Exploring Intangible Assets’ Contribution to Capital Structure in Thailand’s Listed Companies During COVID-19

Abstract

This study examines whether IAS 38-recognized identifiable intangible assets (excluding goodwill) are associated with corporate leverage in Thailand, an emerging bank-dominated financial system, and whether that relationship changed after the COVID-19 shock. Panel on listed firms supports a stepwise design. Estimation begins with firm fixed-effects models, then proceeds to stricter specifications that add year fixed effects and, in the preferred model, industry-by-year fixed effects; dynamic robustness is evaluated using System GMM. In baseline firm fixed-effects specifications, recognized intangible intensity is positively associated with leverage, and the post-COVID-19 interaction is also significant under lighter controls. Statistical significance, however, fades after accounting for broader macro-financial and sector-specific financing conditions, and System GMM results similarly yield weak coefficients for recognized intangibles once leverage persistence is accounted for. The findings imply that apparent financing relevance for recognized intangibles depends strongly on the surrounding financing regime rather than on a robust independent debt-capacity effect.

1. Introduction

Intangible assets have become an increasingly important driver of firm value creation, yet their role in corporate financing remains conceptually and empirically unsettled (Uribe 2025). In classic capital-structure logic, assets that are more verifiable and recoverable in distress should expand debt capacity by improving pledgeability and reducing expected loss given default (Altomonte et al. 2022). However, intangible assets—despite their economic relevance—often exhibit uncertain liquidation values, limited transferability, and wide valuation dispersion, which can intensify information frictions and weaken secured borrowing. This tension is most acute for recognized, identifiable intangibles that are measurable enough to be recorded under IAS 38 but may still fail to serve as effective collateral. Against this setting, the present study asks a focused question: Do recognized intangible assets translate into higher leverage once firm heterogeneity and time-varying financing conditions are tightly controlled, and did this mapping change after COVID-19?

Prior research offers competing predictions and mixed evidence, largely because “intangibles” are heterogeneous and because institutional environments shape whether accounting recognition becomes contracting-relevant. In developed-market settings, evidence suggests that the financing implications of intangibles depend on asset type and measurability. For example, work in the banking-and-finance literature shows that separating identifiable intangible assets from residual components can materially change inference about leverage, consistent with a debt-capacity channel when creditors can value and enforce intangible claims with some reliability (e.g., Lim et al. 2020). Causal evidence also supports the view that pledgeable intangible capital can expand debt capacity: shocks that erode intangible capital—such as patent-related events—have been associated with lower leverage in settings where intellectual property is monetizable and enforceable (Horsch et al. 2021). These studies collectively imply that when intangible assets are verifiable and transferable, they may become debt-supportive in ways closer to tangible collateral.

At the same time, a substantial body of evidence highlights the opposite mechanism: even when intangibles are economically valuable, they may remain difficult to collateralize, thereby exacerbating financing frictions. This view is especially salient in bank-centered environments where lenders emphasize monitoring and recovery values and may discount intangible collateral relative to hard assets. Consistent with this mechanism, evidence from SME settings finds that higher intangible intensity is associated with lower leverage, suggesting that low pledgeability can dominate debt-capacity considerations (Norkio 2024). More broadly, research on relationship lending suggests that credit supply in emerging markets may depend on soft information and bank–borrower ties rather than on-the-balance-sheet stock of intangible assets, thereby muddying the mapping from recognized intangibles to leverage (e.g., Shehadeh et al. 2024). These findings imply that the sign and magnitude of the intangible–leverage relationship are likely to be institution- and regime-dependent, rather than universal.

The study also examines whether the COVID-19 shock altered the contracting relevance of recognized intangibles. The pandemic represents a regime change in financing conditions, characterized by heightened uncertainty, shifts in bank screening incentives, and large liquidity interventions. Existing crisis-era evidence highlights that firms relied heavily on bank liquidity and credit lines, while bank lending responses varied with bank exposure and balance-sheet strength (Li et al. 2020; Almeida 2021). These features motivate a “slope-shift” hypothesis: even if the average intangible–leverage relation is weak in normal times, a systemic shock may reweight pledgeability and monitoring considerations, thereby changing how recognized intangible assets map into observed leverage. At the same time, identification remains difficult because financing choices may affect intangible accumulation, unobserved firm characteristics may influence both, and leverage itself is dynamically persistent. These challenges motivate the stepwise empirical strategy adopted in this study.

Whether recognized intangible assets enhance debt capacity remains an open empirical question, particularly in emerging, bank-centered financial systems where creditor screening, collateral dependence, and enforcement conditions may differ materially from those in more market-based settings. Rather than assuming a universally positive or negative relationship, this study adopts a competing-mechanisms perspective and examines whether IAS 38-recognized identifiable intangible assets (excluding goodwill) are associated with leverage after carefully controlling for firm heterogeneity and time-varying sectoral financing conditions. In this framework, recognized intangibles may increase debt capacity when they are sufficiently contractible, verifiable, and creditor-relevant, but their financing relevance may also be muted when low pledgeability, valuation uncertainty, and conservative bank screening dominate. The analysis therefore focuses specifically on the recorded, balance-sheet-recognized component of intangible assets and its mapping into leverage under different institutional conditions.

This study contributes to the capital-structure and intangible-assets literature in three ways. First, it sharpens construct clarity by distinguishing recognized identifiable intangibles from broader unrecognized intangible resources, thereby reducing conceptual conflation between balance-sheet-recorded intangible assets and other related but distinct firm capabilities. Second, the empirical design emphasizes within-firm identification through a stepwise specification strategy: the analysis begins with firm fixed effects, then evaluates robustness under stricter fixed-effects structures, including year fixed effects and a preferred specification with industry-by-year fixed effects, and finally assesses dynamic robustness using System GMM. Third, the study frames the leverage relevance of recognized intangibles as a boundary-condition question rather than a universal effect, showing that any observed association should be interpreted in light of creditor screening environments, collateral dependence, and post-COVID-19 financing conditions. In this sense, the paper’s contribution lies not in claiming that intangibles uniformly increase or decrease leverage, but in clarifying the institutional and contracting conditions under which recognized intangible assets become more—or less—relevant to debt capacity.

The remainder of the paper proceeds as follows. Section 2 develops the theoretical framework and sets out the competing mechanisms through which recognized intangible assets may affect leverage, including the role of post-COVID-19 financing conditions as a potential slope-shift setting. Section 3 describes the data, variable construction, and empirical strategy, including the stepwise specification design, stricter fixed-effects structures, and dynamic robustness tests. Section 4 presents the empirical results, beginning with baseline within-firm estimates, then examining post-COVID-19 slope heterogeneity, robustness under stricter fixed effects, alternative leverage measures, and additional identification checks. Section 5 discusses the findings and their implications for the financing relevance of recognized intangibles in a bank-centered environment. The final section concludes the paper.

2. Literature Review

2.1. Theoretical Tension on Intangible Assets and the Leverage Ambiguity

Capital-structure theory implies that leverage is determined by a trade-off between the benefits of debt (e.g., tax shields) and the costs of debt that arise when cash flows are risky and contracting is imperfect. Two primitives matter for “debt capacity”: (i) the degree to which assets are contractible/pledgeable (so lenders can price and recover value in distress) and (ii) the severity of information frictions (so external finance is costly and subject to adverse selection). These primitives naturally generate competing predictions: debt can increase when assets support contracting (a collateral/contractibility channel), but debt can fall when financing frictions dominate, and firms prioritize internal funds (a pecking-order channel). This basic tension is central to Stewart C. Myers’s contrast between the trade-off and pecking-order views, and to the asymmetric-information framework of Nicholas S. Majluf and Myers, which formalizes why firms may avoid equity and rely on internal funds and safer financing sequences (Myers and Majluf 1984).

This ambiguity is amplified when “intangibles” are measured, as in most datasets. In this research, INT is balance-sheet recognized intangible assets—a subset of intangible capital that satisfies accounting recognition thresholds. Under IAS 38, intangible assets are recognized only when they are identifiable (separable or arising from contractual/legal rights) and meet control/benefit and reliable measurement criteria (IFRS Foundation 2021); by construction, large parts of internally generated intangible capital (e.g., brand-building, organizational know-how and many research-phase expenditures) are expensed rather than capitalized. This distinction matters for debt contracting: recognized/identifiable intangibles are, in principle, more verifiable than broad “intangible capital”, yet they still differ widely in transferability, liquidation value, and valuation risk—features that determine whether lenders treat them as collateral-like. Consistent with this measurement logic, Lim et al. (2020) show that leverage implications are interpretable only after distinguishing among intangible types: identifiable intangibles can behave more like contracting capital, whereas residual components, such as goodwill, exhibit greater valuation uncertainty and weaker collateral properties that can dampen borrowing incentives.

