1 s2.0 S0040162522003092 Main

Technological Forecasting & Social Change 182 (2022) 121785
Contents lists available at ScienceDirect
Technological Forecasting & Social Change

journal homepage: www.elsevier.com/locate/techfore
The effects of digital transformation on innovation and productivity:

Firm-level evidence of South African manufacturing micro and
small enterprises
Cyrielle Gaglio a, *, Erika Kraemer-Mbula b, Edward Lorenz c
a
University of Helsinki, Department of Economics and Management, Finland, and Sciences Po, OFCE, France
b
DSI/NRF/Newton Fund Trilateral Chair in Transformative Innovation, the Fourth Industrial Revolution and Sustainable Development, University of Johannesburg,
South Africa
c
University of Côte d'Azur, France, University of Johannesburg, South Africa and Aalborg University, Denmark
A R T I C L E I N F O A B S T R A C T
JEL classification: This paper studies the relationships among the use of digital communication technologies, innovation perfor
O14 mance and productivity, using an extended version of the Crepon-Duguet-Mairesse (1998) model, for a sample of
O31 micro and small enterprises (MSEs) in a middle-income country, South Africa. Based on the results of an original
O4
survey carried out in 2019, we investigate these links for a sample of 711 manufacturing MSEs located in
Keywords: Johannesburg. We estimate the relationships sequentially, firstly estimating the relationship between digitali
Digital communication technologies
zation and innovation, and secondly the relationship between innovation and productivity. Our results show that
Product and process innovations
Productivity
selected digital communication technologies including the use of social media and of a business mobile phone for
Micro and small enterprises surfing the internet have a positive effect on innovation, and that innovation conditional on the use of these
South Africa technologies has a positive effect on labor productivity. The findings suggest that public programs aimed at
Informal economy fostering inclusive digitalization must consider the types of digital technologies that are most accessible and
beneficial to small firms, including those operating informally.
1. Introduction the predominant segment of micro and small enterprises (MSEs), which
includes firms operating informally.
Digitalization is transforming economies across the world and The African Union's “Digital Transformation Strategy for Africa 2020-
altering the way firms develop and market goods and services. It holds 2030” argues that digital transformation is a driving force for innova
many promises to spur innovation, generate efficiencies, and improve tive, inclusive, and sustainable growth. This strategic vision sees the
economic prospects (Dahlman et al., 2016). However, the dynamics of current moment as offering a leapfrogging opportunity for the continent
digitalization have not been equally spread across regions, over time, or and observes that African countries with fewer legacy challenges are
even across firms within countries. While differences in digitalization potentially able to adopt digitized solutions faster (African Union,
between high- and low-income countries remain large, spurred by the 2020). South Africa, an upper-middle-income country, has made digital
declining costs of broadband from the mid to late-2000s and the transformation a key component of its National Development Plan for
increasing ease of internet access with inexpensive mobile phones, some eliminating poverty and reducing inequalities by 2030. The “2017-2030
developing countries like South Africa have expanded their digital National e-Strategy” aims to position South Africa as a significant player
economies exponentially over the past decade or so.1 This paper ex in the development of information and communication technologies
plores the effects of the adoption of digital communication technologies (ICTs) throughout the value chain. Similarly, the “2020 National Digital
on firm-level innovation and productivity in South Africa, focusing on and Future Skills Strategy” sets out a roadmap for digital skills
* Corresponding author at: University of Helsinki, Department of Economics and Management, Latokartanonkaari 5, FI-00014 Helsinki, Finland.
E-mail addresses: [email protected] (C. Gaglio), [email protected] (E. Kraemer-Mbula), [email protected] (E. Lorenz).
1
See the country-level digital economy diagnostics undertaken by the World Bank in the context of its Digital Economy for Africa Initiative: https://www.wo
rldbank.org/en/programs/all-africa-digital-transformation. World Bank research estimates that through the digital transformation Sub Saharan Africa can in
crease its rate of growth by 2 % per year (World Bank, 2019a).
https://doi.org/10.1016/j.techfore.2022.121785
Received 21 May 2021; Received in revised form 27 March 2022; Accepted 31 May 2022
Available online 22 June 2022
0040-1625/© 2022 The Authors. Published by Elsevier Inc. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/).
C. Gaglio et al. Technological Forecasting & Social Change 182 (2022) 121785
development and stakeholder involvement in the adoption and use of

new digital technologies (South African Department of Telecommuni
cations, 2017, 2020).
There has been clear evidence of digital transformation in South
Africa over the last decade or so linked to the wide adoption of mobile
phones and improvements in the infrastructure for broadband internet
access. Fig. 1 below shows the share of the population in South Africa
using the internet and compares this to the average for high-income
countries between 1995 and 2017.2 While internet use increased
rapidly in high-income countries between 1995 and 2000, it remained
low in South Africa. This was linked to the limited investments in the
infrastructure for fixed-line telephone systems since at the time, internet
access required dialing up a connection using a modem.
Fig. 2 shows the number of fixed-line and mobile telephone sub
scriptions per 100 persons in South Africa compared to the average for
high-income countries from 1995 to 2019. While the number of fixed- Fig. 2. Number of fixed-line and mobile cellular subscriptions per 100 persons
line telephone subscriptions in South Africa was under 10 per 100 per of population.
sons in the mid-1990s, it was close to 50 per 100 persons on average in Source: https://data.worldbank.org/IT.CEL.SETS.P2.
high-income countries. The evidence points to South Africa not
benefiting substantially from the so-called ICT revolution of the 1990s monopoly on international internet access combined with the landing on
that was taking hold in developed countries based on the use of personal the African continent of several new undersea cables, resulting in an
computers and the copper wire fixed-line telecommunications infra open and competitive international connectivity market contributing to
structure for internet access. significant reductions in bandwidth prices (World Bank, 2019b).3
The number of mobile cellular subscriptions increased rapidly in In South Africa, MSEs are key drivers of the economy in terms of
South Africa after the year 2000, overtaking the number of fixed-line contributions to overall entrepreneurship, livelihoods, and play a role in
subscriptions. This increase occurred at a slightly faster rate than in alleviating poverty through employment generation (Booyens, 2011;
high-income countries while starting from a lower level. In 2000, there SEDA, 2016). While there is evidence that MSEs including informal
were about 18 mobile subscriptions per 100 persons in South Africa economy businesses are creative and can contribute to the generation of
compared to 48 per 100 on average in high-income countries. By 2011, new products (Wunsch-Vincent and Kraemer-Mbula, 2016), there have
South Africa had caught up fully and subsequently, surpassed the been few quantitative studies focusing on the determinants and effects of
average rate of penetration in high-income countries. The wide diffusion innovation on MSEs in South Africa. The few exceptions (Booyens, 2011;
of mobile phones in combination with better and cheaper access to Booyens et al., 2013; Kraemer-Mbula et al., 2019) are largely descriptive
broadband connectivity provided the basis for a rapid increase in and none of them has focused on manufacturing activities. In part, this
internet use in South Africa after 2009 as shown in Fig. 1. The key reflects the lack of data as the national innovation surveys carried out in
institutional and infrastructural developments were an end to Telkom's South Africa and other African countries are limited to registered firms.
This paper addresses this gap by making three important contribu
tions. Firstly, using new evidence, it provides an analysis of the rela
tionship between digitalization, innovation, and productivity at the firm
level for a sample of 711 manufacturing MSEs in South Africa, including
businesses in the informal sector. This unique database captures the
innovative activities of MSEs in the Johannesburg conurbation and their
use of digital technologies. Our main findings point (i) to a positive
relationship between the use of selected digital communication tech
nologies and product and process innovation, and (ii) to a positive
relationship between innovation performance and labor productivity
conditional on the use of these technologies. Secondly, it provides firm-
level empirical evidence that informs the debates around the readiness
of manufacturing MSEs in Africa to adopt more advanced fourth in
dustrial revolution automation technologies. It does so by exploring the
adoption of more accessible digital communication technologies, such as
social media, internet surfing and online sales/e-commerce, which can
be seen as providing an “entry-level” to more advanced technologies.
