Adoption of twin transition technologies in developing countries: a bivariate analysis

doi:10.21203/rs.3.rs-4968284/v1

Industrial firms worldwide face two major technological challenges: digitalizing their processes and increasing the environmental sustainability of their production. Both challenges can be addressed separately or in an integrated way. This paper empirically examines the extent to which the adoption of digital and green technologies is done in a complementary way in developing countries’ industrial firms and identify certain characteristics that tend to be present when such complementarity takes place. The analysis is based on a firm-level survey conducted by UNIDO in Ghana, Thailand, and Vietnam. The results suggest that adopting green and digital technologies are interrelated and complementary. Larger firms, more innovative firms and firms participating in global value chains are more likely to jointly adopt green and digital technologies.

JEL: 014, 033, O25, Q55

twin transition

Fourth Industrial Revolution

green technologies

technology adoption

industrial development

Firms worldwide face two major technological challenges: going digital and becoming environmentally sustainable. In this context, it becomes increasingly imperative for firms to adopt innovations that simultaneously address both challenges to achieve what the literature has commonly labelled as the “twin transition”. But, what are the skills, capabilities and framework conditions required to achieve this transition? And, to what extent do both transformations happen together? These are empirical questions that have not yet been fully addressed by the literature due to lack of empirical data on the adoption of digital and green technologies by firms.

Despite limited data, a growing literature has explored the relationship between the adoption of digital technologies -containing ICT and Fourth Industrial Revolution technologies (4IR, onwards) and green technologies - aiming to enhance efficiency, reducing material usage and optimising production and consumption processes, at the firm level. Within this literature, some studies have examined the potential of digital technologies to enable sustainable processes. These studies underscore that digital technologies play a crucial role in promoting environmental sustainability across various levels of interaction (Mickoleit, 2010; Williams, 2011), and hence, their adoption might contribute to firms finding a more sustainable production path. Other studies show that digital transformation increases the propensity to adopt eco-innovations in developed countries (Ferreira et al., 2023; Horbach, 2023; Montresor and Vezzani, 2023). This process, however, is not automatic and typically requires supportive measures including regulations and government incentives (Beier et al., 2022).

Another strand of the empirical literature, which has been less explored, has attempted to study the joint adoption of green and digital technologies to better understand the complementary-simultaneous nature of these technologies (Ardito, 2023; Ardito et al., 2021; Denicolai et al., 2021). Studying the synergies of these technologies is a key factor in the face of the twin transition, as the adoption objectives of both technologies may not be necessarily aligned (Wen et al., 2022). While one is mostly profitability-oriented, the other is driven by sustainability purposes (Ferreira et al., 2023) and competes for limited organizational resources (Ardito et al., 2021). These studies have found mixed results, providing insights into the impact of adopting both technologies on firms’ ability to pursue positive trajectories during the twin transition.

Studies on the joint adoption of green and digital technologies have primarily concentrated on firms in developed countries. For developing countries, this relationship remains as unexplored territory. To the best of our knowledge, there is no empirical study that investigates the synergies, in terms of complementarity/substitutability, related to the adoption of twin transition technologies in the context of developing countries. Two main reasons explain this research gap. First, the still limited adoption of advanced digital technologies in developing countries (Lorenz and Kraemer-Mbula, 2022). Second, the lack of data sources on the adoption of both types of technology by developing countries’ firms.

This paper aims to fill this gap by empirically examining whether the adoption of digital and green innovations complement each other, and what are the main factors that explain the joint adoption of these technologies. This is done through an econometric bivariate analysis that contemplates the correlation between the adoption of green and digital technologies due to the existence of unobserved characteristics that determine the adoption of both on a sample of 658 manufacturing firms operating in the manufacturing sector from Ghana, Thailand, and Viet Nam. The data is sourced from the United Nations Industrial Development Organization (UNIDO) Survey “Adoption of Digital Technologies by Industrial Firms” which contains specific modules related to energy and sustainability² and current and expected use of digital technologies. ³

Our results suggest that the adoption of green and digital technologies is interrelated and complementary. Some of the salient firm characteristics associated with a joint adoption of these technologies include size, previous investments on capabilities, innovativeness and participation in global value chains (GVCs). Drawing on the insights form prior research; the results affirm a prevailing inclination towards technological complementarity rather than substitution. This pattern is consistent with the earlier findings introduced by Kesidou and Ri (2021) and Kren and Lawless (2023), who played a pivotal role in providing initial insights into the interdependencies of green and digital technologies adoption.

The findings reveal that the firms refraining from adopting both green and digital technologies are primarily smaller in scale, constituting a significant segment of the sampled firms. To inclusively integrate these firms into the twin transition path, our results underscore the critical need to remove barriers impeding innovation activities, promote heightened investment in R&D, and encourage active engagement in global value chains that facilitate the transfer of digital technologies while fostering sustainable practices.

The paper is structured as follows. Section 2 presents the literature review. Section 3 focuses on data sources and introduces the model specification. Section 4 presents the empirical results. Section 5 discusses the implications of the results. Finally, section 6 concludes.

2.1 Firm’s adoption of twin transition technologies: towards a synergistic approach

A large body of literature that has analysed technology diffusion across firms. Broadly, this literature distinguishes four types of theoretical models to explain technology diffusion (Geroski, 2000; Karshenas and Stoneman, 1993). Ranks (or probit) models relate firm’s decision to adopt technologies to firm’s characteristics such as firm size, firm ownership, labour force skills, output growth rate, and marketing expenditure, among others (Bartoloni and Baussola, 2001). Epidemic models connect technology diffusion with firms learning due to the reduction of cost and risk as a consequence of the availability of information. Stock models associate the reduction of profits from technology adoption as a result of an increase in the number of adopters. Finally, order models connect the benefits of technologies diffusion with the position of the firm in the adoption provoking a competence among firms for being early adopters rather than follower (Fusaro, 2009). This paper builds on some of this literature to analyse the adoption of green and digital technologies focusing mainly on rank effects.

In second place, and more specifically, this paper draws on various studies conducted to identify the factors that influence firms adopting digital technology building on firm-level surveys specifically designed to capture the diverse range of technologies (or generations of technologies) used by firms in different countries (Albrieu et al., 2019; Avenyo et al., 2022; Calza et al., 2023; Cirera et al., 2021a; Cirera et al., 2021b; Ferraz et al., 2020). In general, these studies have shown that firms’ adoption of digital technologies, especially the most advanced, in developing countries is still scarce (Lorenz and Kraemer-Mbula, 2022). Some of the factors identified as relevant for adoption are include GVCs participation (Delera et al., 2022), foreign ownership, export orientation and past innovativeness (Avenyo et al., 2022). On the opposite side, lack of capabilities such as managerial capabilities and human capital (Cirera et al., 2021b) or lack of skills in Science Technology, Engineering and Mathematics (STEM) (Ferraz et al., 2020) are some of the main factors limiting the adoption of digital technologies.

Furthermore, from the perspective of the adoption of innovation to reduce environmental impacts, some recent studies have examined the main determinants of adoption in developing countries (Fernández et al., 2021; Han and Chen, 2021; Torrecillas et al., 2023; Yurdakul and Kazan, 2023). These studies highlight specific technological factors (e.g.: R&D intensity, investment intensity, external/internal R&D, etc.), market factors (e.g.: customer demand, reduction of energy/material, new markets, etc.), firm characteristics (e.g.: knowledge transfer mechanism, networking, etc.) and institutional factors (e.g.: regulations, subsidies, norms/standards etc.) as drivers of eco-innovation.

In third place, this paper build on a large body of research dedicated to investigating the determinants of the adoption of interconnected technologies (Stoneman and Kwon, 1994), which suggests that technological innovations hardly function in isolation (Rosenberg, 1979). According to this literature, technologies may either complement or substitute each other as innovation strategies (Kamutando and Tregenna, 2023).