These considerations motivate competing, theory-consistent channels for INT. The collateral/contractibility channel predicts a positive INT–LEV relation if recognized, identifiable intangibles are sufficiently contractible and transferable, thereby expanding debt capacity—pushing leverage upward in a trade-off logic. The information-friction/pecking-order channel predicts a negative INT–LEV relation if intangibles remain hard to pledge and intensify asymmetric information, making external finance (especially riskier debt) more costly and increasing reliance on internal funds. Importantly, even in developed markets, the sign is therefore not theoretically obvious because “intangibles” are heterogeneous in pledgeability and valuation risk; the net average association can be weak or null if these forces offset, or if institutions mute the collateral channel—an issue that is particularly relevant in Thailand’s bank-centered environment, which is examined in this research.

2.2. Prior Empirical Evidence in Developed Markets vs. Emerging/Bank-Based Markets

In developed markets, the sign of the intangible–leverage relation is not theoretically obvious because “intangibles” bundle asset types with fundamentally different pledgeability, valuation risk, and transferability. This is a first-order measurement problem: internally generated intangible capital is largely absent from financial statements, while acquired intangibles are recognized but often mix identifiable assets with residual components (especially goodwill). As a result, empirical INT–LEV estimates can flip sign depending on whether researchers separate collateralizable/contractible intangibles from high-uncertainty, weak-collateral components, and depending on the institutional environment that shapes creditor rights and information quality (Lim et al. 2020).

2.2.1. Developed-Market Evidence with a Positive INT–LEV Channel (Debt Capacity/Collateral Channel)

A central positive channel is the debt-capacity mechanism: if a subset of intangibles is separately identifiable, transferable, and reasonably valued (i.e., more contractible), then it can expand borrowing capacity in ways more akin to tangible collateral. Evidence from Lim et al. (2020) illustrates why disaggregation is central: intangible components differ markedly in valuation uncertainty and collateralizability. Goodwill and other residual components are difficult to value and pledge in distress, whereas identifiable intangibles are separately recognized and valued in arm’s-length transactions, potentially leading to different leverage effects across intangible types. Empirically, they find a strong positive relation between identifiable intangibles and leverage, concluding that identifiable intangibles can support debt financing comparably to tangibles—especially when tangible collateral is scarce.

Importantly, quasi-experimental evidence based on plausibly exogenous shocks to intangible capital also supports a debt-capacity channel, consistent with the view that pledgeable intangible assets expand borrowing capacity. Exploiting patent invalidations as plausibly exogenous shocks to intangible capital, Horsch et al. (2021) show that adverse shocks to intangible capital are associated with a material decline in leverage, consistent with a debt-capacity interpretation when intangible assets are sufficiently contractible and valuable to creditors. Taken together, evidence from developed markets suggests that once the empirical construct aligns with contractible/valued intangible components, the INT–LEV relation can become meaningfully positive.

2.2.2. Developed-Market Evidence with a Negative INT–LEV Channel (Pecking Order/Low Pledgeability/Uncertainty Channel)

A second developed-market stream highlights a negative channel in which intangible-intensive firms face greater information asymmetry (Intara and Suwansin 2024), weaker collateral value, and more uncertain long-horizon payoffs, raising the expected costs of debt and shifting financing toward internal funds or equity in line with pecking-order logic. This interpretation is consistent with evidence that the leverage implications of “intangibles” depend on their contractibility and valuation risk: intangible components that are difficult to value and weakly collateralizable (e.g., goodwill-like residuals) are less likely to support debt capacity (Lim et al. 2020). In this view, intangibility proxies not for “collateral” but for valuation opacity and financing friction, which can depress leverage. Recent evidence consistent with this logic comes from settings in which intangibility is interpreted as increasing risk and reducing debt capacity. For example, Rajaiya (2023) shows that patenting success is associated with lower leverage and a greater propensity to raise funds via equity rather than debt, and supports causal interpretation using an IV based on patent-examiner leniency. Complementing this, evidence from Finnish SMEs also indicates that higher intangible-capital intensity is associated with lower leverage, consistent with an interpretation of a lender-valuation/pledgeability constraint in smaller, bank-dependent firms, where collateral and verification frictions are particularly salient (Norkio 2024). Crucially, this negative stream does not “contradict” the positive stream—because it often reflects different intangible constructs (e.g., knowledge intensity vs. recognized, identifiable intangibles) and/or environments in which uncertainty and verification frictions dominate the debt-capacity channel.

2.2.3. Emerging/Bank-Based/Relationship-Lending Evidence

In emerging, bank-oriented systems, the INT–LEV relation is frequently shaped less by arm’s-length collateral valuation and more by screening, monitoring, relationship lending, and macro-financial constraints. In underdeveloped equity-market environments, lenders may value certain intangibles—especially identifiable ones—because banks can develop informational advantages and can sometimes monetize/assess such assets better than informal equity financiers. Evidence from Vietnam SME financing supports this institutional asymmetry: intangible assets are argued (and tested) to have stronger positive effects on debt finance than on equity finance in underdeveloped equity markets (Le et al. 2024).

Relationship-lending institutions can also generate systematic distortions in bank credit allocation. Using credit-registry data from the Central Bank of Tunisia, Shehadeh et al. (2024) examine 2529 bank–firm relationships for 403 firms (2012–2018) and show that relationship characteristics are strongly linked to loan outcomes. Relationship intensity is associated with a higher likelihood of high-quality loans, whereas longer relationship duration increases the probability of low-quality loans; moreover, low- and non-performing firms tend to maintain longer, closer bank relationships, consistent with adverse selection and moral hazard in a banking system concentrated around a small number of lenders. While their outcome is loan quality rather than leverage, these findings imply that in emerging, relationship-oriented banking systems, the pricing and allocation of credit may depend heavily on relationship dynamics rather than transparent collateral valuation alone—potentially attenuating the marginal role of recognized intangible assets in explaining cross-firm variation in leverage.

Finally, emerging-market corporate leverage is highly sensitive to macro-financial conditions that can dominate firm-level asset-composition channels. Using an exceptionally large sample of roughly 800,000 listed and non-listed firms across 28 emerging markets, Alter and Elekdag (2020) show that accommodative global financial conditions—proxied initially by U.S. monetary policy—are associated with faster leverage growth, with stronger effects for financially constrained firms (e.g., SMEs) and for countries whose domestic monetary policy is more aligned with the United States. The results imply that global liquidity affects leverage not only through domestic interest-rate channels but also by relaxing borrowing constraints, and that leverage rises disproportionately for weaker firms when global conditions loosen. This macro-financial sensitivity provides a clean institutional reason for why firm-level mechanisms linking recognized intangibles to debt capacity may be attenuated in emerging markets, particularly when credit supply and sectoral conditions shift at the industry-year level (Felix et al. 2020).

2.2.4. Theoretical Mechanisms Linking Intangible Assets to Higher Leverage During COVID-19

Although traditional trade-off and agency perspectives often imply a negative relationship between intangibles and leverage, pandemic conditions may alter how lenders evaluate firms with higher stocks of recognized, identifiable intangible assets. In crisis settings, the financing relevance of recorded intangible assets may change not because the accounting measure directly captures broader firm capabilities, but instead because systemic shocks reweight creditor screening, liquidity needs, and financing constraints.

Three mechanisms are particularly relevant. First, lenders may place greater weight on expected cash-flow continuity and borrower-specific information during systemic stress, which can temporarily support borrowing by firms with stronger recorded intangible positions if these firms are viewed as more stable or more transparent within existing lending relationships. Second, relationship banking may become more important during crisis periods, allowing some firms to access liquidity support even when hard collateral remains limited. Third, when equity financing becomes more costly or more dilutive, firms may rely more heavily on debt financing, which can mechanically strengthen the observed association between recognized intangible intensity and leverage during the pandemic period. These mechanisms motivate a slope-shift perspective rather than a simple level effect. Even if the average intangible–leverage relation is weak in normal periods, the COVID-19 shock may change how recognized intangible intensity translates into leverage by altering creditors’ risk tolerance, banks’ screening behavior, and firms’ financing choices. Importantly, this interpretation remains tied to the recorded, balance-sheet-recognized component of intangible assets rather than to broader unobserved capabilities such as innovation or digital transformation.

2.3. Hypothesis Development

Section 2.2 establishes a central point for identification and interpretation: the intangible–leverage relation is not sign-obvious because “intangibles” contain components with very different contractibility, valuation risk, and pledgeability, and because institutional features determine whether lenders can translate accounting recognition into usable collateral value (Lim et al. 2020). This ambiguity is particularly relevant for recognized, identifiable intangibles—the construct in this research—because recognition improves measurability but does not guarantee enforceable collateral value. Consequently, the appropriate hypothesis strategy is not to force a single directional prediction, but to derive competing, theory-consistent mechanisms and then to treat a null average association as interpretable when mechanisms offset or institutions mute the channel (Lim et al. 2020; Horsch et al. 2021).

