Financial Sustainability in the Maritime Industry: Sub-Sectoral Evidence from an Emerging Economy

Abstract

This study examines the determinants of financial sustainability in Turkish maritime industry by analyzing firm-level panel data from 190 ship and boat maintenance firms and 208 coastal shipping companies for the 2010–2022 period, comprising 5174 firm-year observations. Fixed-effects models with Driscoll–Kraay robust standard errors are employed to evaluate how asset structure, liquidity, and energy efficiency jointly affect firm profitability across subsectors, using the Operating Return on Assets (OROA) as the principal indicator of operational performance. The empirical results indicate substantial heterogeneity between maintenance and shipping firms. For maintenance firms, OROA shows a positive association with the Non-Current Assets to Total Assets ratio (NCATA) and the Economic Efficiency Ratio (EER) but a negative association with the Current Ratio (CR), suggesting that capital deepening and operational efficiency tend to correlate with stronger performance, whereas excess liquidity is associated with weaker outcomes. For shipping firms, OROA is positively associated with EER and Total Asset Turnover (TATR) but negatively associated with Fixed Asset Turnover (FATR) and CR, indicating relationships consistent with efficiency gains from energy management and asset utilization but linkages suggesting challenges from fleet aging and liquidity mismanagement. Overall, the findings suggest that the drivers of financial sustainability are associated with different structural conditions across maritime subsectors, highlighting the importance of targeted modernization, port efficiency, and energy-transition investment strategies.

1. Introduction

Financial sustainability pressures increasingly challenge the maritime sector’s long-term resilience, despite its pivotal role in global trade and economic growth. High capital intensity, volatile freight revenues, and tightening environmental regulations compel maritime companies to balance operational efficiency with sound financial management [1,2]. Beyond the firm level, these challenges bear national policy relevance in emerging economies such as Türkiye, where restricted access to long-term financing and persistent macroeconomic volatility amplify sectoral vulnerabilities [3]. Strengthening financial sustainability in maritime firms has thus become both a strategic business imperative and a policy priority, particularly in light of the sector’s central role in financing the green transition.

Despite this importance, the existing literature often treats the maritime industry as a homogeneous entity, overlooking the structural and financial heterogeneity that characterizes its subsectors. Ship and boat maintenance companies face capital constraints related to infrastructure investment and irregular cash flows, whereas shipping firms confront fleet renewal demands, asset utilization challenges, and regulatory adaptation costs [2,4]. Uniform analytical and policy approaches run the risk of obscuring these differences, resulting in frameworks that fail to capture the subsector-specific determinants of financial sustainability. Moreover, financial sustainability in maritime industries is inherently linked to their exposure to energy markets and environmental regulation. Fuel expenses typically account for 40–60% of voyage costs, making bunker price volatility a direct threat to profitability. Regulatory shocks such as the International Maritime Organization’s (IMO) 2020 sulfur cap and the forthcoming decarbonization pathways require substantial investment in cleaner technologies, including scrubbers, alternative fuels, and fuel-efficient vessels. In emerging economies like Türkiye, these developments heighten financing pressures: constrained access to long-term capital complicates fleet modernization and the adoption of green technologies. By explicitly examining how firm-level financial variables interact with capital-intensive energy-transition requirements, the present study contributes to the intersection of corporate finance, energy economics, and sustainability regulation.

Turkish maritime economy is indicative of these challenges. Maintenance firms frequently function within fragmented industrial structures and encounter working-capital inefficiencies, while shipping companies grapple with aging fleets and constrained operational scalability. Across both subsectors, restricted access to affordable credit and heightened exposure to exchange rates and inflationary shocks undermine competitiveness. Addressing these obstacles necessitates the implementation of diversified financial strategies, encompassing asset management, liquidity optimization, and value creation. These strategies must also serve to inform financial and regulatory frameworks. Therefore, a complex understanding of firm-level financial behavior is essential for designing policy instruments that enhance resilience without imposing uniform standards across heterogeneous firms.

Given this context, the principal research question can be articulated as follows: the financial sustainability of Turkish maritime industry is shaped by multiple factors, including high capital intensity, volatile input and freight markets, and increasingly stringent environmental regulations. Yet the maintenance and shipping subsectors experience distinct constraints related to asset composition, utilization, and liquidity. Recognizing these structural differences is essential, as treating the industry as a homogeneous whole risks misidentifying the underlying determinants of operational performance and generating overly generic policy prescriptions. At the same time, firm-level empirical evidence for Türkiye remains limited due to restricted access to official microdata, leaving a significant gap in understanding how asset structure, utilization, liquidity, and financing jointly influence operational profitability—measured by Operating Return on Assets (OROA)—under conditions of macroeconomic volatility.

To address this gap, this study conducts a firm-level, subsector-comparative analysis of financial sustainability in Turkish maritime industry using official microdata from the Türkiye Statistical Institute (TUIK). The dataset comprises 190 ship and boat maintenance firms and 208 shipping companies observed over the 2010–2022 period, yielding 5174 firm-year observations. This extensive coverage enables a detailed examination of financial performance heterogeneity across subsectors. The empirical analysis applies fixed-effects regression models with Driscoll–Kraay robust standard errors to account for heteroskedasticity, serial correlation, and cross-sectional dependence—issues commonly encountered in maritime industry datasets. This specification captures within-firm variation over time and produces consistent estimates of how key financial ratios affect operational profitability.

Building on the reviewed literature, this study hypothesizes that firm-level indicators of asset efficiency, liquidity management, and capital structure systematically influence Operating Return on Assets (OROA) across maritime subsectors. We expect enhanced asset utilization and effective cash-flow management to improve financial performance, while excessive liquidity or leverage may diminish profitability. The fixed-effects models described in Section 3.3 empirically test these hypotheses, which are summarized in Appendix A. The analysis further distinguishes between maintenance and shipping firms to examine whether financial drivers operate differently across asset-intensive and service-oriented subsectors, thereby accounting for observed variations in profitability and resilience.

The contribution of this study is twofold. First, it advances academic debates by demonstrating that financial sustainability in maritime industries is not uniform but shaped by subsector-specific dynamics of asset allocation, liquidity, and efficiency. Second, it offers direct implications for policymakers, financial institutions, and regulators by showing that differentiated financial management strategies—rather than uniform frameworks—are essential to enhance resilience, competitiveness, and sustainable growth in Turkish maritime economy. In doing so, this study provides one of the first firm-level, subsector-comparative empirical assessments of maritime financial sustainability in an emerging-economy context, bridging theoretical insights with policy relevance.

2. Literature Review

The academic discourse on corporate sustainability spans environmental, social, and governance (ESG) dimensions, yet financial sustainability remains a decisive factor in ensuring long-term viability, particularly in capital-intensive industries such as maritime transport. While ESG practices have received growing scholarly attention, the financial dimension provides the empirical foundation for this study. Specifically, financial ratios serve as rigorous instruments for evaluating profitability, liquidity, leverage, and efficiency, thereby offering a reliable framework to assess resilience in volatile and capital-intensive markets.

2.1. Financial Ratios as Determinants of Corporate Sustainability

Financial ratios are widely recognized as core indicators of corporate financial health and long-term sustainability, reflecting a firm’s capacity to generate stable cash flows, manage risk, and absorb external shocks. Consequently, they find extensive application in both academic research and managerial practice. Prior studies, however, demonstrate that the explanatory power of these ratios is context dependent. Kliestik et al. [5] identified profitability, activity, liquidity, and leverage ratios as key predictors of financial strengths and vulnerabilities. Similarly, Siew et al. [6] found a positive association between sustainability reporting and profitability/liquidity among Australian construction firms, while noting the absence of a significant relationship with ESG scores. Together, these findings underscore the complementary role of quantitative financial metrics alongside qualitative sustainability assessments.

The maritime transport economics literature provides a comprehensive framework for analyzing the interaction between financial ratios, market dynamics, and sustainability performance. Jacks and Stuermer [7] traced the historical evolution of dry-bulk transportation expenses from 1850 to 2020, showing that fluctuations in fuel prices and demand-supply shocks exert a decisive influence on freight rates. Extending this line of inquiry, Adland and Cullinane [8] demonstrated that spot freight rates in the tanker market follow a nonlinear and partially mean-reverting process, indicating that volatility intensifies during high-rate periods. This stochastic behavior implies that maritime firms’ revenues and profitability are inherently exposed to cyclical fluctuations, underscoring the need to evaluate financial sustainability using performance indicators responsive to market variability. Similarly, Lim et al. [9] analyzed implied volatility in forward freight agreements (FFA) for Capesize and Panamax vessels, showing that expectations of economic growth and uncertainty shape risk premiums. These findings establish the FFA market as a leading indicator of investor sentiment and risk perception within global maritime trade.

Examining the structural sources of freight rate volatility, ref. [10] analyzed the dry-bulk shipping market using GARCH and EGARCH models and found that market shocks exhibit asymmetric effects, with ship segments such as Capesize, Panamax, and Handysize displaying varying sensitivities to volatility. The results indicate that fluctuations in global trade volumes heighten the short-term instability of freight prices. Extending this literature to the container shipping context, ref. [11] investigated volatility and leverage effects using data from the Shanghai Containerized Freight Index (SCFI). Employing an EGARCH (1,1,1) specification, the study identified strong volatility persistence (approximately 99%) and negative leverage effects, implying that negative market news triggers greater volatility responses than positive news. These findings suggest that the container freight market tends to overreact to adverse global shocks, producing prolonged volatility cycles. Collectively, this evidence provides invaluable information for risk management, pricing strategies, and financial stability policy design within the global container shipping industry.

In the context of sustainable financial performance, ref. [12] evaluated the impact of fuel choice on long-term financial sustainability by comparing the return on investment and emission reduction costs of alternative marine fuels such as LNG, methanol, and ammonia. The findings indicate that fuel selection is decisive not only for environmental outcomes but also for economic viability, highlighting the need to balance capital expenditure and investment efficiency during the green transformation process. Complementing this financial perspective, Chen et al. [13] developed a Life Cycle Cost-Effectiveness Analysis (LCEA) model to assess the economic and environmental feasibility of green alternative fuels in maritime transportation. By comparing methanol and LNG on a life-cycle basis, they found that methanol yields higher short-term returns, whereas LNG provides a more balanced outcome in long-term environmental sustainability. This framework integrates financial sustainability with environmental performance by considering fuel choice through the lenses of operational cost, social cost, energy efficiency, and regulatory compliance. Furthermore, studies on environmental upgrading in global maritime value chains reinforce this perspective. Poulsen et al. [14], for instance, emphasize that ports and shipping companies encounter institutional, financial, and governance constraints during the greening process, providing theoretical foundation for viewing financial sustainability as a function of adaptability to the energy transition.

