Inventory Policy Choice and Cost of Debt: A Private Debtholders’ Perspective

Abstract

Prior research suggests that last-in-first-out (LIFO) inventory policy produces higher accruals quality than first-in-first-out (FIFO), leading to lower information risk for LIFO firms. Equity investors, in turn, price lower information risk by giving a premium to LIFO firms. Prior research also suggests that LIFO inventory policy, by understating aggregate earnings and net assets, creates hidden reserves that can be released to produce higher future earnings. Prior research, however, is silent on how debtholders price inventory policy choice. Focusing on private debtholders, we explore three research questions: (a) Do private debtholders price LIFO differently from FIFO? (b) If so, how does downside risk affect the pricing of LIFO? and (c) What is the underlying reason for the pricing effect, accounting quality, or hidden reserves? First, we find that LIFO is associated with a lower loan spread than FIFO. Not surprisingly, the pricing effect of LIFO is larger in high-inflation years than in other years. Second, we find that the pricing effect of LIFO is much larger for unrated firms than for rated firms. Third, we find that the negative association between LIFO and loan spread is driven primarily by hidden LIFO reserves rather than accounting quality. Overall, our evidence shows that inventory policy choice affects private debtholders’ pricing decisions, and this effect is primarily due to future earnings implications of hidden LIFO reserves.

Keywords

LIFO/FIFO inventory policy cost of debt loan spread asymmetric payoffs downside risk unconditional conservatism accruals quality hidden LIFO reserves

Introduction

A large body of accounting research investigates the implications to firm welfare of managers’ inventory policy choice—for example, the choice between last-in-first-out (LIFO) and first-in-first-out (FIFO). Although most of the literature focuses on the market valuation of the cash flow effects of tax shields from inventory policy (see, for example, Abdel-Khalik & McKeown, 1978; Brown, 1980; Sunder, 1973, 1975), Krishnan, Srindhi, and Su (2008) focus on the lower information risk associated with LIFO inventory policy. They argue that LIFO earnings are less variable (e.g., cost of goods sold are better matched with revenue) and have higher accruals quality than FIFO earnings, resulting in lower information risk à la Francis, LaFond, Olsson, and Schipper (2005). Equity investors, in turn, favorably price LIFO inventory policy choice by according a premium to LIFO firms compared with FIFO firms. In addition, Penman and Zhang (2002) point out that LIFO inventory policy creates “hidden” LIFO reserves by understating current earnings and book value of net assets. They document that these hidden reserves, in fact, result in higher future earnings.¹ Interestingly, equity investors fail to correctly price these hidden reserves; Penman and Zhang (2002) document a positive association between LIFO reserves and future stock returns.

Evidence in prior research on inventory policy choice is centered around equity investors, and as such, it overlooks other firm stakeholders such as debtholders. To our knowledge, there is no study that investigates inventory policy choice from the perspective of debtholders. Because debtholders and equity investors have asymmetric claims to firms’ assets and profits, one might expect that debtholders may view inventory policy differently from equity holders. To fill the gap, we revisit the issue and investigate how debtholders price inventory policy choice. Given the size of debt markets, investigation of debtholder perspective is particularly important.²

There are important differences between equity investors and debtholders that might lead to differential pricing of LIFO/FIFO choice. First, debtholders face asymmetric payoffs, whereas equity holders face symmetric payoffs (as long as firms stay solvent). Therefore, debtholders are more concerned about downside risk, whereas equity holders are more concerned about upside potential. Because of the different payoff structures, equity investors and debtholders are likely to price securities differently. For example, Goh, Lim, Lobo, and Tong (2017) document that conditional conservatism is associated with a reduction in cost of equity but not in cost of debt. However, because LIFO inventory policy is unconditional conservatism, it is not clear a priori whether or how debtholders will price lower information risk associated with LIFO and the implication of hidden LIFO reserves for future earnings. On one hand, if a firm is close to default, debtholders would be particularly concerned about information risk and hidden LIFO reserves. On the other hand, if a firm is financially healthy, neither of these features may have any effect on debtholders even though they have effects on equity holders.

Second, some debtholders, in particular, private debtholders such as banks, have better access to the private information of their borrowers. Coupled with superior information-processing abilities, such access to private information enables them to reduce adverse selection costs (Bharath, Sunder, & Sunder, 2008). Debtholders are, therefore, expected to be more sophisticated than equity investors. Similarly, Even-Tov (2017) argues that debtholders are less constrained than equity investors by either limited attention or difficulties in interpreting information. Therefore, the lower information risk associated with LIFO may not have any pricing effect for debtholders even though it has a pricing effect for equity investors. In sum, although it is priced by equity investors, whether inventory policy choice is priced by debtholders is ultimately an empirical question.

We examine three specific research questions in this article. The first research question we investigate is as follows:

Research Question 1: Do private debtholders price LIFO inventory policy differently from FIFO?

Although debt instruments may be public or private, we focus primarily on private debtholders in this study. Firms that issue public debt (e.g., bonds) tend to be larger and more profitable than those that issue only private debt. Public debtholders are, therefore, less likely to be concerned with downside risk than private debtholders. That is, they are less likely to price inventory policy. If LIFO is priced by debtholders, we should find stronger evidence in private debt. In addition, private debtholders (e.g., banks) have access to private information and superior information-processing abilities. Prior research suggests that due to factors such as relationship banking, better processing of information, and economies of scale, banks have better information about future prospects of the borrowers (e.g., Diamond, 1991; Sharpe, 1990). Consistent with this argument, James (1987) finds that the stock market responds more positively to the announcement of bank loans compared with the announcement of other sources of debt. Therefore, private debtholders should be able to better assess future earnings implications of LIFO reserves and greater information risk associated with FIFO than public debtholders. Therefore, we investigate private debtholders’ pricing of inventory policy choice by examining loan spreads over the period from 1987 to 2017, using data from DealScan and Compustat. We find that LIFO firms are, on average, associated with a loan spread 14.04 bps lower than FIFO firms, suggesting that private debtholders price inventory policy in a similar manner as equity investors. This result reveals that LIFO firms, on average, make an annual saving of US$635,781 (14.07 bps multiplied by 451.87 million, the mean loan amount for LIFO firms) in interest compared with FIFO firms.

We also find that the effect of LIFO is stronger in high-inflation years than in other years, confirming the expectation that the difference in pricing of LIFO and FIFO firms by private debtholders is smaller when inflation is low. In addition, we evaluate whether the pricing effect of LIFO on loan spreads can be fully explained by the negative effect of LIFO on reported profitability and other financial variables. We adjust profitability and other financial variables as if a firm used FIFO (“as-if FIFO” inventory policy) and find that the coefficient estimate of LIFO when we use as-if FIFO variables is similar to that in our main results, suggesting that the negative impact of LIFO on current profitability and other financial variables does not fully explain the pricing effect of LIFO on loan spread.³ Our results are robust to controlling for endogeneity and selection bias. We control for endogeneity by using two-stage least squares (2SLS) and self-selection bias by using the Heckman two-stage procedures. In addition, as an alternative method to deal with endogeneity, we use an instrumental variable (IV) approach and obtain similar results.

The second research question we investigate is as follows:

Research Question 2: How does pricing of LIFO by private debtholders differ by the level of downside risk?

In general, debtholders are more concerned about downside risk than upside potential. Therefore, for firms that are close to default, lower information risk and future earnings implications of hidden LIFO reserves may have greater effects on the pricing of LIFO by private debtholders. However, for firms that are not close to default, they may have little effect. To answer the second research question, we proxy default risk by whether a firm is rated or unrated. Firms that issue public debt (rated firms) tend to be larger and more profitable than those that do not issue public debt (unrated firms). Unrated firms are, therefore, more likely to face higher downside risk than rated firms. If so, the existence of hidden reserves and/or higher accrual quality would make a greater impact on the pricing of unrated firms than rated firms. Consistent with these arguments, we find that LIFO is negatively and significantly associated with the loan spreads of unrated firms but not significantly associated with the loan spreads of rated firms. These findings suggest that LIFO is priced by private debtholders for only firms with higher downside risk.

The third research question we investigate is as follows:

Research Question 3: Can we identify the underlying reason why LIFO is priced by private debtholders: Is it accounting quality or hidden LIFO reserves?

If LIFO firms have higher quality accruals, then private debtholders may favorably price their debts. Alternatively, because hidden LIFO reserves imply that earnings could be higher sometime in the future, debtholders might take that potential into account in pricing their debts. To answer this research question, we examine how accounting quality and hidden LIFO reserves affect the negative association between LIFO and loan spread. If the negative association is due to higher accounting quality, then it should disappear once we control for accounting quality. Alternatively, if the negative association is due to the future earnings implications of hidden reserves, the negative association should vanish once we control for these reserves. Our analysis shows that the negative association is not affected much by accounting quality, but it vanishes once we control for hidden LIFO reserves. These results suggest that the pricing effect of LIFO inventory policy choice on private debtholders is largely attributable to hidden LIFO reserves and not to accounting quality. In addition, we find that the LIFO inventory policy is positively associated with future (1 year ahead) earnings growth. However, once we control for LIFO reserves, the positive association between LIFO and future earnings growth is reduced substantially. This result suggests that the higher future earnings growth associated with LIFO is largely driven by hidden LIFO reserves.

This article contributes to several areas of research. First, it contributes to the extant research that investigates the valuation of inventory policy choice. We extend prior research, which focuses on equity investors (Krishnan et al., 2008; Lev & Thiagarajan, 1993), by presenting evidence from the perspective of private debtholders on inventory policy choice. Given that private debtholders face asymmetric payoffs and have better access to private information along with superior ability to process information that can reduce adverse selection costs, their perspective is particularly of relevance when ascertaining the value implications of inventory policy choice. Our findings also have implications for standard setters. As Holthausen and Watts (2001) point out, accounting standards should pay heed to various stakeholders and not just to equity investors; our study addresses this concern. Coupled with evidence in prior studies on equity markets, our findings raise an interesting question: Why is LIFO prohibited in most countries? While U.S. generally accepted accounting principles (GAAP) allows the use of LIFO, International Financial Reporting Standards (IFRS) do not. In fact, the United States is virtually the only country that allows the use of LIFO.⁴ Although there may be an argument for disallowing LIFO in the United States to be consistent with IFRS, our findings show that both equity and debt investors appear to favorably value LIFO over FIFO, suggesting that both types of investors consider LIFO to be a more desirable inventory policy choice than FIFO.

Second, our study extends the literature about the implications of asymmetric payoff structure of debtholders on their valuations. Consistent with the asymmetric payoff structure of debtholders, Easton, Monahan, and Vasvari (2009) document that earnings news has a greater impact on speculative-grade bonds than investment-grade bonds because the former are exposed to a greater downside risk than the latter. We document evidence consistent with Easton et al. (2009) that the impact of LIFO on private debtholders is greater for unrated firms than rated firms, suggesting that the impact of LIFO on private debtholders is greater for firms with greater downside risk.

