Enterprises’ national innovation system performance in Egypt, Côte d’Ivoire and Cameroon

ABSTRACT

This study aims to analyse enterprises’ performance in terms of the National Innovation System – NIS in three Africans countries: Egypt, Côte d’Ivoire and Cameroon. Drawing inspiration from the Schumpeterian approach of innovation and using the World Bank Enterprises Survey (WBES-2016) data, we estimated a multinomial logit model. The results show that innovation system’s different spheres (academic, governmental, enterprise, societal and environmental) of the quintuple helix contribute more to improving the performance of technological innovation in Egypt than in other countries. In addition, the analyses enabled us to highlight the embryonic state of the NIS in Africa. Thus, underlying innovation policies have been proposed to improve enterprises’ performance in terms of the NIS in the different countries studied.

KEYWORDS:

• National Innovation System
• technological innovation
• multinomial logit
• Africa
• enterprise‌

JEL Classification:

• O3
• O33
• C1
• O55

1. Introduction

The concept of Innovation System – IS came from the transition of innovation as a linear innovation process (Solow 1956; Romer 1986, 1990) to that of actors inserted into different institutional networks. From that view, innovation necessarily involves interactions between actors (enterprises, laboratories, universities, etc.) and their environment (Amable and Petit 2001). Several IS developments can be distinguished: the Technological Innovation System – TIS (Carlsson and Stankiewicz 1991), the Sectoral Innovation System – SIS (Breschi and Malerba 1997), the Regional Innovation System – RIS (Asheim and Isaksen 1997; Cooke, Uranga, and Etxebarria 1997; Storper 1995) and the National Innovation System – NIS (Freeman 1995; Lundvall 1992; Nelson 1993). That last version of the IS involves a number of institutions whose interactions determine enterprises results mainly in terms of technological innovation.

According to the knowledge economy, the importance of technological innovation, as a third market mechanism in addition to supply and demand, has led to the neo-institutional development: the triple helix model (Etzkowitz and Leydesdorff 1995, 1997; Leydesdorff and Etzkowitz 2000). This approach encourages interactions between the government, enterprises and universities as the technological innovation basis. Moreover, in order to take into account the contributions of civil society and the natural environment, the triple helix has been extended to the quadruple helix and quintuple helix model of innovation, respectively (Carayannis and Campbell 2009, 2010). The quintuple helix innovation system makes it possible to highlight the potential of the environment where the enterprise is located, in terms of the ability to generate innovation.

The analysis of IS in Africa reveals the inexistence of interactions between economic agents to generate technological innovation. Studies have shown that IS in Africa is still at a primitive stage, unlike those in developed countries (Gu 1999). Like developing countries, African countries’ IS are considered inadequate, incomplete, underdeveloped and characterized by very weak interactions between their spheres (Casadella 2006). As proof, it emerges some controversial results in terms of innovation performance for the above-mentioned countries. African countries’ innovation capabilities analysis reveals a gradual decline in the performance of Côte d’Ivoire and Cameroon, as compare to that of Egypt. Hence, we note decreasing progression scores of 139, 127, 88and 44 for Côte d’Ivoire, and 110, 105, 64 and 46 for Cameroon against an increasing progression of 80, 111, 131 and 132 for Egypt (WEF 2020). Moreover, none of these countries ranks in the top 10 countries in terms of innovation according to the categories of high-income countries, upper and lower-middle-income countries and low-income countries (Index 2020). However, in terms of overall innovation performance, Egypt ranks highest, 96th ahead Côte d’Ivoire and Cameroon, which rank the 112th and 119th, respectively, out of a total of 131 countries considered in the World Intellectual Property Organization’s report. Egypt’s innovation performance as compared to the other countries can be explained by the existence of tax subsidy policies that favour enterprise’s innovation. Thus, the main question to be examined in that study is: what effects does the NIS have in terms of the ability to generate technological innovation by Egyptian enterprises compared to the Côte d’Ivoire, and Cameroonian ones? To answer this question, the objective of this study is to make a comparative analysis of the different components’ (spheres) contributions of the NIS to the Egyptian, Côte d’Ivoire and Cameroonian’ enterprises technological innovation.

This study is the first ever attempt to empirically analyse the NIS at the enterprises level in Africa. No study to our knowledge has been interested in formulating economic policies toward technological innovation (product innovation, process innovation and both product and process innovation) based on the role of the different spheres of the NIS in Africa. Our work fills this gap in the literature by verifying which spheres (academic, governmental, industrial, societal and environmental) of the quintuple helix contribute more to the improvement of technological innovation (products innovation, process innovation and both product and process innovation) performance in Egypt, Côte d’Ivoire and Cameroon; and also by accessing differences in NIS performance between these three countries.

The rest of the paper is presented as follows: after this introduction, we present a synthesis of the literature followed by the study hypothesis (section 2). Section 3 presents the methodology and the data. The main results and discussion are presented in section 4, and finally, section 5 conclude and formulate some innovation policies to improve enterprises’ innovation performance.

2. Literature review

Theoretical developments on IS began with Lundvall (1985) on the concept of the NIS before Freeman’s innovation policy in Japan in 1987. Two groups of studies emerge from the debates on the NIS. The first group led by Lundvall (1992) focused on research about the NIS’s analytical content. The author investigates the roles played by users, the public sector and financial institutions. The second group led by Nelson (1993) deals with empirical illustrations. That group has tried to understand and make a comparative analysis of the links and differences between institutional infrastructures as a determinant of a nation’s economic performance. More specifically, Nelson focused on the institutions whose interactions determine domestic enterprises’ innovative performance. Thus, he delimits the NIS’s scope to R&D activities by associating other dimensions related to science and technology policies or intellectual property legislation (Garrouste and Kirat 1995). Nelson’s studies have also described the main characteristics of the NIS in countries with different income levels. Freeman, Lundvall and Nelson’s work has thus focused on the application of IS at the country level, given that, this dimension is governed by the sharing of the same language, culture and specific institutional rules related to the knowledge producers. The NIS’s different actors are thus bound by a certain cognitive and institutional proximity.

In his analyses, Lundvall (1992) basically identifies two IS’s approaches. On the one hand, IS in the narrow sense, which is limited to taking into account the contributions of enterprises, universities and the government. This systemic approach of innovation is similar to those of the NIS and the triple helix model, relying on universities as basis. On the other hand, the broad concept extends to all economic and institutional structures that affect the production system. Considering that view, the contributions of the civil society, as well as those of the natural environment have been taken into account in order to reinforce the IS. Thus, the triple helix model has been extended to the quadruple helix and quintuple helix model of innovation. These different systemic models of innovation encourage interactions between the government, universities, enterprises, civil society and the natural environment in order to generate technological innovation, with universities as the IS’s main actor (see Figure 1).

Figure 1. Representation of the Quintuple Helix model of innovation. Source: Authors’ construction, 2019.

