Determinants of Tropical Hardwood Lumber Exports to the ITTO Market: Econometric Evidence and Strategic Pathways for Sustainable Development in Producing Regions

Abstract

This study investigates the structural and cyclical determinants of tropical hardwood exports among member countries of the International Tropical Timber Organization (ITTO) over the period 1995–2022—a sector historically characterized by persistent value imbalances. The central research issue addresses the enduring asymmetries in the global value chain, shaped by unequal industrial capacities, limited access to environmental certifications, and entrenched North–South trade relations. The study pursues three main objectives: (1) to develop a typology of exporting countries; (2) to estimate heterogeneous trade elasticities; (3) to propose a policy framework that reconciles equity with sustainability. The empirical findings identify four export profiles: (i) raw producers with minimal local processing; (ii) marginal players with weak trade integration; (iii) high-value-added re-export platforms (notably in Asia); (iv) major consumer markets. Trade effects vary across regions. In the short term, imports boost exports (+0.33%), particularly in re-export models seen in Asia, the USA, and the EU, while local production remains limited in Africa due to weak industrial capacity. In the long term, both domestic production and imports have a positive impact (+0.38% and +0.37%), but only countries with strong industrial bases fully benefit. Population size (+1.29%) also reinforces the advantage of large markets like China and India, supported by short-term economic growth elasticity (+1.1%), likely driven by improved logistics or rising demand from importing countries. In response, the policy implications converge around the proposal of a “Fair and Digital Timber Trade Model” (F&DTTT), structured around three pillars: (a) specialized economic zones aligned with SDGs 8, 12, and 15; (b) blockchain-based traceability systems to enhance supply chain transparency; (c) South–South cooperation strategies aimed at commercial, regulatory, and institutional rebalancing, including potential cartelization initiatives among Southern countries. Supported by a robust methodological framework, this study provides a forward-looking pathway for transforming the tropical timber trade into a vector of equity and sustainability.

1. Introduction

The international trade in tropical hardwoods operates within a complex web of economic, environmental, and geopolitical dynamics. For decades, tropical wood products—particularly sawnwood—have constituted a core element of North–South trade relations, underpinned by deeply entrenched structural asymmetries. Producing countries, mainly located in Africa, Southeast Asia, and Latin America, continue to specialize in the export of raw or semi-processed timber, while consumer markets in OECD economies concentrate on high-value processing and downstream profit appropriation (Karsenty, 2007; Dauvergne & Lister, 2011; [1,2]). This trade architecture reflects a persistent and unequal international division of labor (Amin, 1972; Rodney, 1972; [3,4]), shaped historically by colonial extractive institutions and reinforced by global value chain (GVC) dynamics (Gereffi et al., 2005; Ponte & Sturgeon, 2014; [5,6]).

Although globalization has expanded market access and increased traded volumes, it has also exacerbated the structural dependency of tropical forest economies on volatile external demand and price instability (Kaplinsky, 2005; [7]). In response, many producing countries have adopted policies to promote Local Timber Processing (LTP), including log export bans, tax incentives, and public–private investment strategies. Yet, these initiatives have yielded limited industrial upgrading due to enduring structural bottlenecks such as technological gaps, underdeveloped infrastructure, low access to finance, and institutional weaknesses (Putzel et al., 2014; Clapp, 1995; [8,9]). These constraints raise critical questions about the capacity of tropical nations to manage their forest resources as engines for inclusive development and environmental sustainability (Seymour & Busch, 2016; [10]).

In this regard, tropical hardwood sawnwood represents a particularly strategic product for examining integration into global timber value chains. Situated between raw log exports and finished wood products, sawnwood is subject to a range of regulatory, industrial, and environmental pressures—including traceability standards, legality verification, and sustainability certification (Cashore et al., 2004; [11]). Yet, a substantial analytical gap persists. Most existing studies on the tropical timber trade adopt a case study or qualitative approach, often limited to individual countries, which restricts cross-national comparison and generalizability (Cerutti et al., 2011; [12]). Furthermore, few empirical studies capture the dynamic and endogenous causal relationships between key trade determinants—such as forest production, demographic pressure, or infrastructure quality—despite their central role in shaping path-dependent trade trajectories (Karsenty, 2007; Hausmann et al., 2007; [1,13]).

This study seeks to address these limitations through an original empirical strategy that combines econometric modeling with typological analysis. It applies the Common Correlated Effects–Autoregressive Distributed Lag (CCE-ARDL) approach, developed by Pesaran (2006) [14], which corrects for cross-sectional dependence and heterogeneous slopes—common features in global panel data settings. This framework allows for the identification of both short-term and long-term elasticities across countries, as well as latent global factors influencing export dynamics. In parallel, Granger causality testing is employed to capture feedback loops and bidirectional relationships between variables such as GDP, population size, timber production, and import dependency (p < 0.01), offering a more realistic representation of trade behavior in tropical forestry contexts.

The article’s contribution is not only methodological, but also conceptual. It draws on a hybrid theoretical framework, combining perspectives from international political economy (Ocampo, 2002; Bernstein & Campling, 2006; [15,16]), political ecology (Peluso, 1992; Robbins, 2012; [17,18]), and evolutionary economics (Hidalgo et al., 2007; Mehlum et al., 2006; [19,20]). Two key analytical innovations are proposed:

First, the notion of “pathological dependence”, which captures the enduring influence of colonial institutions, transnational corporate structures, and structural lock-ins on tropical timber trade patterns. In other words, the concept of “pathological dependence” refers here to the institutional and economic inertia that perpetuates the specialization of producing countries in the export of unprocessed raw materials, despite industrial policies. This notion is based on the work of Karsenty & Ongolo (2021) [21] and is operationalized through three indicators: (i) the local transformation rate (<20% in Africa); (ii) the concentration of forest concessions in the hands of foreign actors; (iii) the asymmetric elasticity of exports to demand shocks.

Second, the design of the “Fair and Digital Tropical Timber Trade Model (F&DTTT)”, which integrates traceability, value chain transparency, and South–South cooperation as levers for structural rebalancing. The “Fair and Digital Trade Model (F&DTTT)” proposes a normative framework that integrates the following: (1) blockchain traceability (through EUTR compliance indicators); (2) mirror clauses in trade agreements (measured by their adoption in FLEGT VPAs); (3) South–South industrial alliances (monitored through intra-regional flows of processed wood).

Research Issues and Rationale

This paper responds to three overarching issues:

• To what extent do macroeconomic, demographic, forestry, and structural factors influence tropical sawnwood exports among ITTO member states?
• Why do some countries succeed in upgrading within timber value chains while others remain stuck in raw commodity exports?
• What policy frameworks could support a more equitable, digitally governed model of the tropical timber trade?

The research questions derived from these issues include the following:

• What are the primary structural and cyclical determinants of tropical sawnwood exports?
• How do macroeconomic (e.g., GDP, population), forestry (e.g., production and processing rates), and commercial (e.g., infrastructure, pricing) variables interact to shape trade performance?
• Can countries be grouped into typologies based on specialization patterns and integration levels within GVCs?
• What instruments—fiscal, digital, and industrial—can promote structural transformation and sustainability in tropical timber-producing regions?

Hypotheses

H1: Higher economic growth expands export capacity by improving production and processing infrastructure.

H2: Larger populations reduce net exports by increasing domestic consumption and forest pressure.

H3: A higher share of local timber transformation is positively associated with export performance.

H4: Countries with strong logistics and traceability infrastructure are more competitive in value-added timber markets.

H5: Colonial legacies and corporate path dependency continue to structure trade geography.

H6: Governance innovations—including blockchain, mirror clauses, and regional alliances—can foster long-term rebalancing.

Study Objectives and Contributions

This article aims to shed new light on the persistent inequalities in the tropical timber trade by applying a robust dynamic panel model to a dataset covering 58 ITTO member countries from 1995 to 2022. Three key contributions are made:

• Identifying value capture asymmetries, particularly between producing regions (e.g., Central Africa, Southeast Asia) and major consuming markets (the EU, China), and uncovering how neglected variables—such as postcolonial institutional inertia and vertical market segmentation—affect export performance.
• Developing an innovative econometric framework, which performs the following:
  • Corrects for cross-sectional and slope heterogeneity (CD tests, Pesaran–Yamagata [22]).
  • Separates short- and long-term effects via dynamic elasticities.
  • Combines macroeconomic modeling with typological classification of trade behaviors.
• Proposing a forward-looking policy model, the Fair and Digital Tropical Timber Trade Model (F&DTTT), based on the following:
  • Blockchain-based traceability systems;
  • Mirror clauses to ensure fairness in trade agreements;
  • Regional industrial alliances to promote South–South collaboration.

Through this multidimensional approach, the study offers empirically grounded and policy-relevant insights into how the tropical timber trade can evolve toward a more inclusive, digitally transparent, and environmentally sustainable system—aligned with SDGs 8, 12, and 15.

1.1. Market Contextual Framework

Effects of Tropical Timber Exports on Economic and Demographic Variables

Tropical timber exports significantly influence forest production, economic growth, trade flows, and demographic dynamics [23]. In Cameroon, an export surtax on logs has encouraged domestic processing, with the forestry sector growing by 12.2% by the end of 2023 (La Voix du Centre, 2024; [24]). However, a complete log export ban—as implemented in Gabon—may undermine competitiveness and employment in the short term (Mpabe Bodjongo & Fotso Mbobda, 2021; [25]; FAO, 2001; [26]). These exports are a major source of foreign currency, contributing to a 0.5-point increase in Cameroon’s GDP in 2023 (Cerutti et al., 2011; [12], La Voix du Centre, 2024; [24]), yet they expose exporting countries to global market volatility and the negative effects of the “resource curse” (Sachs & Warner, 1999; [27]) or “Dutch disease” [28].

On the trade front, European imports peaked at USD 1.05 billion in early 2022 before dropping by 18% in volume and 27% in value in 2023 due to geopolitical shocks (FAO, 2001; [26]). Some Southeast Asian countries, however, have integrated these flows into high-value-added processing chains [22,23]. From a socio-environmental perspective, illegal logging contributes to deforestation, internal migration, and increased health risks (La Voix du Centre, 2024; [24]; Mpabe Bodjongo & Fotso Mbobda, 2021; [25] Du, Li & Zou, 2024 [29]). Local processing emerges as a key strategic lever [26] but remains hindered by structural barriers: inadequate infrastructure, high costs, limited access to finance, and FSC/PEFC certification requirements [12]. Despite proactive policies in Gabon, most local firms still struggle to meet international standards (La Voix du Centre, 2024; [24]).

Finally, flow typologies reveal a divide between raw product exporters (e.g., Gabon, Congo) and high-value-added re-exporters (e.g., Vietnam, China), depending on their level of industrial integration [21,30,31]. These trade patterns are shaped by trade agreements, environmental regulations (EUTR, FLEGT), and sectoral competitiveness [32,33].

2. Literature Review: Structural Inequalities and Sustainable Pathways in the Tropical Timber Trade

2.1. Theoretical Foundations of Resource-Based Development

The analysis of the tropical timber trade draws upon multiple theoretical frameworks that seek to explain the persistent structural inequalities affecting timber-rich developing countries.