Fig. 1. Individuals using internet (in % of population). Thirdly, by including informal businesses in our analysis, we provide
Source: https://data.worldbank.org/IT.CEL.SETS.P2. important insights into the “inclusivity” of the digital transformation in
South Africa, which are of interest to a broader set of countries in Africa
and developing countries at large. The findings also provide valuable
policy reflections that can guide more inclusive approaches to innova
2 tion policy and digital transformation.
The data for internet use and for mobile and fixed-line subscriptions are
taken from the World Development Indicators series. For more information, see:
https://datatopics.worldbank.org/world-development-indicators/.
3
At present, the three largest mobile network operators in terms of market
share are MTN, Vodacom and Cell C. Telekom is number four. A World Bank
Report (2019b) notes that while prices in South Africa compare well to other
Sub-Saharan African nations in terms of the cost of broadband as a share of per
capita income, the absolute price of 1GB mobile broadband data is higher than
in Kenya, Nigeria, or Mauritius.
2
The remainder of the paper is structured as follows. Section 2 dis business survey data from Germany covering the period from 2001 to
cusses the background literature on the relationship between innovation 2003, Bertschek et al. (2013) focus narrowly on the effects of broadband
and productivity, including the effect of digital technologies on inno internet access and find that it has a positive and significant effect on
vation and possible direct and complementary effects of ICTs on pro innovation.5 In a study using Community Innovation Survey (CIS) sur
ductivity. Section 3 presents the model and explains the two-step vey data from 2004 for the Netherlands, Van Leeuwen and Farooqui
estimation approach we adopt, firstly estimating the relationship be (2008) argue that digitalization can enable innovation for several rea
tween the use of digital communication technologies and innovation, sons, including the use of e-commerce to roll out new products, the use
and secondly the relationship between innovation and productivity. of broadband internet for capturing and processing knowledge devel
Section 4 describes the data and presents the sample. Section 5 discusses oped elsewhere and for managing knowledge flows within and between
the results. Section 6 concludes. firms. They test this in a model using two measures of digital technology
use: (i) the share of sales done electronically and (ii) the firms' level of
2. The background literature broadband intensity use. Their results show that digitalization signifi
cantly increases the chances of successful product innovation and has
This paper connects with two debates in the literature. The first one indirect effects on productivity. Polder et al. (2010) also extend the
is around the complex relationships between innovation, digitalization, classic CDM model by investigateing the effect of digitalization on
and productivity. The second relates to the effect of ICTs on firm-level process and organizational innovation as well as product innovation in
productivity. Dutch firms. They find that in services all types of innovation are posi
tively affected by more e-purchasing, although only marginally in the
2.1. Innovation, digitalization, and productivity case of process and organizational innovation. Their results are more
nuanced for manufacturing, broadband being an important driver of
Innovation as a driver of firm performance in terms of productivity is both product and organizational innovation in manufacturing while e-
well established in the literature. There have been several quantitative commerce is positively related to process innovation.
studies establishing these links based on the use of innovation survey Few studies on developing countries have used extended CDM
data for both developed and developing countries. Many of these studies models to include ICTs as an input to innovation. Unlike the studies on
adopt the Crepon-Duguet-Mairesse (CDM) sequential modeling developed countries referred to above, due to data limitations, ICTs are
approach (1998), for which we draw inspiration in this paper.4 In an measured solely in terms of investments in computer hardware and
overview of studies on eight developing countries using the CDM software. These studies also lack measures of the newer forms of digi
framework, Fagerberg et al. (2010) observed that statistically significant talization such as broadband intensity or e-commerce. In a study of
effects were confirmed on labor productivity of at least one of the Uruguayan firms using the 2004 to 2006 and the 2007 to 2009 waves of
innovation measures used. In a study of six Latin American countries, the Service Innovation Survey, Aboal and Tacsir (2018) focus on the
Crespi and Zuñiga (2012) found that innovation has a significant effect distinction between technological (i.e. product and process) and non-
on productivity while noting that the determinants of firm-level in technological (i.e. organizational and marketing) innovation. They
vestments in innovation are much more heterogeneous in the Latin find that ICT capital investments are more important for product and
American countries studied than in OECD countries. In a later study of process innovation in services than in manufacturing. The reverse is true
service firms in Chile, Colombia, and Uruguay, Crespi et al. (2014) for their influence on organizational and marketing innovation. In a
confirm the positive effect of innovation on productivity in services, study of Chilean business using the 2007 and 2009 Longitudinal En
although firm size appears to be a less relevant determinant of innova terprise Surveys, Alvarez (2016) finds that ICT capital investments have
tion in services as compared to manufacturing. In a study adopting the a positive effect on both technological and non-technological in
CDM approach using the 2005–2007 and 2008–2010 waves of the novations in services and manufacturing. When predicted ICT in
Nigerian National Innovation Survey, Edeh and Acedo (2021) find evi vestments are introduced into the productivity equation, however, the
dence for SMEs of a positive effect of product, process and marketing effect of innovation on productivity disappears. On this basis, it is
innovation on productivity. In one of the rare studies adopting the CDM concluded that the effect of ICT capital on productivity is direct rather
framework that includes informal or non-registered businesses, Fu et al. than being indirect through innovation.
(2018) found for a sample of Ghanaian manufacturing firms a positive In summary, the econometric research based on national innovation
relationship between innovation and labor productivity. Our study, surveys in developed countries finds support for the positive effect of
which is like theirs in including informal economy firms in the sample, digitalization in the form of e-commerce and broadband internet access
extends the CDM model by including the use of digital communication on at least certain measures of innovation. Further, it supports the
technologies as an input to the innovation process. positive influence of innovation on productivity contingent on the use of
Several studies have sought to extend the CDM modeling approach in these digital technologies. While the developing country studies
this way to include ICTs as inputs to innovation. However, there is no reviewed do support the positive relationship between innovation and
uniform definition of what is meant by ICTs in these studies and in productivity, the analysis of the determinants of innovation does not
interpreting the results it is important to differentiate between in include measures of the firms' internet bandwidth intensity use or their
vestments in conventional ICT capital, consisting of the computer use of e-commerce. This limits their relevance for understanding the
hardware and software used in the firm's production process, from impact of the current digital transformation in developing regions of the
newer forms of digitalization consisting in e-commerce and broadband world (including Africa), which has witnessed an unprecedented in
internet access. E-commerce took off in developing countries from the crease in broadband internet access based on mobile telephony (see
mid-1990s with the creation of online sites like Amazon and eBay, while Figs. 1 and 2). As the study by Polder et al. (2010) argues for the case of
broadband only began to replace dial-up in the early 2000s. Several the Netherlands, broadband internet access can be a means of acquiring
studies on developed countries using data from the 2000s focus on these new knowledge inputs for innovation and for sharing knowledge be
new digital developments and find support for the positive effects of tween partners, while the use of e-commerce may contribute to
these forms of digitalization on innovation outcomes. In a study using
5
In explaining their use of data for 2001 to 2003 the authors observe that as
4
For a discussion of the evolution of research based on the original CDM only 60 % of firms in Germany had adopted broadband at this time implying
model over the 20 years following the publication of Crepon et al. (1998), see that there was sufficient variation in this variable across firms to make it useful
Lööf et al. (2017). for identifying the impact of broadband internet on innovation performance.
3
successfully rolling out new products and services. Furthermore, there is innovations. The second stage of the analysis focuses on the link be
a large theoretical and case study-based literature on business man tween innovation and productivity conditional on the use of digital
agement regarding the role of social media in driving and enabling communication technologies. As discussed below, to address problems
innovation (Bhimani et al., 2019). Several cases show how social media of endogeneity between innovation and productivity, we employ two-
can be used to support knowledge sharing and open innovation stage least squares (2SLS) using one or both of two digital communi
(Brandtzaeg et al., 2016; Hitchen et al., 2017). Social media may pro cation technologies (i.e. social media and internet surfing with a mobile
mote innovation by providing a tool for interacting with and drawing on phone) as instruments for innovation in the second stage equation. Fig. 3
users' ideas (Dong and Wu, 2015). Muninger et al. (2019) develop an below describes the extended CDM model.
organizational capabilities perspective, arguing that social media sup
ports agile processes that facilitate rapid decision making and knowl 3.1. The knowledge production function
edge flows across teams within the firm.