Enterprises’ contribution to the IS is based on both evolutionary and neoclassical theories. Evolutionists such as Nelson and Winter (1982) consider the enterprise as the technological change primary agent, while Romer (1990) suggests knowledge accumulation through investment in R&D to improve economic performance. Enterprises-based studies emphasize, in addition to R&D spending (Mohnen and Röller 2005), patents, technology licences, etc. as technological innovation basis (Crépon and Duguet 1994; Okubo 1997). Thus, the enterprise’s own investments in R&D have an impact on its innovative capacity. Barasa et al. (2019) analysis of the effect of innovation inputs on the manufacturing firm’s efficiency in Sub-Saharan Africa (SSA) shows that internal R&D have a positive impact against a negative impact on foreign ownership. Hussen and Çokgezen (2021) argued that firms’ R&D contributes only to product innovation and technological innovation in Africa. Analysing the Swedish enterprises, Baum et al. (2017) also emphasized that R&D contributes to innovation for both manufacturing and services. Kasongo, Sithole, and Buchana (2021) documented that patent positively enhances enterprises’ probability to innovate while it has no effect on their innovation intensity in SSA.

Since universities are considered to be crucial in knowledge production and dissemination, many enterprises are engaged in training or upgrading programmes for their staff. Those trainings are offered by universities for new knowledges production and dissemination in order to reinforce enterprises’ performance. The role of universities is therefore focused on education and human capital training (Romer 1990) that favour enterprises. Affes and Chouaibi (2007) study on 61 Tunisian enterprises operating in the agro-food sector reveals that the more enterprises are involved in training programmes, the more they are encouraged to invest in innovative projects. Barasa et al. (2019)’s study shows that human capital development and educational attainment have a negative impact on the enterprises efficiency in SSA. However, Hussen and Çokgezen (2021)’s results reveal that staff training contributes to the firm’s probability to generate product and/or process innovation. High-skilled workers seem to affect negatively product and process innovation and positively technological innovation but the effect is non-significant. Kasongo, Sithole, and Buchana (2021) show that scientific information does not contribute to enterprises innovation intensity in SSA.

Governments constitute an essential link in the IS through their actions in terms of the economic activities development. Following Barro’s (1990) works, governments ensure public spending on physical capital and social infrastructures for economic performance improvement. Thus, they contribute to the IS not only through their supreme function, but also by granting subsidies and/or credit to enterprises in order to encourage their innovation activity development. That is why Aerts and Czarnitzki (2006), Mohnen and Lokshin (2010) and Czarnitzki and Fier (2003) emphasize the importance of public subsidies for technological innovation. Although they have worked in different sectors, they all found a positive effect of public support on enterprises’ technological innovation. Hussen and Çokgezen (2021) find from their micro-evidence in Africa that government-owned firms contribute to both probability and intensity of innovation but have a negative effect on product innovation. The effect of foreign ownership is non-significant on the previous innovation variables. Bekana (2020) shows that government effectiveness and regulatory quality role contribute to global innovation output in SSA. Kasongo, Sithole, and Buchana (2021) show that public information and public finance support have a positive effect on the enterprises’ innovation intensity. Fu, Mohnen, and Zanello (2018) find that subsidised loan contributes to both technological and non-technological innovation in Ghana.

The civil society contributions to the IS rely on the Neo-Keynesian works (Kaldor 1966, Pasinetti 1981, Dosi and Nelson 1994) which consider in particular the internal demand; i.e. the consumers’ degree of orientation towards local products, as a basis for improving enterprises’ innovation performance. This demand is better controlled by the enterprise through various communication relations that it develops with its customers. Habib and Jamel’s study (2007) shows that collaboration with clients favours innovation, because the greater the commitment and collaboration, the greater the enterprise’s incentive to innovate. Barasa et al. (2019)’s study shows that enterprises’ labour force provided by the civil society contributes to their efficiency in SSA. Hussen and Çokgezen (2021) show that the experience of the enterprise top manager which corresponds to labour force contribution to the IS has a positive effect on both probability and intensity to innovate while it has a negative effect on product and process innovation. Fu, Mohnen, and Zanello (2018) show that skilled workers contribute to both technological and non-technological innovation in Ghana.

Finally, the natural environment’s contributions to the IS have been studied by Carayannis and Campbell (2010) and Peris-Ortiz et al. (2016) who identify the natural environment as both a resource and a constraint that affects innovation activities. Thus, empirical works’ results are contradictory. While some studies show a positive effect of the local environment on the enterprise’s ability to innovate, other studies, do not find any significant effect. Guesnier’s (1994) study on Small and Medium-sized Enterprises (SMFs) in the weakly industrialized Poitou-Charente region in France is noteworthy. This region, characterized by a dynamic based on support for innovation, highlighted local, regional, national and even global variables. The author finds that the weight of the local environment in the enterprises’ innovation strategies remained low, and that, the first thing to emerge was extraterritorial relations, with a rather national or even international dimension. Bougrain’s (1999) work on enterprises in the Centre region of France confirm the findings that technological cooperation takes place only in a rather national or even international space, whether it is with customers, suppliers, competitors, or technical or scientific institutions. Similarly, Fort, Rastoin, and Temri (2005), in their study of agro-food companies, have highlighted those findings on enterprises’ abilities to innovate, according to their local environment and the technological learning methods they implement. However, the local environment of innovation sources does not seem to influence the propensity of enterprises to innovate. However, Maillat and Lecoq (1992) study based on a survey of 137 SMFs located in the Jura region of Switzerland, identified a strong involvement of the local level in enterprises’ innovation strategies. Baum et al. (2017) find that regional resource contributes to innovation in Swedish manufacturing and service enterprises. Moreover, they note that the origin of the entrepreneur, the development of activities, traditional know-how and the consumption of current or specialized services are also factors that determine enterprises’ innovation performance. Hussen and Çokgezen (2021) show that enterprises’ localization in a large city has a positive but non-significant effect on innovation performance in Africa. Fu, Mohnen, and Zanello (2018) also find that localization contributes to non-technological innovation in Ghana.

Regarding these theoretical and empirical developments, it emerges, on the one hand, a lack of empirical studies on enterprises’ NIS in Africa. Existing case studies in Africa link innovation to firm’s productivity or firm’s performance. For instance, Fu, Mohnen, and Zanello (2018) focused on innovation and productivity in formal and informal firms in Ghana; and Barasa et al. (2019) studied the relationship between innovation inputs and efficiency by focusing on manufacturing firms in Sub-Saharan Africa. An empirical analysis of innovation and productivity in services firms in South Africa was conducted by Kassongo, Shahsavari, and Ball (2021). Hussen and Çokgezen (2021) analysed the relationship between innovation, regional institutions and firm performance in Africa.

On the other hand, none of the studies on NIS to our knowledge has been interested in formulating economic policies toward technological innovation (product innovation, process innovation and both product and process innovation) based on the role of the NIS’s different spheres in Africa. Thus, our work is the first ever attempt to fill this gap in the literature by verifying the hypothesis according to which: the NIS’s different spheres (academic, governmental, industrial, societal and environmental) of the quintuple helix contribute more to the improvement of technological innovation (product innovation, process innovation and both product and process innovation) performance in Egypt than in Côte d’Ivoire and Cameroon. This hypothesis can be justified by the fact that Egyptian enterprises are characterised by the existence of direct subsidies from the government on contrary to the Cameroonian and Côte-d’Ivoire enterprises.

