Unveiling Innovation Interdependence in Emerging Economies: A Bivariate Copula Analysis of Open and Closed Strategies in ECA and MENA
Examines the joint determinants of product/service and process innovation among 9,710 private firms across 41 ECA and MENA economies using a bivariate copula model. The Joe copula (AIC = 3,136,340) reveals strong upper-tail dependence (τ = 0.914), with closed R&D strategies (β = 4.278) and international market orientation (β = 3.717) driving output innovation, while OIR+OIK (β = 2.697) leads process innovation. Findings advance open innovation theory and inform SDG 9-aligned policy for ECA and MENA.
Abstract
This study investigates the joint determinants of product/service and process innovation among 9,710 private firms across 41 economies in Europe and Central Asia (ECA) and the Middle East and North Africa (MENA), using a bivariate copula model to capture the interdependence of innovation outcomes. Leveraging World Bank Enterprise Survey data (2019–2020), we examine the roles of open knowledge innovation (OIK), open R&D innovation (OIR), and closed R&D innovation (CIR), alongside firm characteristics like size, costs, and market orientation. The Joe copula model (AIC = 3,136,340, BIC = 3,136,476) reveals strong upper-tail dependence (\tau = 0.914), with CIR-driven strategies (\beta = 4.278) and international market orientation (\beta = 3.717) significantly boosting product/service innovation, while OIR+OIK (\beta = 2.697) drives process innovation. Unexpectedly, perceived skill shortages (\beta = 1.031) spur innovation, whereas larger firm size (\beta = -0.003) and technological inputs (\beta = -0.689) constrain it, particularly in MENA. These findings advance resource-based view and open innovation theories, offering actionable insights for firms to balance R&D and external collaborations, and informing SDG 9-aligned policies to foster sustainable industrialization through targeted R&D incentives, workforce development, and export promotion in ECA and MENA.
Keywords: Open Innovation, Closed Innovation, Bivariate Copula, Product Innovation, Process Innovation, Sustainable Development
JEL Classification: O31, O32, O33, L25, C35, Q55
Introduction
In a rapidly evolving global economy, innovation fuels firm competitiveness and sustainable growth, acting as the cornerstone of economic transformation in emerging regions (Schumpeter, 1934). Europe & Central Asia (ECA) and the Middle East & North Africa (MENA) stand at a critical juncture, where harnessing open and closed innovation strategies can unlock unprecedented opportunities for firms navigating dynamic markets. In ECA, transition economies have embraced market-driven systems, boosting innovation in manufacturing and services (Orlic et al., 2018). In MENA, small and medium enterprises (SMEs) grapple with resource constraints, yet hold immense potential for innovation through external collaborations (Haddoud et al., 2023). The pivotal role of organizational ambidexterity in ECA, as shown by (Leitão et al., 2024), reveals how balancing internal R&D (Closed R&D Innovation, CIR) with external partnerships (Open Knowledge Innovation, OIK; Open R&D Innovation, OIR) drives product and eco-innovation in countries like Estonia and the Czech Republic, setting the stage for this study’s exploration of innovation interdependence.
Open innovation, a paradigm shift pioneered by (Chesbrough, 2003; Chesbrough, 2006), underscores the power of external knowledge flows, particularly for MENA’s SMEs, where collaborative networks amplify service innovation (Hippel, 2005). In ECA, firms leverage robust industrial capabilities to excel in process innovation, while MENA firms prioritize product innovation through open strategies (Haddoud et al., 2023; Orlic et al., 2018). This is evident in MENA’s hospitality sector, where (Hameed et al., 2021) demonstrate that external knowledge enhances service innovation, reinforcing OIK’s role in resource-constrained settings. Yet, regional disparities persist, with ECA’s structured economies contrasting MENA’s fragmented markets, necessitating tailored innovation frameworks (Gassmann et al., 2010).
Despite the growing emphasis on open innovation, research on innovation orientations in ECA and MENA remains fragmented, with most studies focusing on developed economies (Crepon et al., 1998; Laursen & Salter, 2006). The interplay of OIK, OIR, and CIR, and their differential impacts on product and process innovation, is underexplored, particularly in MENA’s SME-driven context (Haddoud et al., 2023). As (Edwards-Schachter, 2018) note, the complexity of innovation typologies demands comprehensive frameworks like the Innovation Orientation Index (IOI), which integrates OIK, OIR, and CIR to bridge these gaps and capture the nuanced dynamics of emerging economies.
This study investigates how OIK, OIR, and CIR shape product and process innovation across 9,710 ECA and MENA firms, employing a bivariate copula regression model to capture their interdependent effects. Using World Bank Enterprise Survey data (2019–2020), it tests hypotheses on the distinct and joint impacts of innovation orientations, uncovering regional differences to inform context-specific strategies (World Bank Group, 2020).
By quantifying innovation orientation effects, this study advances Schumpeterian and Resource-Based View (RBV) theories, offering novel insights into how firms balance internal and external resources (Barney, 1991; Schumpeter, 1934). It provides actionable strategies for ECA and MENA firms to optimize innovation practices and informs policy aligned with SDG 9 (Industry, Innovation, and Infrastructure). This aligns with (Ritala, 2024), who advocate for platform ecosystems to address societal challenges, reinforcing the study’s relevance to sustainable development. The semi-parametric copula model introduces a methodological breakthrough, capturing correlated innovation outcomes to address critical research gaps.
The rest of the paper is therefore organized as follows: Section 2 reviews literature on innovation orientations and their impacts. Section 3 details the methodology, including theoretical framework, empirical model, and data. Section 4 presents empirical results, Section 5 discusses implications for theory, practice, and policy, and Section 6 concludes with limitations and future research directions.
Literature Review
This review synthesizes the relevant literature to construct a robust foundation for understanding innovation orientations in the ECA and MENA regions. Drawing on seminal works in innovation theory and empirical studies of emerging economies (Chesbrough, 2003; Crepon et al., 1998; Haddoud et al., 2023), it is axed around three critical themes: (i) Innovation Orientations and Knowledge Creation, (ii) Output Innovation Drivers, and (iii) Process Innovation Drivers. These themes elucidate how OIK, OIR, and CIR shape innovation outcomes in ECA’s industrial economies and MENA’s SME-driven markets.