A first mechanism predicts a positive association between recognized intangibles and leverage through a debt-capacity channel. If identifiable intangibles are sufficiently verifiable and transferable, they can function as collateral-like inputs in debt contracting: they improve lenders’ expected recovery, reduce contracting frictions, and relax borrowing constraints. Empirical evidence from developed markets shows that once researchers isolate identifiable intangible assets from residual components such as goodwill, the leverage implication becomes more positive, consistent with the idea that identifiable intangibles support debt financing in a manner closer to that of tangible assets (Lim et al. 2020). Causal evidence further supports this debt-capacity interpretation: shocks that reduce intangible capital—such as patent invalidations—lower leverage, suggesting that pledgeable intangible capital can expand debt capacity when creditor valuation and enforceability are credible (Horsch et al. 2021). These findings motivate the following directional hypothesis:

H1a (Trade-off/collateral channel).

Recognized intangible asset intensity (INT) is positively associated with leverage (LEV).

A second mechanism predicts a negative association through low pledgeability and financing-friction channels. Even when recognized, many intangible assets remain difficult to liquidate, hard to value in distress, and weak as secured collateral relative to physical assets. If lenders discount intangible collateral—especially under uncertainty—intangibility can amplify information asymmetry and raise expected costs of debt (tighter covenants, rationing, or higher spreads), pushing firms toward internal funds and away from debt, consistent with pecking-order logic. Evidence consistent with this view is documented in bank-dependent contexts and smaller-firm settings where verification and collateral constraints are more binding, with intangible intensity linked to lower leverage (Crouzet et al. 2022; Norkio 2024). More broadly, evidence from relationship-oriented banking environments suggests that screening and monitoring incentives can dominate arm’s-length collateral valuation, making debt availability depend on lender–borrower frictions rather than on the accounting stock of recognized intangibles (Shehadeh et al. 2024). This motivates the competing directional hypothesis:

H1b (Pecking order/low pledgeability channel).

Recognized intangible asset intensity (INT) is negatively associated with leverage (LEV).

The critical implication—especially for Thailand—is that both mechanisms can operate simultaneously, producing an attenuated net effect in equilibrium. Thailand’s financing environment is often characterized as bank-centered, with meaningful relationships, and emerging-market leverage is demonstrably sensitive to macro-financial conditions and credit-supply regimes that can overwhelm firm-level asset-composition channels (Alter and Elekdag 2020). In such settings, even identifiable intangibles may not translate into incremental debt capacity if lenders prioritize tangible collateral, borrower relationships, and conservative screening; as a result, offsetting collateral and low-pledgeability forces can yield a statistically weak or null average association even when the underlying mechanisms are real (Lim et al. 2020; Shehadeh et al. 2024).

This institutional logic also informs the empirical strategy. By using firm fixed effects and industry × year fixed effects, the analysis removes time-invariant firm traits and absorbs industry-specific time-varying shocks to demand, risk, and credit conditions—factors that could jointly influence both intangible recognition and leverage. This design intentionally focuses inference on within-firm variation in INT relative to firms facing the same industry-year environment, where an attenuated (or null) net effect would be expected if institutional features mute the debt-capacity channel on average (Alter and Elekdag 2020; Lim et al. 2020). Accordingly, the study advances a competing-mechanisms hypothesis set in which the null is an outcome under offsetting channels and emerging-market institutions:

• Competing-mechanisms prediction, The association between recognized intangible asset intensity (INT) and leverage (LEV) is theoretically ambiguous: it may be positive (H1a) if identifiable intangibles expand debt capacity, or negative (H1b) if low pledgeability and information frictions dominate; in an emerging, bank-based and relationship-oriented lending system, the net effect may be attenuated, resulting in a statistically weak or null average association (Lim et al. 2020; Shehadeh et al. 2024).

Importantly, the aim of this study is not to argue that intangible assets should universally increase or decrease leverage. Instead, adopting a competing-mechanisms perspective, we examine which channel dominates the financing relevance of recognized intangible assets in an emerging, bank-centered setting—i.e., whether accounting-recognized identifiable intangibles operate as marginal debt capacity through contractibility/collateral-like effects, or whether information frictions and limited pledgeability mute this channel such that the net INT–LEV association becomes weak or null once firm heterogeneity and industry-year financing conditions are tightly controlled.

2.4. Post-COVID-19 Slope Shift

Section 2.2 implies that the average INT–LEV association can be weak or null because collateral/contractibility forces and low-pledgeability/information-friction forces can offset each other (Lim et al. 2020). COVID-19 is precisely the kind of macro shock that can re-weight those forces. The pandemic created an abrupt cash-flow and liquidity shock and triggered unusually large, system-wide financing responses—firms drew heavily on credit lines and relied on banks as “lenders of first resort”, while policy interventions and market disruptions altered the menu and pricing of external finance (Li et al. 2020; Almeida 2021). Consistent with a regime shift in capital-structure dynamics, international evidence shows that the COVID-19 period is associated with changes in how quickly firms adjust leverage toward targets, and U.S. evidence links greater COVID-19 exposure to reduced financial flexibility and altered leverage adjustment behavior (Vo et al. 2022; Rehman et al. 2024). In short, even if pre-COVID-19 mechanisms offset to produce a small average effect, COVID-19 can plausibly change the slope.

The key question is how COVID-19 affects lenders’ and firms’ financing constraints, and how these constraints interact with recognized intangibles. On the creditor side, crises tend to increase uncertainty, elevate downside risk, and sharpen underwriting and monitoring incentives. This can push lenders toward harder collateral and more conservative valuations, mechanically weakening any debt-capacity channel tied to intangibles—especially when recovery values are difficult to verify in distress. The “type matters” insight in Lim et al. (2020) is important as even among recorded intangibles, components differ sharply in collateral properties and valuation risk, and these differences become more binding when lenders and regulators become more conservative (Lim et al. 2020). Related evidence on pandemic-era bank behavior shows that lending responses varied with banks’ exposure and balance-sheet strength, consistent with tighter constraints and heterogeneity in supply (Dursun-de Neef and Schandlbauer 2021). At the loan-allocation margin, evidence from a credit-supply shock around COVID-19 indicates that when bank liquidity is strained, credit can contract and be rationed in ways that depend on bank-level shocks and borrower characteristics, implying that access to debt is shaped by screening and allocation incentives rather than by “paper” asset values alone (Michelson 2025). These patterns jointly motivate an attenuation prediction: in the post-COVID-19 environment, any positive INT-based borrowing capacity is less likely to pass through, and the net association can move closer to zero (or turn more negative) as lenders discount intangibles more heavily.

A bank-based financial system can, in principle, dampen crisis-induced financing frictions through relationship lending. When banks rely on soft information and repeated interactions, they may be willing to continue supplying liquidity during stress, especially via pre-committed facilities. Early-pandemic evidence from the U.S. documents large credit-line drawdowns accommodated by banks, underscoring how relationships and committed liquidity can stabilize funding in crisis conditions (Li et al. 2020). This buffering mechanism, however, does not imply that recognized intangibles become more debt-supportive during stress. Even in relationship-oriented systems, crises can heighten screening intensity and shift credit allocation toward borrowers that are easier to monitor and better protected by hard collateral, reducing the marginal weight placed on assets whose liquidation value and enforceability remain uncertain despite formal recognition (Dursun-de Neef and Schandlbauer 2021). In parallel, crisis periods may amplify precautionary behavior among lenders and borrowers, altering the pricing and availability of debt and thereby changing how balance-sheet composition maps into observed leverage (Lim et al. 2020; Vo et al. 2022). These forces imply that the post-crisis environment can re-weight pledgeability and monitoring considerations, so the relationship between recognized intangible intensity and leverage need not be stable across regimes. Even if the average (unconditional) association is weak, a systemic shock can induce a slope shift—i.e., the sensitivity of leverage to recognized intangibles changes after the onset of COVID-19 because contracting, screening, and collateral priorities are repriced under stress.

H2 (post-COVID-19 slope shift).

The association between recognized intangible asset intensity (INT) and leverage (LEV) differs in the post-COVID-19 period relative to the pre-COVID-19 period, reflecting a regime shift in financing conditions and in the contracting relevance of recognized intangibles (Li et al. 2020; Lim et al. 2020; Vo et al. 2022; Dursun-de Neef and Schandlbauer 2021).