Following the approach in the CDM literature, we model the rela
2.2. ICTs and productivity tionship between inputs to the innovation process and innovation out
puts with a knowledge production function (KPF), of which an early
A related literature on firm heterogeneity in their investment pat discussion can be found in Griliches (1979) and a more recent presen
terns has investigated the direct effects of ICT capital on productivity. tation in Wagner (2006). At the firm i level, the general form of the
Much of this literature adopts a production function approach in which equation is the following:
the standard model is augmented to include ICT capital in addition to
ΔKi = f (Hi , Di , Zi ) (1)
non-ICT capital (Draca et al., 2009; Stiroh, 2005). For the case of
emerging market countries, Commander et al. (2011) look at the con
where ΔKi is growth in the knowledge of the firm, Hi equals the firm's
sequences of ICT capital adoption and use on firm performance in Brazil
investment in developing new knowledge (i.e. R&D expenditures), Di
and India. They find a strong positive association between investments
equals the activities taken to source ideas and knowledge from outside
in ICT capital, measured as the shares of non-production workers using
the firm, and Zi equals a vector of variables that may be important for the
either PCs or ICT-controlled machinery, and productivity in
firms' capacity to develop new knowledge (i.e. its stock of capital, sector
manufacturing firms. Based on these results, we expect that the intensity
of activity, size). The inclusion of the variable Di reflects that R&D is
of ICT capital (measured as investments in computer hardware) would
only one of several possible inputs to the process of knowledge creation
have a positive effect on productivity.
and that firms may also draw from external sources of knowledge,
A strand of this literature investigates possible complementarities
including the knowledge they acquire through interaction with users
between investments in ICT capital, organizational innovation and R&D,
(Lundvall, 1988).
arguing that the productivity gains from ICT capital would be enhanced
In our empirical estimation of Eq. (1), we measure the creation of
by workplace reorganization allowing for more decentralized decision
new knowledge through the introduction of a new product or a new
making. ICT capital is usually captured by the intensity of computer use,
process. The firm's internal investment in new knowledge is proxied by
through indicators such as the share of employees using a PC or in
whether it undertakes R&D, and accessing ideas and knowledge from
vestments in computer hardware and software. In a study for the period
external sources is measured through the firm's use of various new
1995–1996 of firms in the USA, Brynjolfsson and Hitt (2002) find evi
internet-based digital communication technologies as well as its coop
dence of complementarities where firm productivity is higher when
eration with other firms in the same sector of activity. In the first stage,
investments in ICT capital are combined with work reorganization and
our dependent variable (Innovation) refers to the introduction of a new
the use of skilled labor. Black and Lynch (2001), who measure ICT in
or significantly improved product, or alternatively a new or significantly
terms of the percentage of non-managerial employees using computers,
improved process, during the fiscal year 2019.
find evidence of positive complementarities between ICT and workplace
Since our dependent variable is dichotomous, we fit the model with a
innovations. Mohnen et al. (2018), using firm-level data for the
maximum likelihood probit model at the firm i level as specified in Eq.
Netherlands for the period from 2008 to 2010, similarly find that the use
(2):
of ICTs measured in terms of investments in computer hardware are
complementary with R&D and organizational innovation in the sense Innovationi = c + β1 Log(sizei ) + β2 Log(agei ) + β3 Fixed capitali
(2)
that joint investments lead to higher total factor productivity growth. In + β4 ICT capitali + β5 RDi + β6 Xi + β7 Yi + εi
a recent study using panel data for 5511 Spanish industrial firms, Bal
lestar et al. (2020) extend this literature to include complementarities where Innovationi is a dichotomous variable equals to 1 if the firm has
linked to the use of robots and e-commerce. They find evidence of introduced into the market a new or significantly improved product or
positive complementarities for the combined use of robots and engaging alternatively has implemented a new or significantly improved process,
in process innovation, and for the combination of using robots and a c is the constant term, Log(sizei) is the number of employees, and Log
composite measure of e-commerce. (agei) is the age of the firm. Both are expressed in natural logarithms.
As discussed in Section 3.2 below, our data allow for the investiga Fixed capitali is the intensity of fixed capital defined as the value of ve
tion of selected complementarities related to the use of online sales or e- hicles, furniture, and machinery (excluding ICT equipment) per
commerce. We test for complementarities between this digital technol employee and ICT capitali is the value of ICT capital per employee where
ogy and ICT capital, and both process and product innovation. ICT capital includes the firm's stock of computers, fixed-line telephones,
printers, scanners, and fax machines. As discussed above in Section 2.2,
3. Digitalization, innovation, and productivity: a sequential a variety of research has shown that computerization can result in
modeling approach increased productivity by substituting for the use of manual labor in
both manual and information processing tasks. Thus, we expect a higher
Our empirical analysis draws inspiration from the sequential intensity of use of ICT capital to result in higher labor productivity.6
modeling approach associated with the so-called Crepon-Duguet-Mair RDi refers to whether the firm has engaged in any R&D activities for
esse (1998), often known as the CDM model. The CDM literature has the purposes of innovation. Xi is a vector of binary variables measuring
focused on investigating sequentially the link between R&D and inno
vation and the link between innovation and productivity. We extend this
framework by including in the first stage of the analysis the effect of 6
See the literature on routine-biased technical change associated with the
digital communication technologies on product and process work of Autor et al. (2003).
4
Digital communicaon
technologies
Innovaon
Producvity
(product and process)
R&D
Fig. 3. The extended CDM model with digital communication technologies.
the use of four digital communication technologies: social media, and therefore faces its own downward sloping demand curve. The log of
making use of a business mobile phone for surfing the internet, making real revenue is denoted by ri and the log of the firm's output price by pi
use of a mobile phone for interacting with customers and doing online giving: ri = pi + qi. The isoelastic demand equation facing the firm i in
sales. As discussed above, we expect the digital communication tech logarithmic form is as follows:
nology variables to have a positive effect on innovation. Yi is a vector of
qi = ηpi (5)
controls including whether the firm is formal (viz. registered with the
South African Revenue Service), whether its sector of activity is classi where η is the (negative) demand elasticity. Combining Eqs. (4) and (5)
fied as high-tech (HT) or low-tech (LT), and whether it has cooperated yields the following expression for the revenue as a function of the
with other firms in the same industry. As widely discussed in the inno inputs:
vation systems literature, an important mechanism for increasing access
to knowledge that can contribute to better innovation performance is ri =
η+1
(a + αci + βli + γki ) (6)
inter-firm cooperation promoting interactive learning (Lundvall, 2010; η
Jensen et al., 2007). Several studies focusing on African countries sup The equation implies that the knowledge stock may have a positive
port the importance of inter-firm cooperation for innovation, including effect on revenue and also productivity if the value of the coefficient γ is
for small firms (Van Dijk, 2002; Oyelaran-Oyeyinka and McCormick, positive. As Hall (2011) notes, the equation also implies that the esti
2007). Correspondingly, we expect that cooperation between firms will mated coefficients on the inputs will be biased downward if demand is
result in improved access to knowledge and improved innovation per elastic (η < − 1).
formance. β1 to β7 are the coefficients associated with the previous In the CDM literature using CIS data, the estimation of Eq. (6) is
variables and εi~N(0, 1) is the error term. usually done by regressing the log of revenue per employee on the log of
ICTs are transforming both the way firms produce/develop new capital expenditures per employee. Firm size is measured by the number
products and the way they interact with other firms or with consumers. of employees and innovation activity proxied either by the share of
We expect that while the use of computers and other forms of ICT capital innovative sales or by a binary variable that equals to 1 if the firm has
may have a direct effect on productivity, the influence of internet- innovated either a new product or a new process.8 This estimation is
dependent digital communication technologies would be indirect likely to present a problem of endogeneity due to simultaneity since
through their effect on the development and marketing of new products more productive firms may be better placed to invest additional re
and services.7 sources in innovation activities. As it is common in the CDM literature,
we address this problem by employing 2SLS using one or both of the
3.2. The link between innovation and productivity following excluded exogenous variables from the first stage equation as
instruments: (i) the use of social media and (ii) the use of a business
In the second stage of the sequential model, we estimate the rela mobile phone for surfing the internet. As discussed above, we expect
tionship between innovation and productivity. We model productivity these variables to have a positive and statistically significant effect on
with an extended version of the production function, adding a measure innovation. In addition, they should meet the exclusion restriction
of the knowledge created by innovative activity to the standard pro condition of only affecting productivity indirectly through their effect
duction function. The equation takes the following general form: on innovation.
