Financial Development and Labour Productivity in G7 Economies: The Conditional Roles of FDI and Capital Formation

Abstract

This study investigates whether the long-run relationship between financial development and labour productivity in the G7 economies is conditional on foreign direct investment (FDI) and gross fixed capital formation. Using a balanced panel of seven countries over 2000–2024, the analysis employs panel fully modified ordinary least squares (FMOLS), dynamic ordinary least squares (DOLS), and canonical cointegrating regression (CCR), complemented by the common correlated effects mean group (CCEMG) and augmented mean group (AMG) estimators to assess robustness to cross-sectional dependence and slope heterogeneity. The results reveal substantial estimator sensitivity in the finance-productivity relationship: domestic credit to the private sector is positively associated with labour productivity under FMOLS and CCR but negatively associated under DOLS. By contrast, digital connectivity and research and development exhibit comparatively consistent positive associations with productivity, whereas gross fixed capital formation is negatively associated and FDI displays no robust unconditional relationship. Interaction estimates provide limited evidence that the finance-productivity association depends on the investment environment. The credit-FDI interaction is negative and statistically significant under FMOLS but not robust across alternative estimators, while the credit-capital formation interaction becomes insignificant when both interaction terms are included. Further robustness analyses underscore considerable cross-country heterogeneity. Overall, the findings qualify conventional expectations regarding the productivity-enhancing role of financial development and suggest that its long-run association with productivity is sensitive to model specification and country-specific conditions. By distinguishing relatively robust correlates from estimator-dependent results, this study contributes nuanced evidence on the conditional and heterogeneous finance-productivity nexus in advanced economies. The findings are interpreted as long-run associations rather than causal effects.

1. Introduction

Labour productivity is a central determinant of long-term economic performance because sustained improvements in output per worker underpin higher incomes, international competitiveness, and living standards (Salimova et al., 2021). Its importance has increased as advanced economies face demographic ageing, slower labour-force growth, and weaker productivity gains. Recent evidence indicates that productivity growth across OECD economies has slowed markedly relative to previous decades (Broberg et al., 2026). Although labour productivity levels in virtually all OECD economies were higher in 2024 than two decades earlier, the pace of improvement has approximately halved since the early 2000s. Had pre-2007 trends continued, labour productivity levels in 2024 could have been approximately 10–20% higher. These developments renew interest in the mechanisms through which advanced economies can mobilize finance, investment, technology, and productive capital to support productivity growth (Fernald et al., 2024; Goldin et al., 2024).

The G7 economies-Canada, France, Germany, Italy, Japan, the United Kingdom, and the United States-provide a relevant setting for examining these mechanisms. Despite their relatively developed financial systems and technological capabilities, these economies exhibit differences in productivity trajectories, investment structures, and economic performance (Sekmen et al., 2026; Sha, 2023). Recent differences in productivity growth, including the stronger performance of the United States relative to several European economies, illustrate that high levels of financial development do not necessarily translate into comparable productivity outcomes. These differences are particularly important in an environment characterized by changing financing conditions, geopolitical uncertainty, and evolving investment patterns. Although financial conditions have eased in some advanced economies following the monetary tightening associated with post-pandemic inflation, uncertainty surrounding trade policy, fiscal positions, financial markets, and geopolitical developments continues to affect investment decisions (Clarida, 2025). The close integration of G7 economies through trade, international financial markets, and multinational production networks further means that common shocks may influence national productivity and investment outcomes simultaneously.

Financial development may support labour productivity by mobilizing savings, improving capital allocation, reducing financing constraints, facilitating investment, and enabling firms to adopt new technologies. Financial institutions can improve information production and risk diversification, allowing firms to undertake productive projects that might otherwise remain financially constrained. Nevertheless, the relationship between financial development and productivity is not necessarily positive or uniform. Domestic credit to the private sector measures the scale of financial intermediation but does not directly capture the efficiency or destination of lending. Credit expansion may support productive investment, but it may also finance relatively low-productivity activities, excessive leverage, or asset price-driven expansion (Azevedo et al., 2021). Consequently, the productivity implications of financial development may depend on how financial resources are allocated and whether they complement productive investment and technological upgrading (Magazzino & Santeramo, 2023).

These considerations have become increasingly relevant as production systems incorporate digital technologies, artificial intelligence, automation, and intangible assets. Such developments are changing both the composition of productive capital and the channels through which financial resources may affect productivity (Zhang & Piao, 2025). Investment in software, data, digital infrastructure, and complementary organizational capabilities may be particularly important for realizing the productivity benefits of technological change. However, the distribution of digital investment and its associated productivity gains remains uneven across advanced economies. Financial development should therefore be considered in relation to the broader productive environment rather than treated as an independent guarantee of higher productivity.

Two potential conditioning factors are foreign direct investment (FDI) and domestic capital accumulation. FDI can affect labour productivity through capital formation, technology transfer, managerial practices, international production networks, competition, and knowledge spillovers (Barboza et al., 2025). Multinational enterprises may introduce technologies and organizational practices that improve production efficiency and connect domestic firms to international markets and supply chains (Teece, 2025). However, these benefits are not automatic: their realization may depend on domestic technological capabilities, labour skills, absorptive capacity, and linkages between foreign and domestic firms. Financial development may influence the capacity of domestic firms to engage with foreign investors and absorb transferred technologies, while FDI may expand the range of investment opportunities and productive activities available to domestic financial systems (Pal & Bandyopadhyay, 2022). These mechanisms provide a rationale for examining whether the association between financial development and labour productivity varies with FDI, while leaving the direction and magnitude of that relationship to empirical investigation.

Gross fixed capital formation (GFCF) provides a complementary perspective by capturing investment in fixed assets. Financial development may ease credit constraints and facilitate investment, but the productivity consequences of finance may depend on whether financial resources are associated with the accumulation of productive capacity. This issue is particularly relevant when firms require complementary investments in machinery, infrastructure, software, and other assets to benefit from technological progress. As also emphasized in the closely related study by Dwivedi (2026), the relationship between financial development and labour productivity warrants consideration alongside FDI and capital accumulation. Accordingly, the present study does not claim that introducing these two moderators constitutes an entirely new conceptual approach. Instead, it examines their conditional associations with labour productivity within a focused G7 panel and evaluates how the resulting estimates vary across alternative long-run and heterogeneous-panel estimators. This positioning places the emphasis on the empirical consistency and interpretation of the relationships rather than on an unsupported claim of conceptual priority.

The G7 setting also makes it important to account for cross-sectional dependence and heterogeneous country responses. These economies are exposed to common developments in international financial conditions, trade, energy prices, geopolitical risks, technological change, and global demand. Such shocks may generate correlated movements across countries and complicate the interpretation of panel estimates if they are not adequately considered. At the same time, differences in national financial structures, industrial composition, investment patterns, and technological capabilities may produce heterogeneous responses to financial development. These issues are especially pertinent in the context of slower potential growth, demographic pressures, and the need to strengthen investment and technological diffusion. OECD assessments place potential GDP per capita growth across OECD economies at approximately 1.3%, compared with around 2.4% in the 1990s, reinforcing the importance of understanding the conditions associated with sustained productivity improvement.

Against this background, the research gap is empirical and methodological rather than a claim that the finance-productivity relationship or its investment-related moderators have not previously been studied. First, the sign and magnitude of the long-run association between financial development and labour productivity require assessment across alternative cointegration estimators rather than reliance on a single specification. Second, although related work, including Dwivedi (2026), considers FDI and capital accumulation as potential moderators, the consistency of these conditional associations across alternative long-run estimators and approaches designed to accommodate common factors and heterogeneous country responses remains an important issue for the G7 context. Third, the close integration of these economies creates a need to assess whether the results remain broadly consistent when cross-sectional dependence and slope heterogeneity are explicitly considered. These issues motivate a comparative analysis that distinguishes findings that are relatively stable across estimation strategies from those that are sensitive to model specification.

Accordingly, this study examines the long-run association between financial development and labour productivity in seven G7 economies over 2000–2024, focusing on the potential moderating roles of FDI and GFCF. Labour productivity is measured as GDP per person employed, and financial development is proxied by domestic credit to the private sector provided by banks as a percentage of GDP. FDI and GFCF enter the empirical model as both direct explanatory variables and interaction terms with financial development. The analysis also controls for trade openness, internet use, research and development expenditure, inflation, and human capital. Four research questions guide the investigation: (1) Is financial development associated with labour productivity in the G7 economies in the long run? (2) Does the estimated association vary with FDI? (3) Does it vary with GFCF? (4) How sensitive are these relationships to alternative long-run estimators and approaches that account for cross-sectional dependence and heterogeneous country responses?

To address these questions, the study employs fully modified ordinary least squares (FMOLS), dynamic ordinary least squares (DOLS), and canonical cointegrating regression (CCR) to estimate long-run relationships. The common correlated effects mean group (CCEMG) and augmented mean group (AMG) estimators provide complementary assessments that account for unobserved common factors and heterogeneous country responses. This combination enables a comparative evaluation of whether the estimated direct and conditional relationships are robust across alternative empirical approaches. The estimates are interpreted as long-run associations; the econometric design does not, by itself, establish causal effects.

The study’s contribution is deliberately bounded. First, it provides a comparative assessment of the long-run finance-productivity association in the G7 using three conventional cointegration estimators and two approaches designed to address common factors and heterogeneous responses. Second, it evaluates whether the associations involving financial development and labour productivity vary with FDI and GFCF, while explicitly recognizing the overlap of this research question with prior work, including Dwivedi (2026). Third, it distinguishes comparatively consistent findings from estimator-sensitive interaction results and country-level heterogeneity. The contribution therefore lies in the comparative empirical assessment and qualification of these relationships in the G7 setting, rather than in claiming that the proposed moderators or their underlying theoretical mechanisms are novel. This approach provides a more cautious basis for discussing the role of finance in productivity performance across advanced economies.

The remainder of the paper is organized as follows. Section 2 reviews the theoretical and empirical literature and develops the research hypotheses. Section 3 describes the data and variables, and Section 4 presents the econometric methodology. Section 5 reports the empirical results, including diagnostic tests, long-run estimates, moderation analysis, and robustness checks. Section 6 discusses the findings and their implications, Section 7 presents policy recommendations, and Section 8 concludes the study.

2. Literature Review

2.1. Conceptual Foundations: Financial Development and Labour Productivity

The relationship between financial development and economic performance is traditionally examined through the finance-growth literature, which emphasizes the role of financial systems in mobilizing savings, allocating capital, facilitating investment, diversifying risk, and reducing information and transaction costs. However, aggregate economic growth does not necessarily translate into improvements in productive efficiency, as GDP growth may reflect increases in employment or factor accumulation rather than more efficient combinations of labour and capital. Labour productivity, measured as output per employed person, therefore provides a more direct outcome for assessing whether financial development is associated with improvements in productive capacity and technological performance.

Theoretically, financial development may be associated with labour productivity through two interconnected mechanisms. The first is capital accumulation: access to credit and improved financial intermediation can ease financing constraints and support investment in physical and intangible assets. The second is resource allocation efficiency: financial institutions can improve information production, evaluate investment opportunities, diversify risk, and direct resources toward firms and projects with higher expected productivity. Finance may consequently influence not only the amount of capital available to workers but also the efficiency with which productive resources are allocated and used (Dwivedi, 2026; Agu et al., 2025).

Nevertheless, financial depth should not be equated with financial efficiency. Domestic credit to the private sector captures the scale of financial intermediation but does not directly indicate whether credit supports productive investment. Credit expansion can finance technological upgrading and productive capacity, but it can also encourage excessive leverage, asset price expansion, or investment in activities with limited productivity-enhancing potential. The relationship between financial development and labour productivity is therefore theoretically ambiguous, particularly when financial deepening occurs without corresponding improvements in the allocation of capital.

This distinction is also relevant to the relationship between finance, capital accumulation, and total factor productivity (TFP). Financial integration may improve productivity through more efficient resource allocation even when its direct contribution to capital accumulation is limited; conversely, financial crises can damage both investment and productivity (Bonfiglioli, 2006). Financial development should thus not be evaluated solely by whether it expands the capital stock. Its productivity implications also depend on which firms and technologies receive financing and how effectively the resulting resources are transformed into output.

These mechanisms may operate differently across stages of economic and financial development. In economies with relatively underdeveloped financial systems, expanding financial intermediation can relax substantial financing constraints and facilitate basic capital accumulation. In advanced economies, where firms generally have broader access to financial markets and capital stocks are already relatively large, the marginal productivity contribution of additional financial depth may be weaker. Under these circumstances, the quality of credit allocation and the financing of innovation, technological adoption, and organizational improvements may be more important than the expansion of credit alone. Evidence from high-income economies supports this intensive-margin interpretation, indicating that financial development can increasingly operate through innovation and productivity-related channels rather than basic capital accumulation (Radmehr et al., 2022).

Recent research on labour productivity further suggests that the finance-productivity relationship varies across countries and economic contexts. Dwivedi (2026) directly examines financial development and labour productivity while incorporating the complementary roles of FDI and capital formation. Related evidence also associates finance, capital formation, and FDI with productivity, although the magnitude and direction of these relationships depend on institutional conditions, absorptive capacity, and national economic characteristics (Li & Tanna, 2019; Anyachebelu, 2025). Collectively, this literature suggests that financial development may be productivity-enhancing, but that its relationship with labour productivity cannot be assumed to be uniformly positive.

2.2. Financial Development in the G7 and Other Advanced Economies

The G7 economies-Canada, France, Germany, Italy, Japan, the United Kingdom, and the United States-provide a useful setting for examining the conditional relationship between financial development and labour productivity. Although these economies share relatively high levels of development, sophisticated financial systems, and extensive integration into international markets, they differ in financial structure, industrial composition, investment patterns, technological capabilities, and productivity performance. This combination of broadly comparable development levels and meaningful structural heterogeneity allows the study to investigate whether the productivity implications of financial development are consistent across advanced economies.

Evidence specific to the G7 challenges the assumption that deeper financial systems necessarily generate stronger productivity outcomes. Garg (2025) finds that the relationship between financial development and sectoral productivity differs between emerging economies and the G7. Whereas financial development is associated with positive productivity effects in emerging economies, the corresponding estimates for the G7 are negative or statistically insignificant, with additional variation between manufacturing and market services. These findings suggest that the relationship between finance and productivity may depend on both the maturity of financial systems and the sectors receiving financial resources.

Threshold-based studies provide a possible explanation for these differences. Osei and Kim (2020) find that the growth-enhancing effects of financial development weaken beyond higher levels of financial depth and report that the positive effect of FDI becomes negligible when private sector credit exceeds approximately 95.6% of GDP. Dua and Verma (2024) similarly discuss the possibility of a “vanishing effect of finance,” whereby additional financial development no longer produces proportional improvements in economic performance after financial systems reach sufficiently high levels of depth. Although these findings are not confined to the G7, they provide a basis for expecting the marginal association between financial depth and productivity to be weaker in mature financial systems.

Importantly, a negative or statistically insignificant coefficient on domestic credit should not automatically be interpreted as evidence that financial intermediation is intrinsically harmful. Rather, such an estimate may indicate that further credit expansion does not necessarily translate into additional productive capacity when financial systems are already developed. The distinction between the scale and the allocation of finance is therefore central to interpreting the finance-productivity relationship in advanced economies.

Research on high-income countries reinforces this interpretation. Radmehr et al. (2022) report a positive long-run relationship between FDI, capital, and growth in high-income economies and emphasize the importance of innovation and productivity-enhancing activities in understanding the relationship. Their findings imply that, in advanced economies, financial development may contribute to productivity by supporting innovation, intangible investment, technological adoption, and organizational improvements rather than merely expanding conventional physical capital.

The G7 evidence also indicates that financial development operates within an interconnected international investment and institutional environment. Belgacem et al. (2024) find that G7 outward investors are more likely to invest in host economies with stronger financial systems, better institutional quality, and lower institutional frictions, particularly when investment is motivated by asset-seeking considerations involving technology, trademarks, and intellectual property. Sha (2023) similarly identifies joint associations among FDI, financial development, innovation, trade openness, and economic growth in G7 economies. These findings indicate that finance, international investment, and innovation should be considered interconnected components of the productive environment rather than isolated determinants.