Innovation Orientations and Knowledge Creation
This theme examines how OIK, OIR, and CIR contribute to firms’ knowledge stock, a cornerstone of innovation in emerging economies (Chesbrough, 2003; Griliches, 1979). In this regard, (Griliches, 1979) analyzed 1,000 U.S. firms, linking internal R&D (CIR) to productivity gains, a model relevant to ECA’s manufacturing-driven economies (Orlic et al., 2018). In contrast, (Chesbrough, 2003; Chesbrough, 2006) pioneered the open innovation paradigm, demonstrating that OIK and OIR, through external collaborations, enhance knowledge creation, particularly for MENA’s resource-constrained SMEs (Haddoud et al., 2023). For instance, (Hippel, 2005) emphasize user-driven OIK, showing how customer engagement amplifies innovation, a strategy critical for MENA’s service sectors. (Laursen & Salter, 2006) surveyed 2,707 U.K. firms, finding that OIK with diverse partners boosts innovation but risks diminishing returns due to coordination costs, a caution for ECA’s complex innovation ecosystems (Livieratos et al., 2022). (Livieratos et al., 2022) studied 106 European SMEs, highlighting OIK and OIR’s knowledge creation benefits, tempered by “attention capital” constraints in MENA (Mehtap et al., 2019). Similarly, (Odei & Appiah, 2023) used World Bank Enterprise Survey data to show that CIR drives technological innovation in the Czech Republic, reinforcing its role in ECA. (Cuevas-Vargas et al., 2022) applied SEM to 145 Colombian SMEs, revealing that OIK via ICT strengthens absorptive capacity, a key enabler for ECA firms navigating digital transformation (Orlic et al., 2018). These findings underscore the need for balanced innovation strategies to maximize knowledge creation in ECA and MENA.
Output Innovation Drivers
This theme investigates how CIR’s proprietary focus and OIK’s collaborative approach drive product and service innovation, mediated by absorptive capacity (Cohen & Levinthal, 1990). In this regard, (Crepon et al., 1998) applied the CDM model to 5,000 French firms, linking CIR to new product development, a dynamic evident in ECA’s industrial firms (Orlic et al., 2018). (Cohen & Levinthal, 1990) introduced absorptive capacity, critical for MENA SMEs leveraging OIK to overcome resource limitations (Haddoud et al., 2023). (Tether, 2002) analyzed 1,200 U.K. firms, showing that OIK through partnerships enhances product innovation, a strategy applicable to ECA’s collaborative networks (Orlic et al., 2018). In MENA, (Hameed et al., 2021) found that external knowledge drives service innovation in Pakistan’s hospitality sector, highlighting OIK’s role in SME-dominated markets. (Edwards-Schachter, 2018) emphasize the diverse typologies of innovation, underscoring the IOI’s role in capturing CIR and OIK’s contributions to output innovation. (Fisher et al., 2024) studied 200 U.S. firms, showing that OIK extends product lifespans, a benefit relevant to ECA’s competitive markets. These insights highlight the interplay of internal and external resources in driving output innovation, particularly in the MENA region.
Process Innovation Drivers
This theme explores OIR’s role in enhancing process innovation, critical for operational efficiency in ECA and MENA. (Reichstein & Salter, 2006) analyzed 1,000 U.K. firms, demonstrating that OIR with suppliers improves process efficiency, a strategy vital for ECA’s manufacturing sector (Orlic et al., 2018). (Damanpour, 1991) conducted a meta-analysis of 23 studies, identifying managerial and technological drivers of process innovation, more prevalent in ECA’s large firms than MENA’s SMEs (Mehtap et al., 2019). (Evangelista & Vezzani, 2010) studied 10,000 European firms, linking OIR to efficiency gains in services, relevant for ECA’s service-oriented economies. (Patrucco et al., 2022) analyzed nine collaborative projects, finding that OIR enhances process innovation, though MENA firms face partnership barriers (Haddoud et al., 2023). (Abdullah et al., 2016) identified collaboration barriers in 153 Malaysian firms, mirroring MENA’s challenges in adopting OIR due to resource constraints (Mehtap et al., 2019). (Gassmann & Enkel, 2004) outline OIR’s process archetypes, emphasizing supplier-driven innovation, which holds potential for ECA and MENA firms seeking efficiency gains. These findings highlight OIR’s transformative potential, tempered by regional constraints.
Emerging Hypotheses and Announced Methodology
The synthesized literature suggests that OIK enhances both product and process innovation, OIR drives process efficiency, and CIR excels in product innovation, with effects varying by region. In ECA, CIR and OIR leverage industrial capabilities, while in MENA, OIK mitigates resource constraints (Chesbrough, 2006; Haddoud et al., 2023; Orlic et al., 2018). A semi-parametric bivariate copula regression model, using World Bank Enterprise Survey data (2019–2020), will test these hypotheses, capturing the interdependent effects of innovation outcomes to address gaps in ECA and MENA research (World Bank Group, 2020).