Finally, this paper also builds on literature related to the joint adoption of digital and green technologies, typically referred as “twin transition”. The potential synergies between the adoption of both technologies have been described in the field of innovation studies as a “very convenient marriage”⁴. Studying the synergies of these technologies is of prime relevance, as the adoption objectives of both technologies may not be necessary aligned (Wen et al., 2022). While one is mainly profitability-oriented, the other is driven by sustainability purposes (Ferreira et al., 2023) and there may be competition for limited organisational resources (Ardito et al., 2021).

The empirical evidence about possible synergies among both technologies and their impact on firms’ performance is still scarce and results are mixed. On the one hand, Antonioli et al. (2014) find the existence of complementarities between ICT and environmental innovation in the economic performance of manufacturing companies in the northeast of Italy. The authors claim that the impact of the joint adoption of ICT and environmental innovations on a firm’s performance is sector-specific and innovation-specific and the simple combination of technologies does not have a direct impact on economic performance. Complementarities are found between ICT and certain environmental innovations linked to organisational changes to meet green-certificate requirements (e.g. EMAS, ISO14001) among firms that belong to the most polluting sectors and are subject to Emission Trading System (ETS) regulation. In contrast, less regulated firms show higher gains in productivity by adopting green or ICT separately. In a similar vein, Upadhayay et al. (2024) find that both digital and green servitization processes enable firms to develop organisational and adaptation capabilities within the firms that enhance a learning-oriented culture that supports innovation activities for a broad sample of companies in OECD countries and beyond.

On the other hand, a recent study for Canadian SMEs finds that being both digitally and environmentally oriented⁵ may not be beneficial for firms’ innovativeness (Ardito et al., 2021). The authors find evidence that digitalization and sustainability orientation have a positive effect on SMEs innovation when they are considered as distinct strategies. A related study for a sample of 14,000 firms from the Flash Euro-barometer 486 survey finds that the interplay among digitalization and sustainability practices is not significant to enhance both social and environmental innovations and it is negatively associated with the launch of environmental and social innovations separately (Ardito, 2023). These results suggest the existence of the non-complementarity in the adoption of both digitalisation and sustainability practices on sustainable innovation performance.

Beyond the effect of green and digital technologies on innovation, Denicolai et al. (2021) focus on examining the impact of the interaction between artificial intelligence (AI) readiness⁶ and environmental sustainability readiness⁷ on Italian SMEs internationalisation. The study finds that digitalization and sustainability are interrelated and positively correlated, but they become competing assets when international performance⁸ is taken into consideration. This would be explained by the firm’s limitations on resources and capabilities. Inserting virtuously abroad only occurs when firms focus on either digitalization or sustainability.

In addition to the works on the complementary adoption of green and digital technologies on firm performance, other works have focused on the existence of complementarities-in-use. In this regard,

Kesidou and Ri (2021) study the complementarities-in use among digital⁹ and net-zero¹⁰ practices based on the Business Futures Survey for over 1000 UK firms. The authors find only low intense synergies between digital technologies and net-zero practices. These are stronger only in the case of the joint use of digital systems to improve customer experience with practices aimed at reducing, measuring and reporting environmental impacts, and in the joint use of digital technologies (augmented and virtual reality, artificial intelligence (AI) and machine learning) and investment in environmental R&D and organisational net zero practices. Another study carried out by Veugelers et al (2023) using the Survey on Investment and Investment Finance (EIBIS) from the European Investment Bank examines the characteristics of firms that jointly adopt twin transition technologies, understood as those firms that have implemented at least one advanced digital technology and have invested to address the impacts of physical and transition risks from climate change. The authors find that firms that invest in both digital and green are more likely to be innovation leaders and expand and train their workforce. In a similar vein, Kren and Lawless (2023) examine the effect of the joint impact of business climate investment and digital adaptation on Irish companies in manufacturing, information and communication and other internationally traded services sectors with more than 10 employees. The research finds that larger companies, that export, that are more productive and, that invest in R&D are more likely to have both a digital and climate change plan.

In summary, the existing literature has shed light on the interaction between green and digital transitions at the firm level, but the results in terms of their interrelationship (complementarity/substitutability) are still mixed and the few existing empirical works on this matter focus on a limited group of advanced economies. There are still gaps in the empirical literature at the firm level, especially in the context of developing countries. This paper aims at addressing this gap by examining whether the adoption of digital and green innovations complement each other, and to what extent this complementarity is related to observable (e.g.: size, age, participation in GVC, type of ownership, among others) or unobservable (e.g.: environmental regulations, subsidies for technology adoption) characteristics.

3.1 Model specification

This paper uses multivariate analysis, a technique vastly used in the modelling of correlated firm decisions concerning innovation options (Agwu et al., 2020; Freitas et al., 2011; van Leeuwen and Mohnen, 2017; Ziegler and Seijas Nogareda, 2009). More specifically, we examine the correlation of firms’ decisions concerning innovation options employing a bivariate probit model with two dependent observed binary variables: adoption of digital technologies ($\:{Y}_{i,digital}^{*}$) and adoption of green technologies ($\:{Y}_{i,green}^{*}$). The main assumption of the model is that there are unobserved firm characteristics affecting the choice of adopting both technologies. Consequently, the model recognises the correlation in the error terms by simultaneously estimating a set of probit models as follows:

$\:{Y}_{i,digital}^{*}={\beta\:}_{i,digital}^{{\prime\:}}{X}_{i,digital}+{u}_{digital}$ $\:{Y}_{i,digital}=\left\{\begin{array}{c}1\\\:0\end{array}\begin{array}{c}\:if\\\:\:if\end{array}\begin{array}{c}{\:Y}_{i,digital}^{*}>0\\\:{\:Y}_{i,digital}^{*}\le\:0\end{array}\right.$ $\:{Y}_{i,green}^{*}={\beta\:}_{i,green}^{{\prime\:}}{X}_{i,green}+{u}_{green}$ $\:{Y}_{i,green}=\left\{\begin{array}{c}1\\\:0\end{array}\begin{array}{c}\:if\\\:\:if\end{array}\begin{array}{c}{\:Y}_{i,green}^{*}>0\\\:{\:Y}_{i,green}^{*}\le\:0\end{array}\right.$ $\:\left(1\right)$

where u are distributed as standard bivariate normal variables u ∼ N (0, 0, 1, 1, ρ).

Under this specification $\:{Y}_{i,digital}^{*}$ and $\:{Y}_{i,green}^{*}$ are the latent variables that are unobserved and represent the propensity or inclination of a firm to adapt or not green or digital technologies. $\:{Y}_{i,digital}$ and $\:{Y}_{i,green}$ are the observed binary dependent variables that depend on whether the latent variables surpass a certain threshold. The subscript $\:i$ represents the individual firms. $\:{X}_{i,digital}$ and $\:{X}_{i,green}$ constitute the vector observed characteristics of the firms, including size, country, type of product, participation in GVCs, innovativeness, previous investments on capabilities, etc. (see next section for the details).

The model parameter estimates ($\:{\beta\:}_{i,digital}^{{\prime\:}}$, $\:{\beta\:}_{i,green}^{{\prime\:}}$ and ρ ¹¹ are obtained by the maximization of log-likelihood. The log-likelihood of this model is obtained estimating four joint probabilities P_0,0, P_0,1, P_1,0 and P_1,1, where the subscript indicates if the firms adopt digital technologies (first subscript) and/or green technologies (second subscript). For example, P_0,1 calculates the bi-variate predicted probability $\:\text{P}\text{r}({Y}_{digital}=0,\:{Y}_{green}=1)$.

If a significant non-zero correlation coefficient (ρ) between the error terms of the two equations is observed, this indicates that the adoption of green and digital technologies is intermediated by unobserved firm characteristics (e.g., financial constraints and policy instruments such as environmental regulations, subsidies for technology adoption, represented in the correlated error terms), and hence are not independent. Such a finding is interpreted as complementarity between digital and green innovation.

Complementarity can result from the existence of both observed and unobserved attributes. For instance, if we find that both green and digital innovation are easier for large firms, then we will observe complementarity due to the observed size characteristic. Due to the existence of unobserved complementarity effects, it would be incorrect to treat the adoption of digital and green technologies as an independent decision. Consequently, the adoption of these technologies cannot be estimated using two separate probit models as they only capture the complementarity derived from the observed firm’s characteristics.