In the second stage, our dependent variable (Productivity) refers to
Q = ACa Lb K g (3)
labor productivity and is measured as the natural logarithm of the value
of the firm's turnover per employee in 2019. Due to missing observa
where Q is output, AC is the level of capital stock, L is labor and K is a
tions, data on the absolute value of turnover are only available for 273
measure of the knowledge stock. For estimation purposes the logarithm
firms. We have interval data on turnover for 318 firms, but not absolute
of Eq. (3) is taken at the firm i level:
values. On this basis, we make use of multiple imputations to generate
qi = a + αci + βli + γki (4) absolute turnover values for the 318 firms for which only interval data is
available. We exclude firms for which we have neither the absolute
Following the procedure presented in Hall (2011), Eq. (4) can be
value nor an interval range for turnover. Our resulting sample for the
expressed in revenue terms under the assumption of an isoelastic de
second stage productivity regression comprises 591 firms. The second
mand equation. Each firm is assumed to produce differentiated products
stage of the 2SLS regression model takes the following form at the firm i
level as specified in Eq. (7):
7
OECD (2019) observes that even if the gains from digitalization have been
substantial, there is no consensus on the direct causality between digital
communication technologies and productivity. For example, more productive
firms may benefit from digitalization because they are more likely to have
8
access to knowledge for developing new products or implementing new orga The literature is vast. Highly cited studies using either revenue or value
nization methods than other firms. Our sequential approach allows us to added per employee in estimating the productivity relation include Crepon
explore the possible indirect effect of digitalization on productivity through et al. (1998), Hall et al. (2008), Lööf and Heshmati (2006), Mairesse and Robin
innovation. (2017), Griffith et al. (2006), Mairesse et al. (2005).
5
Productivityi = c+β1 Innovationi +β2 Log(sizei )+β3 Log(agei ) Table 1

′ Descriptive statistics.
+β4 Fixed capitali +β5 ICT capitali +β6 RDi +β7 Xi +β8 Yi + εi
Description of variables Type of Mean
(7)
variables (over full
sample)
where c is the constant term, Innovationi is the instrumented value for
either product or process innovation derived from the first stage inno Dependent variables
Product Whether the firm has introduced Binary 0.49
vation Eq. (2). Xi′ is a vector with the included digital communication innovation entirely new or significantly
technologies that are not used to instrument product or process inno improved products
vation.9 The vector of control variables remains the same. β1 to β8 are Process Whether the firm has introduced Binary 0.08
the coefficients associated with these variables and εi is the error term. innovation entirely new or significantly
improved processes
In the estimation of Eq. (7), we include selected interaction terms to
Labor Average value of the log of Nominal 9.73
investigate the presence of possible complementarities. This exercise is productivity a
productivity over 10 imputations
limited by the variables we have available in the data set. We lack a
measure of organizational innovation that would allow us to test for
Independent variables
complementarity between ICT capital and a measure of organizational Sizeb Natural logarithm of the number Nominal 1.27
innovation as developed in the literature reviewed above on ICT com of employees (full-, part-time
plementarities. We do investigate for complementarity between ICT and occasional)
Age Natural logarithm of the age of Nominal 2.35
capital and new digital technology in the form of online sales/e-
the firm in year
commerce under the assumption that by increasing the volume of Fixed capital Total value of fixed assets (i.e. Nominal 22.42
sales, online sales could reduce underutilization of ICT capital and so vehicles, furniture, machinery,
much like capital deepening result in measured increases in labor pro etc.) in tens of thousands of
ductivity. We also examine possible complementarities between online Rands/Number of employees
ICT capital Total value of ICT equipment (i. Nominal 1.58
sales and both product and process innovation. By helping firms to roll
e. computers, telephones,
out and commercialize their new product developments, online sales printers, scanners, fax machines,
might increase the measured productivity gains from product innova etc.) in tens of thousands of
tion. Similarly for ICT capital, greater sales volumes from online sales Rands/Number of employees
Registration Whether the firm is registered Binary 0.27
may help amortize the costly investment in new production methods
(formal or with South African Revenue
reflected in increases in measured labor productivity. informal firm) Service
Cooperation Whether the firm has cooperated Binary 0.48
with other firms in the same
4. Data source, sample, and descriptive statistics
industry
Sector Main manufacturing activities Binary 0.09
4.1. Data source conducted by the firm –
Following the OECD
This paper uses a unique dataset of MSEs located in Johannesburg, classification (Hatzichronoglou,
1997), an activity is either
South Africa, which we refer to as the “MSE Survey”10. The survey
considered as high-tech (HT) or
focused on the central business district of Johannesburg, which is the low-tech (LT)c
capital of Gauteng province and accounts for 16 % of South Africa's GDP R&D Whether the firm has engaged in Binary 0.15
and 40 % of Gauteng's economic activity. It focuses on manufacturing R&D activities for innovation
firms and aims to understand some of the challenges faced by MSEs in
their innovation activities as well as understanding the environment in Digital communication technologies
which these firms operate. This survey consists of a set of 74 questions Social media Whether the firm uses social Binary 0.34
media for the business
capturing a range of information, from the background of the owner to
Internet surfing Whether the firm uses a mobile Binary 0.07
the characteristics of the workforce and financial issues. The data phone for surfing the internet
collection spanned a period of three months, from June to August 2019. Mobile customer Whether the firm uses a mobile Binary 0.28
Finally, the sample covered 711 MSEs. phone to interact with customers
The full description of the different variables built from the survey Online sales/E- Whether the firm uses e- Binary 0.13
commerce commerce or online sales
and associated descriptive statistics are reported in Table 1. Innovation
a
refers to the introduction onto the market of a new or significantly The survey results provide the absolute value of sales in 2019 for 273 firms
improved product, or alternatively a new or significantly improved and interval data for a further 318 firms. Based on these variables and the other
process. To count as an innovation, the product (or the process) needs to variables in the productivity equation, we use multiple imputations to estimate
the value of turnover for the 318 firms for which only interval information on
be new to the firm but not necessarily new to the firm's market and, as
2019 sales exists. This provides a sample of 591 firms.
many authors have observed, innovations introduced by firms located in b
In calculating the total employment of the firm the number of part-time
low- and middle-income countries often have an imitative and incre workers is weighted by a factor of 0.5.
mental nature because these firms are far from the technological frontier c
See Tables A1. HT covers high- and medium-high tech sectors. LT covers low-
(Crespi and Peirano, 2007; Goedhuys, 2007; Fagerberg et al., 2010; and medium-low tech sectors.
Srholec, 2011). By including incremental and imitative activities, this Source: MSE Survey, authors' calculations.
definition of innovation can help account for what may appear to be an

9
We use one variable, social media, to instrument product innovation. For exceedingly high rate of innovation success, with 49 % of the sample of
process innovation, we add to this the variable measuring internet surfing with 711 classified as product innovators.
a mobile phone as it serves to strengthen the instrument in terms of the value of
The MSE survey includes both measures of investments in ICT capital
the standard F-statistic.
