reports the results of estimating equation (4e). The positive association between convertible preferred stock and the three idiosyncratic risk measures is only evident for distressed firms. The standardized coefficient is greater for distressed firms than nondistressed firms when the dependent variable is ΔIRISK_t and ΔIRISK_t5 and is not statistically different for ΔIRISK_t3. The implication is that the positive association observed for the entire sample (Table 5, Panel B) is driven by financially distressed firms. There is no difference in the association of nonconvertible preferred stock and idiosyncratic risk between distressed and nondistressed firms.

5. Discussion and conclusion

Our study examines the association of idiosyncratic risk with specific categories of liabilities and preferred stock and whether that association conforms to the order on the balance sheet. The balance sheet traditionally reports current liabilities separately from noncurrent liabilities and preferred stock after liabilities, in accordance with the FASB codification and SEC Regulation S-X. These categories reflect that the risk associated with current liabilities is greater than noncurrent liabilities, and that preferred stock is a junior claim to liabilities. In addition, we examine whether operating liabilities (amounts due to suppliers, employees and government) differ in risk from financial liabilities (amounts due to financial institutions and other holders of debt securities). Operating liabilities impose more operating risk because operating creditors can threaten to discontinue operations, while bankruptcy laws constrain financial creditors from liquidating assets. We also examine individual accounts which have been controversial (deferred income taxes and capital leases) and whether our results differ between financially distressed firms and nondistressed firms. We measure firm-specific risk borne by common stockholders using idiosyncratic risk.

We find that current liabilities have a stronger association with idiosyncratic risk than noncurrent liabilities. Further, we find that preferred stock is positively associated with idiosyncratic risk, but it has a smaller standardized coefficient than current liabilities and noncurrent liabilities. The result is consistent with the ordering of liabilities before preferred stock on the balance sheet and that preferred stock has less associated financial risk because of the dividend deferral feature. We also find that operating liabilities have a stronger positive association with idiosyncratic risk than financial liabilities. The result suggests that operating creditors are riskier but an important source of credit because operating creditors hold the threat to cease supplying inventory, labor and other services.

We also find that there is significant variation of the association of idiosyncratic risk with individual categories. Current operating liabilities (excluding income taxes), other noncurrent liabilities, short-term debt, long-term debt and capital leases are positively associated with idiosyncratic risk. On the other hand, income taxes payable and deferred taxes are not statistically associated with idiosyncratic risk. While deferred taxes have been debated in the past for whether they are liabilities, these results suggest that common shareholders do not view them as risky and liability-like. Further analysis reveals that three classes of preferred stock (redeemable, nonredeemable and convertible) are positively associated with idiosyncratic risk. Nonconvertible preferred stock is only positively associated with one of the three idiosyncratic risk measures; thus, we cannot draw an inference.

When we compare financially distressed firms to nonfinancially distressed firms, we find that total liabilities and preferred stock have a stronger association with idiosyncratic risk for financially distressed firms. The results suggest that risks such as bankruptcy, financial and liquidity risk are greater for financially distressed firms because of greater risk of business failure, higher interest rates and shorter maturity terms. Further analysis shows that distressed firms have a stronger positive association of current operating liabilities and short-term debt with idiosyncratic risk than nondistressed firms. The implication is that when firms are financially distressed, current operating credit (e.g. suppliers and employees) and short-term debt are riskier (e.g. bankruptcy and liquidity) sources of financing but also important sources because creditors gain bargaining power. We also find that preferred stock is only positively associated with idiosyncratic risk for distressed firms and is not statistically associated with idiosyncratic risk for nondistressed firms. The result suggests that the preferred stock of financially distressed firms bears more risk (e.g. bankruptcy and liquidity) than nondistressed firms and is an important source of financing by providing dividend deferral and leverage reduction. The result is more pronounced for redeemable preferred stock and convertible preferred stock. Redeemable and convertible preferred stock are only positively associated with idiosyncratic risk for distressed firms and are not for nondistressed firms, suggesting that these two classes of preferred stock are an important source of financing for financially distressed firms.

When we compare financially distressed firms to nonfinancially distressed firms, we also find that there are several categories that are no different between distressed and nondistressed firms in the association with idiosyncratic risk. They include noncurrent operating liabilities, noncurrent financial liabilities, nonredeemable preferred stock and nonconvertible preferred stock, suggesting that they are not uniquely used by distressed firms. Noncurrent operating liabilities include long-term deferred revenue that requires cash outflows to earn future revenue. Noncurrent financial liabilities include long-term debt and capital leases that require periodic payments for long-term operating assets. Nonredeemable preferred stock and nonconvertible preferred stock do not have redemption or conversion features that attract investors of distressed firms desiring an early payout or an option to share in future profits.

Our study is relevant to standard setters. The fact that we observe distinct differences between the positive association of specific classes of preferred stock and idiosyncratic risk supports the initiatives of the FASB project, Distinguishing Liabilities and Equity. We find that three classes of preferred stock (redeemable, nonredeemable and convertible) are positively associated with risk and thus are liability-like, but nonconvertible preferred stock is not statistically associated with risk. Redeemable and convertible preferred stock are only positively associated with idiosyncratic risk for distressed firms and are not associated for nondistressed firms.

Overall, our study finds that the stock market does not view all classes of claims equally and that the order of their presentation on the balance sheet is relevant and consistent with current accounting standards. Current liabilities are viewed as riskier than noncurrent liabilities, and preferred stock is the least risky, consistent with the order of presentation. The common stockholders’ view of the riskiness varies by whether the firm is financially distressed or not.

Table 1

Distribution statistics for 1986-2016 (31 years) (N = 102,928)

	Mean	Std	Median	Percentile
				1%	10%	25%	75%	90%	99%
Panel A
IRISKt	0.384	0.254	0.313	0.088	0.146	0.206	0.481	0.714	1.426
IRISKt3	0.388	0.226	0.332	0.099	0.159	0.221	0.492	0.692	1.223
IRISKt5	0.390	0.213	0.342	0.108	0.168	0.230	0.496	0.680	1.150
CL/CEt	0.515	0.957	0.216	0.000	0.041	0.097	0.493	1.131	6.793
NCL/CEt	0.632	1.383	0.185	−0.002	0.002	0.031	0.588	1.481	9.815
TOL/CEt	0.495	0.804	0.246	−0.004	0.048	0.112	0.521	1.074	5.603
TFL/CEt	0.669	1.484	0.190	0.000	0.000	0.022	0.603	1.581	10.430
COL/CEt	0.368	0.585	0.181	−0.085	0.036	0.084	0.389	0.831	3.914
NCOL/CEt	0.121	0.342	0.027	−0.352	−0.005	0.000	0.108	0.301	2.519
STD/CEt	0.210	0.639	0.024	0.000	0.000	0.000	0.123	0.439	4.867
NCFL/CEt	0.482	1.058	0.123	0.000	0.000	0.003	0.449	1.187	7.193
COLx/CEt	0.359	0.576	0.174	−0.102	0.034	0.079	0.378	0.814	3.843
IP/CEt	0.008	0.019	0.000	0.000	0.000	0.000	0.008	0.023	0.122
LTD/CEt	0.462	1.017	0.116	0.000	0.000	0.002	0.432	1.142	6.938
CapL/CEt	0.015	0.055	0.000	0.000	0.000	0.000	0.002	0.025	0.416
DTX/CEt	0.033	0.074	0.000	0.000	0.000	0.000	0.030	0.101	0.450
ONCL/CEt	0.084	0.306	0.008	−0.410	−0.014	0.000	0.059	0.205	2.297
TL/CEt	1.193	2.265	0.485	0.013	0.077	0.193	1.157	2.627	15.980
PS/CEt	0.017	0.088	0.000	0.000	0.000	0.000	0.000	0.001	0.695
PSR/CEt	0.005	0.035	0.000	0.000	0.000	0.000	0.000	0.000	0.299
PSNR/CEt	0.007	0.041	0.000	0.000	0.000	0.000	0.000	0.000	0.337
PSC/CEt	0.011	0.061	0.000	0.000	0.000	0.000	0.000	0.000	0.490
PSNC/CEt	0.002	0.016	0.000	0.000	0.000	0.000	0.000	0.000	0.136
STD_ROAt	0.072	0.110	0.034	0.002	0.008	0.016	0.077	0.170	0.697
BID_ASKt	0.028	0.049	0.008	0.000	0.000	0.001	0.033	0.079	0.286
ALTMANt	4.110	6.490	3.116	−15.190	0.052	1.658	5.166	9.079	39.070
Note(s): Variable definitions are provided in Appendix 2