The resource curse hypothesis (Sachs & Warner, 1995 [34], 1999 [27]; Auty, 1993, [35] 2001 [28]) remains a foundational paradigm, arguing that natural resource abundance is often associated with poor economic growth due to mechanisms such as Dutch disease, volatility of commodity prices, and profit-seeking behavior. These dynamics create disincentives for economic diversification and institutional reform.

However, this deterministic view has been contested by institutional economists, who emphasize the mediating role of governance structures and institutional quality (Acemoglu et al., 2001 [36]; Mehlum et al., 2006 [20]). Countries with inclusive political and economic institutions can escape the resource curse and harness natural wealth for long-term development (North, 1990 [37]; Rodrik et al., 2004 [38]).

Evolutionary and complexity economists have added a structural dimension to this debate by focusing on the role of productive capabilities and knowledge accumulation. Hausmann, Hwang, and Rodrik (2007) [13] and Hidalgo et al. (2007) [19] argue that the ability to transition towards more sophisticated exports depends on the “product space” and the evolutionary proximity between existing and potential new activities. This theory is particularly relevant to timber-exporting countries seeking to move from log exports to value-added processing.

From a political ecology and ecological economics standpoint, scholars like Martinez-Alier (2002) [39], Gerber (2011) [40], and Hornborg (2009) [41] have highlighted the biophysical and environmental limits of resource-based growth, emphasizing the ecological unequal exchange between core and periphery nations. This literature stresses how extractive industries such as tropical timber logging are embedded in unsustainable socio-metabolic regimes (Fischer-Kowalski & Haberl, 2007 [42]).

In parallel, the global value chain (GVC) literature has revealed how structural inequalities are reproduced through asymmetrical power relations between lead firms and suppliers. Gereffi et al. (2005) [5] and Ponte & Sturgeon (2014) [6] describe how buyer-driven governance structures in timber value chains often marginalize producers in the Global South, reinforcing dependency through limited upgrading opportunities and price volatility. This has been corroborated in forestry-specific studies (Lesniewska & McDermott, 2014 [43]; Cerutti et al., 2013 [44]), which document the challenges faced by timber-exporting countries in capturing greater value within the chain.

Furthermore, critical development scholars (Bebbington, 2012 [45]; Watts, 2015 [46]) emphasize the territorial and socio-political dimensions of extractive development, warning against path-dependent models that reinforce local disempowerment and elite capture. These perspectives are particularly salient in the context of Special Economic Zones (SEZs) and forest concessions, where exclusionary governance and opaque land deals often prevail (Corson, 2011 [47]; Pritchard et al., 2013 [48]).

Finally, sustainable development and transition studies provide normative frameworks for rethinking timber trade trajectories. The concept of just transitions (Newell & Mulvaney, 2013 [49]; Scoones et al., 2020 [50]) advocates for a reconfiguration of economic systems that integrates social justice, ecological sustainability, and economic viability. In the forestry sector, this implies shifting from an extractivist paradigm to one based on circular economy principles (Geissdoerfer et al., 2017 [51]) and equitable governance mechanisms (Ostrom, 2009 [52]).

2.2. Historical Roots of Contemporary Trade Structures

Current patterns in the tropical timber trade are deeply embedded in long-standing colonial legacies of extraction and asymmetry (Amin, 1972 [3]; Rodney, 1972 [4]). Historical institutionalist analyses (Karsenty, 2010 [53]; Putzel et al., 2013 [54]) have traced how colonial concession systems evolved into present-day forestry regimes, many of which still reproduce extractive logics and elite control.

Postcolonial scholars such as Mbembe (2000) [55] and Nkrumah (1965) [56] have emphasized the persistence of institutional frameworks designed for resource exploitation, often masked under neoliberal governance forms. From a dependency theory perspective (Frank, 1966 [57]; Cardoso & Faletto, 1979 [58]), the structural position of the Global South in the world economy continues to inhibit endogenous development and reinforces a form of neo-extractive dependency.

Comparative regional studies reveal contrasting trajectories. In Southeast Asia, some countries have made significant progress in value-added processing and domestic industrialization (Dauvergne, 1997 [59]; Barr, 2001 [60]; Resosudarmo et al., 2014 [61]). In contrast, many African timber-exporting countries remain locked into primary commodity exports, especially unprocessed logs (Oyono et al., 2005 [62]; Duguma, L. A., et al., 2018 [63]), making them highly vulnerable to price volatility and global demand shocks.

These divergent pathways are shaped in part by the long-standing influence of transnational corporations (Hart, S. L., 1995 [64]; Dauvergne & Lister, 2011 [2]), as well as international financial institutions (Rich, 1994 [65]; Adams, 2004 [66]), both of which have been widely criticized for reinforcing unsustainable development models based on raw material extraction.

Recent scholarship (Borras et al., 2011 [67]; Larder et al., 2017 [68]) has identified a contemporary reconfiguration of extractivism through megaprojects, Special Economic Zones (SEZs), and South–South trade alliances. Despite some shifts, these new forms often reproduce old patterns of dependency and territorial dispossession. Critical approaches from historical geography and political ecology (Moore, 2000 [69]; Mbembé & Hofmeyr, 2021 [70];) argue that tropical timber trade is not merely an economic process but a geopolitical project that has helped to structure global North–South relations for centuries.

2.3. Contemporary Trade Dynamics and Governance Challenges

In the contemporary globalized economy, tropical timber trade is deeply embedded in complex and asymmetric global value chains (Kaplinsky, 2005 [7]; Neilson et al., 2014 [71]). These chains are governed by lead firms and institutional frameworks that shape trade flows and forest governance outcomes.

The proliferation of certification schemes (Auld, G et al., 2015 [72]; Bartley, 2018 [73]) and legality verification mechanisms (Tacconi, 2007 [74]; Overdevest & Zeitlin, 2014 [75]) has contributed to reshaping forest governance. However, these instruments often focus on technical compliance rather than addressing the root causes of inequality, elite capture, and power asymmetries (Lemos & Agrawal, 2006 [76]; Auld, G et al., 2015 [72]).

The emergence of Asian markets, especially China, has reconfigured global trade routes and intensified demand for tropical timber (Sun et al., 2016 [77]; Hurmekoski et al., 2022 [78]). This shift has created new dependencies and regulatory blind spots, particularly in Africa and Southeast Asia, where state capacity and monitoring remain limited (Canby et al., 2008 [79]; Cerutti et al., 2011 [12]).

In parallel, growing climate and biodiversity concerns (Seymour & Busch, 2016 [10]; McDermott et al., 2015 [80]) have led to the adoption of stricter sustainability standards and zero-deforestation pledges. Yet, these often prioritize carbon metrics over local livelihood concerns and territorial rights (Leach, M. et al. 2018 [81]; Turnhout et al., 2017 [82]), raising critical questions about the equity of global environmental governance.

Despite these evolving dynamics, several persistent and interrelated governance challenges continue to affect the tropical timber trade:

• Widespread illegal logging and weak enforcement mechanisms (Contreras-Hermosilla, 2002 [83]; Kleinschmit et al., 2016 [84]);
• Unequal benefit distribution among stakeholders, often excluding indigenous and local communities (Pacheco et al., 2012 [85]; Hajjar et al., 2021 [86]; Hajjar et al., 2022 [87]);
• Limited effectiveness and market uptake of certification schemes (Gulbrandsen, 2014 [88]; van der Ven & Cashore, 2018 [89]);
• Land tenure insecurity and conflict over customary rights, particularly in forest frontier zones (Colchester, 2007 [90]; Larson et al., 2016 [91]; Rights and Resources Initiative, 2020 [92]).

New policy instruments such as the EU Deforestation Regulation (EUDR) and digital timber traceability platforms have emerged to address these issues. However, concerns remain regarding implementation gaps, cost burdens on producers, and data sovereignty in digital governance (Giurca & Di Nucci, 2022 [93]; De Pryck & Wiersum, 2017 [94]).

2.4. Emerging Solutions and Innovation Pathways

In response to longstanding governance and development challenges in the tropical timber sector, a variety of technological, political, and economic innovations are being explored to support more sustainable and equitable forest economies.

On the technological front, tools such as blockchain-based traceability systems (Howson et al., 2019 [95]) and satellite-based remote sensing (Tyukavina et al., 2022 [96]) are opening new possibilities for real-time monitoring, transparency, and supply chain accountability. Artificial intelligence and machine-learning algorithms are also being used to predict illegal logging hotspots and optimize forest management (Grantham et al., 2020 [97]; Dubayah et al., 2020 [98]).

In terms of governance innovations, policy mechanisms like the FLEGT Voluntary Partnership Agreements (Ochieng et al., 2013 [99]) and REDD+ programs (Angelsen et al., 2018 [100]) attempt to reconcile forest conservation with climate mitigation objectives. However, their effectiveness remains uneven, and some scholars have raised concerns over state capture, elite control, and weak enforcement mechanisms (Myers et al., 2017 [101]; Chomba et al., 2016 [102]).

From an economic perspective, strategies such as value-added timber processing (Sun et al., 2025 [103]; Hansen et al., 2015 [104]), community forestry enterprises (Baynes et al., 2015 [105]), and circular bioeconomy models (Korhonen et al., 2018 [106]; D’Amato et al., 2017 [107]) represent promising approaches to foster inclusive, climate-resilient, and decentralized forest-based development.

Nevertheless, political ecologists (Peluso, 1992 [17]; Robbins, 2012 [18]; Fairhead et al., 2012 [108]) caution against overly technocratic or market-oriented solutions that fail to confront underlying structural and power asymmetries in forest governance and trade. These scholars highlight the risk of green grabbing, carbon colonialism, and the erosion of local control over resources.

Institutional economists advocate for polycentric governance systems (Ostrom, 2009 [52]; Agrawal, 2007 [109]), which rely on diverse actors operating at multiple levels—local, national, and global—to address the complexity of forest management challenges. This model emphasizes local participation, adaptive learning, and institutional diversity.

Finally, critical trade and development scholars (Bernstein & Campling, 2006 [110]; Deacon, 2011 [111]) call for deeper structural transformations in global economic governance, challenging the current terms of trade that disadvantage commodity-exporting countries. Integrating social justice, ecological limits, and redistributive mechanisms into global trade regimes is increasingly seen as essential to achieving genuine sustainability.

2.5. Knowledge Gaps and Future Research Opportunities

Despite a rich body of literature, several critical gaps remain. There is a lack of dynamic econometric analyses of timber trade flows, and ecological variables are still insufficiently integrated with economic indicators. Furthermore, few studies have systematically compared regional trajectories, and South–South cooperation mechanisms in the forest sector remain underexplored.

This study contributes to addressing these gaps through the use of a novel econometric approach (CCE-ARDL modeling) and a comprehensive policy framework that integrates trade, environmental, and institutional dimensions.

3. Materials and Methods

3.1. General Methodological Framework

1. Integrated Analytical Framework (Figure A2; Appendix A)

Our conceptual foundation draws on a dual theoretical lineage: global political economy (Ponte & Sturgeon, 2014 [112]) and critical political ecology (Robbins, 2012 [18]). The proposed framework articulates the following core dimensions (Figure A2; Appendix A):

Logics of structural dependency, or pathological dependencies (Karsenty & Ongolo, 2012 [21]), encompassing the enduring effects of colonial legacies and persistent extractive specialization paths.

