Book–Tax Differences and Earnings Persistence: The Moderating Role of Sales Decline

Abstract

This study investigates why firms with large book–tax differences (BTDs) exhibit lower earnings persistence, particularly during periods of revenue declines. While prior literature has linked BTDs, especially large positive BTDs (LPBTDs), to earnings management, we propose an alternative explanation rooted in operational disruptions. Using a large panel of U.S. firms from 1995 to 2016, we examine whether short-term earnings persistence is affected by sales trends and the direction of BTDs. Our findings reveal that both large positive and large negative BTDs are significantly associated with reduced earnings persistence when sales decline. The effect is pronounced in both accrual and cash flow components of earnings. We develop and test a framework based on “operations theory,” which attributes this reduction to real business shocks, such as asset write-downs, facility closures, and reserve adjustments, that arise during sales decline periods. These results highlight the importance of distinguishing operationally driven BTDs from those arising through discretionary accruals. Our findings have implications for investors, regulators, and researchers seeking to interpret BTDs more accurately in volatile economic environments.

1. Introduction

The reasons why significantly large positive and negative book–tax differences (BTDs) signal the short-term changes in the persistence levels of accounting earnings remain underexplored in the accounting research world. On the other hand, few studies have explored how operational changes (e.g., restructuring, write-downs, and cash flow shocks) triggered by declines in revenue may affect both accruals and cash flow components of earnings in the context of large BTDs.

Prior studies have primarily examined accruals-based earnings management to account for the reduced persistence of next-year earnings linked to significant book–tax differences (Hanlon, 2005; Blaylock et al., 2012; Hanlon & Heitzman, 2010). However, a recent paper by Oler and Coyne (2024) has documented how sales volatility impacts the persistence of operating cash flows, especially in firms flagged for manipulation. It shows that cash flow predictability deteriorates during episodes of irregular sales performance. Drawing on insights from the management accounting literature, we aim to bridge these research gaps by using sales decreases as a moderator variable that triggers operational changes impacting cash flows and accruals, ultimately affecting earnings persistence. In doing so, we respond to calls in the literature for a broader, economically grounded interpretation of BTDs (Hanlon & Heitzman, 2010; Jackson, 2015; Linsmeier & Wheeler, 2021).

BTDs potentially reflect the discrepancy between income reported for financial accounting purposes and income calculated for tax reporting. Most of the literature on BTDs focuses on temporary or timing differences—that is, differences in the period when specific items are recognized for book purposes versus when they are recognized for tax purposes (Blackburne & Blouin, 2016). In the United States, the tax code recognizes income and expenses under different principles than financial reporting. Under U.S. tax regulations, revenue is typically recognized when earned and expenses are deducted when paid, subject to specific rules, such as capitalization requirements (IRC §263A) and depreciation schedules (MACRS under IRC §168). During periods of declining sales, firms may reduce capital expenditures or dispose of assets, both of which alter the timing and magnitude of tax deductions. Additionally, tax law allows for the acceleration of deductions through mechanisms like bonus depreciation or inventory write-downs (IRC §167 and §471), which can create or amplify temporary BTDs. These provisions interact with financial reporting standards (e.g., GAAP) in ways that affect earnings persistence: while GAAP may spread costs over time, tax rules may allow immediate recognition. Consequently, operational responses to sales declines—such as restructuring, asset impairments, and cost reductions—can create large temporary BTDs that reduce the stability of earnings streams. Understanding how these tax treatments apply in downturns provides critical context for interpreting BTDs and their relation to short-term earnings persistence.

Adopting a financial reporting paradigm, the tax literature suggests that BTDs are informative about earnings management because managers have greater flexibility to adjust book income than tax income (Chen et al., 2012; Wahab & Holland, 2015). Moreover, managers typically have incentives to manage book earnings up and taxable income down. The EM theory predicts that earnings persistence will be lower when the BTD is high (large positive BTD) because large differences indicate income-increasing accruals that must reverse in the near future. The EM explanation for lower earnings persistence is supported by Hanlon (2005), who documents that a significant part of the reduction in earnings persistence associated with LPBTDs and LNBTDs comes through accruals, and is further supported by Blaylock et al. (2012), who discriminate between EM and tax avoidance as sources of LPBTDs.

When firms experience declining sales, managers must identify the underlying causes and respond by either reversing the decline or reducing costs to maintain profitability at lower levels of activity. The management accounting literature documents a range of strategic responses to sales downturns (Anderson et al., 2003; Banker & Chen, 2006; Banker et al., 2014, 2017). Common actions include closing underperforming factories or retail locations, reducing workforce size, modifying product offerings to address weaknesses, and reorganizing operations to enhance efficiency. Even when a decline in sales is driven by external economic or industry-wide factors, it often triggers firm-level operational adjustments aimed at shedding unproductive segments and improving overall performance. Consequently, we predict that such operational changes induced by declining demand will lead to reduced earnings persistence. This effect is expected to hold across both directions of BTDs.

The tax–earnings research has primarily focused on accruals-based earnings management (EM) since BTDs capture timing variations in recognizing items for financial versus tax purposes. Consequently, the research has aimed to connect managerial decisions influencing BTDs, such as earnings manipulation, to the persistence of earnings. As EM predictions pertain mostly to financial reporting rather than operational activities, they are more closely tied to accruals than to cash flows, highlighting a research gap. The EM framework does not extend to real earnings management (REM) (Roychowdhury, 2006), as REM impacts both book and tax income.

An alternative approach considers whether factors influencing BTDs also interact with other drivers of earnings persistence (Drake, 2012; Hanlon & Heitzman, 2010). Drake (2012) links BTDs to life-cycle stages, arguing that capital expenditures and asset disposals—which create timing differences in depreciation—vary systematically across these stages. Depreciation is an important component of BTDs that is likely to influence the sorting of firms into LPBTDs and LNBTDs (Poterba et al., 2011). Growth firms (high capex) are more likely to have LPBTDs and decline firms (low capital expenditures) are more likely to have LNBTDs than other firms. Growth firms—particularly in high-tech industries—face intense competition and rapid technological change that can fundamentally reshape their markets. Therefore, a reduction in sales for growth firms is likely to cause greater disruption and repositioning than for more mature firms. Firms that are past maturity are also sensitive to changing market conditions. Such firms are likely to discontinue some operations as specific markets deteriorate. Thus, the combination of SD and large BTDs in either direction may be informative about changes in operations that affect short-term earnings persistence.

To test the predictions of the competing theories, we analyze four scenarios: large positive book–tax differences with sales-down (SD), LPBTDs with sales-up (SU), LNBTDs with sales decline, and LNBTDs with sales increase. The earnings management theory posits a negative relationship between LPBTDs and earnings persistence in both sales-decline and sales-increase scenarios, as earnings management may occur regardless of sales trends. It suggests earnings management might be more pronounced in sales-decline cases, though this lacks empirical confirmation. OT, however, predicts a negative relationship between both LPBTDs and LNBTDs and earnings persistence in sales-decline scenarios. In summary, (1) both theories expect lower earnings persistence with LPBTDs during revenue declines scenarios, (2) only the EM theory anticipates lower persistence with LPBTDs during revenue increase cases, (3) only OT foresees lower persistence with LNBTDs during sales declines, and (4) neither theory predicts lower persistence with LNBTDs during sales increases.

We estimate short-term persistence models, consistent with prior tax and earnings management studies (Hanlon, 2005; Blaylock et al., 2012), and observe: (1) a significant decline in earnings persistence linked to LPBTDs during sales downturns, (2) a minor decrease in earnings persistence tied to LPBTDs during sales upturns, (3) a substantial reduction in earnings persistence associated with LNBTDs when sales are declining, and (4) a slight reduction in earnings persistence linked to LNBTDs when sales are increasing. Finding (1) supports both theories, finding (2) provides weak support for the tax theory, finding (3) strongly supports operations theory (OT), and finding (4) does not support either theory. Overall, our analysis of pre-tax book income (PTBI) persistence suggests OT is more dominant than the EM theory.

Hanlon (2005) separates the effects of BTDs on the persistence of PTBI, showing that both LPBTDs and LNBTDs significantly reduce the persistence of cash flows and accruals. While the earnings management theory focuses on accruals, operations theory suggests that revenue drops lead to operational changes that disrupt the entire earnings stream, affecting both cash flows and accruals. In particular, managerial responses to anticipated declines in customer demand often include cuts to capital expenditures, which in turn reduce the persistence of depreciation-related accruals due to the absence of new assets replacing old ones. Demand-driven asset write-downs—especially for newer assets with high book values—further weaken accrual stability. Similarly, impairments of intangible assets and other restructuring-related charges impact both accruals and cash flows. Asset disposals, common during such adjustments, affect both components as well, and may also alter BTDs. For example, write-downs typically reduce book income but not taxable income, thereby narrowing BTDs, while asset disposals can create mismatches between book and tax treatments, increasing BTDs. These disruptions often contribute to the classification of firms as LNBTDs. For a detailed breakdown of how these operational changes relate to BTDs, see the summary in

Table A3. Summary table of special disruption items.