	Mean	Std	Median	Percentile
				1%	10%	25%	75%	90%	99%
Panel B
ΔIRISKt	0.011	0.164	−0.001	−0.497	−0.144	−0.059	0.063	0.178	0.680
ΔIRISKt3	0.007	0.079	−0.001	−0.221	−0.066	−0.027	0.031	0.088	0.344
ΔIRISKt5	0.005	0.056	−0.001	−0.145	−0.046	−0.019	0.020	0.061	0.257
ΔCL/CEt	0.079	0.628	0.004	−1.931	−0.223	−0.052	0.087	0.366	4.016
ΔNCL/CEt	0.068	0.781	0.000	−2.966	−0.281	−0.053	0.083	0.406	4.766
ΔTOL/CEt	0.062	0.508	0.007	−1.662	−0.210	−0.054	0.092	0.331	3.106
ΔTFL/CEt	0.097	0.871	0.000	−2.906	−0.289	−0.054	0.093	0.467	5.573
ΔCOL/CEt	0.046	0.391	0.004	−1.315	−0.167	−0.042	0.069	0.266	2.315
ΔNCOL/CEt	0.013	0.184	0.000	−0.718	−0.062	−0.012	0.017	0.081	1.166
ΔSTD/CEt	0.048	0.456	0.000	−1.401	−0.107	−0.013	0.023	0.164	3.239
ΔNCFL/CEt	0.047	0.632	0.000	−2.502	−0.238	−0.042	0.058	0.333	3.709
ΔCOLx/CEt	0.045	0.386	0.004	−1.300	−0.163	−0.040	0.068	0.262	2.275
ΔIP/CEt	0.000	0.014	0.000	−0.060	−0.008	−0.001	0.000	0.008	0.067
ΔLTD/CEt	0.001	0.036	0.000	−0.158	−0.017	−0.001	0.001	0.018	0.179
ΔCapL/CEt	0.012	0.165	0.000	−0.667	−0.045	−0.006	0.011	0.063	1.064
ΔDTX/CEt	0.045	0.607	0.000	−2.423	−0.228	−0.040	0.055	0.320	3.526
ΔONCL/CEt	0.001	0.025	0.000	−0.114	−0.004	0.000	0.000	0.003	0.158
ΔTL/CEt	0.171	1.363	0.010	−4.217	−0.479	−0.111	0.194	0.788	8.812
ΔPS/CEt	0.003	0.045	0.000	−0.163	0.000	0.000	0.000	0.000	0.337
ΔPSR/CEt	0.001	0.015	0.000	−0.053	0.000	0.000	0.000	0.000	0.127
ΔPSNR/CEt	0.001	0.016	0.000	−0.065	0.000	0.000	0.000	0.000	0.126
ΔPSC/CEt	0.002	0.029	0.000	−0.104	0.000	0.000	0.000	0.000	0.219
ΔPSNC/CEt	0.000	0.005	0.000	−0.028	0.000	0.000	0.000	0.000	0.038
ΔSTD_ROAt	0.003	0.059	0.000	−0.220	−0.039	−0.012	0.013	0.045	0.288
ΔBID_ASKt	0.002	0.038	0.000	−0.139	−0.026	−0.005	0.005	0.029	0.182
ΔALTMANt	−0.629	5.175	−0.076	−29.450	−3.495	−0.959	0.522	1.888	18.990

Note(s): Δ signifies the difference in the level of the variable between period t and t−1. Appendix 2 provides variable definitions

Table 2

Correlation analysis for 1986–2016 (31 years; N = 102,928)

		(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)	(9)	(10)	(11)	(12)
Panel A
(1)	ΔIRISKt		0.514	0.487	0.296	0.242	0.300	0.270	0.299	0.146	0.191	0.218	0.092
(2)	ΔIRISKt3	0.631		0.693	0.170	0.126	0.177	0.146	0.174	0.081	0.110	0.112	0.073
(3)	ΔIRISKt5	0.632	0.780		0.179	0.133	0.190	0.152	0.185	0.094	0.113	0.116	0.073
(4)	Δ(CL/CE)t	0.334	0.263	0.272		0.513	0.825	0.682	0.897	0.316	0.642	0.472	0.197
(5)	Δ(NCL/CE)t	0.265	0.190	0.196	0.408		0.642	0.795	0.549	0.504	0.297	0.900	0.177
(6)	Δ(TOLC/E)t	0.342	0.266	0.277	0.781	0.642		0.612	0.886	0.575	0.403	0.515	0.209
(7)	Δ(TFL/CE)t	0.313	0.238	0.245	0.716	0.774	0.649		0.607	0.303	0.587	0.832	0.191
(8)	Δ(COL/CE)t	0.336	0.256	0.267	0.871	0.490	0.849	0.637		0.317	0.400	0.514	0.203
(9)	Δ(NCOL/CE)t	0.177	0.142	0.147	0.272	0.535	0.638	0.334	0.254		0.164	0.266	0.105
(10)	Δ(STD/CE)t	0.240	0.201	0.208	0.784	0.214	0.467	0.621	0.458	0.213		0.282	0.121
(11)	Δ(NCFL/CE)t	0.232	0.156	0.161	0.369	0.904	0.484	0.793	0.474	0.260	0.162		0.165
(12)	Δ(PS/CE)t	0.014	0.012	0.009	0.036	0.034	0.037	0.037	0.035	−0.003!	0.035	0.030
Note(s): All coefficients are significant at the 5% level, except those with # – not significant and ! – significant at the 10% level. The upper right corner reports Pearson correlation coefficients and the left lower corner reports Spearman correlation coefficients. Δ signifies the difference in the level of the variable between period t and t−1. Appendix 2 provides variable definitions