3. Methodology

This research employs an archival methodology, which is appropriate for investigating firm-level financing behavior using historical, publicly disclosed data. Archival designs allow inference from actual corporate decisions embedded in audited financial statements and market data, ensuring reliability, objectivity, and cross-firm comparability. The multi-year panel setting further enables analysis of time variation around structural disruptions, including the COVID-19 shock, that plausibly altered financing conditions and corporate leverage decisions.

3.1. Population and Sample

This research focuses on listed firms in Thailand. The population comprises all firms listed on the Stock Exchange of Thailand (SET) during 2016–2023. Listed firms are required to disclose audited, standardized financial statements, which enable consistent measurement of recognized intangible assets, leverage, and firm characteristics. Thailand provides a relevant emerging-market setting in which corporate financing may reflect both standard capital-structure forces and the institutional features of bank-centered credit allocation. The COVID-19 period (2020–2023) is treated as a regime-shift interval relative to the pre-pandemic period (2016–2019), reflecting a macro-financial disruption that plausibly affected credit supply and firms’ financing constraints.

3.2. Research Models

This research follows the empirical spirit of recent capital-structure tests that evaluate whether asset composition—particularly recognized intangibles—maps into debt capacity after controlling for firm heterogeneity and common sector-year shocks. The baseline specification relates leverage to lagged recognized intangible intensity and standard controls:

$$\begin{array}{cccc}& LE{V}_{i,t}=\alpha +\omega IN{T}_{i,t-1}+\beta TA{N}_{i,t-1}+{\gamma }^{\prime }{X}_{i,t-1}+{\mu }_i+{\epsilon }_{i,t}& & \end{array}$$

(1)

where $LE{V}_{i,t}$ denotes leverage. The main dependent variable is long-term debt scaled by market value ($LTD\_MV$); robustness analyses use alternative leverage proxies including total debt scaled by market value ($D\_MV$), total debt scaled by assets ($D\_A$), and long-term debt scaled by assets ($LTD\_A$). The key explanatory variable $IN{T}_{i,t-1}$ is recognized (balance-sheet) identifiable intangible assets scaled by total assets (excluding goodwill) and lagged by one period. Consistent with collateral-based capital-structure mechanisms, the baseline model also includes the asset tangibility $TA{N}_{i,t-1}$ measured as net property, plant, and equipment (PP&E) scaled by total assets, lagged by one period. The vector $X_{i,t-1}$ contains standard controls—firm size (log total assets), growth opportunities (EV/EBIT), operating profitability (OPROA: defined as operating profit divided by total assets), and liquidity (quick ratio)—each lagged to mitigate simultaneity and align predictors with leverage realized in period $t$.

All baseline regressions include firm fixed effects $\left({\mu }_i\right)$ to absorb time-invariant firm heterogeneity, such as persistent business model, governance, and risk profile. To strengthen identification, subsequent specifications introduce progressively stricter fixed-effects structures by additionally absorbing year fixed effects and, in the preferred strict specification, industry-by-year fixed effects. This stepwise design allows the INT–leverage relation to be evaluated after removing increasingly rich forms of common time-varying heterogeneity.

To test whether the intangible–leverage relation shifts after the COVID-19 shock, the model is extended with an interaction term:

$$LE{V}_{i,t}=\alpha +\omega IN{T}_{i,t-1}+\delta (IN{T}_{i,t-1}\times Pos{t}_t)+\beta TA{N}_{i,t-1}+{\gamma }^{\prime }{X}_{i,t-1}+{\mu }_i+{\lambda }_{j\times t}+{\epsilon }_{i,t}$$

(2)

where $Pos{t}_t$ equals 1 for 2020–2023 and 0 for 2016–2019. The coefficient δ captures a post-COVID-19 slope change in the INT–leverage relation.

In specifications that include firm fixed effects and industry-by-year fixed effects, the Post_t indicator is absorbed by the industry-by-year fixed effects and therefore does not contribute independent identifying variation. The coefficient on the interaction term, δ, is instead identified from within-firm changes in recognized intangible intensity across the pre- and post-COVID-19 periods, relative to firms facing the same industry-year financing conditions. Intuitively, δ captures whether firms whose recognized intangible intensity increases (or decreases) more than their industry-year peers experience a differential change in leverage in the post period. Accordingly, the interaction coefficient is interpreted as a slope shift under common sector-year credit conditions, rather than as a level effect attributable to the post-period dummy itself.

The inclusion of industry-by-year fixed effects in the strict specification is motivated by the possibility that the financing relevance of recognized intangibles may differ across sectors because contracting environments vary systematically by industry. In particular, industries differ in their reliance on hard collateral, tangibility, and creditor screening intensity. Rather than estimating separate industry subsamples, the preferred specification addresses this source of sector-level heterogeneity by absorbing industry-by-year fixed effects, which control for shocks common to firms operating in the same industry-year, including sector-specific credit conditions, demand fluctuations, and policy exposure.

3.3. Data Collection

Secondary data are obtained from Refinitiv Eikon for Thai-listed firms over 2016–2023. The period 2016–2019 is classified as pre-COVID-19 and 2020–2023 as post-COVID-19. The dataset contains the balance-sheet items needed to construct recognized intangible assets and tangibility, leverage measures based on both book and market values, and standard firm characteristics (size, profitability, liquidity, and growth). To avoid a moving-target sample across dependent variables, the baseline estimation is conducted on a main estimation sample with complete covariates, and dependent-variable-specific samples are used only when required for robustness measures.

Table 1. Variable definitions and measurement.

Variable	Variable Name	Role	Measurement
Market-based long-term leverage	LTD_MV	Main DV	Long-term debt/ Market value of assets
Market-based total leverage	D_MV	Robust DV	Total debt/market value of assets
Book-based long-term leverage	LTD_A	Robust DV	Long-term debt/total assets
Book-based total leverage	D_A	Robust DV	Total debt/total assets
Recognized intangible intensity	INT; Int_w; L1_Int_w	Key IV	(Recognized identifiable intangible assets−goodwill)/total assets
Post-COVID-19 interaction	INT_Post; L1_INT_Post	Key IV (interaction)	INT × post-COVID-19 indicator
Tangible assets intensity	L1_Tan	Key IV	PPE/total assets
Firm size	Size	Control (log)	ln(total assets)
Growth opportunities	Growth_w; L1_Growth_w	Control	Enterprise value/EBIT
Operating profitability	OPROA; OPROA_w; L1_OPROA_w	Control	Operating income/total assets
Liquidity	Liquidity_w; L1_Liquidity_w	Control	Quick ratio

summarizes variable definitions, roles and measurement.

3.4. Data Analysis

3.4.1. Sample Construction and Variable Treatment

To avoid a “moving-target sample” across specifications, the baseline analysis is estimated on a constant sample with complete regressors where firm identifiers, year, industry classification, and all lagged regressors are simultaneously observed. Alternative dependent variables (D_MV, D_A, LTD_A) are estimated on DV-available subsamples, while keeping the conditioning set on the right-hand side fixed. This design ensures that coefficient changes across specifications reflect modeling choices rather than compositional shifts in observations. The pre-/post-COVID-19 periods are defined as 2016–2019 and 2020–2023, respectively, and the interaction INT_(t − 1) × Post_t captures a slope shift after the shock.

Several accounting variables exhibit extreme right tails. This research applies a transparent rule: INT, growth, operating profitability (OPROA), and liquidity are winsorized at the 1st/99th percentiles by year (variables with suffix _w in STATA) to limit the undue influence of extreme realizations while preserving the panel structure. Leverage measures that are mechanically bounded in [0,1] in our sample (LTD_MV, D_MV, LTD_A) are retained in raw form. For the asset-based ratio D_A, which can exceed one under stress or denominator effects, we winsorize by year and verify robustness to alternative treatments (capping at one or excluding D_A > 1 observations). Results are reported in Appendix A 

Table A1. Variable diagnostics, theoretical bounds, and outlier treatment.