Special Item	BTD Type	Reason	Examples
Write-downs	Negative BTDs	Expensed for book purposes but deferred for tax purposes.	Inventory write-downs, goodwill impairments, and intangible asset write-offs.
Restructuring Costs	Negative/Positive	Timing differences between book and tax recognition.	Severance payments (negative) and prepayments for future expenses (positive).
Asset Disposals	Positive/Negative	Gains/losses are recognized differently under tax and book rules.	Gains on machinery sales (positive) and losses disallowed for tax purposes (negative).
Legal Settlements, Fines and penalties	Negative BTDs	Non-deductible for tax purposes but expensed for book purposes.	Penalties for regulatory violations and litigation settlements.
Non-Recurring Items	Variable	It depends on the nature of the item and tax vs. book treatment.	One-time insurance pay (positive) and non-deductible sponsorship expenses (negative).

in Appendix A.

Following Hanlon’s approach, we decompose the persistence of pre-tax book income into its cash flow and accrual components. The interaction effects between sales decline and both LPBTDs and LNBTDs yield particularly insightful results. When sales increase, reductions in PTBI persistence are small to moderate. However, when sales decline, the reductions—especially in cash flow persistence—are substantially larger for both LPBTD and LNBTD firms. Notably, accrual persistence remains unaffected under sales growth but drops significantly when sales fall, highlighting the disruptive impact of operational stress on both earnings components.

While existing literature largely attributes book–tax differences (BTDs) to earnings management, this paper introduces operations theory as an alternative explanation. Operations theory argues that BTDs can also result from legitimate operational responses to declining sales—such as facility closures, workforce reductions, product line changes, or asset disposals—which disrupt normal business activities and introduce volatility into both accruals and cash flows. These disruptions, in turn, reduce the predictability and persistence of short-term earnings. Examples include non-recurring accruals from asset impairments, severance costs, inventory write-downs, and unpredictable gains or losses from asset sales.

Unlike earnings management, which implies discretionary manipulation, operations theory emphasizes real economic adjustments. This distinction is important: misinterpreting operationally driven BTDs as evidence of manipulation can lead to misguided regulatory, investment, or policy decisions (see

Table A2. Comparison table—earnings management vs. operations theory.

Aspect	EM Theory	Operations Theory
Argument	BTDs result from opportunistic reporting, where managers manipulate earnings for their objectives.	BTDs arise due to operational decisions, such as changes in business activities, strategies, or market conditions.
Positive BTDs?	YES	YES
Negative BTDs?	NO	YES
Cause of BTDs	Intentional manipulation of accounting or tax items.	Natural variation due to operational decisions.
Objective	Meet earnings targets or reduce tax liability.	Respond to market conditions, strategy shifts.
Implications	Signals opportunism or risk of future restatements.	Reflects legitimate business decisions.
Examples	Use of accruals, income smoothing.	Sales declines, inventory buildup.

). Moreover, recent research (e.g., Floropoulos et al., 2024) questions whether higher book–tax conformity necessarily curbs earnings management, further supporting the need for a broader framework like operations theory.

The primary objective of this study is to examine why large BTDs are associated with lower short-term earnings persistence, particularly during periods of sales decline. The paper aims to extend the literature by introducing an alternative explanation, termed operations theory, which attributes changes in earnings persistence to real operational disruptions rather than solely to earnings management via accruals. By integrating concepts from tax accounting and management accounting, the study investigates whether sales declines amplify the destabilizing effects of BTDs on both accrual and cash flow components of pre-tax book income. This approach allows us to evaluate whether operational adjustments (e.g., cost restructuring, asset write-downs, and reserve changes) offer additional explanatory power beyond traditional earnings management frameworks. We document that the operations theory for reduced persistence due to changes in operations resulting from declines in sales is incrementally and differentially informative relative to the EM theory that is more closely aligned with the financial reporting literature. Operations theory explains the persistent effects of both LPBTDs and LNBTDs during sales declines, whereas the EM theory is limited in scope to accrual-based manipulations.

Drawing on alternative perspectives on large BTDs, and building on the author’s Ph.D. thesis (Rahiminejad, 2022) we provide empirical evidence that both large positive and negative BTDs are significantly more associated with reduced earnings persistence during periods of sales decline, driven by both accrual and cash flow effects. This has practical implications for investors, regulators, and standard-setters, who risk misinterpreting operationally driven BTDs as opportunistic behavior. The study findings underscore the importance of operational changes in understanding the persistence of earnings (please see Appendix A, Table A2). The paper is organized as follows. Three avenues of literature relevant to this study, BTD, earnings management, and operations theory, are reviewed in Section 2. Methodology and materials and hypotheses development are stated in Section 3. The data and empirical models are also described in Section 3. Empirical results of estimating the models are presented and discussed in Section 4. Findings and conclusions and contributions and implications are described at the end of the article, in Section 5 and Section 6.

2. Previous Literature

2.1. Book–Tax Differences

Understanding the relationship between current and future earnings—and how future earnings are priced—remains a central question in financial accounting. The limitations of current financial reports in reliably forecasting income sustainability have led investors and policymakers to seek alternative indicators of earnings quality (Alves, 2023). A growing body of research has highlighted the widening gap between financial accounting income and taxable income since the 1990s (Desai, 2003; Manzon & Plesko, 2001). The rationale for using BTDs as a signal lies in the notion that taxable income is subject to more rigid rules and offers less managerial discretion than financial reporting. While GAAP allows managers considerable flexibility in selecting accounting methods and making estimates (Liu et al., 2014; Black et al., 2017), tax reporting is generally more constrained, making BTDs a potentially informative measure of earnings quality.

It is important to distinguish between permanent and temporary book–tax differences. Permanent differences affect the calculation of income for either tax or financial reporting purposes, but never both. In contrast, temporary differences arise due to timing discrepancies between the two reporting systems. These differences stem from mismatches between book-basis and tax-basis balance sheet accounts. Under financial reporting standards, revenue is recognized when earned, and expenses are matched to revenue or recorded in the period incurred. For tax purposes, however, income is often recognized when cash is received, and expenses are typically deducted when paid. Poterba et al. (2011) provide a detailed breakdown of BTD components, identifying temporary differences—particularly those related to depreciation—as the primary drivers of deferred tax liabilities.

A line of research suggests that BTDs reveal the degree to which companies lower taxable income to minimize tax liabilities (Mills & Newberry, 2001). This research explores the link between significant and abnormal BTDs and audit fees, proposing BTDs act as an indicator of tax risk, thereby heightening the chance of IRS audits (Hanlon et al., 2012). The results show that substantially large BTDs signal potential tax avoidance, drawing increased attention from tax authorities. Desai (2003) and Desai and Dharmapala (2006) contend that larger BTDs reflect aggressive tax avoidance, such as sheltering. Rahiminejad (2025) demonstrates that firms with large positive BTDs have a higher probability of financial distress or bankruptcy, pointing to real operational inefficiencies underlying BTDs. A meta-analysis study highlights the value of precise data in understanding opportunistic behaviors and supports the role of book–tax conformity (BTC) in reducing such practices. It documents the BTC approach as being effective in mitigating aggressive reporting, supporting calls for tighter conformity regulations (Evers et al., 2016).

Wahab and Holland (2015) drive attention to understanding of the sources of BTDs and their properties. Another research stream documents associations between BTDs and future performance of firms. Lev and Nissim (2004) and Hanlon (2005) study the information properties of BTDs with respect to the predictability, persistence, growth, and quality of pre-tax book income. Lev and Nissim (2004) use the tax to book ratio, their logic being that construction of the comprehensive tax fundamental reflects all three components, pretax discretionary accruals, discretionary tax accruals, and nondeductible pretax accruals. Hanlon (2005) investigates earnings, accruals, and cash flow persistence for firms with large positive and negative BTDs. Her results suggest that the presence of large BTDs, regardless of their direction, decreases persistence. She argues that the reduction in persistence caused by the presence of large positive book–tax differences is consistent with positive book–tax differences indicating lower earnings quality. A more recent study by Jackson (2015) finds the temporary component of total BTD to better future changes in earnings, and the permanent component to predict future changes in the tax expense (Jackson, 2015).

Blaylock et al. (2012) examine how large positive BTDs reduce earnings persistence and conclude that BTDs driven by earnings management, rather than tax avoidance, are primarily responsible for this decline. Their findings account for the effects of GAAP changes, macroeconomic conditions, and discretionary reporting practices. In contrast, other studies propose alternative explanations. For instance, Wilson (2009) argues that large positive temporary BTDs may reflect aggressive tax planning strategies. However, such aggressive tax positions are often unsustainable over time, which aligns with the observed decline in earnings persistence documented in the broader literature (Frank et al., 2009).