10 and the use of digital communication technologies that depend on
This survey was conducted in 2019 under the project “Community of
Practice in Innovation and Inclusive Industrialisation”, hosted by the DSI/NRF having internet access and may be used to increase the firm's visibility
South African Research Chair in Industrial Development, University of Johan on the market or for communication and exchange with other firms and
nesburg. Accession date: February 2020. clients. These digital communication technologies include making use of
6
social media (34 %), doing online sales (13 %), using a mobile phone to note that the manufacture of machinery and equipment accounts for 7.1
browse the internet (7 %) and using a mobile phone to interact with % of total ICT capital, the fourth-highest percentage.
customers (28 %). As noted in the above discussion of the choice of Fifthly, looking at the use of digital communication technologies, it
instruments, we assume that the influence of using social media and seems that a significant share of MSEs in South African manufacturing
using a mobile phone to browse the internet on productivity would be sectors use social media for their business (34 % of the total population,
indirect, through their effects on innovation. see Table 1). Social media may be used not only to increase the exposure
of the firm to prospective clients but also for purposes of information
4.2. Differences by sector exchange with other firms and organizations. In the context of South
African manufacturing firms, the manufactures of wearing apparel and
The data collected cover only manufacturing firms, so we split our furniture – which have the most weight in this sample in terms of em
sample into high-tech (HT) and low-tech (LT) sectors (Hatzichronoglou, ployees and capital intensities – have the most intensive use of these
1997) as explained in Table 1. Table A1 in the appendix provides digital communication technologies: respectively, 29 % and 14 % for
descriptive statistics according to sector for the firms' size in terms of social media, 18 % and 12 % for internet surfing, 29.1 % and 12.1 % for
employment, its age, the capital intensities, and the use of digital mobile customer, 19.6 % and 22.8 % for online sales. The manufacture
communication technologies. A few salient characteristics emerge from of textiles uses these four digital technologies in a very complementary
the summary statistics. Firstly, most of the firms (90.7 %) belong to LT way, between 10 % and 12.1 %. The other manufacturers mainly use one
sectors: 27.6 % of them are active in the manufacture of wearing apparel of the four digital communication technologies rather than all four
and 17.9 % in the manufacture of furniture. The manufacturers of tex combined. For example, for manufactures of chemicals and wood, the
tiles and basic metals represent 10.1 and 9.4 % of the firms respectively. use of mobile phones for surfing the internet prevails (respectively 10 %
More than half of the firms in the HT sectors are distributed between the and 14.1 %). For manufactures of wood and printing and reproduction
manufacture of chemicals (3.2 % of the firms) and other manufacturing of recorded media, online sales are the most used (5.4 % for each). For
(2.5 %). So, the distribution of the manufacturing firms located in the manufacture of basic metals, the use of mobile phones for interacting
Johannesburg is left-skewed towards LT sectors. with customers is clearly more used than the other three (11.1 %). So,
Secondly, firms have on average 16 years in the LT sectors and 21 the use of digital communication technologies is relatively heteroge
years in the HT sectors. In general, firms in the sample seem to be mature neous between sectors.
and well-established in the productive tissue of the city. Almost half of These shares motivate our study of the relationship between digita
them (0.48 cf. Table 1) cooperate with other firms in the same industry, lization and innovation especially compared with the R&D efforts
and most of them (0.72) work predominantly in local markets, with engaged by South African MSEs: only 15 % of the firms in our sample
customers located in the inner city or surrounding suburbs. So, in other have engaged in R&D activities for innovation (see Table 1). This might
words, these firms penetrate the market and develop distribution be because R&D requires substantial capital investments that are out of
channels. the reach for most MSEs. However, digital communication technologies
Thirdly, a quarter of the total workforce is employed in the manu that rely only on broadband connection seem to be more accessible for
facture of furniture (1349 workers). Manufactures of wearing apparel, South African MSEs.
wood, basic metals, textiles, and food products employ between 7.2 %
and 12.5 % of the remaining workforce. The workforce in all sectors is 5. Econometric results
made up predominantly of full-time contracts. Part-time and occasional
contracts concern 14 % and 8 % of the workers respectively. As proposed 5.1. The effect of digital communication technologies on product and
by Fu et al. (2018), another way to evaluate this workforce refers to a process innovations
decomposition between micro (less than or equal to 9 employees), small
(10 to 29 employees), medium (30 to 99 employees) and large firms In Table 2 below, we provide the results of estimating Eq. (2) for the
(equal or >100 employees). For each sector, we have a proxy of this determinants of product and process innovations. The regressions in
decomposition by relating the number of employees to the number of columns (a) and (d) include only the four digital communication tech
total firms: 78.1 % of the firms are micro-ones and 18.4 % are small ones. nologies. Columns (b) and (e) add the variables for R&D, cooperation
So, the distribution of the number of employees is left-skewed towards between firms and the measures of capital intensities. Columns (c) and
MSEs (maybe artisanal ones). (f) present the complete specifications, including the different controls.
Fourthly, the South African firms differ significantly in terms of the Of the four digital communication technologies, internet surfing has a
relative intensity of use of fixed capital and ICT capital. For example, the positive and statistically significant effect on both product and process
manufacture of rubber and plastic products represents 12.3 % of total innovation, and this result is robust when adding the different controls
fixed capital but a very small part of total ICT capital (1.4 %). The and other covariates. Surfing the internet, as we noted above, is a way of
manufacturers of other non-metallic mineral products show the same acquiring knowledge and ideas that can feed into the process of devel
gap between fixed and ICT capital (6.2 % for the former, 1.1 % for the oping a new product or a new process. Social media has a statistically
latter). The manufactures of furniture and wearing apparel, the two significant effect on product and process innovations in columns (a) and
largest sectors both in terms of the numbers of firms and total employ (d). It can be used as a tool for communities to establish online groups
ment, also differ from one another in their capital intensity. While the for purposes of discussion and knowledge exchange. The coefficient on
manufacture of furniture accounts for 51.3 % of total fixed capital but social media is no longer significant for process innovation after adding
only 4.9 % of total ICT capital, the manufacture of wearing apparel controls while it remains significant in the case of product innovation,
covers 15.2 % of total fixed capital and 24.5 % of total ICT capital. suggesting that social media may be of greater use in accessing external
Conversely, certain sectors account for a larger share of ICT capital and a knowledge for developing new products than it is for developing new
smaller share of fixed capital: this means that firms belonging to these processes.
sectors are relatively advanced in terms of computerization. Moreover, The coefficient on the variable measuring communicating with cus
four sectors (the manufacture of food products, the manufacture of tomers with a mobile phone is positive but, contrary to our expectation,
machinery and equipment, the manufacture of wearing apparel and the is not significant Firms that cooperate with other firms in the same in
manufacture of wood) account for 76.4 % of the ICT capital while they dustry or trade have a greater propensity for product innovation, and
account for only 23.5 % of total fixed capital and only 28.6 % of the total undertaking R&D has a positive and significant efffect on both product
employment. So, the diffusion of ICT capital is very uneven. Even if firms and process innovation. The measure of non-ICT or fixed capital in
in the HT sectors represent a small part of our sample, it is interesting to tensity is not statistically significant for either product or process
7
Table 2
Probit regression predicting product and process innovations.
(a) (b) (c) (d) (e) (f)
Product innovation Process innovation
Social media 0.177*** 0.172*** 0.160*** 0.0391* 0.0325 0.0198

(0.0421) (0.0422) (0.0436) (0.0232) (0.0218) (0.0202)
Internet surfing 0.219*** 0.178** 0.186** 0.265*** 0.185*** 0.144**
(0.0758) (0.0837) (0.0829) (0.0751) (0.0692) (0.0630)
Mobile customer 0.0574 0.0532 0.0521 0.00395 0.00186 0.00342
(0.0450) (0.0452) (0.0454) (0.0225) (0.0215) (0.0197)
Online sales/E-commerce 0.0415 0.00636 0.000847 0.0455 0.0278 0.0101
(0.0614) (0.0627) (0.0646) (0.0346) (0.0289) (0.0239)
R&D 0.131** 0.132** 0.0960** 0.0515*
(0.0579) (0.0599) (0.0376) (0.0300)
Cooperation 0.0919** 0.0941** 0.00981 0.00558
(0.0390) (0.0390) (0.0181) (0.0165)
Fixed capital 8.19e− 09 8.19e− 09 5.20e− 09 4.03e− 09
(1.25e− 08) (1.24e− 08) (3.32e− 09) (2.69e− 09)
ICT capital − 3.97e− 08 − 2.29e− 08 4.40e− 08 3.18e− 08
(1.36e− 07) (1.27e− 07) (4.11e− 08) (3.51e− 08)
Size 0.00353 0.0264***
(0.0214) (0.00841)
Age − 0.0247 − 0.0119
(0.0239) (0.0106)
Registration 0.0443 0.00826
(0.0475) (0.0191)
Sector 0.103 − 0.0477**
(0.0678) (0.0227)
Observations 711 711 711 711 711 711
Pseudo R2 0.0424 0.0554 0.0595 0.126 0.163 0.201
Wald χ 2 39.49*** 50.47*** 57.52*** 51.09*** 66.62*** 75.87***
Correctly classified 60.76 % 61.74 % 61.74 % 91.84 % 91.84 % 92.12 %
Note: Marginal effects are reported in this table. Robust standard errors are given in parentheses.