Third, this article contributes to extant research on the implications of accounting conservatism for debtholders. In the context of conditional conservatism, Goh et al. (2017) document that conditional conservatism has no impact on debtholders, although it has an impact on equity holders. Of course, LIFO inventory policy choice is unconditional conservatism, which typically leads to lower retained aggregate earnings and book value of net assets (Penman & Zhang, 2002; Ruch & Taylor, 2015). We document that private debtholders, like equity investors, positively price this particular form of unconditional conservatism. Thus, our evidence indicates that the difference between equity investors and debtholders concerning the pricing of conditional conservatism does not apply to the pricing of unconditional conservatism. Finally, our article extends the results in Penman and Zhang (2002), who document that hidden reserves created by LIFO positively predict future earnings, but equity investors fail to understand the positive earnings implications of these hidden reserves. We find that private debtholders, in contrast to equity investors, seem to understand these implications and positively price LIFO reserves. Better access to private information, together with the superior ability to reduce adverse selection costs, makes an important difference in the valuation of inventory policy choice for private debtholders compared with equity investors.

The remainder of the article is organized as follows: The next section presents hypotheses and research design. We then present the sample selection and descriptive statistics, followed by the results. We conclude in the last section.

Hypotheses Development and Research Design

The Pricing Effect of Inventory Policy Choice on Loan Spread

We focus on two findings from prior literature on inventory policy: (a) Because LIFO is associated with higher accounting quality and lower information risk, equity investors positively price inventory policy choice by giving an equity premium to LIFO firms as compared with FIFO firms (Krishnan et al., 2008); (b) LIFO leads to hidden reserves, which are positively associated with future earnings (Penman & Zhang, 2002). The implication of these findings to the pricing of inventory policy by private debtholders is not clear, however. Although inventory policy choice is priced by equity investors, private debtholders might price LIFO/FIFO choice differently because, facing different payoffs, they are concerned about different firm attributes. In addition, private debtholders such as banks have advantages over equity investors because they have better access to the private information of their borrowers and possess superior information-processing abilities. Therefore, unlike equity investors, private debtholders may not place value on lower information risk (resulting from higher accrual quality) associated with LIFO. Furthermore, higher future earnings implied by hidden LIFO reserves may not have any effect on the downside risk if a firm is financially healthy. Given that debtholders’ concerns about a firm are quite different from those of equity holders, we cannot a priori predict how inventory policy choice will be priced by private debtholders. Thus, we state the following null hypothesis.

Hypothesis 1 (H1): Loan spread of LIFO firms is not lower than that of FIFO firms.

To investigate the relationship between inventory policy choice and loan spread, we estimate the following pooled cross-sectional ordinary least squares (OLS) regression:

\begin{matrix} SPREA D_{it + 1} = β_{0} + β_{1} LIF O_{it} + β_{2} RO A_{it} + β_{3} PP E_{it} + β_{4} CURREN T_{it} + β_{5} RAT E_{it} + β_{6} LM V_{it} \\ + β_{7} LE V_{it} + β_{8} B M_{it} + β_{9} Z_SCO R_{it} + β_{10} LOS S_{it} + β_{11} PRF_PR C_{it + 1} \\ + β_{12} REVOLV E_{it + 1} + β_{13} MATURIT Y_{it + 1} + β_{14} SECUR E_{it + 1} + β_{15} AMOUN T_{it + 1} \\ + β_{16} TERM_LOA N_{it + 1} + β_{17} SENIO R_{it + 1} + β_{18} # LENDERS + ε_{it + 1} . \end{matrix}

(1)

The definitions of the variables in Equation 1 are presented in the appendix. H1 predicts that the coefficient estimate of LIFO in Equation 1 to be insignificant. We include a number of control variables in Equation 1. We add a profitability measure, ROA, to control for operating performance because superior operating performance would lead to lower spreads by reducing default risk. We add PPE to control for the tangibility of assets, which can be used as collateral. We expect a negative sign on PPE (Bharath et al., 2008). We add CURRENT to control for liquidity risk; firms with high liquidity have a lower cost of debt (Bharath et al., 2008). Myers (1977) shows that firms with a higher credit risk are likely to have a higher cost of debt. We measure credit risk with S&P bond ratings. We expect that the higher the bond rating, the lower the cost of debt; we expect a negative sign on RATE (Beatty, Weber, & Yu, 2008). The cost of debt is expected to be lower for larger firms because they are deemed to be less risky (Fama & French, 1992); we expect a negative sign on LMV (Beatty et al., 2008; Zhang, 2008). The cost of debt increases with leverage (Petersen & Rajan, 1994); we expect a positive sign on LEV. Firms with greater growth options (lower book-to-market) are expected to have a lower cost of debt; we expect a positive sign on BM (Beatty et al., 2008; Bharath et al., 2008). Altman’s Z_SCOR increases with the financial health of the firm; we expect a negative sign on the coefficient (Elliott, Ghosh, & Moon, 2010). We expect a positive sign on LOSS. Easton et al. (2009) document that bond market reactions to earnings news are stronger for loss firms than for profit firms, indicating that bondholders view losses as bad news. We add PRF_PRC and REVOLVE. Asquith, Beatty, and Weber (2005) find that loans are more likely to include performance-pricing features when firms are more likely to have moral hazard and adverse selection costs. They also argue that borrowers with revolving loans have an incentive to increase borrowing amounts if their credit quality declines and, thus, revolving loans increase the lender’s agency costs. We add MATURITY because longer maturity debt will have higher agency costs (Flannery, 1986). Beatty et al. (2008) point out that loans that are large relative to a firm’s assets are riskier. Therefore, we include AMOUNT to reflect the loan amount. In addition, consistent with prior research, we include SECURE, TERM_LOAN, SENIOR, and #LENDER to control for loan characteristics (Beatty et al., 2008; Bharath et al., 2008; Zhang, 2008).

Endogeneity and Self-Selection

In Equation 1, loan-related variables from DealScan are measured in year t+ 1, whereas other firm-specific variables from Compustat are measured in year t. When a loan is initiated in year t+ 1, the inventory policy choice would have already been made in year t. Endogeneity, therefore, may be less of a concern for our estimation. However, a firm might have chosen LIFO in expectation of favorable credit terms for future loans. In that case, a firm’s inventory policy is affected by its financing decision. Therefore, inventory policy choice and borrowing decisions may be endogenously determined.⁵ To address endogeneity, we implement a 2SLS estimation, following Cushing and LeClere (1992) and Krishnan et al. (2008).

In the first stage of 2SLS, we estimate a pooled cross-sectional probit regression to identify the determinants of LIFO choice as follows:

\begin{matrix} LIF O_{it} = β_{0} + β_{1} TAXSAV E_{it} + β_{2} INVMA T_{it} + β_{3} LOSSCAR Y_{it} + β_{4} INVVA R_{it} \\ + β_{5} RD I_{it} + β_{6} LSAL E_{it} + β_{7} LE V_{it} + β_{8} CURREN T_{it} + ε_{it} . \end{matrix}

(2)

The definitions of the variables in Equation 2 are presented in the appendix. We winsorize the independent variables in Equation 2 at the top and bottom 1% of respective distributions. Equation 2 is estimated using all firm-year observations that have data for Equations 1 and 2. After estimating Equation 2 in the first stage of 2SLS, we generate the predicted probability of LIFO choice (predLIFO). In the second stage of 2SLS, we use predLIFO in the estimation of Equation 1 instead of the actual inventory policy choice.

Similar to the issue of endogenous choice in financing and inventory policies, there may be an issue of self-selection. The choice of inventory policy is likely to be affected by managerial incentives, such as tax savings and other firm characteristics. Therefore, there could be a selection bias in the estimation of Equation 1, which could bias the coefficient estimate of LIFO in the OLS estimation. To mitigate this selection bias, we estimate a Heckman two-stage model (Heckman, 1979). In the first stage, we estimate Equation 2 using a bivariate probit estimation and generate an Inverse Mills Ratio (IMR), which is included in the second stage as an additional control variable in the estimation of Equation 1.

As an alternative method to address endogeneity, we use an IV approach instead of 2SLS used by Cushing and LeClere (1992). We use the industry-mean frequency of LIFO as our instrument. As Larcker and Rusticus (2010) point out, industry instruments are frequently used in the accounting literature (e.g., Lev & Sougiannis, 1996; Xue, 2007). To implement an IV approach, we first calculate the mean frequency of LIFO choice for each industry and year using the Fama and French (1997) 48 industry definitions and use this variable as the IV. We then estimate, in the first-stage of 2SLS, the following pooled cross-sectional probit regression.

\begin{matrix} LIF O_{it} = β_{0} + β_{1} INDLIF O_{it} + β_{2} RO A_{it} + β_{3} PP E_{it} + β_{4} CURREN T_{it} + β_{5} RAT E_{it} + β_{6} LM V_{it} \\ + β_{7} LE V_{it} + β_{8} B M_{it} + β_{9} Z_SCO R_{it} + β_{10} LOS S_{it} + β_{11} PRF_PR C_{it + 1} \\ + β_{12} REVOLV E_{it + 1} + β_{13} MATURIT Y_{it + 1} + β_{14} SECUR E_{it + 1} + β_{15} AMOUN T_{it + 1} \\ + β_{16} TERM_LOA N_{it + 1} + β_{17} SENIO R_{it + 1} + β_{18} # LENDERS + ε_{it} . \end{matrix}

(3)

The definitions of the variables in Equation 3 are presented in the appendix. Following Larcker and Rusticus (2010), we include all exogenous variables in Equation 1 along with our instrument, INDLIFO, to generate the predicted value of LIFO (predLIFO). In the second stage, we estimate Equation 1 with predLIFO instead of the actual value of LIFO. We do not include industry-fixed effects in the second stage of 2SLS because our instrument in the first stage of 2SLS is based on industry groups.

Pricing of LIFO for Rated and Unrated Firms

Although the first hypothesis is in null form, conditional on H1 being rejected, we next explore how downside risk affects the pricing of LIFO by private debtholders. In contrast to equity holders who are residual claimants, debtholders have a fixed claim.⁶ In cases of financial distress, however, their returns are likely to be reduced. Thus, debtholders would be more concerned with a firm’s probability of default and what they can recover if the firm defaults (loss-given-default). In bankruptcy, debtholders have a higher priority claim to the firm’s assets, whereas equity investors are likely to receive little or nothing. That is, the payoff functions of equity investors and debtholders are quite different. Therefore, debtholders are more concerned about downside risk than upside potential. Easton et al. (2009) point out that nonlinear payoff structures faced by debtholders can be replicated by a long position in the debtholders’ assets and a short position on a call option on those assets. We can, therefore, predict how debtholders respond to news about a firm by assessing how the news affects the riskiness of the firm. For example, when a call option is deep in the money, news about the firm is less relevant to debtholders than when the call option comes close to being out of the money (Easton et al., 2009). Consistent with these arguments, DeFond and Zhang (2014) document that bond prices react more strongly to bad news than good news, whereas stock prices react more strongly to good news than bad news.