		(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)	(9)	(10)	(11)	(12)	(13)	(14)	(15)
Panel B
(1)	ΔIRISKt		0.514	0.487	0.299	0.038	0.191	0.242	0.215	0.120	0.083	0.121	0.050	0.078	0.080	0.045
(2)	ΔIRISKt3	0.631		0.693	0.174	0.023	0.110	0.126	0.111	0.061	0.025	0.078	0.043	0.059	0.066	0.032
(3)	ΔIRISKt5	0.632	0.780		0.185	0.032	0.113	0.133	0.115	0.067	0.038	0.089	0.044	0.060	0.067	0.033
(4)	Δ(COLx/CE)t	0.336	0.256	0.267		0.202	0.400	0.549	0.509	0.266	0.255	0.237	0.111	0.174	0.177	0.093
(5)	Δ(IP/CE)t	0.068	0.053	0.055	0.203		0.044	0.092	0.077	0.041	0.061	0.071	0.028	0.033	0.033	0.023
(6)	Δ(STD/CE)t	0.240	0.201	0.208	0.458	0.079		0.297	0.275	0.228	0.162	0.104	0.064	0.105	0.104	0.058
(7)	Δ(NCL/ CE)t	0.265	0.190	0.196	0.490	0.127	0.214		0.903	0.282	0.356	0.382	0.096	0.155	0.152	0.089
(8)	Δ(LTD/CE)t	0.229	0.152	0.157	0.464	0.110	0.148	0.906		0.308	0.284	0.183	0.088	0.144	0.140	0.083
(9)	Δ(CapL/CE)t	0.132	0.084	0.091	0.284	0.064	0.201	0.289	0.304		0.112	−0.179	0.049	0.077	0.077	0.044
(10)	Δ(DTX/CE)t	0.100	0.046	0.052	0.196	0.068	0.116	0.323	0.291	0.103		0.088	0.041	0.061	0.053	0.047
(11)	Δ(ONCL/CE)t	0.157	0.134	0.138	0.211	0.083	0.188	0.467	0.211	−0.136	0.088		0.044	0.079	0.075	0.036
(12)	Δ(PSR/CE)t	0.009	0.008	0.005#	0.027	0.017	0.028	0.027	0.024	0.023	0.000!	−0.009		0.043	0.373	0.456
(13)	Δ(PSNR/CE)t	0.048	0.038	0.040	0.084	0.017	0.076	0.069	0.058	0.021	0.007	0.052	−0.000!		0.706	0.333
(14)	Δ(PSC/CE)t	0.064	0.054	0.057	0.113	0.018	0.093	0.101	0.076	0.041	0.012	0.057	0.033	0.836		−0.002!
(15)	Δ(PSNC/CE)t	0.007	0.006#	0.003!	0.022	0.016	0.025	0.022	0.021	0.021	−0.000!	−0.011	0.998	0.003!	0.001!

Note(s): All coefficients are significant at the 5% level, except those with # – not significant, and ! – significant at the 10% level. The upper right corner reports Pearson correlation coefficients, and the left lower corner reports Spearman correlation coefficients. Δ signifies the difference in the level of the variable between period t and t−1. Appendix 2 provides variable definitions

Table 3

Association of leverage components and idiosyncratic risk – First-difference OLS regression (1986–2016) with lagged levels and fixed effects for size deciles, book to market deciles, industry and year (N = 102,928) – Current and noncurrent liabilities

Dependent/Independent variable	Equation (1)
	ΔIRISKt	ΔIRISKt3	ΔIRISKt5
Δ(CL/CE)t	0.2108*** (36.9960)	0.1526*** (29.3066)	0.1543*** (28.5292)
Δ(NCL/CE)t	0.1204*** (23.2095)	0.0831*** (17.5306)	0.0875*** (17.7724)
Δ(STD_ROA)t	0.0661*** (14.5734)	0.0988*** (22.4402)	0.0769*** (16.6499)
Δ(BID_ASK)t	0.1813*** (31.5573)	0.1718*** (34.1477)	0.1827*** (35.0870)
Lagged level terms	Included	Included	Included
Industry FE	Included	Included	Included
Size Decile FE	Included	Included	Included
BM Decile FE	Included	Included	Included
Year FE	Included	Included	Included
Coefficient F-tests
Δ(CL/CE)t = Δ(NCL/CE)t	1401.0***	899.4***	865.8***
Maximum VIF	1.74	1.74	1.74
Adjusted R2	0.230	0.306	0.309
F-value	187.6***	189.0***	173.0***

Note(s): For each independent variable, we report the standardized coefficient and the related t statistic based on the standard error clustered by firm (Petersen, 2009). Coefficient F-tests compare the stated standardized coefficients with the null hypothesis of equality (one tailed significance of null rejection). ^***, ^**, ^*significant at 0.01, 0.05 and 0.10, respectively. Δ signifies the difference in the level of the variable between period t and t−1. Lagged level terms (t−1) of independent variables (excluding dummy variables) are added to control for feedback effects and are not reported for brevity. For brevity, the intercept, and the coefficients on industry dummies, firm size dummies, book to market dummies, and year dummies are not reported. The maximum variance inflation factor excludes dummy variables and lagged terms. Appendix 2 provides variable definitions

Table 4

Association of leverage components and idiosyncratic risk – First-difference OLS regression (1986–2016) with lagged levels and fixed effects for size deciles, book to market deciles, industry and year (N = 102,928) – Operating and financial liabilities and other categories

Dependent/Independent variable	Equation (2a)			Equation (2b)			Equation (2c)
	ΔIRISKt	ΔIRISKt3	ΔIRISKt5	ΔIRISKt	ΔIRISKt3	ΔIRISKt5	ΔIRISKt	ΔIRISKt3	ΔIRISKt5
Δ(TOL/CE)t	0.1843*** (26.3928)	0.1358*** (22.9938)	0.1399*** (22.4458)
Δ(COL/CE)t				0.1738*** (26.0986)	0.1245*** (21.5381)	0.1301*** (22.1968)
Δ(COLx/CE)t							0.1709*** (25.5755)	0.1232*** (21.0416)	0.1282*** (21.6365)
Δ(IP/CE)t							0.0022 (0.5797)	0.0037 (1.0404)	0.0035 (0.9934)
Δ(NCOL/CE)t				0.0556*** (11.0913)	0.0461*** (10.6451)	0.0470*** (10.1565)
Δ(DTX/CE)t							0.0009 (0.2522)	−0.0054 (−1.6310)	−0.0018 (−0.5334)
Δ(ONCL/CE)t							0.0603*** (11.6763)	0.0481*** (10.7816)	0.0490*** (10.2432)
Δ(TFL/CE)t	0.1244*** (18.6886)	0.0858*** (14.9791)	0.0889*** (14.6341)
Δ(STD/CE)t				0.0884*** (15.9127)	0.0674*** (13.4346)	0.0657*** (12.6078)	0.0864*** (15.3855)	0.0669*** (13.1948)	0.0648*** (12.3021)
Δ(NCFL/CE)t				0.0675*** (11.8205)	0.0394*** (7.6746)	0.0427*** (8.1076)
Δ(LTD/CE)t							0.0618*** (10.5156)	0.0385*** (7.1709)	0.0398*** (7.2717)
Δ(CapL/CE)t							0.0273*** (5.7467)	0.0130*** (3.0651)	0.0171*** (3.9583)
Δ(STD_ROA)t	0.0665*** (14.6759)	0.0991*** (22.5218)	0.0770*** (16.6983)	0.0649*** (14.3745)	0.0977*** (22.2926)	0.0757*** (16.4841)	0.0644*** (14.2636)	0.0972*** (22.2132)	0.0752*** (16.4038)
Δ(BID_ASK)t	0.1831*** (31.8680)	0.1736*** (34.6059)	0.1844*** (35.4070)	0.1796*** (31.1305)	0.1706*** (33.8478)	0.1816*** (34.7306)	0.1791*** (31.0351)	0.1702*** (33.7705)	0.1812*** (34.6630)
Lagged level terms	Included	Included	Included	Included	Included	Included	Included	Included	Included
Industry FE	Included	Included	Included	Included	Included	Included	Included	Included	Included
Size Decile FE	Included	Included	Included	Included	Included	Included	Included	Included	Included
BM Decile FE	Included	Included	Included	Included	Included	Included	Included	Included	Included
Year FE	Included	Included	Included	Included	Included	Included	Included	Included	Included
Coefficient F-tests
Δ(TOL/CE)t = Δ(TFL/CE)t	1356.0***	871.3***	855.9***
Maximum VIF	1.92	1.92	1.92	1.82	1.82	1.82	1.90	1.90	1.90
Adjusted R2	0.230	0.306	0.310	0.232	0.307	0.311	0.232	0.307	0.311
F-value	187.5***	190.3***	174.4***	181.7***	182.7***	167.9***	172.5***	173.9***	159.7***