Evidence from emerging markets also points to sectoral variation. Santis et al. [15] compared companies listed in Brazil’s Corporate Sustainability Index with those in the São Paulo Stock Exchange Index, concluding that sectoral affiliation outweighed sustainability investments in explaining profitability and liquidity patterns.

In summary, financial ratios are dynamic, context-sensitive measures that enrich the broader sustainability discourse. Their interpretation must reflect industry-specific structures, institutional conditions, and market environments. Building on this literature, the next section narrows the focus to maritime transport, where empirical studies, especially in emerging markets, remain limited.

2.2. Operating Return on Assets (OROA) as an Indicator of Sustainable Financial Performance

Financial sustainability refers to a company’s ability to maintain stable operations, preserve capital integrity, and generate long-term value in the face of economic volatility. Within this framework, the Operating Return on Assets (OROA)—defined as EBIT over total assets—provides a robust measure of operational efficiency independent of capital structure. By excluding financing and taxation effects, OROA enables consistent comparisons across companies [16].

Prior studies confirm OROA’s validity as a performance metric. Strouhal et al. [17] emphasized that EBIT-based ROA improves comparability across firms with heterogeneous capital structures.

OROA has also been shown to capture structural and contextual nuances. Kaen and Baumann [18] demonstrated a nonlinear relationship between company size and EBIT/total assets in U.S. manufacturing firms. Kannadhasan [19] and Al-Dalaien and Alhroob [20] found OROA to be a significant predictor of operational efficiency and financial sustainability within the Altman Z-score framework. Unlike traditional ROA, OROA incorporates depreciation and amortization, thus offering a more comprehensive measure of asset utilization.

Recent research highlights OROA’s governance and managerial relevance. Pimenow et al. [21] documented a positive association between sustainability practices, transparency, and OROA in Ukrainian companies, while Lambertides and Louca [22] showed that foreign and institutional ownership improved OROA in maritime transport firms. Bennedsen et al. [23] further demonstrated that even short CEO absences reduced OROA by around 7%, underscoring its sensitivity to leadership capacity and managerial stability.

Other indicators frequently used in the literature to assess long-term financial sustainability include the Economic Efficiency Ratio (EER), Return on Equity (ROE), and solvency ratios, which are argued to provide broader insights into capital costs, financial risk, and long-term value creation [24,25]. However, since the Operating Return on Assets (OROA) is derived from cash-like earnings components such as operating profit or EBITDA, it is less affected by accounting adjustments and accrual distortions than measures based on net income [26]. Accordingly, OROA more accurately reflects firms’ true operational efficiency and effectiveness in resource utilization [27]. Moreover, prior studies suggest that OROA is particularly suitable for examining operational performance because of its sensitivity to sector-specific structures and responsiveness to external shocks such as regulatory changes [28,29]. For these reasons, OROA has become an increasingly preferred indicator in empirical analyses exploring the relationship between corporate governance, operational efficiency, and financial sustainability, owing to its robust cash-flow orientation and regulatory relevance.

Taken together, these findings establish OROA as a reliable and sector-relevant metric that integrates the operational and financial dimensions of sustainability. Its robustness across industries, combined with its sensitivity to governance and managerial factors, makes OROA particularly appropriate for analyzing financial sustainability within maritime subsectors. In addition to OROA, this study employs the Economic Efficiency Ratio (EER) as a complementary indicator of operational efficiency. Unlike the Economic Value Added (EVA) metric, EER does not require weighted average cost of capital (WACC) adjustments, rendering it compatible with firm-level accounting data provided by TUIK. Nevertheless, to enhance robustness in future research, comparing OROA-based results with alternative indicators such as EVA, ROE, or solvency measures would help capture broader aspects of long-term financial sustainability.

2.3. Financial Sustainability Challenges in the Maritime Industry

The maritime industry faces distinctive financial sustainability challenges stemming from high capital intensity, cyclical volatility, and exposure to fluctuating freight and bunker markets. Regulatory pressures and limited access to financing further exacerbate these risks, particularly in emerging economies. Prior studies have emphasized the value of financial indicators in capturing such dynamics. Brlečić et al. [30] emphasized the role of Return on Equity (ROE), Return on Assets (ROA), and working-capital management in post-crisis recovery. Lee et al. [31] and Kang et al. [32] found that leverage and debt ratios significantly influence investor confidence, while Park et al. [33] demonstrated that combining cash-flow measures with market indicators enhances the prediction of default risk. Moreover, governance and ownership structures have been shown to affect maritime performance [22]. Collectively, these findings suggest that no single indicator can fully capture maritime financial sustainability. A comprehensive framework integrating OROA, ROE, EVA, Tobin’s Q, and governance-related dimensions is therefore required. In emerging markets such as Türkiye, firm-level analyses employing OROA remain limited, highlighting both the empirical gap and the policy relevance of this study.

2.4. Research Gap and Study Contribution: A Sub-Sectoral Maritime Perspective

Although the maritime industry is often conceptualized as a homogeneous sector, such an approach overlooks the pronounced operational and financial heterogeneity across its subsectors. Firms operate under divergent risk profiles and sustainability pressures shaped by institutional environments, market conditions, and regulatory frameworks—particularly within emerging economies. Existing research largely adopts aggregated or macroeconomic perspectives, thereby overlooking comparative financial analyses between subsectors. In Türkiye, this gap is further compounded by limited access to firm-level financial data, as most maritime enterprises are privately owned. The restricted availability of microdata from the Türkiye Statistical Institute (TUIK), especially within the NACE Rev. 2 classification, has constrained empirical research that could inform differentiated financial and policy strategies for enhancing sectoral resilience.

This study addresses these shortcomings by conducting a comparative company-level panel data analysis of two key subsectors: (i) ship repair and maintenance and (ii) maritime freight transportation. The results highlight substantial heterogeneity in the determinants of financial sustainability, reinforcing the need for customized financial management practices and sector-specific policy interventions.

This study makes four key contributions:

• It provides one of the earliest subsector-specific empirical analyses of Turkish maritime industry using TUIK company-level data.
• It constructs a balanced firm-year panel (2010–2022) for maintenance and coastal shipping, enabling longitudinal analysis of financial sustainability.
• It operates financial sustainability through OROA, offering a robust and comparable indicator across heterogeneous firms.
• It delivers policy-relevant insights by showing how subsectoral differences in assets, liquidity, and profitability shape resilience, especially under the maritime energy transition.

By integrating subsectoral differentiation into maritime finance research, this study bridges a critical gap in the literature and enhances the relevance of sustainability assessments in emerging markets. The following section outlines the methodological framework adopted to operationalize this contribution.

3. Materials and Methods

To establish a robust empirical foundation, this section outlines the methodological framework of this study. It begins with the sample selection criteria and official data sources, followed by the operational definitions of variables, including key financial ratios used as sustainability indicators. Finally, the econometric techniques are presented, with particular emphasis on panel data estimations suited to company-level analyses in emerging markets. This structure ensures transparency, replicability, and alignment with established standards in transportation economics research.

3.1. Sample Selection and Data Sources

This study investigates company-level financial sustainability across two strategically significant subsectors of Turkish maritime industry: (i) ship and boat maintenance and (ii) coastal freight shipping.

Financial data were obtained from the Türkiye Statistical Institute (TUIK), the official authority for national statistics, covering the period 2010–2022. The dataset originates from the Annual Industry and Service Statistics, which report company-level financial statements classified under standardized four-digit NACE codes (Rev. 2). At the time of data retrieval (January 2025), the most recent harmonized dataset extended to fiscal year 2022, consistent with TUIK’s established release schedule. Accordingly, the empirical scope of this study is constrained to the pre-2023 period.

To ensure comparability and robustness, only firm-year observations with complete coverage of essential financial items—including assets, liabilities, EBIT, and cash flow—were retained. The unit of analysis is the firm-year observation. Each row in the dataset corresponds to a single company’s annual financial indicators, classified according to its four-digit NACE code. This structure enables fixed-effects panel estimations that account for unobserved heterogeneity across firms. These classifications are fully aligned with the Statistical Classification of Economic Activities in the European Community (NACE Rev. 2), as adopted by the European Commission for Competition and Industrial Policy [34].

The final dataset consists exclusively of private maritime firms continuously observed over the 2010–2022 period; therefore, there is no firm entry or exit, and survivorship bias is not a concern. To ensure panel balance, only firms with uninterrupted financial records for the 2010–2022 period were retained. Firms with missing or inconsistent financial statements were excluded to maintain variable completeness and comparability across years. This procedure resulted in a balanced panel comprising 190 maintenance firms and 208 shipping firms. A comparison between included and excluded firms revealed no significant differences in average total assets, leverage, or profitability (p > 0.10), suggesting that the final balanced sample is broadly representative and not subject to material survivorship bias. For Model I (ship and boat maintenance), the panel includes 190 firms (N = 190) observed over 13 years (T = 13), resulting in 2470 firm-year observations. For Model II (shipping), the panel comprises 208 firms (N = 208) over the same period, yielding 2704 firm-year observations. This balanced design facilitates comparability while enabling subsector-specific analysis. Within the panel framework, N denotes the number of firms and T the number of years. Accordingly, the fixed-effects estimations are conducted at the firm-year level rather than at the aggregate sector level. The overall sample construction process is summarized in Figure 1.

Figure 1. Flow chart of the methodological framework for dataset preparation.

3.2. Estimation Strategy

This study employs panel data estimation techniques designed to capture both firm-specific heterogeneity and cross-sectional dependence (CSD) typical of maritime industries. The baseline specification is estimated using the Fixed-Effects (FE) estimator, which controls for unobserved, time-invariant firm characteristics that could bias the results. The Hausman [35] test, which confirmed systematic differences between the two estimators, justifies the choice of FE over Random Effects (RE).

Prior to estimation, all variables were tested for stationarity using the Cross-Sectionally Augmented Dickey–Fuller (CADF) test proposed by ref. [36]. The results indicate that the dependent variable, OROA, and most firm-level financial ratios are stationary in levels (I(0)), whereas FATR and FDTA display weak unit-root behavior and were therefore first-differenced. Given the relatively short time dimension (T = 13) and the bounded nature of financial ratios, panel cointegration was not detected. Consequently, the estimations capture short- to medium-run dynamics rather than long-run equilibrium relationships.