Easton et al. (2009) further argue that earnings are more relevant to holders of speculative-grade debt than to holders of investment-grade debt. The call option embedded in the speculative-grade bond is closer to being out of the money than that in the investment-grade bond. In a similar vein, Elliott et al. (2010) investigate the responses of bondholders and equity holders to sustained earnings growth; they find that such growth leads to lower bond yields for high-risk bonds and higher equity returns for low-risk firms. By contrast, if a firm is financially healthy, higher future earnings implied by hidden LIFO reserves or lower information risk associated with LIFO may not have any bearing on the downside risk. Therefore, hidden LIFO reserves and lower information risk may have little effect on private debtholders of firms that are financially more stable. To proxy the level of downside risk, we focus on whether a firm has public debt (rated firms) or relies solely on private debt (unrated firms). Firms that do not issue public debt tend to be smaller and less profitable, and, therefore, are more risky than those that issue public debt; we interpret the absence of public debt as a manifestation of the riskiness of unrated firms. Given that the downside risk of unrated firms, on average, is greater than that of rated firms, the pricing effect of inventory policy choice for unrated firms is likely to be stronger than that for rated firms. Hence, we state the following hypothesis.

Hypothesis 2 (H2): The pricing effect of LIFO inventory policy on loan spread by private debtholders is greater for unrated firms than for rated firms.

To test H2, we first partition our sample into two groups: (a) firms without bond ratings, UNRATE = 1 in Equation 1a; and (b) firms with bond ratings (UNRATE = 0) from S&P. We then estimate Equation 1 for each group. H2 predicts that the coefficient on LIFO for unrated firms is greater in the absolute magnitude than that for rated firms. To test for the difference in the pricing effect of LIFO between rated and unrated firms, we estimate the following model:

\begin{matrix} SPREA D_{it + 1} = β_{0} + β_{1} LIF O_{it} + β_{2} UNRAT E_{it} + β_{3} LIF O_{it} \times UNRAT E_{it} + β_{4} RO A_{it} \\ + β_{5} PP E_{it} + β_{6} CURREN T_{it} + β_{7} RAT E_{it} + β_{8} LM V_{it} + β_{9} LE V_{it} \\ + β_{10} B M_{it} + β_{11} Z_SCO R_{it} + β_{12} LOS S_{it} + β_{13} PRF_PR C_{it + 1} + β_{14} REVOLV E_{it + 1} \\ + β_{15} MATURIT Y_{it + 1} + β_{16} SECUR E_{it + 1} + β_{17} AMOUN T_{it + 1} \\ + β_{18} TERM_LOA N_{it + 1} + β_{19} SENIO R_{it + 1} + β_{20} # LENDERS + ε_{it + 1} . \end{matrix}

(1a)

The definitions of the variables in Equation 1a are presented in the appendix. The coefficient estimate of LIFO in Equation 1a shows the pricing effect of LIFO for rated firms. The coefficient estimate of LIFO × UNRATE shows the difference between rated and unrated firms in the pricing effect of LIFO. The sum of the coefficients LIFO and LIFO × UNRATE shows the pricing effect of LIFO for unrated firms.

The Impact of Accounting Quality and Hidden Reserves on the Pricing of LIFO

If LIFO is priced by private debtholders, we next attempt to identify the driver of the pricing effect. Is it higher accounting quality or hidden LIFO reserves? If accounting quality drives the pricing effect, then we should not see any pricing effect of LIFO by private debtholders once we control for accounting quality. Hence, we state the following hypothesis:

Hypothesis 3a (H3a): The pricing effect of LIFO inventory policy on private debtholders is driven by higher accounting quality.

As discussed above, by understating cumulative aggregate earnings and the current book value of net assets, LIFO creates hidden reserves, which can be liquidated to result in higher earnings in the future. If the pricing effect of LIFO for private debtholders is driven by hidden LIFO reserves, then we should not see any pricing effect of LIFO by private debtholders once we control for hidden LIFO reserves. Hence, we state the following hypothesis:

Hypothesis 3b (H3b): The pricing effect of LIFO inventory policy on private debtholders is driven by hidden LIFO reserves.

To test H3a and H3b, we first calculate AQ by estimating the following equation by year and industry based on Francis et al. (2005):

TC A_{it} = β_{0} + β_{1} CF O_{it - 1} + β_{2} CF O_{it} + β_{3} CF O_{it + 1} + β_{4} Δ RE V_{it} + β_{5} PP E_{it} + v_{it} .

(4)

The definitions of the variables in Equation 4 are presented in the appendix. We include all firm-year observations from Compustat with available data in the estimation of Equation 4 over 5 years from t− 4 to t. We do not require firm-year observations to have data from DealScan for the estimation of Equation 4 because Equation 4 does not include any loan-related variables. In addition, following Francis et al. (2005), we require each industry to have at least 20 observations in year t to be included in the estimation of Equation 4 and calculate accounting quality as the standard deviation of residual (v_it) defined as AQ, from Equation 4 over the years t− 4 to t. Note that accounting quality is inversely related to AQ. Once we generate accruals quality as described above, we add AQ and LIFO reserves (LIFORSV) into Equation 1 and estimate the following equation:

\begin{matrix} SPREA D_{it + 1} = β_{0} + β_{1} LIF O_{it} + β_{2} A Q_{it} + β_{3} LIFORS V_{it} + β_{4} RO A_{it} + β_{5} PP E_{it} \\ + β_{6} CURREN T_{it} + β_{7} RAT E_{it} + β_{8} LM V_{it} + β_{9} LE V_{it} + β_{10} B M_{it} + β_{11} Z_SCO R_{it} \\ + β_{12} LOS S_{it} + β_{13} PRF_PR C_{it + 1} + β_{14} REVOLV E_{it + 1} + β_{15} MATURIT Y_{it + 1} \\ + β_{16} SECUR E_{it + 1} + β_{17} AMOUN T_{it + 1} + β_{18} TERM_LOA N_{it + 1} + β_{19} SENIO R_{it + 1} \\ + β_{20} # LENDERS + ε_{it + 1} . \end{matrix}

(5)

The definitions of the variables in Equation 5 are presented in the appendix. H3a predicts that AQ explains the pricing of LIFO inventory policy choice. If H3a is supported, then the coefficient estimate of LIFO should be insignificant once we include AQ in the estimation. We expect a positive coefficient on AQ because prior research shows that higher accounting quality, from Equation 4 (lower AQ) leads to a lower cost of debt (Bharath et al., 2008). H3b predicts that LIFO reserves explain the pricing of LIFO inventory policy choice. If H3b is supported, the coefficient estimate of LIFO should be insignificant once we include LIFORSV in the estimation. We expect a negative coefficient on LIFORSV based on the evidence in Penman and Zhang (2002) that LIFO reserves are positively associated with future earnings.

Sample and Descriptive Statistics

Sample Selection

We obtain loan data from the DealScan Loan Pricing Corporation (LPC). The LPC database provides loan spreads and other loan information, such as loan maturity, amount, securitizations, performance-pricing grid, types of loans, and seniority. Because we use loan spread in year t+1, our sample period covers the years 1986 to 2016 for financial variables, and 1987 to 2017 for DealScan variables. We obtain financial data from Compustat Annual Files. We also require that firm-year observations have data available for all the variables in Equation 1. Finally, we exclude firm-year observations with a negative book value of equity and those missing value of LIFO reserves. We obtain 24,331 firm-year-facility observations with predominantly FIFO or LIFO inventory policy choice that have data in Compustat and DealScan (we exclude firm-year observations with other inventory methods).⁷ We match Compustat and DealScan with the unique identifier GVKEY, using a linking table provided by Chava and Roberts (2008).

Descriptive Statistics and Pearson Correlations

Panel A of Table 1 presents the descriptive statistics (i.e., M, SD, P25, median, and P75) for the variables used in Equation 1 for the whole sample. We winsorize the ratio variables at 1% from the top and bottom of the respective distributions to eliminate the impact of extreme observations (i.e., BM, CURRENT, PPE, Z_SCOR, and LEV).⁸ Panel B of Table 1 presents the mean and median values for LIFO and FIFO firms. The mean value of loan spread is 166.34 bps for LIFO firms and 226.45 bps for FIFO firms. Descriptive statistics, therefore, show that private debtholders charge LIFO firms a loan spread 60.11 bps lower than FIFO firms. That is, private debtholders appear to price LIFO/FIFO inventory policy choice. This result suggests that H1 is likely to be rejected.

Table 1.

Descriptive Statistics.

Panel A: Descriptive statistics for the full sample.
	M	SD	P25	Median	P75
SPREAD	212.20	148.75	100.00	200.00	300.00
LIFO	0.2370	0.4253	0.0000	0.0000	0.0000
LIFORSV	0.0073	0.0189	0.0000	0.0000	0.0025
ROA	0.0310	0.0969	0.0093	0.0431	0.0768
PPE	0.2939	0.2008	0.1374	0.2458	0.4110
CURRENT	2.0845	1.2440	1.3008	1.7792	2.5024
RATE	5.0365	6.2633	0.0000	0.0000	11.0000
LMV	6.2737	2.2631	4.6519	6.2712	7.8972
LEV	0.2804	0.1777	0.1478	0.2706	0.4001
BM	0.6255	0.5621	0.2943	0.4773	0.7604
Z_SCOR	3.6339	2.8704	1.9668	2.9573	4.4573
LOSS	0.2041	0.4031	0.0000	0.0000	0.0000
PRF_PRC	0.3655	0.4816	0.0000	0.0000	1.0000
REVOLVE	0.5766	0.4941	0.0000	1.0000	1.0000
MATURITY	3.6492	0.7449	3.2581	4.0254	4.0943
SECURE	0.5455	0.4979	0.0000	1.0000	1.0000
AMOUNT	18.2250	1.8793	17.0344	18.4207	19.5193
TERM_LOAN	0.1569	0.3637	0.0000	0.0000	0.0000
SENIOR	0.9969	0.0554	1.0000	1.0000	1.0000
#LENDERS	7.0759	8.0401	1.0000	4.0000	10.0000