Note(s): For each independent variable, we report the standardized coefficient and the related t-statistic based on the standard error clustered by firm (Petersen, 2009). Coefficient F-tests compare the stated coefficients with the null hypothesis of equality (one-tailed significance of null rejection). ^***, ^**, ^*significant at 0.01, 0.05 and 0.10, respectively. Δ signifies the difference in the level of the variable between period t and t−1. Lagged level terms (t−1) of independent variables (excluding dummy variables) are added to control for feedback effects and are not reported for brevity. For brevity, the intercept and the coefficients on industry dummies, firm size dummies, book to market dummies and year dummies are not reported. The maximum variance inflation factor excludes dummy variables and lagged terms. Appendix 2 provides variable definitions

Table 5

Preferred stock analysis

Panel A: Association of leverage components and idiosyncratic risk – First-difference OLS regression (1986–2016) with lagged levels and fixed effects for size deciles, book to market deciles, industry and year (N = 102,928) – Total preferred stock added
Dependent/Independent variable	Equation (3a)			Equation (3b)
	ΔIRISKt	ΔIRISKt3	ΔIRISKt5	ΔIRISKt	ΔIRISKt3	ΔIRISKt5
Δ(CL/CE)t	0.2076*** (36.3738)	0.1492*** (28.5524)	0.1514*** (27.8067)
Δ(NCL/CE)t	0.1176*** (22.5868)	0.0801*** (16.9040)	0.0849*** (17.2361)
Δ(TOL/CE)t				0.1815*** (25.9177)	0.1327*** (22.4362)	0.1372*** (21.9465)
Δ(TFL/CE)t				0.1220*** (18.2776)	0.0831*** (14.5003)	0.0866*** (14.2233)
Δ(PS/CE)t	0.0229*** (4.9616)	0.0246*** (5.7697)	0.0207*** (4.7342)	0.0219*** (4.7322)	0.0241*** (5.6699)	0.0200*** (4.6032)
Δ(STD_ROA)t	0.0658*** (14.5345)	0.0985*** (22.4221)	0.0766*** (16.6382)	0.0663*** (14.6396)	0.0989*** (22.5055)	0.0768*** (16.6890)
Δ(BID_ASK)t	0.1807*** (31.4766)	0.1712*** (34.0666)	0.1822*** (35.0163)	0.1826*** (31.7952)	0.1730*** (34.5290)	0.1839*** (35.3426)
Lagged level terms	Included	Included	Included	Included	Included	Included
Industry FE	Included	Included	Included	Included	Included	Included
Size Decile FE	Included	Included	Included	Included	Included	Included
BM Decile FE	Included	Included	Included	Included	Included	Included
Year FE	Included	Included	Included	Included	Included	Included
Coefficient F-tests
Δ(CL/CE)t = Δ(NCL/CE)t	1317***	832.2***	798.3***
Δ(CL/CE)t = Δ(PS/CE)t	697.5***	447.0***	421.2***
Δ(NCL/CE)t = Δ(PS/CE)t	280.4***	167.0***	167.3***
Δ(TOL/CE)t = Δ(TFL/CE)t				1279***	813.3***	795.5***
Δ(TOL/CE)t = Δ(PS/CE)t				358***	278***	262.2***
Δ(TFL/CE)t = Δ(PS/CE)t				185***	126.6***	116.9***
Maximum VIF	1.74	1.74	1.74	1.93	1.93	1.93
Adjusted R2	0.231	0.307	0.309	0.231	0.306	0.310
F-value	184.4***	185.5***	169.7***	184.4***	186.9***	171.1***
Note(s): For each independent variable, we report the standardized coefficient and the related t-statistic based on the standard error clustered by firm (Petersen, 2009). Coefficient F-tests compare the stated coefficients with the null hypothesis of equality (one-tailed significance of null rejection). ***, **, *significant at 0.01, 0.05 and 0.10, respectively. Δ signifies the difference in the level of the variable between period t and t−1. Lagged level terms (t−1) of independent variables (excluding dummy variables) are added to control for feedback effects and are not reported for brevity. For brevity, the intercept and the coefficients on industry dummies, firm size dummies, book to market dummies and year dummies are not reported. The maximum variance inflation factor excludes dummy variables and lagged terms. Appendix 2 provides variable definitions

Panel B: Association of leverage components and idiosyncratic risk – First-difference OLS regression (1986-2016) with lagged levels and fixed effects for size deciles, book to market deciles, industry and year (N = 102,928) – preferred stock classes added
Dependent/Independent variable	Equation (3c)			Equation (3d)
	ΔIRISKt	ΔIRISKt3	ΔIRISKt5	ΔIRISKt	ΔIRISKt3	ΔIRISKt5
Δ(COLx/CE)t	0.1689*** (25.2025)	0.1206*** (20.5806)	0.1261*** (21.2139)	0.1687*** (25.2038)	0.1206*** (20.6110)	0.1259*** (21.2323)
Δ(IP/CE)t	0.0021 (0.5461)	0.0035 (0.9835)	0.0033 (0.9547)	0.0021 (0.5708)	0.0036 (1.0372)	0.0035 (1.0106)
Δ(DTX/CE)t	0.0008 (0.2314)	−0.0056* (−1.6721)	−0.0019 (–0.5556)	0.0009 (0.2605)	−0.0054 (−1.6136)	−0.0017 (−0.4872)
Δ(ONCL/CE)t	0.0598*** (11.5758)	0.0476*** (10.6478)	0.0485*** (10.1328)	0.0598*** (11.5920)	0.0476*** (10.6871)	0.0485*** (10.1627)
Δ(STD/CE)t	0.0856*** (15.2564)	0.0661*** (13.0112)	0.0641*** (12.1684)	0.0856*** (15.2480)	0.0660*** (13.0180)	0.0640*** (12.1762)
Δ(LTD/CE)t	0.0605*** (10.2850)	0.0367*** (6.8356)	0.0383*** (6.9946)	0.0605*** (10.2926)	0.0371*** (6.9033)	0.0386*** (7.0460)
Δ(CapL/CE)t	0.0267*** (5.6388)	0.0124*** (2.9178)	0.0166*** (3.8437)	0.0268*** (5.6510)	0.0125*** (2.9471)	0.0167*** (3.8662)
Δ(PSR/CE)t	0.0102** (2.4199)	0.0172*** (4.5712)	0.0147*** (3.8751)
Δ(PSNR/CE)t	0.0148*** (3.2593)	0.0148*** (3.4999)	0.0112*** (2.6172)
Δ(PSC/CE)t				0.0188*** (4.0976)	0.0205*** (4.7843)	0.0187*** (4.4828)
Δ(PSNC/CE)t				0.0057 (1.4783)	0.0085** (2.5345)	0.0055 (1.6361)
Δ(STD_ROA)t	0.0642*** (14.2211)	0.0969*** (22.1619)	0.0750*** (16.3741)	0.0642*** (14.2301)	0.0970*** (22.2310)	0.0750*** (16.4297)
Δ(BID_ASK)t	0.1788*** (30.9855)	0.1697*** (33.7069)	0.1808*** (34.6139)	0.1787*** (30.9954)	0.1698*** (33.7609)	0.1808*** (34.7088)
Lagged level terms	Included	Included	Included	Included	Included	Included
Industry FE	Included	Included	Included	Included	Included	Included
Size Decile FE	Included	Included	Included	Included	Included	Included
BM Decile FE	Included	Included	Included	Included	Included	Included
Year FE	Included	Included	Included	Included	Included	Included
Maximum VIF	1.92	1.92	1.92	1.92	1.92	1.92
Adjusted R2	0.232	0.308	0.311	0.233	0.308	0.312
F-value	166.9***	168.3***	154.5***	167.0***	168.4***	154.7***

Note(s): For each independent variable, we report the standardized coefficient and the related t-statistic based on the standard error clustered by firm (Petersen, 2009). ^***, ^**, ^*significant at 0.01, 0.05 and 0.10, respectively. Δ signifies the difference in the level of the variable between period t and t−1. Lagged level terms (t−1) of independent variables (excluding dummy variables) are added to control for feedback effects and are not reported for brevity. For brevity, the intercept, and the coefficients on industry dummies, firm size dummies, book to market dummies and year dummies are not reported. The maximum variance inflation factor excludes dummy variables and lagged terms. Appendix 2 provides variable definitions