Global value chain (GVC) dynamics, marked by asymmetric power relations and fragmented governance architectures (Gereffi, 2005 [5]).

Ecological and normative constraints, especially those linked to deforestation, forest certification schemes, and international sustainability commitments (e.g., SDG 15).

2. Mixed Methodological Design

The empirical strategy integrates both quantitative modeling and typological analysis, structured into three interrelated components:

A. Econometric Modeling

The choice of the CCE-ARDL model is justified by the following:

• Analytical objective: Capturing both short- and long-term dynamics.
• Statistical robustness: Adjustment for cross-sectional dependence (e.g., Pesaran, 2006 [14]).
• Empirical validity: Confirmed through pre-estimation diagnostics (e.g., Pesaran–Yamagata).

B. Typological Analysis

A hierarchical clustering approach is used to categorize countries based on the following:

• Degree of local timber processing.
• Population size.
• Volume of tropical timber trade (imports/exports).
• Income level (GDP per capita).

C. Triangulation and Validation

Empirical robustness is reinforced through the following:

• Robust standard error estimates (e.g., Driscoll–Kraay).
• Comparative case studies (e.g., Gabon vs. Vietnam).
• Alignment with SDG metrics using FAO indicators (2023) [113].

D. Fair and Digital Tropical Timber Trade Model (F&DTTT), based on the following:

• Blockchain-based traceability systems and mirror clauses to ensure fairness in trade agreements;
• Regional industrial alliances to promote South–South collaboration (cf. Figure A2; Appendix A).

3.2. Data, Sources and Variables

The analysis is based on an unbalanced panel of 58 ITTO (International Tropical Timber Organization) member countries, observed over the period 1995–2022, i.e., a total of over 1500 observations. The choice of this long period enables us to capture both long-term structural dynamics (e.g., industrialization, economic transition) and cyclical effects (crises, global demand shocks).

3.2.1. Data Sources

Three major international databases are used to ensure the comparability and reliability of economic and sectoral statistics:

• ITTO—International Tropical Timber Organization: the main source for sectoral data on tropical timber. This database provides detailed annual series on production volumes, trade (import/export) and primary wood processing for each member country. Data are derived from national declarations harmonized to ITTO standards.
• FAO—Food and Agriculture Organization (FAOSTAT Forestry): used to complete data on forestry capacity, domestic production, and exploitable stocks. It can also be used to cross-reference certain environmental data with trade flows.
• World Bank Open Data: source of annual macroeconomic data (real GDP, population, growth rates, inflation), provided worldwide and harmonized to international standards. These data are used to introduce socio-economic determinants into models (domestic market size, aggregate demand, growth dynamics).

3.2.2. Description of Variables

All volume variables are expressed in natural logarithm in order to achieve the following:

• Reduce data variance (

  Table 1. Description of variables.

  Variable	Rating	Description	Source
  Exports	Log (ESNCit)	Log of export volume of non-coniferous tropical sawn timber	ITTO
  Domestic production	Log (P_SNCit)	Log of national production of processed tropical woods	FAO/ITTO
  Imports	Log (ISNCit)	Log volumes of imported tropical sawn timber	ITTO
  Population	Log (POPit)	Log of total population (proxy for domestic demand)	World Bank
  GDP growth	GDP_git	Real annual GDP growth rate (in %)	World Bank

  ),
• Make it easier to interpret coefficients in terms of elasticities,
• Satisfy stationarity conditions (variables I (1) before cointegration).

Additional control variables, such as per capita income or trade openness (imports + exports/GDP), were tested upstream but discarded from the final model for reasons of multicollinearity or robust insignificance.

3.2.3. Panel Structure

The panel is unbalanced, reflecting the heterogeneous reality of statistical and reporting capacities between countries in the South. However, a minimum completeness threshold (22 years out of 28) was set to include a country in the final analysis, guaranteeing sufficient representativeness.

3.2.4. Typological Analysis

This study employs a multivariate analytical approach to examine the structural determinants of international trade in non-coniferous hardwoods. The methodological framework is organized into three complementary stages: (1) a correlation analysis; (2) dimensionality reduction via principal component analysis (PCA); (3) hierarchical country classification.

Correlation Analysis

A Pearson correlation matrix is used to quantify the strength and direction of linear relationships between the explanatory variables. The Pearson correlation coefficient r between two variables X and Y is computed as follows (Pearson, 1895 [114]):

$$r_{X,Y}={\displaystyle \frac{\mathrm{Cov}\left(X,Y\right)}{{\sigma }_X\cdot {\sigma }_Y}}$$

where $\mathrm{Cov}\left(X,Y\right)$ notes the covariance between the variables, and ${\sigma }_X$, ${\sigma }_Y$ are their respective standard deviations. The statistical significance of correlations is tested at the conventional threshold of p < 0.05 (Student, 1908 [115]).

Principal Component Analysis (PCA)

Principal component analysis is applied to standardized data (mean-centered and scaled). This technique transforms a set of correlated variables into a new set of uncorrelated components (principal axes). Each principal component $C_k$ is a linear combination of the original variables (Hotelling, 1933 [116]):

$$C_k=\sum _{i=1}^p{w}_{ki}{X}_i$$

where $w_{ki}$ is the loading (eigenvector) associated with variable $X_i$, and pp is the number of original variables. Each component is ranked based on the variance it explains, measured by its eigenvalue λ_k:

$$\mathrm{Explained}\mathrm{Variance}={\displaystyle \frac{{\lambda }_k}{{\sum }_{i=1}^p{\lambda }_i}}\times 100$$

Typically, the first two components are retained if they capture a substantial proportion of the total variance (Jolliffe, 2002 [117]).

Hierarchical Clustering Analysis (HCA)

Hierarchical Ascendant Clustering is used to group countries into homogeneous clusters based on their Euclidean distances computed from standardized variables:

$$D\left(x_i,x_j\right)=\sqrt{\sum _{k=1}^p{\left(x_{ik}-x_{jk}\right)}^2}$$

where x_i and x_j represent the feature vectors of two countries. Ward’s method is applied to merge observations by minimizing the increase in within-cluster variance:

$$\mathsf{\Delta }\left(C_1,C_2\right)={\displaystyle \frac{\left|C_1\right|\cdot \left|C_2\right|}{\left|C_1\right|+\left|C_2\right|}}\cdot |{\mu }_1-{\mu }_2{|}^2$$

where μ₁ and μ₂ are the centroids of clusters C₁ and C₂, and ∣C∣ indicates cluster size. A dendrogram is used to determine the optimal number of clusters (Ward, 1963 [118]).

Validation and Interpretation

The coherence between the PCA structure and the clustering typology reinforces the methodological robustness. The results are interpreted in relation to the existing literature on tropical timber trade dynamics. This analytical strategy provides an integrated understanding of global trade patterns while allowing the identification of distinctive national profiles.

3.3. Econometric Modeling

3.3.1. Basic ARDL Model (PMG)

The ARDL (Auto-Regressive Distributed Lag) model is written as follows:

$$\begin{array}{cc}\mathsf{\Delta }\mathit{log}\left(ESN{C}_{it}\right)=& {\varphi }_i[\mathit{log}\left(ESN{C}_{i,t-1}\right)-{\theta }_1\mathit{log}\left(P\_SN{C}_{i,t-1}\right)-{\theta }_2\mathit{log}\left(ISN{C}_{i,t-1}\right)\hfill \\ & -{\theta }_3\mathit{log}\left(PO{P}_{i,t-1}\right)-{\theta }_{4}GDP\_g_{i,t-1}]+{\displaystyle \sum _{j=1}^p}{\gamma }_{1ij}\Delta \log \left(ESN{C}_{i,t-j}\right)\hfill \\ & +{\displaystyle \sum _{j=0}^q}{\delta }_{kij}\Delta X_{i,t-j}+{\mu }_i+{\epsilon }_{it}\hfill \end{array}$$

where

• ${\varphi }_i=speedofadjustment;$
• ${\theta }_k=long\text{-}termcoefficients;$
• ${\gamma }_{1ij},{\delta }_{kij}=short\text{-}termeffects;$
• $X_{it}=vectorofexplanatoryvariables;$
• ${\mu }_i,{\epsilon }_{it}=specificeffectsanderrorterm.$

3.3.2. Taking into Account Common Shocks (CS-ARDL-CCE)

To correct for unobserved factors, we use the CS-ARDL-CCE model (Pesaran, 2006 [14]):

$$\begin{array}{cc}\mathsf{\Delta }\mathit{log}\left(ESN{C}_{it}\right)=& {\alpha }_i+{\varphi }_i\left(\mathit{log}\left(ESN{C}_{i,t-1}\right)-X_{i,t-1}^{’}{\theta }_i\right)+{\displaystyle \sum _{j=1}^p}{\gamma }_{ij}\Delta \log \left(ESN{C}_{i,t-j}\right)\hfill \\ & +{\displaystyle \sum _{j=0}^q}{\delta }_{ij}\Delta X_{i,t-j}+{\lambda }_i^{’}{F}_t+{\epsilon }_{it}\hfill \end{array}$$

with

F_t = unobserved common factors represent common shocks affecting all countries, such as financial crises or global trends;

λ_i = joint effects coefficients.

3.3.3. Final Specification: CS-ARDL-CCE Model

The final specification retained for our empirical analysis is the Cross-Section Augmented ARDL model (CS-ARDL-CCE) without a common mean restriction. This model was selected based on a rigorous sequence of residual cross-sectional dependency tests, which are critical for identifying latent common factors across units in panel data. Specifically, we applied four main tests: the Pesaran CD test (2004, 2006, 2015, 2021), which detects basic cross-sectional dependence by rejecting the null hypothesis of independence when significant; the CDw test (Juodis & Reese, 2021 [119]), which adjusts for dynamic panel structures and remains robust in contexts where both N (cross-sections) and T (time periods) are large; the CDw+ test (Fan et al., 2015 [120]), an enhanced version of CDw designed to amplify power and detect weak dependencies; and the CD* test (Pesaran & Xie, 2006 [121]; Pesaran, 2021 [122]), which is particularly suitable for models like CCE, as it explicitly corrects for unobserved common factors through principal component analysis—here using four components.

The CS-ARDL-CCE model was preferred for its ability to accommodate fully heterogeneous slope coefficients, in line with the approaches proposed by Pesaran & Yamagata (2008) [123] and Kapetanios et al. (2011) [22], thereby avoiding any imposed average response across units. This specification allows us to simultaneously correct for cross-sectional dependence, capture unobserved common effects, and model asymmetric dynamic adjustments, all of which are essential features in explaining the heterogeneous behavior observed across countries. The decision was further supported by the results of the dependency tests mentioned above (Pesaran, 2004 [124]; CDw, CDw+; Pesaran, Ullah & Yamagata, 2008 [125]) and consistent with the broader empirical literature on institutional and structural heterogeneity (e.g., Knack & Keefer, 1995 [126]).

3.4. Preliminary Hypotheses Tests

Before estimating, we perform the following tests:

Stationarity: PESCADF (Pesaran, 2007 [127]) → confirmation of I (1) series.

Cointegration: Tests by Westerlund (2007) [128] and Pedroni (1999) [129] → existence of long-term relationships.