Methodology
Theoretical Framework
Drawing on Schumpeterian innovation theory (Schumpeter, 1934) and the Resource-Based View (Barney, 1991), we posit that firm innovation performance—output (new products/services) and process (new production/delivery methods)—depends on innovation orientations: Open Knowledge Innovation (OIK), Open R&D Innovation (OIR), and Closed R&D Innovation (CIR). The Technology-Organization-Environment framework (Tornatzky & Fleischer, 1990) contextualizes regional influences. We define:
\begin{align*} Y_{io} \in \{0,1\} &: \text{binary indicator for output innovation}, \\ Y_{ip} \in \{0,1\} &: \text{binary indicator for process innovation}, \\ Z_i = [OIK_i, OIR_i, CIR_i, X_i] &: \text{vector of firm-level determinants}. \end{align*}
Latent innovation propensities are formulated as:
\begin{align} Y_{io}^* &= \alpha_0 + \alpha_1 OIK_i + \alpha_2 OIR_i + \alpha_3 CIR_i + \alpha_4 X_i + \varepsilon_{io}, \\ Y_{ip}^* &= \beta_0 + \beta_1 OIK_i + \beta_2 OIR_i + \beta_3 CIR_i + \beta_4 X_i + \varepsilon_{ip}, \end{align}
Substituting the binary indicators with the comprehensive Innovation Orientation Index (IOI) as developed elsewhere with eight qualitative factor levels, the bivariate latent system becomes:
\begin{align} Y_{io}^* &= \alpha_0 + \sum_{k=1}^7 \alpha_k \text{IOI}_k + \alpha_8 X_i + \varepsilon_{io}, \\ Y_{ip}^* &= \beta_0 + \sum_{k=1}^7 \beta_k \text{IOI}_k + \beta_8 X_i + \varepsilon_{ip}, \end{align}
where \text{IOI}_k are dummy variables for each level (0_0_0 as reference), and X_i includes control variables, while the errors (\varepsilon_{io}, \varepsilon_{ip}) follow a joint distribution allowing correlation. In this representation, the observed outcomes are:
\begin{align*} Y_{io} &= \begin{cases} 1 & \text{if } Y_{io}^* > 0, \\ 0 & \text{otherwise}, \end{cases} \\ Y_{ip} &= \begin{cases} 1 & \text{if } Y_{ip}^* > 0, \\ 0 & \text{otherwise}. \end{cases} \end{align*}
Empirical Model
We estimate the system using a semi-parametric bivariate copula regression model via the gjrm() function in R’s GJRM package (Wojtyś et al., 2018). This model accommodates flexible marginal distributions, nonlinear covariate effects, and copula-based dependence structures:
\begin{align} g_o(\mathbb{E}[Y_{io}]) &= \eta_{io} = f_o(Z_i), \\ g_p(\mathbb{E}[Y_{ip}]) &= \eta_{ip} = f_p(Z_i), \end{align}
where g_o(\cdot), g_p(\cdot) are logit link functions, and f_o(\cdot), f_p(\cdot) are additive covariate functions. The joint distribution is:
\begin{equation} \text{Pr}(Y_{io} = y_o, Y_{ip} = y_p) = C_\theta(F_o(y_o|Z_i), F_p(y_p|Z_i)), \end{equation}
where F_o, F_p are marginal CDFs of product/service innovation (ProdServInnov) and process innovation (ProcessInnov). C_\theta is a copula function with parameter \theta, that captures the dependence between the two outcomes, based on three alternative tested specifications: Gaussian, Gumbel, and Joe. Model fit is assessed using AIC and Vuong’s test. Adopting a Probit model representation, the empirical implementation for j = 1, 2, is given as:
P(\texttt{ProdServInnov} = 1, \texttt{ProcessInnov} = 1 | X) = C(F_1(\mu_1(X)), F_2(\mu_2(X)); \theta),
with:
\begin{equation} \begin{split} \mu_j &= \beta_{j0} + \beta_{j1}\texttt{InovStratOrient} + \beta_{j2}\texttt{nFulTimEmplyLFY} + \beta_{j3}\texttt{laborCost} \\ &\quad + \beta_{j4}\texttt{LaborReg} + \beta_{j5}\texttt{InadqEducWorkforce} + \beta_{j6}\texttt{ExptdFutSales} \\ &\quad + \beta_{j7}\texttt{ElectricityCost} + \beta_{j8}\texttt{FuelCost} + \beta_{j9}\texttt{extAudit} \\ &\quad + \beta_{j10}\texttt{taxAuditLFY} + \beta_{j11}\texttt{legalStat} + \beta_{j12}\texttt{MainProdServLFY} \\ &\quad + \beta_{j13}\texttt{outMktOrient} + \beta_{j14}\texttt{techInpMktOrient} + \beta_{j15}\texttt{MangYrExpSect} \\ &\quad + \beta_{j16}\texttt{topManagfem} + \beta_{j17}\texttt{femOwner} + \beta_{j18}\texttt{PercSenManTimGovReg} \\ &\quad + \beta_{j19}\texttt{AccsToFinObstOP} + \beta_{j20}\texttt{BigestObstOP} \\ &\quad + \beta_{j21}\texttt{ProsPubSpendPriorty} + s(\texttt{CNTnameID}), \end{split} \end{equation}
where F_1 and F_2 are probit cumulative distribution functions, and \mu_j for j = 1, 2 are linear predictors for each outcome. A smooth term for country effects (CNTnameID), and survey weights (wstrict) for standard error correction, are also included in the estimation of the means functions \mu_j.
Data Source and Variables
This study utilizes data from the World Bank’s Enterprise Surveys (WBES) conducted in 2019–2020, released on March 29, 2024 (World Bank Group, 2020). The harmonized dataset employs standardized WBES methodology to capture firms’ business environment, including innovation activities, financial access, and operational characteristics. The final sample comprises 9,710 private firms across 41 economies, representing a 34.62% retention rate from the initial 28,042 companies. Table 1 details the sample distribution: 4,378 firms (45.09%) from 22 European economies, 2,475 firms (25.49%) from 12 Central Asian economies, and 2,857 firms (29.42%) from 7 MENA economies. This section describes the quantitative and qualitative variables pivotal to understanding innovation dynamics, with descriptive statistics presented in Table 2, Table 3, and Table 4.
| Region | Countries | Freq. | % |
|---|---|---|---|
| Europe | Albania, Bosnia and Herzegovina, Bulgaria, Croatia, Cyprus, Czechia, Estonia, Greece, Hungary, Italy, Kosovo, Latvia, Lithuania, Malta, Montenegro, North Macedonia, Poland, Portugal, Romania, Serbia, Slovak Republic, Slovenia | 4,378 | 45.09 |
| Central Asia | Armenia, Azerbaijan, Belarus, Georgia, Kazakhstan, Kyrgyz Republic, Moldova, Mongolia, Russia, Tajikistan, Ukraine, Uzbekistan | 2,475 | 25.49 |
| MENA | Egypt, Jordan, Lebanon, Morocco, Tunisia, Turkey, West Bank and Gaza | 2,857 | 29.42 |
| Total | 41 economies | 9,710 | 100.00 |
Quantitative Variables
The quantitative variables encompass firm characteristics and operational metrics relevant to innovation, including size, costs, labor constraints, and market expectations. Table 2 provides means, medians, and standard deviations for the 9,710 firms.