3. Methodology

In this section, we present successively the theoretical benchmark model and the underlying econometric specification, the model validity criteria and robustness analysis, and finally the data.

3.1. Theoretical benchmark model and econometric specification

In order to analyse enterprises’ performance in terms of NIS (which encourages the interaction between different spheres in order to generate technological innovation), we build our theoretical model relying on the Schumpeterian approach in which the innovation technology of an operative enterprise is given by: (1) ${\mu }_j=\varphi Z_j^{{\delta }_z}$(1) with ${\mu }_j$ the probability of innovation of enterprise $j$, $Z_j$ the enterprise R&D expenditures (which here indicate the enterprise’s contribution in the NIS), ${\delta }_z$ the elasticity of innovation with respect to R&D investment, and $\varphi >0$ a parameter representing R&D sector productivity. Thus, following our systemic approach, the dynamics of innovation will be determined by interactions between the NIS’s different spheres according to Equations (2) and (3) below: (2) ${\mu }_j=\mathrm{\varphi \omega }(V_j,W_j,X_j,Y_j,Z_j)$(2) (3) $\omega (.)=V_j^{{\delta }_v}{W}_j^{{\delta }_w}{X}_j^{{\delta }_x}{Y}_j^{{\delta }_y}{Z}_j^{{\delta }_z}$(3) where $V_j,W_j,X_j$ and $Y_j$ respectively enable us to account for the contribution of the IS’s academic, governmental, societal and environmental spheres of the quintuple helix. Thus, by substituting $\omega (.)$ in Equation (2) and then taking the log on both sides of the equation, we end up with the econometric transformation of the model specified below: (4) $I{T}_j={\delta }_c+{\delta }_v{v}_j+{\delta }_w{w}_j+{\delta }_x{x}_j+{\delta }_y{y}_j+{\delta }_z{z}_j+{\mathcal{E}}_j$(4) In this equation, ${\mathcal{E}}_j$ corresponds to the error term and $I{T}_j$ corresponds to the NIS output indicating enterprises’ different modes of innovation. Given the qualitative nature of our dependent variable, specification (4) is in fact a multinomial logit specification. In particular, $I{T}_j$ corresponding to the NIS output is captured by a multinomial variable with different modes such as: 0 = ‘do not innovate’, 1 = ‘product innovation’, 2 = ‘process innovation’ and 3 = ‘product and process innovation’. The variables that account for enterprises, universities, government, society and environment’s contributions to the NIS are respectively captured by expenditures on Research and Development (R&D), Staff Training (Fperso), Delay of obtaining operating licences (Dlexploi), Internal Demand (Dinterne) and enterprises’ Land Ownership (DT) (see Tables A2–A4 in the Appendix).

The R&D variable corresponds to the enterprise’s financial contribution to the innovation activity. Staff training indicates interactions between enterprises and academics which provides skills for innovation. The delay in obtaining operating licences reflects the efficiency of public services. We use internal demand to capture the civil society contribution because the consumer’s degree of orientation toward internal products is relevant for innovation. Land Ownership corresponds to the percentage of the land occupied by the establishment. The choice of that variable is related to the fact that the increase of the land occupied percentage is related to enterprise size, and thus, may determine its probability to innovate.

We therefore expect NIS’s spheres to have a positive effect on the different types of technological innovation, except for the variable Dlexploi, from which we may get a positive or negative effect. Although a negative sign would seem more intuitive, a positive effect of this variable would therefore suggest that enterprises will have a greater incentive to innovate in order to offset the losses related to the delay in obtaining operating licences.

3.2. Model validity

The multinomial logit model parameters’ estimation is performed using log-likelihood function maximization algorithms. Coefficients are not directly interpretable in terms of marginal propensity, only the signs of the coefficients indicate a favourable or unfavourable contribution of the different spheres to the IS. The coefficient’s significance can be assessed using the z-statistic ratios and the overall significance of the adjustment $(\mathrm{Hypothesis}{H}_o:{\delta }_c={\delta }_v={\delta }_w={\delta }_y={\delta }_x={\delta }_z=0)$ by the Likelihood Ratio statistic which follows, under the null hypothesis $H_o$ a ${\chi }^2$ distribution with 5 degrees of freedom, and that we compare to a ${\chi }^2$ read from the table at a threshold of 0.95% and at 5 degrees of freedom.

Notice that our econometric transformation of the model specified in Equation (4) only includes the Innovation System variables, that are variables related to enterprises, universities, the government, the civil society and the environment. However, many other factors may influence the innovation system’s performance. In other words, some enterprise-specific variables may influence the enterprise’s innovation activities, whether it is product, process, or both product and process innovation combined.

3.3. Robustness check

For the robustness check, we account for many control variables such as the enterprise’s industry sector, size and ownership. We add one additional variable accounting for the environment’s contribution to the NIS which is the enterprise’s geographical localization (Local). The variable localization indicates whether the enterprise is located in the main business city or not. The idea behind that variable is that, generally, the main business city of each country is characterized by a highly skilled labour, an access to financing, a high connected network, so that enterprises may interact with other enterprises and easily develop new products/process. We also add a quantitative variable related to human capital (HighSchool) indicating the percentage of full-time workers who completed high school within the enterprise. That variable helps to account for universities’ contributions to the NIS. For Egypt, we consider whether the enterprise receive direct subsidies (subs) from the government, which also accounts for the government contribution to the NIS. That subsidy policy variable is only available for Egypt, and we have 37 over 1814 enterprises that benefit from subsidies.

Moreover, we perform an estimation by Instrumental Variables – IV. The need for this analysis stems from the potential existence of inverse causality between R&D spending and technological innovation. R&D spending generates technological innovation just as the incentive to innovate leads enterprises to spend on R&D. It is therefore important to find a source of exogenous variation for R&D, i.e. instruments that influence innovation only through R&D. The validity of the IV estimator depends on the validity of the instrumental variables. This requires two things: (i) a strong correlation between the instruments and the variable to be explained and (ii) an exogenous instrument. With respect to our identification strategy, the variables: holding technology licenses and enterprises’ operating licenses are therefore retained as instruments. The reason is that, technology licenses reflect a property right on an innovation that has already occurred, and operating licenses are related to an authorization to exploit an existing technology. In addition, the possession of technology and exploitation licences promotes learning and enterprises’ incentive to spend on R&D in order to innovate.

3.4. Data

We used data from the World Bank Enterprises Survey¹ (WBES). These surveys are conducted in collaboration with the European Bank for Reconstruction and Development (EBRD), the European Investment Bank (EIB) and the European Union (EU). This database provides information on the NIS characteristic variables, which are: R&D, operating licences, grants, annual sales, staff training, technological innovation, etc. It covers more than 3900 African enterprises in 9 countries: Benin, Cameroon, Côte d’Ivoire, Egypt, Guinea, Lesotho, Mali, Togo and Zimbabwe. Hundred and fifty enterprises were surveyed in Benin, Guinea, Lesotho, Mali and Togo, 361 enterprises in Cameroon and Côte d’Ivoire, 600 in Zimbabwe and 1814 in Egypt. Our empirical study only includes Egypt, Cameroon and Côte d’Ivoire. The choice of Cameroon, Côte d’Ivoire and Egypt is based on the fact that these countries have the latest survey in the same year (2016) and that their sample size is relatively consistent for the analysis (see descriptive statistics tables A7–A12 in the Appendix) (Tables A10–A12).