***
P < 0.01.
**
P < 0.05.
*
P < 0.1.
innovation, and we find no significant influence for the different control the intensity of ICT capital use has a positive and statistically significant
variables including the size or the age of the firms on product innova effect on productivity. This supports the literature cited above on the
tion. There is a positive and statistically significant effect of size on positive effect of the computerization of work processes and internal
process innovation which may reflect that larger firms are better placed knowledge flows on productivity. The coefficient of the measure of fixed
to finance costly investments in new equipment. In the context of SMEs capital intensity is surprisingly negative but not statistically significant.
in Sri Lanka, De Mel et al. (2009) similarly find that firm size plays a In column (a) neither of the two included digital communication
larger role in process and organizational innovations than in product technology variables (viz. the use of a mobile phone for surfing the
innovation. So, our findings are in line with their results. Furthermore, internet and for communicating with customers) has a statistically sig
as in the case of our sample, for manufacturing firms, they do not find a nificant effect on productivity. In the results presented in column (e), the
significant correlation between innovation and the age of the firm. Being one included digital communication technology variable (i.e. using a
registered has a small positive but statistically insignificant effect on mobile phone to communicate with customers) similarly does not have a
innovation. significant effect on productivity. These results are consistent with our
We find convincing evidence in support of the endogeneity of both argument that digital communication technologies, unlike ICT capital,
product and process innovation to productivity. Based on both the only have an indirect effect on productivity through their positive in
Durbin χ 2 test and Wu-Hausman F statistic, the null hypothesis of exo fluence on innovation. This is also the case with R&D and inter-firm
geneity is rejected at the 0.06 level or better. The value of the F-statistic cooperation, which our results suggest only have a significant influ
for the instruments is over 11 in the case of product innovation and over ence on productivity through their effect on innovation. This latter
24 in the case of process innovation, showing that the instruments are finding is consistent with studies suggesting that while small African
not weak. firms may draw from inter-firm relationships to innovate, their ability to
transform external sources of knowledge into superior firm performance
is contingent on having complementary firm-specific capabilities (skills
5.2. The effect of product and process innovations on labor productivity and managerial capabilities) to do so (Zulu-Chisanga et al., 2021). This
may explain why inter-firm cooperation has a positive effect on product
Table 3 presents the results of estimating the second stage of the two- innovation, but not on productivity in our sample. The coefficient
stage least squares IV regression. Columns (a) and (e) show the results capturing the effect of firm registration is positive but only statistically
for the two regressions estimating the effect of product and process in significant in the case of the regression estimating the effect of process
novations respectively on productivity. The results confirm positive and innovation on productivity. The age of the firm has a positive and sig
statistically significant effects of both product and process innovations nificant influence on productivity, while the size of the business has a
on labor productivity. Consistent with our expectations, the measure of
8
Table 3
2nd stage Instrumental variable regression explaining labor productivity.
(a) (b) (c) (d) (e) (f) (g) (h)
Labor productivity
Product innovation 0.234** 0.221* 0.208* 0.205*

(0.119) (0.114) (0.126) (0.118)
Process innovation 0.270*** 0.273*** 0.371*** 0.373***
(0.0984) (0.0967) (0.125) (0.125)
Internet surfing 0.0260 0.0345 0.00842 0.0281
(0.0405) (0.0389) (0.0483) (0.0412)
Mobile customer − 0.0296 − 0.0280 − 0.0337 − 0.0296 − 0.0127 − 0.0123 − 0.0188 − 0.0181
(0.0206) (0.0200) (0.0222) (0.0203) (0.0156) (0.0157) (0.0167) (0.0167)
Online sales/E-commerce − 0.0483* − 0.0645** 0.0871 − 0.00377 − 0.0478** − 0.0538** − 0.260* − 0.277*
(0.0271) (0.0280) (0.167) (0.123) (0.0227) (0.0230) (0.144) (0.145)
R&D − 0.0375 − 0.0329 − 0.0466 − 0.0361 − 0.0350 − 0.0341 − 0.0399 − 0.0378
(0.0303) (0.0292) (0.0340) (0.0301) (0.0254) (0.0255) (0.0264) (0.0265)
Cooperation − 0.00598 − 0.00593 − 0.00123 − 0.00366 0.0153 0.0147 0.0150 0.0142
(0.0211) (0.0206) (0.0225) (0.0211) (0.0142) (0.0142) (0.0147) (0.0147)
ICT capital 1.23e− 07** 1.00e− 07* 1.49e− 07** 1.07e− 07* 9.83e− 08* 9.06e− 08* 9.71e− 08* 8.61e− 08
(5.98e− 08) (5.86e− 08) (6.99e− 08) (6.03e− 08) (5.05e− 08) (5.06e− 08) (5.23e− 08) (5.25e− 08)
Fixed capital − 5.77e− 09 − 4.80e− 09 − 8.43e− 09 − 5.85e− 09 − 4.79e− 09 − 4.56e− 09 − 6.80e− 09 − 6.46e− 09
(5.27e− 09) (5.08e− 09) (6.47e− 09) (5.51e− 09) (4.25e− 09) (4.27e− 09) (4.63e− 09) (4.62e− 09)
Size 0.0150 0.0151* 0.0226* 0.0188 0.00740 0.00738 0.00477 0.00486
(0.00935) (0.00912) (0.0135) (0.0117) (0.00864) (0.00862) (0.00913) (0.00912)
Age 0.0208** 0.0210** 0.0194* 0.0204** 0.0247*** 0.0248*** 0.0249*** 0.0250***
(0.0102) (0.00993) (0.0108) (0.0101) (0.00873) (0.00872) (0.00902) (0.00901)
Registration 0.0278 0.0258 0.0242 0.0236 0.0365** 0.0355** 0.0359** 0.0345*
(0.0216) (0.0212) (0.0231) (0.0217) (0.0173) (0.0173) (0.0179) (0.0179)
Sector − 0.0462 − 0.0376 − 0.0411 − 0.0332 − 0.0113 − 0.00846 − 0.0119 − 0.00830
(0.0297) (0.0289) (0.0316) (0.0302) (0.0264) (0.0262) (0.0271) (0.0270)
Online sales * ICT capital 1.20e− 06** 1.48e− 06** 4.18e− 07 6.29e− 07
(4.88e− 07) (7.35e− 07) (3.94e− 07) (4.23e− 07)
Product innovation * Online sales − 0.335 − 0.159
(0.408) (0.314)
Process innovation * Online sales 0.253 0.262
(0.168) (0.167)
Constant 2.184*** 2.173*** 2.183*** 2.170*** 2.193*** 2.188*** 2.193*** 2.187***
(0.0710) (0.0704) (0.0742) (0.0710) (0.0581) (0.0577) (0.0598) (0.0595)
Observations 591 591 591 591 591 591 591 591
Wald χ 2 38.96*** 43.88*** 36.16*** 43.68*** 52.09*** 56.22*** 50.65*** 54.30***
Note: Standard errors are given in parentheses.

***
P < 0.01.
**
P < 0.05.