Panel B: M and median values for LIFO and FIFO firms.
	M			MEDIAN
	LIFO	FIFO	p value of difference	LIFO	FIFO	p value of difference
SPREAD	166.34	226.45	.00	137.50	200.00	.00
LIFORSV	0.0276	0.0010	.00	0.0166	0.0000	.00
ROA	0.0443	0.0268	.00	0.0471	0.0415	.00
PPE	0.3363	0.2807	.00	0.3188	0.2242	.00
CURRENT	1.8512	2.1570	.00	1.6711	1.8229	.00
RATE	7.7654	4.1887	.00	9.0000	0.0000	.00
LMV	6.8980	6.0798	.00	7.0011	6.0608	.00
LEV	0.2883	0.2780	.00	0.2779	0.2676	.00
BM	0.6536	0.6167	.00	0.5055	0.4691	.00
Z_SCOR	3.3790	3.7130	.00	2.9966	2.9507	.04
LOSS	0.1302	0.2271	.00	0.0000	0.0000	.00
PRF_PRC	0.4025	0.3541	.00	0.0000	0.0000	.00
REVOLVE	0.5764	0.5767	.97	1.0000	1.0000	.96
MATURITY	3.6505	3.6488	.88	4.0943	3.9890	.00
SECURE	0.4004	0.5906	.00	0.0000	1.0000	.00
AMOUNT	18.8141	18.0420	.00	18.9513	18.1975	.00
TERM_LOAN	0.1325	0.1645	.00	0.0000	0.0000	.00
SENIOR	0.9967	0.9970	.73	1.0000	1.0000	.74
#LENDERS	9.0730	6.4555	.00	6.0000	4.0000	.00
N	5,767	18,564

Note. Panel A presents the descriptive statistics (M, SD, P25, median, and P75) for the whole sample. P25 (P75) is the bottom (top) 25%. Panel B presents the mean and median for LIFO and FIFO firms. The difference column for mean shows the p value of t test for equality of means between LIFO and FIFO firms. The difference column for median shows the p value of Wilcoxon rank sum test for equality of medians between LIFO and FIFO firms. LIFO (FIFO) firms are those that use predominantly LIFO (FIFO) in year t. The definitions of variables are in the appendix. LIFO = last-in-first-out; FIFO = first-in-first-out.

LIFO firms are, on average, quite different from FIFO firms in many dimensions. They are more profitable (the mean ROA is 0.0443 for LIFO firms and 0.0268 for FIFO firms) and invest more in tangible assets than FIFO firms (the mean PPE is 0.3363 for LIFO firms and 0.2807 for FIFO firms). LIFO firms have higher bond ratings than FIFO firms, suggesting that FIFO firms are riskier than LIFO firms (the mean RATE is 7.7654 for LIFO firms and 4.1887 for FIFO firms). LIFO firms are larger in market capitalization (the mean LMV is 6.8980 for LIFO firms and 6.0798 for FIFO firms) and are slightly more leveraged than FIFO firms (the mean LEV is 0.2883 for LIFO firms and 0.2780 for FIFO firms). Loss frequency is much higher for FIFO firms than LIFO firms, which is consistent with LIFO firms being more profitable (the mean LOSS is 0.1302 for LIFO firms and 0.2271 for FIFO firms). Debtholders are more likely to secure their debt with collateral for FIFO firms than LIFO firms, which is consistent with FIFO firms being riskier than LIFO firms (the mean value of SECURE is 0.4004 for LIFO firms and 0.5906 for FIFO firms). The frequency of firm-year observations with FIFO inventory policy choice in our sample is 3.22 times greater than that for LIFO, indicating that FIFO is a more widespread inventory policy choice than LIFO (our sample consists of 18,564 for FIFO firms and 5,767 for LIFO firms). The mean LIFORSV for LIFO firms is 0.0276; that is, hidden reserves constitute 2.76% of total assets for LIFO firms. LIFORSV is 0.0010 for FIFO firms; FIFO firms do not have economically significant LIFO reserves. Overall, the descriptive statistics demonstrate that there are important differences between FIFO and LIFO firms in financial and loan characteristics, pointing to the need to control for these differences (self-selection) in exploring the impact of inventory policy choice on the cost of debt.

Table 2 presents Pearson correlations for the sample. A negative correlation between LIFO and SPREAD again suggests that H1 is likely to be rejected. Of course, this correlation fails to control for other factors that might affect the relationship between inventory policy choice and the cost of debt. SPREAD has significant correlations with all the control variables in Equation 1, pointing to the need to control for them in our estimation. LIFO is significantly correlated with all the independent variables in Equation 1, illustrating that there are significant differences between LIFO and FIFO firms in most of the firm characteristics (except maturity). As expected, there is a strong positive correlation between LIFO reserves and LIFO—the correlation between LIFO and LIFORSV is .60. Moreover, there is a negative correlation between SPREAD and LIFORSV, suggesting that LIFO reserves are positively priced by private lenders.

Table 2.

Pearson Correlations.

		1	2	3	4	5	6	7	8	9	10	11	12	13	14	15	16	17	18	19
1	SPREAD	1.00
2	LIFO	−.17	1.00
3	LIFORSV	−.13	.60	1.00
4	ROA	−.31	.08	.05	1.00
5	PPE	−.08	.12	.07	.01	1.00
6	CURRENT	.01	−.10	−.02	.09	−.29	1.00
7	RATE	−.28	.24	.10	.12	.07	−.24	1.00
8	LMV	−.40	.15	.04	.30	.03	−.15	.66	1.00
9	LEV	.16	.02	−.05	−.21	.21	−.31	.14	−.07	1.00
10	BM	.22	.03	.05	−.26	.07	.02	−.16	−.43	.06	1.00
11	Z_SCOR	−.17	−.05	.00	.44	−.16	.47	−.16	.09	−.57	−.29	1.00
12	LOSS	.31	−.10	−.05	−.70	−.01	−.05	−.15	−.27	.16	.27	−.27	1.00
13	PRF_PRC	−.22	.04	.01	.09	.00	−.01	.05	.12	−.04	−.05	.04	−.08	1.00
14	REVOLVE	−.18	.00	.03	.00	−.01	.06	−.08	−.08	−.11	.04	.06	.00	.13	1.00
15	MATURITY	−.01	.00	−.01	.09	.05	−.01	.01	.10	.09	−.05	−.06	−.09	.14	.08	1.00
16	SECURE	.46	−.16	−.13	−.21	−.07	.05	−.29	−.37	.14	.16	−.11	.22	.04	−.06	.13	1.00
17	AMOUNT	−.37	.17	.07	.21	.07	−.18	.54	.75	.08	−.17	−.08	−.19	.22	.01	.18	−.28	1.00
18	TERM_LOAN	.15	−.04	−.03	−.05	.04	.00	−.11	−.16	.06	.07	−.04	.04	−.08	−.50	.11	.07	−.24	1.00
19	SENIOR	−.12	.00	.00	.00	−.02	.00	.02	.01	.00	−.01	.01	.00	.04	.06	.01	.01	.00	−.01	1.00
20	#LENDERS	−.27	.14	.03	.11	.06	−.16	.40	.48	.08	−.13	−.08	−.13	.24	−.02	.12	.16	.57	−.09	.02

Note. This table presents Pearson correlations. Correlations in bold are significant at 5%. The definitions of variables are presented in the appendix.

Results

The Pricing Effect of Inventory Policy Choice on Loan Spread (H1)

Table 3 presents the results of the pooled cross-sectional OLS estimation of Equation 1. We cluster firm-year observations by firm to eliminate autocorrelation as per Petersen (2009). We present two estimations to explore the impact of industry-fixed effects on the results. The first estimation, Model 1, when we do not control for industry-fixed effects, shows a coefficient estimate of LIFO of −18.87 (p < .01), whereas Model 2, when we control for industry effects, shows a coefficient estimate of −14.07 (p < .01). These results suggest that LIFO has a significant impact on debtholders even after controlling for the industry effects. This finding shows the strength of the impact of LIFO on private debtholders; H1 is rejected.⁹

Table 3.

The Pricing Effect of Inventory Policy on Loan Spread.

	Model 1		Model 2
	Coefficient	t stat	Coefficient	t stat
LIFO	−18.87	−6.38***	−14.07	−4.58***
ROA	−111.96	−6.29***	−109.66	−6.21***
PPE	−54.57	−7.94***	−90.22	−10.12***
CURRENT	−2.78	−2.41**	−1.90	−1.64*
RATE	−1.09	−3.50***	−1.03	−3.37***
LMV	−2.96	−2.06**	−3.51	−2.41**
LEV	66.70	7.27***	71.08	7.72***
BM	18.40	5.56***	20.08	6.21***
Z_SCOR	−1.16	−1.86*	−1.21	−1.91*
LOSS	32.67	8.05***	31.22	7.78***
PRF_PRC	−42.83	−21.21***	−42.36	−21.18***
REVOLVE	−45.35	−19.08***	−45.04	−19.01***
MATURITY	7.01	3.82***	7.00	3.88***
SECURE	91.08	33.77***	88.94	32.90***
AMOUNT	−9.85	−7.37***	−9.79	−7.27***
TERM_LOAN	−9.87	−2.53**	−9.30	−2.42**
SENIOR	−299.29	−8.13***	−301.85	−8.37***
#LENDERS	−1.07	−6.94***	−1.10	−7.23***
Industry-fixed effects	No		Yes
R ²	.3960		.4042
N	24,331		24,331

\begin{matrix} SPREA D_{it + 1} = β_{0} + β_{1} LIF O_{it} + β_{2} RO A_{it} + β_{3} PP E_{it} + β_{4} CURREN T_{it} + β_{5} RAT E_{it} + β_{6} LM V_{it} + β_{7} LE V_{it} + β_{8} B M_{it} \\ + β_{9} Z_SCO R_{it} + β_{10} LOS S_{it} + β_{11} PRF_PR C_{it + 1} + β_{12} REVOLV E_{it + 1} + β_{13} MATURIT Y_{it + 1} + β_{14} SECUR E_{it + 1} \\ + β_{15} AMOUN T_{it + 1} + β_{16} TERM_LOA N_{it + 1} + β_{17} SENIO R_{it + 1} + β_{18} # LENDERS + ε_{it + 1} . \end{matrix}

(1)

Note. This table presents the pooled cross-sectional OLS estimation results of Equation 1. The definitions of the variables are in the appendix. Model 1 (Model 2) presents the estimation results of Equation 1 without (with) industry-fixed effects. Industry-fixed effects are based on Fama and French (1997) 48 industry definitions. Firm-year observations are clustered by firm to eliminate autocorrelation as per Petersen (2009). OLS = ordinary least squares.

, **, and *** indicate statistical significance at the 10%, 5%, and 1% levels, respectively.