Variable	Role	Theoretical Bound/ Expected Range	Distribution Diagnostics (P1, P99, Max; Skew; Kurt)	Out-of-Range Check	Baseline Decision
LTD_MV	Main DV	[0,1]	P1 = 0.000; P99 ≈ 0.851; max = 1.000; skew ≈ 1.45; kurt ≈ 4.41	LTD_MV > 1:0	No winsorization
D_MV	Robust DV	[0,1]	P1 = 0.000; P99 ≈ 0.926; max = 1.000; skew ≈ 0.81; kurt ≈ 2.58	D_MV > 1:0	No winsorization
LTD_A	Robust DV	[0,1]	P1 = 0.000; P99 ≈ 0.548; max ≈ 0.937; skew ≈ 1.43; kurt ≈ 4.61	LTD_A > 1:0	No winsorization
D_A	Robust DV	Typically ≤1; may exceed 1 in distress	P99 ≈ 0.771; max ≈ 1.887; skew ≈ 0.75; kurt ≈ 3.55	D_A > 1:11	Winsorization + robustness
INT	Key IV	Ratio; potentially skewed	P99 ≈ 0.402; max ≈ 0.728; skew ≈ 5.52; kurt ≈ 40.00; includes small negatives	—	Winsorization
TAN	Key IV	[0,1]	P1 ≈ 0.0036; P99 ≈ 0.891; max ≈ 0.992; skew ≈ 0.44; kurt ≈ 2.20	—	No winsorization
SIZE	Control (log)	Log scale (unbounded)	min ≈ 15.64; max ≈ 29.14; skew ≈ 0.82; kurt ≈ 4.00	—	No winsorization
GROWTH	Control	Unbounded; can be extreme	P99 ≈ 597; max ≈ 44,052; min ≈ −73.6; skew ≈ 43.23; kurt ≈ 1958	—	Winsorization
OPROA	Control	Can be extreme	P99 ≈ 17.67; max ≈ 651; min ≈ −37.1; skew ≈ 30.57; kurt ≈ 1221	—	Winsorization
LIQUIDITY	Control	≥0 in theory; may contain data noise	P99 ≈ 11.34; max ≈ 66.26; min ≈ −0.117; skew ≈ 8.78; kurt ≈ 150.5	—	Winsorization (+optional check)

Notes: P1 and P99 denote the 1st and 99th percentiles. Winsorization is performed by fiscal year to prevent a small number of extreme observations from a single year dominating estimates while maintaining comparability over time.

.

3.4.2. Baseline Estimation Strategy

Because leverage may reflect unobserved firm-specific traits that are plausibly correlated with asset composition, pooled OLS and random-effects estimators may be biased if the orthogonality assumption fails. We therefore conduct standard panel model-selection diagnostics by estimating (1) a pooled OLS benchmark, (2) a random-effects model, and (3) a fixed-effects model, followed by a Hausman specification test comparing fixed versus random effects. The Hausman test rejects the random-effects orthogonality condition, indicating that the fixed-effects estimator is preferred.

Consistent with this result, the empirical design proceeds in stages. The baseline analysis begins with firm fixed-effects estimation, which isolates within-firm variation in leverage associated with changes in recognized intangible intensity while absorbing time-invariant firm heterogeneity. We then evaluate robustness under progressively stricter fixed-effects structures that additionally absorb year fixed effects and, in the preferred strict specification, industry-by-year fixed effects. This progression is intended to assess whether the estimated INT–leverage relation persists after removing increasingly rich forms of common time-varying heterogeneity. The preferred strict specification absorbs industry-by-year fixed effects to control for shocks common to firms operating in the same industry-year, such as sector-specific credit conditions, demand shifts, and policy exposure. This is particularly important because the contracting relevance of recognized intangibles may vary across sectors depending on differences in collateral dependence, tangibility, and creditor screening intensity. By absorbing industry-by-year fixed effects, the model does not estimate separate industry effects; rather, it identifies the INT coefficient from within-firm variation, net of sector-year financing conditions. Standard errors are clustered at the firm level throughout to ensure valid inference under heteroskedasticity and within-firm serial correlation.

3.4.3. Dynamic Robustness: System GMM

Because leverage is persistent and financing decisions may adjust dynamically, we additionally estimate a System GMM specification as a robustness check. This approach is appropriate because the inclusion of lagged leverage introduces dynamic panel bias in fixed-effects estimation, while the relation between leverage and recognized intangible intensity may also be affected by reverse causality and omitted time-varying firm characteristics. In the dynamic specification, lagged leverage is treated as endogenous, while recognized intangible intensity is treated as predetermined (i.e., potentially endogenous but not strictly exogenous). Identification relies on internal instruments constructed from lagged levels and lagged differences in the regressors under the standard System GMM framework. To avoid instrument proliferation, the instrument matrix is deliberately restricted using a parsimonious lag structure, and instrument counts are reported relative to the number of firms. We report the standard diagnostic tests—AR(1), AR(2), Hansen and Sargan tests, and the number of instruments—to assess whether the dynamic specification is statistically well-behaved and whether the internal-instrument strategy provides credible robustness to dynamic endogeneity concerns.

4. Results

4.1. Descriptive Statistics

Table 2. Descriptive statistics of key variables.

Variable	N	Mean	SD	P25	Median	P75	Min	Max
LTD_MV	5482	0.168	0.215	0.005	0.067	0.272	0	1
D_MV	7126	0.259	0.262	0.015	0.18	0.442	0	1
D_A	6869	0.245	0.217	0.036	0.211	0.408	0	1.887
LTD_A	6869	0.115	0.143	0.003	0.052	0.188	0	0.937
INT	5392	0.025	0.07	0.001	0.004	0.014	0	0.728
Tan	5475	0.324	0.246	0.092	0.298	0.514	0	0.992
Size	5480	22.235	1.861	20.936	21.99	23.231	15.637	29.138
Growth	3705	66.623	895.975	11.723	19.704	34.554	−73.59	44,052.02
OPROA	5005	1.218	13.672	−0.081	−0.03	0.201	−37.117	651.325
Liquidity	4336	1.735	2.637	0.507	0.985	1.968	−0.117	66.269

reports raw (pre-winsorization) descriptive statistics for the main variables used in the analysis. We present the raw distribution to transparently document the degree of skewness and tail behavior in accounting ratios, which is informative in settings where a small number of distressed observations or denominator effects can generate extreme realizations. However, in the regression analysis, we mitigate the influence of outliers using a consistent trimming rule described below and summarized in Appendix A Table A1.

Table 2 presents descriptive statistics for raw (pre-winsorization) variables. For regression analyses, INT, growth, OPROA, and liquidity are winsorized at the 1st/99th percentiles by year to limit the influence of extreme realizations while preserving the panel structure (variables with suffix _w). Each observed variable comprises sample size (N), mean, standard deviation, 25th percentile, median, 75th percentile, minimum, and maximum. Variables include LTD_MV, D_MV, D_A, LTD_A, INT, Tan, size, growth, OPROA, and liquidity. Observations per variable vary, reflecting an unbalanced panel. Means and dispersion measures indicate heterogeneity across variables (e.g., size mean ≈ 22.235, growth mean ≈ 66.623 with a large SD).

4.2. Correlation Analysis

This section presents the correlation analysis of the key variables, including recognized intangible intensity (Int), tangible asset intensity (Tan), leverage ratios (D_A, D_MV, LTD_A, LTD_MV), and the control variables (firm size, growth, profitability, and liquidity). It also considers the post-COVID-19 indicator and the interaction term used in the slope-shift specification. The correlation analysis provides descriptive evidence on pairwise associations before the multivariate regression analysis.

Table 3. Correlation analysis.

Variable	LTD_MV	D_MV	D_A	LTD_A	Int	Tan	Size	Growth	OPROA	Liquidity
Int	0.0685 **	0.0375 **	0.1084 **	0.1706 **	1.0000
Tan	0.1100 **	0.0928 **	0.1653 **	0.2274 **	−0.0939 **	1.0000
Size	0.4759 **	0.3853 **	0.2501 **	0.4210 **	0.0608 **	−0.0003	1.0000
Growth	0.0374 **	0.0271	0.0174	0.0214	0.0066	−0.0042	0.0042	1.0000
OPROA	−0.0031	−0.0038	−0.0178	−0.0141	−0.0100	−0.0077	0.0143	−0.0054	1.0000
Liquidity	−0.0415 **	−0.0699 **	−0.0893 **	−0.0410 **	0.0097	−0.0224	−0.0073	0.0191	−0.0219	1.0000

** Denotes significant level at 5%.

presents the Pearson correlation coefficients among key variables. The results illustrate a statistically significant positive correlation at the 5% level between intangible assets and the leverage ratios, as well as between tangible assets and the leverage ratios across all four leverage measurements. However, the correlation coefficients for tangible assets are consistently higher than those for intangible assets across all leverage measures, indicating that firms with a greater proportion of tangible assets rely more heavily on debt financing than those with higher intangible assets. The table also shows a positive correlation between firm size and liquidity at the 5% level of significance, and a statistically significant positive correlation at the same level between firm size and intangible assets, indicating that larger firms tend to possess higher levels of intangible assets. On the other hand, a negative correlation has been observed between liquidity and the leverage ratios, indicating that firms with higher liquidity generally use less debt financing.

Multicollinearity was assessed using Variance Inflation Factors (VIF) across the regression specifications; all VIF values were below 3, suggesting no serious multicollinearity concerns. Given the panel structure of the data and the potential for heteroskedasticity and within-firm serial correlation, all regressions are estimated using firm fixed effects with firm-clustered robust standard errors. This approach ensures consistent inference under general forms of heteroskedasticity and intra-firm dependence.