To study the existence of complementarities/substitution in the adoption of green and digital technologies we estimate a “complementarity statistic” that measures the difference between the probability of being both green and digital and the product of the probabilities of adopting green and digital technologies:

$$\:\text{Pr}\left(Digital=1,Green=1\right)-\text{Pr}\left(Digital=1\right)*\text{Pr}\left(Green=1\right)\:\:\:\:\left(2\right)$$

In Eq. 2, the bivariate probability of jointly adopting green and digital technologies (either observed in the raw data or as estimated in the bivariate probit model of Eq. 1) includes the effect of complementarities, while the product of the univariate (marginal) predicted probabilities of adopting digital and green technologies (again, either observed or estimated) represents the probability of joint adoption exclusive of any complementarities. If the difference is positive, the relationship between green and digital adoption must be considered complementary, and if negative, substitution rules.

Previous studies on innovation adoption have used a similar approach to Eq. 2, but instead used predicted conditional probabilities derived from the multivariate analysis to calculate “treatment effects” (Agwu et al., 2020; Calia and Ferrante, 2013). Contrary to the complementarity statistic in Eq. 2, such treatment effects have a direction (the effect of green on digital, and the effect of digital on green). Such a directional approach seems more tailored for a context where causality can be identified, which is not case for our data. Therefore, we prefer the related approach of Eq. 2, which does not assume any (causal) direction of the complementarity/substitution effect. In line with Agwu et al (2020), we assume that the costs and benefits of the adoption of each type of innovation are not considered, and therefore the existence of complementarity is not defined in terms of firm performance but in terms of the compatibility of innovations in use.

3.2 Data

The analysis of this paper makes use of the UNIDO Survey “Adoption of Digital Technologies by Industrial Firms” conducted in Ghana, Thailand, and Vietnam in 2019. This survey provides data on the adoption of digital technologies by 658 manufacturing firms with more than 20 employees. It comprises five modules¹²: company profile, current and expected use of digital technologies, employment and skills, location, and energy and sustainability.

3.2.1 Dependent variables

Following Eq. 1, our model comprises two dependent variables. The first one (“green”) is a dummy variable that captures the environmental profile of the firms. It takes the value 1 if the firm declares to use instruments of life-cycle assessment (e.g. EU ecolabel, ISO 14020) and/or declares to have a certified energy management standard. These questions are consistent with the conceptual framework for measurement and data collection on green innovation for developing countries proposed by Grazzi, Sasso, and Kemp (2019) and also adopted by Beier et al. (2022). In our sample, 37% of firms can be considered as green (see Fig. 1).

The second dependent variable (“digital”) is a dummy variable that captures the digital profile of the firm. It takes value 1 if the firm reports adopting the latest vintages of digital production technologies (Generations 2 to 4) included in the survey questionnaire (see Table 1 for the details on how the survey organises the digital production technologies that can be used by firms). In our sample, 18% of firms can be considered as digital (see Fig. 1).

Table 1

Digital Production Technologies (DPTs) generations
Generation 4	Smart production: DPTs allow for fully integrated, connected, and smart production processes, where information flows across operations and generates real- time feedback to support decision-making
Generation 3	Integrated production: DPTs integrated across different activities and functions, allowing for the interconnection of the whole production process
Generation 2	Lean production: DPTs involve and connect different functions and activities within the firm.
Generation 1	Rigid production: DPTs limited to a specific purpose in a specific function
Generation 0	Analog production: No used throughout the whole production process.
Source: extracted from UNIDO (2019)

Figure 2 displays the green and digital adoption by firm’s size and sector. Overall, the share of firms adopting digital technologies is lower than that adopting green technologies, across all groups displayed. When it comes to digital adoption, size seems to be more important than the sector of operation. Large firms in medium-high- and high-technology (MH&HT) sectors are by far the ones showing more adoption of digital technologies (more than 20 percentage points higher than the average). Digital adoption across SMEs falls below the average and is not significantly different between the two broad technology sectors shown in the figure. In the case of green technologies, the opposite seems to happen: the sector of operation seems to be more important than the size when it comes to the adoption of these technologies. Firms in MH&HT sectors tend to show significant larger shares of adoption of green technologies than those in medium-low- and low-technology (ML&LT) sectors. As a matter of fact, SMEs in sectors with higher technology intensity show larger rates of adoption than large firms in sectors with lower technology intensity (49% versus 34%).

3.2.2 Covariates

The covariates of the model ($\:{X}_{i,digital}$ and $\:{X}_{i,green}$) have been selected based on the literature reviewed in Section 2 as follows.

To begin, we include variables linked to inherent characteristic of the firm such as the number of employees, age, and type of ownership. For the number of employees, we include a dummy variable to distinguish larger firms from the rest. The threshold used in this paper is 100 employees. Firms with less than this number of employees are considered SMEs. Being a large firm is a relevant characteristic in adoption models due to the fact they have higher access to financial resources required to invest in new technologies and better position to hire or train in-house workers with the skills needed for adopting and using the new technologies (Lorenz and Kraemer-Mbula, 2022). Firm’s age is measured as the years in operation an reflects it can be considered as an approximation for its technological experience (Grazzi and Jung, 2019). For the type of ownership, we include a dummy variable indicating whether the firm is foreign-owned as these firms are more likely to be engaged in international networks and connected with other affiliates overseas. For example, Wang et al. (2023) find that different types of ownership identity have distinct effects on the propensity to adopt environmental management systems (EMS) such as ISO 14001.

Furthermore, we include a dummy variable related to whether the firm has invested in capabilities such as R&D, machinery and/or training. The different sources of knowledge investment are considered together as R&D in developing countries tends to be low and does not constitute the main knowledge investment (Cirera, Lage, and Sabetti, 2016). Additionally, information about technological innovation by incorporating a variable whether the firm has carried out a product/process innovation, new to the firm of the market.

Given the positive impact of participation in GVCs on the adoption of advanced digital technologies (Delera et al., 2022); we added a variable capturing whether the firm produces for a foreign actor or a multinational. In this regard, Agostino et al. (2023) find a positive association between firms' participation in GVCs and the adoption of energy-related sustainable practices, although the effects are heterogeneous, as this association is attenuated by poorer institutional frameworks, less human and financial capital, smaller size and fewer years of operation. Additionally, a recent article from Siewers et al. (2024) finds that GVCs are more prone to adopt cleaner production practices, face stricter external constraints—such as environmental regulation and standards—and, in some cases, also monitor CO2 emissions.

Also, we incorporate a variable to capture the level of high-skilled workers, that is, workers with a background in science, technology, engineering, and mathematics (STEM). A highly skilled workforce may be positively associated with the adoption of more advanced technologies. The importance of digital skills in hiring new employees is taken into account in the model. Several authors have pointed out the correlation between digital skills and the twin transition. For instance, a study conducted by Santoalha et al. (2021) at the European regional level found that possessing digital skills has a positive impact on regions’ ability to specialise in new technological domains, particularly in green activities. Additionally, aligned with the literature that considers a prerequisite building digital infrastructure a prerequisite for adopting digital technologies (World bank, 2024), we include a dummy variable indicating whether the enterprise has access to high-speed internet connection.

Finally, we include sectoral and countries effects through dummies. We expect significant gaps in the probability of adoption between countries and sectors due to the existence of different institutional frameworks and technological opportunities, among others (Delera et al., 2022).

Table 2 provides details about the main descriptive statistics of all variables used in the analysis.