Guenther (2011), using data techniques, identifies a small subset of observations that he claims account for Hanlon’s (2005) findings. He argues that the observed relationship between earnings persistence and large book–tax differences (BTDs) is primarily driven by firms that are young, small in size, have high returns on assets (ROAs), and exhibit large transitory items. Building on this, Drake (2012) employs a life-cycle measure developed by Dickinson (2011) to classify firm-year observations into distinct life-cycle stages. He suggests that the relationship between book and tax income provides useful information about a firm’s life-cycle position, enhancing the informativeness of tax disclosures. Tang and Firth (2012) further demonstrate that abnormal negative and positive book–tax spreads contain incremental explanatory power for earnings persistence, beyond what is captured by traditional earnings management measures, such as discretionary and total accruals. Their findings support the notion that BTDs contribute additional insight under the information content hypothesis and warrant further investigation.

Overall, prior literature largely supports the earnings perspective, suggesting that large positives (LPBTDs) reflect higher discretionary accruals and that earnings persistence declines due to the short-term reversal of those accruals. However, there is comparatively limited evidence explaining why reduced earnings persistence is also observed in the presence of large negative BTDs (LNBTDs), or why the cash flow components of earnings show lower persistence for both LPBTD and LNBTD firms. While studies such as Guenther (2011) and Drake (2012) offer partial insights, the field lacks a unifying theoretical framework to integrate these findings. This study addresses that gap by emphasizing the importance of distinguishing between BTDs arising from earnings management and those stemming from operational disruptions. Failure to make this distinction may mislead policymakers, investors, and researchers—potentially resulting in overregulation or market mispricing. By proposing a more comprehensive framework, this paper contributes new insights that enhance both academic understanding and practical interpretation of BTDs.

2.2. Direction of Sales Change

Prior research has shown that the direction of sales change serves as a meaningful signal of operational disruption within firms (Banker & Chen, 2006; Banker et al., 2014, 2017). Banker and Chen (2006) demonstrate that incorporating sales decline as a moderating variable enhances the predictive power of time-series models for future earnings. Similarly, Banker et al. (2014) find that during economic downturns, firms with declining sales tend to implement more aggressive cost-cutting strategies than those with rising sales, often resulting in improved margins. Lev and Thiagarajan (1993) interpret sales declines as negative and abnormal signals, whereas sales increases are seen as routine and non-disruptive. Further, Banker et al. (2017) observe that sales downturns often lead to asset write-downs and goodwill impairments, reinforcing the idea that sales direction reflects underlying managerial responses (Linsmeier & Wheeler, 2021; Han et al., 2021). These findings suggest that sales decline is not merely a financial outcome but a trigger for real operational changes that directly affect earnings quality.

The trajectory of revenue changes significantly affects earnings, cash flow, and accrual persistence. A drop in sales serves as a warning to managers, signaling the need to refine or overhaul the firm’s business model. In contrast, a firm experiencing sales growth is more likely to maintain its existing strategy. Approximately 75% of revenue declines are followed by an increase in the next period, while about 62% of revenue increases are followed by further increases. This indicates that sales declines are often short-lived, with managers acting swiftly to reverse the downturn and steer the company toward recovery. In defining our sales-down moderating construct, we follow Banker and Chen (2006), who classify firms into “sales decline” and “sales increase” categories using a simple dummy variable (negative vs. positive sales change). This binary dual approach avoids the need for an arbitrary percentage threshold in the model equations.

2.3. Earnings Management

Earnings management theory holds that managers intentionally adjust accounting estimates and timing of accruals to achieve reporting objectives (Healy & Wahlen, 1999; Dechow & Dichev, 2002). In the tax context, EM is often associated with large positive BTDs, where book income is inflated through discretionary accruals, while taxable income is minimized. These practices reduce earnings persistence, as the accruals reverse in subsequent periods (Hanlon, 2005; Blaylock et al., 2012). The EM literature largely focuses on the accrual component of earnings and assumes that cash flows remain unaffected. Therefore, earnings volatility under EM theory arises not from real business changes, but from temporary manipulations of reported income within GAAP limits. While this framework explains reduced persistence for LPBTDs, it does not fully account for LNBTDs or disruptions to cash flows.

2.4. Operations Theory

While the traditional earnings management literature attributes BTDs to managerial discretion over accruals, the management accounting literature offers a complementary perspective by focusing on how operational changes affect the structure and timing of reported earnings. This alternative view, which we refer to as operations theory, is grounded in studies that examine how firms respond to shifts in demand, particularly through cost behavior and restructuring activities. In short, operations theory suggests that the root cause lies in real economic disruptions, such as factory closures, workforce reductions, and asset write-downs, that follow a sales downturn. These operational changes disrupt business processes and create variability in both accrual and cash flow components, thus lowering earnings predictability (Linsmeier & Wheeler, 2021).

A growing body of research shows that sales declines often prompt firms to make significant changes in their operations—such as facility closures, asset disposals, layoffs, and alterations in product mix—which introduce disruptions in both cash flows and accruals (Anderson et al., 2003; Banker & Chen, 2006). These operational responses are typically non-discretionary and represent strategic efforts to maintain profitability amid declining revenue. As a result, the earnings stream becomes less stable and less predictable, leading to reduced earnings persistence that is not the result of manipulation but of real economic adaptation.

Banker et al. (2014) show that cost reductions during downturns can improve profit margins, but also introduce significant volatility into the earnings process, especially when combined with fixed-cost structures. Banker et al. (2017) further demonstrate that impairment decisions and asset write-downs—common in response to negative demand shocks—create large, one-time accruals that diminish the persistence of reported earnings. These findings are supported by Linsmeier and Wheeler (2021), who argue that asset impairments and restructuring charges reflect meaningful signals about operational change, rather than opportunism.

Unlike EM theory, which centers on accrual-based discretion, OT emphasizes the role of real activities and structural shifts in the firm’s business model. Events such as inventory obsolescence, severance payments, and divestitures may all trigger large accruals or unusual cash flows that decrease earnings persistence, while also affecting temporary BTDs. For example, restructuring costs may be recognized immediately for book purposes but deferred for tax, contributing to negative BTDs (Han et al., 2021). This literature suggests that large BTDs may arise not from manipulation, but from the timing misalignment between tax and book recognition of economically justified decisions. Therefore, interpreting BTDs as solely indicative of earnings management overlooks the operational realities that many firms face, especially in periods of declining sales. Operations theory, thus, provides a more holistic framework by incorporating economic substance and business strategy into the analysis of BTDs and earnings persistence.

3. Materials and Methods

3.1. Hypothesis: Operations Theory

The management accounting literature provides a valuable foundation for understanding how operational adjustments, triggered by sales declines, impact short-term earnings persistence. For example, studies on cost behavior (e.g., Anderson et al., 2003; Banker & Chen, 2006) illustrate how cost-cutting measures and capacity adjustments introduce variability in accruals and cash flows. These disruptions—such as restructuring costs, asset impairments, and changes in inventory levels—result in one-time charges that undermine earnings predictability. By linking these findings to earnings persistence, we bridge the gap between management accounting and tax research, offering a more comprehensive understanding of the factors influencing earnings variability. Operational changes, such as factory closures, often lead to asset write-downs or impairments, which result in large, one-time accruals. Workforce reductions incur severance costs that create short-term accrual liabilities, while product mix adjustments may lead to inventory write-offs or changes in reserve estimates. Similarly, asset disposals may produce gains or losses that disrupt cash flows and accrual patterns. These changes reduce the stability and persistence of earnings. These operational changes disrupt the continuity of earnings by introducing large, non-recurring accruals and irregular cash flows. For example, asset write-downs and severance payments create accruals that reverse in subsequent periods, while disrupted cash flows reflect the operational instability caused by declining sales. This variability reduces short-term earnings persistence.

While large BTDs have been linked to tax avoidance and real earnings management (REM) in prior literature, this study offers a distinct perspective through the lens of operations theory. Unlike REM, which involves deliberate alterations to operations to influence earnings or tax outcomes, operations theory attributes earnings variability to involuntary changes driven by sales declines, such as asset write-downs or restructuring activities. These operational adjustments disrupt accruals and cash flows, undermining short-term earnings persistence (Ahnan & Murwaningsari, 2019).

3.1.1. First Hypothesis: The Impact of Sales Decline on Earnings Persistence

Prior literature in management accounting suggests that a decline in sales is a strong indicator of operational disruption, which in turn affects the stability and persistence of earnings (Banker & Chen, 2006; Banker et al., 2014). Such disruptions typically lead to non-recurring costs like asset write-downs, severance payments, and impairments, which introduce volatility into both accruals and cash flows (Banker et al., 2017; Anderson et al., 2003). Therefore, consistent with prior studies, we expect sales-down conditions to reduce short-term earnings persistence:

H1: 

Persistence of pre-tax book income (PTBI) is negatively associated with sales-down.