| Variable | Units | Mean | Median | SD |
|---|---|---|---|---|
FT Employees (nFulTimEmplyLFY) |
People | 83.3 | 20.0 | 681.5 |
Labor Cost (laborCost) |
Local (M) | 100.2 | 1.2 | 1110.0 |
Labor Regulation (LaborReg) |
0-4 Scale | 1.04 | 1.00 | 1.12 |
Unskilled Workforce (InadqEducWorkforce) |
0-4 Scale | 1.46 | 1.00 | 1.30 |
Expected Sales (ExptdFutSales) |
1-3 Scale | 1.76 | 1.00 | 0.89 |
Electricity (ElectricityCost) |
Local (M) | 12.79 | 0.08 | 166.04 |
Fuel (FuelCost) |
Local (M) | 18.48 | 0.05 | 304.16 |
Manager Experience (MangYrExpSect) |
Years | 22.21 | 20.00 | 11.41 |
The number of full-time employees varies significantly (mean = 83.3, median = 20, SD = 681.5), reflecting a mix of SMEs and larger firms, particularly in ECA’s manufacturing sector (Orlic et al., 2018). Labor costs exhibit substantial dispersion (mean = 100.2M, median = 1.2M, SD = 1110.0M), highlighting economic heterogeneity across ECA and MENA (Haddoud et al., 2023). Labor regulation obstacles (LaborReg, mean = 1.04) and inadequately educated workforce (InadqEducWorkforce, mean = 1.46) indicate moderate constraints, with MENA facing greater skill shortages (Aljanabi, 2018). Expected future sales (ExptdFutSales, mean = 1.76, median = 1) suggest cautious optimism. High variability in electricity (SD = 166.04M) and fuel costs (SD = 304.16M) reflects diverse energy demands, especially in ECA’s industrial base (Evangelista & Vezzani, 2010). Managers’ extensive sector experience (mean = 22.21 years, median = 20) underscores robust human capital across regions.
Qualitative Variables
The qualitative variables capture categorical firm characteristics, innovation orientations, outcomes, market orientation, gender diversity, and operational obstacles. These are detailed in Table 3 and Table 4, with absolute and relative frequencies.
Innovation Orientation Index (IOI): The IOI, based on combinations of Open Knowledge Innovation (OIK), Open R&D Innovation (OIR), and Closed R&D Innovation (CIR), reveals that 76.95% of firms (7,472) report no innovation activity (0_0_0), reflecting resource constraints, particularly in MENA (Haddoud et al., 2023). CIR-inclusive levels (e.g., 1_0_0, 8.11%; 1_1_0, 3.30%; 1_1_1, 3.07%) dominate in ECA due to stronger R&D infrastructure (Orlic et al., 2018), while OIK-inclusive levels (e.g., 0_0_1, 3.87%) are more prevalent in MENA’s SMEs (Aljanabi, 2018).
Innovation Outcomes: Product/service innovation (ProdServInnov, 23.61%, 2,293 firms) and process innovation (ProcessInnov, 13.95%, 1,355 firms) indicate limited innovation activity, with process innovation facing higher capital barriers, especially in MENA (Abdullah et al., 2016).
Audit Practices: External audits (extAudit, 47.23%, 4,586 firms) and tax audits (taxAuditLFY, 45.96%, 4,462 firms) are common, with higher prevalence in ECA due to stricter regulations (Evangelista & Vezzani, 2010).
Legal Status and Sector: Private shareholding companies (49.97%, 4,852 firms) and sole proprietorships (22.17%, 2,153 firms) dominate, with publicly traded firms less common (6.78%, 658 firms). Manufacturing leads (55.77%, 5,415 firms), followed by retail trade (14.90%, 1,447 firms) and services (9.19%, 892 firms), reflecting ECA’s industrial strength and MENA’s SME-driven economy (Aljanabi, 2018; Orlic et al., 2018).
Market Orientation: Output market orientation (outMktOrient) shows 49.97% of firms (4,852) targeting national markets, 33.99% (3,301) local markets, and 16.04% (1,557) international markets. Technology input market orientation (techInpMktOrient, 13.60%, 1,320 firms) indicates limited technological engagement, particularly in MENA (Haddoud et al., 2023).
Gender Variables: Female top managers (topManagfem, 14.99%, 1,456 firms) and female ownership (femOwner, 28.26%, 2,744 firms) suggest moderate gender diversity. Higher female ownership in MENA reflects SME-driven markets, where women play a significant role in entrepreneurship (Aljanabi, 2018; Mehtap et al., 2019).
Biggest Obstacle: The primary obstacles to operations (BigestObstOP) are tax rates (21.95%, 2,131 firms), inadequately educated workforce (13.56%, 1,316 firms), and political instability (12.07%, 1,172 firms), followed by access to finance (10.27%, 997 firms). Less prevalent obstacles include access to land and courts (1.08% each, 105 firms). These findings highlight fiscal, human capital, and institutional barriers, particularly in MENA, where skill shortages and political instability hinder innovation (Aljanabi, 2018; Haddoud et al., 2023).