Table 6

Financial distress analysis

Panel A: Association of leverage components and idiosyncratic risk – First-difference OLS regression (1986–2016) with lagged levels and fixed effects for size deciles, book to market deciles, industry and year (N = 102,928) – Comparison of financially distressed firms (N = 28,079) to non-distressed firms (N = 74,849) – Total liabilities
Equation (4a)
Dependent/Independent variable	ΔIRISKt			ΔIRISKt3			ΔIRISKt5
	Distressed	Non-distressed	Diff. (p-value)	Distressed	Non-distressed	Diff. (p-value)	Distressed	Non-distressed	Diff. (p-value)
Δ(TL/CE)t	0.2921*** (38.5449)	0.2093*** (28.9193)	0.0828*** (0.0000)	0.2209*** (32.0554)	0.1251*** (17.8885)	0.0958*** (0.0000)	0.2283*** (31.8976)	0.1265*** (16.9214)	0.1018*** (0.0000)
Δ(STD_ROA)t	0.0844*** (11.9629)	0.0226*** (4.2872)	0.0618*** (0.0000)	0.0931*** (14.3050)	0.0856*** (17.0392)	0.0075 (0.3630)	0.0828*** (12.4545)	0.0506*** (10.1290)	0.0322*** (0.0001)
Δ(BID_ASK)t	0.1783*** (20.8677)	0.1687*** (22.1971)	0.0096 (0.3980)	0.1778*** (23.1255)	0.1511*** (22.8205)	0.0267*** (0.0093)	0.1833*** (23.0170)	0.1683*** (24.8301)	0.0150 (0.1520)
Lagged level terms	Included	Included		Included	Included		Included	Included
Industry FE	Included	Included		Included	Included		Included	Included
Size Decile FE	Included	Included		Included	Included		Included	Included
BM Decile FE	Included	Included		Included	Included		Included	Included
Year FE	Included	Included		Included	Included		Included	Included
Maximum VIF	1.62	1.88		1.62	1.88		1.62	1.88
Adjusted R2	0.267	0.175		0.339	0.248		0.334	0.249
F-value	96.05***	116.80***		98.50***	119.50***		88.15***	106.30***
Note(s): Financially distressed firms are distinguished from nondistressed firms by an Altman z-score less than 1.81 (Altman, 1968). For each independent variable, we report the standardized coefficient and the related t-statistic based on the standard error clustered by firm (Petersen, 2009). The p-value (one-tailed) is reported for the difference between standardized coefficients based on a chi-squared test. ***, **, *significant at 0.01, 0.05 and 0.10 respectively. Δ signifies the difference in the level of the variable between period t and t−1. Lagged level terms (t−1) of independent variables (excluding dummy variables) are added to control for feedback effects and are not reported for brevity. For brevity, the intercept, and the coefficients on industry dummies, firm size dummies, book to market dummies and year dummies are not reported. The maximum variance inflation factor excludes dummy variables and lagged terms. Appendix 2 provides variable definitions

Panel B: Association of leverage on idiosyncratic risk – First-difference OLS regression (1986-2016) with lagged levels and fixed effects for size deciles, book to market deciles, industry and year (N = 102,928) – Comparison of financially distressed firms (N = 28,079) to non-distressed firms (N = 74,849) – Total liabilities and total preferred stock
Equation (4b)
Dependent/Independent variable	ΔIRISKt			ΔIRISKt3			ΔIRISKt5
	Distressed	Non-distressed	Diff. (p-value)	Distressed	Non-distressed	Diff. (p-value)	Distressed	Non-distressed	Diff. (p-value)
Δ(TL/CE)t	0.2863*** (37.1271)	0.2084*** (28.8023)	0.0779*** (0.0000)	0.2158*** (30.7980)	0.1235*** (17.5851)	0.0923*** (0.0000)	0.2231*** (30.5148)	0.1266*** (16.9115)	0.0965*** (0.0000)
Δ(PS/CE)t	0.0257*** (3.7218)	0.0047 (0.8746)	0.0210** (0.0168)	0.0223*** (3.6607)	0.0092* (1.7441)	0.0131* (0.0999)	0.0230*** (3.7236)	−0.0012 (−0.2199)	0.0242*** (0.0029)
Δ(STD_ROA)t	0.0842*** (11.9487)	0.0226*** (4.2910)	0.0616*** (0.0000)	0.0930*** (14.3027)	0.0856*** (17.0085)	0.0074 (0.3680)	0.0827*** (12.4551)	0.0506*** (10.1160)	0.0321*** (0.0001)
Δ(BID_ASK)t	0.1777*** (20.8179)	0.1686*** (22.1893)	0.0091 (0.4170)	0.1774*** (23.0965)	0.1509*** (22.7950)	0.0265*** (0.0098)	0.1828*** (22.9921)	0.1683*** (24.8329)	0.0145 (0.1640)
Lagged level terms	Included	Included		Included	Included		Included	Included
Industry FE	Included	Included		Included	Included		Included	Included
Size Decile FE	Included	Included		Included	Included		Included	Included
BM Decile FE	Included	Included		Included	Included		Included	Included
Year FE	Included	Included		Included	Included		Included	Included
Maximum VIF	1.62	1.88		1.62	1.88		1.62	1.88
Adjusted R2	0.267	0.175		0.340	0.249		0.334	0.249
F-value	94.33***	114.7***		96.90***	117.4***		86.65***	104.2***
Note(s): Financially distressed firms are distinguished from non-distressed firms by an Altman z-score less than 1.81 (Altman, 1968). For each independent variable, we report the standardized coefficient and the related t-statistic based on the standard error clustered by firm (Petersen, 2009). The p-value (one-tailed) is reported for the difference between standardized coefficients based on a chi-squared test. ***, **, *significant at 0.01, 0.05 and 0.10, respectively. Δ signifies the difference in the level of the variable between period t and t−1. Lagged level terms (t−1) of independent variables (excluding dummy variables) are added to control for feedback effects and are not reported for brevity. For brevity, the intercept, and the coefficients on industry dummies, firm size dummies, book to market dummies and year dummies are not reported. The maximum variance inflation factor excludes dummy variables and lagged terms. Appendix 2 provides variable definitions