Table 2

Descriptive analysis
Variable	Description	Obs	Mean	Std. dev.	Min	Max
Digital	= 1 if firm adopts Gen III-IV technologies	658	0.179	0.384	0	1
Green	= 1 if the firm adopts life cycle and/or energy management instruments	658	0.368	0.483	0	1
Age (in logs)*	= Years in operation (logs)	658	2.674	0.666	0.693	4.779
Large*	=firms with more than 99 employees	658	0.488	0.500	0	1
Foreign Ownership*	=firms with at least 10% foreign ownership	658	0.415	0.493	0	1
Internet Speed*	= 1 if firm has an internet connection of more than 30 Mibt/s	658	0.429	0.495	0	1
GVC* Human Capital	= 1 if firm either: export intermediates. or whose import and export shares are higher than 10%, or that are importers or exporters and are also outsourced from = Share of workers with STEM background	658 658	0.467 0.125	0.499 0.174	0 0	1 1
Digital Skills	= 1 for firms that consider important digital skills when hiring a new employee	658	0.283	0.451	0	1
Investment Capabilities*	= 1 if the firm made any investment in (R&D), training or machinery	658	0.764	0.425	0	1
Technological Innovation*	= 1 if the firm has carried out a product/process innovation, new to the firm of the market.	658	0.605	0.489	0	1
Finished Goods	= 1 if the firm produces finished goods	658	0.764	0.425	0	1
Vietnam	=if firm is from Vietnam	658	0.397	0.490	0	1
Ghana	= 1 if the firm is from Ghana.	658	0.299	0.458	0	1
Thailand	= 1 if the firm is from Thailand.	658	0.304	0.460	0	1
Food	= 1 if firm operates in Food and beverages sector	658	0.278	0.448	0	1
Textiles	= 1 if firm operates in Textile and apparels sector	658	0.187	0.390	0	1
Information	= 1 if the firm operates in Electronic and ICT sector	658	0.166	0.372	0	1
Automotive	= 1 if firm operates in Automotive and auto-parts sector	658	0.193	0.395	0	1
Metal	= 1 if firm operates in metals sector	658	0.061	0.239	0	1
Note: *-The value of the statistics coincides with Delera et al., 2022, see Table 4.

4.1 Preliminary correlations

We start our analysis by looking at simple correlations between our variables of interest¹³. Table 3 shows the sample distribution of the adoption of advanced digital production technologies and the adoption of green innovations among enterprises. Most firms (56%) are not adopters of any of the two technologies. More than one-fourth of the sample adopts only green technologies, whereas firms adopting only ADP technologies are only 7.6%. Firms adopting both green and digital technologies stand at around 10% of the sample¹⁴.

Table 3

Sampling distribution of the adoption of advanced digital production technologies and the adoption of green innovations
Category	Description	Firms	Share
Green-Digital (1,1)	This category comprises those firms that adopt Generation III and Generation IV innovations, as well as green innovations measured by the adoption of life cycle assessment and/or energy management practices.	68	10.3%
Only Green (1,0)	This category comprises those firms that adopt green innovations by the adoption of life cycle assessment or energy management practices but they adopt technologies from Generation I and Generation II.	174	26.4%
Only Digital (0,1)	This category comprises firms that adopt Generation III and Generation IV innovations, but do not adopt any green innovation.	50	7.6%
Neither Green nor Digital (0,0)	This category comprises firms that adopt Generation I and Generation II and do not adopt any green innovation.	366	55.6%

Figure 3 provides more disaggregated information on the distribution of firms adopting digital and green technologies and the complementarity index according to firm characteristics. The difference between the probabilities of green and digital adoption is in the percentage of firms adopting green and those adopting digital. Positive values mean the likelihood of adopting green is higher than that of digital in our sample. Our estimates by firm characteristics confirm that, except in Ghana and the metal sector, green technology adoption is more likely than digital technology adoption in all categories considered. In addition, the statistical assessment of the Complementarity Index confirms that the joint probability of adopting green and digital technologies is higher than the product of the univariate predicted adoption probabilities (see Eq. 2). This is simply the difference between the share of firms adopting both green and digital and the product of the separate green and digital shares. Positive values of the complementarity index indicate that these technologies are more likely to be adopted together than separately, thus suggesting complementarity in adoption. Firms in sectors such as information and communication and motor vehicles are more likely to complement green-digital technologies. In addition, enterprises producing finished goods have a higher proportion of firms complementing both technologies. Furthermore, green and digital technologies are more likely to be used together in enterprises with at least 50% foreign ownership. Finally, firms adopting green and digital technologies are more likely to introduce new product innovations and invest in training.

In summary, the descriptive statistics presented in this section suggest that there is complementarity between the adoption of green and digital technologies. In the next section, we will try to confirm this preliminary evidence using econometric techniques and we will also examine to what extent this complementarity is determined by observable (e.g.: size, age, participation in GVC, type of ownership, among others) or by unobservable (e.g.: environmental regulations, subsidies for technology adoption) characteristics.

4.2 Regression Results

4.2.1 Determinants of the adoption of green and digital technologies

Our results shed light on whether firms that adopt twin transition technologies do so in a complementary or substitutive manner and which firm characteristics - both observable and unobservable - explain this complementarity. First, we model the prevalence of green and digital technologies adoption separately (univariate probit model (UVP)) (see columns 1 and 2, Table 4) and jointly (bivariate probit model (BVP)) (see columns 3 and 4, Table 4). The UVP estimates ignore the possible correlation between the unobservable characteristics that affect the adoption of green and digital technologies. The results of this model indicate that observable factors such as firm size, producing final goods, belonging to a GVC, being innovative, investing in capabilities are significant determinants of the adoption of green and digital technologies. After controlling for firm characteristics, Vietnamese enterprises are more likely to adopt green and digital technologies than Ghanaian and Thai companies.

The fact that the estimated ρ is around 0.177¹⁵ for both models and statistically significant at 5% in our BVP model and of a magnitude comparable to the correlation coefficient reported in Table A 1 of the Annex¹⁶ confirms that the adoption of green and digital technologies are interrelated and the use of a BVP is a better option than the use of separate probits. BVP estimation confirms that the errors are correlated and therefore part of the complementarity between green and digital technology adoption can be explained by the existence of some unobserved firm-level heterogeneities related to both dependent variables. This means that any positive change in the unobserved factors that increases the likelihood of adopting digital technologies will also increase the probability of adopting green technologies. In the BVP, the predicted probability of joint adoption of green and digital technologies is 10.4%, higher than under UVP (9.2%), confirming the impact of unobserved characteristics captured by ρ. Moreover, our model specification rejects the Wald test null hypothesis meaning that the impact or effect of firm characteristics is different between the digital and green sets. Similar in magnitude and significance to the UVP model, the BVP model estimates reveal that observable characteristics such as larger firms, firms that participate in a GVC, firms that are innovative, firms that invest in skills, and firms that belong to Vietnam are more likely to adopt green technologies, but also more likely to be digital¹⁷.

Table 4

Univariate and Bivariate Probit: Determinants of the adoption of green and digital technologies
	UVP		BVP
VARIABLES	(1) Digital	(2) Green	(3) Digital	(4) Green
Age (in Logs)	0.0330	0.121	0.0350	0.125
	(0.114)	(0.0946)	(0.114)	(0.0939)
Large	0.411***	0.227*	0.407***	0.224*
	(0.143)	(0.128)	(0.143)	(0.128)
Foreign Ownership	0.0210	0.227*	0.0227	0.229*
	(0.142)	(0.125)	(0.142)	(0.125)
Finished Goods	0.538***	-0.00792	0.543***	-0.00887
	(0.185)	(0.170)	(0.185)	(0.170)
GVC	0.439***	0.350**	0.439***	0.344**
	(0.157)	(0.147)	(0.157)	(0.147)
Internet Speed	-0.0681	-0.0601	-0.0760	-0.0580
	(0.144)	(0.134)	(0.144)	(0.134)
Human Capital	-0.103	-0.164	-0.0769	-0.161
	(0.417)	(0.361)	(0.422)	(0.362)
Technological Innovation	0.258*	0.491***	0.263*	0.493***
	(0.139)	(0.121)	(0.139)	(0.121)
Investment Capabilities	0.615***	0.305*	0.630***	0.311**
	(0.216)	(0.156)	(0.214)	(0.156)
Digital Skills	0.527***	0.153	0.533***	0.159
	(0.164)	(0.148)	(0.164)	(0.148)
Countries
Vietnam	0.567**	0.744***	0.579**	0.742***
	(0.267)	(0.226)	(0.268)	(0.226)
Thailand	0.421	0.314	0.435	0.319
	(0.294)	(0.228)	(0.295)	(0.228)
Sectors
Food	0.299	0.737**	0.289	0.722**
	(0.357)	(0.304)	(0.358)	(0.303)
Textiles	-0.210	0.717**	-0.214	0.705**
	(0.387)	(0.315)	(0.390)	(0.314)
Information	0.597	1.237***	0.593	1.225***
	(0.389)	(0.343)	(0.391)	(0.343)
Automotive	0.431	1.046***	0.428	1.033***
	(0.383)	(0.333)	(0.385)	(0.332)
Metal	1.022***	-0.293	1.019***	-0.344
	(0.371)	(0.505)	(0.372)	(0.498)
Constant	-3.414***	-2.797***	-3.446***	-2.803***
	(0.508)	(0.414)	(0.509)	(0.413)
Observations	658	658	658	658
Log-likelihood	-255.52761	-332.18387	585.54662
Pseudo-R-Squared	0.1744	0.2325
ρ (athrho)			0.177**
			(0.0871)
Rho			0.176