As described above, the tax literature focuses on explaining how large book–tax differences relate to earnings persistence (Hanlon, 2005; Blaylock et al., 2012). This literature divides firm-year observations into quintiles based on the size of BTDs deflated by total assets and labels the top quintile LPBTDs and the bottom quintile LNBTDs. By reviewing the components of BTDs described by Poterba et al. (2011), we find useful information about the types of items that are likely to cause firms to have LPBTDs and LNBTDs. An important component of BTDs is the difference between depreciation for book purposes and for tax purposes. As Drake (2012) observed, growth firms have high deferred tax liabilities due to depreciation, so LPBTD firms are likely to be early-stage companies. In fact, Guenther (2011) observed that young, small firms with high ROAs are among the firms with high BTDs. On the other hand, firms that are past the maturity stage and are no longer replacing physical assets are likely to have LNBTDs. Drake (2012) included life-cycle as an additional explanatory variable in her model, taking away some of the information in BTDs.

We follow an alternative approach and use the information in BTDs in conjunction with sales-down, as suggested by management accounting research, to examine whether operating changes precipitated by declining sales are more strongly associated with lower earnings persistence for firms with LPBTDs and LNBTDs. As a robustness test, and based on literature findings, we also examine whether SD is associated with the components of earnings, meaning cash flow and accruals (Lev et al., 2021).

The use of a joint interaction term between the two dummy variables (SD × LN(P)BTD) aligns with the theoretical framework of this study, since we posit that the combined effect of these factors has a distinct impact on earnings persistence beyond their individual effects. By using an interaction term, we can directly test these hypotheses. Our key prediction is that the types of firms that have LPBTDs and LNBTDs are more likely to have disruptions associated with sales-down that lead to changes in operations and less persistent earnings.

3.1.2. Second Hypothesis: Sales Decline and Earnings Persistence for LPBTD Firms

Firms with large positive BTDs are often associated with income-increasing accruals, interpreted by the literature as a sign of earnings management (Hanlon, 2005; Blaylock et al., 2012). However, LPBTDs can also arise in growth-stage firms, where aggressive capital expansion leads to deferred tax liabilities and timing differences related to depreciation (Drake, 2012; Poterba et al., 2011). When such firms experience a sales decline, they are more likely to adjust operations—e.g., by cutting capital expenditures, adjusting reserves, or restructuring—to respond to shrinking demand. These actions disrupt normal operations and introduce accrual shocks and timing mismatches between book and tax reporting. Empirical evidence shows that restructuring activities (e.g., impairment charges and write-downs) reduce the predictability of earnings streams (Banker et al., 2017; Linsmeier & Wheeler, 2021).

Firms in growth stages are particularly vulnerable to external shocks, such as intensifying competition and technological changes, which can disrupt markets and challenge their position in the industry. A decline in sales is especially destabilizing for these firms, often triggering operational adjustments, such as reductions in capital expenditures, reserve revisions, or restructuring initiatives (Banker & Chen, 2006; Anderson et al., 2003). In addition, sales downturns may lead to bad debt write-offs or warranty expenditures that exceed newly established reserves—since these reserves are typically based on current sales volume, a drop in revenue constrains provisioning, while past quality or credit issues surface as charges. These dynamics can reinforce or exacerbate large positive BTDs and contribute to reduced earnings persistence.

Sales-down periods may amplify the effects of LPBTDs on earnings volatility. For instance, during a downturn, firms may face higher bad debt write-offs or warranty expenditures that exceed the additions to related reserves, leading to earnings volatility. Since these reserve additions are typically based on current sales volume, a decline in sales can reduce new provisioning, while past issues—such as poor receivables collection or product quality failures—surface as charges. These adjustments affect book income without proportionately affecting tax income, reinforcing or exacerbating LPBTDs. Thus, the combination of LPBTDs and declining sales reflects a compound disruption (SD × LPBTD)—not just due to accrual reversals but due to real operational stress—resulting in lower earnings persistence. Based on this logic, we expect:

H2: 

Persistence of pre-tax book income is lower for LPBTD companies when sales decline than for companies that do not have large BTDs.

3.1.3. Third Hypothesis: Sales Decline and Earnings Persistence for LNBTD Firms

Another significant finding of the paper is the LNBTD×SD interaction, where we find that firms with large negative BTDs face lower persistence due to restructuring costs, impairments, or special items, which affect book income without immediate tax consequences (Jackson, 2015; Tang & Firth, 2012). Large negative BTDs often occur in mature or declining firms, where book income is reduced by non-recurring charges, such as asset write-downs, impairment losses, or special items, but taxable income remains higher due to differences in recognition rules (Jackson, 2015; Han et al., 2021). Research by Guenther (2011) highlights how these accounting events relate to abnormal BTDs and reduced earnings quality. In the context of sales declines, LNBTD firms are particularly vulnerable to performance shocks and may engage in restructuring, layoffs, or divestitures to manage costs and re-align operations (Anderson et al., 2003). Studies show that firms in decline stages exhibit greater volatility and lower earnings predictability when undergoing operational changes (Drake, 2012; Banker et al., 2014). These responses affect both accruals and cash flows, thus reducing earnings persistence (SD × LNBTD × PTBI):

H3: 

Persistence of pre-tax book income is lower for LNBTD companies when sales decline than for companies that do not have large BTDs.

Because our hypotheses are based on operations as opposed to earnings management, they apply to persistence of both cash flows and accruals. Accrual persistence may be particularly sensitive to operating changes. For instance, when companies in growth stages encounter dropping demand, their capital expenditures slow down or stop, reducing the persistence of accruals. Other items described above, such as write-downs, impairment charges, and changes to reserves, may also affect the persistence of accruals.

3.2. Methods

PTBI represents pre-tax book income (Compustat variable PI), scaled by average assets (year-over-year change in Compustat variable AT). Following established literature (Dechow & Dichev, 2002), we scale variables by total average assets. We see that firms with sales increases have significantly higher current and future PTBI, as well as current PTACC, PTCF, and ETR. Sales-up firms also appear to be smaller than sales-down firms. The univariate results show that book–tax differences are not statistically different between sales-up and sales-down firms.

We employ ordinary least squares (OLS) regression to examine the relationship between BTDs and earnings persistence. OLS is chosen due to its ability to estimate linear autoregressive time-series persistence relationships effectively, ease of interpretation, and widespread use in literature persistence models (e.g., Hanlon, 2005; Blaylock et al., 2012). To establish a baseline, we begin by estimating the basic persistence of pre-tax earnings by modeling future pre-tax earnings as a function of current pre-tax earnings for all three samples (full sample, sales-up, and sales-down), as shown in the OLS equation:

$${PTBI}_{i,t+1}={\beta }_0+{\beta }_1{PTBI}_{i,t}+{\epsilon }_{i,t}$$

(1)

Then, we add our moderating variable, sales-down, which is a dummy variable equal to one if a firm’s change in sales (Compustat SALE) from time t − 1 to t is negative, and zero otherwise. Our first hypothesis predicts that firms with sales decreases have lower earnings persistence than firms with sales increases. To test our first hypothesis, we estimate the following equation:

$${PTBI}_{i,t+1}={\beta }_0+{\beta }_1{PTBI}_{i,t}+{\beta }_{2}SD+{\beta }_{3}SD\times {PTBI}_{i,t}+{\epsilon }_{i,t}$$

(2)

Based on our hypothesis, we expect β₃ to be significantly negative. We then replicate Hanlon’s (2005) model by introducing interactions for firms with large positive and large negative book–tax differences. LNBTD is a dummy variable equal to one for firm-years with scaled temporary BTDs in the lowest quintile of firms in each year, and zero otherwise. Similarly, LPBTD is a dummy variable equal to one for firm-years with scaled temporary BTDs in the highest quintile of firms in each year, and zero otherwise:

$$\begin{array}{c}{PTBI}_{i,t+1}={\beta }_0+{\beta }_1{PTBI}_{i,t}+{\beta }_2{LNBTD}_{i,t}+{\beta }_{3}LPBTD+{\beta }_{4}LNBTD\times {PTBI}_{i,t}\\ +{\beta }_{5}LPBTD\times {PTBI}_{i,t}+{\epsilon }_{i,t}\end{array}$$

(3)

Based on Hanlon (2005), we expect β₄ and β₅ to be significantly negative. To test our Hypotheses 2 and 3, we interact our dummy variable for sales-down with LNBTD and LPBTD:

$$\begin{array}{ll}{PTBI}_{i,t+1}={\beta }_0+& {\beta }_1{PTBI}_{i,t}+{\beta }_{2}lNBTD+{\beta }_{3}LPBTD+{\beta }_{4}LNBTD\times {PTBI}_{i,t}\\ & +{\beta }_{5}LPBTD\times {PTBI}_{i,t}+ {\beta }_{6}SD+{\beta }_{7}SD\times {PTBI}_{i,t}\\ & +{\beta }_{8}LNBTD\times SD+{\beta }_{9}LPBTD\times SD\\ & +{\beta }_{10}LNBTD\times {PTBI}_{i,t}\times SD+ {\beta }_{11}LPBTD\times {PTBI}_{i,t}\times SD\\ & +{\epsilon }_{i,t}\end{array}$$

(4)

Based on our predictions, we expect significant negative coefficients on the three-way interaction terms between sales-down, pre-tax book income, and LNBTD and LPBTD, respectively. Thus, we expect β₁₀ and β₁₁ to be significantly negative.