Taken together, the literature establishes a distinction between the expected benefits of financial development and the conditions under which those benefits materialize. For the G7, the relevant question is not simply whether financial development is associated with productivity, but whether that association depends on complementary investment and the broader economic environment.

2.3. FDI, Financial Development, and Productivity: The Moderating Role of Foreign Investment

FDI may contribute to labour productivity by facilitating technology transfer, introducing managerial and organizational knowledge, connecting domestic firms to international production networks, increasing competitive pressure, and supporting human-capital development (Khalikchaeva et al., 2026). Multinational enterprises can introduce technologies and production practices that domestic firms might otherwise adopt more slowly. These channels underpin the expectation that FDI can support productivity improvements (Majeed et al., 2021; Emako et al., 2022; Nosova, 2022; Anyachebelu, 2025).

However, foreign investment does not automatically generate productivity spillovers. Domestic firms must possess the capacity to absorb, adapt, and use the technologies and managerial practices introduced by foreign investors. Financial development may influence this process by providing domestic firms with financing for complementary investment, technological adoption, workforce development, and participation in international production networks. In this sense, financial development and FDI may be complementary: FDI can expand access to knowledge and productive opportunities, while domestic financial systems can help firms mobilize the resources needed to benefit from them.

Alfaro et al. (2009) provide important evidence for this mechanism, finding that countries with more developed financial markets benefit from FDI primarily through improvements in TFP rather than through physical- or human-capital accumulation. Their results suggest that the finance-FDI relationship can operate through efficiency-enhancing channels. Financial development may help domestic firms respond to foreign investment opportunities, while FDI can introduce technologies and management practices that increase the productivity of the resources financed domestically.

The effectiveness of these complementary mechanisms also depends on institutional and absorptive conditions. Li and Tanna (2019) find that the direct effect of FDI on TFP may be weak but becomes more robust when institutional and human-capital conditions are considered. Anyachebelu (2025) likewise emphasizes the importance of domestic economic and institutional conditions in shaping the productivity consequences of FDI. Related studies identify substantial variation in the productivity implications of FDI, reflecting differences in absorptive capacity and the characteristics of foreign investment (Emako et al., 2022; Nosova, 2022; Dua & Verma, 2024).

Nevertheless, a positive moderating relationship is not inevitable. Dwivedi (2026) finds that FDI negatively moderates the relationship between financial development and labour productivity on average, although the dampening effect is stronger in developing than in developed economies. This result suggests that FDI may sometimes substitute for domestic financial intermediation, direct productivity gains toward foreign-owned firms, or produce limited additional benefits in economies where financial systems are already mature. The composition of FDI may also matter: greenfield investment can create new productive capacity, whereas mergers and acquisitions may operate through different channels and need not generate equivalent additions to productive assets. Emako et al. (2022) and Nosova (2022) emphasize that FDI outcomes depend partly on investment type and sectoral orientation. In advanced economies, the strength of productivity spillovers may also depend on technology-intensive ownership advantages and intra-industry knowledge transfer.

The synthesis of this evidence yields a conditional proposition. FDI may strengthen the association between financial development and labour productivity when foreign investment creates opportunities for technology transfer and domestic firms have the financial capacity to exploit them. However, the strength of this complementarity may be limited where foreign investment substitutes for domestic financing or where its productivity benefits remain concentrated. Given the evidence of positive absorptive capacity mechanisms and the potential for negative moderation, a positive interaction is a theoretically grounded expectation to be tested rather than an established outcome.

2.4. Capital Accumulation and the Financial Development-Productivity Relationship

Capital accumulation provides a second potential mechanism through which financial development may be associated with labour productivity. By mobilizing savings, easing financing constraints, improving risk-sharing, and expanding access to external finance, financial systems can facilitate investment in productive assets. Gross fixed capital formation (GFCF) captures investment in fixed assets and therefore provides an indicator of the extent to which financial and other resources are translated into productive capacity. Recent research associates capital formation with labour productivity and broader economic performance, although the strength of this relationship varies across economic and institutional settings (Dwivedi, 2026; Zoaka & Güngör, 2023; Anyachebelu, 2025).

The relationship between finance and capital accumulation is likely to differ according to the level of financial development. In economies where financial markets are relatively shallow, improvements in financial intermediation can relax substantial financing constraints and facilitate investment that would otherwise not occur. Evidence from Sub-Saharan Africa indicates positive associations among financial development, capital accumulation, and labour productivity, illustrating the importance of investment and financial access in settings where constraints on productive capital may be substantial (Zoaka & Güngör, 2023).

In advanced economies, however, the productivity implications of additional financial depth may depend more heavily on the composition and productivity of investment. Firms typically have access to a broader range of financing sources and operate with larger existing capital stocks. Under these conditions, additional credit may have limited productivity implications unless it supports investment that enhances productive efficiency. Radmehr et al. (2022) emphasize the growing importance of innovation-oriented activities in high-income economies, suggesting that finance may be increasingly relevant to the quality and technological content of investment rather than simply to the quantity of physical capital.

This argument is particularly pertinent to technological transformation. Labour productivity can depend on complementary investment in machinery, information and communication technologies, software, data, infrastructure, research, and organizational capabilities. The presence of productive capital may therefore determine whether firms can transform financial resources into higher output per worker. GFCF can consequently be considered not only an independent determinant of productivity but also a potential moderator of the financial development-productivity relationship.

Evidence from Dwivedi (2026) directly supports this proposition, indicating that capital formation positively moderates the relationship between financial development and labour productivity in developed economies. This finding implies that financial resources may be more strongly associated with productivity when accompanied by investment in productive assets. In theoretical terms, financial development provides access to resources, whereas capital formation provides a channel through which those resources can be embodied in productive capacity.

At the same time, capital accumulation does not guarantee that financial deepening will improve productivity. Agu et al. (2025) report a negative productivity effect of financial development in Nigeria despite positive effects associated with FDI, emphasizing the importance of regulatory quality and the measurement of financial development. Siddikee and Rahman (2020) identify a negative long-run relationship between financial development and GDP in Bangladesh and discuss inefficient credit use and weaknesses in banking-sector quality as potential explanations. Although these studies focus on developing economies, they reinforce the general point that the productivity implications of financial development depend on how financial resources are allocated and used. L. Olorogun et al. (2020) similarly show that FDI and banking-sector financial development can predict economic growth under stable long-run relationships, while broader institutional and macroeconomic conditions remain relevant.

Overall, the literature supports the expectation that capital accumulation can strengthen the productivity association of financial development when investment expands or improves productive capacity. The interaction between financial development and GFCF therefore provides a direct test of whether the relationship between finance and labour productivity is stronger in economies and periods characterized by higher levels of capital formation.

2.5. Financial Openness, Globalization, and the Broader Investment Environment

Financial openness and globalization provide additional context for understanding the finance-productivity relationship. Financial integration can widen firms’ access to capital, facilitate international risk-sharing, and support cross-border investment (Beruniy et al., 2026). However, the productivity implications of these processes need not be proportional to the amount of international capital available. Bonfiglioli (2006) found that financial integration can raise productivity without producing an equivalent direct increase in capital accumulation, while financial crises can adversely affect both investment and TFP. This evidence reinforces the distinction between capital deepening and efficiency-enhancing channels.

Evidence from developing economies also cautions against assuming that openness necessarily improves productive performance. Damasceno and Guedes (2024) found no evidence that financial openness systematically stimulates capital stock growth or productivity growth after accounting for institutional and macroeconomic conditions. The consequences of financial integration may therefore depend on domestic institutions, financial market structures, macroeconomic stability, and absorptive capacity.

These considerations are particularly relevant to the G7 because its economies are deeply integrated through trade, FDI, financial markets, and multinational production networks. Sha (2023) found that FDI, financial development, innovation, and trade openness are jointly associated with economic performance in G7 economies. Nevertheless, the effects of globalization may differ across structural and institutional settings, and some characteristics, including natural resource dependence, may weaken the expected benefits of international integration.

This literature informs the specification of the present study in two ways. First, it supports considering FDI within the broader international investment environment rather than treating it only as an unrelated control. Second, it motivates accounting for other factors that may be associated with labour productivity, including trade openness, internet use, research and development expenditure, inflation, and human capital. These controls help distinguish the focal finance-productivity relationship from associations involving the wider economic environment. They do not, however, eliminate the possibility that unobserved common shocks or country-specific differences influence the estimates.

2.6. Synthesis of the Literature and Research Gap

The reviewed literature points to three substantive conclusions. First, financial development may be associated with labour productivity through capital accumulation, improved resource allocation, technological adoption, and efficiency gains. However, the direction and magnitude of this association are not uniform. In mature financial systems, additional credit may yield limited productivity gains when it does not support productive investment or innovation (Garg, 2025; Osei & Kim, 2020; Dua & Verma, 2024).

Second, the relationship between financial development and labour productivity may depend on complementary forms of investment. FDI can provide technology, managerial knowledge, and access to international production networks, but its benefits depend on financial capacity, institutions, human capital, and domestic absorptive capacity (Alfaro et al., 2009; Li & Tanna, 2019; Majeed et al., 2021; Emako et al., 2022; Nosova, 2022; Anyachebelu, 2025). Moreover, the negative moderating effect reported by Dwivedi (2026) indicates that FDI cannot automatically be assumed to amplify the productivity implications of financial development. Capital accumulation offers a distinct potential complementarity because it provides a channel through which financial resources can be translated into productive assets (Radmehr et al., 2022; Dwivedi, 2026).

Third, the finance-productivity relationship must be considered within the broader context of globalization and the heterogeneous economic structures of advanced economies. The G7-specific literature examines financial development, FDI, innovation, institutions, and international integration, but these strands do not fully resolve whether the association between financial development and labour productivity is conditional on FDI and capital accumulation (Belgacem et al., 2024; Sha, 2023; Garg, 2025).

The present study addresses this issue by examining the relationship between financial development and labour productivity in the G7 and explicitly testing the moderating roles of FDI and GFCF. This positioning acknowledges that Dwivedi (2026) has already examined the finance-productivity relationship and the roles of these complementary factors. Accordingly, the contribution is not the claim that these moderators have never been studied; rather, it is the focused assessment of their conditional associations in the G7 using a complementary set of long-run estimators and robustness approaches.

The empirical strategy combines FMOLS, DOLS, and CCR to assess long-run relationships and uses CCEMG and AMG to examine the robustness of the results while accommodating cross-sectional dependence and heterogeneous country responses. This is relevant because G7 economies are interconnected and exposed to common global shocks, while their financial structures and productive systems remain heterogeneous. Estimates that do not adequately address these features may conflate common movements with country-specific relationships.

The study therefore examines three linked questions: whether financial development is positively associated with labour productivity in the G7; whether FDI strengthens this association; and whether capital accumulation strengthens it. These questions are evaluated as long-run conditional associations, not as causal effects. The hypotheses below follow from the dominant theoretical mechanisms while recognizing that the existing evidence-especially for financially mature economies-does not establish that the expected relationships must hold in every specification.

Sectoral heterogeneity is central to understanding labour productivity in the G7 economies. Baumol (1967) and Nordhaus (2006) explain how differences in sectoral productivity growth and the expansion of relatively stagnant activities can constrain aggregate productivity growth. Structural transformation further affects aggregate outcomes through changes in employment and value-added shares across agriculture, manufacturing, and services (Herrendorf et al., 2014; Rodrik, 2016). Empirical studies demonstrate that productivity convergence and growth vary across industries and countries (Bernard & Jones, 1996), with differences in ICT adoption and market services contributing to productivity divergence between the United States and Europe (O’Mahony & Timmer, 2009; Van Ark et al., 2008; Stiroh, 2002; Jorgenson et al., 2008). Moreover, services are internally heterogeneous (Duernecker et al., 2024), productivity growth has slowed across advanced economies (Bergeaud et al., 2016; Cette et al., 2016; Goldin et al., 2024), and productivity differences persist both between frontier and other firms and among firms within the same industry (Andrews et al., 2016; Syverson, 2011; Bartelsman & Doms, 2000). The productivity benefits of digital technologies may also depend on complementary intangible investment (Brynjolfsson et al., 2021; Kurbonov et al., 2023). Sectoral differences are equally relevant to financial development and FDI: industries vary in their dependence on external finance (Rajan & Zingales, 1998), financial development may influence convergence differently across development levels (Aghion et al., 2005), excessive financial depth may weaken growth outcomes (Arcand et al., 2015), and FDI benefits may depend on domestic financial development (Alfaro et al., 2004). Methodologically, common correlated effects estimators (Pesaran, 2006) and heterogeneous production-function approaches (Eberhardt & Teal, 2011) help account for common factors and cross-country heterogeneity. Taken together, this literature suggests that aggregate labour productivity reflects both within-sector developments and changes in sectoral composition. Therefore, the estimated relationships between financial development, FDI, capital accumulation, and labour productivity in this study should be interpreted as aggregate long-run associations that may conceal differences across agriculture, industry, and services, rather than as uniform sector-specific relationships.

2.7. Hypothesis Development

2.7.1. Financial Development and Labour Productivity

Financial development may raise labour productivity by easing financing constraints, improving capital allocation, and enabling investment in productive assets and technology. These mechanisms motivate the expectation of a positive association. Evidence from mature financial systems, however, suggests that additional financial depth can have weaker or even negative associations with productivity when credit is allocated inefficiently or does not support productivity-enhancing activities (Garg, 2025; Osei & Kim, 2020; Radmehr et al., 2022; Dua & Verma, 2024). For the G7 economies, a positive association is therefore treated as a testable baseline proposition, not as an assumption that credit expansion necessarily improves productivity.

H1. 

Financial development is positively associated with labour productivity in G7 economies.

H1 is evaluated using the coefficient on financial development ($F{D}_{it}$) in the baseline long-run specification, estimated by FMOLS, DOLS and CCR. CCEMG and AMG assess whether the result holds under cross-sectional dependence and heterogeneous country responses. A positive and statistically significant coefficient supports H1. An insignificant or negative coefficient does not.

2.7.2. Moderating Role of Foreign Direct Investment

A positive moderating role of FDI rests on the complementarity between domestic financial capacity and foreign investment. FDI can bring technologies, management practices and knowledge, and a developed domestic financial system may help local firms finance complementary investment and absorb these capabilities (Alfaro et al., 2009). Evidence on institutional quality, human capital and absorptive capacity also shows that the productivity effects of FDI depend on the conditions under which it operates (Li & Tanna, 2019; Emako et al., 2022; Nosova, 2022; Dua & Verma, 2024). This implies that the productivity association of financial development may be stronger at higher FDI levels.

An important qualification comes from Dwivedi (2026), who finds that FDI can negatively moderate the finance-productivity relationship. FDI may substitute for domestic financial intermediation, and its benefits may be limited when domestic firms do not capture the spillovers. The expected positive moderation is therefore a theoretically motivated hypothesis to be tested, not a predetermined result.

H2. 

FDI positively moderates the relationship between financial development and labour productivity in G7 economies, such that the association becomes stronger at higher levels of FDI.

H2 is tested through the interaction term $F{D}_{it}\times FD{I}_{it}$ in the extended specification, which also includes $F{D}_{it}$, $FD{I}_{it}$ and the controls. The conditional association is

$${\displaystyle \frac{\partial L{P}_{it}}{\partial F{D}_{it}}}={\beta }_1+{\beta }_3\hspace{0.17em}FD{I}_{it},$$

where $L{P}_{it}$ is labour productivity and ${\beta }_3$ is the interaction coefficient. H2 predicts ${\beta }_3>0$. A positive and significant ${\beta }_3$ supports H2. A negative one indicates moderation in the opposite direction, and an insignificant one provides no evidence of systematic moderation.