Robust standard errors in parentheses *** p < 0.01, ** p < 0.05, * p < 0.1

Note :We are not including in this model the wood and furniture sector, sectors that do not contain any firm categorised as green and/or digital and the plastic and rubber sector, as only comprises Ghanaian firms. Country Category equal to Ghana is omitted.

The BVP model also allows calculating the marginal effects of joint probabilities and marginal probabilities. The marginal effects measure the effect of a change in the explanatory variables on the outcome variable, to explore the magnitude of the impact of firm characteristics on the adoption of each technology. We estimate the marginal effects at the means (MEMs), which means varying one variable at a time (e.g.: Large from 0 to 1) while keeping values of the other covariates at their mean value. The comparison of the predicted probability between the two values of the first variable then gives the marginal effect. The marginal effect expresses how the probability of being green or digital (e.g.: $\:\text{P}\text{r}({Y}_{green}=1)$ or $\:\text{Pr}\left({Y}_{digital}=1\right)$ changes as the explanatory variable changes in one unit (e.g.: Large changes from zero to 1).

Marginal effects estimations (see Table 5) confirm that larger firms have an 8,44% and 7,86% higher probability of adopting digital and green technologies respectively having the mean value for the other independent variables in the model. Moreover, firms that produce finished goods have a higher probability of adopting digital technologies. Firms that engaged in GVCs show higher likelihood of adopting green and digital technologies: 12% and 9,1% respectively. Finally, having foreign ownership increases the likeliness of adopting green innovation. Investing in R&D, training or machinery increases the probability of adopting digital and green technologies at 13,1% and 10,9% respectively¹⁸. Furthermore, more innovative firms show a higher likelihood to adopt green technologies (17,3%) and to a lesser degree digital technologies (5,46%)¹⁹. Firms adopting digital technologies are mainly from medium and high technological sectors such as information and metals. The likelihood of adopting green technologies is higher in advanced technological sectors such as the automotive and information sectors and in less magnitude in the food and textiles sectors. This might be reflecting the widespread of green technologies across sectors with different technological advancements. Neither having a workforce with STEM experience nor access to broadband Internet to support high-speed Internet seems to be decisive when entering a green and/or digital business technological trajectory. Finally, firms that consider important digital skills when hiring a new employee are more likely to adopt digital technologies but not relevant when adopting green technologies.

Table 5

Marginal effects: adoption of green and digital technologies
	Marginal Effects-BVP		Marginal effects joint probabilities-BVP
	(1)	(2)	(3)	(4)	(5)	(6)
VARIABLES	Digital	Green	Neither	Only Green	Only Digital	Green and Digital
Age (in Logs)	0.00727	0.0437	-0.0421	0.0348	-0.00169	0.00896
	(0.0236)	(0.0329)	(0.0318)	(0.0297)	(0.0155)	(0.0100)
Large	0.0851***	0.0786*	-0.120***	0.0348	0.0413**	0.0438***
	(0.0299)	(0.0448)	(0.0431)	(0.0395)	(0.0193)	(0.0146)
Foreign Ownership	0.00471	0.0811*	-0.0725*	0.0678*	-0.00859	0.0133
	(0.0297)	(0.0447)	(0.0440)	(0.0390)	(0.0185)	(0.0140)
Finished Goods	0.0954***	-0.00311	-0.0559	-0.0396	0.0590***	0.0365***
	(0.0280)	(0.0597)	(0.0577)	(0.0553)	(0.0183)	(0.0131)
GVC	0.0928***	0.121**	-0.160***	0.0671	0.0392*	0.0536***
	(0.0336)	(0.0513)	(0.0487)	(0.0453)	(0.0221)	(0.0163)
Internet Speed	-0.0157	-0.0203	0.0272	-0.0115	-0.00692	-0.00876
	(0.0297)	(0.0466)	(0.0456)	(0.0407)	(0.0191)	(0.0135)
Human Capital	-0.0159	-0.0564	0.0583	-0.0423	-0.00192	-0.0140
	(0.0875)	(0.127)	(0.127)	(0.110)	(0.0554)	(0.0395)
Technological Innovation	0.0530*	0.167***	-0.177***	0.125***	0.0104	0.0426***
	(0.0271)	(0.0393)	(0.0390)	(0.0351)	(0.0183)	(0.0123)
Investment Capabilities	0.108***	0.104**	-0.161***	0.0530	0.0567***	0.0510***
	(0.0285)	(0.0493)	(0.0515)	(0.0445)	(0.0198)	(0.0124)
Digital Skills	0.126***	0.0567	-0.124**	-0.00132	0.0676**	0.0580***
	(0.0425)	(0.0533)	(0.0505)	(0.0453)	(0.0282)	(0.0209)
Countries
Vietnam	0.113**	0.258***	-0.291***	0.178***	0.0328	0.0800***
	(0.0492)	(0.0739)	(0.0711)	(0.0665)	(0.0306)	(0.0252)
Thailand	0.0775	0.0993	-0.140*	0.0623	0.0405	0.0370*
	(0.0499)	(0.0689)	(0.0716)	(0.0614)	(0.0343)	(0.0196)
Sectors
Food	0.0644	0.266**	-0.262***	0.197*	-0.00452	0.0689
	(0.0844)	(0.112)	(0.102)	(0.104)	(0.0461)	(0.0466)
Textiles	-0.0411	0.265**	-0.211*	0.252**	-0.0536	0.0126
	(0.0695)	(0.120)	(0.110)	(0.116)	(0.0354)	(0.0389)
Information	0.151	0.458***	-0.446***	0.295**	-0.0117	0.163*
	(0.115)	(0.117)	(0.0992)	(0.123)	(0.0446)	(0.0834)
Automotive	0.102	0.388***	-0.376***	0.273**	-0.0124	0.115*
	(0.103)	(0.120)	(0.105)	(0.119)	(0.0454)	(0.0680)
Metal	0.312**	-0.110	-0.144	-0.168**	0.254**	0.0579
	(0.137)	(0.142)	(0.138)	(0.0792)	(0.118)	(0.0769)
Observations	658	658	658	658	658	658

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

Note:We are not including in this model the wood and furniture sector, sectors that do not contain any firm categorised as green and/or digital and the plastic and rubber sector, as only comprises Ghanaian firms. Country Category equal to Ghana is omitted. The marginal effects were calculated are their means.

PPR = Estimated Predicted Probability.

As mentioned above, the BVP model offers information on joint and conditional probabilities. In the predicted joint probabilities from BVP, the strategy of non-adoption dominates all the strategies. Looking at the marginal effects (at the mean values of the covariates) on the joint probabilities of this strategy (see Table 5, column 3) confirms our previous results from the opposite view (not adopting). The probability of not adopting is 12% lower for larger enterprises or firms that consider digital skills important, 16% lower for firms that invest in R&D or participate in GVCs, and 17,7% lower for innovative firms. These results indicate that a virtuous inclusion of developing countries’ firms in the twin transition requires the removal of barriers to innovation activity, increased R&D investment and participation in GVCs involving the transfer of digital technologies and the promotion of sustainable technologies.