Panel C: Association of leverage components and idiosyncratic risk – First-difference OLS regression (1986-2016) with lagged levels and fixed effects for size deciles, book to market deciles, industry and year (N = 102,928) – Comparison of financially distressed firms (N = 28,079) to non-distressed firms (N = 74,849) – Liability components and total preferred stock
Equation (4c)
Dependent/Independent variable	ΔIRISKt			ΔIRISKt3			ΔIRISKt5
	Distressed	Non-distressed	Diff. (p-value)	Distressed	Non-distressed	Diff. (p-value)	Distressed	Non-distressed	Diff. (p-value)
Δ(COL/CE)t	0.1828*** (19.6064)	0.1381*** (16.3540)	0.0447*** (0.0002)	0.1454*** (18.1849)	0.0722*** (9.1471)	0.0732*** (0.0000)	0.1540*** (19.1132)	0.0754*** (9.4627)	0.0786*** (0.0000)
Δ(NCOL/CE)t	0.0604*** (8.7369)	0.0459*** (6.9318)	0.0145 (0.1180)	0.0543*** (8.9659)	0.0378*** (6.2459)	0.0165** (0.0487)	0.0511*** (7.8487)	0.0461*** (7.3999)	0.0050 (0.5680)
Δ(STD/CE)t	0.0940*** (11.9678)	0.0533*** (7.8878)	0.0407*** (0.0001)	0.0658*** (9.1928)	0.0309*** (4.4397)	0.0349*** (0.0005)	0.0647*** (8.8323)	0.0301*** (4.0493)	0.0346*** (0.0009)
Δ(NCFL/CE)t	0.0705*** (8.9739)	0.0456*** (6.1766)	0.0249** (0.0156)	0.0365*** (5.0383)	0.0331*** (4.6372)	0.0034 (0.7360)	0.0437*** (5.9271)	0.0298*** (3.9571)	0.0139 (0.1760)
Δ(PS/CE)t	0.0243*** (3.5354)	0.0048 (0.8964)	0.0195** (0.0259)	0.0216*** (3.5528)	0.0099* (1.8656)	0.0117 (0.1390)	0.0218*** (3.5209)	−0.0004 (−0.0686)	0.0222*** (0.0064)
Δ(STD_ROA)t	0.0797*** (11.3740)	0.0220*** (4.1767)	0.0577*** (0.0000)	0.0892*** (13.7660)	0.0852*** (16.9844)	0.0040 (0.6270)	0.0788*** (11.9209)	0.0502*** (10.0752)	0.0286*** (0.0004)
Δ(BID_ASK)t	0.1728*** (20.1100)	0.1676*** (22.0075)	0.0052 (0.6480)	0.1735*** (22.6119)	0.1510*** (22.6910)	0.0225** (0.0291)	0.1790*** (22.5737)	0.1689*** (24.7898)	0.0101 (0.3300)
Lagged variable terms	Included	Included		Included	Included		Included	Included
Industry FE	Included	Included		Included	Included		Included	Included
Size Decile FE	Included	Included		Included	Included		Included	Included
BM Decile FE	Included	Included		Included	Included		Included	Included
Year FE	Included	Included		Included	Included		Included	Included
Maximum VIF	1.74	2.27		1.74	2.27		1.74	2.27
Adjusted R2	0.271	0.176		0.343	0.249		0.338	0.250
F-value	89.95***	109.0***		91.98***	111.5***		83.25***	98.98***
Note(s): Financially distressed firms are distinguished from nondistressed firms by an Altman z-score less than 1.81 (Altman, 1968). For each independent variable, we report the standardized coefficient and the related t-statistic based on the standard error clustered by firm (Petersen, 2009). The p-value (one-tailed) is reported for the difference between standardized coefficients based on a chi-squared test. ***, **, *significant at 0.01, 0.05 and 0.10, respectively. Δ signifies the difference in the level of the variable between period t and t−1. Lagged level terms (t−1) of independent variables (excluding dummy variables) are added to control for feedback effects and are not reported for brevity. For brevity, the intercept, and the coefficients on industry dummies, firm size dummies, book to market dummies and year dummies are not reported. The maximum variance inflation factor excludes dummy variables and lagged terms. Appendix 2 provides variable definitions

Panel D: Association of leverage components and idiosyncratic risk – First-difference OLS regression (1986-2016) with lagged levels and fixed effects for size deciles, book to market deciles, industry and year (N = 102,928) – Comparison of financially distressed firms (N = 28,079) to non-distressed Firms (N = 74,849) – Liability components and redeemable/nonredeemable preferred stock classes
Equation (4d)
Dependent/Independent variable	ΔIRISKt			ΔIRISKt3			ΔIRISKt5
	Distressed	Non-distressed	Diff. (p-value)	Distressed	Non-distressed	Diff. (p-value)	Distressed	Non-distressed	Diff. (p-value)
Δ(COL/CE)t	0.1837*** (19.7068)	0.1382*** (16.3913)	0.0455*** (0.0002)	0.1459*** (18.2197)	0.0722*** (9.1544)	0.0737*** (0.0000)	0.1545*** (19.1729)	0.0753*** (9.4453)	0.0787*** (0.0000)
Δ(NCOL/CE)t	0.0605*** (8.7416)	0.0459*** (6.9299)	0.0146 (0.1160)	0.0544*** (8.9616)	0.0378*** (6.2405)	0.0166** (0.0486)	0.0511*** (7.8424)	0.0461*** (7.3957)	0.0050 (0.5720)
Δ(STD/CE)t	0.0942*** (11.9899)	0.0534*** (7.8979)	0.0408*** (0.0001)	0.0659*** (9.2043)	0.0309*** (4.4348)	0.0350*** (0.0005)	0.0648*** (8.8450)	0.0301*** (4.0457)	0.0347*** (0.0008)
Δ(NCFL/CE)t	0.0710*** (9.0479)	0.0458*** (6.2016)	0.0252** (0.0147)	0.0366*** (5.0590)	0.0332*** (4.6411)	0.0034 (0.7310)	0.0439*** (5.9632)	0.0299*** (3.9645)	0.0140 (0.1730)
Δ(PSR/CE)t	0.0150** (2.3970)	−0.0018 (−0.3792)	0.0168** (0.0313)	0.0192*** (3.3835)	0.0059 (1.2753)	0.0133* (0.0766)	0.0200*** (3.5372)	−0.0006 (−0.1314)	0.0206*** (0.0057)
Δ(PSNR/CE)t	0.0138** (2.0248)	0.0046 (0.8356)	0.0092 (0.2940)	0.0093 (1.5364)	0.0084 (1.5776)	0.0009 (0.9100)	0.0090 (1.4867)	−0.0002 (−0.0344)	0.0092 (0.2510)
Δ(STD_ROA)t	0.0796*** (11.3643)	0.0221*** (4.1921)	0.0575*** (0.0000)	0.0892*** (13.7493)	0.0853*** (16.9896)	0.0039 (0.6360)	0.0788*** (11.9160)	0.0502*** (10.0759)	0.0286*** (0.0005)
Δ(BID_ASK)t	0.1730*** (20.1339)	0.1676*** (22.0013)	0.0054 (0.6320)	0.1737*** (22.6417)	0.1509*** (22.6771)	0.0228** (0.0267)	0.1792*** (22.6111)	0.1688*** (24.7850)	0.0100 (0.3160)
Lagged level terms	Included	Included		Included	Included		Included	Included
Industry FE	Included	Included		Included	Included		Included	Included
Size Decile FE	Included	Included		Included	Included		Included	Included
BM Decile FE	Included	Included		Included	Included		Included	Included
Year FE	Included	Included		Included	Included		Included	Included
Maximum VIF	1.73	2.27		1.73	2.27		1.73	2.27
Adjusted R2	0.271	0.176		0.343	0.249		0.337	0.250
F-value	88.62***	106.9***		90.70***	109.6***		81.93***	97.17***
Note(s): Financially distressed firms are distinguished from nondistressed firms by an Altman z-score less than 1.81 (Altman, 1968). For each independent variable, we report the standardized coefficient and the related t-statistic based on the standard error clustered by firm (Petersen, 2009). The p-value (one-tailed) is reported for the difference between standardized coefficients based on a chi-squared test. ***, **, *significant at 0.01, 0.05 and 0.10, respectively. Δ signifies the difference in the level of the variable between period t and t−1. Lagged level terms (t−1) of independent variables (excluding dummy variables) are added to control for feedback effects and are not reported for brevity. For brevity, the intercept, and the coefficients on industry dummies, firm size dummies, book to market dummies and year dummies are not reported. The maximum variance inflation factor excludes dummy variables and lagged terms. Appendix 2 provides variable definitions