2.7.3. Moderating Role of Capital Accumulation

Capital accumulation may strengthen the productivity association of financial development because financial resources are more likely to raise output per worker when they finance productive assets. Financial development mobilises savings and eases financing constraints, while gross fixed capital formation (GFCF) is the channel through which resources become physical productive capacity. This is particularly relevant where productivity depends on complementary investment in machinery, infrastructure and technology.

The expectation is consistent with evidence that capital formation positively moderates the finance-productivity relationship in developed economies (Dwivedi, 2026) and with research linking capital formation to productivity and economic performance (Zoaka & Güngör, 2023; Anyachebelu, 2025; Radmehr et al., 2022). However, evidence of inefficient financial allocation and negative financial development associations shows that greater financial depth does not always translate into more productive capital (Agu et al., 2025; Siddikee & Rahman, 2020; L. Olorogun et al., 2020). The proposition therefore concerns the interaction between finance and investment, not only the separate association of GFCF with productivity.

H3. 

Capital accumulation positively moderates the relationship between financial development and labour productivity in G7 economies, such that the association becomes stronger at higher levels of GFCF.

H3 is tested through $F{D}_{it}\times GFC{F}_{it}$ in the extended specification, with the conditional association

$${\displaystyle \frac{\partial L{P}_{it}}{\partial F{D}_{it}}}={\beta }_1+{\beta }_3\hspace{0.17em}GFC{F}_{it}.$$

H3 predicts ${\beta }_3>0$. The interpretation of positive, negative and insignificant coefficients follows H2.

2.7.4. Integrated Hypothesis Framework and Estimation Strategy

The three hypotheses form one framework. Financial development is the principal explanatory variable, and FDI and GFCF are examined as potential moderators. The baseline model assesses the direct association (H1), and two extended models assess whether it varies with FDI (H2) and with capital accumulation (H3).

Baseline specification:

$$L{P}_{it}={\alpha }_i+{\beta }_{1}F{D}_{it}+{\mathsf{\gamma }}^{\prime }{\mathrm{X}}_{it}+{\epsilon }_{it}$$

FDI moderation specification:

$$L{P}_{it}={\alpha }_i+{\beta }_{1}F{D}_{it}+{\beta }_{2}FD{I}_{it}+{\beta }_3\left(F{D}_{it}\times FD{I}_{it}\right)+{\mathsf{\gamma }}^{\prime }{\mathrm{X}}_{it}+{\epsilon }_{it}$$

Capital accumulation moderation specification:

$$L{P}_{it}={\alpha }_i+{\beta }_{1}F{D}_{it}+{\beta }_{2}GFC{F}_{it}+{\beta }_3\left(F{D}_{it}\times GFC{F}_{it}\right)+{\mathsf{\gamma }}^{\prime }{\mathrm{X}}_{it}+{\epsilon }_{it}$$

Here, ${\mathrm{X}}_{it}$ contains trade openness, internet use, research and development expenditure, inflation and human capital. ${\alpha }_i$ captures country-specific components, and ${\epsilon }_{it}$ is the error term. In each model, the interaction coefficient shows whether the association between financial development and labour productivity changes with the level of the moderator.

FMOLS, DOLS and CCR estimate the long-run relationships under alternative procedures, and CCEMG and AMG provide robustness estimates under cross-sectional dependence and heterogeneous country responses. Agreement across estimators would raise confidence in the stability of the associations. Differences would indicate sensitivity to estimation assumptions or cross-country heterogeneity. The hypotheses are evaluated on the signs and significance of the financial development coefficient and the two interaction coefficients. The framework therefore tests, and does not assume, that the expected relationships hold in the G7 economies.

3. Data

3.1. Data Sources and Sample

Table 1 summarizes the operationalization of the variables used in the empirical analysis, including their measurement, proxy indicators, expected analytical role, and data sources. The study employs annual data for the seven G7 economies-Canada, France, Germany, Italy, Japan, the United Kingdom, and the United States-for 2000–2024, producing a balanced panel of 175 country-year observations. The selection of the variables follows the theoretical framework developed in the preceding sections. Labour productivity is used as the dependent variable, while domestic credit provided by banks serves as the principal proxy for financial development. FDI and gross fixed capital formation are introduced as moderating variables, reflecting the external investment and domestic capital accumulation dimensions of the productivity process. Trade openness, Internet use, R&D expenditure, inflation, and human capital are included as control variables to account for international integration, digitalization, innovation, macroeconomic conditions, and labour quality differences.

Table 1. Variable Measurement and Data Sources.

Variable	Symbol	Role	Measurement/Proxy	Unit	Data Source	Period
Labour productivity	lnLP	Dependent variable	Natural logarithm of GDP per person employed, measured in constant-price terms	Log of constant-price output per employed person	World Bank WDI	2000–2024
Financial development	dcps	Main explanatory variable	Domestic credit provided by banks to the private sector, relative to GDP	% of GDP	World Bank WDI	2000–2024
Foreign direct investment	fdi	Moderator	Net inflows of foreign direct investment relative to GDP	% of GDP	World Bank WDI	2000–2024
Gross fixed capital formation	gfcp	Moderator	Gross fixed capital formation relative to GDP	% of GDP	World Bank WDI	2000–2024
Trade openness	trade	Control	Sum of exports and imports of goods and services relative to GDP	% of GDP	World Bank WDI	2000–2024
Internet use	internet	Control	Individuals using the Internet as a share of the population	% of population	World Bank WDI	2000–2024
Research and development	rd	Control	Research and development expenditure relative to GDP	% of GDP	World Bank WDI	2000–2024
Inflation	inf	Control	Annual inflation rate based on the consumer price index	Annual %	World Bank WDI	2000–2024
Human capital	hc	Control	Human capital index based on education and returns to education	Index	Penn World Table (PWT)	2000–2024

Notes: The sample consists of Canada, France, Germany, Italy, Japan, the United Kingdom, and the United States. All variables are observed annually over 2000–2024, yielding 175 observations. lnLP denotes the natural logarithm of labour productivity. FDI and gross fixed capital formation are used as moderators through their interaction with financial development.

3.2. Rationale for Variable Selection

Labour productivity (lnLP) is specified as the dependent variable and measured as output per employed person in natural logarithms, a choice that aligns the empirical specification with the theoretical focus on worker efficiency rather than aggregate output growth and is consistent with cross-country labour productivity studies that operationalize productivity as output or GDP per worker and estimate log-linear relationships (Javed, 2025; Haider et al., 2025; Rani & Kumar, 2019; Yeboah et al., 2025). Financial development (dcps) is proxied by domestic credit provided by banks as a percentage of GDP, following the literature that treats private or domestic credit as a standard measure of financial depth and as a channel through which financial systems can ease borrowing constraints, support investment, and improve the allocation of capital and labour across firms (Dwivedi, 2026; Javed, 2025; Fonseca & Van Doornik, 2021; Bai et al., 2018;). This variable is interpreted as capturing the scale of intermediation rather than the full quality of the financial system, since evidence from multi-country productivity studies shows that aggregate financial depth can have weak or specification-sensitive effects when it does not map cleanly onto productivity-enhancing finance (İmamoğlu, 2026; Osei & Kim, 2020).

Foreign direct investment (FDI) is measured as net FDI inflows relative to GDP in order to normalize cross-country differences in economic size and to capture the macroeconomic relevance of foreign capital within each G7 economy, consistent with panel studies that model FDI as net inflows and treat it as a source of capital, technology transfer, managerial know-how, and productivity spillovers (Javed, 2025; Yeboah et al., 2025; Saharti, 2026; Hermes & Lensink, 2003). FDI is entered not only as an explanatory variable but also as a moderator of the finance-productivity relationship because a large literature argues that the productivity effects of foreign capital depend on domestic absorptive capacity, especially financial development and institutional quality, and newer labour productivity evidence explicitly models FDI as moderating the association between financial development and labour productivity (Dwivedi, 2026; Saharti, 2026; Hermes & Lensink, 2003; Fetahi-Vehapi et al., 2015; Osei & Kim, 2020). Gross fixed capital formation (gfcp), measured as a percentage of GDP, is included as the second moderator because it proxies the intensity of investment in fixed productive assets and captures whether financial resources are being translated into capital accumulation, a mechanism widely emphasized in growth and productivity models (Yeboah et al., 2025; Wani, 2022; Fetahi-Vehapi et al., 2015; Dwivedi, 2026). Its interaction with financial development is theoretically and empirically justified by evidence that capital formation is positively associated with productivity and that its moderating role differs across country groups, with recent estimates indicating a positive moderating effect in developed economies (Dwivedi, 2026; Yeboah et al., 2025; Wani, 2022; Najarzadeh et al., 2014).

Trade openness (trade) is measured as total trade in goods and services as a percentage of GDP, following standard macro-panel practice (Rani & Kumar, 2019; Yeboah et al., 2025; Wani, 2022). It is included as a control because trade integration can influence productivity through specialization, resource reallocation, scale effects, and technology diffusion, although the magnitude and even the sign of this effect vary across settings and may depend on complementary conditions such as capital formation, FDI, and domestic capabilities (Raghutla, 2020; Wani, 2022; Fetahi-Vehapi et al., 2015; Duong et al., 2023). Internet use (internet), measured as the percentage of the population using the Internet, is incorporated to capture digital connectivity and the diffusion of information technologies, since cross-country evidence shows that Internet penetration can improve labour productivity directly, mediate part of the effect of human capital, and facilitate knowledge spillovers, equipment investment, and broader innovation performance (Najarzadeh et al., 2014; Hsieh & Goel, 2019; Baker et al., 2020; Paunov & Rollo, 2016). Research and development expenditure (rd), measured as a percentage of GDP, is added to control for innovation intensity because R&D is routinely included in labour productivity models and recent evidence shows that it strengthens the productivity gains associated with digital adoption (Javed, 2025; Haider et al., 2025). Inflation (inf) is retained as a macroeconomic control to separate the estimated finance-productivity relationship from broader macroeconomic conditions, consistent with panel evidence that models inflation jointly with financial development, trade openness, technology, and growth outcomes (Raghutla, 2020). Finally, human capital (hc) is measured using the Penn World Table human-capital index and included because worker capabilities shape the ability of firms and economies to absorb finance, adopt new technologies, and transform foreign and domestic investment into higher labour productivity (Javed, 2025; Baker et al., 2020).

3.3. Interaction Variables

To test the proposed moderating relationships, the study constructs two interaction terms:

$$FD\times FD{I}_{it}=dcp{s}_{it}\times fd{i}_{it}$$

and

$$FD\times GFC{F}_{it}=dcp{s}_{it}\times gfc{p}_{it}.$$

The first interaction examines whether the association between financial development and labour productivity varies with the intensity of foreign direct investment, whereas the second evaluates whether it varies with the level of domestic capital formation. The corresponding marginal effects can be expressed as:

$${\displaystyle \frac{\partial lnL{P}_{it}}{\partial dcp{s}_{it}}}={\beta }_1+{\beta }_{2}FD{I}_{it},$$

and

$${\displaystyle \frac{\partial lnL{P}_{it}}{\partial dcp{s}_{it}}}={\beta }_1+{\beta }_{3}GFC{F}_{it}.$$

Thus, the interaction coefficients are not interpreted independently as direct effects. Instead, their statistical significance indicates whether the association between financial development and labour productivity changes systematically with FDI or capital accumulation. This specification is consistent with the study’s hypotheses concerning the conditional rather than exclusively direct role of financial development.

3.4. Descriptive Statistics

Table 2 reports the descriptive statistics for all variables. Labour productivity has a mean value of 11.628 with a standard deviation of 0.152, ranging from 11.265 to 11.945. The relatively narrow dispersion of lnLP is consistent with the relatively high and broadly comparable levels of economic development among the G7 economies. Financial development (dcps) has a mean of 122.657% of GDP, with a standard deviation of 43.381 percentage points. The variable ranges from 58.257% to 219.126% of GDP, indicating substantial differences in the scale of bank-based financial intermediation across the sample. This considerable variation provides useful empirical variation for examining the relationship between financial development and labour productivity. FDI inflows average 2.121% of GDP, with a standard deviation of 2.131, and range from −0.894% to 12.609% of GDP. The negative minimum reflects periods of net disinvestment or negative net FDI inflows. The relatively wide range also indicates considerable differences in the importance and volatility of foreign investment across the G7 economies. Gross fixed capital formation (gfcp) averages 21.466% of GDP, with a standard deviation of 2.820, and ranges between 16.181% and 29.720% of GDP. This variation indicates meaningful differences in investment intensity across countries and over time. Trade openness has a mean of 53.106% of GDP and a standard deviation of 17.503, with values ranging from 19.275% to 88.786% of GDP, reflecting substantial differences in the degree of international integration among the G7 economies. Internet use averages 72.840% of the population, with a standard deviation of 19.546 percentage points. Its minimum and maximum values of 14.308% and 96.197%, respectively, indicate considerable changes in digital connectivity across the sample period. R&D expenditure averages 2.300% of GDP, ranging from 1.001% to 3.487%, demonstrating differences in innovation intensity across the G7 economies. Inflation has a mean of 1.904, with a standard deviation of 1.620, and ranges from −1.353 to 8.201. The negative minimum indicates that some country-year observations experienced deflationary conditions, while the upper bound reflects periods of substantially higher consumer price inflation. Finally, the human-capital index has a mean of 3.463 and a standard deviation of 0.291, ranging from 2.749 to 3.846, indicating relatively smaller but still meaningful variation in human-capital accumulation across the sample.

Table 2. Descriptive Statistics.

Variable	Obs.	Mean	Std. Dev.	Min.	Max.
lnLP	175	11.628	0.152	11.265	11.945
dcps	175	122.657	43.381	58.257	219.126
fdi	175	2.121	2.131	−0.894	12.609
gfcp	175	21.466	2.820	16.181	29.720
trade	175	53.106	17.503	19.275	88.786
internet	175	72.840	19.546	14.308	96.197
rd	175	2.300	0.683	1.001	3.487
inf	175	1.904	1.620	−1.353	8.201
hc	175	3.463	0.291	2.749	3.846

3.5. Correlation Analysis

Table 3 presents the pairwise correlation matrix. The correlation between labour productivity and financial development is −0.295, indicating a moderate negative unconditional association between lnLP and dcps. This correlation should not, however, be interpreted as evidence of a negative causal or conditional relationship, since it does not account for country-specific effects, common time shocks, control variables, or the potential nonlinear and moderating relationships examined in the subsequent econometric analysis. Labour productivity is negatively correlated with gross fixed capital formation (−0.489) and R&D expenditure (−0.286), while its correlations with FDI (−0.007) and Internet use (−0.030) are close to zero. Trade openness exhibits a relatively weak positive correlation with labour productivity (0.185), whereas inflation is positively correlated with labour productivity (0.361). Human capital has a negative unconditional correlation with labour productivity (−0.238). These unconditional relationships further demonstrate the importance of employing multivariate panel estimators rather than relying on simple correlations. The explanatory variables also exhibit several notable correlations. The strongest is between financial development and trade openness (−0.777), suggesting substantial overlap between bank-based financial depth and international integration within the sample. Human capital is relatively strongly correlated with Internet use (0.656) and R&D expenditure (0.531), while financial development is positively correlated with R&D (0.570) and human capital (0.502). Internet use and R&D are also positively correlated (0.414). Importantly, the correlation matrix does not by itself establish problematic multicollinearity because pairwise correlations do not account for the joint explanatory structure of the regression models. Nevertheless, the relatively high correlations among dcps, trade, rd, internet, and hc warrant formal multicollinearity diagnostics in the subsequent empirical analysis. In particular, the strong negative correlation between financial development and trade openness and the relatively high correlation between Internet use and human capital should be considered when interpreting the coefficients of the fully specified models.

Table 3. Correlation Matrix.