| Variable | Level | Freq. | % |
|---|---|---|---|
Innovation Orientation InovStratOrient |
No innovation (0_0_0) | 7,472 | 76.95 |
| OIK only (0_0_1) | 376 | 3.87 | |
| OIR only (0_1_0) | 112 | 1.15 | |
| OIR+OIK (0_1_1) | 53 | 0.55 | |
| CIR only (1_0_0) | 787 | 8.11 | |
| CIR+OIK (1_0_1) | 291 | 3.00 | |
| CIR+OIR (1_1_0) | 321 | 3.30 | |
| Full (1_1_1) | 298 | 3.07 | |
Innovation Outcomes ProdServInnov |
No (0) | 7,417 | 76.39 |
| Yes (1) | 2,293 | 23.61 | |
ProcessInnov |
No (0) | 8,355 | 86.05 |
| Yes (1) | 1,355 | 13.95 | |
Audit Practices extAudit |
No (0) | 5,124 | 52.77 |
| Yes (1) | 4,586 | 47.23 | |
taxAuditLFY |
No (0) | 5,248 | 54.04 |
| Yes (1) | 4,462 | 45.96 | |
Legal Status legalStat |
Public shareholding (1) | 658 | 6.78 |
| Private shareholding (2) | 4,852 | 49.97 | |
| Sole proprietorship (3) | 2,153 | 22.17 | |
| Partnership (4) | 857 | 8.83 | |
| Limited partnership (5) | 1,074 | 11.06 | |
| Other (6) | 116 | 1.19 |
| Variable | Level | Freq. | % |
|---|---|---|---|
Sector MainProdServLFY |
Manufacturing (1) | 5,415 | 55.77 |
| Retail trade (2) | 1,447 | 14.90 | |
| Wholesale trade (3) | 1,447 | 14.90 | |
| Construction (4) | 793 | 8.17 | |
| Hotel/Restaurant (5) | 443 | 4.56 | |
| Services (6) | 892 | 9.19 | |
Market Orientation outMktOrient |
Local (1) | 3,301 | 33.99 |
| National (2) | 4,852 | 49.97 | |
| International (3) | 1,557 | 16.04 | |
techInpMktOrient |
No (0) | 8,390 | 86.40 |
| Yes (1) | 1,320 | 13.60 | |
Gender Variables topManagfem |
No (0) | 8,254 | 85.01 |
| Yes (1) | 1,456 | 14.99 | |
femOwner |
No (0) | 6,966 | 71.74 |
| Yes (1) | 2,744 | 28.26 | |
Biggest Obstacle BigestObstOP |
Access to finance (1) | 997 | 10.27 |
| Access to land (2) | 105 | 1.08 | |
| Business licensing and permits (3) | 330 | 3.40 | |
| Corruption (4) | 629 | 6.48 | |
| Courts (5) | 105 | 1.08 | |
| Crime, theft and disorder (6) | 118 | 1.22 | |
| Customs and trade regulations (7) | 210 | 2.16 | |
| Electricity (8) | 406 | 4.18 | |
| Inadequately educated workforce (9) | 1,316 | 13.56 | |
| Labor regulations (10) | 350 | 3.60 | |
| Political instability (11) | 1,172 | 12.07 | |
| Practices of competitors in the informal sector (12) | 860 | 8.86 | |
| Tax administration (13) | 516 | 5.31 | |
| Tax rates (14) | 2,131 | 21.95 | |
| Transport (15) | 465 | 4.79 |
Expected Effects
The regression model in equation (8) includes variables expected to influence product/service (ProdServInnov) and process (ProcessInnov) innovation outcomes. Innovation orientation (InovStratOrient) is anticipated to positively affect both outcomes, with stronger effects for Closed R&D Innovation (CIR)-inclusive levels due to higher R&D intensity (Orlic et al., 2018). Firm size (nFulTimEmplyLFY) and managerial experience (MangYrExpSect) are expected to enhance innovation through greater resource availability and sector-specific expertise (Barney, 1991). Costs, including labor (laborCost), electricity (ElectricityCost), and fuel (FuelCost), may have mixed effects, potentially constraining innovation due to resource allocation trade-offs, particularly in MENA (Haddoud et al., 2023).
Labor regulations (LaborReg) and an inadequately educated workforce (InadqEducWorkforce) are likely to negatively impact innovation by increasing operational constraints, especially in MENA (Aljanabi, 2018). Conversely, positive sales expectations (ExptdFutSales) should encourage innovation investment (Aljanabi, 2018). Audit practices (extAudit, taxAuditLFY) may positively influence innovation by signaling robust governance (Evangelista & Vezzani, 2010). Legal status (legalStat) and sector (MainProdServLFY) are expected to shape innovation, with manufacturing and public shareholding firms potentially showing stronger innovation due to scale and structure (Orlic et al., 2018). International market orientation (outMktOrient) and technological inputs (techInpMktOrient) should foster innovation through knowledge spillovers (Chesbrough, 2003).
Gender diversity, including female top managers (topManagfem) and female ownership (femOwner), may positively affect innovation, particularly in MENA’s SME-driven markets, where women contribute significantly to entrepreneurship (Aljanabi, 2018; Mehtap et al., 2019). The percentage of senior management time spent on government regulations (PercSenManTimGovReg) is expected to negatively impact innovation by diverting resources (Orlic et al., 2018). Access to finance obstacles (AccsToFinObstOP) and the biggest operational obstacles (BigestObstOP), such as tax rates and political instability, are likely to hinder innovation, particularly in resource-constrained MENA (Aljanabi, 2018; Haddoud et al., 2023). Public spending priority perceptions (ProsPubSpendPriorty) may have mixed effects, potentially supporting innovation through infrastructure investment or constraining it due to misaligned priorities (Haddoud et al., 2023). The smooth term for country-specific effects (s(CNTnameID)) is expected to capture unobserved heterogeneity across economies, influencing innovation outcomes (Odei & Appiah, 2023).
Econometric Results
Model Fit and Convergence
Three bivariate copula models were estimated: Gaussian (\theta = 0.988, \tau = 0.901), Gumbel (\theta = 9.12, \tau = 0.89), and Joe (\theta = 22.1, \tau = 0.914). All models converged successfully, with trust region iterations ranging from 41–69 and smoothing loops from 3–8. The Joe model exhibits the best fit (AIC = 3,136,340, BIC = 3,136,476), followed by Gaussian (AIC = 3,515,160, BIC = 3,515,268) and Gumbel (AIC = 8,351,085, BIC = 8,351,179), suggesting superior modeling of upper-tail dependence in innovation outcomes.
Estimated Effects
Table 5 presents key coefficients from the Joe copula model for variables of interest, focusing on statistically significant effects (p < 0.05) unless otherwise noted.
Innovation Orientation (IOI): Most IOI levels significantly affect both innovation outcomes, with InovStratOrient1_0_0 (CIR only, \beta = 4.278) showing the strongest positive effect on product/service innovation, and InovStratOrient0_1_1 (OIR+OIK, \beta = 2.697) for process innovation. However, InovStratOrient0_0_1 (\beta = -0.985) and InovStratOrient1_1_0 (\beta = -3.272) negatively affect product/service innovation, suggesting resource allocation trade-offs (Orlic et al., 2018).
Firm Size: nFulTimEmplyLFY has a small negative effect on both outcomes (\beta = -0.003, \beta = -0.000), contradicting expectations and indicating that larger firms may face bureaucratic inefficiencies (Barney, 1991).
Costs: laborCost has a negligible negative effect on product/service innovation (\beta = -0.000) but is insignificant for process innovation (p = 0.384). ElectricityCost and FuelCost show small positive effects for product/service innovation but are insignificant or weak for process innovation, reflecting mixed cost impacts (Haddoud et al., 2023).
Labor Constraints: LaborReg positively affects product/service innovation (\beta = 0.340) but negatively affects process innovation (\beta = -0.015). InadqEducWorkforce positively affects both outcomes (\beta = 1.031, \beta = 0.042), contrary to expectations, suggesting that perceived skill shortages may drive compensatory innovation efforts (Aljanabi, 2018).