Given the presence of heteroskedasticity, serial correlation, and cross-sectional dependence (CSD) identified through the Breusch–Pagan LM, Wooldridge, and Pesaran CD tests, conventional standard errors would be biased and inconsistent. Accordingly, the fixed-effects (FE) estimator was employed with Driscoll–Kraay standard errors, which are robust to general forms of cross-sectional and temporal dependence when N is large and T is moderate [37,38]. To verify the robustness of the findings, two additional model specifications were estimated. The first applies the FE estimator with firm-clustered robust standard errors to correct for heteroskedasticity and serial correlation within firms. The second employs ref. [39] Common Correlated Effects (CCE) estimator, which explicitly addresses unobserved common factors and cross-sectional dependence by incorporating the cross-sectional averages of both dependent and independent variables.

In addition to FE with Driscoll–Kraay SEs, we estimate Pesaran’s CCE–MG by augmenting the regressions with the cross-section averages of y (OROA) and X (and their lags), thereby purging unobserved common factors. As shown in the robustness tests (Appendix B, Table A5 and Table A6), the coefficient signs and magnitudes remained consistent across all three estimation methods, confirming the stability of the Driscoll–Kraay specification. For panels with T = 13, the lag length for the Driscoll–Kraay correction was set to two (T^(1/3) ≈ 2), following the rule of thumb for short- to medium-term persistence in annual financial data.

To further examine the temporal stability of the estimated relationships, all models were re-estimated by interacting with each firm-level ratio with a post-2020 dummy variable under firm and year fixed effects. This specification tests whether slope coefficients differ between the pre-2020 (2010–2019) and post-2020 (2020–2022) periods, capturing potential structural shifts associated with the COVID-19 pandemic and exchange-rate volatility. The results reveal heterogeneous coefficient stability across subsectors. In the maintenance model (Model I), the joint significance test of all post-2020 interactions is insignificant, indicating that the relationships between OROA and firm-level determinants remained stable over time. By contrast, in the shipping model (Model II), the joint test is highly significant (F = 111.96, p < 0.001), suggesting partial slope adjustments after 2020. Specifically, the responsiveness of OROA to NCATA increased, while its sensitivity to TATR, EER, and FDTA declined. These results suggest that structural shocks—such as the pandemic and exchange-rate crises—marginally reweighted profitability drivers toward capital allocation efficiency, although the overall direction of relationships remained consistent. Full estimation results for both subsectors are presented in Appendix B (Table A7 and Table A8).

Figure 2 illustrates the sharp volatility in bunker and LNG fuel prices around 2020, coinciding with the structural break identified in the models. The 2020 oil price collapse and subsequent rebound provide an economic narrative for the post-2020 profitability shifts, reflecting cost pass-through and temporary capacity layups.

Figure 2. Bunker and LNG fuel prices (2011–2021) [40].

To address cross-sectional dependence arising from unobserved common factors—such as global freight cycles, exchange-rate shocks, and fuel price volatility—this study also estimated models using ref. [39] Common Correlated Effects (CCE) approach. This method augments the standard FE specification by incorporating cross-sectional averages of both dependent and explanatory variables, thereby filtering out the influence of unobserved common components that may otherwise bias coefficient estimates. In particular, both the CCE Mean Group (CCEMG) and Pooled CCE (CCEP) estimators were implemented as robustness checks. As shown in the robustness results (Appendix B, Table A7 and Table A8), the coefficient signs and magnitudes are consistent with those obtained from the baseline Driscoll–Kraay specification, confirming that the main findings are not driven by unobserved common shocks or cross-sectional spillovers.

Potential endogeneity and simultaneity between OROA and certain explanatory variables—such as ROE, cash-flow ratios, and EER—were also considered. The FE specification inherently reduces bias stemming from unobserved, time-invariant firm characteristics by utilizing within-firm variation. Given the moderate time dimension (T = 13) and sample size (N ≈ 190–208), dynamic panel estimators such as system-GMM were not employed to avoid weak-instrument and overfitting concerns. Instead, robustness checks were performed using one-year lagged explanatory variables (e.g., ROE, EER, and liquidity measures). The results remained directionally and statistically consistent with the baseline estimations, suggesting that simultaneity does not materially bias the conclusions. This study’s empirical focus, therefore, lies in identifying structural relationships rather than asserting strict causal effects.

To capture heterogeneity across maritime subsectors, separate fixed-effects estimations were performed for the maintenance and shipping industries. This design allows slope coefficients to differ between subsectors, reflecting their structural contrasts in asset intensity, capital turnover, and liquidity management. During preliminary estimations, several interaction terms (e.g., NCATA × FATR, CR × asset intensity) were also tested to explore potential cross-effects among asset composition and efficiency measures. However, these terms were excluded from the final specifications due to moderate-to-high multicollinearity (variance inflation factors exceeding 5) and unstable coefficient signs across alternative estimators. Consequently, the final models were kept parsimonious to preserve parameter interpretability, ensure estimator stability, and facilitate a clear comparison between the two subsectors.

Accordingly, the results reported in the empirical section are based primarily on the FE estimator with Driscoll–Kraay standard errors (lag = 2), which provides reliable inference in the presence of both serial and cross-sectional dependence. To ensure transparency and replicability, detailed robustness diagnostics and extended model specifications are reported in Appendix C Table A9, Table A10, Table A11 and Table A12, covering alternative lag structures, subsector-level estimations, and post-2020 sensitivity checks.

3.3. Definition of Variables and Correlation Matrix

The dependent variable in this study is the Operating Return on Assets (OROA), calculated as the ratio of earnings before interest and taxes (EBIT) to total assets. OROA provides a precise assessment of operational profitability by excluding taxation, interest expenses, and non-operating items. This makes it particularly relevant for asset-intensive industries such as maritime transport and marine services, where long-term financial sustainability depends on operational efficiency rather than short-term accounting adjustments [41,42].

Maritime companies typically operate with substantial fixed assets—vessels, port facilities, and maintenance infrastructure—while being exposed to high revenue volatility driven by freight rates, bunker fuel prices, and seasonal demand. Leverage, tax regimes, or exchange-rate fluctuations can distort conventional profitability measures like ROA. In contrast, OROA provides a stable and policy-neutral measure of operational performance, acting as a dependable benchmark for assessing long-term financial sustainability [43]. Its managerial relevance is further underscored by [23], who showed that even temporary CEO absences can reduce OROA by approximately 7%, highlighting its sensitivity to leadership and decision-making.

All firms in the sample are non-listed and therefore lack share price and outstanding share information. However, book-value equity data are available in the TUIK micro dataset, enabling the calculation of accounting-based profitability ratios. Profitability indicators based on net income alone would capture financial structure and taxation effects that vary considerably across firms, thereby distorting the measurement of pure operational performance. In contrast, OROA—defined using EBIT relative to total assets—provides a cleaner measure of core operating efficiency that is comparable across maritime subsectors with different leverage and fiscal regimes. Accordingly, OROA was selected as the most consistent indicator of firm-level operational sustainability in the dataset.

In addition to OROA, the empirical models include a set of company-level financial ratios that capture asset structure, liquidity, profitability, capital growth, and leverage. Appendix A presents the operational definitions and calculation methods of these variables in Table A1 (Model I) and Table A2 (Model II). For transparency, correlation matrices are also reported: Appendix A, Table A3 provides the results for Model I, and Table A4 for Model II. Year-fixed effects (2010–2022) are included to control for unobserved time-specific shocks, including macro-financial disturbances such as exchange-rate fluctuations, fuel-price changes, and regulatory adjustments.

The variables Economic Efficiency Ratio (EER) and Net Cash Flow to Current Liabilities (NCFCL) are also incorporated as firm-level financial performance indicators. EER is defined as Earnings Before Tax (EBT) divided by Total Assets, reflecting the firm’s ability to generate operating earnings from its asset base. NCFCL is defined as Net Cash Flow divided by Current Liabilities, where Net Cash Flow = Cash Inflows from Operating Activities − Cash Outflows from Operating Activities. The cash-flow data are obtained directly from firms’ official cash-flow statements in the TUIK micro-dataset, ensuring accuracy in measuring operational liquidity. Both ratios are expected to exhibit positive associations with OROA. All continuous variables, including EER and NCFCL, are winsorized at the 1st and 99th percentiles to mitigate the influence of outliers. Detailed operational definitions and descriptive statistics appear in Appendix A Table A1 and Table A2.

3.4. Model Specification

To assess the financial sustainability of maritime companies in Türkiye, two sector-specific panel regression models are estimated. Recognizing the structural and operational heterogeneity across subsectors, separate models are developed for the ship and boat maintenance sector (Model I) and the coastal freight shipping sector (Model II). Both models explain variations in financial sustainability, measured by the Operating Return on Assets (OROA), as a function of company-level financial ratios that capture asset structure, liquidity, profitability, growth, and leverage.

The econometric specifications are presented as follows:

Model I—Maintenance Firms:

$${OROA}_{it}={\beta }_0+{\beta }_1{QR}_{it}+{\beta }_2{NCATA}_{it}+{\beta }_3{OMR}_{it}+{\beta }_4{ROE}_{it}+{\beta }_5{CAR}_{it}+{\beta }_6{EER}_{it}+{\beta }_7{NCFCL}_{it}+{\beta }_8{CR}_{it}+{\beta }_9{EGR}_{it}+{\beta }_{10}{ROS}_{it}+{\beta }_{11}{COGSR}_{it}+{\mu }_{i }+{\epsilon }_{it}$$

(1)

Model II—Shipping Firms:

$${OROA}_{it}={\alpha }_0+{\alpha }_1{TATR}_{it}+{\alpha }_2{NCATA}_{it}+{\alpha }_3{FATR}_{it}+{\alpha }_4{EER}_{it}+{\alpha }_5{NPGR}_{it}+{\alpha }_6{FDTA}_{it}+{\alpha }_7{NCFCL}_{it}+{\alpha }_8{CR}_{it}+{\alpha }_9{EGR}_{it}+{\alpha }_{10}{ROE}_{it}+{\delta }_{i }+{\nu }_{it}$$

(2)

In these specifications, $ORO{A}_{it}$ denotes the operating return on assets for firm i in year t. The terms ${\alpha }_i$ and ${\mu }_i$ capture firm-specific fixed effects, while ${\delta }_t$ represents year fixed effects (2010–2022). Error terms ${\epsilon }_{it}$ and ${\nu }_{it}$ denote idiosyncratic disturbances.

3.5. Multicollinearity Diagnostics

Potential multicollinearity among the explanatory variables was assessed using Variance Inflation Factor (VIF) analysis, conducted separately for the maintenance (Model I) and shipping (Model II) subsectors. Multicollinearity can inflate standard errors and reduce statistical precision, thereby undermining the credibility of regression estimates. Given that company-level financial datasets are often constructed from accounting ratios, some degree of correlation is expected.