4.3. Baseline Fixed-Effect Regression Analysis

We begin by presenting baseline firm fixed-effects estimates that characterize the within-firm association between recognized intangible intensity and leverage. These specifications control for time-invariant firm heterogeneity but do not fully absorb year- or industry-level financing shocks. Accordingly, the estimates should be interpreted as baseline evidence of the INT–LEV relation prior to imposing stricter identification structures. Subsequent specifications incorporate additional fixed effects and robustness checks to assess whether the documented patterns persist after controlling for broader macro and sector-year financing conditions. This stepwise approach allows us to evaluate how the financing relevance of recognized intangibles evolves under increasingly stringent controls.

Table 4. Firm fixed-effects regressions of leverage on intangible asset intensity.

Variable	D_A	D_MV	LTD_A	LTD_MV
Constant	−2.2211 ***	−3.8107 ***	−1.5584 ***	−3.3036 ***
	(0.3057)	(0.3202)	(0.2439)	(0.320)
L1_Tan	0.3285 ***	0.2323 ***	0.2889 ***	0.2486 ***
	(0.0352)	(0.0337)	(0.0309)	(0.0382)
L1_Int_w	0.2892 **	0.3624 ***	0.2653 ***	0.4336 ***
	(0.1297)	(0.1012)	(0.0907)	(0.1218)
L1_Size	0.1037 ***	0.1768 ***	0.0697 ***	0.1496 ***
	(0.0134)	(0.014)	(0.0107)	(0.0139)
L1_Growth_w	−0.000	−0.000	−0.000 **	0.000 *
	(0.000)	(0.000)	(0.000)	(0.000)
L1_OPROA_w	−0.000	−0.000	−0.000 *	−0.000 **
	(0.000)	(0.000)	(0.000)	(0.000)
L1_Liquidity_w	−0.0014 *	−0.0000459	0.0001089	0.0009608
	(0.00082)	(0.000757)	(0.00058)	(0.000814)
Within R2	0.2152	0.4147	0.1885	0.2849
Overall R2	0.1174	0.1312	0.2608	0.2239
N (firm-years)	2933	2933	2933	2933
Firms	575	575	575	575

Note: Robust standard errors clustered at the firm level are reported in parentheses. *, **, and *** denote statistical significance at the 10%, 5%, and 1% levels, respectively.

reports the baseline firm fixed-effects estimates linking recognized intangible intensity to leverage. Under this baseline specification, both tangibility and recognized intangible intensity are positively associated with leverage across the reported measures, with the strongest coefficient appearing for long-term market-based leverage (LTD_MV). These estimates indicate that, within firms, increases in recorded intangible intensity are associated with higher leverage under the lighter fixed-effects structure. However, these baseline results should be interpreted cautiously, because they do not yet absorb broader year-level or industry-year financing conditions. Accordingly, Table 4 is best read as baseline within-firm evidence rather than as definitive confirmation of a robust debt-capacity effect.

4.4. Post-COVID-19 Slope Heterogeneity Within the Firm FE Framework

This section examines whether the baseline INT–leverage relation changes during the post-COVID-19 period within the same-firm fixed-effects framework.

While Table 4 documents the baseline within-firm association between recognized intangible intensity and leverage, it does not address whether this relation is stable across financing regimes. To examine whether the COVID-19 shock altered the slope of the INT–LEV relation,

Table 5. Fixed-Effects Regressions on Intangible Asset Intensity and Leverage in the Post–COVID–19 Period.

Variable	D_A	D_MV	LTD_A	LTD_MV
Constant	−2.2913 ***	−3.8007 ***	−1.5870 ***	−3.3538 ***
	(0.3989)	(0.3865)	(0.3174)	(0.3898)
L1_Tan	0.3263 ***	0.2307 ***	0.2882 ***	0.2469 ***
	(0.0349)	(0.0337)	(0.031)	(0.0383)
L1_Int_w	0.2258 *	0.3356 ***	0.2421 **	0.3857 ***
	(0.1245)	(0.1008)	(0.0962)	(0.1148)
L1_Size	0.1070 ***	0.1765 ***	0.0710 ***	0.1519 ***
	(0.000)	(0.000)	(0.000)	(0.000)
L1_Growth_w	0.0000	0.0000	0.0000 **	0.0000 *
	(0.000)	(0.000)	(0.000)	(0.000)
L1_OPROA_w	0.0000	0.0000	−0.0000 *	−0.0000 **
	(0.000)	(0.000)	(0.000)	(0.000)
L1_Liquidity_w	−0.0012	0.0000	0.0002	0.0011
	(0.00082)	(0.000775)	(0.000597)	(0.000853)
PostCOVID	−0.0096	−0.0024	−0.0036	−0.0072
	(0.00672)	(0.00575)	(0.00547)	(0.00656)
L1_INT_Post	0.1988 ***	0.0946 **	0.0717	0.1508
	(0.0663)	(0.0477)	(0.0705)	(0.1092)
Within R2	0.2223	0.4159	0.19	0.2883
Overall R2	0.1182	0.1313	0.2613	0.2243
N (firm-years)	2933	2933	2933	2933
Firms	575	575	575	575

Note: Robust standard errors clustered at the firm level are reported in parentheses. *, **, and *** denote statistical significance at the 10%, 5%, and 1% levels, respectively.

augments the baseline specification with a post-period indicator and an interaction term (INT × PostCOVID). This specification does not introduce a new identification layer but instead allows us to test whether the financing relevance of recognized intangibles changes under crisis-era credit conditions within the same firm fixed-effects framework. Accordingly, Table 5 should be interpreted as a structured extension of Table 4 that evaluates slope heterogeneity rather than as an alternative baseline model.

The empirical estimates from this extended specification are presented in Table 5. Recognized intangible intensity remains positively associated with leverage in the pre-COVID-19 period within the baseline firm fixed-effects specification. Moreover, the interaction term INT × PostCOVID is positive and statistically significant for total leverage measures (D_A and D_MV), indicating that the slope of the INT–LEV relation becomes more positive during 2020–2023 under this lighter control structure. These results are consistent with a setting in which recognized identifiable intangibles exhibit financing relevance in total leverage adjustments during the pandemic period. However, the absence of a statistically significant interaction for long-term debt measures suggests that this strengthening effect is maturity-dependent rather than universal. The differences across leverage measures are economically informative. First, the stronger evidence for market-value leverage relative to book-value leverage is consistent with the pandemic period being characterized by heightened valuation sensitivity and risk repricing, such that financing decisions and balance-sheet adjustments are more clearly reflected when leverage is scaled by market-valued assets. Second, the fact that the interaction term is significant for total leverage (D_A and D_MV) but not for long-term debt ratios (LTD_A and LTD_MV) suggests a maturity-specific response in a bank-centered system. During periods of elevated uncertainty, banks may be more willing to accommodate short-maturity borrowing tied to liquidity management (e.g., revolving credit and working-capital facilities) than to expand long-duration commitments, especially when recovery values associated with intangible-intensive borrowers are more difficult to verify. Thus, these patterns imply that the post-COVID-19 strengthening in the INT–LEV relation, where present, is contingent on leverage measurement and debt maturity, reflecting institutional boundary conditions rather than a uniform shift in capital structure. Although Table 5 provides evidence on post-COVID-19 slope heterogeneity within the baseline firm fixed-effects framework, the estimates do not yet absorb broader time-varying sectoral shocks. We therefore next examine whether the documented patterns persist under stricter fixed-effects structures.

4.5. Robustness: INT–Leverage Under Stricter Fixed Effects and the Post-COVID-19 Slope Test

The specifications reported thus far establish the within-firm INT–leverage relation under the baseline fixed-effects framework and, in Table 5, its potential post-COVID-19 slope variation. However, these estimates may still reflect residual common shocks that vary over time across sectors. To strengthen identification, we next re-estimate the INT–leverage relation under progressively stricter fixed-effects structures, first absorbing firm and year fixed effects, and then adopting a preferred specification that absorbs firm and industry-by-year fixed effects. This progression allows us to assess whether the baseline patterns remain robust after controlling for increasingly rich forms of common time-varying heterogeneity.

Table 6. Baseline INT–LEV and post-COVID-19 slope shift.