Sales Expectations: ExptdFutSales negatively affects both outcomes (\beta = -1.827, \beta = -0.453), indicating that optimistic sales expectations may divert resources from innovation.
Audits: extAudit negatively affects both outcomes (\beta = -1.172, \beta = -0.189), while taxAuditLFY has a small positive effect on product/service innovation (\beta = 0.106) but is insignificant for process innovation (p = 0.0793), suggesting governance costs outweigh benefits (Evangelista & Vezzani, 2010).
Legal Status and Sector: Private shareholding (legalStat2) and sole proprietorships (legalStat3) strongly negatively affect product/service innovation, while limited partnerships (legalStat5) positively affect process innovation. Non-manufacturing sectors (MainProdServLFY2,4) show strong negative effects on product/service innovation, but wholesale trade (MainProdServLFY3) and services (MainProdServLFY6) are positive (Orlic et al., 2018).
Market Orientation: outMktOrient strongly positively affects both outcomes (\beta = 3.717, \beta = 0.495), supporting open innovation theory (Chesbrough, 2003). techInpMktOrient negatively affects both (\beta = -0.689, \beta = -0.263), indicating resource diversion in MENA (Haddoud et al., 2023).
Managerial Experience and Gender: MangYrExpSect negatively affects both outcomes (\beta = -0.098, \beta = -0.021). topManagfem negatively affects both, while femOwner positively affects product/service innovation (\beta = 0.304) but negatively affects process innovation (\beta = -0.132).
Governance and Finance: PercSenManTimGovReg and AccsToFinObstOP have small positive effects, while ProsPubSpendPriorty negatively affects product/service innovation (\beta = -0.365) but positively affects process innovation (\beta = 0.192).
| Variable | ||||
|---|---|---|---|---|
| Estimate | p-value | Estimate | p-value | |
InovStratOrient0_0_1 |
-0.985 | <2e-16 | 0.518 | <2e-16 |
InovStratOrient0_1_0 |
0.338 | <2e-16 | 0.125 | 2.31e-13 |
InovStratOrient0_1_1 |
2.104 | <2e-16 | 2.697 | <2e-16 |
InovStratOrient1_0_0 |
4.278 | <2e-16 | 1.435 | <2e-16 |
InovStratOrient1_0_1 |
0.562 | <2e-16 | 1.749 | <2e-16 |
InovStratOrient1_1_0 |
-3.272 | <2e-16 | 1.093 | <2e-16 |
InovStratOrient1_1_1 |
1.946 | <2e-16 | 1.750 | <2e-16 |
nFulTimEmplyLFY |
-0.003 | <2e-16 | -0.000 | 6.05e-16 |
laborCost |
-0.000 | <2e-16 | 0.000 | 0.384 |
LaborReg |
0.340 | <2e-16 | -0.015 | 0.00147 |
InadqEducWorkforce |
1.031 | <2e-16 | 0.042 | <2e-16 |
ExptdFutSales |
-1.827 | <2e-16 | -0.453 | <2e-16 |
ElectricityCost |
0.000 | <2e-16 | 0.000 | 5.09e-05 |
FuelCost |
0.000 | <2e-16 | 0.000 | 0.918 |
extAudit |
-1.172 | <2e-16 | -0.189 | <2e-16 |
taxAuditLFY |
0.106 | <2e-16 | 0.014 | 0.0793 |
legalStat2 |
-5.518 | <2e-16 | -0.407 | <2e-16 |
legalStat3 |
-3.570 | <2e-16 | -0.218 | <2e-16 |
legalStat4 |
-1.437 | <2e-16 | -0.221 | <2e-16 |
legalStat5 |
-1.094 | <2e-16 | 0.210 | <2e-16 |
legalStat6 |
10.780 | <2e-16 | 0.021 | 0.211 |
MainProdServLFY2 |
-10.540 | <2e-16 | -0.763 | <2e-16 |
MainProdServLFY3 |
1.304 | <2e-16 | -0.632 | <2e-16 |
MainProdServLFY4 |
-4.896 | <2e-16 | -0.206 | <2e-16 |
MainProdServLFY5 |
0.157 | <2e-16 | -0.364 | <2e-16 |
MainProdServLFY6 |
1.426 | <2e-16 | -0.589 | <2e-16 |
outMktOrient |
3.717 | <2e-16 | 0.495 | <2e-16 |
techInpMktOrient |
-0.689 | <2e-16 | -0.263 | <2e-16 |
MangYrExpSect |
-0.098 | <2e-16 | -0.021 | <2e-16 |
topManagfem |
-0.438 | <2e-16 | -0.075 | 2.67e-11 |
femOwner |
0.304 | <2e-16 | -0.132 | <2e-16 |
PercSenManTimGovReg |
0.043 | <2e-16 | 0.004 | <2e-16 |
AccsToFinObstOP |
0.076 | <2e-16 | 0.145 | <2e-16 |
ProsPubSpendPriorty |
-0.365 | <2e-16 | 0.192 | <2e-16 |
Findings in Context
The Joe model’s superior fit (\tau = 0.914) aligns with prior studies modeling asymmetric dependence in innovation outcomes (Wojtyś et al., 2018). The positive effects of IOI and outMktOrient corroborate RBV and open innovation predictions (Barney, 1991; Chesbrough, 2003), while negative effects of techInpMktOrient and nFulTimEmplyLFY highlight context-specific barriers in MENA (Haddoud et al., 2023). The unexpected positive effect of InadqEducWorkforce suggests adaptive innovation strategies, extending prior findings (Aljanabi, 2018; Evangelista & Vezzani, 2010).
Implications
This study’s objectives were to identify the joint determinants of product/service and process innovation among 9,710 private firms in the Europe and Central Asia (ECA) and Middle East and North Africa (MENA) regions, using bivariate copula models, and to address gaps in the literature regarding the interdependence of innovation outcomes and their drivers in emerging economies. The hypotheses posited positive effects of innovation orientation, firm size, and market orientation, mixed effects of cost-related factors, and context-specific influences of labor constraints and governance mechanisms. The Joe copula model’s results (\theta = 22.1, \tau = 0.914) provide robust insights for theoretical, practical, policy, and sustainable development implications.