Variable	lnLP	dcps	fdi	gfcp	trade	internet	rd	inf	hc
lnLP	1.000
dcps	−0.295	1.000
fdi	−0.007	−0.121	1.000
gfcp	−0.489	0.288	−0.234	1.000
trade	0.185	−0.777	0.282	−0.399	1.000
internet	−0.030	0.249	−0.116	0.045	0.188	1.000
rd	−0.286	0.570	−0.225	0.432	−0.361	0.414	1.000
inf	0.361	−0.055	0.130	−0.167	0.250	0.168	−0.093	1.000
hc	−0.238	0.502	0.135	−0.012	−0.023	0.656	0.531	0.156	1.000

Overall, the descriptive evidence indicates considerable cross-country and temporal variation in financial development, FDI, capital formation, trade openness, digital connectivity, innovation, macroeconomic conditions, and human capital within the balanced G7 panel. This variation provides the empirical basis for examining both the direct relationship between financial development and labour productivity and the proposed moderating roles of FDI and capital accumulation. The relatively strong correlations among several explanatory variables also motivate the subsequent diagnostic and estimation procedures.

4. Methodology

4.1. Empirical Strategy

This study examines the relationship between financial development and labour productivity in the G7 economies and investigates whether foreign direct investment (FDI) and capital accumulation condition this relationship. The empirical strategy is designed to account for the principal econometric characteristics of the dataset, including the relatively small cross-sectional dimension ($N=7$), the moderate time dimension ($T=25$), potential cross-sectional dependence, parameter heterogeneity across countries, non-stationarity, and the possibility of long-run relationships among the variables.

The methodological framework proceeds in several stages. First, the data are examined using descriptive statistics and pairwise correlations. Second, cross-sectional dependence and slope heterogeneity are assessed to determine whether conventional first-generation panel estimators are appropriate. Third, second-generation panel unit-root tests are employed to account for potential common factors across the G7 economies. Fourth, panel cointegration tests are conducted to establish whether a stable long-run relationship exists among the variables. Fifth, long-run coefficients are estimated using Fully Modified Ordinary Least Squares (FMOLS), Dynamic Ordinary Least Squares (DOLS), and Canonical Cointegrating Regression (CCR). Finally, Common Correlated Effects Mean Group (CCEMG) and Augmented Mean Group (AMG) estimators are employed as robustness approaches that explicitly account for unobserved common factors and cross-country heterogeneity. The moderating roles of FDI and capital accumulation are assessed through interaction terms. This sequential approach is important because the G7 economies are highly integrated through international trade, financial markets, technological diffusion, and common macroeconomic shocks. Consequently, treating the countries as cross-sectionally independent may lead to biased coefficients and misleading statistical inference. At the same time, imposing identical slope coefficients across the seven economies may conceal economically meaningful heterogeneity. The methodology therefore places particular emphasis on estimators that accommodate both common shocks and heterogeneous country-specific responses.

4.2. Baseline Econometric Specification

The baseline relationship between financial development and labour productivity is specified as:

$$\begin{gathered} lnL{P}_{it}={\alpha }_i+{\lambda }_t+{\beta }_1{DCPS}_{it}+{\beta }_2{FDI}_{it}+{\beta }_3{GFCF}_{it}+{\beta }_4{TRADE}_{it}+{\beta }_5{INTERNET}_{it} \\ +{\beta }_6{RD}_{it}+{\beta }_7{INF}_{it}+{\beta }_8{HC}_{it}+{\epsilon }_{it}, \end{gathered}$$

where $i=1,\dots ,7$ denotes the G7 countries and $t=2000,\dots ,2024$ denotes the year. $lnLP$ represents the natural logarithm of labour productivity, while $DCPS$ represents financial development measured by domestic credit provided by banks as a percentage of GDP. $FDI$ denotes net FDI inflows as a percentage of GDP, $GFCF$ represents gross fixed capital formation as a percentage of GDP, and $TRADE$ measures trade openness. $INTERNET$, $RD$, $INF$, and $HC$ represent Internet use, R&D expenditure, inflation, and human capital, respectively. The specification includes country-specific effects (${\alpha }_i$) to capture time-invariant structural characteristics of individual G7 economies and year effects (${\lambda }_t$) to account for common temporal shocks. The error term is denoted by ${\epsilon }_{it}$. The baseline model provides an estimate of the conditional association between financial development and labour productivity after accounting for international integration, digitalization, innovation, macroeconomic conditions, human capital, and investment intensity.

4.3. Modelling the Moderating Role of FDI

To examine whether FDI changes the relationship between financial development and labour productivity, an interaction term between financial development and FDI is introduced:

$$\begin{array}{cc}lnL{P}_{it}=& {\alpha }_i+{\lambda }_t+{\beta }_{1}DCP{S}_{it}+{\beta }_{2}FD{I}_{it}+{\beta }_3\left(DCP{S}_{it}\times FD{I}_{it}\right)\hfill \\ & +{\beta }_{4}GFC{F}_{it}+{\beta }_{5}TRAD{E}_{it}+{\beta }_{6}INTERNE{T}_{it}+{\beta }_{7}R{D}_{it}+{\beta }_{8}IN{F}_{it}+{\beta }_{9}H{C}_{it}+{\epsilon }_{it}.\hfill \end{array}$$

The coefficient ${\beta }_3$ captures the moderating component of FDI. The marginal association between financial development and labour productivity is therefore:

$${\displaystyle \frac{\partial lnL{P}_{it}}{\partial DCP{S}_{it}}}={\beta }_1+{\beta }_{3}FD{I}_{it}.$$

Accordingly, the coefficient of the interaction term should not be interpreted as the direct effect of FDI. Rather, it indicates whether the conditional association between financial development and labour productivity varies with the level of FDI. A statistically significant interaction would provide evidence that the productivity implications associated with financial development are conditional on the degree of foreign investment. Because the existing literature provides arguments for both complementary and potentially offsetting effects, the analysis does not impose an a priori sign on ${\beta }_3$.

4.4. Modelling the Moderating Role of Capital Accumulation

The second moderating relationship is examined by interacting financial development with gross fixed capital formation:

$$\begin{array}{cc}lnL{P}_{it}=& {\alpha }_i+{\lambda }_t+{\beta }_{1}DCP{S}_{it}+{\beta }_{2}GFC{F}_{it}+{\beta }_3\left(DCP{S}_{it}\times GFC{F}_{it}\right)\hfill \\ & +{\beta }_{4}FD{I}_{it}+{\beta }_{5}TRAD{E}_{it}+{\beta }_{6}INTERNE{T}_{it}+{\beta }_{7}R{D}_{it}+{\beta }_{8}IN{F}_{it}+{\beta }_{9}H{C}_{it}+{\epsilon }_{it}.\hfill \end{array}$$

The corresponding marginal association is:

$${\displaystyle \frac{\partial lnL{P}_{it}}{\partial DCP{S}_{it}}}={\beta }_1+{\beta }_{3}GFC{F}_{it}.$$

Here, ${\beta }_3$ determines whether the relationship between financial development and labour productivity varies according to the intensity of capital formation. A significant coefficient would indicate that the association between financial development and productivity is conditional on the level of domestic investment. Finally, a fully specified model incorporates both interaction terms simultaneously:

$$\begin{array}{cc}lnL{P}_{it}=& {\alpha }_i+{\lambda }_t+{\beta }_{1}DCP{S}_{it}+{\beta }_{2}FD{I}_{it}+{\beta }_{3}GFC{F}_{it}\hfill \\ & +{\beta }_4\left(DCP{S}_{it}\times FD{I}_{it}\right)+{\beta }_5\left(DCP{S}_{it}\times GFC{F}_{it}\right)\hfill \\ & +{\beta }_{6}TRAD{E}_{it}+{\beta }_{7}INTERNE{T}_{it}+{\beta }_{8}R{D}_{it}+{\beta }_{9}IN{F}_{it}+{\beta }_{10}H{C}_{it}+{\epsilon }_{it}.\hfill \end{array}$$

This specification allows the two moderating mechanisms to be evaluated jointly while controlling for the other determinants of labour productivity.

4.5. Cross-Sectional Dependence

Given the close economic and financial integration of the G7 economies, cross-sectional dependence is a central consideration in the empirical analysis. Common global shocks-including financial crises, changes in international monetary conditions, technological developments, commodity-price movements, and geopolitical events-may affect several countries simultaneously. Ignoring such dependence can lead to biased standard errors and unreliable statistical inference. The study therefore employs the Pesaran cross-sectional dependence (CD) test (Pesaran, 2014) to examine whether the residuals or variables exhibit significant dependence across countries. The general form of the CD statistic is:

$$CD=\sqrt{{\displaystyle \frac{2T}{N\left(N-1\right)}}}{\displaystyle \sum _{i=1}^{N-1}}{\displaystyle \sum _{j=i+1}^N}{\widehat{\rho }}_{ij}$$

where ${\widehat{\rho }}_{ij}$ represents the pairwise correlation of the relevant residuals across countries. The null hypothesis assumes cross-sectional independence. Rejection of the null indicates that shocks affecting one country may be correlated with outcomes in other countries. Evidence of cross-sectional dependence provides a methodological justification for using second-generation panel procedures and common-factor estimators rather than relying exclusively on conventional first-generation panel techniques.

4.6. Testing for Slope Heterogeneity

The assumption that the coefficients linking financial development, investment, and productivity are identical across all G7 economies may also be restrictive. Although the countries share high levels of economic development, their financial structures, industrial compositions, innovation systems, labour markets, and institutional arrangements differ considerably. To examine whether the slope coefficients are homogeneous, the study employs the Hashem Pesaran and Yamagata (2008) slope heterogeneity test. The null hypothesis assumes homogeneous slope coefficients across countries, whereas rejection indicates that the underlying relationships differ across cross-sectional units. This test is particularly relevant to the present study because the estimated effect of financial development may reasonably differ between, for example, economies with bank-dominated financial structures and those characterized by deeper market-based financial systems. Similarly, the productivity implications of FDI and capital accumulation may vary according to country-specific technological and institutional conditions. Evidence of slope heterogeneity supports the use of estimators such as CCEMG and AMG, which permit country-specific coefficients rather than imposing a common slope across the entire panel.

4.7. Panel Unit-Root Tests

Before estimating long-run relationships, the time-series properties of the variables are examined. Conventional first-generation panel unit-root tests assume cross-sectional independence and may therefore be inappropriate when common shocks are present. Given the expected integration among G7 economies, the study employs second-generation panel unit-root tests, particularly the cross-sectionally augmented Dickey-Fuller (CADF) and cross-sectionally augmented IPS (CIPS) procedures developed by Pesaran (2007).

The CADF approach augments the conventional ADF regression with cross-sectional averages:

$$\mathsf{\Delta }{y}_{it}=a_i+b_i{y}_{i,t-1}+c_i{\overline{y}}_{t-1}+d_i\mathsf{\Delta }{\overline{y}}_t+{\displaystyle \sum _{j=1}^p}{\varphi }_{ij}\mathsf{\Delta }{y}_{i,t-j}+{\epsilon }_{it},$$

where ${\overline{y}}_t$ represents the cross-sectional average of the dependent variable. The inclusion of cross-sectional averages allows the test to account for unobserved common factors affecting multiple countries. The CIPS statistic is obtained by averaging the individual CADF statistics across countries. The null hypothesis corresponds to the presence of a unit root, while rejection provides evidence of stationarity. The unit-root analysis is important for determining whether the variables are stationary in levels or become stationary after first differencing and, consequently, whether subsequent cointegration analysis is warranted.

4.8. Panel Cointegration Analysis

If the variables are found to be integrated of compatible orders, panel cointegration tests are employed to determine whether a stable long-run relationship exists among labour productivity, financial development, FDI, capital accumulation, and the control variables. The study employs Pedroni (1999) and Westerlund (2006) cointegration tests as complementary approaches. Pedroni’s framework allows for heterogeneous intercepts and slope coefficients across cross-sectional units, whereas Westerlund’s approach is based on error-correction dynamics and provides evidence on whether deviations from the long-run equilibrium are corrected over time.

The general long-run relationship can be represented as:

$$lnL{P}_{it}={\alpha }_i+{\beta }_i^{\prime }{X}_{it}+u_{it},$$

where $X_{it}$ is the vector of explanatory variables and $u_{it}$ represents the long-run equilibrium error.

Evidence of cointegration would indicate that the variables share a stable long-run relationship despite potentially being non-stationary individually. This provides the econometric basis for estimating long-run coefficients using cointegration-based estimators. Given the small cross-sectional dimension of the present study, the cointegration results are interpreted cautiously. The tests are used jointly rather than relying on a single statistical criterion, and the subsequent long-run estimates are treated as complementary evidence rather than definitive causal estimates.

4.9. Long-Run Estimation: FMOLS, DOLS, and CCR

Following evidence of long-run cointegration, the study estimates the long-run relationship using Fully Modified Ordinary Least Squares (FMOLS), Dynamic Ordinary Least Squares (DOLS), and Canonical Cointegrating Regression (CCR). FMOLS modifies the conventional OLS estimator to account for endogeneity and serial correlation associated with the cointegrating relationship. Its general representation is:

$$lnL{P}_{it}={\alpha }_i+{\beta }^{\prime }{X}_{it}+u_{it}.$$

The FMOLS transformation corrects the OLS estimates using estimates of the long-run covariance structure between the regressors and the disturbance term.

DOLS provides an alternative correction by augmenting the cointegrating regression with leads and lags of the first differences in the explanatory variables:

$$lnL{P}_{it}={\alpha }_i+{\beta }^{\prime }{X}_{it}+{\displaystyle \sum _{j=-q}^q}{\mathsf{\Gamma }}_j\mathsf{\Delta }{X}_{i,t-j}+u_{it}.$$

The inclusion of leads and lags helps address potential simultaneity and endogeneity between the dependent and explanatory variables. CCR, in turn, transforms the variables in the cointegrating regression to account for endogeneity and serial correlation. Employing FMOLS, DOLS, and CCR provides complementary estimates of the long-run association and allows the robustness of the estimated coefficients to be assessed across alternative cointegration estimators (L. A. Olorogun, 2024). Importantly, these estimators are interpreted as long-run conditional associations, not as evidence of causal effects (Smailov et al., 2026). This distinction is particularly important given the possibility of reverse relationships between financial development, investment, and productivity.

4.10. Common Correlated Effects Mean Group Estimation

Because the G7 economies are exposed to common global shocks and exhibit potentially heterogeneous responses, the Common Correlated Effects Mean Group (CCEMG) estimator is employed as a principal robustness estimator. The CCEMG approach augments country-specific regressions with cross-sectional averages of the dependent and explanatory variables, thereby controlling for unobserved common factors (Bui, 2020; Shanyazov et al., 2026).

For country $i$, the general specification can be expressed as:

$$y_{it}={\alpha }_i+{\beta }_i^{\prime }{X}_{it}+{\gamma }_i^{\prime }{\overline{Z}}_t+{\epsilon }_{it},$$

where ${\overline{Z}}_t$ denotes cross-sectional averages of the dependent and explanatory variables.

The country-specific coefficients are subsequently averaged:

$${\widehat{\beta }}_{CCEMG}={\displaystyle \frac{1}{N}}{\displaystyle \sum _{i=1}^N}{\widehat{\beta }}_i.$$

The principal advantage of CCEMG in the present context is that it simultaneously accommodates unobserved common factors and heterogeneous slope coefficients. This is particularly appropriate for the G7, where common international shocks coexist with substantial differences in national economic structures. The interaction specifications described above are estimated within the CCEMG framework to assess whether the moderating effects identified in the conventional models remain after accounting for common factors and heterogeneous country responses.

4.11. Augmented Mean Group Estimation

The Augmented Mean Group (AMG) estimator is employed as an additional robustness procedure. AMG allows for heterogeneous country-specific coefficients while incorporating a common dynamic process that captures unobserved common shocks (Badmus et al., 2022). The estimator is particularly useful when countries may respond differently to common global developments (Xia et al., 2021; Cengiz & Manga, 2024). In the present study, such common processes may include international financial conditions, technological change, global investment cycles, and other shocks affecting the G7 economies simultaneously. The AMG estimates therefore provide a complementary assessment of whether the financial development-productivity relationship and the proposed moderating effects remain evident when both country heterogeneity and common dynamic processes are explicitly considered. The comparison of CCEMG and AMG estimates is especially informative. If an interaction effect remains statistically significant and broadly similar in magnitude across these estimators, stronger evidence exists that the relationship is not driven solely by omitted common factors or the imposition of homogeneous slopes. Conversely, substantial differences across estimators indicate that the estimated relationship is sensitive to common-factor treatment or heterogeneous country dynamics.