Residual Cross-Sectional Dependence: Pesaran test (CD test) → need to integrate common effects.

3.4.1. Stationarity

Tests used: PESCADF (Pesaran, 2007 [127]).

Objective: To verify whether the time series are stationary, that is to say whether their statistical properties (mean, variance, autocorrelation) are constant over time.

General ADF (Augmented Dickey–Fuller) test formula:

$$\mathsf{\Delta }{y}_{it}={\alpha }_i+{\beta }_i{y}_{it-1}+\sum _{j=1}^{p_i}{\gamma }_{ij}\Delta y_{it-j}+{\epsilon }_{it}$$

• $\mathsf{\Delta }{y}_{it}:firstdifferenceofthevariableofinterest{y}_{it};$
• ${\alpha }_i:individualconstant;$
• $$\begin{gathered} {\beta }_i:coefficienttested(hypothesis:{\beta }_i=0\Rightarrow {\beta }_i=0\Rightarrow unitroot,thereforenon\text{-} \\ stationary); \end{gathered}$$
• $p_i:numberofdelays;$
• ${\epsilon }_{it}:errorterm.$

Hypothesis of the unit root test:

H₀: The series has a non-stationary unit root.

H₁: The series is stationary.

PESCADF (Pesaran, 2007 [127]): This test is a version of the ADF test that takes into account cross-sectional dependence (cross-sectional dependence) via a group averaging method.

Conclusion: The tests conclude to integrated series of order 1, i.e., I (1), which means that the data becomes stationary after differentiation once.

3.4.2. Co-Integration

Tests used: Westerlund (2007) [128] and Pedroni (1999) [129].

Objective: To verify the existence of a long-term equilibrium relationship between non-stationary variables I (1).

(a) Pedroni test (1999)

The basic co-integration model is

$$y_{it}={\alpha }_i+{\delta }_{i}t+{\beta }_{1i}{x}_{1it}+{\beta }_{2i}{x}_{2it}+\dots +{\beta }_{Ki}{x}_{Kit}+{\epsilon }_{it}$$

• $y_{it}:dependentvariable;$
• $x_{kit}:independentvariables;$
• ${\alpha }_i,{\delta }_i:individueleffectsandtrend;$
• ${\epsilon }_{it}:residual\left(whichmustbestationaryifcointegration\right).$

H₀: No cointegration.

H₁: Present cointegration (stationary residues).

(b) Westerlund test (2007)

Based on the short-term adjustments of an ECM model (error correction model):

$$\mathsf{\Delta }{y}_{it}={\alpha }_i+{\varphi }_i\left(y_{it-1}-{\beta }_i{x}_{it-1}\right)+\sum {\gamma }_{ij}\mathsf{\Delta }{y}_{it-j}+\sum {\delta }_{ij}\mathsf{\Delta }{x}_{it-j}+{\epsilon }_{it}$$

• ${\varphi }_i:adjustmenttermtowardsequilibrium;$
• $Nullhypothesis:\varphi i=0\Rightarrow noreturntoequilibrium\to nocointegration.$
• $Alternativehypothesis:\varphi i<0\Rightarrow adjustmenttoequilibrium\to cointegration.$
• Conclusion: Both tests confirm the existence of long-term relationships between variables.

3.4.3. Dependency Between Cross Sections

Test used: Pesaran CD (cross-sectional dependence) test.

Formula:

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

• $\widehat{{\rho }_{ij}}:correlationofresidualsbetweenunitsiandj;$
• $T:numberofperiods;$
• $N:numberofcountries.$

Hypothesis:

H₀: Independence between cross units.

H₁: Dependence between units (non-zero correlations).

Conclusion: The rejection of H₀ indicates the presence of dependence between countries. It is therefore necessary to integrate common effects or global factors into the models

3.5. Causality Tests

3.5.1. Dumitrescu–Hurlin Test (2012)

Granger-type causality in a heterogeneous panel:

$$y_{it}={\alpha }_i+\sum _{k=1}^K{\gamma }_{ik}{y}_{it-k}+\sum _{k=1}^K{\beta }_{ik}{x}_{it-k}+{\epsilon }_{it}$$

Hypothesis:

• $H_0:{\beta }_{i1}={\beta }_{i2}=\dots ={\beta }_{iK}=0\hspace{1em}\forall i\Rightarrow nocausality.$
• $H_1:\exists i\in \left[1,N\right],\hspace{1em}{\mathsf{\beta }}_{ik}\ne 0foratleastonek\Rightarrow causalitypresent.$
• $H_0:x_{it}\nrightarrow y_{it}\left(absenceofcausalityfromxtoy\right).$
• $H_1:x_{it}\to y_{it}\left(presenceofcausality\right).$

3.5.2. HPJ Test (Het Panel Joint)

• Proposed to detect a joint causality in a heterogeneous panel, robust to individual specificities. Its general formulation is based on the aggregation of individual causality test statistics, taking into account the structural heterogeneity of the panel.

HPJ test benefits:

• Suitable for panels with transverse dependency.
• More robust than conventional Granger tests in a heterogeneous context.
• Allows us to conclude a global causality while allowing different effects according to the units.

3.6. Methodological Contributions

• Dynamic ARDL models coupled with robust CCE models.
• Advanced panel causality tests.
• An approach tailored to the specific needs of tropical markets.
• Dynamic and structural modeling of the tropical timber trade.
• Consideration of international interdependencies.
• Explicit decoupling of short- and long-term effects.
• Multi-level validation of economic relationships (stationarity, cointegration, causality).

4. Results

4.1. Exploratory and Structural Analysis of the ITTO Market

4.1.1. Principal Component Analysis (PCA)

To better understand the dynamics between the main variables in the non-coniferous hardwood trade, a Pearson correlation matrix was calculated (

Table 2. Pearson correlation matrix (with significance).

Variable	Log (ESNCit)	Log (ISNCit)	Log (P_SNCit)	Log (POP_Tit)	GDP_git
Log (ESNCit)	1.00
Log (ISNCit)	0.14 **	1.00
Log (P_SNCit)	0.66 ***	0.40 ***	1.00
Log (POP_Tit)	0.14 **	0.48 ***	0.53 ***	1.00
GDP_git	−0.12 **	0.00	0.08	0.15 ***	1.00

Stars indicate significance: **: p < 0.01 and ***: p < 0.001.

). The results show a strong positive correlation between national production (logP_SNC) and exports (logESNC) (r = 0.66, p < 0.001), suggesting that countries with higher production are also those that export more. Moderately positive correlations also appear between total population (logPOP_T) and imports (logISNC) (r = 0.48, p < 0.001), which may indicate that a large population increases demand for non-coniferous wood. In contrast, GDP growth (GDP_g) is weakly correlated with the other variables, and even slightly negatively correlated with logESNC (r = −0.12, p < 0.01), highlighting a certain independence between macroeconomic performance and timber trade flows in the observed sample (Figure 1). PCA on centered-reduced data shows that the first two components together explain 78.2% of the total variance, as expanded on below.

• Axis 1 (52.3% of variance):

This first factor contrasts countries (Figure 2) with high production/export (logP_SNC, logESNC) with those whose structure is based more on domestic demand (logISNC, logPOP_T).

• Axis 2 (25.9% of variance):

This secondary factor is dominated by GDP growth (GDP_g), which appears to vary independently of the other trade dimensions. Projection onto the factorial plane shows a clear distinction between tropical producing countries (more to the left of the F1 axis) and importing countries (to the right of F1), confirming the results of the CAH typology. These results highlight an asymmetrical pattern of trade in non-coniferous hardwood, in which tropical African countries appear as suppliers, and the major demographic powers (Asia and the West) as the main consumers. The influence of population on import volumes confirms the role of demographics as a driver of demand, while the absence of a strong link with GDP growth suggests that the timber trade remains dependent on structural (resources, industry, logistics) rather than cyclical logics.

4.1.2. Typological Analysis (Hierarchical Ascending Classification)

The use of hierarchical ascending classification (HAC) on standardized indicators enabled the identification of four main country profiles in the tropical sawnwood trade (Figure 3):

• Group 1: Export-oriented producing countries—This group includes nations like Gabon and the Republic of the Congo, which demonstrate high levels of production (logP_SNC) and exports (logESNC), coupled with relatively small populations (logPOP_T), suggesting an economy heavily reliant on external markets. Brazil, although highly populated, shows limited domestic use due to factors such as a low degree of industrial development, modest average income levels, and internal structural constraints.
• Group 2: Countries with limited processing and marginal trade integration—Comprising nations such as Angola and Cameroon, this group is characterized by low levels of production, imports, and exports. These countries are either weakly integrated into international timber value chains or function as minor re-exporters with minimal processing capacity and low value addition.
• Group 3: High-value-added re-exporting hubs—These countries act as pivotal intermediaries in the tropical hardwood trade by importing raw or semi-processed timber for transformation, storage, and re-export. Typically, they generate substantial added value in the process. The group includes key Asian economies (e.g., Vietnam) and European players like Germany and France.
• Group 4: Major consumer markets and densely populated countries—Encompassing nations such as China and India, this group is defined by intense domestic demand (high logISNC) and large population sizes. It also includes countries with strong processing and re-export capacity like the Netherlands, Belgium, and Singapore. While these actors play a central role in the globalization of timber trade, their prominence raises concerns related to illicit deforestation practices and disproportionate value capture to the detriment of producing countries.

This typology reveals the heterogeneous positioning of countries in the tropical sawnwood market and draws attention to the varied economic functions and environmental stakes associated with their participation.

4.2. Dynamic Analysis of the Effects of Non-Coniferous Hardwood Exports

4.2.1. Descriptive Analysis in N and Large T Panels

The descriptive analysis provides a detailed interpretation of the international trade in non-coniferous timber. The analysis is structured according to the sections and what each test brings to the large N and T panel study.

Table 3. Characteristics of panel variables.

Variable	Average (Total Standard Deviation)	Standard Deviation Between (Cross-Sectional)	Intra Standard Deviation (Over Time)
ESNCit	256,260.4 (527,390.4)	527,390.4	204,238.6
P_SNCit	2,061,532 (5,304,556)	5,304,556	2,978,146
ISNCit	341,003.7 (1,011,724)	1,011,724	4,983,961
POP_Tit	100 M (272 M)	272 M	24.3 M
GDP_git	5.38% (14.02%)	14.02%	3.97%

summarizes the characteristics of the panel variables.

Descriptive analysis of the data (Table 3) reveals considerable heterogeneity between ITTO member countries in terms of exports, production, and imports of sawn non-coniferous timber. The high standard deviations between countries (e.g., 5,304,556 m³ for production) reflect marked structural differences in industrial capacities and trade dynamics. In addition, significant temporal variations, particularly within exports and imports, underline the need to model these data using dynamic approaches.

4.2.2. Cross-Sectional Dependence (CSD) Test

The results of cross-sectional dependence (CSD) tests reveal significant interdependence (

Table 4. Cross-sectional dependence test for variables of interest.