Next, we investigate the impact of LIFO during high- and low-inflation years. Results are reported in Table 4. We define a high-inflation year as the year in which the inflation rate is in the top 20% of our sample period (6 out of 31 years are considered high-inflation years). During high-inflation years, the differences in cost of goods sold and inventory values between FIFO and LIFO firms will be much larger, and the impact of the LIFO/FIFO inventory policy choice on private debtholders will be greater. The first estimation presents the results for the low-inflation years. The coefficient estimate of LIFO is −11.97 (p < .01). The second estimation presents the results for the high-inflation years; the coefficient estimate of LIFO is −24.39 (p < .01). These results show that LIFO has a 12.42 bps (i.e., the difference in the coefficient estimates between high- and low-inflation subperiods is significant at p = .01) higher impact on private debtholders during the high-inflation years than during the low-inflation years. Moreover, the association between LIFO and loan spread is more than twice higher in the high-inflation period than in the low-inflation period, indicating that the difference is economically significant. The last estimation in Table 4 presents the results when we interact HINF with LIFO. HINF is an indicator variable, which equals 1 if a year is a high-inflation year and 0 otherwise. The coefficient on the interaction variable LIFO × HINF is −15.18 and marginally significant, suggesting that the pricing effect of LIFO is greater in high-inflation years than in low-inflation years. These results show that the difference in the coefficient estimates of LIFO between high- and low-inflation years is both economically and statistically significant.¹⁰ In an untabulated analysis, we partition our sample into three subperiods (1987-1997, 1998-2007, and 2008-2017) to see whether there is any discernable time trend. We find that the coefficients on LIFO are negative and significant only in the first two subperiods. Because the last decade had inflation significantly lower than the earlier period, this result is consistent with our expectation.

Table 4.

The Pricing Effect of Inventory Policy on Loan Spread: High- and Low-Inflation Years.

	Low inflation		High inflation		Full sample
	Coefficient	t stat	Coefficient	t stat	Coefficient	t stat
LIFO	−11.97	−4.07***	−24.39	−2.81***	−12.65	−4.23***
HINF					21.30	4.72***
LIFO × HINF					−15.18	−1.81*
ROA	−125.21	−6.62***	−49.58	−1.15	−113.40	−6.44***
PPE	−82.50	−9.50***	−107.19	−4.19***	−91.93	−10.32***
CURRENT	−1.44	−1.21	−4.78	−1.40	−2.04	−1.77*
RATE	−1.28	−4.18***	0.79	1.02	−1.10	−3.58***
LMV	−2.06	−1.43	−13.29	−3.07***	−2.70	−1.80*
LEV	68.44	7.19***	58.88	2.36**	68.98	7.50***
BM	19.19	6.20***	20.35	1.98**	20.14	6.28***
Z_SCOR	−1.05	−1.63	−2.33	−1.16	−1.21	−1.90*
LOSS	29.23	6.98***	39.46	3.68***	31.67	7.92***
PRF_PRC	−44.85	−21.82***	88.16	6.99***	−39.78	−18.92***
REVOLVE	−44.95	−18.96***	−52.88	−6.22***	−45.92	−19.33***
MATURITY	16.83	8.71***	−28.27	−7.26***	7.69	4.27***
SECURE	87.88	30.88***	81.23	12.31***	89.84	33.13***
AMOUNT	−10.75	−7.91***	−3.22	−1.01	−9.62	−7.14***
TERM_LOAN	−11.01	−2.67***	−2.80	−0.27	−10.54	−2.74***
SENIOR	−275.44	−6.37***	−333.88	−5.67***	−300.40	−8.37***
#LENDERS	−1.14	−7.83***	−0.81	−1.32	−1.15	−7.45***
Industry-fixed effects	Yes		Yes		Yes
R ²	.4155		.4050		.4057
N	21,078		3,253		24,331

Note. This table presents the pooled cross-sectional OLS estimation results of Equation 1 (presented in Table 3) for high- and low-inflation years. High-inflation years are the fiscal years during which inflation rates are in the top quintile of the annual inflation rates. Low-inflation years are the remainder of fiscal years. HINF is an indicator variable that equals 1 if a year is high-inflation year and 0 otherwise. The definitions of the variables are presented in the appendix. Industry-fixed effects are based on Fama and French (1997) 48 industry definitions. Firm-year observations are clustered by firm to eliminate autocorrelation as per Petersen (2009). OLS = ordinary least squares.

, **, and *** indicate statistical significance at the 10%, 5%, and 1% levels, respectively.

Our analysis so far is based on the variable values reported in the financial statements of our sample firms. We have not taken into account that different inventory valuation methods generate different financial numbers that are not comparable. Although lenders might price loans by using these unadjusted figures because firms provide disclosures on inventory policy and LIFO reserves, lenders, in principle, could adjust LIFO earnings and inventory numbers to those under FIFO. After the adjustment, LIFO may not have any impact on loan spread. To evaluate whether the pricing effect of LIFO on loan spread is driven by the negative effect of LIFO on unadjusted financial numbers, we try to undo the impact of LIFO on reported financial variables. Specifically, we recalculate ROA, PPE, CURRENT, LEV, and BM on the FIFO basis (“as-if FIFO”). We then estimate the following pooled cross-sectional OLS regression:

\begin{matrix} SPREA D_{it + 1} = β_{0} + β_{1} LIFO + β_{2} ROA_F_{it} + β_{3} PPE_F_{it} + β_{4} CURRENT_F_{it} + β_{5} RAT E_{it} + β_{6} LM V_{it} \\ + β_{7} LEV_F_{it} + β_{8} BM_F_{it} + β_{9} Z_SCO R_{it} + β_{10} LOS S_{it} + β_{11} PRF_PR C_{it + 1} \\ + β_{12} REVOLV E_{it + 1} + β_{13} MATURIT Y_{it + 1} + β_{14} SECUR E_{it + 1} + β_{15} AMOUN T_{it + 1} \\ + β_{16} TERM_LOA N_{it + 1} + β_{17} SENIO R_{it + 1} + β_{18} # LENDERS + ε_{it + 1} . \end{matrix}

(1b)

The definitions of the variables in Equation 1b are presented in the appendix. If the pricing impact of LIFO on loan spread is determined by the way different inventory methods generate different financial numbers, once we adjust them on the FIFO basis, there should be no pricing effect of LIFO on loan spread. That is, the coefficient on LIFO in Equation 1b should be insignificant once we replace variable values with adjusted numbers. In untabulated results, we find that the mean value of ROA_F for LIFO firms is 0.0438, similar to the mean value of ROA, 0.0443, in Table 1. Adjusting to the FIFO basis does not make much of a difference in ROA even though it increases profits.¹¹

The pooled cross-sectional OLS regression results of Equation 1b are presented in Table 5. The sample size in Table 5 is 23,976, slightly smaller than the 24,331 in Table 3 because we require firm-year observations to have LIFO reserves in both years t− 1 and t to calculate the change in LIFO reserves. The first estimation does not include industry-fixed effects, but the second estimation includes them. The coefficient estimate of LIFO in the second estimation is −14.92 and statistically significant (p < .01). Compared with the coefficient estimate of LIFO in Table 3, −14.07, the results in Table 5 suggest that adjusting profitability and other financials to the FIFO basis has a very little impact on the pricing impact of LIFO on loan spread. That is, the impact of LIFO on loan spread in Table 5 using as-if FIFO is similar to that in Table 3. These results in Table 5 suggest that the pricing impact of LIFO on loan spread is not driven by the impact of LIFO on current profitability or other financial variables. Although private lenders may adjust (LIFO) financial numbers on a FIFO basis, it appears that their analysis goes beyond adjusted numbers and incorporates the implications of LIFO reserves (which are disclosed in footnotes).

Table 5.

The Pricing Effect of Inventory Policy on Loan Spread As-If FIFO Inventory Policy.

	Model 1		Model 2
	Coefficient	t stat	Coefficient	t stat
LIFO	−19.31	−6.50***	−14.92	−4.86***
ROA_F	−110.68	−6.21***	−108.62	−6.15***
PPE_F	−55.03	−7.90***	−91.23	−10.04***
CURRENT_F	−3.03	−2.66***	−2.12	−1.85*
RATE	−1.07	−3.43***	−1.02	−3.32***
LMV	−2.87	−1.98**	−3.40	−2.30**
LEV_F	67.20	7.29***	71.71	7.79***
BM_F	18.17	5.62***	19.67	6.24***
Z_SCOR	−1.13	−1.80*	−1.18	−1.85*
LOSS	32.17	7.89***	30.71	7.62***
PRF_PRC	−42.83	−21.19***	−42.39	−21.14***
REVOLVE	−45.15	−18.83***	−44.83	−18.72***
MATURITY	6.94	3.75***	6.89	3.79***
SECURE	91.66	33.84***	89.48	32.95***
AMOUNT	−9.91	−7.38***	−9.84	−7.28***
TERM_LOAN	−9.21	−2.35**	−8.60	−2.23**
SENIOR	−302.89	−8.15***	−305.57	−8.40***
#LENDERS	−1.14	−7.56***	−1.17	−7.91***
Industry-fixed effects	No		Yes
R ²	.3974		.4056
N	23,976		23,976

\begin{matrix} SPREA D_{it + 1} = β_{0} + β_{1} LIFO + β_{2} ROA_F_{it} + β_{3} PPE_F_{it} + β_{4} CURRENT_F_{it} + β_{5} RAT E_{it} + β_{6} LM V_{it} + β_{7} LEV_F_{it} \\ + β_{8} BM_F_{it} + β_{9} Z_SCO R_{it} + β_{10} LOS S_{it} + β_{11} PRF_PR C_{it + 1} + β_{12} REVOLV E_{it + 1} + β_{13} MATURIT Y_{it + 1} \\ + β_{14} SECUR E_{it + 1} + β_{15} AMOUN T_{it + 1} + β_{16} TERM_LOA N_{it + 1} + β_{17} SENIO R_{it + 1} \\ + β_{18} # LENDERS + ε_{it + 1} . \end{matrix}

(1b)

Note. This table presents the pooled cross-sectional OLS estimation results of Equation 1b. The definitions of the variables are in the appendix. Industry-fixed effects are based on Fama and French (1997) 48 industry definitions. Firm-year observations are clustered by firm to eliminate autocorrelation as per Petersen (2009). FIFO = first-in-first-out; OLS = ordinary least squares.

, **, and *** indicate statistical significance at the 10%, 5%, and 1% levels, respectively.

Results on Endogeneity and Self-Selection

Table 6 presents the pooled cross-sectional estimation results of OLS, 2SLS, the Heckman two-stage procedures, and the IV approach. We require firm-year observations to have nonmissing values for all the variables in the first-stage estimation of 2SLS in Equation 2.¹² We also exclude from our sample the firm-year observations with a negative value of operating income (OIBDP from Compustat) because operating income is used as the denominator in calculating LOSSCARY in Equation 2.