Also, marginal effects provide us with insightful results regarding firms that engage in the twin transition by adopting only digital or green technologies (See Table 5, Columns 4 and 5). Some interesting contrasts are the following. Size, GVC participation, previous investments in capabilities and importance given to digital skills seem to be particularly important in the adoption of digital technologies but not necessarily in the adoption of green technologies. On the opposite side, innovativeness and foreign ownership seem to be significant determinants of green technology adoption but not in digital technology adoption.

To conclude this section, the findings reported so far support the notion that the adoption of green and digital technologies is a complementary strategy by firms, hence confirming the appropriate choice of the empirical model. In turn, part of this complementarity may be because the joint adoption of green and digital technologies depends on similar firm characteristics, even though unobservable effects remain significant. Finally, we find that the different adoption strategies present marked differences in the characteristics of the firms that significantly affect them.

4.2.2 Twin transition technologies: complementary or substitute paths?

In this section, instead of calculating the raw probabilities (see Fig. 2), we rely on the predicted probabilities of the BVP to confirm that complementarities exist between green and digital technologies beyond simply using the raw probabilities. Firstly, our two-sample t-test analysis with equal variances rejects the hypothesis that the complementarity statistic (see Eq. 2) using BVP predictions is zero, either for the entire sample or for a particular subsample of firms, suggesting that complementarity exists in the adoption of green and digital technologies. in most of the subsamples we use (except medium-sized firms and privately held firms with less than 10% foreign ownership) (See Table A.4.1 in the Appendix). Secondly, we calculate the MEMs of the covariates on the different components of the complementarity statistics as well as on the complementarity effect itself (i.e., the difference between those components). Table 6 displays the effect of the covariates on the adoption of green and digital technologies jointly and separately and their difference.

In a nutshell, our results confirms that the joint adoption of green and digital technologies becomes more complementary as the firm increases in size, produces finished goods, participates in a GVC, adopts innovation, invests in capabilities and considers relevant digital skills when hiring new employees. These are all observable factors. We have confirmed above that error terms of the adoption of green and digital technologies equations are correlated to each other and therefore, other unobservable factors co-explain this complementarity. For example, in column 9, the estimated MEM, significant and equal to 0.00720, implies that the likelihood of adopting both technologies is greater (by 0.72% points) for larger firms. In addition, the marginal effect on the probability of adopting both technologies together rather than separately are larger (by 1%) for firms that invest in R&D, training or machinery than for firms that do not invest in any of these areas. The magnitude of the MEMs on complementarity is slightly smaller for firms that participate in GVCs and consider relevant digital skills when hiring new employees.

Table 6

Marginal effects of covariates on the probabilities of innovation adoption
	(7)	(8)	(9)
VARIABLES	Pr(Green & Digital)	Pr(Green)×Pr(Digital)	Difference
Age (in Logs)	0.00896	0.00775	0.00121
	(0.0100)	(0.00832)	(0.00181)
Large	0.0438***	0.0366***	0.00720*
	(0.0146)	(0.0119)	(0.00418)
Foreign Ownership	0.0133	0.0118	0.00154
	(0.0140)	(0.0117)	(0.00238)
Finished Goods	0.0365***	0.0289***	0.00754*
	(0.0131)	(0.0102)	(0.00443)
GVC	0.0536***	0.0453***	0.00829*
	(0.0163)	(0.0138)	(0.00446)
Internet Speed	-0.00876	-0.00732	-0.00145
	(0.0135)	(0.0112)	(0.00243)
Human Capital	-0.0140	-0.0120	-0.00203
	(0.0395)	(0.0328)	(0.00691)
Technological Innovation	0.0426***	0.0361***	0.00651*
	(0.0123)	(0.0100)	(0.00353)
Investment Capabilities	0.0510***	0.0412***	0.00985*
	(0.0124)	(0.00963)	(0.00506)
Digital Skills	0.0580***	0.0491***	0.00889*
	(0.0209)	(0.0178)	(0.00505)
Countries
Vietnam	0.0800***	0.0686***	0.0114*
	(0.0252)	(0.0215)	(0.00635)
Thailand	0.0370*	0.0295*	0.00745
	(0.0196)	(0.0152)	(0.00544)
Sectors
Food	0.0689	0.0620	0.00698
	(0.0466)	(0.0411)	(0.00656)
Textiles	0.0126	0.0141	-0.00158
	(0.0389)	(0.0334)	(0.00601)
Information	0.163*	0.154*	0.00848
	(0.0834)	(0.0792)	(0.00690)
Automotive	0.115*	0.107*	0.00782
	(0.0680)	(0.0629)	(0.00694)
Metal	0.0579	0.0508	0.00713
	(0.0769)	(0.0685)	(0.00919)
Observations	658	658	658

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

We summarize the main results of our econometric analysis in Table 7. This table displays the signs of the marginal effects reported in Table 5 and Table 6 and discussed above.

Table 7

Marginal effect reported sign
	Neither	Joint Probabilities Only GreenOnly Digital	Green and Digital	Complementarity Statistic
Age (in logs)
Large	-	+		+
Foreign ownership	-	++
Finished goods		+	+	+
Internet speed
GVC	-	+	+	+
Human Capital
Technological Innovation	-	++	+	+
Investment Capabilities	-	+	+	+
Digital Skills	-	+	+	+

Industrial production worldwide is being reshaped by two major megatrends: advanced digitalization of production and the transition towards environmentally sustainable production practices. In this paper we have empirically analysed the interplay between these two megatrends by looking into developing countries industrial firm technology adoption decisions. This was done based on primary data from three developing countries: Ghana, Thailand and Viet Nam.

Our work suggests the existence of complementarities between green and digital technologies, which is explained by observed and unobserved firm characteristics that have an impact on the adoption of both technologies. Firm size, participation in GVCs, innovativeness and previous investments in capabilities are some of the main characteristics associated with twin technology adoption in the sample studied in this paper.

Some of the results obtained in this paper can inform policymakers in the development of strategies to promote twin technology adoption. Consistent with Avenyo et al. (2022), the heterogeneity of factors determining digital and green transformation in manufacturing countries implies that targeted policies to enhance investment capabilities and digital skills seem to be key aspects to orient SMEs towards sustainable digital industrialisation. The integrated approach to the technologies that make up the twin transition advocated by international organisations needs to take into account the importance of firm heterogeneity in technology adoption, and that these characteristics may determine whether firms complement green and digital technologies.

This study is limited in several ways. First, the relationship between green and digital technologies is endogenous and it is not a question of looking for causalities that control this true endogeneity, but of making the analytical effort to study the adoption synergies of the core technologies of the twin transition, a key aspect to understand the feasibility of an integrated agenda of adoption of both technologies for developing countries. This implies a challenge of methodological change, given the lack of data. Second, to capture the effect of the extent to which the adoption of both technologies benefited from this type of support, the survey lacks information on environmental regulation and/or government support for digitalisation. These factors are highly relevant to the concept of complementarity between industrial and environmental policies, at national and international levels. For instance, Meckling (2021) argues that industrial policy can support environmental policy by reducing technology costs and mobilising stakeholders. Similarly, environmental policy could support green industrial policy through more ambitious GHG targets. Finally, due to the cross-sectional nature of the data is not possible to study the sequential complementarities in the adoption of green and digital technologies (Battisti, Colombo, and Rabbiosi, 2015).

Ethical Approval

Not applicable.

Funding

Not applicable.

Author Contribution

M.M and A.L wrote the main manuscript. M.M prepared all the Tables (with the descriptives and with the regression models)A.L. prepared figures 1-3All the authors reviewed the manuscript

Acknowledgement

This paper was presented in several Workshops and conferences. We thank the participants for their critical remarks. We are specifically grateful to Bart Verspagen, Önder Nomaler and Nanditha Matthews from UNU-MERIT and Carlos Bianchi and Adriana Peluffo from IECON-FCEA-UDELAR for their careful reading of the manuscript and comments to improve the paper. The views expressed in this document are those of the authors and do not necessarily reflect the views of the Secretariat of UNIDO.

Data Availability

The data that support the findings of this study are available from UNIDO, but restrictions apply to the availability of these data, which were used under licence for the current study and so are not publicly available. The data are, however, available from the authors upon reasonable request and with the permission of UNIDO.