Variable	CD	CDw	CDw+	CSD
Log (ESNCit)	9.17 (0.000)	−2.62 (0.009)	2627.31 (0.000)	6.71 (0.000)
Log (P_SNCit)	10.47 (0.000)	−1.71 (0.087)	3462.08 (0.000)	6.35 (0.000)
Log (ISNCit)	−1.83 (0.067)	−1.82 (0.069)	3654.34 (0.000)	−0.99 (0.324)
Log (POP_Tit)	66.13 (0.000)	−2.93 (0.003)	7270.86 (0.000)	−1.25 (0.211)
GDP_git	78.51 (0.000)	1.63 (0.104)	3296.22 (0.000)	−2.70 (0.007)

) between countries for the model’s main variables (exports, imports, GDP growth, etc.), confirmed by significant CD, CDw+, and CD* statistics at conventional thresholds (p < 0.01). This dependence can be explained by trade or macroeconomic spillover effects, reinforcing the need to use estimators robust to CSD, such as Common Correlated Effects (CCEs) or Driscoll–Kraay.

4.2.3. Slope Heterogeneity Test

Furthermore, tests for slope heterogeneity (Pesaran–Yamagata, Blomquist–Westerlund) show highly significant p-values (<0.05), indicating that coefficients vary substantially from country to country (

Table 5. Multiple slope heterogeneity tests.

Test	Statistics	p-Value	Adjusted Statistics	Adjusted p-Value
Pesaran–Yamagata (CSA)	1.327	0.185	2.064	0.039
PY (AR adjusted)	2.807	0.005	3.285	0.001
Blomquist and Westerlund	2.807	0.005	3.285	0.001
PY (Single)	−7.222	0.000	2.319	0.020

). Thus, the relationships between macroeconomic determinants and timber trade flows cannot be assumed to be homogeneous. This characteristic justifies the use of panel models with heterogeneous coefficients, such as Mean Group (MG), CCE, or Pooled Mean Group (PMG), to better capture the structural diversity of trade behavior within the panel.

4.2.4. Stationarity Tests (Unit Root)

In order to verify the stationarity properties of the series used in the dynamic modeling of non-coniferous sawnwood exports, several unit root tests adapted to panel data were applied, comprising the PESCADF test.

PESCADF Test Results

Table 6. PESCADF test.

Statistics		Independent Variables
		Log (ESNCit)	Log (P_SNCit)	Log (ISNCit)	Log (POP_Tit)	Log (GDP_git)
Level	Z[t-bar]	0.566	0.712	−0.38	0.811	−2.996
	p-value	0.714	0.762	0.352	0.765	0.001
First difference	Z[t-bar]	−5.233	−4.241	−4.806	−5.481	−10.557
	p-value	0.000	0.000	0.000	0.000	0.000

of the PESCADF test shows that, for all variables except GDP growth (GDP_g), the Z[t-bar] statistics regarding level are positive and insignificant (p-values > 0.05), indicating non-stationarity regarding level.

At first difference, all variables, including GDP_g, become significant with zero p-values, suggesting that they become stationary after differentiation. This shows that the logESNC, logP_SNC, logISNC, logPOP_T, and GDP_g series are integrated with order 1, i.e., I (1).

4.2.5. Cointegration Tests

Confirmation of the order of integration of the variables enables us to examine the existence of a cointegrating relationship between exports of non-coniferous sawn timber (logESNC) and the explanatory variables. Two types of tests are used: the Westerlund test and the combined Westerlund and Pedroni test.

Westerlund Test Results (ECM)

The results of Westerlund’s ECM test (see

Table 7. Westerlund test (ECM).

Statistics		Independent Variable
		Log (P_SNCit)	Log (ISNCit)	Log (POP_Tit)	Log (GDP_git)
Gt	Value	−2.578	−2.224	−2.663	−2.288
	z-value	6.783	3.788	7.506	4.325
	p-value	0.000	0.000	0.000	0.000
Ga	Value	−9.523	−6.586	−4.283	−6.874
	z-value	3.33	0.779	4	0.376
	p-value	0.000	0.782	1	0.647
Pt	Value	−16.172	−14.129	−12.923	−14.923
	z-value	5.177	3.122	1.909	3.921
	p-value	0.000	0.001	0.028	0.000
Pa	Value	−7.861	−6.643	−4.423	−7.542
	z-value	6.228	4.138	0.33	5.681
	p-value	0.000	0.000	0.371	0.000

) reveal that several forms of the test (Gt, Pt, Ga, Pa) detect a cointegrating relationship for the explanatory variables logP_SNC, logISNC, logPOP_T and GDP_g:

The Gt and Pt statistics are highly significant for all variables (p < 0.01), indicating the presence of a long-term adjustment mechanism in the export equation. The Ga test reveals strong significance for the variable logP_SNC (p = 0.000), suggesting a consistent long-term relationship for this variable across countries; however, it is not significant for others, such as logISNC (p = 0.782), pointing to possible heterogeneity in long-term dynamics among countries. Additionally, while most Pa values are statistically significant, the exception is logPOP_T, which has a p-value of 0.371, indicating a potential absence of cointegration between population and exports in certain countries. Taken together, these results support the existence of a stable long-term relationship between exports and key explanatory variables, particularly domestic production, imports, and economic growth.

Westerlund and Pedroni Combined Test

The combined results provide further support for the existence of cointegration.

The Westerlund variance ratio (

Table 8. Westerlund and Pedroni combined test.

Tests	Statistics	Values	p-Value
Westerlund test for cointegration	Variance ratio	−2.0431	0.0205
Pedroni test for cointegration	Modified Phillips–Perron t	4.0429	0.000
	Phillips–Perron t	−7.0769	0.000
	Augmented Dickey–Fuller t	−6.4662	0.000

) test yields a significant result (p = 0.0205), confirming the existence of a long-term relationship between the variables. Additionally, all three versions of the Pedroni test—Modified Phillips–Perron, Phillips–Perron t, and PESCADF—are significant at the 1% level, providing strong evidence of cointegration among the variables in the export model. Stationarity analyses show that all series are integrated with order one (I (1)), while the consistent outcomes of the cointegration tests further confirm the presence of stable long-term relationships between non-coniferous sawnwood exports (linked to processing) and key economic explanatory variables. These findings support the appropriateness of employing a long-term dynamic panel model—such as PMG or CCE—to analyze international timber trade among ITTO member countries.

4.2.6. Estimates

Estimates of export determinants from the dynamic models ARDL_PMG, ARDL_FE, NoCS_ARDL_CCE, and CS_ARDL_CCE (

Table 9. Dynamic model estimates.

VARIABLES	NoCS_ARDL_CCE	CS_ARDL_CCE (Optimal)	ARDL_PMG	ARDL_FE
Dependent Variable (logESNC)
LD.logESNC	−0.078 ** (0.039)	−0.106 *** (0.035)
Short-Term Effects
D.logP_SNC	0.066 (0.130)	0.120 (0.093)	0.115 (0.078)	0.208 *** (0.074)
LD.logP_SNC	0.286 ** (0.129)	0.261 ** (0.117)
D.logISNC	0.269 *** (0.058)	0.333 *** (0.061)	0.257 *** (0.046)	0.117 *** (0.022)
LD.logISNC	0.042 (0.049)	0.072 (0.049)
D.logPOP_T	3.743 (5.802)	3.065 (4.422)	−5.923 (7.097)	−0.014 (0.043)
D.GDP_g	0.016 ** (0.008)	0.011 * (0.007)	0.001 (0.005)	0.005 (0.006)
LD.GDP_g	−0.001 (0.006)	0.005 (0.005)
Constant		0.011 (0.020)	−4.802 *** (0.473)	2.098 *** (0.625)
Long-Term Effects
Adjust. Term (lr_logESNC)	−1.078 *** 0.000	−1.106 *** (0.035)	−0.336 *** (0.026)	−0.392 ***(0.021)
lr_logP_SNC	0.360 *** (0.132)	0.380 *** (0.111)	0.001 (0.005)	0.378 *** (0.108)
lr_logISNC	0.286 *** (0.071)	0.365 *** (0.076)	−0.001 (0.022)	−0.001 (0.049)
lr_logPOP_T	1.201 *** (0.322)	1.287 ***(0.188)	1.389 *** (0.210)	0.041 (0.054)
lr_GDP_g	0.003 (0.013)	0.011 (0.010)	0.016 ** (0.008)	−0.019 (0.019)
lr__cons		0.019 (0.015)	−4.802 *** (0.473)	2.098 *** (0.625)
Comments (N/T)	1507 (58/26)	1507 (58/26)	1565 (58/26)	1565 (58/26)
CD-Statistic (residual)	−1.48	0.72	1.16	0.47
PESCADF (P)	−13.468 ***	14.013 ***	5.61 ***	7.818 ***
ADF-Fisher	387.821 ***	340.2267 ***	409.0523 ***	414.126 ***

Standard errors in parentheses, *** p < 0.01, ** p < 0.05, * p < 0.1.

).

Residual Cross-Sectional Dependency Test (CD Test)

Table 10. Residual cross-sectional dependence tests for each model (CD test).

Model	CD (Pesaran)	CDw (Juodis-Reese)	CDw+ (Fan et al.)	CD* (Pesaran-Xie)
CS_ARDL_CCE (Optimal)	12.28 (0.000)	2.09 (0.037)	1707.99 (0.000)	−1.64 (0.101)
NoCS_ARDL_CCE	7.41 (0.000)	−0.72 (0.472)	3151.67 (0.000)	3.23 (0.001)
ARDL_PMG	12.24 (0.000)	1.16 (0.245)	1503.45 (0.000)	3.41 (0.001)
ARDL_FE	53.62 (0.000)	−2.03 (0.042)	4306.45 (0.000)	0.47 (0.637)

presents a clear and structured interpretation of the results of the cross-sectional dependence (CD) test applied to several dynamic models in a panel context with large N and T, based on the various test statistics reported.

Justification for Choosing the Optimal Model: CS_ARDL_CCE

In analyzing the determinants of non-coniferous timber exports among ITTO member countries, the CS_ARDL_CCE model (Auto-Regressive Distributed Lag with Common Correlated Effects and heterogeneous slopes) emerged as the most appropriate specification. This model is particularly suited for panels with a large number of countries (N = 58) and moderate time series (T = 22–28 years), as it accounts for both structural heterogeneity across countries (Pesaran & Yamagata, 2008 [123]; Kapetanios et al., 2011 [22]) and cross-sectional dependence (Pesaran, 2006 [14]).

To validate this model choice, we conducted four residual cross-sectional dependence tests (Table 10):

• The Pesaran CD test (2004) indicates strong dependence (CD = 12.28, p < 0.01).
• The CDw test (Juodis & Reese, 2021 [119]), tailored for dynamic panels, also detects dependence (CDw = 2.09, p = 0.037).
• The CDw+ test (Fan et al., 2015 [120]), designed to capture weak dependence, reveals high correlation (CDw+ = 1707.99, p < 0.01).

However, the CD* test (Pesaran & Xie, 2025 [130]), which adjusts for unobserved common factors via principal component analysis (four components retained), indicates no significant remaining cross-sectional dependence (CD* = −1.64, p = 0.101). These results confirm that the CS_ARDL_CCE model effectively controls for latent common factors, while highlighting the importance of remaining cautious about persistent structural dependence. By contrast, alternative specifications (NoCS_ARDL_CCE, ARDL_PMG, ARDL_FE) exhibit even higher levels of residual dependence, despite adjustments, casting doubt on the reliability of their estimates.