4.12. Methodological Considerations

The methodological design reflects the distinctive characteristics of the G7 panel. With only seven countries and 25 annual observations per country, estimators developed primarily for large-(N) panels must be applied cautiously. Accordingly, the study does not rely on System GMM as its principal estimator, given the limited cross-sectional dimension and the potential difficulties of applying conventional dynamic-panel approaches in this setting. The empirical strategy instead combines FMOLS, DOLS, and CCR with CCEMG and AMG to examine the stability of the estimated long-run relationships across alternative specifications. This approach also reflects the importance of benchmarking alternative empirical methods rather than assuming that greater methodological complexity necessarily produces more reliable results, as illustrated by Kurylek (2025), although that study focuses on forecasting rather than panel estimation. The methodology also recognizes the potential for cross-sectional dependence and heterogeneous economic relationships across countries. Du et al. (2026) demonstrate the relevance of spatial spillovers and regional heterogeneity in their analysis of digital consumption across Chinese provinces. Although their spatial modelling framework differs from the present study, their findings highlight the importance of considering interdependence and heterogeneous responses in economic analysis. In the G7 context, CCEMG and AMG are used to account for common factors and heterogeneous country-specific relationships; however, these estimators do not explicitly identify spatial spillovers between countries. Furthermore, the study does not interpret statistical associations as causal effects in the absence of a credible identification strategy. Financial development may itself respond to labour productivity, while FDI and capital formation may be jointly determined with economic performance. The estimated coefficients are therefore interpreted as conditional long-run associations. The use of multiple estimators, cross-sectional-dependence corrections, and heterogeneous-coefficient approaches is intended to assess the consistency of these associations rather than establish definitive causal mechanisms. Overall, the methodology integrates second-generation panel diagnostics, cointegration analysis, long-run estimation, and common-factor heterogeneous estimators to assess whether financial development is associated with labour productivity in the G7 and whether this relationship varies with FDI and capital accumulation.

5. Results

Table 4 reports the results of Pesaran’s CD test for cross-sectional dependence across the variables included in the empirical analysis. The null hypothesis of the test is that the cross-sectional units are independently distributed. The results provide evidence of substantial cross-sectional dependence for most variables. In particular, the null hypothesis is strongly rejected for lnLP, FDI, GFCP, trade, internet use, R&D, inflation, and human capital, with CD statistics ranging from 4.81 for GFCP to 22.58 for human capital and corresponding p-values of 0.000. The magnitude of the average pairwise correlations further indicates meaningful interdependence across countries. Internet use exhibits the strongest average correlation at 0.918, followed by human capital at 0.985, while inflation and lnLP also display relatively high average correlations of 0.758 and 0.475, respectively. These results suggest that developments in labour productivity, digitalisation, human-capital accumulation, inflation, and other structural variables are not isolated across the countries in the sample but may reflect common regional or international shocks and interconnected economic processes. By contrast, DCPS does not exhibit statistically significant cross-sectional dependence, with a CD statistic of 0.83 and a p-value of 0.409. Its average pairwise correlation is also comparatively low at 0.036, although the average absolute correlation of 0.405 indicates that some degree of bilateral comovement may still exist across individual country pairs. Importantly, failure to reject the null for DCPS does not imply that the variable is unrelated to the remaining variables or that cross-sectional dependence is absent from the panel as a whole. Finally, the evidence indicates that cross-sectional independence cannot be assumed for the majority of variables in the panel. This finding has important methodological implications because conventional first-generation panel estimators and diagnostic procedures that ignore cross-sectional dependence may produce inefficient or potentially misleading statistical inference when countries are exposed to common shocks. The detected dependence is economically plausible given the countries’ geographical proximity, interconnected trade and investment relationships, shared exposure to international financial and commodity market conditions, and common regional and global shocks. Accordingly, the subsequent empirical analysis should employ estimation and testing procedures that explicitly account for cross-sectional dependence and, where appropriate, heterogeneity across countries. This provides a methodological justification for incorporating second-generation panel procedures and cross-sectionally robust estimators in the subsequent analysis.

Table 4. Pesaran CD Test for Cross-Sectional Dependence.

Variable	CD Statistic	p-Value	Average Correlation	Average Absolute Correlation
lnLP	10.89	0.000	0.475	0.782
DCPS	0.83	0.409	0.036	0.405
FDI	5.02	0.000	0.219	0.255
GFCP	4.81	0.000	0.210	0.396
Trade	8.95	0.000	0.391	0.630
Internet	21.03	0.000	0.918	0.918
R&D	8.26	0.000	0.360	0.651
Inflation	17.36	0.000	0.758	0.758
Human capital	22.58	0.000	0.985	0.985

Notes: The test is based on Pesaran’s CD test. The null hypothesis is cross-sectional independence. The panel comprises seven cross-sectional units and is balanced, with an average of 29.17 observations per group.

Table 5 reports the Pesaran cross-sectionally augmented Dickey-Fuller test (2007) results for the stationarity properties of the variables. The CADF procedure is particularly appropriate in the present setting because the preceding Pesaran CD test indicates substantial cross-sectional dependence across the majority of the variables. The null hypothesis of the CADF test is that the series contains a unit root, while rejection of the null indicates stationarity. The results indicate that lnLP, DCPS, FDI, GFCP, trade, Internet, and R&D are non-stationary in levels but become stationary after first differencing, supporting their classification as I(1) processes. For example, lnLP has a level t-bar statistic of −1.369 with a p-value of 0.857, while its first difference yields a t-bar statistic of −3.053 with a p-value below 0.001. Similarly, trade and R&D are non-stationary in levels but become stationary after first differencing, with p-values of 0.001 and below 0.001, respectively. GFCP also exhibits a statistically significant first-difference result, with a p-value of 0.001. DCPS requires somewhat more cautious interpretation. Its level result does not reject the unit-root null, while the first-difference statistic of −2.291 is significant at the 10% level with p = 0.074. Thus, DCPS can be classified as I(1) under a 10% significance criterion, although the evidence is weaker than for the other I(1) variables. Inflation similarly displays borderline evidence in levels, with a p-value of 0.066, indicating stationarity at the 10% level but not at the 5% level. Under the conventional 5% decision rule, it is therefore treated as I(1) because stationarity is established after first differencing. Overall, the unit-root results indicate that the variables are predominantly integrated of order one, with no evidence in the reported results of a variable requiring second differencing. This is an important prerequisite for the subsequent cointegration analysis because the presence of I(1) variables permits investigation of whether a stable long-run equilibrium relationship exists among labour productivity and its explanatory variables. Given the detected cross-sectional dependence and slope heterogeneity, the stationarity evidence also supports the use of second-generation panel procedures rather than relying exclusively on first-generation unit-root tests.

Table 5. Pesaran CADF Panel Unit Root Tests.

Variable	Level: t-bar	p-Value	First Difference: t-bar	p-Value	Integration Order
lnLP	−1.369	0.857	−3.053	0.000	I(1)
DCPS	−1.192	0.939	−2.291	0.074	I(1)
FDI	−3.268	0.125	−5.099	0.000	I(1)
GFCP	−2.223	0.103	−2.189	0.001	I(1)
Trade	−1.816	0.439	−2.861	0.001	I(1)
Internet	−2.671	0.186	−3.594	0.000	I(1)
R&D	−1.390	0.844	−3.300	0.000	I(1)
Inflation	−2.312	0.066	−4.568	0.000	I(1)

Notes: Pesaran’s CADF test includes a constant and one lag. Critical values for the t-bar statistic are −2.210, −2.330, and −2.570 at the 10%, 5%, and 1% levels, respectively. The null hypothesis is that the series contains a unit root.

Table 6 reports the Hashem Pesaran and Yamagata (2008) slope homogeneity test, which examines whether the slope coefficients of the underlying panel model can be assumed to be identical across the seven cross-sectional units. The null hypothesis states that the slope coefficients are homogeneous across countries, whereas rejection of the null indicates systematic cross-country heterogeneity in the estimated relationships. The results provide strong evidence against slope homogeneity. The standard Delta statistic is 5.953 with a p-value of 0.000, while the adjusted Delta statistic is 7.685 with a p-value of 0.000. Thus, the null hypothesis of homogeneous slope coefficients is rejected at the 1% significance level by both test statistics. This finding indicates that the relationships between labour productivity and the explanatory variables are unlikely to be identical across the countries included in the sample. The evidence of slope heterogeneity is economically relevant because the seven countries may differ in terms of financial system development, openness to foreign investment, capital accumulation, digital infrastructure, human-capital conditions, and macroeconomic structures. Consequently, imposing a common slope coefficient across all countries could conceal important differences in the underlying relationships. The result therefore provides an additional methodological justification for employing estimation procedures that allow for cross-sectional heterogeneity when assessing the long-run relationships in the subsequent analysis. Taken together with the evidence of cross-sectional dependence reported in Table 5, the rejection of slope homogeneity suggests that the empirical framework should account for both cross-sectional dependence and heterogeneous slope coefficients. Accordingly, the subsequent estimation strategy relies on panel methods that are designed to provide more appropriate inference under these characteristics of the data.

Table 6. Hashem Pesaran and Yamagata (2008) Slope Homogeneity Test.

Test Statistic	Statistic	p-Value
Delta	5.953	0.000
Adjusted Delta	7.685	0.000

Notes: The test follows Hashem Pesaran and Yamagata (2008). The null hypothesis is that the slope coefficients are homogeneous across cross-sectional units. The constant is partialled out. Both the Delta and adjusted Delta statistics strongly reject slope homogeneity, indicating significant heterogeneity in the slope coefficients across the seven countries.

The Pedroni (1999) panel cointegration tests in Table 7 provide evidence of a long-run equilibrium relationship among the variables. The modified variance ratio rejects the null of no cointegration at the 1% level, while the modified Phillips-Perron and Phillips-Perron statistics are significant at the 5% and 1% levels, respectively. The panel-specific statistics similarly reject the null, with the modified Phillips-Perron, Phillips-Perron, and ADF statistics significant at conventional levels. Taken together, these results support the presence of a stable long-run association among labour productivity, financial development, FDI, capital formation, trade, digitalization, R&D, inflation, and human capital, thereby providing a basis for the subsequent estimation of long-run parameters using FMOLS, DOLS, and CCR.

Table 7. Pedroni Panel Cointegration Tests.

Test Statistic	Statistics	p-Value
Between
Modified variance ratio	−2.5690 ***	0.0051
Modified Phillips-Perron t	2.6012 ***	0.0147
Phillips-Perron t	3.2769 ***	0.0009
Augmented Dickey-Fuller t	0.0468 *	0.0589
Within
Modified Phillips-Perron t	1.8622 **	0.0313
Phillips-Perron t	−2.4478 ***	0.0083
Augmented Dickey-Fuller t	−4.6611 **	0.0243

Notes: * p < 0.10; ** p < 0.05; *** p < 0.01, indicating statistical significance at the 10%, 5%, and 1% levels, respectively.

Table 8 reports the baseline long-run estimates for the G7 economies over 2000–2024 using Panel FMOLS, DOLS, and CCR. Given the potential for reverse causality, joint determination, and omitted-variable bias, the estimated coefficients are interpreted as conditional long-run associations rather than causal effects. Although the use of alternative estimators provides a useful assessment of specification sensitivity, it does not, by itself, resolve endogeneity or establish causal identification. Financial development (DCPS) exhibits a positive and statistically significant association with labour productivity under FMOLS and CCR, whereas its coefficient is negative and significant under DOLS. This variation indicates that the estimated relationship is sensitive to the estimation method, preventing a definitive conclusion about the direction of the long-run association across specifications. Moreover, the estimates do not establish whether financial development is associated with subsequent productivity improvements or whether more productive economies tend to have deeper financial systems. Gross fixed capital formation (GFCF) has a negative and statistically significant coefficient across all three estimators, with a substantially larger magnitude under DOLS. This consistent negative association should be interpreted cautiously. It does not imply that capital accumulation reduces labour productivity because the estimates may reflect differences in investment composition, adjustment periods, capital utilisation, or other factors not fully captured by the model. The sensitivity of the coefficient’s magnitude also suggests that the estimated relationship depends on the estimation procedure. Foreign direct investment (FDI) is statistically insignificant across all three specifications, providing no robust evidence of an independent long-run association with labour productivity after accounting for the included variables. This finding should not be interpreted as evidence that FDI has no productivity implications. Its relationship with productivity may depend on complementary domestic conditions, including financial development and absorptive capacity, which are examined in the subsequent interaction models. Trade openness is statistically insignificant under FMOLS and CCR but negative and significant under DOLS. This pattern indicates that the estimated association between trade openness and labour productivity is not robust across the three estimators. In contrast, Internet use exhibits a positive and highly statistically significant association across all specifications. This consistency suggests that greater digital connectivity is associated with higher labour productivity, conditional on the included regressors. Nevertheless, the estimates do not establish that increased Internet use directly causes productivity growth, as digital adoption may also reflect broader economic development and technological capacity. Research and development expenditure (R&D) is positively associated with labour productivity across the specifications, although its statistical significance is weaker under DOLS. This pattern is consistent with a positive relationship between research activity and productivity, but the results should not be interpreted as identifying the causal return to R&D expenditure. Similarly, inflation exhibits a positive and statistically significant association across all three estimators. This unexpected sign warrants cautious interpretation, as it may reflect the sample context, concurrent macroeconomic conditions, or other factors not fully captured by the empirical specification; it should not be taken to imply that higher inflation improves productivity. Human capital exhibits a positive and statistically significant association with labour productivity under FMOLS and CCR, whereas the DOLS estimate is negative and statistically significant. This divergence indicates substantial estimator sensitivity and does not provide robust evidence of a consistently positive or negative long-run association. Differences in specification, the treatment of long-run dynamics, or potential endogeneity may contribute to the variation, although the baseline results alone cannot determine the underlying explanation. Overall, Table 8 reveals relatively consistent positive associations for Internet use and R&D, as well as a consistent negative coefficient for GFCF. However, the magnitude and interpretation of these relationships require caution, and the estimates do not establish causal effects. The coefficients for financial development, trade openness, and human capital vary across estimators, while FDI remains statistically insignificant in the baseline specifications. These results therefore provide a benchmark for the subsequent interaction models, which examine whether the conditional association between financial development and labour productivity varies with FDI and capital formation.

Table 8. Baseline Long-Run Estimates from Panel FMOLS, DOLS, and CCR.

Variables	Panel FMOLS	Panel DOLS	Panel CCR
DCPS	0.000669 ***	−0.002264 ***	0.000655 ***
	(0.000049)	(0.000507)	(0.000122)
FDI	−0.000458	0.003076	−0.000446
	(0.000865)	(0.004572)	(0.001207)
GFCP	−0.003498 ***	−0.031744 ***	−0.003460 ***
	(0.000531)	(0.003544)	(0.001152)
Trade	0.000414	−0.005951 ***	0.000437
	(0.000345)	(0.001208)	(0.000806)
Internet	0.001256 ***	0.002963 ***	0.001254 ***
	(0.000101)	(0.000667)	(0.000180)
R&D	0.045625 ***	0.034914 *	0.045355 ***
	(0.004812)	(0.018035)	(0.008740)
Inflation	0.005545 ***	0.037200 ***	0.005619 ***
	(0.000484)	(0.005415)	(0.001261)
Human Capital	0.818703 ***	−0.175929 ***	0.822342 ***
	(0.023628)	(0.051089)	(0.053782)
Constant	8.464392 ***	12.186460 ***	8.451986 ***
	(0.070424)	(0.135606)	(0.155751)
Observations	175	175	175

Notes: * p < 0.10; *** p < 0.01, indicating statistical significance at the 10% and 1% levels, respectively.