Variables	(1) Firm FE + Year FE	(2) Firm FE + Ind × Year FE (Preferred)	(3) Firm FE + Year FE	(4) Firm FE + Ind × Year FE (Preferred)
L1_Int_w	0.1658	0.1694	0.1402	0.1264
	(0.1427)	(0.1438)	(0.1579)	(0.1620)
L1_Int_w × Post			0.0468	0.0791
			(0.1174)	(0.1245)
L1_Tan	0.1686 ***	0.1726 ***	0.1684 ***	0.1724 ***
	(0.0430)	(0.0432)	(0.0429)	(0.0432)
L1_Size	0.1025 ***	0.1026 ***	0.1021 ***	0.1018 ***
	(0.0163)	(0.0168)	(0.0165)	(0.0170)
L1_Growth_w	0.00000409	0.00000459	0.00000402	0.00000463
	(0.0000470)	(0.0000481)	(0.0000471)	(0.0000482)
L1_OPROA_w	0.001316	0.001172	0.001288	0.001115
	(0.001207)	(0.001177)	(0.001214)	(0.001183)
L1_Liquidity_w	0.001161	0.001247	0.001163	0.001249
	(0.001565)	(0.001647)	(0.001567)	(0.001650)
N (observations)	2241	2241	2241	2241
Clusters (firms)	458	458	458	458
Within R2	0.0951	0.0969	0.0953	0.0975

Notes: Robust standard errors in parentheses. *** denote statistical significance at the 1% levels.

reports the core estimates linking recognized intangible asset intensity (INT) to long-term leverage (LTD_MV) under increasingly stringent fixed effects. Across both the firm-and-year specification and the preferred firm-and-industry × year specification, the INT coefficient is positive in sign but statistically indistinguishable from zero. Specifically, INT is 0.1658 (SE = 0.1427) under firm FE + year FE and 0.1694 (SE = 0.1438) under firm FE + industry × year FE, implying that once time-invariant firm heterogeneity and common (or industry-specific) macro shocks are absorbed, balance-sheet recognized intangibles do not exhibit a systematic average association with leverage in Thailand.

The trade-off/collateral channel predicts that contractible, identifiable intangibles can expand debt capacity, while pecking-order and low-pledgeability mechanisms predict lower reliance on debt when intangibles exacerbate valuation and information frictions. In a relationship-oriented lending environment, these forces can plausibly offset in the cross-section, attenuating the average slope. Consistent with this interpretation, the coefficient remains small relative to its uncertainty and does not stabilize into a directional effect even under the preferred fixed-effect design that removes industry-year shocks.

Table 6 also evaluates whether the INT–leverage slope shifts in the post-COVID-19 period by adding the interaction term INT × Post. The estimated interaction is again statistically weak under both FE structures: 0.0468 (SE = 0.1174) with firm FE + year FE and 0.0791 (SE = 0.1245) with firm FE + industry × year FE. This suggests that, conditional on stringent absorption of industry-year conditions, the post-COVID-19 regime change does not translate into a systematic change in how recognized intangibles map into long-term leverage in Thailand. Put differently, once sector-year financing conditions and credit supply shifts are netted out, there is limited evidence that the pandemic period altered the marginal leverage implications of recognized intangible intensity.

Importantly, the core asset-composition regressors behave in a theoretically coherent way and provide an internal validity check on the specification. Tangibility (Tan) is consistently positive and highly significant (e.g., 0.1686–0.1726 across specifications), aligning with collateral-based debt capacity. Firm size is also strongly positive (approximately 0.1018–0.1026), consistent with lower default risk and better access to formal credit markets. In contrast, growth, profitability, and liquidity are not robustly associated with long-term leverage once fixed effects and industry-year shocks are absorbed, which is plausible in settings where leverage is shaped more by stable firm traits and sector-year financing conditions than by short-horizon performance variation.

4.6. Robustness: Alternative Leverage Proxies (Market- and Book-Based Leverage)

Having evaluated the sensitivity of the baseline findings to stricter identification controls, we next examine whether the results are robust to alternative leverage definitions based on book- and market-value scaling.

Table 7. Robustness to leverage measurement (alternative dependent variables).

Variables	(1) LTD_MV	(2) LTD_A	(3) D_MV	(4) D_A
L1_Int_w	0.1250	0.0844	0.1856	0.1105
	(0.1625)	(0.1132)	(0.2010)	(0.2022)
L1_INT_Post	0.0771	0.0949	0.1566	0.2478
	(0.1243)	(0.1325)	(0.2095)	(0.2180)
L1_Tan	0.1728 ***	0.0404	−0.0038	0.0139
	(0.0433)	(0.0348)	(0.0398)	(0.0506)
L1_Size	0.1023 ***	0.0027	0.0048	0.0086
	(0.0171)	(0.0102)	(0.0148)	(0.0168)
L1_Growth_w	0.00000481	0.0000329	0.0000359	0.0000316
	(0.0000483)	(0.0000345)	(0.0000356)	(0.0000441)
L1_OPROA_w	0.0013	−0.0007	−0.00000435	0.0006
	(0.0012)	(0.0012)	(0.0012)	(0.0033)
L1_Liquidity_w	0.0014	0.0010	0.0021	0.00000905
	(0.0017)	(0.0019)	(0.0031)	(0.0028)
Constant	−2.1980 ***	0.0283	0.1497	0.0253
	(0.3899)	(0.2338)	(0.3373)	(0.3870)
Observations	2196	2196	2196	2196
Firms (clusters)	448	448	448	448
Fixed effects	Firm + Industry * Year	Firm + Industry * Year	Firm + Industry * Year	Firm + Industry * Year
SE clustered by firm	Yes	Yes	Yes	Yes

Note: Robust standard errors in parentheses. *** denotes statistical significance at the 1% levels.

reports robustness checks across four alternative leverage-dependent variables for a common-support sample (the intersection of firms observed on all proxies). Columns (1)–(4) correspond to LTD_MV (long-term debt to market value), D_MV (debt to market value), LTD_A (long-term debt to assets), and D_A (debt to assets). Regressors include lagged interaction terms (L1_Int_w, L1_INT_Post), and controls such as lagged tangibility (L1_Tan), firm size (L1_Size), growth, operating profit (L1_OPROA_w), and liquidity. All specifications employ firm fixed effects and industry-by-year fixed effects, with stand-ard errors clustered by firm. The sample comprises 2196 observations from 448 firms.

The table presents key findings: L1_Tan is positive and highly significant for LTD_MV (column 1), indicating that tangible asset intensity increases market-based long-term leverage. L1_Size is likewise positive and significant in column 1, suggesting larger firms hold higher long-term market leverage. Most coefficients on the intangible interaction terms (L1_Int_w, L1_INT_Post) are small and statistically insignificant across models, implying limited or inconsistent effects of intangible intensity on leverage when using alternative measures. Constants, observation counts, and clustering are reported consistently.

4.7. Additional Robustness and Identification Tests

To further assess whether the documented post-COVID-19 slope shift reflects a genuine regime-dependent effect rather than a spurious time trend, we conduct a timing placebo test within the pre-COVID-19 subsample. Specifically, we restrict the sample to the period 2016–2019 and define a pseudo-post indicator that takes on a value of 1 starting in 2018. We then re-estimate the preferred specification with firm and industry-by-year fixed effects, interacting recognized intangible intensity with the pseudo post indicator. If the observed post-COVID-19 interaction were driven by generic pre-existing trends, a similar slope change would be detected in the placebo period. However, as reported in Appendix A 

Table A2. Timing placebo test (pre-COVID-19 pseudo shock), dependent variable: LTD_MV (market-value long-term debt ratio).

Variables	(A2-1) LTD_MV
L1_Int_w	0.2351 (0.3885)
L1_INT_pseudo (INT × PseudoPost)	0.0603 (0.0794)
Controls	Yes
Firm FE	Yes
Industry × Year FE	Yes
Std. errors	Clustered by firm
Observations	791
Firms (clusters)	290
Within R2	0.0306

Notes: Robust standard errors clustered at the firm level are reported in parentheses. The placebo sample is restricted to 2016–2019 with PseudoPost = 1 for 2018–2019 and 0 for 2016–2017. The key placebo coefficient is INT × PseudoPost.

, the interaction term (INT × PseudoPost) is statistically insignificant for the main leverage measure (LTD_MV), suggesting that the interaction observed under baseline specification is unlikely to reflect spurious pre-trends.

4.8. Dynamic Robustness and Endogeneity: System GMM

A key concern in the intangible leverage setting is endogeneity. Recognized intangible intensity may be jointly determined with financing choices through omitted time-varying firm conditions (e.g., shifts in risk, growth opportunities, or collateral quality), and leverage is inherently dynamic given adjustment frictions and persistence. To address these issues, we complement the high-dimensional fixed-effects evidence with a dynamic panel System GMM specification that explicitly includes lagged leverage and employs internal instruments to mitigate concerns about simultaneity and reverse causality.