Table 6.

The Pricing Effect of Inventory Policy on Loan Spread: Endogeneity and Self-Selection.

	OLS		2SLS		Heckman		IV
	Coefficient	t stat	Coefficient	t stat	Coefficient	t stat	Coefficient	t stat
LIFO	−13.90	−3.42***	−29.06	−2.13**	−12.99	−3.15***	−45.37	−7.88***
ROA	−87.01	−2.00**	−85.51	−1.96**	−85.49	−1.95**	−110.91	−6.24***
PPE	−96.84	−7.18***	−102.81	−7.81***	−95.99	−7.12***	−50.07	−7.23***
CURRENT	−0.42	−0.23	−0.88	−0.49	−0.53	−0.29	−3.07	−2.67***
RATE	−1.43	−3.51***	−1.42	−3.45***	−1.39	−3.38***	−0.70	−2.19**
LMV	3.09	1.48	3.59	1.73*	3.30	1.58	−3.43	−2.40**
LEV	68.40	5.16***	65.94	4.93***	69.31	5.20***	64.99	7.09***
BM	36.31	6.30***	36.94	6.41***	36.71	6.33***	19.81	5.99***
Z_SCOR	−2.03	−1.67*	−2.18	−1.78*	−2.07	−1.69*	−1.16	−1.87*
LOSS	33.53	5.21***	33.59	5.18***	33.53	5.18***	30.86	7.62***
PRF_PRC	−41.30	−15.53***	−41.84	−15.70***	−41.59	−15.64***	−42.38	−20.92***
REVOLVE	−44.34	−13.64***	−43.98	−13.51***	−44.05	−13.53***	−45.33	−19.23***
MATURITY	13.24	5.03***	12.98	4.93***	13.39	5.13***	7.01	3.82***
SECURE	94.85	25.25***	95.57	25.61***	95.25	25.35***	88.96	32.93***
AMOUNT	−11.66	−6.26***	−11.43	−6.07***	−11.56	−6.15***	−9.46	−7.10***
TERM_LOAN	1.28	0.23	1.59	0.28	1.75	0.31	−9.98	−2.57***
SENIOR	−418.33	−8.90***	−418.41	−8.91***	−418.15	−8.91***	−299.82	−8.10***
#LENDERS	−1.22	−6.78***	−1.24	−6.96***	−1.21	−6.73***	−1.03	−6.63***
IMR					3.15	0.98
Industry-fixed effects	Yes		Yes		Yes		No
R ²	.4221		.4215		.4235		.3972
N	12,259		12,259		12,209		24,331

Note. This table presents the pooled cross-sectional estimation results of Equation 1 (presented in Table 3). OLS is ordinary least squares. 2SLS is two-stage least squares. Heckman is Heckman two-stage procedure. IMR is Inverse Mills Ratio generated from the estimation of Equation 2 in the first stage of the Heckman procedure. IV is the instrumental variable approach. In the first stage of the IV approach, we generate the predicted value of LIFO (predLIFO) by regressing LIFO on INDLIFO (industry-mean LIFO) and other exogenous variables using Equation 3. In the second stage, we estimate Equation 1 using predLIFO instead of the actual value of LIFO. INDLIFO is industry-mean LIFO in year t. Industry groups are based on Fama and French (1997) 48 industry definitions. Firm-year observations are clustered by firm to eliminate autocorrelation as per Petersen (2009). The definitions of the variables are presented in the appendix.

, **, and *** indicate statistical significance at the 10%, 5%, and 1% levels, respectively.

The first estimation in Table 6 presents the pooled cross-sectional OLS results to ensure that the negative association between LIFO and loan spread still holds for this reduced sample. The coefficient estimate of LIFO is −13.90 (p < .01), indicating that the negative association still holds for this sample. The second estimation in Table 6 presents the results for 2SLS. The coefficient estimate of the predicted value of LIFO (predLIFO) is −29.06 (p < .05), showing that the negative association is robust to controlling for endogeneity using 2SLS. The third estimation presents the results of the Heckman two-stage procedure. We add IMR as an additional control variable in the estimation of Equation 1. The coefficient estimate of LIFO is −12.99 (p < .01), suggesting that the negative association between loan spread and LIFO is robust to controlling for self-selection using the Heckman two-stage procedure. The last estimation uses the industry-mean LIFO frequency as our IV. Note that the sample size is substantially larger for the IV approach than for the other three models because we use the same sample as Table 3 in the IV approach. We do not include industry-fixed effects in the IV approach because using industry LIFO as our instrument precludes us from using industry-fixed effects (Larcker & Rusticus, 2010). The coefficient estimate of the predicted value of LIFO (predLIFO) is −45.37 (p < .01), which indicates that the negative association between LIFO and loan spread documented in Table 3 is robust to using an IV approach to address endogeneity. In sum, these results show that our results are robust to controlling for endogeneity and self-selection.

Results: The Pricing of LIFO for Rated and Unrated Firms (H2)

Table 7 presents the pooled cross-sectional OLS estimation results of Equation 1 for rated and unrated firms. Approximately 57% of our sample consists of firms without any public debt, indicating that unrated firms constitute the majority of our sample. We include industry-fixed effects in all estimations. The first estimation presents results for rated firms. The coefficient estimate of LIFO is −19.25 (p < .01), indicating that inventory policy choice is priced by private debtholders for unrated firms. The next estimation presents the results for rated firms. The coefficient estimates of LIFO in the second estimation is −2.68 and insignificant, suggesting that inventory policy choice is not priced by private debtholders for rated firms.¹³ The difference in the coefficient estimates of LIFO between rated and unrated firms is 16.57 (i.e., the difference is significant at p < .01). In the last estimation, we interact UNRATE and LIFO. The coefficient estimate of LIFO in the last estimation is −4.95 and insignificant, suggesting that the association between LIFO and loan spread is not significant for rated firms. The coefficient estimate of LIFO × UNRATE is –15.39 (p < .01), indicating that the association between LIFO and loan spread is 15.39 bps smaller for rated firms than unrated firms. These results show that the association between LIFO and loan spread is −19.34 bps (= −4.95−15.39) for unrated firms.¹⁴ These results suggest that there is a significant difference in the pricing of LIFO between rated and unrated firms supporting H2. Overall, the findings in Table 7 support H2, showing that the pricing effect of LIFO is stronger for unrated firms than rated firms. These findings are consistent with the notion that unrated firms face greater downside risk than rated firms.

Table 7.

The Pricing Effect of Inventory Policy on Loan Spread for Unrated and Rated Firms.

	Unrated firms		Rated firms		Full sample
	Coefficient	t stat	Coefficient	t stat	Coefficient	t stat
LIFO	−19.25	−4.64***	−2.68	−0.62	−4.95	−1.16
UNRATE					−83.75	−9.98***
LIFO × UNRATE					−15.39	−2.69***
ROA	−107.53	−5.54***	−52.99	−1.22	−109.65	−6.21***
PPE	−80.50	−7.65***	−72.78	−4.76***	−85.70	−9.81***
CURRENT	−2.65	−1.97**	−2.24	−1.11	−3.11	−2.72***
RATE			−10.92	−11.63***	−7.80	−10.60***
LMV	−13.15	−6.97***	11.83	6.11***	−2.05	−1.37
LEV	46.65	4.11***	45.44	3.01***	46.92	5.10***
BM	7.85	2.16**	35.56	5.78***	19.17	5.91***
Z_SCOR	−0.44	−0.62	−3.29	−2.50**	−1.16	−1.84*
LOSS	34.34	6.85***	22.08	3.40***	27.85	7.02***
PRF_PRC	−44.16	−15.91***	−35.51	−12.34***	−42.85	−21.73***
REVOLVE	−59.41	−17.14***	−33.37	−10.69***	−44.23	−18.70***
MATURITY	1.42	0.65	6.17	2.10**	4.96	2.70***
SECURE	75.54	23.90***	91.12	18.63***	84.45	29.98***
AMOUNT	−2.64	−1.71*	−16.95	−8.52***	−11.04	−8.03***
TERM_LOAN	−23.80	−4.96***	14.92	2.44**	−6.92	−1.80*
SENIOR	−274.67	−6.30***	−345.18	−6.15***	−299.69	−8.34***
#LENDERS	−0.98	−2.66***	−1.09	−6.71***	−1.06	−7.10***
Industry-fixed effects	Yes		Yes		Yes
R ²	.3530		.4781		.4121
N	13,982		10,349		24,331

\begin{matrix} SPREA D_{it + 1} = β_{0} + β_{1} LIF O_{it} + β_{2} UNRAT E_{it} + β_{3} LIF O_{it} \times UNRAT E_{it} + β_{4} RO A_{it} + β_{5} PP E_{it} + β_{6} CURREN T_{it} + β_{7} RAT E_{it} \\ + β_{8} LM V_{it} + β_{9} LE V_{it} + β_{10} B M_{it} + β_{11} Z_SCO R_{it} + β_{12} LOS S_{it} + β_{13} PRF_PR C_{it + 1} + β_{14} REVOLV E_{it + 1} \\ + β_{15} MATURIT Y_{it + 1} + β_{16} SECUR E_{it + 1} + β_{17} AMOUN T_{it + 1} + β_{18} TERM_LOA N_{it + 1} + β_{19} SENIO R_{it + 1} \\ + β_{20} # LENDERS + ε_{it + 1} . \end{matrix}

(1a)

Note. This table presents the pooled cross-sectional OLS estimation results of Equation 1 (presented in Table 3) for rated and unrated firms and Equation 1a. Rated (unrated) firms are those with (without) a bond rating from S&P. UNRATED is an indicator variable, which equals 1 for unrated firms and 0 otherwise. Firm-year observations are clustered by firm to eliminate autocorrelation as per Petersen (2009). The definitions of the variables are in the appendix. OLS = ordinary least squares.

, **, and *** indicate statistical significance at the 10%, 5%, and 1% levels, respectively.

As further analysis, we also separate the rated firms into investment-grade (i.e., S&P bond rating is greater than equal to BBB−) and non–investment-grade firms (i.e., S&P bond rating is less than BBB−) to investigate the association between LIFO and loan spread separately for each group. In untabulated results, we find that the coefficient estimate of LIFO in Equation 1 is not significant for either group, suggesting that LIFO is not priced for either investment-grade or non–investment-grade firms.