*
P < 0.1.
positive but non statistically significant effect on productivity. communication technologies could also explain the unanticipated
Our data allow for selected testing for the presence of ICT comple negative coefficient on the variable measuring the direct use of online
mentarities influencing productivity. The regression results in columns sales. If this interpretation is correct, then it points to the need for pol
(b), (c) and (d) estimating the effect of product innovation on produc icies and training programs designed to develop digital skills and com
tivity, test for complementarity between online sales and ICT capital, petencies. We return to the question of policies to promote the adoption
and between online sales and product innovation. Even though the and effective use of digital technologies in the conclusions.
estimated direct effect of online sales on productivity is negative, we
find support for the argument that the measured productivity benefits of 6. Conclusion
investments in ICT capital are increased by using online sales. The
complementarity between ICT capital and online sales, however, is not This paper explores the relationship between digitalization, inno
present in the case of the regressions estimating the effect of process vation, and productivity for a sample of micro and small manufacturing
innovation on productivity in columns (f) and (h). Furthermore, neither firms in Johannesburg, South Africa. The survey of MSEs on which this
the coefficient on the interaction term between process innovation and study is based includes indicators of the use of social media and the use
online sales in column (g) nor the coefficient on the interaction term of a business mobile phone for surfing the internet, communicating with
between product innovation and online sales in column (c) are statisti customers, and accessing markets through online sales. To our knowl
cally significant. edge, it is the only survey in South Africa that specifically measures the
These unexpected results may reflect the limited experience that adoption of these digital technologies by micro-enterprises, including
MSEs in South Africa have in doing online sales, which is an emerging non-registered businesses. Our findings show, firstly, that selected dig
technology in their context. Their limited experience with digital ital communication technologies, including the use of social media and
9
using a mobile phone to browse the internet, have a positive effect on such as those in the informal sector.
innovation. These results support the literature arguing that social Our original contributions are threefold. (i) We utilize a novel and
media and using the internet can enable innovation by supporting up-to-date database that addresses the existing data gap in the national
interaction and knowledge exchange among firms and with consumers. innovation surveys carried out in South Africa and other African coun
Secondly, innovation conditional on the use of these digital communi tries, which are limited to registered firms. (ii) We analyze the effect of
cation technologies has a positive effect on labor productivity. This digital transformation on firm performance. In this regard, by using the
result is consistent with the large literature cited above in both devel CDM approach, we pay attention to the role of innovation in enhancing
oped and developing countries, showing a positive relationship between productivity, since these firms are creative and contribute to the gen
innovation and productivity. eration of new products. Digitalization changes the production process,
The results of the analysis show that MSEs in a developing country but it is also an engine of product quality improvement. In the context of
context can benefit in their innovation performance from the use of digitalization, firms are more likely to have access to knowledge for
these digital communication technologies and that innovation condi developing new products or implementing new organization methods.
tional on their use increases the level of labor productivity. In MSEs, (iii) By exploring MSEs' adoption of accessible digital communication
where the owner-manager plays a central role in making strategic de technologies, we raise important issues related to their “readiness” that
cisions, this finding supports the existing prioritization of digital tech can inform broader debates about achieving an inclusive fourth indus
nology adoption in South Africa as an important driver of firm trial revolution in South Africa (Mazibuko-Makena and Kraemer-Mbula,
performance. 2020).
However, despite the increased affordability and improved access to There are several ways in which the results of this study could be
the internet and related digital technologies, there remains a “digital usefully extended. Firstly, we have only examined the effect of a limited
divide” (UNCTAD, 2019). Larger and technologically sophisticated firms range of digital communication technologies that are accessible to MSEs.
are more likely to have better access to digital resources. Although Other technologies that are highlighted in the literature on digitalization
digital technologies have positive effects on MSEs' innovation and pro include cloud computing and the use of services available on digital
ductivity, barriers to access remain and tend to be more pervasive in the platforms. Secondly, the analysis could be extended to larger pop
context of developing countries. They typically include a combination of ulations of MSEs, including service sector firms which are some of the
a lack of high-quality and affordable infrastructure (including access to most active users of online digital services. This points to the need for a
reliable electricity), a shortage of digital skills, high costs and poor ac large-scale measurement program in Africa and other developing
cess to financing for smaller firms, and relatively high cost of data country regions designed to investigate the adoption and effects of
(Chege and Wang, 2020). These barriers can be ameliorated through digital communication technologies, which account for most firms and a
policy strategies that deliberately address the needs of MSEs, especially large share of employment. Thirdly, the survey design is cross-sectional,
those operating informally which are the most vulnerable. and we only study the relationships between the use of digital
There is a considerable debate about the impact of digitalization and communication technologies, innovation performance and productivity
Industry 4.0 technologies. A major focus in the literature on Industry 4.0 in 2019, but the analysis could be extended to a longer time period. With
has been around automation and the use of cyber-physical systems based more appropriate multi-year dataset, an interesting question to explore
on the use of advanced robotics in combination with big data and arti would be the nature of possible time lags in the effect of the introduction
ficial intelligence resulting in the development of “smart factories”. of digital technologies on innovation and productivity in South African
These technologies typically involve large capital investments and are manufacturing firms.
adapted to the needs of large firms engaged in large scale production,
notably in sectors like automobiles, chemicals and plastics and con CRediT authorship contribution statement
sumer electronics. The limited existing survey evidence available for
developing countries shows these advanced manufacturing technologies The three co-authors contributed equally to the development of the
are only adopted to a limited extent and rarely in small firms (Kupfer Conceptualization, Writing - original draft, Writing - review & editing.
et al., 2019).11 The debate on the adoption of Industry 4.0
manufacturing technologies, however, ignores the wider impact of the Acknowledgments
current digitalization process underway in developing countries which
involves the use of technologies based on mobile telephones and the This work has been partially supported by the National Research
internet for accessing and exchanging knowledge that involve smaller Foundation of South Africa (Grant Number: 118873) through the DSI/
capital outlays and are within reach for MSEs. NRF/Newton Fund Trilateral Chair in Transformative Innovation, the
A fundamental question that emerges from the analysis is how gov Fourth Industrial Revolution and Sustainable Development, at the Uni
ernments can use these findings to guide the design of public programs versity of Johannesburg. The survey used in this paper was conducted
aimed at fostering digital-technology adoption among MSEs? This is under the project ‘Community of Practice in Innovation and Inclusive
particularly important at a time when much of the post-pandemic re Industrialisation’ (grant number 110691), hosted by the DSI/NRF South
covery hinges on the ability of MSEs to survive and be productive, taking African Research Chair in Industrial Development (grant number
advantage of the accelerated digitalization trends. Public policy needs to 98627), University of Johannesburg. We are grateful to the DSI/NRF
be concerned with the types of digital technologies that are accessible to South African Research Chair in Industrial Development for giving us
MSEs so that policy interventions and limited public resources can be access to the data emerging from this survey.
most beneficial to firms that are in the most vulnerable circumstances,
11
Kupfer et al. (2019) summarize the results of recent surveys undertaken by UNIDO on the adoption of advanced manufacturing technology is Argentina, Brazil,
Ghana, Viet Nam, and Thailand. The results show that at most 1.5 % of small firms (defined as firms with less that 100 employees) have adopted advanced technology
involving the use of robots, artificial intelligence, and big data.
10
Appendix A
Table A1
Sample description of firms according to sector for size, age, capital intensities and digital communication technologies.
Number of Number of Mean of Share of Share of Share of each digital communication technology
firms (share) employees (share) age fixed capital ICT capital
Social Internet Mobile Online
media surf customer sales
HT sectors
Manufacture of chemicals 23 (3.2) 305 (5.7) 20 1.4 3.4 4.1 10.0 4.5 6.5
Manufacture of pharmaceuticals 10 (1.4) 39 (0.7) 24 0.04 0.1 0.8 2.0 0.5 0.0
Manufacture of electrical 5 (0.7) 22 (0.4) 36 0.3 0.1 0.4 2.0 1.0 1.1
equipment
Manufacture of machinery and 6 (0.8) 57 (1.1) 26 1.1 7.1 0.8 2.0 1.0 1.1
equipment
Manufacture of motor vehicles 1 (0.1) 8 (0.2) 15 0.0 0.0 0.0 0.0 0.0 0.0
Other manufacturing (includes 18 (2.5) 205 (3.9) 12 0.1 0.9 4.1 8.0 4.0 2.2
jewelry, musical instruments, etc.)