The VIF results, presented in Table 1, indicate that all values fall well below the conventional threshold of 5. This outcome confirms that the explanatory variables contribute uniquely to OROA in both subsectors and supports the stability, interpretability, and robustness of the estimated models [44,45].

Table 1. Multicollinearity Diagnostics: Variance Inflation Factors (VIFs) for Both Sectoral Models.

Variable	VIF—Model I (Maintenance)	Variable	VIF—Model II (Shipping)
NCATA	1.05	NCATA	1.04
QR	3.06	TATR	1.07
CR	2.95	CR	1.07
NCFCL	1.73	NCFCL	1.07
ROS	1.45	-----	-----
OMR	1.35	-----	-----
CAR	1.27	-----	-----
COGSR	1.19	-----	-----
-----	-----	NPGR	1.05
-----	-----	FATR	1.04
-----	-----	FDTA	1.03
ROE	1.00	ROE	1.02
EER	1.08	EER	1.10
EGR	1.03	EGR	1.01

Note: All VIF values are well below the conventional threshold of 5. The highest observed VIF is 3.28 for CATA in the maintenance sector model. Source: Authors’ calculations based on TUIK data.

3.6. Cross-Sectional Dependence Tests

Testing for cross-sectional dependence (CSD) is essential given the structure of the panel dataset, which groups company-level observations by sector over time. The presence of CSD indicates that unobserved common shocks or macroeconomic influences simultaneously affect multiple companies, violating the assumption of independence across cross-sectional units. If this issue is not dealt with, it could lead to parameter estimates that are wrong and not very useful. CSD was tested using the Pesaran CD test [46], which is particularly suitable for datasets with a relatively large number of cross-sectional units (N) and a shorter time dimension (T). The test evaluates the average pairwise correlation of residuals to determine whether cross-sectional dependence exists.

Results reported in Table 2 indicate that the maintenance subsector (Model I) exhibits statistically significant cross-sectional dependence (CSD) for nearly all variables, with p-values below the 1% level. This finding confirms that firms within this subsector are jointly influenced by common sectoral and macroeconomic shocks, thereby validating the application of estimation techniques that are robust to CSD. The only exception was the Return on Sales (ROS) variable, which did not display statistically significant dependence. In line with methodological recommendations [46,47], first-generation panel unit root tests were applied to ROS, whereas second-generation tests were used for all other variables.

Table 2. Pesaran CD Test for Cross-Sectional Dependence in Model I (Maintenance Firms).

Variable	CD Test	p-Value	corr
OROA	11.71	0.000	0.024
QR	3.50	0.000	0.007
CR	8.02	0.000	0.017
NCFCL	7.06	0.000	0.015
NCATA	11.56	0.000	0.024
CAR	8.74	0.000	0.018
ROS	1.41	0.159	0.003
OMR	6.11	0.000	0.013
COGSR	7.63	0.000	0.016
ROE	5.93	0.000	0.012
EER	10.03	0.000	0.021
EGR	93.06	0.000	0.193

Note: The CD test confirms significant cross-sectional dependence on Panel Model I, which was addressed using robust estimation techniques. Source: Authors’ calculations based on TUIK data.

As reported in Table 3, all variables in the shipping sector (Model II) exhibit statistically significant cross-sectional dependence at the 1% level. This outcome reinforces the need to apply estimation methods that account for such dependence.

Table 3. Pesaran CD Test for Cross-Sectional Dependence in Model II (Shipping Firms).

Variable	CD Test	p-Value	corr
OROA	10.17	0.000	0.019
CR	11.75	0.000	0.022
NCFCL	15.62	0.000	0.030
NCATA	76.40	0.000	0.628
NPGR	10.87	0.000	0.021
FATR	26.00	0.000	0.083
FDTA	12.43	0.000	0.040
ROE	6.13	0.000	0.012
TATR	11.88	0.000	0.038
EER	12.21	0.000	0.023
EGR	60.90	0.000	0.115

Note: The CD test confirms significant cross-sectional dependence in Panel Model II, which was addressed using robust estimation techniques. Source: Authors’ calculations based on TUIK data.

3.7. Panel Unit Root Tests

Panel datasets frequently exhibit cross-sectional dependence (CSD), which can compromise the reliability of unit root tests and result in invalid statistical inferences. As emphasized by Nelson and Plosser [48], many macroeconomic and firm-level time series follow non-stationary processes that resemble random walks. Incorporating such variables into regression models without appropriate transformations may lead to spurious relationships and misleading conclusions. Therefore, assessing the stationarity properties of the variables constitutes an essential preliminary step in panel data analysis.

Given the presence of CSD, as confirmed by the Pesaran CD tests reported in Section 3.5, the selection of suitable unit root tests becomes critical. Baltagi [47] suggests using second-generation tests to detect significant cross-sectional dependence, a common problem in firm-level financial panels. Accordingly, second-generation panel unit root tests were applied to all variables exhibiting CSD, while first-generation tests were reserved for those without evidence of dependence.

In the maintenance sector model (Model I), the Pesaran CD test results revealed significant cross-sectional dependence (CSD) for all variables except Return on Sales (ROS). Consequently, ROS was examined using first-generation unit root tests—namely the Levin, Lin, and Chu (LLC) and Im, Pesaran, and Shin (IPS) tests—given the absence of CSD. All other variables in Model I were assessed using the second-generation Cross-sectionally Augmented Dickey–Fuller (CADF) test proposed by ref. [36], which incorporates cross-sectional averages into the regression to account for interdependencies across units. The results presented in Table 4 indicate that all variables in Model I are stationary at levels, including ROS, as validated by the first-generation tests.

Table 4. Panel Unit Root Test Results for Panel Model I.

Variable	Test Type	Test Method	Test Statistic	p-Value	Stationarity	Integration Order
OROA	2nd	CADF	−13.169	0.000	Yes	I(0)
QR	2nd	CADF	−9.002	0.000	Yes	I(0)
CR	2nd	CADF	−8.042	0.000	Yes	I(0)
NCFCL	2nd	CADF	−15.206	0.000	Yes	I(0)
NCATA	2nd	CADF	−6.536	0.000	Yes	I(0)
CAR	2nd	CADF	−11.444	0.000	Yes	I(0)
ROS	1st	LLC	−17.442	0.000	Yes	I(0)
ROS	1st	IPS	−15.149	0.000	Yes	I(0)
OMR	2nd	CADF	−13.279	0.000	Yes	I(0)
COGSR	2nd	CADF	−11.782	0.000	Yes	I(0)
ROE	2nd	CADF	−15.443	0.000	Yes	I(0)
EER	2nd	CADF	−15.714	0.000	Yes	I(0)
EGR	2nd	CADF	−14.523	0.000	Yes	I(0)

All variables were tested for stationarity using second-generation CADF tests, except for ROS, which was evaluated with first-generation methods (LLC and IPS) due to the absence of cross-sectional dependence. All tests reject the null hypothesis of a unit root at the 1% level, confirming stationarity. Source: Authors’ calculations based on TUIK data.

For the shipping sector model (Model II), the same stationarity-testing framework was applied to maintain methodological consistency with Model I. As all variables exhibited statistically significant cross-sectional dependence (see Section 3.5), the second-generation Cross-sectionally Augmented Dickey–Fuller (CADF) test proposed by Pesaran [36] was employed. The results reported in Table 5 indicate that most variables are stationary at levels [I(0)], while the Fixed Asset Turnover Ratio (FATR) and Financial Debt-to-Total Assets ratio (FDTA) attain stationarity only after first differencing [I(1)]. These outcomes confirm that the dataset satisfies the requirements for subsequent panel regression analyses based on second-generation tests.

Table 5. Panel Unit Root Test Results for Panel Model II.

Variable	Test Type	Test Method	Test Statistic	p-Value	Stationarity	Integration Order
OROA	2nd	CADF	−9.156	0.000	Yes	I(0)
NCATA	2nd	CADF	−5.470	0.000	Yes	I(0)
NPGR	2nd	CADF	−21.184	0.000	Yes	I(0)
FATR *	2nd	CADF	−18.485	0.000	No	I(1)
FDTA *	2nd	CADF	−13.766	0.000	No	I(1)
ROE	2nd	CADF	−17.583	0.000	Yes	I(0)
TATR	2nd	CADF	−2.331	0.010	Yes	I(0)
EER	2nd	CADF	−7.871	0.000	Yes	I(0)
EGR	2nd	CADF	−6.578	0.000	Yes	I(0)
CR	2nd	CADF	−9.861	0.000	Yes	I(0)
NCFLC	2nd	CADF	−15.771	0.000	Yes	I(0)

Note: All variables were tested for stationarity using the second-generation CADF approach, which accounts for cross-sectional dependence. All tests reject the null hypothesis of a unit root at the 1% significance level. Variables marked with * are integrated of order one, I(1), and were transformed by first differencing prior to regression estimations. Source: Authors’ calculations based on TUIK data.

3.8. Model Specification Test: Hausman Approach

The Hausman specification test was conducted to identify the appropriate panel estimation technique for both sector-specific models. This test evaluates whether unobserved firm-specific heterogeneity is correlated with the explanatory variables. In the presence of such correlation, the random-effects (RE) estimator becomes inconsistent, and the fixed-effects (FE) estimator is preferred for yielding unbiased and consistent coefficients [35].

Table 6 reports the Hausman test results. For both the ship and boat maintenance model (Model I) and the coastal freight shipping model (Model II), the null hypothesis—that RE and FE estimators do not differ systematically—is rejected at the 1% level. The chi-square statistics are 23.00 (p < 0.05) for Model I and 45.04 (p < 0.05) for Model II.

Table 6. Hausman Specification Test Results.

Model	Chi Square Statistics	p-Value	Preferred Estimator
Model I: Maintenance	23.00	0.017	Fixed Effects (FE)
Model II: Shipping	45.04	0.000	Fixed Effects (FE)

Note: The null hypothesis of no systematic difference between RE and FE is rejected at the 1% significance level for both models, confirming the use of the fixed-effects (FE) estimator. Source: Authors’ calculations based on TUIK data.

These results confirm a statistically significant correlation between unobserved firm-specific effects and the explanatory variables in both subsectors. Accordingly, the FE estimator is adopted for both models, ensuring robust and consistent parameter estimates in the subsequent regression analyses.