Theoretical Implications
The study advances RBV and open innovation theories by modeling the joint dependence of innovation outcomes, with the Joe copula capturing strong upper-tail dependence (Barney, 1991; Chesbrough, 2003). The positive effect of outMktOrient (\beta = 3.717 for product/service innovation) confirms open innovation’s role in leveraging external knowledge, while the negative effect of techInpMktOrient (\beta = -0.689) challenges assumptions of universal benefits, suggesting resource diversion in MENA (Haddoud et al., 2023). The unexpected negative effect of nFulTimEmplyLFY (\beta = -0.003) refines RBV, indicating that larger firms may face innovation inefficiencies due to bureaucracy (Barney, 1991). The positive effect of InadqEducWorkforce (\beta = 1.031) suggests that perceived skill shortages spur adaptive innovation, extending labor market theories (Aljanabi, 2018). The methodological contribution of the Joe copula enhances innovation research by offering a robust framework for joint modeling (Wojtyś et al., 2018).
Practical Implications
Firms in ECA and MENA should prioritize international market engagement (outMktOrient, \beta = 0.495) to access knowledge spillovers, particularly SMEs (median nFulTimEmplyLFY = 20) (Chesbrough, 2003). The negative effect of techInpMktOrient suggests firms should balance external technology adoption with internal R&D to avoid resource strain (Haddoud et al., 2023). The positive effect of InovStratOrient1_0_0 (\beta = 4.278) underscores the value of CIR-focused strategies; managers should invest in R&D to drive product innovation. The negative effect of extAudit (\beta = -1.172) indicates governance costs; firms should streamline audit processes to minimize innovation disruptions. Addressing perceived skill shortages (InadqEducWorkforce) through training can enhance innovation, particularly in MENA (Aljanabi, 2018).
Policy Implications
Policymakers should promote export-oriented policies to leverage outMktOrient effects, aligning with SDG 9 (Chesbrough, 2003). R&D incentives, particularly for CIR-focused firms, can amplify innovation, especially in MENA (29.42% of sample) (Orlic et al., 2018). Labor market reforms to address skill shortages (InadqEducWorkforce) are critical, with vocational training programs to support innovation (Aljanabi, 2018). Policies mitigating finance constraints (AccsToFinObstOP, \beta = 0.145) through SME financing can enhance innovation. Streamlining external audits (extAudit) can reduce governance burdens, fostering innovation ecosystems (Evangelista & Vezzani, 2010).
Sustainable Development Implications
The study supports SDG 9 by identifying innovation drivers, promoting sustainable industrialization. The positive effects of outMktOrient and InovStratOrient encourage knowledge-driven growth, while addressing InadqEducWorkforce supports inclusive industrialization. Policies reducing finance and governance barriers can enhance economic resilience in ECA and MENA, aligning with SDG 9’s focus on innovation infrastructure (Haddoud et al., 2023; Orlic et al., 2018).
Conclusions and Future Research
This study aimed to identify the joint determinants of product/service and process innovation among 9,710 private firms across 41 economies in the Europe and Central Asia (ECA) and Middle East and North Africa (MENA) regions, using bivariate copula models to address the literature gap in modeling the interdependence of innovation outcomes in emerging economies. The hypotheses posited that innovation orientation, firm size, and market orientation would positively influence innovation, while cost-related factors, labor constraints, and governance mechanisms would have mixed effects, moderated by regional contexts. The Joe copula model, which outperformed Gaussian and Gumbel specifications (AIC = 3,136,340, BIC = 3,136,476), provides robust evidence partially supporting these hypotheses, offering theoretical, practical, and policy insights. Below, we summarize the findings, discuss limitations, propose future research directions, and close with reflections on the study’s broader impact.
Summary
The Joe copula model best captures the joint determinants of product/service innovation (ProdServInnov) and process innovation (ProcessInnov) among 9,710 firms in ECA (70.58% of the sample) and MENA (29.42%), fulfilling the study’s objective to model innovation interdependence. The model’s strong upper-tail dependence (\theta = 22.1, \tau = 0.914) highlights that firms excelling in one innovation type are highly likely to excel in the other, confirming the hypothesis of interdependent innovation outcomes (Damanpour et al., 2018). Key findings, as detailed in Table 5, underscore the critical role of innovation orientation, output market orientation, and perceived skill shortages, while revealing constraints posed by firm size, technological inputs, and governance mechanisms, particularly in MENA.
Innovation orientation (InovStratOrient), particularly CIR-inclusive levels (e.g., InovStratOrient1_0_0, \beta = 4.278 for product/service innovation, \beta = 1.435 for process innovation), strongly predicts both innovation types, supporting the hypothesis that R&D-focused strategies enhance innovation (Orlic et al., 2018). This aligns with the resource-based view (RBV), which posits that internal capabilities drive competitive advantage (Barney, 1991). Output market orientation (outMktOrient, \beta = 3.717 for product/service innovation, \beta = 0.495 for process innovation) reinforces open innovation principles, confirming the hypothesis that international market engagement fosters knowledge spillovers (Chesbrough, 2003). Unexpectedly, an inadequately educated workforce (InadqEducWorkforce, \beta = 1.031 for product/service innovation, \beta = 0.042 for process innovation) positively influences innovation, suggesting that perceived skill shortages spur adaptive innovation strategies, challenging the hypothesis of negative labor constraints (Aljanabi, 2018).
Conversely, firm size (nFulTimEmplyLFY, \beta = -0.003 for product/service innovation, \beta \approx 0.000 for process innovation) negatively affects both outcomes, contradicting the hypothesis that larger firms leverage greater resources and indicating bureaucratic inefficiencies (Haddoud et al., 2023). Technological input market orientation (techInpMktOrient, \beta = -0.689 for product/service innovation, \beta = -0.263 for process innovation) reduces innovation, challenging open innovation theory and supporting the hypothesis of context-specific barriers in MENA, where external technology reliance may divert resources (Chesbrough, 2003; Haddoud et al., 2023). External audits (extAudit, \beta = -1.172 for product/service innovation, \beta = -0.189 for process innovation) negatively affect innovation, contradicting the hypothesis that governance mechanisms foster innovation and highlighting compliance costs (Evangelista & Vezzani, 2010). Cost-related factors, including labor costs (laborCost), electricity costs (ElectricityCost), and fuel costs (FuelCost), show negligible or mixed effects, partially supporting the hypothesis of resource constraints (Aljanabi, 2018).