Agostino, M., Giunta, A., Ruberto, S., & Scalera, D. (2023). Global value chains and energy-related sustainable practices. evidence from enterprise survey data. Energy Economics, 127, 107068. https://doi.org/10.1016/j.eneco.2023.107068
Agwu, G. A., Agbanike, T., Uwajumogu, N., & Ogbuagu, R. A. (2020). How do firms combine different types of innovation? a multivariate probit approach. African Journal of Science, Technology, Innovation and Development, 12 (2), 173–185. https://doi.org/10.1080/20421338.2019.1624312
Albrieu, R., Brest López, C., Rapetti, M., Ferraz, J. C., Nogueira de Paiva Britto, J., Kupfer, D., & Torracca, J. (2019). The adoption of digital technologies in developing countries: Insights from firm level surveys in Argentina and Brazil. UNIDO.
Antonioli, D., Cecere, G., & Mazzanti, M. (2014). Ict and environmental innovations in a complementary fashion. is the joint adoption by firms economically visible? (Tech. rep.). SEEDS, Sustainability Environmental Economics and Dynamics Studies.
Ardito, L. (2023). The influence of firm digitalization on sustainable innovation performance and the moderating role of corporate sustainability practices: An empirical investigation. Business Strategy and the Environment, 32 (8), 5252–5272. https://doi.org/https://doi.org/10.1002/bse.3415
Ardito, L., Raby, S., Albino, V., & Bertoldi, B. (2021). The duality of digital and environmental orientations in the context of SMEs: Implications for innovation performance. Journal of Business Research, 123, 44–56. https://doi.org/10.1016/j.jbusres.2020.09.022
Avenyo, E. K., Bell, J. F., & Nyamwena, J. (2022). Determinants of digital technologies’ adoption in South African manufacturing: Evidence from a firm-level survey. CCRED-IDTT Working Paper, 1.
Bartoloni, E., & Baussola, M. (2001). The determinants of technology adoption in Italian manufacturing industries. Review of Industrial Organization, 19, 305–328. https://doi.org/10.1023/A:1011881516500
Battisti, G., Colombo, M. G., & Rabbiosi, L. (2015). Simultaneous versus sequential complementarity in the adoption of technological and organizational innovations: The case of innovations in the design sphere. Industrial and corporate change, 24 (2), 345–382. https://doi.org/10.1093/icc/dtv003
Beier, G., Matthess, M., Shuttleworth, L., Guan, T., Grudzien, D. I. d. O. P., Xue, B., De Lima, E. P., & Chen, L. (2022). Implications of Industry 4.0 on industrial employment: A comparative survey from Brazilian, Chinese, and German practitioners. Technology in Society, 70, 102028. https://doi.org/10.1016/j.techsoc.2022.102028
Calia, P., & Ferrante, M. R. (2013). How do firms combine different internationalisation modes? A multivariate probit approach. Review of World Economics, 149, 663–696. https://doi.org/10.1007/s10290-013-0162-5
Calza, E., Lavopa, A., & Zagato, L. (2023). Advanced digitalisation and resilience during the COVID- 19 pandemic: Firm-level evidence from developing and emerging economies. Industry and In- novation, 30 (7), 864–894. https://doi.org/10.1080/13662716.2023.2230162
Cirera, X., Comin, D., Cruz, M., & Lee, K. (2021a). Firm-level adoption of technologies in Senegal. Policy Research Working Paper, No. 9657. http://doi.org/10.1596/1813-9450-9657
Cirera, X., Comin, D., Cruz, M., Lee, K., & Soares Martins-Neto, A. (2021b). Firm-level technology adoption in Vietnam. Policy Research Working Paper; No. 9567. http://doi.org/10.1596/1813-9450-9567
Cirera, X., Lage, F., & Sabetti, L. (2016). ICT Use, Innovation, and Productivity: Evidence from Sub-Saharan Africa. The World Bank. https://doi.org/10.1596/1813-9450-7868
Delera, M., Pietrobelli, C., Calza, E., & Lavopa, A. (2022). Does value chain participation facilitate the adoption of Industry 4.0 technologies in developing countries? World Development, 152, 105788. https://doi.org/https://doi.org/10.1016/j.worlddev.2021.105788
Denicolai, S., Zucchella, A., & Magnani, G. (2021). Internationalization, digitalization, and sustainability: Are SMEs ready? A survey on synergies and substituting effects among growth paths. Technological Forecasting and Social Change, 166, 120650. https://doi.org/10.1016/j.techfore.2021.120650
Fernández, S., Torrecillas, C., & Labra, R. E. (2021). Drivers of eco-innovation in developing countries: The case of Chilean firms. Technological Forecasting and Social Change, 170, 120902. https://doi.org/10.1016/j.techfore.2021.120902
Ferraz, J. C., Kupfer, D., Torracca, J., & Britto, J. N. P. (2020). Snapshots of a state of flux: How Brazilian industrial firms differ in the adoption of digital technologies and policy implications. Journal of Economic Policy Reform, 23 (4), 390–407. https://doi.org/10.1080/17487870.2019.1578651
Ferreira, J. J., Lopes, J. M., Gomes, S., & Rammal, H. G. (2023). Industry 4.0 implementation: Environmental and social sustainability in manufacturing multinational enterprises. Journal of Cleaner Production, 404, 136841. https://doi.org/10.1016/j.jclepro.2023.136841
Freitas, I.M.B., Clausen, T.H., Fontana, R. et al. Formal and informal external linkages and firms’ innovative strategies. A cross-country comparison. J Evol Econ 21, 91–119 (2011). https://doi.org/10.1007/s00191-010-0188-y
Fusaro, M. A. (2009). The rank, stock, order and epidemic effects of technology adoption: An empirical study of bounce protection programs. The Journal of Technology Transfer, 34, 24–42. https://doi.org/10.1007/s10961-007-9062-y
Geroski, P. A. (2000). Models of technology diffusion. Research Policy, 29 (4-5), 603–625. https://doi.org/10.1016/S0048-7333(99)00092-X
Grazzi, M., & Jung, J. (2019). What are the drivers of ICT diffusion? Evidence from Latin American firms. Information Technologies & International Development, 15, 15.
Grazzi, M., Sasso, S., & Kemp, R. (2019). A conceptual framework to measure green innovation in Latin America and the Caribbean. Washington DC, New York: Inter-American Development Bank. http://dx.doi.org/10.18235/0002082
Han, M. S., & Chen, W. (2021). Determinants of eco-innovation adoption of small and medium enterprises: An empirical analysis in Myanmar. Technological Forecasting and Social Change, 173, 121146. https://doi.org/10.1016/j.techfore.2021.121146
Horbach, J. (2023). Digitalisation and sustainability strategies at the firm level. RWI - Leibniz-Institut für Wirtschaftsforschung. https://doi.org/10.4419/96973167
Kamutando, G., & Tregenna, F. (2023). Complementarity between product and process innovation in small and micro-enterprises in Johannesburg, South Africa. Economics of Innovation and New Technology, 0 (0), 1–27. https://doi.org/10.1080/10438599.2023.2254241
Karshenas, M., & Stoneman, P. L. (1993). Rank, stock, order, and epidemic effects in the diffusion of new process technologies: An empirical model. The RAND Journal of Economics, 24 (4), 503–528. Retrieved February 24, 2024, from http://www.jstor.org/stable/2555742
Kesidou, E., & Ri, A. (2021). Twin green and digital transitions: Joint adoption of net zero and digital practices by UK SMEs (tech. rep.). ERC Insight Paper October. https://www.enterpriseresearch.ac.uk/wp-content/uploads/2021/10/ERC-Insight-Twin-Green-and-Digital-Transitions.pdf
Kren, J., & Lawless, M. (2023). Firm-level attitudes and actions to the “twin transition” challenges of digitalisation and climate change. ESRI Working Paper 742. https://www.esri.ie/publications/firm-level-attitudes-and-actions-to-the-twin-transition-challenges-of-digitalisation
Lorenz, E., & Kraemer-Mbula, E. (2022). Frontier technology adoption in developing countries: A measurement framework and proposed questionnaire (tech. rep.). UNCTAD Working Paper https://unctad.org/system/files/official-document/dtlstict2021d5_en.pdf
Meckling, J. (2021). Making Industrial Policy Work for Decarbonization. Global Environmental Poli- tics, 21 (4), 134–147. https://doi.org/10.1162/glepa00624
Mickoleit, A. (2010). Greener and smarter: ICTs, the environment and climate change. OECD Green Growth Papers, No. 2010/01, OECD Publishing, Paris, https://doi.org/10.1787/5k9h3635kdbt-en
Montresor, S., & Vezzani, A. (2023). Digital technologies and eco-innovation. Evidence of the twin transition from Italian firms. Industry and Innovation, 1–35. https://doi.org/3662716.2023.2213179
OECD. (2011). ISIC Rev.3 Technology Intensity Definition—Classification of Manufacturing Indus- tries into Categories Based on R&D Intensities.
Rosenberg, N. (1979). Technological interdependence in the American economy. Technology and Culture, 20 (1), 25–50. Retrieved February 24, 2024 from http://www.jstor.org/stable/3103110
Santoalha, A., Consoli, D., & Castellacci, F. (2021). Digital skills, relatedness and green diversification: A study of European regions. Research Policy, 50 (9), 104340. https://doi.org/10.1016/j.respol.2021.104340
Siewers, S., Martínez-Zarzoso, I., & Baghdadi, L. (2024). Global value chains and firms’ environmental performance. World Development, 173, 106395. https://doi.org/10.1016/j.worlddev.2023.106395
Stoneman, P., & Kwon, M.-J. (1994). The diffusion of multiple process technologies. The Economic Journal, 104 (423), 420–431. Retrieved February 24, 2024, from http://www.jstor.org/stable/2234761t
Torrecillas, C., Fernández, S., & García-García, C. (2023). Drivers to increase eco-efficiencies in Uruguay, Peru, and Panama. Energy Policy, 183, 113832. https://doi.org/10.1016/j.enpol.2023.113832
UNIDO. (2019). Industrial Development Report 2020. Industrializing in the digital age (tech. rep.).UNIDO.
Upadhayay, N. B., Rocchetta, S., Gupta, S., Kamble, S., & Stekelorum, R. (2024). Blazing the trail: The role of digital and green servitization on technological innovation. Technovation, 130, 102922. https://doi.org/10.1016/j.technovation.2023.102922
van Leeuwen, G., & Mohnen, P. (2017). Revisiting the porter hypothesis: An empirical analysis of green innovation for the Netherlands. Economics of Innovation and New Technology, 26 (1-2), 63–77. https://doi.org/10.1080/10438599.2016.1202521
Veugelers, R., Faivre, C., Rückert, D., & Weiss, C. (2023). The green and digital twin transition: EU vs US firms. Intereconomics, 58 (1), 56–62. https://doi.org/10.2478/ie-2023-0010
Wang, R., Heugens, P. P., & Wijen, F. (2023). Green by affiliation? Ownership identity and environmental management system adoption in Chinese business groups. Journal of Management, 0 (0), 01492063221143721. https://doi.org/10.1177/01492063221143721
Wen, H., Wen, C., & Lee, C.-C. (2022). Impact of digitalization and environmental regulation on total factor productivity. Information Economics and Policy, 61, 101007. https://doi.org/10.1016/j.infoecopol.2022.101007
Williams, E. (2011). Environmental effects of information and communications technologies. Nature, 479 (7373), 354–358. https://doi.org/10.1038/nature10682
World Bank. (2024). Digital Progress and Trends Report 2023. The World Bank. https://doi.org/doi:10.1596/978-1-4648-2049-6
Yurdakul, M., & Kazan, H. (2023). A study on determining the eco-innovation dynamics in Turkey’s manufacturing companies. Journal of the Knowledge Economy, 1–27. https://doi.org/10.1007/s13132-023-01269-x
Ziegler, A., & Seijas Nogareda, J. (2009). Environmental management systems and technological environmental innovations: Exploring the causal relationship. Research Policy, 38 (5), 885–893. https://doi.org/10.1016/j.respol.2009.01.020