Building on the baseline estimates in Table 8, Table 9 examines whether the conditional long-run association between domestic credit to the private sector (DCPS) and labour productivity varies with FDI and gross fixed capital formation (GFCF). Given the potential for reverse causality, joint determination, and omitted-variable bias, the estimated coefficients are interpreted as conditional associations rather than causal effects. Across all three FMOLS specifications, the DCPS coefficient remains positive and highly statistically significant, indicating a consistently positive association between domestic credit and long-run labour productivity within the estimated models.

Table 9. Interaction Effects in the Long-Run Labour Productivity Model: Panel FMOLS.

Variables	Model 1: DCPS × FDI	Model 2: DCPS × GFCP	Model 3: DCPS × FDI + DCPS × GFCP
DCPS	0.000481 ***	0.000652 ***	0.000475 ***
	(0.000048)	(0.000028)	(0.000039)
FDI	0.024894 ***	0.000282	0.024697 ***
	(0.003522)	(0.000485)	(0.002744)
GFCP	−0.004200 ***	−0.007293 ***	−0.005945 ***
	(0.000450)	(0.001846)	(0.002195)
Trade	−0.000247	−0.000308	−0.000291
	(0.000300)	(0.000192)	(0.000237)
Internet	0.001342 ***	0.001258 ***	0.001340 ***
	(0.000085)	(0.000055)	(0.000066)
R&D	0.034953 ***	0.046999 ***	0.035606 ***
	(0.004243)	(0.002742)	(0.003477)
Inflation	0.006820 ***	0.005433 ***	0.006750 ***
	(0.000436)	(0.000272)	(0.000353)
Human Capital	0.830751 ***	0.816293 ***	0.830732 ***
	(0.019671)	(0.012905)	(0.015392)
DCPS × FDI	−0.000368 ***	-	−0.000363 ***
	(0.000050)		(0.000039)
DCPS × GFCP	-	0.0000622 **	0.0000288
			(0.000036)
Constant	8.471026 ***	8.473316 ***	8.471073 ***
	(0.058380)	(0.038520)	(0.045705)
Observations	175	175	175

Notes: ** p < 0.05; *** p < 0.01, indicating statistical significance at the 5%, and 1% levels, respectively.

In Model 1, the DCPS × FDI interaction coefficient is negative and highly statistically significant. This result indicates that the positive association between domestic credit and labour productivity becomes weaker at higher levels of FDI, conditional on the other variables in the specification. The positive and statistically significant coefficient on FDI indicates that FDI is positively associated with labour productivity when DCPS is zero, given the model’s parameterisation. However, this coefficient should not be interpreted as FDI’s overall association at all levels of domestic credit because the interaction implies that the conditional association varies with DCPS. The negative interaction coefficient remains highly significant in Model 3, where both interaction terms are included, suggesting that the negative conditional relationship between FDI and the DCPS-productivity association is robust to the inclusion of the GFCF interaction. Nevertheless, this finding does not establish that FDI causally weakens the productivity benefits of domestic credit. Model 2 reports a positive and statistically significant DCPS × GFCF interaction, indicating that the positive association between domestic credit and labour productivity becomes stronger at higher levels of capital formation within this specification. However, when both interaction terms are included simultaneously in Model 3, the DCPS × GFCF coefficient becomes statistically insignificant. This change suggests that the estimated complementary association between domestic credit and capital formation is sensitive to model specification and does not provide robust evidence of a distinct moderating relationship after accounting for the FDI interaction. The control-variable estimates remain broadly consistent across the interaction specifications. Internet use, R&D expenditure, inflation, and human capital retain positive and highly statistically significant coefficients, whereas trade openness remains statistically insignificant. These estimates describe conditional relationships within the specified models and should not be interpreted as identifying the causal effects of the respective variables on labour productivity.

Taken together, the FMOLS interaction results suggest that the association between domestic financial development and labour productivity varies with the level of FDI, with the negative DCPS × FDI interaction emerging as the most consistent finding across the relevant specifications. By contrast, the positive DCPS × GFCF interaction is not robust to the simultaneous inclusion of the FDI interaction. These findings provide evidence of heterogeneous conditional associations within the G7 panel, but the absence of a dedicated identification strategy means that they should not be interpreted as establishing causal moderation effects.

Table 10 examines the conditional long-run associations between domestic credit to the private sector (DCPS) and labour productivity using the DOLS specification. The results indicate that the estimated relationship between domestic credit and labour productivity varies with the inclusion of FDI and gross fixed capital formation (GFCF). However, the interaction estimates differ from those obtained using FMOLS, highlighting the sensitivity of the findings to the estimation approach.

Table 10. Interaction Effects on Labour Productivity: Panel DOLS.

Variables	Model 1: DCPS × FDI	Model 2: DCPS × GFCP	Model 3: DCPS × FDI + DCPS × GFCP
DCPS	0.002273 ***	0.002140 ***	0.002150 ***
	(0.000509)	(0.000491)	(0.000492)
FDI	0.003066 ***	0.001169	0.001194 ***
	(0.004584)	(0.004561)	(0.004573)
GFCP	−0.031321 ***	−0.024253 ***	−0.023695 ***
	(0.003781)	(0.003987)	(0.004202)
Trade	−0.005950 ***	−0.005327 ***	−0.005324 ***
	(0.001211)	(0.001178)	(0.001181)
Internet	0.002917 ***	0.002274 ***	0.002214 ***
	(0.000683)	(0.000671)	(0.000687)
R&D	0.036367 *	0.055846 ***	0.057749 ***
	(0.018619)	(0.018319)	(0.018889)
Inflation	0.037108 ***	0.032342 ***	0.032211 ***
	(0.005437)	(0.005391)	(0.005413)
Human Capital	0.173491 ***	0.191966 ***	0.188923 ***
	(0.051763)	(0.049485)	(0.050108)
DCPS × FDI	0.0045 ***	-	0.05759 ***
	(0.000138)		(0.000133)
DCPS × GFCP	-	0.000340 ***	0.000341 ***
		(0.000093)	(0.000093)
Constant	12.178640 ***	12.232060 ***	12.222280 ***
	(0.138051)	(0.131426)	(0.133696)
Observations	175	175	175

Notes: * p < 0.10; *** p < 0.01, indicating statistical significance at the 10% and 1% levels, respectively.

The DCPS × FDI interaction coefficient is positive in Models 1 and 3, suggesting that the positive association between domestic credit and labour productivity becomes stronger at higher levels of FDI within the DOLS specifications. This finding contrasts with the negative and statistically significant interaction estimated under FMOLS. The opposing signs indicate that the direction of the conditional DCPS-FDI relationship is not consistent across the two long-run estimation approaches. Accordingly, the DOLS results do not establish a uniform moderating relationship between FDI and the association between domestic credit and labour productivity. Differences in estimation procedures and model specifications may contribute to this variation, although the estimates alone do not identify its underlying cause. The DCPS × GFCF interaction is positive and statistically significant in Models 2 and 3. This pattern provides evidence that the positive association between domestic credit and labour productivity becomes stronger at higher levels of capital formation within the DOLS specifications. Unlike the FDI interaction, this positive interaction remains statistically significant when both moderating terms are included in Model 3. Nevertheless, this result should be interpreted as a conditional association rather than evidence that capital formation causally strengthens the productivity effects of domestic credit. Potential reverse causality and omitted-variable bias remain relevant limitations. The control variables also display several notable patterns. GFCF has a negative and statistically significant coefficient across all three specifications, while trade openness is consistently negative and statistically significant. These estimates indicate negative conditional associations with labour productivity within the DOLS models, but they do not establish that capital formation or trade openness reduces productivity. Internet use remains positive and highly statistically significant across all specifications, indicating a comparatively consistent positive association with labour productivity. R&D expenditure is also positively associated with labour productivity, with stronger statistical significance in Models 2 and 3. Inflation exhibits a positive and highly statistically significant coefficient throughout the three specifications. The human capital coefficient should be interpreted according to the corrected estimation output, ensuring that its reported sign and statistical significance accurately reflect the underlying results. Finally, the DOLS estimates indicate that the conditional associations between domestic credit and labour productivity vary with FDI and capital formation. The positive DCPS × FDI interaction contrasts with the negative FMOLS estimate, suggesting that the direction of the FDI interaction is estimator-dependent. By comparison, the positive and statistically significant DCPS × GFCF interaction across Models 2 and 3 provides more consistent evidence of a conditional association in the DOLS specifications, although its robustness across other estimators must also be assessed. These findings should therefore be interpreted cautiously, without attributing causal effects to the interaction terms. The differences between FMOLS and DOLS underscore the importance of comparing the CCR, CCEMG, and AMG results before drawing conclusions about the overall consistency of the estimated relationships.

Table 11 presents the CCR estimates for the interaction specifications and provides further evidence on the conditional long-run associations between financial development and labour productivity in G7 economies over 2000–2024. Consistent with the interpretation of the FMOLS and DOLS results, the CCR coefficients are interpreted as conditional associations rather than causal effects, given the potential for reverse causality, joint determination, and omitted-variable bias.

Table 11. Interaction Effects on Labour Productivity: Panel CCR Estimates.

Variables	Model 1: DCPS × FDI	Model 2: DCPS × GFCP	Model 3: DCPS × FDI + DCPS × GFCP
DCPS	0.000484 ***	0.000653 ***	0.000472 ***
	(0.000117)	(0.000108)	(0.000134)
FDI	0.024691 ***	0.000277	0.025343 ***
	(0.005895)	(0.001125)	(0.004979)
GFCP	−0.004231 ***	−0.06920 ***	−0.05532 ***
	(0.001098)	(0.005705)	(0.006801)
Trade	−0.000257	0.000299	−0.000335
	(0.000713)	(0.000614)	(0.000947)
Internet	0.001335 ***	0.001246 ***	0.001330 ***
	(0.000165)	(0.000139)	(0.000175)
R&D	0.035637 ***	0.046714 ***	0.034595 **
	(0.011573)	(0.010718)	(0.016150)
Inflation	0.006786 ***	0.005501 ***	0.006905 ***
	(0.001438)	(0.001675)	(0.002417)
Human Capital	0.828739 ***	0.818737 ***	0.835167 ***
	(0.050090)	(0.048201)	(0.066824)
DCPS × FDI	0.000361 ***	-	−0.000374 ***
	(0.000111)		(0.000089)
DCPS × GFCP	-	0.00556 ***	0.00206 ***
		(0.000114)	(0.000140)
Constant	8.477232 ***	8.465993 ***	8.459287 ***
	(0.141616)	(0.135326)	(0.182453)
Observations	175	175	175

Notes: ** p < 0.05; *** p < 0.01, indicating statistical significance at 1% levels, respectively.

Across all three specifications, domestic credit to the private sector (DCPS) remains positive and highly statistically significant, with coefficients ranging from 0.000472 to 0.000653. This consistency indicates that DCPS maintains a positive estimated long-run association with labour productivity within the CCR framework, even when the interaction terms are included. However, the estimates do not establish that increases in domestic credit cause improvements in labour productivity.

The estimated DCPS × FDI interaction varies across specifications. In Model 1, the interaction coefficient is positive and highly statistically significant (0.000361), suggesting that the positive association between domestic credit and labour productivity becomes stronger at higher levels of FDI within this specification. FDI itself is also positive and highly statistically significant. However, when the DCPS × GFCF interaction is included in Model 3, the DCPS × FDI coefficient becomes negative and highly statistically significant (−0.000374), while the FDI coefficient remains positive and significant. This reversal indicates that the estimated conditional relationship between domestic credit and FDI is sensitive to the inclusion of the capital-formation interaction. Consequently, the CCR results do not establish a consistently positive or negative DCPS × FDI relationship across specifications. By contrast, the DCPS × GFCF interaction is positive and highly statistically significant in both Models 2 and 3. This finding indicates that the positive association between domestic credit and labour productivity becomes stronger at higher levels of gross fixed capital formation within the CCR specifications. The consistency of the interaction coefficient across these two models provides comparatively stronger evidence of a conditional association between domestic credit, capital formation, and labour productivity in the CCR framework. Nevertheless, this pattern should not be interpreted as proof of a causal complementary effect, particularly given the potential endogeneity of both financial development and investment. GFCF itself remains negatively associated with labour productivity across all three specifications, with the magnitude of its coefficient increasing substantially in Models 2 and 3. These results indicate a negative conditional association between GFCF and labour productivity after accounting for the other regressors, including the interaction terms. The negative coefficient on GFCF does not necessarily contradict the positive interaction coefficient: the former captures the association with GFCF conditional on the other model terms, whereas the latter indicates how the estimated association between DCPS and labour productivity varies with GFCF. Neither coefficient, however, identifies a causal effect. Trade openness remains statistically insignificant across all three CCR specifications, providing no robust evidence of an independent conditional association with labour productivity in these models. In contrast, Internet use retains a positive and highly statistically significant coefficient throughout, indicating a consistent positive association between digital connectivity and labour productivity. R&D expenditure is also positively associated with labour productivity across all specifications, although its statistical significance is weaker in Model 3. Inflation remains positive and highly statistically significant, while human capital exhibits a large and positive coefficient across the three models. These patterns are consistent with the corresponding baseline results, although the estimates should not be interpreted as causal effects of the individual control variables. Finally, the CCR results reinforce the positive conditional association between domestic credit and labour productivity and provide comparatively consistent evidence of a positive DCPS × GFCF interaction. The DCPS × FDI interaction, however, changes sign when the capital-formation interaction is introduced, indicating sensitivity to model specification. Considered alongside the FMOLS and DOLS estimates, the results suggest that the direction and statistical significance of the interaction relationships are not uniformly consistent across estimators. The positive DCPS × GFCF interaction is comparatively more consistent in the CCR specifications, but its broader robustness must be assessed against the contrasting results from the other estimation approaches. Accordingly, the findings are best interpreted as evidence of conditional long-run associations rather than definitive evidence of causal moderation or policy effects.

Table 12 reports the Common Correlated Effects Mean Group (CCEMG) estimates as a robustness assessment that accounts for unobserved common factors and allows slope coefficients to vary across the seven G7 economies. The results differ in some respects from those obtained using FMOLS, DOLS, and CCR, indicating that the estimated relationships are sensitive to cross-country heterogeneity and the treatment of common shocks. In line with the study’s approach to endogeneity, the coefficients are interpreted as conditional associations rather than causal effects.

Table 12. Robustness Check: Common Correlated Effects Mean Group Estimates.

Variables	Model 1: Baseline	Model 2: DCPS × FDI	Model 3: DCPS × GFCP	Model 4: Both Interactions
DCPS	0.00631 ***	0.00879 ***	0.01785 ***	0.00825 ***
	(0.000488)	(0.000908)	(0.001037)	(0.001219)
FDI	0.005405 **	−0.015840	0.04478 ***	0.017245 ***
	(0.002525)	(0.011779)	(0.002888)	(0.00337)
GFCP	−0.011172 ***	−0.012648 ***	−0.010735	−0.001019
	(0.002748)	(0.004621)	(0.008921)	(0.017619)
Trade	0.003094	0.003136	0.002346	0.000923
	(0.002014)	(0.002229)	(0.002408)	(0.002809)
Internet	0.000965 ***	0.000459	0.001067 ***	0.03412 ***
	(0.000242)	(0.000545)	(0.000405)	(0.000954)
R&D	0.052387 **	0.063679 ***	0.068651 ***	0.071978 ***
	(0.021902)	(0.004840)	(0.006540)	(0.009941)
Inflation	0.01021 ***	0.00733 ***	0.0215 ***	0.03447 ***
	(0.003048)	(0.003088)	(0.002553)	(0.003785)
Human Capital	1.226406 *	1.780342 **	0.640870 ***	0.649448 ***
	(0.705018)	(0.844686)	(0.070084)	(0.092025)
DCPS × FDI	-	0.01118 ***	-	0.02522 ***
		(0.000238)		(0.000423)
DCPS × GFCP	-	-	0.02055 ***	0.01161 ***
			(0.000160)	(0.000186)
Observations	175	175	175	175
Countries	7	7	7	7

Notes: * p < 0.10; ** p < 0.05; *** p < 0.01, indicating statistical significance at the 10%, 5%, and 1% levels, respectively.