Cash flow is considered a better indicator of companies’ financial performance than net income, since cash flow is subject to less distortion based on different accounting practices (Dechow, 1994; Jia & Li, 2022). In order to more directly test her claims regarding BTDs as indicators of persistence due to earnings management, Hanlon (2005) breaks earnings into pre-tax accruals (PTACCs) and pre-tax cash flow (PTCF) components, as described in the following basic equation:

$${PTBI}_{i,t+1}={\beta }_0+{\beta }_1{PTCF}_{i,t}+{\beta }_2{PTACC}_{i,t}+{\epsilon }_{i,t}$$

(5)

We follow Hanlon’s next empirical model and separate pre-tax earnings into the accrual and cash flow components. We measure PTCF as the sum of total operating cash flows (Compustat variable OANCF) and taxes paid in cash (Compustat variable TXPD), less cash flow due to extraordinary items (Compustat variable XIDOC). PTACC is measured as the difference between PTBI and PTCF. Both PTCF and PTACC are scaled by average assets. We use the following two OLS equations to estimate the persistence of pre-tax cash flows and accruals:

$$\begin{array}{c}{PTBI}_{i,t+1}={\beta }_0+{\beta }_1{PTCF}_{i,t}+{\beta }_2{PTACC}_{i,t}+{\beta }_{3}SD+{\beta }_4\times SD\times {PTCF}_{i,t}\\ + {\beta }_5\times SD\times {PTACC}_{i,t}+{\epsilon }_{i,t}\end{array}$$

(6)

Based on Hypothesis 1 (sales-down’s moderating effect), we expect the coefficients β₄ and β₅ to be significantly negative. To estimate Hanlon’s (2005) expanded OLS model, we take the basic cash/accrual persistence model and expand the model by adding and interacting the components with our variables for large BTDs:

$$\begin{array}{ll}{PTBI}_{i,t+1}={\beta }_0+& {\beta }_1{PTCF}_{i,t}+{\beta }_2{PTACC}_{i,t}+{\beta }_{3}LNBTD+{\beta }_{4}LPBTD\\ & +{\beta }_5\times LNBTD\times {PTCF}_{i,t}+{\beta }_6\times LPBTD\times {PTCF}_{i,t}\\ & +{\beta }_7\times LNBTD\times {PTACC}_{i,t}+{\beta }_8\times LPBTD\times {PTACC}_{i,t}+{\epsilon }_{i,t}\end{array}$$

(7)

We also expect the coefficients β₇ and β₈ to be significant and negative. The final OLS model is the comprehensive integrated model that captures different elements and sources of earnings persistence:

$$\begin{array}{ll}{PTBI}_{i,t+1}={\beta }_0+& {\beta }_1{PTCF}_{i,t}+{\beta }_2{PTACC}_{i,t}+{\beta }_{3}LNBTD+{\beta }_{4}LPBTD\\ & +{\beta }_5\times LNBTD\times {PTCF}_{i,t}+ {\beta }_6\times LPBTD\times {PTCF}_{i,t}\\ & +{\beta }_7\times LNBTD\times {PTACC}_{i,t}+{\beta }_8\times LPBTD\times {PTACC}_{i,t}+{\beta }_{9}SD\\ & +{\beta }_{10}\times SD\times {PTCF}_{i,t}+{\beta }_{11}\times SD\times {PTACC}_{i,t}\\ & +{\beta }_{12}\times SD\times LNBTD+{\beta }_{13}\times SD\times LPBTD\\ & +{\beta }_{14}\times SD\times LNBTD\times {PTCF}_{i,t}\\ & +{\beta }_{15}\times SD\times LPBTD\times {PTCF}_{i,t}\\ & +{\beta }_{16}\times SD\times LNBTD\times {PTACC}_{i,t}\\ & +{\beta }_{17}\times SD\times LPBTD\times {PTACC}_{i,t}+{\epsilon }_{i,t}\end{array}$$

(8)

Our theory suggests that both the accruals and cash flow components of earnings will exhibit lower persistence when sales decline. Thus, we expect that the net earnings persistence coefficients will be smaller for sales-down firms than for sales-up firms (incremental sales-down coefficients will be negative). We predict significantly negative coefficients β₁₆ and β₁₇ (accruals) but also significantly negatively incremental coefficients β₁₄ and β₁₅ (cash flows).

3.3. Sample Selection

The sample in this study excludes firms in the financial and utilities sectors, as their unique regulatory and accounting environments may distort BTDs. Additionally, the sample is restricted to U.S.-incorporated firms to ensure consistency in tax and financial reporting regulations. For comparability, we also follow the procedures in Hanlon (2005) when selecting our sample but expand the time period to include more recent years. We use 1994 as the starting point because this was the year that ASC 740 (formerly SFAS 109) became effective, which allows all observations to have consistent accounting for the tax variables. We begin with the Compustat annual dataset from 1995 to 2016, including 242,024 observations. We deliberately exclude data from 1994 to control for effects of the starting point. We eliminate utilities and financial services and firms not incorporated in the U.S. We drop 108,183 observations with SIC codes from 6000 to 6799. We drop 24,206 observations with SIC codes from 4000 to 4999. We also eliminate firms with missing pre-tax income or pre-tax financial reporting losses (16,408 observations), negative current tax expense, or a net operating loss (71,246 observations). Observations with a negative current tax expense or net operating losses were excluded because these firms are not profitable, and their BTDs arise from structural issues, such as tax credits or carryforwards, rather than typical timing or temporary differences. Including such observations could distort the analysis by introducing outliers and weakening the relationship between book–tax differences and earnings persistence. Moreover, this approach is consistent with prior research (e.g., Hanlon, 2005; Blaylock et al., 2012), which excluded similar observations to ensure reliable and interpretable results. After deleting observations with missing data, our final sample contains 21,981 firm-year observations and 4341 firms.

Consistent with the approach used by Hanlon (2005), we categorize firms based on the size and direction of their BTDs by ranking BTDs into yearly quintiles. Firms in the top quintile are classified as having large positive BTDs, while those in the bottom quintile are designated as having large negative BTDs. The LPBTD subset includes 4396 firm-year observations across 1018 unique firms, whereas the LNBTD subset consists of 4395 firm-year observations from 999 firms. The remaining 13,190 firm-year observations, representing 2324 firms, fall within the middle three quintiles and are considered to have moderate BTDs.

Table 1. Descriptive statistics.

Panel A: Full Sample
Variable	N	Mean	StD	25%	Median	75%
PTBIi,t+1	21,981	0.128	0.099	0.062	0.107	0.17
PTBIi,t	21,981	0.137	0.105	0.067	0.114	0.178
PTCFi,t	21,981	0.163	0.121	0.092	0.148	0.218
PTACCi,t	21,981	−0.026	0.088	−0.070	−0.033	0.005
BTDi,t	21,981	0.006	0.043	−0.011	0.003	0.02
Total Assetsi,t	21,981	3379	18,146	29	210	970
ETRi,t	21,981	0.343	0.106	0.308	0.362	0.390
DTEi,t	21,981	5.735	152	−0.956	0.12	3.138
REVENUEi,t	21,981	3906	17,511	120.6	440	1670
Pre-Tax Income	21,981	447	2321	9.76	37	152
Panel B: Sales-Up Sample
Variable	N	Mean	StD	25%	Median	75%
PTBIi,t+1	17,911	0.135	0.099	0.067	0.114	0.178
PTBIi,t	17,911	0.146	0.107	0.076	0.123	0.188
PTCFi,t	17,911	0.167	0.125	0.093	0.153	0.225
PTACCi,t	17,911	−0.021	0.09	−0.066	−0.030	0.011
BTDi,t	17,911	0.006	0.043	−0.011	0.0028	0.02
Total Assets	17,911	2916	15,656	33	215	917
ETRi,t	17,911	0.343	0.103	0.311	0.363	0.390
DTEi,t	17,911	6.28	141	−0.963	0.117	2.99
REVENUE	17,911	3641	16,571	123	435	1592
Pre-Tax Income	17,911	425	2211	10.6	38.5	149
Panel C: Sales-Down Sample
Variable	N	Mean	StD	25%	Median	75%
PTBIi,t+1	4070	0.0.98	0.089	0.045	0.08	0.129
PTBIi,t	4070	0.096	0.082	0.042	0.077	0.127
PTCFi,t	4070	0.143	0.097	0.085	0.131	0.188
PTACCi,t	4070	−0.046	0.078	−0.081	−0.047	−0.015
BTDi,t	4070	0.0077	0.039	−0.008	0.0035	0.02
ATi,t	4070	5416	26,360	15	185	1312
ETRi,t	4070	0.340	0.122	0.291	0.355	0.390
DTEi,t	4070	3.33	191	−0.9	0.133	3.99
REVENUEi,t	4070	5070	21,120	109	468	2007
Pre-Tax Income	4070	544	2754	6.45	30.5	164

presents statistical descriptives. In the Panel A, the full sample’s descriptive statistics align closely with those reported in the Hanlon findings. Yet, our sample firms are notably larger, with asset averages of around 3379 million, as compared to 1726 million in the Hanlon sample. Our sample also shows slightly higher profitability, with a PTBI of 0.137 versus 0.132. Consistent with prior studies (Dechow, 1994), both mean and median accruals are negative. Mean and median cash flows exceed pre-tax income, at 0.160 and 0.148, compared to Hanlon’s at 0.143 and 0.136. The mean ETR of 34% is near the statutory rate of 35%. Note that following the Tax Cuts and Jobs Act (TCJA) on 20 December 2017, the corporate statutory tax rate was reduced to a flat 21% effective 1 January 2018. ETR is constrained between 0 and 1, as firm-years with negative tax expense were excluded during sample selection. Panels B and C give us descriptive information based on the direction of year-over-year sales change. The sales-decline subsample consists of 4070 firm-years across 587 firms, while the sales-increase (sales-up) group includes 17,911 firm-years, representing 3754 firms.