Therefore, the CS_ARDL_CCE model stands out as the optimal methodological choice for this study because it achieves the following:

(i)

Captures structural heterogeneity through country-specific slope coefficients (Pesaran & Smith, 1995 [131]);

(ii)

Corrects for cross-sectional dependence via the CCE approach (Pesaran, 2006 [14]);

(iii)

Jointly estimates short-term dynamics, long-term relationships, and the speed of adjustment toward equilibrium (Chudik & Pesaran, 2015 [132]).

In sum, the CS_ARDL_CCE model provides a robust empirical framework for examining the structural and dynamic drivers of non-coniferous timber exports across countries, while minimizing the risk of biased inference due to cross-sectional dependence.

Detailed Interpretation of Results (Model CS_ARDL_CCE)

The estimation of the optimal dynamic model, CS_ARDL_CCE, provided a fine-grained reading of the determinants of non-coniferous hardwood lumber exports (logESNC) in a heterogeneous panel framework with cross-dependencies. By combining short- and long-term dynamic effects with structural specificities between countries, this model proved particularly well-suited to the characteristics of the data observed.

• Short-term effects

Analysis of the short-term coefficients highlights several key results. Firstly, the effect of variation in local non-coniferous wood production (ΔlogP_SNC) appears positive but insignificant, suggesting that a one-off increase in domestic supply does not immediately translate into an increase in exports. On the other hand, imports (ΔlogISNC) show a significant and positive effect: a 1% increase in imports is associated with a 0.33% increase in exports in the short term. This result could reflect local processing for re-export or a complementarity effect between local and imported wood. Finally, economic growth (ΔGDP_g) exerts a slightly significant effect: a 1% growth in GDP leads to an increase of around 1.1% in exports, which could be explained by an improvement in logistics infrastructures or increased global demand (

Table 11. Interpretation of results CS_ARDL_CCE.

Short-Term Effects (Short-Term)		Dependent Variable: Log (ESNC)
Explanatory variable	Coefficient	Interpretation
ΔlogP_SNC (Production)	0.120 (ns)	Positive effect, but not significant. In the short term, a one-off increase in local non-coniferous tropical hardwood lumber production (processing) has no statistically assured effect on exports.
ΔlogISNC (Imports)	0.333 *	Highly significant. In the short term, a 1% increase in imports of non-coniferous sawnwood leads to a 0.33% increase in exports. This could suggest a logic of local processing and re-export or a matching effect between local and imported supply.
ΔGDP_g (GDP growth)	0.011 *	Significant at the 10% level. Stronger economic growth slightly boosts exports in the short term (+1% GDP → +1.1% exports). This could reflect an improvement in logistics capacities or a growing global demand effect.
Effect of adjustment towards equilibrium
Term	Coefficient	Interpretation
Adjustment Term (lr_logESNC)	−1.106 ***	The adjustment term associated with the lagged dependent variable is very significant and negative at 1% (−1.106 ***, standard deviation: 0.035), indicating a high speed of convergence to the long-term equilibrium after a shock. This suggests strong resilience of the exporting economic system in the medium term.
Long-term effects		Dependent variable: log (ESNC)
Explanatory variable	Coefficient	Interpretation
lr_logP_SNC	0.380 ***	Highly significant. A 1% increase in national production of non-coniferous hardwood sawn timber leads to a 0.38% increase in long-term exports. This validates a positive structural relationship between local production and export performance.
lr_logISNC	0.365 ***	Also, highly significant. The effect is strong: 1% more imports leads to 0.37% more exports in the long term. This confirms vertical commercial integration in the non-coniferous wood sector.
lr_logPOP_T	1.287 ***	Here, it is highly significant: an increase in total population is positively associated with exports in the long term. This may indicate a structural development effect: the larger a country’s population, the more it develops commercial and productive infrastructures.
lr_GDP_g	0.011 (ns)	Not significant in the long term. Economic growth does not appear to have a structural effect on long-term exports in this sector.
Constant (lr_cons)	0.019 (ns)	No significant long-term effect of the constant.

ns: not significant and *** p < 0.01, * p < 0.1.

).

• Adjustment towards balance

The adjustment term associated with the lagged dependent variable is very significant and negative at 1% (−1.106 ***, standard deviation: 0.035), indicating a high speed of convergence to the long-term equilibrium after a shock. This suggests strong resilience of the exporting economic system in the medium term.

• Long-term effects

Long-term results reveal robust structural relationships. Domestic non-coniferous wood production (logP_SNC) is positively and significantly related to exports (+0.38%), confirming that a sustained increase in production translates into better export performance. Similarly, imports (logISNC) maintain a significant effect (+0.37%), supporting the hypothesis of vertical integration of the sector on an international scale. Total population (logPOP_T), which is only significant in this model, has a significant effect (+1.29%), suggesting that broader demographic structures are correlated with greater trade capacity. On the other hand, neither long-term economic growth (GDP_g) nor the constant show significant effects in this framework.

4.2.7. Testing for Granger Causality

Granger non-causality tests conducted within a heterogeneous and dynamic panel framework further validate and complement the insights derived from the CS_ARDL_CCE model (

Table A6. Granger test from Dumitrescu and Hurlin (2012) [133] (bidirectional).

Relationship Tested	Z-Bar (p-Value)	Z-Tilde (p-Value)	Interpretation
ESNC ↔ P_SNC	6.65 ↔ 8.80	0.0000	Very strong two-way causality between local production and exports: one influences the other. This confirms the dynamic adjustment logic of the ARDL model.
ESNC ↔ ISNC	4.08 ↔ 5.94	0.0000	The relationship is also bidirectional: Exports react to imports and vice versa. This reinforces the idea of an integrated “import–processing–export” cycle.
ESNC ↔ POP_T	7.36 ↔ 20.43	0.0000	Very strong bilateral relationship: Demographic weight affects exports and vice versa. This suggests a structural link between population growth and export specialization.
ESNC ↔ GDP_g	5.14 ↔ 2.58	0.0001 0.0097	Bilateral relationship, but less strong on the GDP → export side. This confirms the ARDL results: GDP has only a marginal long-term role in explaining exports.

 Appendix A). The analysis reveals that bilateral causal relationships are predominant, especially among production, imports, and exports—an economically consistent finding for a wood value chain that relies on processing and vertical integration. Population emerges as a structurally significant variable, serving as a deep-rooted driver of export specialization. While GDP plays a comparatively less central role, it still functions as an important factor in cyclical adjustments within the export dynamics.

Analysis of Dynamic Causality: Granger Tests in Heterogeneous Panels

Dynamic causality relationships between explanatory variables and exports of sawn non-coniferous wood (logESNC) were examined using Granger tests adapted to heterogeneous panel data. Two complementary approaches were mobilized: the HPJ test (Het Panel Joint test with Bootstrap) and the test proposed by Dumitrescu and Hurlin (2012) [133], making it possible to incorporate both individual heterogeneity and cross-sectional dependence.

The results of the HPJ test reveal (

Table A5. HPJ test results (Het Panel Joint test—Bootstrap).

Explanatory Variable (Lagged)	Coefficient	Error-SD	Interpretation
L.logP_SNC (Production)	−0.004	0.063	Trend towards a marginally significant unidirectional relationship towards ESNC. In other words, past production has little influence on current exports.
L.logPOP_T (Population)	−0.979 **	0.044	Significant effect: Past population causes Granger exports. This reinforces the idea of a structural development effect.
L.GDP_g (GDP)	−0.012 *	0.006	Significant effect at 10%: Past economic growth has a causal impact on short-term exports.
L.logISNC (Imports)	−0.077 *	0.039	Significant causal effect of past imports on exports. This suggests an integrated import–processing–export mechanism.
Comments (N/T)	1623 (58/26)
HPJ Wald test:	16.5364
p-value	0.0024

HPJ Wald test overall = 16.53, p = 0.0024: the null hypothesis of no joint causality is rejected. This means that, overall, the explanatory variables (P_SNC, ISNC, GDP_g, POP_T) significantly cause exports (log ESNC) in a heterogeneous panel. ** p < 0.05, * p < 0.1.

, Appendix A) an overall significance of causality, with a Wald test value of 16.53 (p = 0.0024), indicating that the explanatory variables taken together (logP_SNC, logISNC, GDP_g, logPOP_T) significantly influence exports. More specifically, past GDP (GDP_g) and imports (logISNC) show significant effects at the 10% level on contemporary exports, suggesting a plausible economic link between growth, trade integration, and export performance in the short term. The demographic variable (logPOP_T) shows a significant influence at 5%, reinforcing the hypothesis of a structural effect linked to the size of the domestic market. Past production (logP_SNC) shows a weakly significant effect (p = 0.063), reflecting a potentially unidirectional relationship (Dumitrescu et al., 2012 [133]).

The results of the Dumitrescu and Hurlin test (Table A6, Appendix A) confirm and clarify this causal dynamic by identifying significant two-way relationships. All the pairs of variables tested (ESNC ↔ P_SNC, ESNC ↔ ISNC, ESNC ↔ POP_T, ESNC ↔ GDP_g) show highly significant Z-bar and Z-tilde coefficients (p < 0.01), signaling robust bidirectional causalities. These results point to a dynamic interdependence between exports and their structural and cyclical determinants. For example, the relationship between production and exports is doubly causal, corroborating the role of an endogenous adjustment logic in forest value chains. Similarly, import flows, essentially destined for local processing, appear to be a key link in export performance (Gereffi, et al., 2005 [5]; Sun, L., & Canby, K., 2021 [134]).

Cross-referenced with the results of the CS_ARDL_CCE model (Table 9, Appendix A), this analysis supports the idea that export dynamics in ITTO countries are based on a complex interaction between structural (population, production) and cyclical (growth, imports) factors, with significant short- and long-term elasticities. Economic growth, although weakly linked to exports in the long term, contributes to their dynamism in the short term via opportunity or infrastructure effects. Thus, the robustness of the causal relationships detected reinforces the economic and empirical validity of the ARDL model selected and suggests concrete implications for development strategy and trade policy in the forestry sector.

Analysis of Endogeneity Biases and Interpretational Implications

Our study acknowledges three potential sources of endogeneity that warrant detailed discussion. First, Granger causality tests (Dumitrescu–Hurlin) reveal significant bidirectional relationships (p < 0.01) between (1) exports and domestic production (Z = 8.80) and (2) exports and imports (Z = 5.94). Second, residual analysis shows potential correlations (ρ ≈ 0.35–0.45) between error terms and certain explanatory variables. Third, the Hausman test (H = 68.3, df = 3) strongly rejects (p < 0.001) the strict exogeneity hypothesis (

Table A4. Advanced endogeneity.

Test	Statistic	Interpretation
Hausman (FE vs. CCE)	H = 68.3 ***	CCE model preferred

***: Means that the Hausman test is significant at the threshold of 0.01.

, Appendix A).

The CS-ARDL-CCE model mitigates these biases through three mechanisms: (1) inclusion of dynamic lags (up to t − 2); (2) cross-sectional common effects (CCE) correction; (3) separate estimation of short/long-term effects. However, the long-term elasticities (0.38 for P_SNC, 0.365 for ISNC) should be interpreted as conditional associations within an interdependent system, with a ±15% margin of error according to our Monte Carlo simulations.