3.9. Robustness and Model Validity Diagnostics

After model specification, a series of diagnostic tests was conducted to assess the robustness and validity of the estimated panel regressions. Two econometric issues frequently encountered in panel data—heteroskedasticity and autocorrelation—were explicitly analyzed, as their existence can undermine efficiency and result in biased statistical inference if inadequately addressed. Heteroskedasticity was tested using the Modified Wald test for groupwise heteroskedasticity [49], whereas first-order autocorrelation was evaluated with the Wooldridge test [45], a widely adopted procedure in panel data analysis. The findings in Table 7 demonstrate statistically significant heteroskedasticity and serial correlation in both models at the 1% significance level, thereby refuting the null hypotheses of homoskedasticity and the absence of autocorrelation.

Table 7. Diagnostic Test Results for Heteroskedasticity and Autocorrelation.

Model	Heteroskedasticity Test (Wald)	Autocorrelation Test (Wooldridge)	Correction Applied
Model I: Maintenance	χ2 = 4.70 × 107, p < 0.01	F = 4.442, p = 0.036	Driscoll–Kraay SE
Model II: Shipping	χ2 = 9.40 × 1011, p < 0.01	F = 513.44, p = 0.000	Driscoll–Kraay SE

Note: Both heteroskedasticity and first-order autocorrelation were detected in the two models. Robustness was ensured by applying company-level clustered standard errors, which are correct for both issues. Source: Authors’ calculations based on TUIK data.

To correct these violations and strengthen inference, final estimations were performed with company-level clustered robust standard errors for each sector. This adjustment provides standard errors robust to both heteroskedasticity and autocorrelation, thereby enhancing the reliability of coefficient estimates and their significance levels.

4. Empirical Results and Interpretation

This section reports the estimation results for the sector-specific panel regression models evaluating the financial sustainability of maritime companies in Türkiye. Based on the Hausman specification tests [35] reported in Section 3.8, the fixed-effects (FE) estimator was selected for both subsectors. To correct for the heteroskedasticity and autocorrelation detected in Section 3.9, Driscoll–Kraay standard errors were employed, thereby enhancing the robustness of coefficient estimates and the reliability of statistical inference.

4.1. Regression Results and Interpretation for Model I

The maintenance sector model (Model I) is estimated on an unbalanced panel comprising 190 firms over 13 years (2010–2022). Such a panel structure typically entails heteroskedasticity, serial correlation, and cross-sectional dependence (CSD), each of which can bias conventional standard errors. The Pesaran CD test confirmed the presence of CSD, necessitating robust estimation techniques.

Following Hoechle [35] and Petersen [38], the Fixed-Effects (FE) estimator was implemented with Driscoll–Kraay standard errors (lag = 2), which are robust to heteroskedasticity, serial correlation, and general forms of cross-sectional dependence when N is large and T is moderate. This approach ensures consistent inference in the presence of macro-level shocks affecting all firms simultaneously—such as fuel price volatility or currency depreciation—common in the Turkish maritime services sector (see Table 8 for estimation results).

Table 8. FE Regression with Driscoll–Kraay SEs (lag = 2): Maintenance Firms (Model I).

Variable	Coefficient	p-Value
NCATA	0.21274	0.002 *
QR	−0.00012	0.933
CR	−0.00990	0.002 *
EER	0.71167	0.000 *
ROS	−0.00585	0.275
OMR	0.00603	0.064 ***
COGSR	−0.07472	0.120
ROE	−0.00038	0.635
NCFCL	0.02076	0.093 ***
EGR	0.00049	0.872
CAR	0.01117	0.675
Cons	0.14288	0.008
R2 (overall): 0.6822
Observations: 2470	Firms (N): 190	Years (T): 13

Note: Robust Driscoll–Kraay standard errors (lag = 2) are applied. Statistical significance denoted as * p < 0.01, *** p < 0.10. Source: Authors’ own calculations based on TUIK firm-level data (2010–2022).

The estimation results highlight several key determinants of financial sustainability in Turkish maritime maintenance industry. The Non-Current Assets-to-Total Assets (NCATA) ratio exhibits a positive and statistically significant coefficient (p = 0.002), indicating that capital investment in dry-dock facilities, repair equipment, and technical infrastructure enhances operational efficiency and profitability. This finding underscores the capital-intensive nature of the maintenance segment, where productive asset deployment is a central driver of long-term performance.

In contrast, the Current Ratio (CR) has a negative and statistically significant effect (p = 0.002), suggesting that excessive liquidity or idle short-term assets may reflect inefficient working-capital management, thereby reducing returns on operating assets. The Economic Efficiency Ratio (EER) displays a strong positive association with profitability (p < 0.001), confirming that firms achieving higher operational efficiency and value creation tend to outperform their peers financially.

The Operating Margin Ratio (OMR) is positively related to profitability at the 10% significance level (p = 0.064), reflecting the contribution of cost control and pricing discipline to sustainable performance. Similarly, the Net Cash Flow-to-Current Liabilities (NCFCL) ratio has a positive, marginally significant coefficient (p = 0.093), implying that firms maintaining positive operating cash flows are more resilient under macroeconomic volatility, inflationary pressures, and credit constraints.

Other financial indicators—such as the Quick Ratio (QR), Return on Sales (ROS), Cost of Goods Sold to Sales Ratio (COGSR), Return on Equity (ROE), Equity Growth Ratio (EGR), and Cash to Assets Ratio (CAR)—do not show statistically significant effects, suggesting that their influence is secondary or context-specific within the observed period.

Overall, the findings suggest that productive asset utilization, operational efficiency, and prudent liquidity management primarily drive financial sustainability in Turkish maritime maintenance sector. Firms that strategically invest in long-term assets while maintaining efficient cash-flow operations are better positioned to sustain profitability and competitiveness in capital-intensive maritime industries.

4.2. Regression Results and Interpretation for Model II

Table 9 presents the fixed-effects (FE) regression results with Driscoll–Kraay standard errors (lag = 2) for the shipping sector, where operating profitability is measured by the Operating Return on Assets (OROA). This estimation procedure corrects for cross-sectional dependence, heteroskedasticity, and serial correlation, thereby ensuring robust statistical inference. Several financial ratios emerge as statistically significant determinants of profitability, revealing the financial sustainability dynamics of Turkish coastal freight shipping companies.

Table 9. FE Regression with Driscoll–Kraay SEs (lag = 2): Shipping Firms (Model II).

Variable	Coefficient	p-Value
TATR	0.04420	0.023 **
NCATA	0.40555	0.000 *
FATR	−9.11000	0.057 ***
EER	0.78194	0.000 *
NPGR	0.00009	0.068 ***
FDTA	0.04518	0.547
NCFCL	−0.00164	0.508
CR	−0.01132	0.001 *
EGR	0.00293	0.344
ROE	−0.00383	0.376
Cons	0.16752	0.000
R2 (overall): 0.3430
Observations: 2496	Firms (N): 208	Years (T): 13

Note: Robust Driscoll–Kraay standard errors (lag = 2) are applied. Statistical significance denoted as * p < 0.01, ** p < 0.05, *** p < 0.10. Source: Authors’ own calculations based on TUIK firm-level data (2010–2022).

The estimation results identify several key drivers of financial sustainability in Turkish coastal freight shipping sector. The Total Asset Turnover Ratio (TATR) exhibits a positive and statistically significant coefficient at the 5% level (p = 0.023), confirming that efficient utilization of the asset base enhances operating profitability. This finding aligns with the resource-based view, emphasizing that effective asset deployment serves as a central source of competitive advantage in capital-intensive industries.

The Non-Current Assets-to-Total Assets (NCATA) ratio is strongly positive and highly significant (p < 0.01), indicating that sustained investments in vessels, port infrastructure, and other long-term assets strengthen operating performance by expanding capacity and productivity. In contrast, the Fixed Asset Turnover Ratio (FATR) shows a small but negative and marginally significant coefficient (p = 0.057), suggesting potential inefficiencies in capital utilization—possibly arising from underused assets, aging fleets, or technical constraints in operations.

The Economic Efficiency Ratio (EER) emerges as the most powerful predictor of profitability, displaying a strong positive effect at the 1% level (p < 0.001). This result highlights the role of earnings efficiency as a comprehensive measure of firm-level financial sustainability and operational effectiveness. Conversely, the Cash Ratio (CR) exhibits a negative and statistically significant coefficient (p = 0.001), implying that excessive liquidity may reflect idle funds or overly conservative financial management practices that suppress profitability.

Other variables—including the Financial Debt-to-Total Assets (FDTA) ratio, Net Cash Flow-to-Current Liabilities (NCFCL), Equity Growth Ratio (EGR), and Return on Equity (ROE)—do not exhibit statistically significant effects, suggesting that their influence is either indirect or context dependent.

Taken together, these findings underscore that financial sustainability in Turkish coastal freight shipping sector depends primarily on efficient asset management, strategic long-term investment in fixed capital, earnings efficiency, and optimized liquidity control. These factors collectively enhance firms’ resilience and profitability under volatile macroeconomic and maritime conditions.

4.3. Policy Implications

The empirical findings reveal that financial sustainability in the maritime industry is primarily driven by asset composition, operational efficiency, and liquidity management. The positive and statistically significant effect of the Non-Current Assets-to-Total Assets (NCATA) ratio on operating profitability (OROA) suggests that investment in fixed assets—such as vessel modernization, port equipment, and maintenance facilities—enhances financial returns. This relationship implies that policies promoting fleet renewal, technological upgrading, and long-term asset financing can substantially strengthen the sector’s financial sustainability.

However, this effect is heterogeneous across subsectors. In the maintenance segment, the estimated NCATA coefficient (0.21274) indicates a moderate but steady return to capital deepening, whereas in coastal transportation, the higher coefficient (0.40555) suggests that long-term investments generate more pronounced financial improvements. The negative coefficients of the Current Ratio (CR) across both models imply that excessive liquidity holdings may constrain operational efficiency, emphasizing the need for targeted liquidity optimization and improved working-capital management.

The findings for the Total Asset Turnover Ratio (TATR) and the Economic Efficiency Ratio (EER) confirm that operational and environmental efficiency are key pillars of sustainable profitability. In particular, the strong positive association between EER and OROA demonstrates that greener operations can be financially beneficial, supporting the case for eco-efficiency incentives and cleaner technology adoption in maritime activities.

To illustrate these policy implications quantitatively, we provide simple numerical examples based on the estimated coefficients. For a typical maintenance firm with ₺500 million in total assets, a 5-percentage-point increase in NCATA corresponds to a 1.06-percentage-point rise in OROA—equivalent to roughly ₺5.3 million in additional EBIT per year, assuming a required CAPEX of about ₺25 million (ROI ≈ 21%). For a coastal transportation firm with ₺2 billion in assets, the same 5 pp increase in NCATA implies an annual EBIT gain of ₺40.6 million (ROI ≈ 41%). Similarly, a 0.10 improvement in TATR yields approximately ₺8.8 million in additional EBIT. These calculations demonstrate that fleet-renewal and efficiency-enhancing policies are economically meaningful and proportionate to the sector’s scale, as reflected in the estimated coefficients reported in Table 8 and Table 9.