4. Results

4.1. Correlations

A Pearson and Spearman correlations comparison can validate the robustness of the observed relationships. Pearson correlation measures the strength and direction of the linear relationship between variables, while Spearman correlation measures the strength and direction of the monotonic relationship between variables and is less sensitive to outliers. The inclusion of both metrics provides a more robust understanding of relationships. Pearson is appropriate for linear relationships, while Spearman captures rank-based associations and can detect non-linear but monotonic trends. If the data contain outliers or potential non-linear relationships, Spearman correlation can offer additional insights.

Table 2. Partial correlations matrix (Pearson and Spearman).

Panel A—Pearson Correlations
	PTBIi,t+1	PTBIi,t	PTCFi,t	PTACCi,t	BTDi,t	ATi,t	ETRi,t
PTBIi,t+1	_
PTBIi,t	0.719 *	_
	<0.0001
PTCFi,t	0.60 *	0.701 *	_
	<0.0001	<0.0001
PTACCi,t	0.035 *	0.231 *	−0.532 *	_
	<0.0001	<0.0001	<0.0001
BTDi,t	−0.040 *	0.021 *	−0.039 *	0.078 *	_
	<0.0001	0.002	<0.0001	<0.0001
ATi,t	−0.048 *	−0.05 *	−0.022 *	−0.030 *	−0.011	_
	<0.0001	<0.0001	0.001	<0.0001	0.103
ETRi,t	0.014 *	−0.026	0.020 *	−0.058 *	0.081 *	−0.061 *	_
	0.0423	0.0001	0.003	<0.0001	<0.0001	<0.0001
Panel B—Spearman Correlations
	PTBIi,t+1	PTBIi,t	PTCFi,t	PTACCi,t	BTDi,t	ATi,t	ETRi,t
PTBIi,t+1	_
PTBIi,t	0.719 *	_
	<0.0001
PTCFi,t	0.575 *	0.674 *	_
	<0.0001	<0.0001
PTACCi,t	0.026 *	0.167 *	−0.53 *	_
	<0.0001	<0.0001	<0.0001
BTDi,t	−0.075 *	−0.028 *	−0.057 *	0.039 *	_
	<0.0001	<0.0001	<0.0001	<0.0001
ATi,t	−0.116 *	−0.147 *	−0.02 *	−0.138 *	0.042 *	_
	<0.0001	<0.0001	0.003	<0.0001	<0.0001
ETRi,t	0.016 *	−0.012	0.011	−0.024 *	0.071 *	−0.146 *	_
	<0.05	<0.1	0.084	0.0004	<0.0001	<0.0001

Table 2 shows the Pearson correlations (below diagonal) and the Spearman correlations (above diagonal). Below the correlations are the p-values. * Represents 1 percent statistical significance, respectively. Refer to Table A1 in Appendix A for variable definitions.

provides a Pearson–Sperman correlation matrix. None of the correlations are large enough to warrant concern regarding multi-collinearity. In addition, in both Spearman and Pearson coefficients (−0.532 * and −0.53 *, both significant at 1% level), cash flows and accruals are negatively correlated with each other, consistent with prior literature (Dechow, 1994). The correlation between book income and cash flows is higher than the correlation between accruals and book income, consistent with current earnings management literature. Further, we see a positive correlation between PTBI and BTD, with a Pearson correlation coefficient of 0.021, statistically significant at the 1% level. This is expected, as firms that report higher earnings are also more likely to engage in timing-based tax planning. However, BTD’s correlation with PTBI_t+1 is negative (coeff_pearson = −0.04 *), signaling BTD’s effect on lower earnings persistence. More interesting is BTD’s negative correlation with cash flows (coeff_pearson = −0.039 */coeff_spearman = −0.057 *), and oppositely, its positive correlation with accruals (coeff_pearson = 0.078 */coeff_spearman = 0.039 *). Firms with larger temporary BTDs tend to have lower cash flows from operations. These firms may rely more on non-cash earnings components and defer taxable income. This pattern is consistent with earnings management or tax deferral strategies. Larger temporary BTDs also show higher accruals. Since temporary BTDs are fundamentally timing differences, it makes sense that they stem from book income being driven more by accruals than cash. This supports Hanlon’s idea that temporary BTDs can act as a signal for accrual-based earnings management.

4.2. Main Results

In

Table 3. Basic tests of earnings persistence.

Panel A: Pre-Tax Book Income
PTBIi,t+1 = β0 + β1 × PTBIi,t + εi,t
Variable	Full Sample (1)		Sales-Up Firms (2)		Sales-Down Firms (3)
PTBIi,t	0.676	***	0.686	***	0.578	***
	(0.000)		(0.000)		(0.000)
Intercept	0.035	***	0.034	***	0.042	***
	(0.000)		(0.000)		(0.000)
No. of Obs.	21,981		17,911		4070
R-squared	0.52		0.55		0.28
Panel B: Pre-Tax Book Income with Cash Flows and Accruals
PTBIi,t+1 = β0 + β1 × PTCFi,t + β2 × PTACCi,t + εi,t
Variable	Full Sample (1)		Sales-Up Firms (2)		Sales-Down Firms (3)
PTCFi,t	0.707	***	0.713	***	0.627	***
	(0.000)		(0.000)		(0.000)
PTACCi,t	0.551	***	0.563	***	0.436	***
	(0.000)		(0.000)		(0.000)
Intercept	0.027	***	0.028	***	0.028	***
	(0.000)		(0.000)		(0.000)
No. of Obs.	21,981		17,911		4070
R-squared	0.53		0.57		0.31

***, **, and * represent 1, 5, and 10 percent statistical significance, respectively.

, Panel A displays the initial findings from estimating the basic persistence model outlined in Equation (1). Column 1 shows that firms in our sample demonstrate persistent earnings behavior, indicated by the positive coefficient on PTBI (coeff = 0.676 ***). This value serves as the reference point for comparing other earnings persistence measures. In the 2nd and 3rd columns, we estimate Equation (1) separately for SU and SD firm-years. The comparisons made across the columns for Table 3 are causal, with statistical tests of differences between SU and SD firm-year observations based on models in later tables that include SD as a moderating variable. We find that both sales-up and sales-down firms show persistent earnings behavior similar to the full sample. However, the coefficient on PTBI is smaller for SD firms compared to SU firms. This means lower persistence for sales-down firm-year observations, supporting the first hypothesis (coeff_SU = 0.686 *** vs. coeff_SD = 0.578 ***).

In Panel B of Table 3, Equation (2) estimation is reported. For the full sample, which is in Panel B, column 1, the results support persistence of both cash flows and accruals through significant coefficients on the earnings component’s coefficients (coeff_CF = 0.707 *** and coeff_ACC = 0.551 ***). A comparison of columns 2 and 3 results for sales-up and sales-down firms reports lower coefficients on both cash flows and accruals for sales-down firms, suggesting lower earnings persistence for these firms (PTCF-SU_coeff = 0.713 *** vs. PTCF-SD_coeff = 0.627 ***).

In

Table 4. Persistence of pre-tax book income. Results of estimating models with sales-down (SD) and BTDs.

Variable	SD Only Model (1)		No SD Model (2)		Integrated Model (3)
PTBIi,t	0.686	***	0.697	***	0.694	***
	(0.000)		(0.000)		(0.000)
LNBTD			0.017	***	0.014	***
			(0.000)		(0.000)
LPBTD			0.007		0.004	*
			(0.000)		(0.059)
LNBTD × PTBIi,t			−0.039	***	−0.018	*
			(0.000)		(0.085)
LPBTD × PTBIi,t			−0.075	***	−0.044	***
			0.000		(0.001)
SD	0.007	***			−0.002
	(0.000)				(0.323)
SD × PTBIi,t	−0.108	***			0.016
	0.000				(0.399)
LNBTD × SD					0.022	***
					(0.000)
LPBTD × SD					0.017	***
					(0.001)
LNBTD × SD × PTBIi,t					−0.231	***
					(0.000)
LPBTD × SD × PTBIi,t					−0.237	***
					(0.000)
Intercept	0.034	***	0.031	***	0.032	***
	(0.000)		(0.000)		(0.000)
No. of Obs.	21,981		21,981		21,981
R-squared	0.519		0.5216		0.5246

This table represents the tests of earnings persistence. Column 1 shows the results of tests based on the direction of sales change. Column 2 shows the results for the effect of large negative and large positive book–tax differences. Column 3 shows the results for the effects of sales-down, large negative and large positive book–tax differences, and interactions. Standard errors are shown in parentheses. ***, **, and * represent 1, 5, and 10 percent statistical significance, respectively. Refer to Table A1 in Appendix A for variable definitions.