For future research, we recommend the following: (1) employing instrumental variables (rainfall as instrument for P_SNC, trade distance for ISNC); (2) applying dynamic panel GMM with cross-sectional dependence correction; (3) incorporating firm-level microeconomic data to better isolate causal effects. These limitations do not invalidate our main findings but call for cautious interpretation of the estimated coefficients.

Extended Validation of the CS-ARDL-CCE Model

Our comprehensive diagnostic testing confirms the robustness of the CS-ARDL-CCE specification. Residual analysis reveals normal distribution (Jarque–Bera: χ² = 3.21, p = 0.201) without significant autocorrelation (Wooldridge: F = 1.87, p = 0.172), though mild heteroskedasticity is detected (Breusch–Pagan: χ² = 28.34, p = 0.059) (

Table A1. Residual diagnostics.

Test	Methodology	Results	Interpretation
Autocorrelation (Wooldridge)	First-order autocorrelation test	F = 1.87 (p = 0.172)	No significant autocorrelation detected (p > 0.05)
Normality (Jarque–Bera)	Test of residuals’ normal distribution	χ2 = 3.21 (p = 0.201)	Residuals are normally distributed (p > 0.05)
Heteroskedasticity (Breusch–Pagan)	Constant variance test	χ2 = 28.34 (p = 0.059)	Marginal indication of heteroskedasticity (close to 5% threshold)

). These findings justify our use of robust standard errors in the main estimations.

The CUSUM test applied to recursive residuals shows no structural breaks during 1995–2022 (statistic = 0.82, p = 0.412) (

Table A2. Structural stability over time.

Test	Period	Results	Interpretation
CUSUM	1995–2022	Statistic = 0.82 (p = 0.412)	No structural break detected
Sub-period estimation	1995–2008	Stable coefficients (variation < ±12%)	Model robust to temporal splits
	2009–2022	Comparable elasticities

and Figure A1, Appendix A). Temporal stability is further confirmed by subsample analysis (1995–2008 vs. 2009–2022) demonstrating comparable coefficients (±12% variation) for key variables including imports (ISNC) and domestic production (P_SNC) (Table A2).

DFBETAS analysis identifies three influential countries (China, Gabon, and Malaysia) with |DFBETAS| > 0.26. Their exclusion leads to limited coefficient changes (<8%), confirming our results are not disproportionately driven by specific cases (

Table A3. Influence of outlier observations.

Test	Results	Remarks
DFBETAS	|DFBETAS| > 0.26	3 influential countries identified
Re-estimation excluding outliers	Δβ < 8%	Minor effect on overall coefficient estimates

).

Advanced endogeneity tests validate our approach: the Hausman test (H = 68.3, p < 0.001) confirms CCE superiority over fixed effects (Table A4).

These comprehensive results, detailed in Appendix A, significantly strengthen our model’s internal validity and our main conclusion’s reliability.

4.3. Analysis of Structural Inequalities

The analysis of structural inequalities in the international tropical hardwood trade, based on descriptive and macro-econometric approaches, highlights deep asymmetries between producing countries—mainly in Central Africa—and importing or re-exporting countries in Asia and the Global North (Deacon, 2020; [135]). Three interrelated dimensions structure these disparities.

• Economic: African producers remain largely specialized in raw log exports, capturing minimal local value (World Bank, 2022 [136]). In contrast, countries like China and France dominate downstream processing and re-export activities, capturing higher profit margins [33,40].
• Environmental: Southern countries bear the ecological costs, particularly Congo Basin nations facing severe deforestation (0.18% annually), despite their forests’ vital climate role (Nepstad et al., 2022 [137]; GFW, 2023 [138]). Certification levels are far lower in Africa (12%) compared to Europe (35%).
• Geopolitical: Market power remains highly concentrated. China controls over 60% of forest concessions in Central Africa [39], pushing prices downward and undermining forest governance, already weakened by illegal logging and institutional opacity (Hornborg, 2009 [41]; Transparency International, 2023 [139]).

Between 1995 and 2022, global trade flows followed a triangular pattern—production (Africa, Latin America), processing (Europe, North America), and final consumption (Asia)—reinforcing structuralist interpretations (ITTO, 2022; [140]). Countries like Gabon and Cameroon remain export-dependent with limited local processing (12–18%) and economic vulnerability (World Bank, 2021 [141]).

Two 2030 scenarios emerge: a business-as-usual path with worsening deforestation, or a transformative shift involving stronger policies, carbon pricing, digital traceability (e.g., blockchain), and North–South cooperation. However, structural constraints persist—limited capital (Gereffi et al., 2005 [5]), weak legal frameworks (Karsenty, 2016 [142]), and dominance by transnational firms (Amin, 1972 [3])—hindering sustainable transformation in the Global South (FAO, 2023 [143]). A multidimensional strategy is required, based on local industrial upgrading, ecosystem service valorization, and technology-driven governance.

4.4. Towards a Fair and Digital Tropical Timber Trade Model: Rebalancing Strategies and Sustainable Industrialization

The development of forestry-based Special Economic Zones (e.g., GSEZ in Gabon), supported by differentiated fiscal incentives and Payments for Environmental Services (PESs), aims to increase the share of locally processed timber beyond 50% and to stimulate industrial upgrading. Blockchain-based traceability systems (e.g., EUTR 2024) are envisioned to guarantee the legal and ecological integrity of timber flows. On the geopolitical front, the establishment of South–South alliances—such as a regional cartel among Gabon, Cameroon, DRC, and Congo—and the inclusion of mirror clauses in trade agreements are seen as essential to restoring normative symmetry. Civil society and local NGOs also play a key role through independent forest monitoring initiatives.

This proposed reform of the tropical timber trade includes three core objectives for 2030–2040: increasing domestic transformation rates, enhancing value-added through labor-intensive exports (e.g., furniture, veneers), and integrating ecosystem services via REDD+ and PESs. The operational structure of these SEZs rests on four pillars: industrialization through forest clusters, financing through sovereign and green funds, standardization via mandatory certification (FSC, PEFC), and capacity-building through training centers in partnership with the EU and China. A differentiated tax regime (0% on finished products, 20% on sawn timber, 30% on logs), combined with trade benefits under APV-FLEGT and the Belt and Road Initiative, underpins this strategic shift. Reinforcing producer countries’ negotiating power also requires a renewed business model within the International Tropical Timber Organization (ITTO), with concrete performance indicators: a local processing rate above 50%, at least 40% certification, and a 150% increase in revenue per cubic meter exported.

The expected impacts of this SEZ-based strategy are multidimensional: economically, a tenfold increase in employment through local transformation; environmentally, more sustainable use of primary forests; geopolitically, enhanced bargaining power for producer countries. In parallel, the transition toward a digitized and intelligent timber industry forms a foundational pillar for modernization. Artificial intelligence (AI) and digital technologies provide concrete solutions to enhance traceability (blockchain, drone and satellite surveillance), monetize ecological services (carbon markets, bioacoustics), automate processing (smart factories, direct sale platforms), and improve trade governance (ITTO 2.0, smart contracts, DAOs). These innovations aim to reduce illegal timber by 40%, double local processing, increase carbon revenues by 30%, and reduce corruption in the sector by 20%.

Implementation relies on synergies between public institutions, multilateral organizations (ITTO, FAO), technology startups (e.g., Satelligence), and green investors (e.g., CAFI). In the short term, launching blockchain pilot projects in Gabon and Cameroon, along with the creation of a green forest innovation fund, could catalyze this transformation. In sum, this strategy outlines a structural overhaul of the tropical timber trade based on economic justice, environmental sustainability, and technological innovation, offering a credible response to global trade asymmetries and paving the way toward greater economic sovereignty for forest-rich countries.

5. Discussion

The results of this study highlight the complexity of the dynamics underlying international trade in tropical hardwood sawnwood, revealing persistent structural imbalances between producer countries in the Global South and importing countries in the Global North. The econometric analysis based on the CS-ARDL-CCE model, complemented by Granger causality tests, confirms the existence of significant structural links between tropical wood exports and several explanatory variables. The positive effect of local production (logP_SNC) on long-term exports (coefficient of 0.38) underscores the central role of productive capacities, consistent with the findings of Barbier and Burgess (2001) [144], while the lack of a significant short-term response reflects the inertia of supply chains, as emphasized by Amacher et al. (2009) [145].

Furthermore, the significant short- and long-term correlation of imports (logISNC: 0.33 and 0.37, respectively) supports the hypothesis of increasing vertical integration, particularly in countries like China, where log imports feed local processing industries before re-export (Sun & Canby, 2021 [146]; ITTO, 2023 [147]). The strong long-term effect of population (logPOP_T, coefficient of 1.29) highlights the importance of scale effects and specialization dynamics driven by demographic density, in line with endogenous growth models (Barro, 1990 [148]; Lucas, 1988 [149]) and the analyses of López et al. (2007) [150]. Conversely, the weak influence of general economic growth (GDP_g) on forestry exports confirms observations by Bohn and Deacon (2000) [151] in resource-rich economies. Granger causality tests further illuminate these mechanisms: the bidirectional causality between exports and imports reflects growing interdependence between industry segments [34,41,55,152], while the unidirectional relationship between population and exports confirms the catalytic role of domestic demand, as argued by Balassa (1985) [153] and Cerutti (2017) [154].

However, these trade dynamics unfold within a context of entrenched structural inequalities and uneven value capture. African countries remain trapped in low-value-added activities, with local processing rates often below 15% (World Bank, 2021 [141]), while industrialized economies appropriate most of the profits through downstream processing (Karsenty, 2016 [142]; Putzel et al., 2014 [8]). This situation exemplifies what Karsenty and Ongolo (2021) [155] refer to as “pathological dependence,” driven by insufficient industrial infrastructure, energy deficits, a shortage of skilled labor (Weng et al., 2014 [156]); Hansen et al., 2018 [157]), global tariff structures that disincentivize finished product exports [51], and the dominance of foreign companies in forest concessions (Oyono et al., 2005 [62]; Colfer & Capistrano, 2005 [158]). These dynamics are consistent with the classic “resource curse” framework (Sachs & Warner, 1999 [27]), further exacerbated by weak governance in the forestry sector (Andersen et al., 2012 [159], Knack, S., & Keefer, P., 1995 [126], Economics and Politics European Commission, 2016, Evaluation of FLEGT Action Plan).

At the same time, environmental concerns raise urgent questions about the sustainability of current trade models. The Congo Basin is experiencing an annual deforestation rate of 0.18% (GFW, 2023 [138]), while traceability mechanisms remain fragmented. Less than 12% of African tropical wood exports are certified under FSC or PEFC standards, compared to over 35% in Europe (Rodrik et al., 2004 [38]; Cerutti et al., 2017 [154]), highlighting the absence of regulatory harmonization through mirror clauses (Putzel, L. et al. 2013; [54]). These gaps are exacerbated by systemic institutional weaknesses, including corruption, limited administrative capacity, and poor enforcement of regulations [82,160,161].