Firms’ are distributed through different sectors in the three countries, Egypt, Cameroon and Côte d’Ivoire (Table A6 in appendix). Firms are categorised into three main industry sectors: manufacturing, retail services and other services. The manufacturing sectors includes: extractives, food, tobacco, textiles, garments, leather, wood, paper, publishing, printing and recorded media, refined petroleum product, chemicals, plastics and rubber, non-metallic mineral products, basic metals, fabricated metal products, machinery and equipment, electronics, precision instruments, transport machines, furniture and recycling. Retail services are stated retail. Others services includes: wholesale, IT, hotel and restaurants, construction, services of motor vehicles, transport, travel agencies and real estate. We count 1193 manufacturing firms in Egypt, 102 in Cameroon and 106 in Côte d’Ivoire. 113 firms operate in retail services in Egypt, 106 in Cameroon and 117 in Côte d’Ivoire. 508 firms do other services in Egypt, 153 in Cameroon and 138 in Côte d’Ivoire (see Table A6 in Appendix).

4. Comparative analyses of enterprises’ NIS performance in Cameroon, Côte d’Ivoire and Egypt

This section presents the descriptive analysis of enterprises’ innovation activity in Egypt, Côte d’Ivoire and Cameroon. It also presents econometric results and discusses about them.

4.1. Enterprises’ representation according to their innovation activity

Chart 1 presents enterprises’ innovation activity. It shows that 115 enterprises conduct product innovation in Egypt, while in Cameroon and Côte d’Ivoire we get 106 and 89 respectively. In terms of process innovation, we register 34 enterprises in Egypt, 20 in Côte d’Ivoire and 12 in Cameroon. And finally, considering both types of innovations simultaneously, the number of innovative enterprises in Egypt is 69 against 42 in Côte d’Ivoire and 40 in Cameroon.

Chart 1. Overall representation of innovative enterprises. Source: Authors using data from WBES-2016.

We can thus notice that Egypt has the largest number of enterprises conducting innovation. While the number of enterprises conducting product innovation in Cameroon is higher than in Côte d’Ivoire, the latter registers more enterprises that conduct process innovation and both product and process innovation (simultaneously) than Cameroon. Product innovations mostly occurred by enterprises in all three countries. Thus, analysing Figure 1 reveals that enterprises’ innovations are fundamentally expressed in terms of product innovation for all countries. Very few enterprises strive to introduce process innovations, which makes the number of enterprises relatively small when considering the two types of innovation simultaneously.

4.2. Econometric results and discussions

We analysed results comparing the three countries studied: Egypt, Côte d’Ivoire and Cameroon. Each NIS actor’s contributions are recorded in the subsequent tables following the different types of innovation: ‘product innovation’, ‘process innovation’ and ‘product and process innovation’ versus do not innovate. Table 1 presents the multinomial logit regression results according to the first type of innovation – ‘product innovation’ (InvProd) versus non-innovation and the relative risk ratio (RRR) associated to each coefficient.

Table 1. m-logit regression results following the ‘Product Innovation’ mode.

Variables	Egypt		Côte d’Ivoire		Cameroon
	InvPrd	RRR	InvPrd	RRR	InvPrd	RRR
RD	1.452***	4.271***	0.209	1.232	0.284	1.328
	(0.265)	(1.134)	(0.488)	(0.601)	(0.407)	(0.541)
Fperso	1.248***	3.483***	0.417	1.517	0.196	1.217
	(0.216)	(0.752)	(0.262)	(0.398)	(0.267)	(0.325)
Dlexploi	0.002	1.002	0.022	1.022	0.025	1.025
	(0.001)	(0.001)	(0.015)	(0.016)	(0.025)	(0.026)
Dinterne	−0.009***	0.991***	0.002	1.002	0.006	1.006
	(0.003)	(0.003)	(0.005)	(0.005)	(0.006)	(0.006)
TD	−0.002	0.998	−0.001	0.999	0.002	1.002
	(0.002)	(0.002)	(0.003)	(0.003)	(0.003)	(0.003)
Constant	−2.353***	0.095***	−2.087**	0.124**	−2.049**	0.129**
	(0.362)	(0.034)	(0.820)	(0.102)	(0.914)	(0.118)
Pseudo R^2	0.1240		0.0561		0.0285
Prob > chi^2	0.000		0.0001		0.1237
LR chi^2	218.83		43.37		21.43
Observations	1814		361		361

Note: Standard errors are expressed in parentheses, level of significance ***p < 0.01, **p < 0.05, *p < 0.10.

Source: Authors’ estimates on STATA from WBES-2016 data.

That table shows that, in Egypt, the probability for an enterprise to conduct product innovation increases, as increases by one unit, the probability of their R&D spending. This result confirms those found by Hussen and Çokgezen (2021). However, Côte d’Ivoire, and Cameroonian enterprises probability of R&D spending has no significant effect on the probability to conduct product innovation. The same finding emerges from the academic sphere’s contribution analysis through the fact that enterprises may set up or not formal training programmes for their staff (Fperso). Enterprises’ probability to conduct product innovation in Egypt increases significantly (we obtained no significant effects for Côte d’Ivoire and Cameroonian enterprises) following a one-unit increase in the probability that enterprises will implement formal training programmes for their staff. In addition, the relative probability ratio of enterprises that invest in R&D is 4271 higher than that of enterprises which do not invest in R&D, while the relative probability ratio of enterprises that implement formal training programmes for their employees is 3483 higher than that of enterprises which do not implement it. These results are not fare from those obtained by Hussen and Çokgezen (2021), Kasongo, Sithole, and Buchana (2021).

However, analysing the societal sphere’s contribution to the NIS, we observe that internal demand (measured by the share of internal sales in the national ones) has a reducing effect on enterprises’ probability to conduct product innovation in Egypt. In Cameroon and Côte d’Ivoire, consumer’s preferences in terms of internal demand do not favour product innovation. Similarly, analysing the governmental and environmental spheres’ contribution to the NIS through the delay of obtaining operating licences (Dlexploi) and the percentage of land held by enterprises (TD) (in Egypt, Côte d’Ivoire and Cameroon), we got no significant effects on the probability to conduct product innovation.

Table 2 presents the multinomial logit estimation results following ‘process innovation’ (InvPrc) mode versus non-innovation. These results reveal that enterprises’ probability to conduct process innovation in Egypt increases as increase R&D and staff training, while in Côte d’Ivoire and Cameroon we obtain no effect. These results are in line with those found by Barasa et al. (2019). Besides, the delay in obtaining operating licences only contributes to improving enterprises’ probability to conduct process innovation in Côte d’Ivoire. However, the increase in domestic demand reduces enterprises’ probability to conduct process innovation in Egypt and Cameroon. Similarly, the proportion of land owned by enterprises contributes to reducing the probability to conduct process innovation in Egypt.

Table 2. m-logit regression results following the ‘Process Innovation’ mode.