Finally, we investigate the pricing of LIFO/FIFO choice by equity investors separately for three groups of firms: firms with investment-grade bonds, firms with non–investment-grade bonds, and unrated firms. We regress stock returns in year t+ 1 on LIFO, size (market value of equity), and book-to-market ratio in year t separately for each group. We use raw, sizes-adjusted, and market-adjusted returns as measures of stock returns. With all three measures of stock returns, in untabulated results, we find that there is no economically significant difference in the pricing of LIFO/FIFO choice by equity investors among the three groups. As discussed above, unlike debtholders who face asymmetric payoffs, equity investors face symmetric payoffs. Therefore, equity investors appear not to price LIFO/FIFO choice differently for unrated, investment-grade, and non–investment-grade firms.

Results: The Impact of Accounting Quality and Hidden Reserves on the Pricing of LIFO (H2a and H2b)

The two possible explanations for the negative association between LIFO and loan spread are accounting quality and hidden reserves, proxied by LIFO reserves (Penman & Zhang, 2002). We investigate the joint effect of LIFO reserves and accounting quality on the negative association between LIFO and loan spread. Table 8 presents the estimation results of Equation 5. We include industry-fixed effects in all estimations and cluster firm-year observations by firm as per Petersen (2009). The accounting quality measure requires a firm to have data for all the variables in Equation 4 over the years t− 4 to t. Thus, the sample size in Table 9 is 16,759, smaller than the 24,331 in Table 3.

Table 8.

The Impact of AQ and LIFORSV on the Pricing Effect of LIFO on Loan Spread.

	(1)		(2)		(3)		(4)
	Coefficient	t stat	Coefficient	t stat	Coefficient	t stat	Coefficient	t stat
LIFO	−12.80	−3.58***	−4.64	−1.08	−11.82	−3.30***	−3.90	−0.90
LIFORSV			−307.84	−3.51***			−299.48	−3.42***
AQ					195.50	4.39***	192.27	4.33***
ROA	−108.77	−4.61***	−108.42	−4.60***	−100.84	−4.28***	−100.63	−4.27***
PPE	−105.11	−10.26***	−104.51	−10.23***	−96.59	−9.25***	−96.15	−9.24***
CURRENT	−1.37	−0.92	−1.26	−0.85	−0.75	−0.50	−0.66	−0.44
RATE	−1.31	−3.69***	−1.30	−3.76***	−1.31	−3.69***	−1.30	−3.76***
LMV	−0.03	−0.02	−0.28	−0.16	1.23	0.68	0.97	0.54
LEV	79.82	7.00***	76.99	6.73***	82.29	7.19***	79.50	6.93***
BM	25.21	6.54***	25.07	6.51***	27.62	7.02***	27.45	6.99***
Z_SCOR	−1.58	−1.81*	−1.56	−1.80*	−1.77	−2.05**	−1.75	−2.04**
LOSS	30.66	6.13***	31.22	6.27***	30.10	6.01***	30.65	6.16***
PRF_PRC	−38.98	−16.24***	−39.20	−16.34***	−38.64	−16.16***	−38.85	−16.26***
REVOLVE	−41.16	−14.18***	−41.00	−14.20***	−41.57	−14.31***	−41.40	−14.32***
MATURITY	9.73	4.30***	9.81	4.36***	10.18	4.48***	10.26	4.54***
SECURE	91.18	28.15***	90.72	28.13***	90.11	27.70***	89.68	27.69***
AMOUNT	−12.38	−7.54***	−12.29	−7.58***	−12.19	−7.42***	−12.10	−7.45***
TERM_LOAN	−3.48	−0.73	−3.49	−0.74	−3.79	−0.80	−3.80	−0.80
SENIOR	−393.13	−8.47***	−393.36	−8.44***	−395.05	−8.50***	−395.23	−8.48***
#LENDERS	−0.93	−5.35***	−0.95	−5.52***	−0.97	−5.56***	−0.99	−5.72***
Industry-fixed effects	Yes		Yes		Yes		Yes
R ²	.4179		.4190		.4197		.4207
N	16,759		16,759		16,759		16,759

\begin{matrix} SPREA D_{it + 1} = & β_{0} + β_{1} LIF O_{it} + β_{2} A Q_{it} + β_{3} LIFORS V_{it} + β_{4} RO A_{it} + β_{5} PP E_{it} + β_{6} CURREN T_{it} + β_{7} RAT E_{it} + β_{8} LM V_{it} \\ + β_{9} LE V_{it} + β_{10} B M_{it} + β_{11} Z_SCO R_{it} + β_{12} LOS S_{it} + β_{13} PRF_PR C_{it + 1} + β_{14} REVOLV E_{it + 1} \\ + β_{15} MATURIT Y_{it + 1} + β_{16} SECUR E_{it + 1} + β_{17} AMOUN T_{it + 1} + β_{18} TERM_LOA N_{it + 1} \\ + β_{19} SENIO R_{it + 1} + β_{20} # LENDERS + ε_{it + 1} . \end{matrix}

(5)

Note. This table presents the pooled cross-sectional OLS estimation results of Equation 5. AQ is accruals quality calculated as standard deviation of residuals over 5 years from t to t− 4 from the estimation of Equation 4. The definitions of other variables are in the appendix. Industry-fixed effects are based on Fama and French (1997) 48 industry definitions. Firm-year observations are clustered by firm to eliminate autocorrelation as per Petersen (2009). LIFO = last-in-first-out; OLS = ordinary least squares.

, **, *** indicate statistical significance at the 10%, 5%, and 1% levels, respectively.

Table 9.

Future Earnings Growth and Inventory Policy.

	Coefficient	t stat	Coefficient	t stat
LIFO	0.0065	6.09***	0.0020	1.71*
LIFORSV			0.1236	6.26***
ROA	−0.2260	−24.05***	−0.2262	−24.07***
PPE	0.0410	9.54***	0.0407	9.49***
CURRENT	−0.0010	−2.22***	−0.0010	−2.23**
RATE	−0.0008	−7.05***	−0.0008	−6.95***
LMV	0.0070	16.98***	0.0070	17.04***
LEV	−0.0306	−6.99***	−0.0297	−6.74***
BM	0.0049	4.11***	0.0049	4.11***
Z_SCOR	−0.0024	−9.22***	−0.0024	−9.21***
LOSS	−0.0109	−4.68***	−0.0109	−4.67***
Industry-fixed effects	Yes		Yes
R ²	.0895		.0896
N	75,746		75,746

\begin{matrix} EARNGRO W_{it} = β_{0} + β_{1} LIF O_{it} + β_{2} LIFORS V_{it} + β_{3} RO A_{it} + β_{4} PP E_{it} + β_{5} CURREN T_{it} + β_{6} RAT E_{it} + β_{7} LM V_{it} \\ + β_{8} LE V_{it} + β_{9} B M_{it} + β_{10} Z_SCO R_{it} + β_{11} LOS S_{it} + ε_{it + 1} . \end{matrix}

(6)

Note. This table presents the pooled cross-sectional OLS estimation results of Equation 6. The definitions of the variables are in the appendix. Industry-fixed effects are based on Fama and French (1997) 48 industry definitions. Firm-year observations are clustered by firm to eliminate autocorrelation as per Petersen (2009). OLS = ordinary least squares.

, **, and *** indicate statistical significance at the 10%, 5%, and 1% levels, respectively.

The first estimation in Table 8 reports the results of Equation 1 without including either AQ or LIFO reserves (LIFORSV). We first estimate this model to see whether we find a significant coefficient on LIFO for this reduced sample. The coefficient estimate of LIFO is −12.80 (p < .05), indicating that there is a negative association between loan spread and LIFO even with this smaller sample. The second estimation presents the results when we include only LIFORSV. As expected, the coefficient estimate of LIFO is −4.64 and insignificant. The inclusion of LIFORSV, therefore, reduces the coefficient estimate of LIFO by 8.16 bps (a reduction of 64% in the coefficient estimate of LIFO), suggesting that LIFORSV explains, to a large extent, the negative association between loan spread and LIFO, supporting H3b. The third estimation presents the result when we include only AQ in the estimation. The coefficient estimate of LIFO is −11.82 (p < .01), indicating that the inclusion of AQ does not have much effect on the coefficient estimate of LIFO. That is, the inclusion of AQ in this estimation reduces the coefficient estimate of LIFO only by 0.98 bps (= 12.80 – 11.82), a reduction of only 8%. This result implies that AQ does not explain the negative association between LIFO and loan spread, rejecting H3a. The coefficient estimate of AQ itself is 195.50 (p < .01), suggesting that AQ itself is priced by private debtholders, consistent with Bharath et al. (2008). The fourth estimation presents the results when we include both LIFORSV and AQ. The coefficient estimate of LIFO is −3.90 and insignificant. That is, the coefficient estimate of LIFO when we include both LIFORSV and AQ is reduced by only 0.74 bps (= 4.64 − 3.90) compared with the estimate in which only LIFORSV is included. These results show that the inclusion of AQ does not have much effect, but the inclusion of LIFORSV makes the coefficient estimate of LIFO insignificant. Finally, the coefficient estimates of LIFORSV and AQ are both significant (p < .01) in the last estimation, suggesting that LIFO reserves and accounting quality are two distinct phenomena priced by private lenders. Inclusion of one does not subsume the pricing effect of the other. Overall, the results in Table 8 show that H3a is rejected, whereas H3b is supported, indicating that it is the future earnings implications of LIFO reserves that drive the negative association between LIFO and loan spread, rather than accounting quality.

The coefficient on LIFORSV is −299.48 in the last estimation of Table 8. Because the mean value of LIFORSV for LIFO firms is 0.0287, the mean effect of LIFORSV on loan spread is, therefore, 8.60 bps (299.48 × 0.0287). The mean value of AQ for LIFO firms is 0.0367. The coefficient estimate of AQ in the last estimation of Table 8 is 192.27. This result suggests that the mean effect of AQ on loan spread for LIFO firms is 7.06 bps (= 192.27 × 0.0367).

As shown in Table 2, there is a high correlation between LIFO and LIFORSV. This raises the possibility that multicollinearity might affect our results presented in Table 8. To assess the severity of multicollinearity, we calculate variance inflation factor (VIF) for coefficient estimates of LIFO and LIFORSV. We find that the VIFs for LIFO and LIFORSV are 1.90 and 1.66, respectively. The cutoff point for severe multicollinearity is 10 (Hair, Anderson, Tatham, & Black, 1995; Kennedy, 1992; Marquardt, 1970; Neter, Wasserman, & Kutner, 1989). Therefore, multicollinearity is not a serious concern for the results presented in Table 8.