, we report the outcomes of estimating models that examine earnings persistence in relation to SD alone (first column results), earnings persistence with the Hanlon variables (column 2), and earnings persistence with the interactions between sales-down and the Hanlon variables LNBTD and LPBTD, which is the third column. The results in column 1 confirm a reduction in earnings persistence linked to sales-down, supporting the hypothesis 1 statement (coeff_SD-PTBI = −0.108 ***). Consistent with the operations theory’s prediction, operational disruption and revenue declines impact the stability and quality of company earnings in a negative way. The 2nd column empirically represents the results for the full sample, producing estimations consistent with Hanlon (2005). Firms with large negative and positive book–tax differences have less persistent earnings, as evidenced by the significantly negative coefficients on LNBTD × PTBI_i,t (−0.039 ***) and LPBTD × PTBI_i,t (−0.075 ***).

In Table 4, the third column shows the outcomes of estimating Equation (4), which incorporates both the sales-down dummy and BTD variables. The findings strongly support a significant reduction in the persistence of PTBI for sales-down firms with either large negative or positive BTDs (SD-LNBTD-PTBI_coeff = −0.231 *** and SD-LPBTD-PTBI_coeff = −0.237 ***). The sales-up coefficients on LNBTD × PTBI_i,t (−0.018 *) and LPBTD × PTBI_i,t (−0.044 ***) are modest in terms of magnitude compared to the incremental coefficients for sales-down observations on LNBTD × SD × PTBI_i,t (−0.231) and LPBTD × SD × PTBI_i,t (−0.237). In other words, these findings corroborate both H2 and H3 predictions. Sales-up situations improve earnings persistence, mitigating the detrimental impact of large BTDs. The impact of large BTDs is notably stronger in companies experiencing sales drops compared to those with growing sales. These results provide compelling evidence for the alternative operations theory, which argues that operational adjustments prompted by declining sales significantly undermine earnings persistence in firms with both large negative and large positive BTDs.

Table 5. Persistence of pre-tax book income with cash flows and accruals. Results of estimating models with sales-down (SD) and BTDs.

Variable	SD Only Model (1)		No SD Model (2)		Integrated Model (3)
PTCFi,t	0.713	***	0.726	***	0.726	***
	(0.000)		(0.000)		(0.000)
PTACCi,t	0.563	***	0.572	***	0.557	***
	(0.000)		(0.000)		(0.000)
LNBTD			0.014	***	0.014	***
			(0.000)		(0.000)
LPBTD			0.006	***	0.006	***
			(0.002)		(0.009)
LNBTD × PTCFi,t			−0.036	***	−0.026	***
			(0.000)		(0.017)
LPBTD × PTCFi,t			−0.073	***	−0.056	***
			(0.000)		(0.000)
LNBTD × PTACCi,t			−0.037	***	0.006
			(0.007)		(0.69)
LPBTD × PTACCi,t			−0.057	***	0.003
			(0.000)		(0.82)
SD	0.007				0.0008
	(0.74)				(0.738)
SD × PTCFi,t	−0.085	***			−0.009
	(0.000)				(0.63)
SD × PTACCi,t	−0.126	***			0.081	***
	(0.000)				(0.003)
LNBTD × SD					−0.003
					(0.558)
LPBTD × SD					0.001
					(0.931)
LNBTD × SD × PTCFi,t					−0.119	***
					(0.001)
LPBTD × SD × PTCFi,t					−0.126	***
					(0.001)
LNBTD × SD × PTACCi,t					−0.363	***
					(0.000)
LPBTD × SD × PTACCi,t					−0.322	***
					(0.000)
Intercept	0.027	***	0.024	***	0.024	***
	(0.000)		(0.000)		(0.000)
No. of Obs.	21,981		21,981		21,981
R-squared	0.5378		0.539		0.5425

This table represents the tests of earnings persistence. Column 1 shows the results of tests based on the direction of sales change. Column 2 shows the results for the effect of large negative and large positive book–tax differences. Column 3 shows the results for the effects of sales-down, large negative and large positive book–tax differences, and interactions. Standard errors are shown in parentheses. ***, **, and * represent 1, 5, and 10 percent statistical significance, respectively. Refer to Appendix A for variable definitions.

displays the outcomes of our analysis of cash flow and accruals persistence. The findings in the first column (sales-down only) confirm the hypothesis that the persistence of PTBI, for both cash flows and accruals, is significantly diminished during sales declines, as shown by SD × PTCF_i,t (−0.085 ***) and SD × PTACC_i,t (−0.126 ***). When a firm faces a sales drop, the cash flow component of earnings becomes less stable and less predictive of future earnings. Likewise, the relationship between sales declines and accruals suggests that the accrual component of earnings also becomes more erratic and less persistent during these periods.

The empirical model estimation in column 2 reproduces Hanlon’s model, including coefficients for the interactions between large BTDs and cash flows and accruals components. Respectively, we have LNBTD × PTCF_i,t (−0.036 ***), LPBTD × PTCF_i,t (−0.073 ***), LNBTD × PTACC_i,t (−0.037 ***), and LPBTD × PTACC_i,t (−0.057 ***). The estimation presented in column 3 clearly highlights the impact of revenue decreases in diminishing the persistence of both the cash and accrual earnings components, PTCF and PTACC. Specifically, regarding cash flows, the coefficients for sales-up (SU) observations are comparatively minor: LNBTD × SU × PTCF_i,t (−0.026 ***) and LPBTD × SU × PTCF_i,t (−0.056 ***), compared to the incremental coefficients for the sales-down (SD) observations: LNBTD × SD × PTCF_i,t (−0.119 ***) and LPBTD × SD × PTCF_i,t (−0.126 ***). This further supports H2 and H3 statements and operations theory, which expects persistence to drop due to operations changes and disturbances associated with sales declines. With respect to accruals, the coefficients for sales-up observations vanish (neither are significant), and the incremental coefficients are especially large: LNBTD × SD × PTACC_i,t (−0.363 ***) and LPBTD × SD × PTACC_i,t (−0.322 ***), given that the base earnings persistence variable is about 0.726 ***. As a robustness test (

Table 6. Persistence of pre-tax book income with earnings management (EM). Results of estimating models with SD, EM, and BTDs.

Variable	Earnings Model (1)		EM-BTD Model (2)		EM-BTD-SD Model (3)
PTBIi,t	0.764	***	0.791	***	0.798	***
	(0.000)		(0.000)		(0.000)
EM	0.019	***	0.019	***	0.020	***
	(0.000)		(0.000)		(0.000)
EM × PTBIi,t	−0.169	***	−0.179	***	−0.182	***
	(0.000)		(0.000)		(0.000)
LNBTD			0.022	***	0.0197	***
			(0.000)		(0.000)
LPBTD			0.007	***	0.009	***
			(0.002)		(0.000)
LNBTD × PTBIi,t			−0.066	***	−0.051	***
			(0.000)		(0.000)
LPBTD × PTBIi,t			−0.075	***	−0.079	***
			(0.000)		(0.000)
LNBTD × EM			−0.002		−0.002
LPBTD × EM			−0.011	*	−0.013	**
					(0.030)
LNBTD × EM × PTBIi,t			0.005		0.007
LPBTD × EM × PTBIi,t			0.105	***	0.114	***
			(0.000)		(0.000)
SD					0.007
					(0.705)
SD × PTBIi,t					−0.046	**
					(0.017)
LNBTD × SD					0.017	***
					(0.000)
LPBTD × SD					−0.003
LNBTD × SD × PTBIi,t					−0.192	***
					(0.000)
LPBTD × SD × PTBIi,t					−0.108	**
					(0.03)
Intercept					0.021	***
No. of Obs.	21,981		21,981		21,981
R-squared	0.556		0.559		0.562

Standard errors are shown in parentheses. ***, **, and * represent 1, 5, and 10 percent statistical significance, respectively. Refer to Table A1 in Appendix A for variable definitions.

), we control for the effect of earnings management, and the negative SD–BTD interaction coefficients stand, LNBTD × SD × PTBI_i,t (−0.192 ***) and LPBTD × SD × PTBI_i,t (−0.108 ***).