Variables	Egypt		Côte d’Ivoire		Cameroon
	InvPrc	RRR	InvPrc	RRR	InvPrc	RRR
RD	1.269***	3.557***	−0.219	0.803	−12.824	0.000
	(0.479)	(1.705)	(1.070)	(0.859)	(554.781)	(0.001)
Fperso	1.022***	2.777***	0.299	1.348	0.225	1.252
	(0.384)	(1.065)	(0.490)	(0.661)	(0.653)	(0.817)
Dlexploi	0.001	1.001	0.028*	1.028*	0.009	1.009
	(0.003)	(0.003)	(0.016)	(0.017)	(0.063)	(0.064)
Dinterne	−0.009*	0.991*	−0.008	0.992	−0.016*	0.984*
	(0.006)	(0.006)	(0.007)	(0.007)	(0.009)	(0.009)
TD	−0.007*	0.993*	−0.002	0.998	0.002	1.002
	(0.004)	(0.004)	(0.006)	(0.006)	(0.007)	(0.007)
Constant	−3.020***	0.049***	−2.893***	0.055***	−1.754	0.173
	(0.581)	(0.028)	(0.977)	(0.054)	(1.893)	(0.328)
Pseudo R^2	0.1240		0.0561		0.0285
Prob > chi^2	0.000		0.0001		0.1237
LR chi^2	218.83		43.37		21.43
Observations	1814		361		361

Note: Standard errors are expressed in parentheses, level of significance ***p < 0.01, **p < 0.05, *p < 0.10.

Source: Authors’ estimates on STATA from WBES-2016 data.

Table 3 presents the m-logit estimation results following ‘product and process innovation’ (InvPrdPrc) mode versus non-innovation. From that table, we observe some increases in enterprises’ probability to conduct product and process innovation: in Egypt, Côte d’Ivoire and Cameroon due to an increase in the probability of implementing formal training programmes for staff; in Egypt and Côte d’Ivoire due to an increase in the probability of conducting R&D and the percentage of land owned by enterprises; and in Egypt due to an increase in the delay of obtaining operating licences. These results are in line with those fund by Barasa et al. (2019) and Hussen and Çokgezen (2021). This positive effect of Dlexploi variable would therefore suggest that enterprises have a greater incentive to innovate in order to offset losses related to the delay of obtaining operating licences by increasing their revenues once they are operational.

Table 3. m-logit regression results following the ‘Product and Process Innovation’ mode.

Variables	Egypt		Côte d’Ivoire		Cameroon
	InvPrdPrc	RRR	InvPrdPrc	RRR	InvPrdPrc	RRR
RD	1.706***	5.505***	1.792***	6.001***	0.727	2.068
	(0.305)	(1.677)	(0.444)	(2.665)	(0.493)	(1.019)
Fperso	1.764***	5.833***	0.957***	2.604***	0.787**	2.198**
	(0.284)	(1.654)	(0.370)	(0.965)	(0.372)	(0.816)
Dlexploi	0.004**	1.004**	−0.010	0.991	−0.024	0.977
	(0.002)	(0.002)	(0.018)	(0.018)	(0.041)	(0.040)
Dinterne	−0.002	0.998	−0.002	0.998	0.003	1.003
	(0.005)	(0.005)	(0.007)	(0.007)	(0.008)	(0.008)
TD	0.013***	1.014***	0.007*	1.007*	0.006	1.006
	(0.005)	(0.005)	(0.004)	(0.004)	(0.004)	(0.004)
Constant	−5.158***	0.006***	−2.099**	0.123**	−1.887	0.152
	(0.690)	(0.004)	(0.988)	(0.121)	(1.373)	(0.208)
Pseudo R^2	0.1240		0.0561		0.0285
Prob > chi^2	0.000		0.0001		0.1237
LR chi^2	218.83		43.37		21.43
Observations	1814		361		361

Note: Standard errors are expressed in parentheses, level of significance ***p < 0.01, **p < 0.05, *p < 0.10.

Source: Authors’ estimates on STATA from WBES-2016 data.

4.2.1. Robustness analysis

The robustness results are presented in the Tables A1–A5 in the appendix. Table A1 presents results obtained by adding controls as well as environmental Local and university HighSchool variables according to the first type of innovation – ‘product innovation’ (InvPrd) versus non-innovation. Results from table 4 reveal positive effects of the delay of obtaining operating licences and the probability that the enterprise is held by foreign ownership. The size of the enterprise contributes to the probability of innovation in Cameroon and Côte d’Ivoire. However, enterprises’ geographical location, which accounts for the environmental contribution to the IS, has a negative effect on product innovation. The highschool variable related to the university contributes positively to the Côte d’Ivoire’s NIS. This latter result is in contrary to those found by Hussen and Çokgezen (2021) and Fu, Mohnen, and Zanello (2018) in African and Ghanaian’s enterprises, respectively. Table A2 presents results for ‘process innovation’ (InvPrc) versus non-innovation. Results confirm the previous ones and emphasized that the internal demand enhances the probability to generate process innovation in Cameroon. Results are the same for Egypt in addition to the size of the enterprise which contributes also positively to the NIS. Table A3 presents results for ‘product and process innovation’ (InvPrdPrc) versus non-innovation. We also confirm previous results in terms of the probability to generate both product and process innovation except for the environmental, society and university variables. In addition, the highschool variable of university sphere and Others services (compared to manufacture services) variables contribute to the decrease in the product and process innovation. Table A4 presents robustness results for Egypt with subsidies variable. Results reveal that the geographical location has a negative effect on the probability to generate both products and process innovation. The manufacturing as compared to other services has a negative effect on both product and process innovation in Egypt. In the contrary, the size of the enterprise favours process innovation in Egypt. Direct subsidies received by enterprises reveal a positive effect on the probability to generate product innovation. Moreover, direct subsidies improve the contribution of all components of the IS in terms of the probability of innovation except for the geographical location variable. These results are in line with those of Hussen and Çokgezen (2021) and Fu, Mohnen, and Zanello (2018). Our findings can be justified by the fact that the analysis focuses mainly on SMEs, which generally face very strong competition in the central business city and are forced to innovate according to their sector of activity in order to avoid bankruptcy.

Furthermore, Table A5 presents the IV regression results. The presumption of endogeneity bias was not confirmed and the results obtained are not better than those provided by the multinomial logit regression (Tables 1–3). In summary, the results analysed above confirm Gu’s (1999) work, who reports the IS primitive stage in Africa, unlike that of developed countries. Results also concur with Casadella (2006) conclusions which highlight marginal, non-existent, few or inappropriate interactions between NIS’s different spheres in developing countries. Several implications for innovation policies emerge from these findings in order to improve enterprises NIS performance in the different countries studied.

5. Conclusion and policies implications

In this paper, we studied enterprises’ NIS performance in Africa. Specifically, we analyse enterprises’ performance in terms of NIS considering three African countries: Egypt, Côte d’Ivoire and Cameroon. The data used are from the World Bank Enterprises Surveys (WBES-2016). Because of the qualitative and multimodal nature of our dependent variable, we estimated a multinomial logit model. Besides, an estimation by instrumental variables was carried out in order to take into account the plausible existence of inverse causality between R&D expenditures and technological innovation.