Panel E: Association of leverage components and idiosyncratic risk – First-difference OLS regression (1986-2016) with lagged levels and fixed effects for size deciles, book to market deciles, industry and year (N = 102,928) – Comparison of financially distressed firms (N = 28,079) to non-distressed Firms (N = 74,849) – Liability components and convertible/nonconvertible preferred stock classes
Equation (4e)
Dependent/Independent variable	ΔIRISKt			ΔIRISKt3			ΔIRISKt5
	Distressed	Non-distressed	Diff. (p-value)	Distressed	Non-distressed	Diff. (p-value)	Distressed	Non-distressed	Diff. (p-value)
Δ(COL/CE)t	0.1833*** (19.6922)	0.1382*** (16.4185)	0.0451*** (0.0002)	0.1461*** (18.2770)	0.0720*** (9.1498)	0.0741*** (0.0000)	0.1544*** (19.2313)	0.0752*** (9.4630)	0.0792*** (0.0000)
Δ(NCOL/CE)t	0.0604*** (8.7457)	0.0460*** (6.9558)	0.0144 (0.1200)	0.0544*** (8.9930)	0.0380*** (6.2824)	0.0164* (0.0503)	0.0511*** (7.8771)	0.0462*** (7.4237)	0.0049 (0.5780)
Δ(STD/CE)t	0.0939*** (11.9523)	0.0536*** (7.9362)	0.0403*** (0.0001)	0.0656*** (9.1690)	0.0312*** (4.4922)	0.0344*** (0.0006)	0.0643*** (8.8017)	0.0303*** (4.0800)	0.0340*** (0.0010)
Δ(NCFL/CE)t	0.0708*** (9.0185)	0.0461*** (6.2525)	0.0247** (0.0166)	0.0369*** (5.0868)	0.0336*** (4.7213)	0.0033 (0.7460)	0.0440*** (5.9776)	0.0302*** (4.0105)	0.0138 (0.1780)
Δ(PSC/CE)t	0.0213*** (3.1229)	0.0023 (0.4281)	0.0190** (0.0272)	0.0171*** (2.8198)	0.0092* (1.7558)	0.0079 (0.3100)	0.0207*** (3.4815)	−0.0008 (−0.1525)	0.0215*** (0.0065)
Δ(PSNC/CE)t	0.0101* (1.7333)	−0.0029 (−0.6451)	0.0130* (0.0780)	0.0098* (1.9044)	0.0015 (0.3560)	0.0083 (0.2250)	0.0081 (1.5313)	−0.0018 (−0.4661)	0.0099 (0.1350)
Δ(STD_ROA)t	0.0797*** (11.3700)	0.0220*** (4.1818)	0.0577*** (0.0000)	0.0892*** (13.7671)	0.0853*** (16.9944)	0.0039 (0.6330)	0.0788*** (11.9302)	0.0502*** (10.0756)	0.0286*** (0.0005)
Δ(BID_ASK)t	0.1729*** (20.1376)	0.1677*** (22.0161)	0.0052 (0.6450)	0.1736*** (22.6726)	0.1511*** (22.7049)	0.0225** (0.0280)	0.1791*** (22.6675)	0.1689*** (24.8135)	0.0102 (0.3280)
Lagged level terms	Included	Included		Included	Included		Included	Included
Industry FE	Included	Included		Included	Included		Included	Included
Size Decile FE	Included	Included		Included	Included		Included	Included
BM Decile FE	Included	Included		Included	Included		Included	Included
Year FE	Included	Included		Included	Included		Included	Included
Maximum VIF	1.73	2.27		1.73	2.27		1.73	2.27
Adjusted R2	0.271	0.176		0.343	0.249		0.338	0.250
F-value	88.77***	107.0***		90.77***	109.5***		82.18***	97.24***

Note(s): Financially distressed firms are distinguished from non-distressed firms by an Altman z-score less than 1.81 (Altman, 1968). For each independent variable, we report the standardized coefficient and the related t-statistic based on the standard error clustered by firm (Petersen, 2009). The p-value (one-tailed) is reported for the difference between standardized coefficients based on a chi-squared test. ^***, ^**, ^*significant at 0.01, 0.05 and 0.10, respectively. Δ signifies the difference in the level of the variable between period t and t−1. Lagged level terms (t−1) of independent variables (excluding dummy variables) are added to control for feedback effects and are not reported for brevity. For brevity, the intercept and the coefficients on industry dummies, firm size dummies, book to market dummies and year dummies are not reported. The maximum variance inflation factor excludes dummy variables and lagged terms. Appendix 2 provides variable definitions

IRISKt	idiosyncratic risk - the standard deviation of the residual (multiplied by 10) from regressing the daily stock return (ret) on the CRSP value-weighted return (vwretd) for the one-year period from the beginning of the fourth month of fiscal year t until the end of the third month of year t+1.
IRISKt3	idiosyncratic risk - the standard deviation of the residual (multiplied by 10) from regressing the daily stock return (ret) on the CRSP value-weighted return (vwretd) for the three-year period from the beginning of the fourth month of fiscal year t-2 until the end of the third month of year t+1.
IRISKt5	idiosyncratic risk - the standard deviation of the residual (multiplied by 10) from regressing the daily stock return (ret) on the CRSP value-weighted return (vwretd) for the five-year period from the beginning of the fourth month of fiscal year t-4 until the end of the third month of year t+1.
CEt	the market value of common equity at the end of year t (csho*prcc_f).
CLt	current liabilities at the end of year t (lct).
NCLt	noncurrent liabilities at the end of year t (lt – lct – mib).
TOLt	total operating liabilities at the end of year t (lt – dlc – dltt – dclo – mib).
TFLt	total financial liabilities at the end of year t (dlc + dltt + dclo).
COLt	current operating liabilities at the end of year t (lct – dlc).
NCOLt	noncurrent operating liabilities at the end of year t (TOLt – COLt).
STDt	current financial liabilities (short-term debt) at the end of year t (dlc). STD includes short-term notes payable and the current portion of long-term debt.
NCFLt	noncurrent financial liabilities at the end of year t (dltt + dclo).
COLxt	current operating liabilities (excluding taxes payable) at the end of year t (lct-txp-dlc). COLx includes accounts payable, accrued expenses, and other current liabilities (current portion of deferred taxes, unearned revenue, unearned premiums, acceptances outstanding, and loans payable on derivatives), according to Compustat.
IPt	income taxes payable at end of year t (txp).
DTXt	deferred tax liability at the end of year t (txdb).
ONCLt	other noncurrent liabilities at the end of year t (lt – lct – dltt –dclo – txdb – mib). It consists of pension liabilities, contingent liabilities, accounts payable due after one year, assigned accounts receivable, customer deposits, negative goodwill, reserves, foreign exchange losses, facility realignment and relocation, reserves for self-insurance, and investment tax credits, according to Compustat.
LTDt	long-term debt at the end of year t (dltt).
CapLt	capital leases at the end of year t (dclo).
TLt	total liabilities at the end of year t (lt-mib)
PSt	preferred stock at the end of year t (pstk).
PSRt	redeemable preferred stock at the end of year t (pstkr).
PSNRt	nonredeemable preferred stock at the end of t (pstkn).
PSCt	convertible preferred stock at the end of year t (pstkc).
PSNCt	non-convertible preferred stock at the end of year t (pstk – pstkc).
STD_ROAt	The standard deviation of ROA for years t, t-1, and t-2. ROA is equal to net income before interest and income taxes (oiadp) scaled by total assets.
BID_ASKt	The bid-ask spread equal to (Aski,t – Bidi,t)/Mi,t; where Aski,t is the ask price of stock i on day t, Bidi,t is the bid price of stock i on day t, and Mi,t is the mean of Aski,t and Bidi,t. We use the bid-ask spread three months after the fiscal year ends, consistent with the return window for computing idiosyncratic risk.
IND_DUMMIES	dummy variables based on Fama and French (1997) 48 industries.
SIZE_DUMMIES	dummy variables based on size deciles using break-points from the market value of NYSE stocks on the 1st trading date of July in each year.
BM_DUMMIES	dummy variables based on book-to-market ratio deciles. Each firm-year’s book-to-market ratio is calculated as the calendar year-end market value divided by the fiscal year-end book equity value.
YEAR_DUMMIES	dummy variables based on calendar year.
ALTMANt	Altman (1968) bankruptcy prediction z-score for year t, computed as: 1.2X1+ 1.4X2+ 3.3X3+ 0.6X4+ 0.999X5; where X1 = working capital (act-lct)/total assets (at), X2 = retained earnings (re)/total assets, X3 = earnings before interest and tax (ebit)/total assets, X4= market value of equity (csho*prcc_f + pstk)/total liabilities (lt), and X5 = sales (sale)/total assets.