Across the four specifications, domestic credit to the private sector (DCPS) remains positive and statistically significant, indicating a consistent positive association with labour productivity within the CCEMG framework. FDI is also positive and statistically significant in Models 1, 3, and 4, suggesting that its conditional association with labour productivity is positive in these specifications. However, these results do not establish that FDI independently improves productivity, particularly given the potential for reverse causality and omitted-variable bias. GFCF is negative and statistically significant in Models 1 and 2 but becomes statistically insignificant when the DCPS × GFCF interaction is included. This change suggests that its estimated association depends on the specification and the inclusion of the interaction term. Internet use, R&D expenditure, inflation, and human capital generally retain positive and statistically significant coefficients across the specifications, although their magnitudes vary. These results are broadly consistent with the baseline findings and indicate positive conditional associations between these variables and labour productivity under the CCEMG approach. Nevertheless, statistical significance across multiple specifications does not, by itself, establish causal relationships.

The interaction estimates provide further evidence that the conditional relationship between domestic credit and labour productivity varies with the levels of FDI and capital formation. The DCPS × FDI interaction is positive and statistically significant in Models 2 and 4, while the DCPS × GFCF interaction is positive and statistically significant in Models 3 and 4. These results suggest that the positive association between domestic credit and labour productivity becomes stronger at higher levels of FDI or GFCF in the respective specifications. The significance of both interactions in Model 4 indicates that these conditional associations remain statistically distinguishable from zero when the two interaction terms are included simultaneously within the CCEMG framework. However, the results should not be interpreted as establishing causal complementarity or demonstrating that either moderator strengthens the causal effect of financial development. Overall, the CCEMG estimates provide additional evidence that the association between domestic credit and labour productivity varies with FDI and capital formation when cross-sectional dependence and heterogeneous country responses are taken into account. At the same time, differences in coefficient magnitudes and statistical significance across FMOLS, DOLS, CCR, and CCEMG indicate that the estimated relationships are sensitive to the estimation framework. The positive interaction coefficients under CCEMG contrast with the negative DCPS × FDI estimates obtained in some FMOLS and CCR specifications, while the positive DCPS × GFCF relationship is more consistent across the DOLS, CCR, and CCEMG results than under FMOLS. These differences preclude a conclusion that the moderating relationships are uniformly robust across estimators. Accordingly, the findings are best understood as evidence of conditional long-run associations within the G7 economies, rather than as causal effects or uniform relationships applicable to every country in the sample.

Table 13 reports the Augmented Mean Group (AMG) estimates as an additional robustness assessment. This estimator accommodates a common dynamic process while allowing slope coefficients to differ across the seven G7 economies. The results provide further evidence on the conditional associations between financial development, investment-related factors, and labour productivity, while highlighting differences across specifications. Consistent with the study’s approach to endogeneity, the estimates are interpreted as conditional associations rather than causal effects.

Table 13. Robustness Check: Augmented Mean Group Estimates.

Variables	Model 1: Baseline	Model 2: DCPS × FDI	Model 3: DCPS × GFCP	Model 4: Both Interactions
DCPS	0.00181 ***	0.00298 ***	0.00066 ***	0.00203 ***
	(0.000266)	(0.000206)	(0.000313)	(0.000231)
FDI	0.004701 **	0.03879 ***	0.003071 **	0.04562 ***
	(0.001887)	(0.008512)	(0.001313)	(0.010134)
GFCP	−0.005279 **	−0.005244 **	−0.003125	−0.004228
	(0.002143)	(0.002580)	(0.003822)	(0.005848)
Trade	0.002822 ***	0.002421 ***	0.002134 ***	0.001520 ***
	(0.000547)	(0.000703)	(0.000319)	(0.000542)
Internet	0.000724 *	0.000679 *	0.00535 ***	0.00622 ***
	(0.000374)	(0.000391)	(0.000507)	(0.000557)
R&D	0.025505 ***	0.023730 ***	0.020829 ***	0.028310 ***
	(0.002225)	(0.009191)	(0.006026)	(0.003868)
Inflation	0.003185 *	0.002587	0.002447 **	0.01539 ***
	(0.001707)	(0.001861)	(0.001013)	(0.001129)
Human Capital	0.393657 **	0.368191 ***	0.328577 ***	0.368559 ***
	(0.171786)	(0.134769)	(0.118949)	(0.114590)
DCPS × FDI	-	0.002598 ***	-	0.000080
		(0.00175)		(0.000204)
DCPS × GFCP	-	-	−0.00168 ***	0.00066 ***
			(0.000080)	(0.000195)
Common dynamic process	1.002996 ***	1.041373 ***	1.081308 ***	1.141086 ***
	(0.165647)	(0.177260)	(0.158902)	(0.157153)
Constant	12.614440 ***	12.510450 ***	12.517740 ***	12.640690 ***
	(0.604453)	(0.473026)	(0.412865)	(0.367066)
Observations	175	175	175	175
Countries	7	7	7	7

Notes: * p < 0.10; ** p < 0.05; *** p < 0.01, indicating statistical significance at the 10%, 5%, and 1% levels, respectively.

Domestic credit to the private sector (DCPS) is positive and statistically significant across all four specifications, with coefficients ranging from 0.000660 to 0.002980. This consistency indicates a positive association between domestic credit and labour productivity within the AMG framework. FDI is also positive and statistically significant across the reported models, although its coefficient varies considerably across specifications. These findings suggest that both variables are positively associated with labour productivity conditional on the included regressors; however, they do not establish that increases in domestic credit or FDI cause productivity improvements.

Gross fixed capital formation (GFCF) is negative and statistically significant in Models 1 and 2, but its coefficient becomes statistically insignificant in Models 3 and 4. This pattern suggests that the estimated association between capital formation and labour productivity is sensitive to the inclusion of additional interaction terms. It does not, by itself, indicate that capital accumulation reduces productivity.

Trade openness exhibits a positive and highly statistically significant association with labour productivity across all four specifications, providing relatively consistent evidence within the AMG framework. Internet use is positive and statistically significant in Models 1 and 2, while its coefficients remain positive in Models 3 and 4. The latter estimates should not be described as statistically significant unless supported by their corresponding standard errors and significance levels. R&D expenditure remains positive and statistically significant throughout the four specifications, indicating a comparatively stable positive association with labour productivity. Inflation is positive and statistically significant in Models 1 and 3, whereas its significance is weaker or absent in the remaining specifications. Human capital is positive and statistically significant across all four models, suggesting a consistent positive conditional association within the AMG framework.

The interaction estimates provide more qualified evidence regarding the conditional relationships. The DCPS × FDI interaction is positive and statistically significant in Model 2 but becomes statistically insignificant when both interaction terms are included in Model 4. This indicates that the estimated relationship between domestic credit and labour productivity conditional on FDI is sensitive to the specification. Similarly, the statistical significance of the DCPS × GFCF interaction should be assessed using the corresponding estimates and standard errors in Models 3 and 4. Taken together, the interaction results do not establish a uniformly robust moderating relationship across specifications. Instead, they suggest that the estimated conditional associations depend on the model specification and the inclusion of the alternative interaction term.

The estimated common dynamic process is positive and highly statistically significant across all four models, with coefficients increasing from 1.003 to 1.141. This finding indicates a substantial common dynamic component in the labour productivity series captured by the AMG specifications. However, the common dynamic coefficient should be interpreted within the structure of the estimator and should not be treated as a direct measure of the causal effect of common shocks on productivity. Overall, the AMG results provide additional robustness evidence for the positive associations of DCPS, FDI, trade openness, R&D expenditure, and human capital with labour productivity. Nevertheless, the changing significance of GFCF, inflation, and the interaction terms demonstrates that some estimated relationships are sensitive to specification. In particular, the DCPS × FDI interaction is not statistically significant when both interaction terms are included, limiting the evidence for a robust conditional relationship involving FDI. These findings complement the FMOLS, DOLS, CCR, and CCEMG results by demonstrating the importance of accounting for common dynamics and cross-country heterogeneity. Given the potential for reverse causality and omitted-variable bias, the overall evidence should be interpreted as indicating conditional associations rather than causal effects of financial development, FDI, or capital formation on labour productivity.

6. Discussion

The empirical findings provide a nuanced picture of the conditional associations between financial development, investment, international integration, technological capabilities, and long-run labour productivity in the G7 economies over 2000–2024. The results do not support a uniform relationship between all explanatory variables and productivity. Instead, the estimated associations vary across estimators and model specifications, suggesting that the interpretation of productivity patterns depends partly on the measurement of economic factors and the treatment of common dynamics and cross-country heterogeneity. These findings are consistent with research emphasizing that the productivity relevance of finance, foreign investment, trade, innovation, and technology depends on country-specific characteristics and absorptive capacity rather than exposure to these factors alone (Girma, 2005; Ndubuisi & Owusu, 2023; Hammar & Belarbi, 2021; Malhotra, 2025). The sensitivity of several estimates is also consistent with evidence that advanced economies close to the technological frontier may exhibit heterogeneous productivity relationships and weaker average associations than catching-up economies (Männasoo et al., 2018; Ndubuisi & Owusu, 2023). The findings also highlight the importance of accounting for cross-country interdependence and methodological sensitivity when interpreting long-run productivity relationships. Du et al. (2026) document significant spatial dependence and heterogeneous spillover patterns in digital consumption across Chinese provinces. Their results provide a useful comparative perspective on how economic relationships may extend beyond individual geographical units, although the specific spillover mechanisms identified in their spatial model cannot be directly inferred for the G7 economies from the present analysis. In this study, CCEMG and AMG provide complementary evidence on common dynamics and heterogeneous country responses, but they do not replace an explicit spatial analysis. Future research could therefore investigate whether trade, financial linkages, and cross-border investment generate measurable productivity spillovers among G7 economies. The sensitivity of several coefficients across FMOLS, DOLS, CCR, CCEMG, and AMG further cautions against relying on a single estimator. This interpretation is consistent with the broader methodological lesson illustrated by Kurylek (2025): more sophisticated analytical approaches do not necessarily produce more reliable results than simpler alternatives. However, because Kurylek examines forecasting rather than long-run panel estimation, the comparison is conceptual rather than a direct empirical validation of the estimators used here. Taken together, these considerations support a cautious interpretation of the findings, with greater emphasis on relationships that remain comparatively consistent across specifications and less emphasis on coefficients whose signs or significance depend on the estimation method.

Domestic credit to the private sector (DCPS) is positively associated with labour productivity under FMOLS and CCR, while the DOLS estimate is negative and statistically significant. The positive estimates are consistent with the possibility that financial intermediation supports capital upgrading, technological adoption, business expansion, innovation, and working-capital financing. Bai et al. (2018), for example, emphasize the role of financial intermediation in reallocating resources toward more productive firms. However, the relationship may depend on the allocation of credit rather than its aggregate volume. In mature financial systems, credit can support productive investment but may also flow toward asset-related or other activities with limited direct implications for productivity (Mazzucato & Wray, 2015). Arestis (2020) similarly discusses how prolonged periods of exceptionally low interest rates may allow relatively unproductive firms to remain active. These mechanisms provide possible interpretations of the results, rather than explanations directly tested by the present models. Given the conflicting FMOLS, DOLS, and CCR estimates, the evidence supports a positive association in some specifications but does not establish a uniformly positive long-run relationship between domestic credit and labour productivity.

The results for foreign direct investment (FDI) are similarly conditional. Its baseline coefficient is statistically insignificant across FMOLS, DOLS, and CCR, providing no robust evidence of an unconditional long-run association with labour productivity. The interaction estimates are also sensitive to the estimator: the DCPS × FDI coefficient is negative and significant in the FMOLS specifications, whereas DOLS and CCR produce different patterns. This variation does not support a general conclusion that FDI either strengthens or weakens the productivity association of domestic credit. A plausible interpretation is that aggregate FDI combines different activities, including greenfield investment, mergers and acquisitions, corporate restructuring, and strategic asset purchases, which may have different implications for productive capacity. The relationship may also depend on domestic firms’ ability to absorb foreign knowledge and technology. Previous studies emphasize the importance of host-country characteristics, absorptive capacity, and domestic financial conditions in determining the productivity relevance of FDI (Ang & Madsen, 2013; Kinfemichael, 2015; Li & Tanna, 2019; Bodman & Le, 2013; Girma, 2005; Ahmed & Kialashaki, 2021; Ahmed & Kialashki, 2017; Sultana & Turkina, 2020). Castellani and Zanfei (2003), for instance, document heterogeneous FDI spillovers across European firms. Other research highlights the importance of global value-chain integration, intangible investment, domestic R&D, and technology-transfer channels (Karpowicz & Suphaphiphat, 2020; Cameron et al., 2005). Access to multinational firms’ internal capital markets and other international financing channels may also alter the relative role of domestic credit. These possibilities are consistent with the observed specification sensitivity, but the aggregate data do not allow the study to distinguish conclusively between financing substitution, technology spillovers, and differences in investment composition. Accordingly, the findings indicate that the association between domestic credit and productivity may vary with FDI, rather than establishing a universal moderating mechanism (Bai et al., 2018; Mazzucato & Wray, 2015; Liao et al., 2016; Hammar & Belarbi, 2021; Sultana & Turkina, 2020; Ndubuisi & Owusu, 2023).

Gross fixed capital formation (GFCF) exhibits a negative and statistically significant baseline association with labour productivity across FMOLS, DOLS, and CCR. Although this finding contrasts with the conventional expectation that productive capital accumulation supports labour productivity, it should not be interpreted as evidence that investment itself reduces productivity. Aggregate GFCF does not distinguish between investment types with different productivity implications, such as replacement capital, construction, infrastructure, ICT, software, and other intangible assets. Its coefficient may therefore reflect the composition and timing of investment, as well as the limitations of an aggregate measure. Research on mature economies emphasizes the importance of distinguishing between overall capital accumulation and the technological content of investment (Giordano & Zollino, 2020; Arestis, 2020; Byrne et al., 2016). Other studies highlight the role of intangible assets, innovation, ICT, and complementary human capital in productivity performance (Karpowicz & Suphaphiphat, 2020; Bronzini & Piselli, 2009; Reljic et al., 2025; Van Ark et al., 2008; Tinta, 2022; Fassio et al., 2019).

The interaction results provide some evidence that the association between GFCF and productivity may depend on domestic financial conditions. The DCPS × GFCF interaction is positive and statistically significant in selected specifications, including one FMOLS model, while its significance is not consistently maintained when both interaction terms are included or across alternative estimators. This pattern is compatible with the possibility that investment is more closely associated with productivity when financing supports productive firm expansion, equipment upgrading, and innovation. Related research emphasizes the importance of credit allocation, intangible investment, and technological orientation (Bai et al., 2018; Karpowicz & Suphaphiphat, 2020; Akaev & Sadovnichiy, 2021). Nevertheless, because the interaction is not uniformly robust, the evidence does not establish a stable complementary relationship. The principal implication is interpretive: the volume of capital formation alone may not adequately represent the productivity relevance of investment.

Trade openness is statistically insignificant under FMOLS and CCR but negative and significant under DOLS in the baseline model. Thus, the estimates do not establish a consistent positive long-run association between trade and labour productivity in the G7 sample. Trade may support productivity through market expansion, specialization, access to imported inputs, and technology transfer, but the magnitude of these associations can depend on industrial structure and participation in global value chains. These mechanisms are discussed in the literature on trade integration and productivity (Ndubuisi & Owusu, 2023; Malhotra, 2025; Lee, 2020). The productivity implications of international integration may also differ between developed and developing economies (Ndubuisi & Owusu, 2023). Adjustment costs, import competition, changes in value-added structures, and exposure to international disturbances provide additional possible explanations for heterogeneous estimates (Aichele & Heiland, 2018; Suleman et al., 2026). However, the present results do not directly identify these channels. The estimator sensitivity therefore warrants caution against treating trade openness as an independent and uniformly productivity-enhancing factor.