The results reveal that under the ‘product innovation’ mode, NIS spheres such as enterprises, universities and civil society (negatively) influence enterprises’ innovation dynamics only in Egypt. In other countries, however, interaction for technological innovation is almost non-existent. Moreover, according to the ‘process innovation’ mode, NIS spheres that influence enterprises’ innovation dynamics in Egypt are enterprises, universities, civil society (negative effect) and the environment (negative effect); the government in Côte d’Ivoire, and civil society in Cameroon. Furthermore, according to the ‘product and process innovation’ mode, the main actors (spheres) are enterprises, universities, the government and the environment in Egypt; enterprises, universities and the environment in Côte d’Ivoire; and enterprises and universities in Cameroon. Thus, the IS’s different spheres (academic, governmental, industrial, societal and environmental) of the quintuple helix contribute more to improving technological innovation performance in Egypt than in Côte d’Ivoire and Cameroon. We undertake robustness analysis by adding controls and others variables related to the IS (the geographical localization and direct subsidies) and the results are stable.

Based on our findings, we strongly suggest some policies implications for the three countries considered in this study. For Côte d’Ivoire and Cameroon, we suggest that Governments should put in place subsidy policies that favour enterprises, as in Egypt, in order to increase their R&D spending incentive. Cameroonian and Côte d’Ivoire enterprises should establish some formal training programmes for staff for more knowledge accumulation. Besides, Egyptian, Côte d’Ivoire and Cameroonian Governments should implement protectionist policies that favour foreign ownership and local enterprises in order to increase internal sales of their product resulting from innovation. Governments should encourage enterprises by improving operating licences’ services, and enterprises’ land ownership regime in order to enhance their capacity to generate product innovation.

The main limitation of that study is the use of some proxy variables to measure the IS components. To apprehend enterprises contribution to the NIS we could also use other variables such as patents or technology licences used by firms. As for the government contribution to the NIS, variables such as public spending on physical capital and social infrastructures should also be interesting to be considered. We use variables related to the natural environment only as a resource and not as constraint that may affect differently innovation activities. Future studies should consider the collected survey in addition to the previous variables in order to adequately analyse the NIS performance.

Acknowledgements

Thanks to the European Commission – EC which initiated the Entrepreneurship, Resources, Management, Innovation and Technologies – ERMIT project and of which we are scholarship recipients.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Notes

1 In Latin America, ES are conducted by the Inter-American Development Bank (IADB). Other institutions that have contributed to ES in the past include the British Department for International Development (BDID), the German Fund for Asset Growth (GFAG), etc.

Appendix

Table A1. Robustness check with controls for ‘Product Innovation’.

Variables	Egypt	Côte d’Ivoire	Cameroon
	InvPrd	InvPrd	InvPrd
RD	1.544***	0.064	0.439
	(0.274)	(0.503)	(0.434)
Fperso	1.320***	0.198	0.009
	(0.225)	(0.280)	(0.284)
Dlexploi	0.003*	0.029*	0.035
	(0.001)	(0.017)	(0.027)
Dinter	−0.008**	0.006	0.003
	(0.004)	(0.006)	(0.007)
TD	−0.002	−0.002	0.001
	(0.003)	(0.003)	(0.003)
Local	−0.373*	−0.397	−1.248***
	(0.216)	(0.309)	(0.299)
Size	0.000	0.003*	0.005*
	(0.000)	(0.002)	(0.003)
Foreign	0.010***	0.001	−0.001
	(0.004)	(0.004)	(0.006)
HighSchool	−0.004	0.009**	0.001
	(0.003)	(0.004)	(0.003)
Rservices	0.237	−0.107	0.402
	(0.439)	(0.359)	(0.348)
Oservices	−0.034	0.082	0.211
	(0.257)	(0.334)	(0.322)
Constant	−2.096***	−2.779***	−1.901*
	(0.453)	(0.928)	(1.013)
Pseudo R^2	0.1619	0.0842	0.0765
Prob > chi^2	0.000	0.001	0.005
LR chi^2	285.64	65.08	57.48
Observations	1814	361	361

Note: Standard errors are expressed in parentheses, level of significance ***p < 0.01, **p < 0.05, *p < 0.10.

Source: Authors’ estimates on STATA from WBES-2016 data.

Table A2. Robustness check with controls for ‘Process Innovation’.

Variables	Egypt	Côte d’Ivoire	Cameroon
	InvPrc	InvPrc	InvPrc
RD	1.246**	−0.405	−16.873
	(0.518)	(1.085)	(4,663.716)
Fperso	1.100***	0.162	0.100
	(0.408)	(0.526)	(0.722)
Dlexploi	0.001	0.034*	0.042
	(0.003)	(0.018)	(0.088)
Dinter	−0.005	−0.003	−0.025**
	(0.006)	(0.008)	(0.011)
TD	−0.007*	−0.004	0.003
	(0.004)	(0.006)	(0.007)
Local	−0.496	−0.403	−1.065
	(0.377)	(0.572)	(0.784)
Size	0.000**	0.002	−0.018
	(0.000)	(0.003)	(0.016)
Foreign	0.011	−0.003	−1.194
	(0.006)	(0.007)	(65.957)
HighSchool	−0.009	0.008	0.002
	(0.005)	(0.007)	(0.009)
Rservices	0.493	−1.263	−1.154
	(0.677)	(0.876)	(0.802)
Oservices	0.121	0.313	−1.164
	(0.442)	(0.573)	(0.765)
Constant	−2.719***	−3.399***	−0.392
	(0.743)	(1.131)	(2.486)
Pseudo R^2	0.1619	0.0842	0.0765
Prob > chi^2	0.000	0.001	0.005
LR chi^2	285.64	65.08	57.48
Observations	1814	361	361

Note: Standard errors are expressed in parentheses, level of significance ***p < 0.01, **p < 0.05, *p < 0.10.

Source: Authors’ estimates on STATA from WBES-2016 data.

Table A3. Robustness check with controls for ‘Product and Process Innovation’.

Variables	Egypt	Côte d’Ivoire	Cameroon
	InvPrdPrc	InvPrdPrc	InvPrdPrc
RD	1.890***	1.740***	0.587
	(0.337)	(0.459)	(0.536)
Fperso	2.037***	0.875**	0.608
	(0.306)	(0.389)	(0.389)
Dlexploi	0.003	0.006	−0.023
	(0.002)	(0.019)	(0.042)
Dinter	−0.000	0.003	0.006
	(0.006)	(0.008)	(0.009)
TD	0.008	0.006	0.005
	(0.005)	(0.004)	(0.004)
Local	−1.063***	−0.368	−0.522
	(0.284)	(0.439)	(0.395)
Size	−0.000	0.004**	0.005**
	(0.000)	(0.002)	(0.003)
Foreign	0.009	0.005	0.009
	(0.006)	(0.005)	(0.007)
HighSchool	−0.013***	0.003	0.004
	(0.004)	(0.005)	(0.005)
Rservices	−0.915	−0.267	−0.103
	(1.043)	(0.524)	(0.522)
Oservices	−1.342**	−0.047	−0.028
	(0.549)	(0.466)	(0.431)
Constant	−3.425***	−3.110***	−2.207
	(0.767)	(1.124)	(1.530)
Pseudo R^2	0.1619	0.0842	0.0765
Prob > chi^2	0.000	0.001	0.005
LR chi^2	285.64	65.08	57.48
Observations	1814	361	361

Note: Standard errors are expressed in parentheses, level of significance ***p < 0.01, **p < 0.05, *p < 0.10.