Overall, the empirical evidence points to several actionable policy directions:

• Fleet renewal and modernization programs should be prioritized in the coastal transportation segment, where asset-based improvements yield the largest profitability effects.
• Targeted financing mechanisms (e.g., green credit lines or asset-based loans) can improve access to capital for sustainable investments.
• Operational efficiency and digitalization incentives—particularly those enhancing TATR and EER—should be supported through fiscal and regulatory instruments.
• Working-capital optimization guidelines may help firms avoid liquidity traps that diminish profitability.

Collectively, these implications offer a roadmap for policymakers and financial institutions to align maritime-sector investment strategies with both profitability and sustainability objectives.

5. Discussion

This study investigates the financial sustainability dynamics of Turkish maritime industry by comparing two structurally distinct subsectors: maintenance (Model I) and coastal shipping (Model II). Using OROA as the main indicator of operational performance, the results demonstrate how asset composition, liquidity management, and energy efficiency jointly determine long-term resilience and profitability.

In the maintenance sector, the strong positive coefficient of NCATA confirms that higher fixed-asset intensity supports operational stability and efficiency, consistent with evidence from capital-intensive industries [16]. In the Turkish context—where access to long-term credit remains limited—strategic investment in technical facilities and maintenance infrastructure is essential for sustaining competitiveness. The positive effect of NCFCL further highlights the importance of liquidity buffers in maintaining financial flexibility, echoing Altman Z-score–based findings [19,20]. Firms with stronger cash-flow positions are better equipped to withstand macroeconomic volatility. In contrast, the negative coefficient of COGSR indicates weaknesses in cost control and operational efficiency, suggesting that firms with higher cost-to-sales ratios face profitability constraints. The significant positive effect of EER underscores that energy efficiency directly contributes to financial sustainability, particularly for maintenance-intensive operations where energy and operational costs are substantial.

In the shipping sector, profitability is reinforced by asset efficiency (TATR), liquidity (CR), and capital structure (NCATA, NCFCL), as well as by EER and NPGR. These results are consistent with prior studies ref. [32], which emphasize that efficient asset utilization and value-oriented strategies strengthen competitiveness in maritime transport. However, the negative coefficient of FATR points to structural inefficiencies in asset deployment, often associated with outdated fleets and fragmented port infrastructure. Overcoming these constraints requires targeted policy interventions that support fleet renewal, green modernization, and energy-efficient port operations aligned with international sustainability standards. The findings also align with ref. [14], who argue that the green transformation of maritime transport and ports is constrained not only by technological or regulatory capacity but also by financial and institutional limitations. Similarly, the capacity of Turkish maritime companies to participate in the green transition is hindered by restricted access to affordable financing and gaps in policy support. Aligning national financing instruments and port-level governance frameworks with international sustainability initiatives is therefore essential to achieving genuine environmental improvement in the sector.

Comparative analysis shows that both subsectors benefit from sound liquidity and value-enhancing investment strategies, yet through different mechanisms. Maintenance firms rely more on physical assets and internal cash flow, whereas shipping firms depend on asset turnover, growth, and modernization. The former face challenges in cost and working-capital efficiency, while the latter confront structural bottlenecks in fleet utilization.

Notably, the robustness of these results was verified through complementary inference diagnostics, including two-way clustered standard errors (firm × year), trimming of the 1% tails of EER and TATR, and a within-firm year-shuffle placebo test. The detailed outcomes, presented in Appendix C (Table A13), confirm that the estimated relationships remain stable and statistically significant across all robustness specifications.

These findings also align with ref. [23], who emphasize that managerial capacity and governance critically shape performance in asset-intensive industries. The evidence presented here reinforces that liquidity management and asset deployment must be strategically aligned with leadership quality and governance standards. Excess liquidity, if left idle, can hinder efficiency, whereas disciplined asset allocation and sound governance enhance operational resilience.

Overall, this dual-sector analysis supports ref. [15], who argue that sector-specific approaches are vital to understanding sustainability dynamics in heterogeneous industries. OROA emerges as a robust and policy-neutral indicator of operational performance [17]. For Turkish maritime economy—characterized by high volatility, limited financing capacity, and fragmented firm structures—the findings underscore the need for subsector-specific financial strategies. Policymakers should prioritize working-capital optimization, asset modernization, and energy-efficiency improvements at the firm level. At the same time, green transition investments—such as fleet renewal, alternative fuels, and emission-reduction technologies—should be treated as long-term strategic commitments rather than short-term cost pressures. For emerging economies like Türkiye, ensuring access to affordable financing for sustainability-oriented expenditures remains a critical policy challenge.

6. Conclusions

This study examined the financial sustainability dynamics of Turkish maritime industry by analyzing firm-level data for the maintenance and coastal shipping subsectors between 2010 and 2022. Using fixed-effects estimators with Driscoll–Kraay corrections, the analysis provides statistically robust associations between asset composition, operational efficiency, and liquidity management in shaping profitability and resilience in a capital-intensive industry exposed to cyclical shocks.

The results indicate that profitability—measured by Operating Return on Assets (OROA)—is associated with productive asset use and efficiency-oriented management. In the maintenance subsector, earnings efficiency and the share of non-current assets in total assets show positive associations with OROA, suggesting that firms emphasizing technical infrastructure and energy-efficient operations tend to perform better in the long run. At the same time, high liquidity levels are associated with lower profitability, implying that idle capital and overly conservative cash holdings correlate with reduced operational efficiency. In the coastal shipping subsector, total asset turnover and earnings efficiency appear positively related to profitability, whereas fixed-asset turnover and the cash ratio display negative associations. These findings suggest that shipping firms benefit from relationships consistent with efficient asset utilization and earnings quality rather than liquidity accumulation.

Overall, the empirical results highlight associations consistent with previous studies emphasizing that asset utilization, financial discipline, and cost control are linked to sustainable performance in capital-intensive industries. The observed relationships between efficiency indicators and profitability align with the maritime finance and industrial sustainability literature, while the negative association of excessive liquidity is consistent with earlier findings on working-capital inefficiencies in emerging economies.

From a managerial and policy perspective, the findings indicate relationships suggesting that firm strategies should be tailored to subsectoral characteristics. Maintenance firms may benefit from capacity expansion, energy efficiency, and cash-flow optimization, whereas shipping firms may prioritize fleet renewal, asset productivity, and balanced financial structures. For policymakers, the results underscore associations pointing to the importance of supporting modernization and green investment through targeted financing mechanisms and fiscal incentives that reward operational efficiency.

While numerical simulations based on estimated coefficients illustrate relationships consistent with potential improvements in operating income, these results should be interpreted as descriptive associations rather than causal effects. Future research could extend this framework to include environmental, social, and governance (ESG) indicators and apply predictive validation or cross-country comparisons to explore how institutional and financial structures relate to long-term resilience within the global blue economy.

Appendix C.1. Time-Effect Controls

Table A9 and Table A10 report the full fixed-effects estimation results with year dummies for Model I (Maintenance) and Model II (Shipping), respectively, covering the 2011–2022 period. Year dummies are included to capture macroeconomic and regulatory shocks such as the 2018 exchange-rate crisis, the COVID-19 pandemic, and the IMO 2020 sulfur-limit adjustment. The joint significance of the time dummies is confirmed in both models (Prob > F = 0.000 < 0.05), indicating that year-specific effects are statistically relevant and should be retained in the specification.

Appendix C.2. Robustness Checks with Winsorized Samples

Table A11 and Table A12 present robustness checks based on winsorized samples for the maintenance and shipping sectors, respectively. All firm-level financial ratios are winsorized at the 1st and 99th percentiles to mitigate the influence of extreme observations and potential outliers. The model specifications otherwise mirror those of the baseline estimations reported in the main text. Across both sectors, the winsorized estimations confirm the robustness of the main results: coefficient signs and relative magnitudes remain stable, and the economic interpretation is unaffected after trimming extreme values.

Table A9. Full fixed-effects estimation with year dummies, 2011–2022 (Model I—Maintenance).

Variable	Coefficient	Std. Error	t	p > |t|	95% Confidence Interval
NCATA	0.1915	0.0509	3.76	0.003	[0.0805, 0.3024]
QR	−0.0014	0.0019	−0.77	0.459	[−0.0056, 0.0027]
EER	0.7117	0.1311	5.43	0.000	[0.4259, 0.9975]
ROS	−0.0065	0.0052	−1.24	0.237	[−0.0180, 0.0049]
OMR	0.0059	0.0029	2.00	0.069	[−0.0005, 0.0123]
COGSR	−0.0723	0.0466	−1.55	0.147	[−0.1739, 0.0292]
ROE	−0.0005	0.0007	−0.66	0.523	[−0.0021, 0.0011]
NCFCL	0.0218	0.0111	1.96	0.074	[−0.0024, 0.0461]
EGR	0.0012	0.0029	0.42	0.683	[−0.0051, 0.0075]
CAR	0.0119	0.0242	0.49	0.631	[−0.0408, 0.0646]
CR	−0.0090	0.0022	−4.11	0.001	[−0.0138, −0.0042]
Year 2011	−0.0055	0.0026	−2.13	0.054	[−0.0112, 0.0001]
Year 2012	0.0254	0.0115	2.20	0.048	[0.0002, 0.0507]
Year 2013	0.0657	0.0035	18.33	0.000	[0.0579, 0.0736]
Year 2014	0.0263	0.0087	3.00	0.011	[0.0072, 0.0455]
Year 2015	0.0141	0.0080	1.76	0.104	[−0.0033, 0.0317]
Year 2016	−0.0127	0.0063	−1.99	0.070	[−0.0265, 0.0011]
Year 2017	−0.0126	0.0049	−2.53	0.026	[−0.0235, −0.0017]
Year 2018	−0.0139	0.0024	−5.65	0.000	[−0.0192, −0.0085]
Year 2019	−0.0097	0.0100	−0.98	0.348	[−0.0315, 0.0120]
Year 2020	−0.0258	0.0055	−4.63	0.001	[−0.0379, −0.0136]
Year 2021	−0.0335	0.0120	−2.79	0.016	[−0.0598, −0.0073]
Year 2022	−0.0392	0.0074	−5.26	0.000	[−0.0555, −0.0229]

Note: Fixed-effects estimation with firm-level clustering for 190 firms (N = 190, T = 12). Year dummies (2011–2022) are included to capture macroeconomic and regulatory shocks such as the 2018 exchange-rate crisis, the COVID-19 pandemic, and the IMO 2020 sulfur-cap regulation. The base year (2010) is omitted to avoid the dummy-variable trap. The joint significance test of time effects (testparm i.year) yields Prob > F = 0.000 < 0.05, confirming the overall importance of year-specific controls.