This paper is the first to empirically explore this module of the survey.
See Appendix A.1.
https://carlotaperez.org/digital-and-green-a-very-convenient-marriage/
Digital oriented is defined as a variety of areas in which digital technologies have been adopted, e.g.: advertising, selling, purchasing, manufacturing. Environmental orientation refers to the variety of areas in which environmentally friendly practices have been adopted, e.g.: caring for the environment, reducing/recycling waste, etc.).
Defined as the ability to implement transformations involving AI-related applications and technology.
Defined as the propensity of a firm to grow without negatively affecting future generations.
Defined as the share of foreign sales.
Digital technologies comprise the most diffused innovations (accounting and HR software, e-commerce, online marketing and social media, and video conferencing) and novelty technologies (computer-aided design software and CRM system, cloud computing, augmented reality and artificial intelligence and matching learning).
Net Zero practices include: Environmental reports or audits; Changed processes or transport/logistics to reduce carbon emissions; Invested in research and development related to the environment; Introduced air pollution monitoring and filtering; Conducted training on environmental matters; Conducted market research related to low carbon products or services; Introduced new low carbon products or services; Switched to more renewable energy.
The sample was stratified by region, sector and size. Once the firms were identified, they were chosen randomly from data banks on industrial firms in each country (such as business registries). For more information about the methodology of the Survey, see Annex A.3 (UNIDO, 2019).
As reported in the Table A.2.1 (See Appendix A.2), the correlation between our dependent variables (green and digital) is 0.202 (at 10% of significance), regardless of any other differences in firms’ characteristics. Overall, the correlations are low to moderate, suggesting that multicollinearity is unlikely to occur.
This share is significantly lower than that reported by other surveys from advanced countries. The EIB survey on the twin transition, for instance, identifies 30% (EU) and 23% (US) of enterprises as green and digital respectively (Veugelers et al., 2023).
We note that the value of the ρ (athrho), considering only the sector and country effects, is higher (0.277) and significant, confirming that the observed characteristics of the firm displayed in the Table 4 and included in our model explain part of the complementarity between green and digital technologies adoption.
We note that the value of the arho, taking into account only the sector and country effects, is higher (0.277) and significant, confirming that the observed characteristics of the firm displayed in the Table 4 and included in our model explain part of the complementarity between green and digital technologies adoption.
As a way to validate bivariate probit as the estimation technique, we checked whether the results might be changed by other different measures of the covariates, but we did not find significant differences (See Appendix A.3, Table A.3.2 and Table A.3.3). The exceptions to this result are found in the models in which we incorporate investment capabilities as variables that count the areas (R&D and/or training) in which the firm invests. In those models the ρ is not significant, but the joint probability of adopting green and digital technologies remains higher than the product of the separate probabilities, which indicates that complementarity is only due to observed firm characteristics. Additionally, we check the results using a multinomial probit that allows including the four different strategies of adoption: not adopting, adopting green, adopting digital, adopting green and digital, confirming that each strategy depends on different firms’ characteristics (See A.3,Table A.3.1).
To check the reliability of our estimates regarding investment capabilities, we estimate the model using the intensity of investing capabilities (measures as the sum of total investments (R&D, training and machinery)/total of employees instead of Investment Capabilities dummy. The marginal effect remains positive for green as well as digital adoption (See Appendix A.3 Appendix, Table A.3.2).
To check the robustness of estimates, we estimate the BVP using separate dummies for product and process innovation. Marginal effects remain negative for process innovation and positive for product innovation. Additionally, estimate the BVP and marginal effects by using a categorical variable that represents the total number of areas (products or processes) in which the firm has innovated (See Appendix A.3 Appendix, Table A.3.2).

No competing interests reported.

Adoption of twin transition technologies in developing countries: a bivariate analysis

Status:

Version 1

Abstract

Figures

1. Introduction

2. Literature Review

2.1 Firm’s adoption of twin transition technologies: towards a synergistic approach

3. Empirical strategy