Source: Authors’ estimates on STATA from WBES-2016 data.

Table A4. Robustness check with controls and subsidies only for Egypt.

Variables	Egypt
	InvPrd	InvPrc	InvPrdPrc
RD	1.547***	1.242**	1.882***
	(0.275)	(0.527)	(0.338)
Fperso	1.285***	1.124***	2.034***
	(0.227)	(0.408)	(0.306)
Dlexploi	0.002*	0.001	0.003
	(0.001)	(0.003)	(0.002)
Dinter	−0.007**	−0.006	−0.000
	(0.004)	(0.006)	(0.006)
TD	−0.002	−0.007*	0.008
	(0.003)	(0.004)	(0.005)
Local	−0.345	−0.532	−1.064***
	(0.217)	(0.379)	(0.285)
Size	0.000	0.000**	−0.000
	(0.000)	(0.000)	(0.000)
Foreign	0.010***	0.010	0.009
	(0.004)	(0.006)	(0.006)
HighSchool	−0.005	−0.009	−0.013***
	(0.003)	(0.005)	(0.004)
Rservices	0.227	0.516	−0.898
	(0.444)	(0.675)	(1.042)
Oservices	−0.032	0.123	−1.335**
	(0.258)	(0.441)	(0.548)
Subs	0.794*	−16.002	−0.065
	(0.460)	(2723.285)	(0.795)
Constant	−2.134***	−2.675***	−3.426***
	(0.454)	(0.744)	(0.769)
Pseudo R^2	0.165
Prob > chi^2	0.000
LR chi^2	291.13
Observations	1814

Note: Standard errors are expressed in parentheses, level of significance ***p < 0.01, **p < 0.05, *p < 0.10.

Source: Authors’ estimates on STATA from WBES-2016 data.

Table A5. Regression results with instrumental variables.

Variables	Egypt	Côte d’Ivoire	Cameroon
RD	0.833**	0.830	1.733
	(0.351)	(1.538)	(1.584)
Fperso	0.373***	0.218	−0.285
	(0.093)	(0.234)	(0.505)
Dlexploi	0.001**	0.009	−0.023
	(0.000)	(0.020)	(0.028)
Dinterne	−0.001	−0.003	−0.007
	(0.001)	(0.003)	(0.005)
TD	0.000	0.001	−0.001
	(0.001)	(0.002)	(0.002)
Constant	0.140	0.378	1.918**
	(0.096)	(1.037)	(0.891)
Observations	1173	106	102
R-squared	0.155	0.058	0
Pob. Durbin	0.5872	0.8934	0.3731
Prob. Wu-Hausman	0.5884	0.8973	0.3904

Note: Standard errors are expressed in parentheses, level of significance ***p < 0.01, **p < 0.05, *p < 0.10.

Source: Authors’ estimates on STATA from WBES-2016 data.

Table A6. Firms’ distribution by industry sector.

	Industry	Egypt	Cote d’Ivoire	Cameroon
Manufacturing	Extractives	5	0	0
	Food	199	30	20
	Tobacco	4	0	0
	Textiles	56	0	2
	Garments	76	14	7
	Leather	99	2	0
	Wood	40	11	8
	Paper	43	1	2
	Publishing, printing, and Recorded media	19	7	15
	Refined petroleum product	8	0	2
	Chemicals	124	9	6
	Plastics & rubber	123	0	6
	Non-metallic mineral products	129	1	0
	Basic metals	20	2	4
	Fabricated metal products	113	10	9
	Machinery and equipment (29–30)	60	2	5
	Electronics (31–32)	9	2	2
	Precision instruments	16	0	0
	Transport machines (34–35)	6	0	0
	Furniture	20	8	13
	Recycling	28	0	1
	Sub-total Manufactures	1193	106	102
Retail services	Retail	113	106	117
Other services	Wholesale	67	40	33
	IT	68	9	10
	Hotel and restaurants: section H	78	16	32
	Construction Section F:	95	20	20
	Services of motor vehicles	22	40	24
	Transport Section I: (60–64)	106	24	23
	Travel Agencies	49	0	0
	Real estate	19	0	0
	Sub-total Others services	508	138	153
	Total	1814	361	361

Source: Authors’ production based on WBES-2016 data.

Table A7. Descriptive statistics for Egypt.

Variables	Questions from Enterprises Survey	Obs.	Mean	S.D.	Min	Max
Innovation	The multinomial variable Innovation is constructed from 2 variables: h1 and h5 in the Enterprises Survey questionnaire. The question relating to these variables is: In the last three years has the establishment introduced new products/services (h1), new processes (h5)?	1814	0.215	0.659	0	3
R&D	In the last fiscal year, did the establishment spend on R&D (excluding market research)? This is a binary variable taking the value 1 if the company invests in R&D and 0 otherwise.	1814	0.076	0.264	0	1
Fperso	Has the establishment introduced formal training programs for permanent full-time employees in the last fiscal year? This variable is also binary taking the value 1 if the company carries out training for staff and 0 otherwise.	1814	0.210	0.407	0	1
Dlexploi	How many days did it take to get your operating license? Days are between 0 and 90.	1814	34.065	45.588	0	995
Dinterne	In the last fiscal year, what was the establishment’s share in the national sales? Sales shares are in percentage.	1814	90.277	24.080	0	100
TD	What is the percentage of the land occupied by the establishment (owned by the establishment)? Lands shares are in percentage.	1814	72.669	44.248	0	100
Local	Is the establishment located in the main business city? This is a binary variable taking the value 1 if the establishment is located in the main business city and 0 otherwise.	1814	0.658	0.474	0	1
Sector	The establishment industry sector. This is a categorical variable taking the value 1 for Manufacturing, 2 for Retail services and 3 for other services.	1814	1.639	0.896	1	3
Size	What is the number of Permanent Full-Time Employees at end of 3 fiscal years ago? This is a numerical variable.	1814	170.793	618.168	0	15000
Foreign	What is the percentage owned by private foreign individuals, companies or organizations? This is a numerical variable.	1814	4.448	18.53	0	100
HighSchool	This is a quantitative variable indicating the % Of full time workers completed high school within the establishment.	1814	67.998	32.302	0	100
Subs	Does the establishment received direct subsidies (including wage, investment, other) over last 3 years? This is a binary variable taking the value 1 if the establishment received direct subsidies and 0 otherwise	1814	0.02	0.141	0	1

Source: Authors’ production based on WBES-2016 data.

Table A8. Descriptive statistics for Côte d’Ivoire.