Table A9. Full fixed-effects estimation with year dummies, 2011–2022 (Model I—Maintenance).

Variable	Coefficient	Std. Error	t	p > |t|	95% Confidence Interval
NCATA	0.1915	0.0509	3.76	0.003	[0.0805, 0.3024]
QR	−0.0014	0.0019	−0.77	0.459	[−0.0056, 0.0027]
EER	0.7117	0.1311	5.43	0.000	[0.4259, 0.9975]
ROS	−0.0065	0.0052	−1.24	0.237	[−0.0180, 0.0049]
OMR	0.0059	0.0029	2.00	0.069	[−0.0005, 0.0123]
COGSR	−0.0723	0.0466	−1.55	0.147	[−0.1739, 0.0292]
ROE	−0.0005	0.0007	−0.66	0.523	[−0.0021, 0.0011]
NCFCL	0.0218	0.0111	1.96	0.074	[−0.0024, 0.0461]
EGR	0.0012	0.0029	0.42	0.683	[−0.0051, 0.0075]
CAR	0.0119	0.0242	0.49	0.631	[−0.0408, 0.0646]
CR	−0.0090	0.0022	−4.11	0.001	[−0.0138, −0.0042]
Year 2011	−0.0055	0.0026	−2.13	0.054	[−0.0112, 0.0001]
Year 2012	0.0254	0.0115	2.20	0.048	[0.0002, 0.0507]
Year 2013	0.0657	0.0035	18.33	0.000	[0.0579, 0.0736]
Year 2014	0.0263	0.0087	3.00	0.011	[0.0072, 0.0455]
Year 2015	0.0141	0.0080	1.76	0.104	[−0.0033, 0.0317]
Year 2016	−0.0127	0.0063	−1.99	0.070	[−0.0265, 0.0011]
Year 2017	−0.0126	0.0049	−2.53	0.026	[−0.0235, −0.0017]
Year 2018	−0.0139	0.0024	−5.65	0.000	[−0.0192, −0.0085]
Year 2019	−0.0097	0.0100	−0.98	0.348	[−0.0315, 0.0120]
Year 2020	−0.0258	0.0055	−4.63	0.001	[−0.0379, −0.0136]
Year 2021	−0.0335	0.0120	−2.79	0.016	[−0.0598, −0.0073]
Year 2022	−0.0392	0.0074	−5.26	0.000	[−0.0555, −0.0229]

Note: Fixed-effects estimation with firm-level clustering for 190 firms (N = 190, T = 12). Year dummies (2011–2022) are included to capture macroeconomic and regulatory shocks such as the 2018 exchange-rate crisis, the COVID-19 pandemic, and the IMO 2020 sulfur-cap regulation. The base year (2010) is omitted to avoid the dummy-variable trap. The joint significance test of time effects (testparm i.year) yields Prob > F = 0.000 < 0.05, confirming the overall importance of year-specific controls.

Table A10. Full fixed-effects estimation with year dummies, 2011–2022 (Model II—Shipping).

Variable	Coefficient	Std. Error	t	p > |t|	95% Confidence Interval
NO	−0.0119	0.0054	−2.19	0.029	[−0.0225, −0.0012]
NCFCL	−0.0014	0.0016	−0.87	0.384	[−0.0045, 0.0017]
NCATA	0.2567	0.0763	3.36	0.001	[0.1070, 0.4064]
EGR	0.0030	0.0012	2.55	0.011	[0.0007, 0.0053]
NPGR	0.00009	0.00006	1.33	0.183	[−0.00004, 0.00021]
EER	0.7868	0.0244	32.20	0.000	[0.7389, 0.8347]
DTATR	0.0165	0.0114	1.44	0.149	[−0.0059, 0.0389]
DFATR	−7.64 × 10−9	7.40 × 10−9	−1.03	0.302	[−2.22 × 10−8, 6.87 × 10−9]
ROE	−0.0032	0.0044	−0.72	0.474	[−0.0119, 0.0055]
FDTA	0.0120	0.0092	1.30	0.194	[−0.0061, 0.0300]
Year 2011	0.0187	0.0625	0.30	0.765	[−0.104, 0.141]
Year 2012	−0.0456	0.0626	−0.73	0.466	[−0.168, 0.077]
Year 2013	−0.0321	0.0626	−0.51	0.609	[−0.155, 0.091]
Year 2014	−0.0407	0.0627	−0.65	0.516	[−0.164, 0.082]
Year 2015	−0.1227	0.0627	−1.96	0.051	[−0.246, 0.000]
Year 2016	−0.0933	0.0630	−1.48	0.139	[−0.217, 0.030]
Year 2017	−0.1193	0.0633	−1.88	0.060	[−0.243, 0.005]
Year 2018	−0.1322	0.0635	−2.08	0.037	[−0.257, −0.008]
Year 2019	−0.1603	0.0638	−2.51	0.012	[−0.285, −0.035]
Year 2020	−0.1944	0.0646	−3.01	0.003	[−0.321, −0.068]
Year 2021	−0.2122	0.0645	−3.29	0.001	[−0.339, −0.086]
Year 2022	−0.2430	0.0650	−3.74	0.000	[−0.371, −0.115]

Note: Fixed-effects estimation with firm-level clustering for 208 firms (N = 208, T = 12). Year dummies (2011–2022) are included to capture macroeconomic and regulatory shocks such as the 2018 exchange-rate crisis, the COVID-19 pandemic, and the IMO 2020 sulfur-cap regulation. The base year (2010) is omitted to avoid the dummy-variable trap. The joint significance test of time effects (testparm i.year) yields Prob > F = 0.000 < 0.05, confirming the overall importance of year-specific controls.

Table A10. Full fixed-effects estimation with year dummies, 2011–2022 (Model II—Shipping).

Variable	Coefficient	Std. Error	t	p > |t|	95% Confidence Interval
NO	−0.0119	0.0054	−2.19	0.029	[−0.0225, −0.0012]
NCFCL	−0.0014	0.0016	−0.87	0.384	[−0.0045, 0.0017]
NCATA	0.2567	0.0763	3.36	0.001	[0.1070, 0.4064]
EGR	0.0030	0.0012	2.55	0.011	[0.0007, 0.0053]
NPGR	0.00009	0.00006	1.33	0.183	[−0.00004, 0.00021]
EER	0.7868	0.0244	32.20	0.000	[0.7389, 0.8347]
DTATR	0.0165	0.0114	1.44	0.149	[−0.0059, 0.0389]
DFATR	−7.64 × 10−9	7.40 × 10−9	−1.03	0.302	[−2.22 × 10−8, 6.87 × 10−9]
ROE	−0.0032	0.0044	−0.72	0.474	[−0.0119, 0.0055]
FDTA	0.0120	0.0092	1.30	0.194	[−0.0061, 0.0300]
Year 2011	0.0187	0.0625	0.30	0.765	[−0.104, 0.141]
Year 2012	−0.0456	0.0626	−0.73	0.466	[−0.168, 0.077]
Year 2013	−0.0321	0.0626	−0.51	0.609	[−0.155, 0.091]
Year 2014	−0.0407	0.0627	−0.65	0.516	[−0.164, 0.082]
Year 2015	−0.1227	0.0627	−1.96	0.051	[−0.246, 0.000]
Year 2016	−0.0933	0.0630	−1.48	0.139	[−0.217, 0.030]
Year 2017	−0.1193	0.0633	−1.88	0.060	[−0.243, 0.005]
Year 2018	−0.1322	0.0635	−2.08	0.037	[−0.257, −0.008]
Year 2019	−0.1603	0.0638	−2.51	0.012	[−0.285, −0.035]
Year 2020	−0.1944	0.0646	−3.01	0.003	[−0.321, −0.068]
Year 2021	−0.2122	0.0645	−3.29	0.001	[−0.339, −0.086]
Year 2022	−0.2430	0.0650	−3.74	0.000	[−0.371, −0.115]

Note: Fixed-effects estimation with firm-level clustering for 208 firms (N = 208, T = 12). Year dummies (2011–2022) are included to capture macroeconomic and regulatory shocks such as the 2018 exchange-rate crisis, the COVID-19 pandemic, and the IMO 2020 sulfur-cap regulation. The base year (2010) is omitted to avoid the dummy-variable trap. The joint significance test of time effects (testparm i.year) yields Prob > F = 0.000 < 0.05, confirming the overall importance of year-specific controls.

Table A11. Robustness check using winsorized sample—Model I (Maintenance sector).

Variable	Coef.	Std. Err.	t	p > |t|	95% Confidence Interval
OROA_W	1.56018	0.31957	4.88	0.000	[0.8638, 2.2564]
NCATA_W	−0.03616	1.92492	−0.02	0.985	[−4.2302, 4.1578]
QR_W	−0.00201	0.00387	−0.52	0.613	[−0.0104, 0.0064]
CR_W	−0.00651	0.01246	−0.52	0.611	[−0.0336, 0.0206]
EER_W	−0.99103	0.20804	−4.76	0.000	[−1.4443, −0.5377]
ROS_W	−0.15857	0.10663	−1.49	0.163	[−0.3909, 0.0737]
OMR_W	−0.10145	0.08345	−1.22	0.248	[−0.2838, 0.0809]
COGSR_W	−0.08796	0.05710	−1.53	0.151	[−0.2120, 0.0368]
ROE_W	−0.00518	0.00225	−2.30	0.040	[−0.0100, −0.0002]
NCFCL_W	−0.01544	0.01089	−1.42	0.182	[−0.0391, 0.0082]
EGR_W	−0.00798	0.00316	−2.52	0.027	[−0.0148, −0.0010]
CAR_W	0.01562	0.07869	0.20	0.846	[−0.1558, 0.1870]
2010	0 (empty)
2011	0.02552	0.02065	1.24	0.240	[−0.0194, 0.0705]
2012	−0.00427	0.00994	−0.43	0.675	[−0.0259, 0.0173]
2013	0.03779	0.00494	7.65	0.000	[0.0270, 0.0485]
2014	−0.00248	0.00992	−0.25	0.806	[−0.0214, 0.0191]
2015	−0.00752	0.01268	−0.59	0.564	[−0.0351, 0.0200]
2016	0.00346	0.00955	0.36	0.723	[−0.0173, 0.0242]
2017	−0.00921	0.01195	−0.77	0.456	[−0.0352, 0.0168]
2018	−0.00874	0.01092	−0.80	0.439	[−0.0325, 0.0150]
2019	0.00015	0.01674	0.01	0.993	[−0.0363, 0.0366]
2020	0.00681	0.01797	0.38	0.711	[−0.0323, 0.0459]
2021	0.02527	0.02876	0.88	0.397	[−0.0374, 0.0879]
2022	0.02845	0.02967	0.96	0.357	[−0.0362, 0.0931]
_cons	−0.01685	0.02764	−0.61	0.554	[−0.0770, 0.0433]

Table A11. Robustness check using winsorized sample—Model I (Maintenance sector).