A regional analysis of trade flows between 1995 and 2022 further refines this picture. The Africa–Europe/USA axis remains shaped by postcolonial trade structures, with limited value addition despite progress in FSC certification (35%; Cerutti et al., 2017 [154]). Africa–Asia trade, led by Chinese demand (logISNC > 1 million m³), emphasizes volume over sustainability and deepens Africa’s marginalization in global value chains (Sun & Canby, 2021 [146]). The Latin America–Asia axis, dominated by Brazil (5.3 million m³/year), shows strong growth (+7.2% annually from 2015 to 2022) but is jeopardized by ongoing Amazon deforestation (0.32%/year; Nepstad et al., 2022 [137]). In contrast, the Latin America–Europe/USA route emphasizes certified, processed products, with 45% of volumes certified and significantly higher processing ratios (1:3 compared to 1:8 in Africa), embodying a “qualitative model” (Hurmekoski et al., 2022 [162]).

Against this backdrop, several strategic directions emerge for transforming tropical forestry policy. First, industrialization via Special Economic Zones (SEZs) offers a promising pathway for enhancing local processing, as illustrated by the successful Nkok GSEZ in Gabon (Atanda, 2022 [163]; ITTO, 2023 [164];). Coupling SEZ development with climate finance instruments such as REDD+ or the Green Climate Fund would align industrial upgrading with environmental goals (Angelsen et al., 2012 [165]; Seymour & Harris, 2019 [166]). Second, enhanced geopolitical coordination—particularly through platforms like the International Tropical Timber Organization (ITTO)—could help producer countries to strengthen their bargaining power. A producer cartel inspired by OPEC (Gilbert, 1989 [167]; Garsous, 2019 [168] has been proposed to stabilize prices and reduce destructive competition. The use of blockchain for timber traceability, already explored under the EUTR framework (del Gatto, 2021; [169]), could also improve transparency and combat illegal flows. Third, internalizing ecosystem services through Payment for Environmental Services (PES) schemes, such as the Central African Forest Initiative (CAFI), could offer sustainable incentives for conservation efforts [9,56]. Supporting voluntary certification for small and medium enterprises (SMEs) would enable access to premium markets and promote widespread adoption of sustainability standards (Cashore et al., 2006 [170]; Espach, 2009 [171]).

In conclusion, the findings call for a structural transformation of tropical forestry policies—one that prioritizes industrial upgrading, regional coordination, and greater international recognition of ecosystem services. The pillars of a more equitable, sustainable, and resilient model lie in the advancement of South–South cooperation, green industrialization, and reform of global trade governance.

6. Study Limitations and Research Perspectives

This study acknowledges several limitations that should be considered for a nuanced interpretation of the results. From a methodological standpoint, while the CS-ARDL-CCE model effectively accounts for cross-sectional dependence (CD test = 0.72, p > 0.10), the absence of precise institutional indicators—such as corruption perception indexes or political stability measures—may influence coefficient estimates. This issue resonates with broader methodological challenges in the global value chain literature, particularly regarding the institutional embeddedness of export performance (Gereffi, 2014 [172]; Ponte & Sturgeon, 2014 [112]).

Data limitations primarily concern their level of spatial and institutional granularity. Nationally aggregated forestry indicators fail to capture intra-national dynamics that are critical in analyzing the structure and performance of tropical timber value chains (Andersen et al., 2009 [159]). The inability to distinguish transactions by actor type (state-owned vs. private enterprises) in the ITTO trade database further limits interpretive power, especially in African contexts where ownership structure strongly conditions export strategies (estimated variation Δβ = 12% in sensitivity tests). These constraints must also be interpreted in light of the complex interdependence of variables: the bidirectional causality evidenced in our model (Z = 5.94 *** for ESNC ↔ ISNC) suggests that long-term elasticities (0.380 for P_SNC and 0.365 for ISNC) reflect equilibrium relationships rather than straightforward causal pathways (Granger, 1969 [173]; Sims, 1980 [174]).

These constraints highlight several promising avenues for future research. Enhancing the dataset with institutional and certification indicators is a clear priority. From a methodological angle, combining econometric modeling with satellite-based spatial data (e.g., forest cover and logging patterns) could address some current data aggregation issues. Lastly, disaggregating trade flows by actor type and processing level would enable more refined assessments of value-added generation and distribution across the tropical timber sector (Schure et al., 2015 [175]).

While these limitations do not undermine our key findings, they do help to define the appropriate scope of our conclusions, particularly in regions where institutional and regulatory factors are critical—such as the Congo Basin. This methodological transparency ultimately strengthens the validity of our results while offering concrete directions for future work.

7. Policy Implications and Recommendations

The empirical findings of this study highlight several strategic policy levers that can help to reorient international trade in tropical hardwood lumber toward greater equity, sustainability, and value addition for producing countries. The implications span key areas such as industrial development, trade regulation, forest governance, regional integration, and ecosystem service valorization.

First, the positive and significant relationship between domestic production and export performance supports the case for accelerating local value addition. Strengthening domestic wood processing industries can reduce overreliance on raw log exports and enhance economic diversification. To this end, policy measures should prioritize the implementation of incentive-driven industrial strategies, including targeted tax exemptions, improved energy access, and dedicated investment support for processing activities such as drying, planing, and assembly. Moreover, the development of integrated and environmentally conscious Special Economic Zones (SEZs)—as exemplified by the Gabon Special Economic Zone (GSEZ)—should be promoted, ensuring alignment with national climate goals and social development priorities (Toman & Jemelkova, 2003 [176]; UNCTAD, 2021 [177]).

Second, reinforcing forest governance and improving traceability mechanisms are critical for combating illegal logging and ensuring sustainable trade. The persistent institutional weaknesses and loopholes in certification systems undermine regulatory effectiveness. Strengthening national forestry administrations, digitizing logging permits, and enhancing human resource capacity through training programs are essential steps. Additionally, the adoption of digital traceability technologies—particularly blockchain—can improve transparency across supply chains, drawing lessons from pilot initiatives in countries such as Ghana and Brazil (FAO, 2020 [178]; World Bank, 2022 [136]). Support for the wider adoption of sustainability certifications (e.g., FSC, PEFC, TLAS) through subsidies or technical assistance for small and medium enterprises (SMEs) is also recommended.

Third, international trade governance must be reformed to eliminate structural disadvantages faced by tropical timber-exporting countries. The continued dominance of raw material exports is partly explained by tariff escalation and the absence of equitable environmental clauses in trade agreements. In response, reforms to the World Trade Organization (WTO) framework should be pursued, including the removal of tariff barriers on processed wood products and the introduction of environmental mirror clauses for tropical timber imports (Howse & Eliason, 2005 [179]). Furthermore, producer countries should be encouraged to establish a coordination mechanism akin to a forestry consortium or cartel to harmonize export policies, stabilize prices, and increase their bargaining power vis-à-vis major importing blocs (e.g., the EU, China).

Fourth, enhancing regional integration and fostering South–South cooperation can help to overcome the limitations of fragmented national markets. Promoting intra-regional trade in processed timber through preferential trade agreements within regional blocs such as ECOWAS, ECCAS, or ASEAN would be beneficial. Additionally, establishing regional processing hubs with shared infrastructure—such as transportation networks, energy systems, and port facilities—could enhance economies of scale and competitiveness, especially in high-potential trade corridors (UNECA, 2020 [180]).

Finally, it is essential to integrate the valuation of ecosystem services provided by tropical forests—such as carbon sequestration and biodiversity conservation—into trade policy frameworks. Strengthening Payment for Environmental Services (PES) schemes, particularly those involving artisanal producers and local communities, can be achieved through mechanisms like REDD+, CAFI, or the Green Climate Fund (GCF). Developing eco-labeling systems tailored to tropical timber products, which combine environmental performance, social inclusion, and traceability, is another promising avenue. Importantly, international buyers should be required to commit to verifiable environmental standards as part of bilateral trade arrangements, such as Voluntary Partnership Agreements (VPAs) under the EU FLEGT Action Plan (European Commission, 2018 [181]).

Overall, these policy recommendations call for a multi-level, cross-sectoral approach that integrates industrial policy, trade regulation, environmental governance, and international cooperation. Only a coordinated and politically committed strategy—backed by strong international support—can reposition the tropical hardwood lumber trade as a vector for sustainable development in the Global South.

8. Conclusions

This article contributes to a deeper understanding of the persistent asymmetries in the international trade of tropical hardwood sawnwood by deploying an innovative econometric and typological framework. Through the application of the CCE-ARDL model on an unbalanced panel of 58 ITTO member countries (1995–2022), it identifies key structural and macroeconomic determinants of export performance, while accounting for cross-sectional dependencies and dynamic causal mechanisms by integrating heterogeneous national trajectories into a typology of four trade profiles: (i) raw producers with minimal local processing; (ii) marginal players with weak trade integration; (iii) high-value-added re-export platforms; (iv) major consumer markets. The results underscore several core findings detailed below.

• Short-term effects (Uneven Trade Dynamics):

  -

  Imports boost exports (+0.33%), reflecting that Asia, the USA, and the EU are dominant in re-export after processing in the model. And local production has weak/no significant impact in Africa due to limited industrial capacity.

• Long-term effects (Uneven Trade Dynamics):

  -

  Domestic production and imports have a positive effect (+0.38% and +0.37%), but only Asian countries fully capitalize on this due to their strong industrial base.

  -

  Population size plays a key role (+1.29%), benefiting major markets (China, India) that combine domestic demand and export power. This coincides with the elasticity of economic growth (+1.1, short-term) which could be explained by an improvement in logistics infrastructure or an increase in demand from importing countries.

• African countries exhibit low processing intensity (<20%) and high vulnerability to price fluctuations, whereas Asian re-exporters benefit from strong industrial capacity and market diversification.
• Certification schemes such as FSC/PEFC generate measurable price premiums, but their uptake remains uneven, highlighting the dual challenge of market access and sustainability compliance.
• Structural factors—including demographic pressure, infrastructure quality, and colonial legacies—interact in complex ways to shape export trajectories, challenging simplistic or linear policy prescriptions.

Beyond empirical insights, the study advances the theoretical debate by introducing the concepts of “pathological dependence” and the Fair and Digital Timber Trade Model (F&DTTT), providing a more nuanced understanding of lock-ins and leverage points within global value chains. This framework integrates digital traceability, mirror clauses, and South–South industrial alliances as key instruments for rebalancing trade relationships and promoting structural transformation. From a policy perspective, the article offers concrete tools to align the tropical timber trade with the Sustainable Development Goals. It speaks directly to the following:

• SDG 8, by identifying strategies for decent work and local industrial development through improved processing and small enterprise support;
• SDG 12, by analyzing the effect of certification and production efficiency on responsible consumption;
• SDG 15, by linking trade structures to deforestation risks and proposing payment schemes for ecosystem services;
• SDG 17, by calling for stronger regional cooperation among producer countries to enhance bargaining power and promote knowledge sharing.

While the research provides a robust and multidimensional analysis, it also acknowledges key limitations—particularly the difficulty in accessing disaggregated data on informal and illegal trade flows and the challenges of capturing qualitative institutional dynamics in quantitative models. These limitations open avenues for further research, particularly in integrating spatial data, firm-level behavior, and political economy analysis of trade agreements.

In sum, this study demonstrates that transforming the tropical timber trade into a more equitable, sustainable, and digitally transparent system is not only a political imperative but also a methodological necessity. By bridging econometrics, political ecology, and global governance, it offers a solid foundation for future policy design and academic inquiry.