We estimate a standard dynamic leverage model in which current leverage depends on its own lag and lagged covariates, and implement System GMM with a conservative instrument strategy to avoid instrument proliferation. Specifically, we use collapsed GMM-style instruments and restrict the lag window to 2–3 for the endogenous/predetermined blocks, while treating the accounting controls and year dummies as standard instruments in the level’s equation. We report both one-step and two-step (Windmeijer-corrected) estimates with robust standard errors and the small-sample correction, and maintain the same baseline sample used in the main analysis.

Table 8. System GMM (dynamic leverage model).

	(1) Two-Step System GMM	(2) One-Step System GMM
L1_LTD_MV	0.7788 ***	0.7353 ***
	(0.0733)	(0.0778)
L1_Int_w	−0.1578	0.0641
	(0.1642)	(0.1968)
L1_INT_Post	0.2018	0.0216
	(0.1467)	(0.1693)
L1_Tan	0.0100	0.0106
	(0.0134)	(0.0129)
L1_Size	0.0121 ***	0.0141 ***
	(0.0043)	(0.0044)
L1_Growth_w	0.00000958	0.00000755
	(0.0000298)	(0.0000319)
L1_OPROA_w	−0.000603	−0.000767
	(0.000848)	(0.000858)
L1_Liquidity_w	−0.000301	−0.000523
	(0.001232)	(0.001237)
Constant	−0.2391 ***	−0.2764 ***
	(0.0874)	(0.0887)
Year effects	YES	YES
Observations	2241	2241
Firms (# groups)	458	458
Instruments	21	21
AR(1) p-value	0.000	0.000
AR(2) p-value	0.192	0.177
Hansen p-value	0.586	0.586
Sargan p-value (not robust)	0.000	0.000
Diff-in-Hansen (levels) p-value	0.410	0.410
Diff-in-Hansen (GMM: LTD_MV) p-value	0.556	0.556

Notes: Robust standard errors in parentheses. *** p < 0.01. AR tests are Arellano–Bond tests in first differences.

shows that leverage is strongly persistent, consistent with partial adjustment: the coefficient on lagged leverage is large and precisely estimated (two-step: L1(LEV) = 0.7788, p < 0.01). In contrast, the coefficients on recognized intangible intensity and its post-period interaction remain statistically indistinguishable from zero (two-step: L1_Int_w = −0.1578, p = 0.337; L1_INT_Post(0.2018, p = 0.170)), corroborating the fixed-effects conclusion that the average INT–LEV relation is weak once confounding is tightly controlled. The diagnostic tests support the specification: AR(1) is significant as expected, while AR(2) is not (two-step p = 0.192), and the Hansen test does not reject the overidentifying restrictions (p = 0.586). Instrument count is kept modest (21 instruments for 458 firms), consistent with a “tight” GMM design intended to preserve test power and reduce finite-sample bias. Overall, the dynamic panel evidence indicates that the baseline inference is not an artifact of leverage persistence or simple reverse causality, and that the post-COVID-19 interaction does not exhibit a robust slope shift in the dynamic setting. Overall, the System GMM results reinforce the main fixed-effects evidence. Leverage remains strongly persistent, while the coefficients on recognized intangible intensity and its post-period interaction remain statistically weak once dynamic adjustment and internal-instrument identification are taken into account. These results suggest that the baseline positive association observed under lighter controls is not robust to stricter dynamic specifications. The broader identification-based and institutional interpretation of this attenuation is discussed in Section 5.

5. Discussion

The comparison between the baseline and stricter specifications is central to interpreting the INT–leverage relation. In the baseline firm fixed-effects models, recognized intangible intensity is positively associated with leverage, and the post-COVID-19 interaction is also significant under lighter controls. However, once year fixed effects and, more stringently, industry-by-year fixed effects are absorbed, both the baseline INT coefficient and the interaction term lose statistical significance. This attenuation is economically informative rather than simply negative. It indicates that the earlier positive association is at least partly embedded in broader macro-financial and sector-specific financing conditions, rather than reflecting a strong autonomous debt-capacity effect of recognized intangibles. The same pattern is reinforced in the dynamic specification, where leverage remains highly persistent while the coefficients on recognized intangibles and their post-period interaction remain statistically weak.

Relative to the prior literature, these findings suggest a more conditional interpretation of the relevance of recognized intangibles to leverage. This interpretation is also consistent with emerging-market evidence emphasizing relationship lending, conservative creditor screening, and weaker reliance on intangible collateral, which together suggest that the leverage relevance of recognized intangibles may be more limited outside stronger enforcement and more market-based financial systems (Shehadeh et al. 2024). Some developed-market studies report that identifiable or creditor-relevant intangible assets can support borrowing when valuation and enforcement are sufficiently strong, particularly when such assets are more contractible and more readily recognized by lenders (Lim et al. 2020). The present evidence does not directly reject that possibility; rather, it qualifies it. In lighter specifications, the Thai data are directionally consistent with a debt-capacity interpretation. Yet once common year-level and industry-year financing conditions are more tightly controlled, the estimated effect is substantially attenuated. This implies that the leverage relevance of recognized intangibles is highly sensitive to institutional setting, identification strategy, and the surrounding financing regime. In this sense, the contribution of the present study is not to claim that recognized intangibles are uniformly debt-supportive or uniformly debt-reducing, but instead to show that their apparent financing relevance depends on the contractual and credit environment in which firms borrow.

This interpretation is especially plausible in Thailand’s bank-dominated financial system. Even when intangible assets are sufficiently measurable to be recognized under IAS 38, lenders may still place greater weight on hard collateral, expected recoverability, sector-specific risk, and relationship-based screening than on the standalone stock of recorded intangibles. As a result, recognized intangibles may co-move with leverage under lighter specifications, particularly when common financing conditions are favorable, but they do not appear to exert a strong independent effect on debt capacity once broader financing regimes are properly accounted for. The same reasoning helps explain the weak post-COVID-19 interaction under stricter models: although the pandemic clearly altered financing conditions, the apparent slope shift in lighter specifications is largely absorbed by common year-level and sector-year shocks rather than emerging as a robust firm-specific change in how recognized intangibles map into leverage. Together, the results support a boundary-condition interpretation in which the financing relevance of recognized intangibles is conditional, limited, and institution-dependent. More broadly, the findings suggest that accounting recognition alone should not be assumed to imply creditor relevance. For recognized intangibles to function as a meaningful driver of debt capacity, the surrounding institutional environment must also support their contractibility, screening relevance, and expected recoverability. In this sense, the stricter fixed-effects and dynamic results do not merely weaken the baseline findings; they clarify that much of the apparent INT–leverage association is absorbed by the broader financing regime in which firms operate.

6. Conclusions and Limitation

This study examines whether IAS 38-recognized identifiable intangible assets (excluding goodwill) are associated with corporate leverage and whether that relation shifts after the COVID-19 shock in Thailand, an emerging, bank-dominated financial system. Using a stepwise empirical design, the analysis begins with baseline firm fixed-effects models, then evaluates robustness under stricter fixed-effects structures, and finally assesses dynamic robustness using System GMM. The results show that recognized intangible intensity is positively associated with leverage in baseline firm fixed-effects specifications, and that the post-COVID-19 interaction is also significant under lighter controls. However, both effects lose statistical significance once year fixed effects and, more stringently, industry-by-year fixed effects are absorbed. The dynamic results point in the same direction: leverage is strongly persistent, while the coefficients on recognized intangible intensity and its post-period interaction remain statistically weak. These findings suggest that the baseline positive association and the apparent post-COVID-19 slope shift are substantially absorbed by broader macro-financial and sector-specific financing conditions, rather than reflecting a strong independent debt-capacity effect of recognized intangibles.

The main implication is that the relevance of leverage for recognized intangibles is best understood as a boundary-condition question rather than a universal effect. In the Thai setting, balance-sheet recognition alone does not appear sufficient to render intangible assets strongly debt-supportive once creditor screening, collateral dependence, and sector-year financing conditions are properly controlled for. This study, therefore, contributes by sharpening construct clarity, strengthening identification through progressively stricter specifications, and showing that the financing relevance of recognized intangibles is conditional on the surrounding institutional and credit environment. Several limitations remain. The analysis focuses on recorded intangible assets rather than broader unrecognized intangible resources. In addition, the evidence is drawn from a single institutional setting. Future research could extend this framework by incorporating alternative proxies for unrecognized intangible investment and comparing the leverage relevance of recognized intangibles across financial systems with different creditor enforcement and lending structures. Overall, the evidence suggests that recognized intangibles may co-move with leverage under lighter controls, but their standalone financing relevance is substantially attenuated once broader, time-varying financing conditions are accounted for. The contribution of the paper therefore lies not in claiming that intangibles uniformly increase or decrease leverage, but in clarifying the institutional conditions under which recognized intangible assets become more—or less—relevant to debt capacity.