Earnings management cannot take away the explanatory power of operations disruptions and significant sales declines, meaning the lower earnings persistence in large BTD groups is a consequence of real economic and business factors driving company rather than management discretionary earnings manipulation. Similarly, in

Table 7. Persistence of pre-tax book income.

Panel A: Results of Estimating Models Based on Large BTD Sampling
Variable	LNBTD Subsample		Full Sample		LPBTD Subsample
PTBIi,t	0.685	***	0.686	***	0.696	***
	(0.000)		(0.000)		(0.000)
SD	0.023	***	0.007	***	0.003
	(0.000)		(0.000)
SD × PTBIi,t	−0.24	***	−0.108	***	−0.092	***
	(0.000)		(0.000)		(0.009)
Intercept	0.047	***	0.034	***	0.032	***
	(0.000)		(0.000)		(0.000)
No. of Obs.	4395		21,981		4396
R-squared	0.461		0.519		0.492
Panel B: Results of Estimating Models Based on Sales-Down Sampling
Variable	Sales-Up Subsample		Full Sample		Sales-Down Subsample
PTBIi,t	0.747	***	0.697	***	0.708	***
	(0.000)		(0.000)		(0.000)
LNBTD	0.021	***	0.017	***	0.039	***
	(0.000)		(0.000)		(0.000)
LPBTD	0.007	***	0.007		0.005
LNBTD × PTBIi,t	−0.062	***	−0.039	***	−0.264	***
					(0.000)
LPBTD × PTBIi,t	−0.051	***	−0.075	***	−0.104	**
	(0.000)		(0.000)		(0.014)
Intercept	0.026	***	0.031	***	0.031	***
	(0.000)		(0.000)		(0.000)
No. of Obs.	17,911		21,981		4070
R-squared	0.589		0.5216		0.314

Standard errors are shown in parentheses. ***, **, and * represent 1, 5, and 10 percent statistical significance, respectively. Refer to Table A1 in Appendix A for variable definitions.

, Panel A, the SD×PTBI_i,t coefficient is negative and significant in both the LNBTD (−0.24 ***) and the LPBTD subset (−0.092 ***), as H1 predicts. Regarding H2 and H3, Panel B reports strong negative coefficients in the SD subset (coeff_LNBTD-PTBI = −0.264 ***/coeff_LPBTD-PTBI = −0.104 ***), compared to SU (coeff_LNBTD-PTBI = −0.062 ***/coeff_LPBTD-PTBI = −0.051 ***) and the full sample (coeff_LNBTD-PTBI = −0.039 ***/coeff_LPBTD-PTBI = −0.075 ***).

The empirical results support all three hypotheses and confirm that large BTDs are associated with reduced short-term earnings persistence, especially during periods of declining sales. This finding validates the central premise of the study and underscores the importance of considering sales-driven operational disruptions in interpreting BTDs. The results are broadly consistent with prior research linking LPBTDs to reduced earnings quality and income-increasing accruals, while also extending the literature by shedding light on LNBTDs, and the interaction between sales volatility and BTD information proxy.

In additional (untabulated) analyses, we find that firm-year observations with large negative BTDs are more likely to report asset write-downs, goodwill impairments, and special items than those with large positive BTDs. Specifically, LNBTD firms show a higher frequency of non-zero values for these variables, indicating that such events are more common among this group. Furthermore, the magnitude of asset write-downs and goodwill impairments is significantly greater for firms with LNBTDs, and special items are, on average, substantially more negative. These results suggest that firms classified with large negative BTDs are undergoing more severe operational disruptions—such as asset impairments, divestitures, and restructuring charges—which are often triggered by declining sales or deteriorating business conditions. This pattern supports the argument that LNBTDs are not solely a product of accounting choices but frequently arise from real economic adjustments in response to performance shocks. The findings are consistent with recent literature emphasizing the role of economic fundamentals—such as restructuring activity, cost variability, and asset volatility—in shaping both BTDs and short-term earnings outcomes (Jackson, 2015; Banker et al., 2017; Linsmeier & Wheeler, 2021). Our results extend this line of research by showing that the accounting signals associated with LNBTDs often reflect underlying changes in the firm’s operating structure and business model, rather than opportunistic manipulation.

Importantly, this evidence challenges the conventional interpretation of BTDs as primarily reflective of earnings management behavior. While prior research has largely focused on discretionary accruals and tax avoidance to explain large BTDs, our study introduces operations theory as a complementary framework that attributes negative BTD observations to tangible operational responses to adverse conditions. By highlighting the connection between sales declines, operational restructuring, and the emergence of LNBTDs, our framework offers a broader lens for understanding the informational content of BTDs.

5. Discussion

Findings, Implications, and Limitations

This paper investigates why large BTDs are particularly problematic for short-term earnings persistence during periods of sales decline. Revenue decline is a key variable in solving this BTD and earnings persistence relation. Our findings indicate that reductions in persistence of pre-tax book income associated with LPBTDs and LNBTDs are much larger when sales decrease than when sales increase. The reduction in earnings persistence attributable to cash flows is greater for both LPBTDs and LNBTDs when sales decrease. Moreover, the reduction in accruals persistence is only observed for both types of BTDs when sales decrease. We propose a complementary framework—referred to as operations theory—to enhance the traditional earnings management perspective. Our findings indicate that firms facing sales declines experience a notable reduction in earnings persistence relative to those with stable or growing sales. This reduction is especially evident among firms with substantial BTDs, where disruptions in accrual components play a key role in diminishing earnings predictability.

The study focuses on U.S.-incorporated firms from 1995 to 2016, a period following the adoption of ASC 740, and excludes financial and utility sectors due to their distinct accounting rules. These design choices improve internal consistency but may limit generalizability to other sectors or international settings. Still, the findings provide a foundation for future research to extend the analysis across different tax systems and industry contexts.

The findings call for a more nuanced understanding of BTDs that distinguishes between financial manipulation and legitimate business responses to economic conditions. Failing to distinguish between EM-driven and operationally driven BTDs has negative consequences. Misattributing operationally driven BTDs to EM may lead to flawed policy decisions (e.g., overregulation or incorrect tax policies). For investors and analysts, this could mean misguided inferences, such as mispricing firms due to perceived opportunism when operational explanations are valid. For policymakers, misinterpreting BTDs as purely EM-driven symptoms could result in overly strict accounting or tax regulations, and disincentivize legitimate operational adjustments. Researchers might draw invalid conclusions if they fail to differentiate between these types of BTDs, undermining the broader understanding of firm behavior.

The findings have practical implications for managers, policymakers, and accountants. Managers must anticipate the financial implications of operational changes, such as workforce reductions or asset write-downs, and integrate these into strategic planning to minimize earnings variability. Policymakers could consider providing targeted tax relief to firms undergoing restructuring to promote financial stability. Accountants and auditors, meanwhile, should focus on accurately capturing the financial impact of these changes, ensuring transparent reporting. Additionally, firms should align their tax planning and earnings management strategies with operational adjustments to mitigate the financial disruptions associated with BTDs.

6. Conclusions and Contributions

This study makes two primary contributions. First, it introduces a novel perspective—operations theory—to the BTD information literature and earnings quality field. While prior research has largely interpreted large BTDs as indicators of earnings management, this study proposes that BTDs may also reflect real operational disruptions triggered by declining sales, such as asset write-downs, restructuring, and changes in cost structure. We provide an alternative explanation, “operations theory,” for the observed association between large BTDs and lower short-term earnings persistence, which broadens the scope beyond accrual-based earnings management. This study introduces the operations theory framework into the earnings–BTD relation, which posits that changes in the business environment—not just managerial discretion—explain the reduced earnings persistence observed in firms with large BTDs, particularly during sales downturns. By incorporating this operational dimension, the study provides a more holistic explanation for the persistence effects of BTDs. By doing so, it also contributes to bridging the gap between tax accounting and management accounting, offering insights into niche areas, such as BTDs, sales accounting, and earnings quality.

This shift in theoretical lens bridges tax research and management accounting, offering a more comprehensive explanation for the observed earnings volatility. Our study builds on the tax literature by providing new evidence on how BTDs relate to the persistence of earnings and the cash flow and accrual components. We extend this research by introducing a perspective from the management accounting literature, specifically how operational disruptions moderated by sales declines influence earnings behavior. We find that the direction of sales change is an important piece of the puzzle and provide an alternative explanation for the finding of Hanlon (2005) on why firm-years with large negative BTDs have lower earnings persistence. Additional tests show that abnormal BTDs and normal BTDs can provide incremental information about earnings persistence beyond the information in discretionary accruals and total accruals, suggesting that the investigation into the components of BTDs adds value to financial analysis.

Finally, we provide a new perspective challenging the dominant earnings management interpretation, by incorporating insights from the management accounting literature on sales declines. However, future research can further investigate the disruption effects, especially for the large negative BTD subsample. We attribute low persistence partly to restructuring, write-downs, or special items. It might be helpful to further test which of these factors are more influential. For example, intangible write-downs, such as goodwill impairments, are more disruptive than physical asset disposals in the form of sudden write-offs or corporate restructures.