These findings address the study’s objective to inform policy and practice for SDG 9, promoting sustainable industrialization and innovation infrastructure in ECA and MENA. By identifying drivers and barriers, the study bridges the literature gap in joint innovation modeling, offering a robust framework for understanding innovation dynamics in emerging economies.
Limitations
Despite its contributions, the study faces several limitations that contextualize its findings and guide future research. First, the cross-sectional nature of the World Bank Enterprise Surveys (WBES) data (2019–2020) limits causal inference, as it captures a snapshot of firm behavior (World Bank Group, 2020). While the Joe copula model accounts for interdependence, it cannot establish temporal causality between predictors (e.g., InovStratOrient, outMktOrient) and innovation outcomes, tempering interpretations of dynamic innovation processes (Damanpour et al., 2018).
Second, the binary measures of innovation (ProdServInnov, ProcessInnov) oversimplify the complexity of innovation activities. With 23.61% of firms reporting product/service innovation and 13.95% reporting process innovation, these dichotomous variables do not capture the intensity, novelty, or scope of innovation, potentially masking nuanced effects (Orlic et al., 2018). For instance, the negative effect of techInpMktOrient may reflect resource diversion rather than innovation failure, but binary measures limit such distinctions.
Third, although the model includes country-specific smooth terms (s(CNTnameID), p < 0.001) to account for regional heterogeneity across 41 economies, it does not fully capture industry-specific dynamics beyond broad sectors (MainProdServLFY, e.g., 55.77% manufacturing). The strong negative coefficients for non-manufacturing sectors (e.g., MainProdServLFY2, \beta = -10.540 for product/service innovation) suggest sector-specific barriers, but finer-grained industry data could reveal additional insights, particularly in MENA’s diverse economies (Evangelista & Vezzani, 2010).
Fourth, missing summary statistics for some variables (e.g., topManagfem, femOwner) limit descriptive insights, constraining the interpretation of gender-related effects (Haddoud et al., 2023). Finally, the study’s reliance on the WBES dataset introduces potential selection bias, as the final sample (9,710 firms) represents a 34.62% retention rate from the original 28,042 observations due to preprocessing. While weights (wstrict) mitigate sampling biases, unobservable factors (e.g., innovation culture) may influence results, limiting generalizability (Haddoud et al., 2023).
Future Research
To address these limitations and build on the study’s findings, several avenues for future research emerge, aligning with the objective to advance innovation scholarship in emerging economies. First, longitudinal data should be employed to establish causality between predictors and innovation outcomes. Panel data from multiple WBES waves or alternative sources (e.g., OECD innovation surveys) could track changes in InovStratOrient, outMktOrient, and InadqEducWorkforce over time, clarifying their dynamic effects (Damanpour et al., 2018). This could also explore lagged effects of governance mechanisms like extAudit, which may impose delayed innovation costs (Evangelista & Vezzani, 2010).
Second, future studies should adopt finer-grained innovation measures to capture the complexity of innovation activities. Continuous or ordinal measures, such as R&D expenditure, patent filings, or innovation novelty scales, could provide deeper insights into the effects of techInpMktOrient and InadqEducWorkforce (Orlic et al., 2018). For instance, distinguishing between incremental and radical innovation could reveal whether negative technological input effects reflect strategic choices or resource constraints in MENA (Haddoud et al., 2023).
Third, industry-specific analyses are warranted to explore heterogeneity beyond broad sectors. Given the significant negative effects of non-manufacturing sectors (e.g., MainProdServLFY4, \beta = -0.206 for process innovation), studies focusing on specific industries (e.g., technology, construction) could uncover tailored drivers and barriers. This is particularly relevant for ECA’s technology-driven economies (e.g., Estonia) versus MENA’s resource-based ones (e.g., Egypt) (Aljanabi, 2018). Industry-level data could also integrate innovation networks, refining open innovation theory (Chesbrough, 2003).
Fourth, alternative copula specifications, such as Clayton or Frank copulas, could be tested to further refine the dependence structure between innovation outcomes. While the Joe copula captures strong upper-tail dependence, the Clayton copula could model lower-tail dependence, where firms with low innovation in one domain struggle in the other (Wojtyś et al., 2018). Comparative analyses could enhance methodological robustness, addressing the study’s objective to advance econometric modeling.
Fifth, future research should explore unobservable firm-level factors, such as innovation culture, leadership styles, or absorptive capacity, which may mediate the effects of InovStratOrient and outMktOrient. Qualitative or mixed-methods approaches could complement quantitative findings, providing richer insights into MENA’s innovation barriers (Haddoud et al., 2023). Additionally, extending the analysis to other emerging regions (e.g., Sub-Saharan Africa, Latin America) could test the generalizability of the Joe model’s findings, contributing to global innovation scholarship.
Sixth, integrating machine learning techniques, such as random forests or neural networks, could uncover non-linear relationships between predictors and innovation outcomes, complementing the copula framework. For instance, machine learning could identify interactions between nFulTimEmplyLFY and AccsToFinObstOP, which may be obscured in linear models (Barney, 1991). Finally, exploring gender-related variables (topManagfem, femOwner) with complete data could clarify their mixed effects, addressing diversity’s role in innovation (Evangelista & Vezzani, 2010).
Closing Remarks
This study underscores the transformative potential of targeted policies and practices to unlock innovation in ECA and MENA, fostering sustainable economic growth and resilience. By demonstrating the strong joint dependence of product/service and process innovation through the Joe copula model, the findings offer a nuanced perspective on the drivers and barriers shaping innovation ecosystems in emerging economies. The positive effects of innovation orientation, output market orientation, and adaptive responses to skill shortages, coupled with constraints from firm size, technological inputs, and governance costs, highlight the need for context-specific strategies to achieve SDG 9’s goals of sustainable industrialization and innovation infrastructure.
In ECA, where 70.58% of the sample operates, policies promoting international market engagement and R&D investment can amplify innovation, leveraging the region’s economic diversity. In MENA, where resource constraints and skill shortages are pronounced, targeted workforce development and financing programs are critical to overcoming barriers. These insights, grounded in rigorous econometric analysis, position ECA and MENA to harness innovation as a catalyst for economic resilience, aligning with global sustainable development priorities. As emerging economies navigate complex global challenges, this study offers a hopeful vision for innovation-led growth, inspiring further research and action to realize this potential.