Variables	Questions from Enterprises Survey	Obs.	Mean	S.D.	Min	Max
Innovation	The multinomial variable Innovation is constructed from 2 variables: h1 and h5 in the Enterprises Survey questionnaire. The question relating to these variables is: In the last three years has the establishment introduced new products/services (h1), new processes (h5)?	361	0.706	1.01	0	3
R&D	In the last fiscal year, did the establishment spend on R&D (excluding market research)? This is a binary variable taking the value 1 if the company invests in R&D and 0 otherwise.	361	0.1	0.3	0	1
Fperso	Has the establishment introduced formal training programs for permanent full-time employees in the last fiscal year? This variable is also binary taking the value 1 if the company carries out training for staff and 0 otherwise.	361	0.424	0.495	0	1
Dlexploi	How many days did it take to get your operating license? Days are between 0 and 90.	361	42.008	20.336	0	365
Dinterne	In the last fiscal year, what was the establishment’s share in the national sales? Sales shares are in percentage.	361	90.224	25.685	0	100
TD	What is the percentage of the land occupied by the establishment (owned by the establishment)? Lands shares are in percentage.	361	24.413	42.154	0	100
Local	Is the establishment located in the main business city? This is a binary variable taking the value 1 if the establishment is located in the main business city and 0 otherwise.	361	0.681	0.467	0	1
Sector	The establishment industry sector. This is a categorical variable taking the value 1 for Manufacturing, 2 for Retail services and 3 for other services.	361	2.114	0.831	1	3
Size	What is the number of Permanent Full-Time Employees at end of 3 fiscal years ago? This is a numerical variable.	361	43.102	115.294	0	1170
HighSchool	This is a quantitative variable indicating the % Of full time workers completed high school within the establishment.	361	28.346	36.484	0	100
Foreign	What is the percentage owned by private foreign individuals, companies or organizations? This is a numerical variable.	361	20.327	37.736	0	100

Source: Authors’ production based on WBES-2016 data.

Table A9. Descriptive statistics for Cameroon.

Variables	Questions from Enterprises Survey	Obs.	Mean	S.D.	Min	Max
Innovation	The multinomial variable Innovation is constructed from 2 variables: h1 and h5 in the Enterprises Survey questionnaire. The question relating to these variables is: In the last three years has the establishment introduced new products/services (h1), new processes (h5)?	361	0.693	0.973	0	3
R&D	In the last fiscal year, did the establishment spend on R&D (excluding market research)? This is a binary variable taking the value 1 if the company invests in R&D and 0 otherwise.	361	0.102	0.304	0	1
Fperso	Has the establishment introduced formal training programs for permanent full-time employees in the last fiscal year? This variable is also binary taking the value 1 if the company carries out training for staff and 0 otherwise.	361	0.335	0.473	0	1
Dlexploi	How many days did it take to get your operating license? Days are between 0 and 90.	361	27.008	4.975	0	90
Dinterne	In the last fiscal year, what was the establishment’s share in the national sales? Sales shares are in percentage.	361	91.186	21.608	0	100
TD	What is the percentage of the land occupied by the establishment (owned by the establishment)? Lands shares are in percentage.	361	28.371	44.151	0	100
Local	Is the establishment located in the main business city? This is a binary variable taking the value 1 if the establishment is located in the main business city and 0 otherwise.	361	0.396	0.49	0	1
Sector	The establishment industry sector. This is a categorical variable taking the value 1 for Manufacturing, 2 for Retail services and 3 for other services.	361	2.1108	0.8158	1	3
Size	What is the number of Permanent Full-Time Employees at end of 3 fiscal years ago? This is a numerical variable.	361	31.378	74.615	0	1000
HighSchool	This is a quantitative variable indicating the % Of full time workers completed high school within the establishment.	361	36.2133	39.87956	0	100
Foreign	What is the percentage owned by private foreign individuals, companies or organizations? This is a numerical variable.	361	5.565	21.063	0	100

Source: Authors’ production based on WBES-2016 data.

Table A10. Correlation matrix.

Panel A: Correlation matrix for Egypt
	inno	RD	fperso	Dlexploi	dinterne	td	local	size	foreign	HighSchool	sector	subs
inno	1.00
RD	0.30***	1.00
fperso	0.28***	0.32***	1.00
dlexploi	0.05*	−0.00	−0.01	1.00
dinterne	−0.10***	−0.13***	−0.13***	0.01	1.00
td	0.06**	0.08***	0.09***	−0.03	−0.02	1.00
local	−0.13***	−0.01	−0.01	0.00	0.05*	−0.13***	1.00
size	0.13***	0.21***	0.21***	−0.02	−0.21***	0.09***	−0.04	1.00
foreign	0.09***	0.05	0.07**	−0.02	−0.19***	−0.06*	−0.04	0.10***	1.00
HighSchool	−0.04	0.12***	0.17***	−0.03	−0.05*	−0.02	0.18***	0.02	−0.00	1.00
sector	−0.10***	−0.06*	−0.01	−0.03	0.06**	−0.32***	0.25***	−0.08***	−0.05*	0.25***	1.00
subs	0.04	0.05*	0.12***	0.02	−0.11***	0.01	−0.04	0.10***	0.04	0.06*	−0.02	1.00

Source: Authors’ production based on WBES-2016 data.

Table A11. Correlation matrix.

Correlation matrix for Côte d’Ivoire
	inno	RD	fperso	dlexploi	dinterne	td	local	size	foreign	HighSchool	sector
inno	1.00
RD	0.23***	1.00
fperso	0.18***	0.11*	1.00
dlexploi	0.03	−0.08	0.04	1.00
dinterne	−0.03	−0.03	−0.04	0.02	1.00
td	0.10	0.06	0.14**	−0.06	0.05	1.00
local	−0.03	0.03	−0.09	0.01	0.09	−0.23***	1.00
size	0.20***	0.14**	0.12*	−0.02	−0.15**	0.11*	0.10	1.00
foreign	0.05	−0.04	0.04	0.00	−0.15**	−0.12*	0.06	0.11*	1.00
HighSchool	0.12*	0.10	0.22***	−0.00	−0.11*	−0.00	0.26***	0.15**	0.11*	1.00
sector	0.02	−0.05	0.16**	0.06	0.09	−0.01	0.06	−0.09	−0.04	0.18***	1.00

Source: Authors’ production based on WBES-2016 data.

Table A12. Correlation matrix.

Correlation matrix for Cameroon
	inno	RD	fperso	dlexploi	dinterne	td	local	size	foreign	HighSchool	sector
inno	1.00
RD	0.10	1.00
fperso	0.15**	0.21***	1.00
dlexploi	−0.01	−0.00	0.06	1.00
dinterne	−0.02	−0.03	−0.06	0.01	1.00
td	0.11*	0.06	0.19***	−0.03	−0.15**	1.00
local	−0.11*	0.08	−0.10	0.02	−0.16**	−0.01	1.00
size	0.14**	0.24***	0.16**	−0.01	−0.20***	0.23***	0.16**	1.00
foreign	0.09	0.06	0.11*	0.01	−0.08	0.02	0.01	0.12*	1.00
HighSchool	0.00	0.04	0.04	−0.10	−0.06	−0.08	0.25***	0.06	0.01	1.00
sector	−0.03	−0.07	0.06	−0.04	0.09	−0.02	0.01	−0.14**	0.01	0.01	1.00

Source: Authors’ production based on WBES-2016 data.