Additional Analysis: Future Earnings Implications of LIFO Reserves

Finally, we investigate the question of the extent to which LIFO reserves are associated with future earnings. By understating aggregate (retained) earnings and book value of net assets, LIFO creates reserves that are not on the balance sheet: that is, hidden reserves. When the reserves are released sometime in the future, earnings will increase. Penman and Zhang (2002) document a positive association between Q score and 1-year-ahead earnings. Although Q score also depends on reserves created by other reserves (R&D and advertising), ceteris paribus, Q score would be higher for firms with larger LIFO reserves. Based on the evidence in Penman and Zhang (2002), we investigate two questions: (a) whether LIFO is associated with higher future earnings growth and (b) to what extent, if any, LIFO reserves explain the association between LIFO and future earnings growth. To explore these questions, we estimate the following equation:

\begin{matrix} EARNGRO W_{it} = β_{0} + β_{1} LIF O_{it} + β_{2} LIFORS V_{it} + β_{3} RO A_{it} + β_{4} PP E_{it} + β_{5} CURREN T_{it} \\ + β_{6} RAT E_{it} + β_{7} LM V_{it} + β_{8} LE V_{it} + β_{9} B M_{it} + β_{10} Z_SCO R_{it} \\ + β_{11} LOS S_{it} + ε_{it + 1} . \end{matrix}

(6)

The definitions of the variables in Equation 6 are presented in the appendix. We estimate pooled cross-sectional OLS regressions using all Compustat firms with data available to estimate Equation 6. We do not require firm-year observations to have data from DealScan as in Table 3 because the analysis here is not related to loan spread. As a result, we have a much larger sample in this analysis compared with that in Table 3. We include industry-fixed effects in all estimations and cluster firm-year observations by firm. The estimation results of Equation 6 are presented in Table 9. The first estimation presents the results when we include only LIFO as an indicator variable. The coefficient estimate of LIFO is 0.0065 (p < .01), suggesting that LIFO firms have higher future earnings growth than FIFO firms. The second estimation presents the results when we include both LIFO and LIFORSV. The coefficient estimate of LIFORSV is 0.1236 and significant (p < .01), indicating that LIFO reserves are associated with higher future earnings growth. However, the coefficient estimate of LIFO is 0.0020 and only marginally significant. The positive association between LIFO and future earnings growth is reduced by 70% when we include LIFORSV in the estimation. Thus, LIFO reserves, to most extent, explain the positive association between future earnings growth and LIFO.

Conclusion

We investigate the relationship between inventory policy choice and the cost of debt, proxied by loan spreads on private debt. Specifically, we investigate three research questions: (a) Do private debtholders price LIFO inventory policy differently from FIFO inventory policy? (b) If so, then how does downside risk affect the pricing of LIFO by private debtholders? and (c) What is the underlying reason for the pricing difference between FIFO and LIFO firms: accounting quality or hidden reserves?

As an answer to the first research question, we find that private debtholders, on average, demand greater loan spreads from FIFO than from LIFO firms. We also find that the impact of LIFO on loan spread is greater in high-inflation years than in low-inflation years. In answer to the second research question, we find that the pricing effect of LIFO on loan spread is 7 times greater for unrated firms than rated firms, suggesting that the impact of LIFO on private debtholders is much greater for firms with higher downside risk. Regarding the third research question, we find that AQ has little impact on the pricing effect of LIFO on loan spread. LIFO reserves, however, account for the negative association between loan spread and LIFO. Our findings suggest that it is the hidden LIFO reserves, rather than the accounting quality, that explain the pricing effect of LIFO on loan spread. Further analysis reveals that LIFO is positively associated with future earnings growth. However, once we control for LIFO reserves, the positive association between LIFO and future earnings growth vanishes.

The findings in this article contribute to several lines of research. First, we contribute to extant research on the valuation of inventory policy choice. Although prior research examined the perspective of equity investors (Krishnan et al., 2008; Lev & Thiagarajan, 1993), it has overlooked the perspective of debtholders. By focusing on private debtholders, we fill the gap in the extant literature. Second, our results contribute to the extant research that investigates the implications of asymmetric payoffs of debtholders. We find that, due to asymmetric payoffs, the impact of LIFO on private debtholders is greater for firms with higher downside risk (i.e., unrated firms) than lower downside risk (i.e., rated firms). Third, our findings contribute to the literature on accounting conservatism. We document that unconditional conservatism (i.e., LIFO) is favorably priced by private debtholders as by equity investors. Finally, our article contributes to the extant literature on the pricing of hidden LIFO reserves. Prior research documents that equity investors fail to correctly price LIFO reserves (Penman & Zhang, 2002). By contrast, we document that private debtholders positively price LIFO reserves.

Footnotes

Appendix

Authors’ Note

Mustafa Ciftci is now affiliated with University of North Carolina at Charlotte, USA.

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

Notes

References

Abdel-Khalik

A. R.

McKeown

J. C.

(1978). Understanding accounting changes in an efficient market: Evidence of differential reaction. The Accounting Review, 53, 851-868.

Armstrong

C. S.

Guay

W. R.

Weber

J. P.

(2010). The role of information and financial reporting in corporate governance and debt contracting. Journal of Accounting & Economics, 50, 127-179.

Asquith

Beatty

Weber

(2005). Performance pricing in bank debt contracts. Journal of Accounting & Economics, 40, 101-128.

Basu

Naughton

J. P.

(2018). The real effects of financial statement recognition: Evidence from corporate credit ratings. Available at SSRN 2980510.

Beatty

Weber

J. J.

(2008). Conservatism and debt. Journal of Accounting & Economics, 45, 154-174.

Bharath

S. T.

Sunder

(2008). Accounting quality and debt contracting. The Accounting Review, 83, 1-28.

Brown

R. M.

(1980). Short-range market reaction to changes to LIFO accounting using preliminary earnings announcement dates. Journal of Accounting Research, 18, 38-63.

Chava

Roberts

M. R.

(2008). How does financing impact investment? The role of debt covenants. The Journal of Finance, 63, 2085-2121.

Cushing

B. E.

LeClere

M. J.

(1992). Evidence on the determinants of inventory accounting policy choice. The Accounting Review, 67, 355-366.

10.

Defond

M. L.

Zhang

(2014). The timeliness of the bond market reaction to bad earnings news. Contemporary Accounting Research, 31, 911-936.

11.

Diamond

D. W.

(1991). Monitoring and reputation: The choice between bank loans and directly placed debt. Journal of Political Economy, 99, 689-721.

12.

Easton

P. D.

Monahan

S. J.

Vasvari

F. P.

(2009). Initial evidence on the role of accounting earnings in the bond market. Journal of Accounting Research, 47, 721-767.

13.

Elliott

J. A.

Ghosh

Moon

(2010). Asymmetric valuation of sustained growth by bond- and equity-holders. Review of Accounting Studies, 15, 833-878.

14.

Even-Tov

(2017). When does the bond price reaction to earnings announcements predict future stock returns? Journal of Accounting & Economics, 64, 167-182.

15.

Fama

E. F.

French

K. R.

(1992). The cross-section of expected stock returns. The Journal of Finance, 47, 427-465.

16.

Fama

E. F.

French

K. R.

(1997). Industry costs of equity. Journal of Financial Economics, 43, 153-193.

17.

Flannery

M. J.

(1986). Asymmetric information and risky debt maturity choice. The Journal of Finance, 41, 19-37.

18.

Francis

LaFond

Olsson

Schipper

(2005). The market pricing of accruals quality. Journal of Accounting & Economics, 39, 295-327.

19.

Goh

B. W.

Lim

C. Y.

Lobo

G. J.

Tong

Y. H.

(2017). Conditional conservatism and debt versus equity financing. Contemporary Accounting Research, 34, 216-251.

20.

Hair

J. F., Jr.

Anderson

R. E.

Tatham

R. L.

Black

W. C.

(1995). Multivariate data analysis. New York, NY: Macmillan.

21.

Heckman

(1979). Sample selection bias as a specification error. Econometrica, 47, 153-161.

22.

Holthausen

R. W.

Watts

(2001). The relevance of the value relevance literature for financial accounting standard setting. Journal of Accounting & Economics, 31, 3-75.

23.

James

(1987). Some evidence on the uniqueness of bank loans. Journal of Financial Economics, 19, 217-235.

24.

Jensen

M. C.

Meckling

W. H.

(1976). Theory of the firm: Managerial behavior, agency costs, and ownership structure. Journal of Financial Economics, 3, 305-360.

25.

Kennedy

(1992). A guide to econometrics. Oxford, UK: Blackwell.

26.

Kraft

(2015). Rating agency adjustments to GAAP financial statements and their effect on ratings and credit spreads. The Accounting Review, 90, 641-674.

27.

Krishnan

G. V.

Srindhi

B. S.

(2008). Inventory policy, accruals quality and information risk. Review of Accounting Studies, 13, 369-410.

28.

Larcker

D. F.

Rusticus

T. O.

(2010). On the use of instrumental variables in accounting research. Journal of Accounting & Economics, 49, 186-205.

29.

Lev

Sougiannis

(1996). The capitalization, amortization and value relevance of R&D. Journal of Accounting & Economics, 21, 107-138.

30.

Lev

Thiagarajan

(1993). Fundamental information analysis. Journal of Accounting Research, 31, 190-216.

31.

Marquardt

D. W.

(1970). Generalized inverses, ridge regression, biased linear estimation, and nonlinear estimation. Technometrics, 12, 591-256.

32.

Myers

(1977). Determinants of corporate borrowing. Journal of Financial Economics, 5, 147-175.

33.

Neter

Wasserman

Kutner

M. H.

(1989). Applied linear regression models. Homewood, IL: Irwin.

34.

Penman

Zhang

H. X.

(2002). Accounting conservatism, the quality of earnings and stock returns. The Accounting Review, 77, 237-264.

35.

Petersen

M. A.

(2009). Estimating standard errors in finance panel data sets: Comparing approaches. Review of Financial Studies, 22, 435-480.

36.

Petersen

M. A.

Rajan

(1994). The benefits of lending relationships: Evidence from small business data. The Journal of Finance, 49, 3-37.

37.

Ruch

G. W.

Taylor

(2015). Accounting conservatism: A review of the literature. Journal of Accounting Literature, 34, 17-38.

38.

Sharpe

S. A.

(1990). Asymmetric information, bank lending, and implicit contracts: A stylized model of customer relationships. The Journal of Finance, 45, 1069-1087.

39.

Sunder

(1973). Relationship between accounting changes and stock prices. Journal of Accounting Research, 11, 1-45.

40.

Sunder

(1975). Stock price and risk related to accounting changes in inventory valuation. The Accounting Review, 50, 305-315.

41.

Xue

(2007). Make or buy new technology: The role of CEO compensation contract in a firm’s route to innovation. Review of Accounting Studies, 12, 659-690.

42.

Zhang

(2008). The contracting benefits of accounting conservatism to lenders and borrowers. Journal of Accounting & Economics, 45, 27-54.