Variable	Coef.	Std. Err.	t	p > |t|	95% Confidence Interval
OROA_W	1.56018	0.31957	4.88	0.000	[0.8638, 2.2564]
NCATA_W	−0.03616	1.92492	−0.02	0.985	[−4.2302, 4.1578]
QR_W	−0.00201	0.00387	−0.52	0.613	[−0.0104, 0.0064]
CR_W	−0.00651	0.01246	−0.52	0.611	[−0.0336, 0.0206]
EER_W	−0.99103	0.20804	−4.76	0.000	[−1.4443, −0.5377]
ROS_W	−0.15857	0.10663	−1.49	0.163	[−0.3909, 0.0737]
OMR_W	−0.10145	0.08345	−1.22	0.248	[−0.2838, 0.0809]
COGSR_W	−0.08796	0.05710	−1.53	0.151	[−0.2120, 0.0368]
ROE_W	−0.00518	0.00225	−2.30	0.040	[−0.0100, −0.0002]
NCFCL_W	−0.01544	0.01089	−1.42	0.182	[−0.0391, 0.0082]
EGR_W	−0.00798	0.00316	−2.52	0.027	[−0.0148, −0.0010]
CAR_W	0.01562	0.07869	0.20	0.846	[−0.1558, 0.1870]
2010	0 (empty)
2011	0.02552	0.02065	1.24	0.240	[−0.0194, 0.0705]
2012	−0.00427	0.00994	−0.43	0.675	[−0.0259, 0.0173]
2013	0.03779	0.00494	7.65	0.000	[0.0270, 0.0485]
2014	−0.00248	0.00992	−0.25	0.806	[−0.0214, 0.0191]
2015	−0.00752	0.01268	−0.59	0.564	[−0.0351, 0.0200]
2016	0.00346	0.00955	0.36	0.723	[−0.0173, 0.0242]
2017	−0.00921	0.01195	−0.77	0.456	[−0.0352, 0.0168]
2018	−0.00874	0.01092	−0.80	0.439	[−0.0325, 0.0150]
2019	0.00015	0.01674	0.01	0.993	[−0.0363, 0.0366]
2020	0.00681	0.01797	0.38	0.711	[−0.0323, 0.0459]
2021	0.02527	0.02876	0.88	0.397	[−0.0374, 0.0879]
2022	0.02845	0.02967	0.96	0.357	[−0.0362, 0.0931]
_cons	−0.01685	0.02764	−0.61	0.554	[−0.0770, 0.0433]

Table A12. Robustness check using winsorized sample—Model II (Shipping sector).

Variable	Coef.	Std. Err.	t	p > |t|	95% Confidence Interval
OROA_W	1.62445	0.19760	8.22	0.000	[1.1939, 2.0549]
NO_W	−0.00168	0.01144	−0.15	0.885	[−0.0266, 0.0232]
NCATA_W	−2.31210	6.84352	−0.36	0.741	[−17.2228, 12.5986]
TATR_W	−0.04131	0.03098	−1.33	0.207	[−0.1088, 0.0261]
EER_W	−1.16000	0.09419	−12.31	0.000	[−1.3652, −0.9547]
ROE_W	−0.01458	0.02460	−0.59	0.564	[−0.0681, 0.0390]
NCFCI_W	−0.03692	0.01793	−2.06	0.062	[−0.0760, 0.0021]
EGR_W	0.01320	0.01161	1.14	0.278	[−0.0121, 0.0385]
FATR_W	0.00016	0.00011	1.37	0.195	[−0.0000, 0.0004]
NPGR_W	−0.00031	0.00072	−0.43	0.672	[−0.0018, 0.0012]
FDTA_W	−0.22093	0.07665	−2.88	0.014	[−0.3879, −0.0539]
2010	0 (empty)
2011	−0.05629	0.01542	−3.65	0.003	[−0.0899, −0.0226]
2012	−0.08754	0.00906	−9.66	0.000	[−0.1072, −0.0677]
2013	−0.14486	0.00629	−23.02	0.000	[−0.1585, −0.1311]
2014	−0.14189	0.00321	−44.10	0.000	[−0.1489, −0.1348]
2015	−0.13603	0.00820	−16.58	0.000	[−0.1539, −0.1181]
2016	−0.13428	0.01499	−8.95	0.000	[−0.1669, −0.1016]
2017	−0.09980	0.01382	−7.22	0.000	[−0.1299, −0.0696]
2018	−0.07793	0.02511	−3.10	0.009	[−0.1326, −0.0232]
2019	−0.06068	0.03123	−1.94	0.076	[−0.1287, 0.0073]
2020	−0.04094	0.03985	−1.03	0.325	[−0.1277, 0.0458]
2021	−0.00042	0.05761	−0.01	0.994	[−0.1259, 0.1251]
2022	−0.01341	0.05178	−0.26	0.800	[−0.1262, 0.0994]
_cons	−0.06421	0.09119	−0.70	0.495	[−0.2629, 0.1344]

Table A12. Robustness check using winsorized sample—Model II (Shipping sector).

Variable	Coef.	Std. Err.	t	p > |t|	95% Confidence Interval
OROA_W	1.62445	0.19760	8.22	0.000	[1.1939, 2.0549]
NO_W	−0.00168	0.01144	−0.15	0.885	[−0.0266, 0.0232]
NCATA_W	−2.31210	6.84352	−0.36	0.741	[−17.2228, 12.5986]
TATR_W	−0.04131	0.03098	−1.33	0.207	[−0.1088, 0.0261]
EER_W	−1.16000	0.09419	−12.31	0.000	[−1.3652, −0.9547]
ROE_W	−0.01458	0.02460	−0.59	0.564	[−0.0681, 0.0390]
NCFCI_W	−0.03692	0.01793	−2.06	0.062	[−0.0760, 0.0021]
EGR_W	0.01320	0.01161	1.14	0.278	[−0.0121, 0.0385]
FATR_W	0.00016	0.00011	1.37	0.195	[−0.0000, 0.0004]
NPGR_W	−0.00031	0.00072	−0.43	0.672	[−0.0018, 0.0012]
FDTA_W	−0.22093	0.07665	−2.88	0.014	[−0.3879, −0.0539]
2010	0 (empty)
2011	−0.05629	0.01542	−3.65	0.003	[−0.0899, −0.0226]
2012	−0.08754	0.00906	−9.66	0.000	[−0.1072, −0.0677]
2013	−0.14486	0.00629	−23.02	0.000	[−0.1585, −0.1311]
2014	−0.14189	0.00321	−44.10	0.000	[−0.1489, −0.1348]
2015	−0.13603	0.00820	−16.58	0.000	[−0.1539, −0.1181]
2016	−0.13428	0.01499	−8.95	0.000	[−0.1669, −0.1016]
2017	−0.09980	0.01382	−7.22	0.000	[−0.1299, −0.0696]
2018	−0.07793	0.02511	−3.10	0.009	[−0.1326, −0.0232]
2019	−0.06068	0.03123	−1.94	0.076	[−0.1287, 0.0073]
2020	−0.04094	0.03985	−1.03	0.325	[−0.1277, 0.0458]
2021	−0.00042	0.05761	−0.01	0.994	[−0.1259, 0.1251]
2022	−0.01341	0.05178	−0.26	0.800	[−0.1262, 0.0994]
_cons	−0.06421	0.09119	−0.70	0.495	[−0.2629, 0.1344]

This appendix summarizes the inference robustness analyses addressing the reviewer’s comments. We conduct three complementary checks for both models (Maintenance and Shipping), with results reported in Table A13. First, we re-estimate the two-way fixed-effects models with two-way clustered standard errors (firm × year). Second, we trim the top and bottom 1% of the key explanatory variables (EER, and TATR for the shipping model) to limit the impact of outliers. Third, we perform a placebo test by shuffling year labels within firms. As shown in Table A13, the main coefficients remain positive and significant across all checks, confirming that the results are robust to clustering choice, outlier influence, and randomization.

Table A13. Inference Robustness Checks.

Panel	Specification	Sample/Adj.	Clustered SE	EER β (p)	TATR β (p)	N
A. Maintenance Model I	Baseline FE–FE	Full (2010–2022)	Firm and Year (two-way)	0.712 (0.0014)	—	2470
	Trimmed FE–FE	Drop top/bottom 1% of EER	Firm and Year (two-way)	[β_trim]	[β_trim]	—
	Placebo FE–FE	Year shuffled within firms	Conservative (max firm and shuf-year)	0.819 (<0.001)	—	2420
B. Shipping Model II	Baseline FE–FE	Full	Firm and Year (two-way)	0.784 (0.00001)	0.053 (0.0216)	2704
	Trimmed FE–FE	Drop top/bottom 1% of EER and TATR	Firm and Year (two-way)	0.562 (0.0140)	0.082 (0.0074)	2272
	Placebo FE–FE	Year shuffled within firms	Conservative (max firm and shuf-year)	0.556 (0.004)	0.082 (0.002)	2265

Table A13. Inference Robustness Checks.

Panel	Specification	Sample/Adj.	Clustered SE	EER β (p)	TATR β (p)	N
A. Maintenance Model I	Baseline FE–FE	Full (2010–2022)	Firm and Year (two-way)	0.712 (0.0014)	—	2470
	Trimmed FE–FE	Drop top/bottom 1% of EER	Firm and Year (two-way)	[β_trim]	[β_trim]	—
	Placebo FE–FE	Year shuffled within firms	Conservative (max firm and shuf-year)	0.819 (<0.001)	—	2420
B. Shipping Model II	Baseline FE–FE	Full	Firm and Year (two-way)	0.784 (0.00001)	0.053 (0.0216)	2704
	Trimmed FE–FE	Drop top/bottom 1% of EER and TATR	Firm and Year (two-way)	0.562 (0.0140)	0.082 (0.0074)	2272
	Placebo FE–FE	Year shuffled within firms	Conservative (max firm and shuf-year)	0.556 (0.004)	0.082 (0.002)	2265