Notes

1. The accounting standard also defines current liabilities as due within the term of the operating cycle, which for most businesses is less than one year.

2. Market frictions include transaction costs, individuals and corporations borrowing at different rates, agency costs, information asymmetry and bankruptcy costs.

3. CFR § 210.5-02 – Balance Sheets applies to commercial and industries companies and excludes investment, insurance and bank holding companies. We exclude financial services firms from our empirical analysis.

4. They include the following. Obligations that have entered into the operating cycle (trade payables, unearned revenue, accruals for wages, salaries, commissions, rentals, royalties and taxes) (ASC 210-10-45-8). “[O]ther liabilities whose regular and ordinary liquidation is expected to occur within a relatively short period, usually 12 months” such as short-term debt (ASC 210-10-45-9).

5. Financially distressed firms that choose to restructure their debt, in half of the cases agree to a privately negotiated plan with their creditors, and in the other half, agree to a court-supervised plan under Chapter 11 (Gilson et al., 1990). Such restructuring may involve selling assets, reformulating terms of the debt contracts and issuing additional debt and common and preferred stock (Hotchkiss et al., 2008). Alternatively, firms can liquidate assets under Chapter 7 of the Code (11 U.S.C.§ 701–784). Nearly all distressed firms choose Chapter 11 restructuring over Chapter 7 liquidation (Warren and Westbrook, 2009).

6. Compared to financial institutions, suppliers visit buyer premises more often, monitor payment discounts not taken (a sign of distress) and monitor the size and timing of purchases (Petersen and Rajan, 1997). Further, companies that secure trade credit can borrow from relatively uninformed banks (more geographically distant, a greater number and a shorter relationship) (Giannetti et al., 2011).

7. Refer to CSSE (CIK 0001679063) FY 2018 Form 10-K filed April 1, 2019, Available from: https://www.sec.gov/Archives/edgar/data/0001679063/000114420419017149/0001144204-19-017149-index.htm.

8. As mentioned, Reg S-X requires “preferred stock subject to mandatory redemption” to be classified as mezzanine equity; however, if the preferred stock meets the definition of ASC 480-10 as mandatorily redeemable, it is classified as a liability (PwC, 2021).

9. FASB ASU 2020-06 simplifies the accounting by requiring fewer cases of bifurcation (FASB, 2020).

10. For instance, Odyssey Marine Exploration, Inc. (OME) has redeemable convertible Series G preferred stock with an embedded derivative reported as a current liability at fair value and an amount reported as mezzanine equity. Refer to OME (CIK 0000798528) FY 2011 Form 10-K filed March 12, 2012, Available from: https://www.sec.gov/Archives/edgar/data/798528/000119312512109723/0001193125-12-109723-index.htm.

11. Compustat reports the OME 2011 redeemable convertible Series G preferred stock as redeemable preferred stock (pskr) and convertible preferred stock (pstkc), and the embedded derivative as a current liability (lct). It reports mandatorily redeemable preferred stock as long-term debt (dltt) for Textron Inc. FY 1996 (CIK 0000217346).

12. For instance, OME FY 2011 Series G preferred stock has voting rights and CSSE FY 2018 Series A preferred stock has voting rights to add two directors if dividends are in arrears for a minimum period.

13. Gebhardt, Lee, & Swaminathan (2001) note that beta is an industry-related risk measure, suggesting its limited usefulness.

14. The holders can redeem it after December 2011, initially at 109% of the liquidation value, and thereafter for an additional 1% per month. OME explores the ocean for shipwrecks and minerals. The redemption feature is attractive to investors who desire to liquidate their investment if prospects are poor.

15. The lower dividend rate frees up additional funds to invest, while offering investors the upside potential of common stock price appreciation (Frankel, 2022). For instance, in 2011, OME Series G preferred stock is convertible into common stock, allowing investors to convert their preferred shares to common shares if the company makes a substantial discovery.

16. See Appendix 2 for all variable definitions.

17. In a robustness test, we add the Altman (1968) bankruptcy prediction z-score (ALTMAN_t) as a control variable and we find that our inferences are unchanged.

18. According to Altman, Iwanicz-Drozdowska, Laitinen, & Suvas (2017), “Even though the Z-Score model was developed more than 45 years ago and many alternative failure prediction models exist, the Z-Score model continues to be used worldwide as a main or supporting tool for bankruptcy or financial distress prediction and analysis both in research and in practice.” We acknowledge that other models have been widely used such as Campbell, Hilscher, & Szilagyi (2008) and Shumway (2001). In robustness tests, we partition by recession and nonrecession years.

19. We find robust results using recession years (1990, 1991, 2001, 2007 to 2009 and 2020) as an alternative to the Altman Z-score cutoff of 1.81, except we do not find a statistically significant difference between recession years and nonrecession years for preferred stock and the individual classes. However, the test seems less valid because not all firms are equally financially distressed during recession years (e.g. Grocers vs. Construction).

20. This is similar to Ohlson and Juettner-Nauroth (2005) without growth.

Appendix 1 Modigliani and Miller (1958, 1963) framework

The purpose of this appendix is to show that from the Modigliani and Miller (1958, 1963) framework the expected rate of return to common stockholders (cost of common equity capital) increases with leverage. The Modigliani and Miller’s (1958, 1963) framework posits that the value of a levered firm equals the value of an unlevered firm plus the tax benefit from debt.

(A1)VL=VU+τDL

where V_L and V_U represent the value of levered and unlevered firms, respectively; D_L represents the value of debt for the levered firm, and τ is the corporate tax rate. By defining V_L = S_L + D_L and V_U = S_U, where S_L and S_U are the market value of equity for levered and unlevered firms, respectively, equation (A1) can be modified as follows:

(A2)SL+DL=SU+τDL

Combining similar terms yields:

(A3)SU=SL+(1−τ)DL

For simplicity, assume that the market value of equity can be expressed as the expected permanent earnings capitalized at the firm’s cost of equity capital (r) [20]. That is:

(A4)SU=EˆU(1−τ)rU

(A4a)SL=(EˆU−iDL)(1−τ)rL

where rU and rL are the cost of equity capital for unlevered and levered firms, respectively. Setting Equations (A4) and (A4a) equal to EˆU(1−τ):

(A5)EˆU(1−τ)=rUSU

(A5a)EˆU(1−τ)=rLSL+iDL(1−τ)

Equating equations (A5) and (A5a) leads to:

(A6)rL=SUSLrU−iDL(1−τ)SL

Change (A3):

(A7)SUSL=1+1−τDLSL

Substituting (A7) for (A6) yields:

(A8)rL=rU+(rU−i)(1−τ)DLSL

Equation (A8) shows that the cost of ‘levered’ equity capital (rL) equals the sum of the cost of ‘unlevered’ equity capital (rU) plus a term that is jointly affected by leverage and the spread between the cost of ‘unlevered’ equity capital and the interest rate (rU−i) and a term that considers the tax deductibility of interest payments.

Appendix 2Variable Definitions

All variables are winsorized at the 1st and 99th percentiles.

All Compustat and CRSP variable names are indicated in italics. Missing values are set to 0 for current liabilities (CL), short-term debt (STD), income taxes payable (IP), deferred taxes (DTX), long-term debt (LTD), capital leases (CapL), preferred stock variables (PS, PSR, PSNR, PSC and PSNC), BID_ASK, mibt, mib, dclo, act, re, and lt.