Among the explanatory variables, Internet use displays one of the most consistent positive associations with labour productivity across FMOLS, DOLS, and CCR and in several interaction specifications. This pattern is consistent with the role of digital connectivity in reducing information and transaction costs, improving coordination, expanding market access, and supporting data-based decision-making. Previous studies document associations between ICT adoption, digitalization, and productivity (Cardona et al., 2013; Varlamova & Larionova, 2020; Amador & Silva, 2025). The productivity relevance of digital technologies may also depend on complementary skills, organizational capabilities, and intangible investment (Liao et al., 2016; Cette et al., 2021). Nevertheless, Internet use is an aggregate indicator and does not directly measure the intensity, quality, or productive application of digital technologies. The results should therefore be interpreted as evidence of a comparatively consistent association with labour productivity, not as a direct estimate of the causal return to digitalization.

R&D expenditure is positively associated with labour productivity across the principal long-run estimators, although its statistical significance is weaker under DOLS. This pattern is compatible with the role of research and innovation in developing technologies, improving production processes, and supporting knowledge-intensive activities. The literature emphasizes the contribution of R&D and technological capabilities to productivity, while also highlighting the importance of country characteristics and complementary factors (Männasoo et al., 2018; Lee, 2020; Hammar & Belarbi, 2021). The estimated association may also reflect the time required for research expenditure to translate into measurable productivity outcomes. Apokin and Ipatova (2016) discuss delayed productivity responses to R&D, while Ciaffi et al. (2024) emphasize the potential complementarity between public innovation spending and private R&D activity. Because the study measures R&D expenditure rather than successful innovation outputs, the results indicate a positive association with research intensity rather than an immediate or uniform productivity return.

Human capital is positively associated with labour productivity under FMOLS and CCR but has a negative coefficient under DOLS. The contrasting estimates indicate that its relationship with productivity is less robust across estimators than the positive results under some specifications might suggest. Human capital may be particularly relevant in economies where productivity depends on advanced skills, management capabilities, digital competencies, and the effective use of technology. Previous research emphasizes its contribution to productivity and its complementarity with innovation and technological adoption (Männasoo et al., 2018; Ndubuisi & Owusu, 2023; Liao et al., 2016). However, aggregate human-capital indicators may not capture skill quality, field-specific expertise, skill matching, or the effective deployment of educated workers. The relationship may also overlap with other slow-moving structural variables, including digitalization and R&D. As the present analysis does not directly test these mechanisms, the mixed coefficients should be treated as evidence of estimator sensitivity rather than as a definitive estimate of the productivity contribution of human capital. The role of absorptive capacity and institutional conditions may also help explain why skills do not translate uniformly into productivity outcomes (Sultana & Turkina, 2020).

Inflation is positive and statistically significant in the baseline FMOLS, DOLS, and CCR estimates, but this association should not be interpreted as evidence that higher inflation improves labour productivity. The coefficient may reflect broader macroeconomic conditions, common trends, or the particular episodes covered by the sample, including the global financial crisis, the COVID-19 pandemic, and subsequent energy and supply shocks. Structural and macroeconomic conditions can influence estimated productivity relationships (Ndubuisi & Owusu, 2023), but the present models do not identify the mechanism underlying the positive inflation coefficient. Given the potential for inflation to increase operating costs and uncertainty, the result should be treated as a conditional statistical association rather than a basis for recommending inflationary policies.

The CCEMG and AMG robustness estimates further qualify the findings by accounting for common factors and allowing for heterogeneous country-level relationships. The estimates generally reinforce the positive association between DCPS and labour productivity, while the significance and direction of several interaction terms vary across estimators and specifications. The AMG results also indicate a strong common dynamic component, suggesting that productivity movements across the G7 share an important common element. Potential sources include global technological change, international business cycles, financial conditions, energy price fluctuations, and geopolitical disruptions, although these specific channels are not separately identified in the models. The heterogeneous estimates are consistent with research emphasizing differences in technological capabilities, innovation systems, and productivity responses across countries (Männasoo et al., 2018; Hammar & Belarbi, 2021; Ndubuisi & Owusu, 2023). Research also highlights the importance of spatial spillovers and ICT-related capabilities in productivity performance (Männasoo et al., 2018; Cette et al., 2021). These findings support the use of estimators that account for cross-sectional dependence and heterogeneous slopes, but they also demonstrate that robustness across estimation frameworks should not be assumed. In particular, the interaction results are less stable than several of the main associations, limiting the extent to which a single moderating mechanism can be generalized across the G7.

Taken together, the findings suggest that labour productivity in the G7 is associated with a combination of financial, investment, technological, and human-capital factors rather than a single dominant determinant. Internet use and R&D expenditure display comparatively consistent positive associations across the principal long-run estimators. Domestic credit is positive under FMOLS and CCR but negative under DOLS, while the results for FDI, trade openness, GFCF, human capital, and the interaction terms vary more substantially across specifications. The negative baseline coefficient for GFCF, in particular, highlights the limitations of using aggregate investment volumes to characterize the productivity relevance of capital accumulation. The results are consistent with the broader literature emphasizing the roles of capital composition, intangible assets, innovation, and complementary capabilities in advanced economies (Byrne et al., 2016; Karpowicz & Suphaphiphat, 2020; Akaev & Sadovnichiy, 2021). Similarly, the findings concerning domestic credit are compatible with research emphasizing the importance of credit allocation and the distinction between financial expansion and productive investment (Bai et al., 2018; Mazzucato & Wray, 2015). The study contributes to the literature by comparing the estimated long-run associations across FMOLS, DOLS, and CCR and examining their sensitivity through CCEMG and AMG specifications, including models with interaction terms. The results highlight the importance of distinguishing relatively consistent associations from findings that depend on the estimator or model specification. In practical terms, they suggest that aggregate indicators of credit, FDI, and capital formation may provide an incomplete picture of productivity-related conditions. However, these findings should not be interpreted as establishing that particular policy interventions will raise productivity. The analysis covers only seven economies, and the estimated coefficients remain subject to potential measurement limitations, specification sensitivity, reverse causality, and omitted-variable bias. Furthermore, the study does not employ a dedicated causal identification strategy. The findings therefore support a cautious, associational interpretation and provide a basis for further research rather than definitive causal policy conclusions. Future studies could examine sector-level or firm-level data, distinguish greenfield FDI from mergers and acquisitions, separate tangible from intangible investment, and investigate country-specific differences in the allocation of credit and the adoption of digital technologies. Such extensions would help clarify which mechanisms account for the heterogeneous associations observed across the G7.

7. Policy Recommendations

The empirical findings suggest several policy considerations for improving labour productivity in the G7 economies. However, given the estimator sensitivity, cross-country heterogeneity, and absence of a dedicated causal identification strategy, these implications should be interpreted as policy directions consistent with the observed conditional associations rather than as evidence that particular interventions will necessarily increase productivity. The results support a selective and context-specific approach rather than a uniform set of policy prescriptions for all seven economies.

First, financial development policy should emphasize credit allocation rather than credit expansion alone. Domestic credit to the private sector (DCPS) is positively associated with labour productivity under FMOLS and CCR, although the DOLS estimate has the opposite sign. This mixed evidence does not establish a uniformly positive relationship between credit and productivity. Nevertheless, it provides a basis for examining whether financial resources are directed toward productive firms, technological upgrading, and innovation-intensive activities. Improving credit information systems, lending assessment, and access to finance for innovative enterprises may be relevant considerations, particularly where financing constraints limit productive investment.

Second, investment policy should distinguish capital formation from its productivity content. The negative association between gross fixed capital formation (GFCF) and labour productivity across the baseline estimators suggests that aggregate investment volume alone may not adequately capture the productivity relevance of capital accumulation. This result should not be interpreted as evidence against investment. Instead, it highlights the value of examining the composition of investment, including digital infrastructure, advanced machinery, ICT assets, research facilities, and other technology-intensive capital. Evaluating investment proposals against transparent criteria for economic viability, expected utilization, and technological relevance may help policymakers better assess the potential contribution of capital projects to productivity.

Third, FDI policy should consider domestic linkages and absorptive capacity alongside investment volumes. The baseline FDI coefficient is statistically insignificant across FMOLS, DOLS, and CCR, while the DCPS × FDI interaction varies across estimators and specifications. Consequently, the findings do not support the assumption that attracting more FDI will automatically improve labour productivity or that FDI consistently strengthens or weakens the productivity association of domestic credit. Supplier-development initiatives, collaboration between multinational enterprises and domestic firms, workforce training, and technology-transfer partnerships are possible areas for policy consideration. Their effectiveness, however, would need to be evaluated using evidence that directly measures domestic linkages and productivity outcomes.

Fourth, digitalization and R&D warrant continued attention in productivity strategies. Internet use and R&D expenditure exhibit comparatively consistent positive associations with labour productivity across the principal long-run estimators. These findings are consistent with the importance of digital connectivity and innovation in advanced economies, although they do not identify the causal returns to specific programmes. Relevant policy priorities may include supporting digital adoption among small and medium-sized enterprises, strengthening digital skills and cybersecurity, facilitating access to research infrastructure, and improving collaboration between universities, research institutions, and businesses. Particular attention could be given to the diffusion and commercialization of existing technologies, rather than relying exclusively on increases in digital infrastructure or R&D expenditure.

Fifth, human-capital and trade policies should reflect the mixed empirical evidence. Human capital is positively associated with labour productivity under FMOLS and CCR but has a negative coefficient under DOLS. Similarly, trade openness is statistically insignificant under FMOLS and CCR but negatively associated with productivity under DOLS. These findings do not justify definitive conclusions about the effects of education or trade liberalization. They instead suggest the importance of examining skill quality, worker-firm matching, technological requirements, and the sectoral composition of international integration. Measures such as lifelong learning, vocational and digital training, worker reskilling, and support for firms adapting to international competition may be worth considering, with their relevance assessed against country-specific evidence.

Finally, productivity strategies should account for differences across the G7 economies. The CCEMG and AMG results indicate that common dynamic factors and heterogeneous relationships are relevant to the analysis. Although these estimates do not identify the specific contribution of individual national characteristics, they caution against assuming that the same policy mix will be equally appropriate across all seven countries. National strategies could therefore be informed by country-specific assessments of credit allocation, investment composition, digital adoption, innovation capacity, and workforce skills. Coordination across financial, investment, innovation, digital, education, and trade policies may also be useful where complementary constraints are identified.

Overall, the findings support a quality-oriented and evidence-based approach to productivity policy. The estimates do not establish that increasing credit, investment, FDI, or trade will necessarily raise labour productivity. Rather, they highlight the relevance of examining how finance, capital, technology, and skills are combined within different national economic settings. The policy considerations presented here should therefore be treated as directions for further assessment, not as guaranteed causal prescriptions. More targeted firm-level, sector-level, and country-specific evaluations would be needed to determine which interventions are effective, for whom, and under what conditions.

8. Conclusions

This study examined the long-run associations between labour productivity and domestic credit to the private sector, foreign direct investment (FDI), gross fixed capital formation (GFCF), trade openness, Internet use, R&D expenditure, inflation, and human capital in the seven G7 economies over 2000–2024. Panel FMOLS, DOLS, and CCR were used as the principal long-run estimators, complemented by interaction specifications and CCEMG and AMG robustness analyses. The findings provide a differentiated picture of productivity relationships, with several associations varying across estimators and model specifications.

Internet use and R&D expenditure exhibit comparatively consistent positive associations with labour productivity across the principal estimators. Domestic credit to the private sector is positively associated with productivity under FMOLS and CCR but negatively associated under DOLS, indicating estimator sensitivity. GFCF displays a negative association across the principal long-run estimators, although this finding does not establish that investment reduces productivity. Rather, the aggregate measure does not distinguish between different types and uses of capital. The baseline estimates provide no robust evidence of an unconditional association between FDI and labour productivity, while the results for trade openness and human capital vary across estimation methods. Inflation is positively associated with productivity in the baseline specifications, but this statistical relationship should not be interpreted as evidence that higher inflation improves productivity.

The interaction analysis further indicates that the estimated relationship between domestic credit and labour productivity depends on model specification. The DCPS × FDI interaction is negative and statistically significant in the FMOLS specifications but displays different patterns under alternative estimators. The DCPS × GFCF interaction is positive in selected specifications, although its statistical significance is not consistently maintained across models. Consequently, the evidence does not establish uniform moderating relationships between domestic credit, FDI, and capital formation. The CCEMG and AMG results additionally highlight the relevance of common dynamics and heterogeneous country responses. Overall, the findings suggest that productivity relationships in the G7 are conditional and sensitive to the estimation framework rather than governed by a single mechanism that operates uniformly across all seven economies.

The study contributes to the existing literature by examining financial development, investment, international integration, digitalization, innovation, and human capital within a unified long-run panel framework. Its comparison of alternative estimators and interaction specifications distinguishes comparatively consistent associations from relationships that are sensitive to model specification. From a practical perspective, the findings provide a basis for examining the allocation of credit, the composition of investment, the adoption of digital technologies, and the role of innovation in advanced economies. However, the results do not establish that specific policy interventions will cause productivity improvements. Policy implications should therefore be treated as directions for further assessment rather than definitive causal prescriptions.

Several limitations qualify these conclusions. First, the sample includes only seven G7 economies. Although the 2000–2024 period provides a relatively long time dimension, the limited cross-sectional coverage restricts generalizability beyond the G7 and may make aggregate estimates sensitive to country-specific characteristics. Second, the study identifies conditional long-run associations rather than causal effects. Reverse causality and omitted-variable bias remain possible because productivity may itself influence credit, FDI, investment, R&D expenditure, and digital adoption. The use of FMOLS, DOLS, CCR, CCEMG, and AMG does not, by itself, eliminate these endogeneity concerns. Third, some findings are sensitive to the estimator and model specification, particularly those concerning domestic credit, human capital, trade, and the interaction terms, limiting the robustness of conclusions about their direction and magnitude.

Fourth, the aggregate indicators have measurement limitations. Domestic credit does not directly capture lending quality or sectoral allocation; gross fixed capital formation (GFCF) does not distinguish adequately between types of capital investment; and aggregate FDI does not differentiate between investment modes or recipient sectors. Similarly, Internet use captures digital connectivity rather than the full extent of digital transformation, while the human-capital measure may not fully reflect skill quality, occupational matching, or the effective utilization of skills. Fifth, the study uses aggregate national-level data and does not distinguish between agriculture, industry, and services. These sectors differ in productivity levels, capital intensity, technological adoption, and exposure to financial development and foreign investment. Moreover, aggregate productivity reflects both changes within sectors and the reallocation of labour across sectors. Consequently, the estimated relationships may partly reflect differences in sectoral composition across G7 economies and cannot establish whether the observed associations hold uniformly across economic activities. Future research using sector-disaggregated data could provide a more detailed assessment of these relationships and clarify the extent to which sectoral heterogeneity contributes to the observed cross-country differences.

Future research could address these limitations through several targeted extensions. Firm-level and sector-level analyses could help establish whether the observed associations differ across industries and firm types. Disaggregating FDI and capital formation would allow researchers to distinguish between investment categories with potentially different productivity implications. More detailed indicators of digital adoption, innovation outputs, credit allocation, and workforce skills could also improve measurement of the mechanisms considered in this study. Finally, future work could employ credible instrumental-variable or quasi-experimental designs where suitable instruments or exogenous policy changes are available, alongside analyses of whether the estimated relationships differ across major economic disruptions or structural periods.

In conclusion, the evidence indicates that labour productivity in the G7 is associated with a combination of financial, investment, technological, and human-capital factors, but the strength and direction of these associations are not uniform across estimators. Digitalization and R&D exhibit comparatively consistent positive associations, whereas the results for domestic credit, FDI, capital formation, trade, and human capital require greater caution. By documenting these differences, the study underscores the importance of estimator sensitivity, measurement choices, and cross-country heterogeneity when interpreting long-run productivity relationships. Further research using more disaggregated data and stronger causal identification is needed to determine which mechanisms underpin these associations and under what conditions they translate into sustained productivity improvements.