Repair and installation of 2 (0.3) 6 (0.1) 16 0.04 0.02 0.0 0.0 1.0 0.0
machinery and equipment
LT sectors
Manufacture of food products 29 (4.1) 383 (7.2) 20 2.6 16.7 6.2 8.0 4.0 2.2
Manufacture of beverages 1 (0.1) 100 (1.9) 12 0.003 0.05 0.4 0.0 0.0 1.1
Manufacture of tobacco products 1 (0.1) 25 (0.5) 7 0.0 0.0 0.4 0.0 0.0 0.0
Manufacture of textiles 72 (10.1) 390 (7.3) 11 1.3 3.4 12.4 10.0 12.1 11.0
Manufacture of wearing apparel 196 (27.6) 666 (12.5) 11 15.2 24.5 29.0 18.0 29.1 19.6
Manufacture of leather and related 34 (4.8) 126 (2.4) 18 0.1 0.3 3.7 4.0 2.0 4.3
products
Manufacture of wood 46 (6.5) 414 (7.8) 14 4.6 28.1 8.3 14.0 5.5 6.5
Manufacture of paper 6 (0.8) 87 (1.6) 18 0.3 0.6 1.2 0.0 1.5 5.4
Printing and reproduction of 13 (1.8) 212 (4.0) 21 0.6 2.1 3.3 4.0 2.0 5.4
recorded media
Manufacture of coke and refined 1 (0.1) 4 (0.1) 39 0.01 0.2 0.0 0.0 0.5 0.0
petroleum products
Manufacture of rubber and plastic 10 (1.4) 129 (2.4) 20 12.3 1.4 2.9 0.0 1.5 4.3
products
Other non-metallic mineral 23 (3.2) 188 (3.5) 14 6.2 1.1 3.3 0.0 3.5 2.2
products
Manufacture of basic metals 67 (9.4) 397 (7.5) 12 0.6 0.4 3.3 2.0 11.1 2.2
Manufacture of fabricated metal 20 (2.8) 205 (3.9) 17 2.0 4.8 1.2 4.0 3.0 2.2
products
Manufacture of furniture 127 (17.9) 1349 (25.4) 13 51.3 4.9 14.0 12.0 12.1 22.8
Total 711 (100) 5317 (100) – 100 100 100 100 100 100
Source: MSE Survey, authors' calculations.
References Commander, S., Harrison, R., Menezes-Filho, N., 2011. ICT and productivity in
developing countries: new firm-level evidence from Brazil and India. Rev. Econ. Stat.
93 (2), 528–541.
Aboal, D., Tacsir, E., 2018. Innovation and productivity in services and manufacturing:
Crepon, B., Duguet, E., Mairesse, J., 1998. Research, innovation, and productivity: an
the role of ICT. Ind. Corp. Chang. 27 (2), 221–241.
econometric analysis at the firm level. J. Econ. Innov. New Technol. 7 (2), 115–158.
African Union, 2020. The Digital Strategy for Africa 2020-2030, Addis Ababa, Ethiopia.
Crespi, G., Peirano, F., 2007. Measuring innovation in Latin America: what we did, where
Alvarez, R., 2016. The impact of R&D and ICT investment on innovation and
we are and what we want to do. In: Paper for the Conference on Micro Evidence on
productivity in Chilean firms. In: Inter-American Development Bank Technical Note
Innovation in Developing, UNU-MERIT.
Series: Washington DC, IDB-TN-1056.
Crespi, G., Tacsir, E., Vargas, F., 2014. Innovation and productivity in services: empirical
Autor, D.H., Levy, F., Murnane, R.J., 2003. The skill content of recent technological
evidence from Latin America. In: UNU-MERIT, Maastricht Economic and Social
change: an empirical exploration. Q. J. Econ. 118 (4), 1279–1333.
Research and Training Centre on Innovation and Technology.
Ballestar, M.T., Díaz-Chao, Á., Sainz, J., Torrent-Sellens, J., 2020. Knowledge, robots,
Crespi, G., Zuñiga, P., 2012. Innovation and productivity: evidence from six latin
and productivity in SMEs: explaining the second digital wave. J. Bus. Res. 108,
american countries. World Dev. 40 (2), 273–290.
119–131.
Dahlman, C., Mealy, S., Wermelinger, M., 2016. Harnessing the Digital Economy for
Bertschek, I., Cerquera, D., Klein, G.J., 2013. More bits–more bucks? Measuring the
Developing Countries. OECD Development Centre. Working Paper No. 334.
impact of broadband internet on firm performance. Inf. Econ. Policy 25 (3),
De Mel, S., McKenzie, D., Woodruff, C., 2009. Innovative firms or innovative owners?
190–203.
Determinants of innovation in micro, small, and medium enterprises. In: World Bank
Bhimani, H., Mention, A.L., Barlatier, P.J., 2019. Social media and innovation: a
Policy Research Working Paper 4934.
systematic literature review and future research directions. Technol. Forecast. Soc.
Dong, J.Q., Wu, W., 2015. Business value of social media technologies: evidence from
Chang. 144, 251–269.
online user innovation communities. J. Strateg. Inf. Syst. 24 (2), 113–127.
Black, S.E., Lynch, L.M., 2001. How to compete: the impact of workplace practices and
Draca, M., Sadun, R., Van Reenen, J., 2009. Productivity and ICTs: a review of the
information technology on productivity. Rev. Econ. Stat. 83 (3), 434–445.
evidence. In: The Oxford Handbook of Information and Communication
Booyens, I., 2011. Are small, medium- and micro-sized enterprises engines of
Technologies, pp. 101–148.
innovation? The reality in South Africa. Sci. Public Policy 38 (1), 67–78.
Edeh, J.N., Acedo, F.J., 2021. External supports, innovation efforts and productivity:
Booyens, I., Molotja, N., Phiri, M.Z., 2013. Innovation in high-technology SMMEs: the
estimation of a CDM model for small firms in developing countries. Technol.
case of the new media sector in Cape Town. Urban Forum 24 (2), 289–306.
Forecast. Soc. Chang. 173, 121189.
Brandtzaeg, P.B., Haugstveit, I.M., Lüders, M., Følstad, A., 2016. How should
Fagerberg, J., Srholec, M., Verspagen, B., 2010. Innovation and economic development.
organizations adapt to youth civic engagement in social media? A lead user
In: Handbook of the Economics of Innovation, Vol. 2. North-Holland, pp. 833–872.
approach. Interact. Comput. 28 (5), 664–679.
Fu, X., Mohnen, P., Zanello, G., 2018. Innovation and productivity in formal and
Brynjolfsson, E., Hitt, L.M., 2002. Computing productivity: firm-level evidence. Rev.
informal firms in Ghana. Technol. Forecast. Soc. Chang. 131 (6), 315–325.
Econ. Stat. 85 (4), 793–808.
Goedhuys, M., 2007. The impact of innovation activities on productivity and firm
Chege, S.M., Wang, D., 2020. Information technology innovation and its impact on job
growth: evidence from Brazil. In: UNU-MERIT Working Papers.
creation by SMEs in developing countries: an analysis of the literature review. Tech.
Griffith, R., Huergo, E., Mairesse, J., Peters, B., 2006. Innovation and productivity in four
Anal. Strat. Manag. 32 (3), 256–271.
european ccuntries. Oxf. Rev. Econ. Policy 22, 483–498.
11
Griliches, Z., 1979. Issues in assessing the contribution of research and development to Van Dijk, M.P., 2002. Innovation and small enterprise development examples from
productivity growth. Bell J. Econ. 92–116. Burkina Faso, Ghana and Zimbabwe. In: Innovation and Small Enterprises in the
Hall, B.H., 2011. Innovation and productivity. In: NBER Working Paper, No. 17178. Third World, Edward Elgar, Northampton, Massachusetts.
Hall, B.H., Lotti, F., Mairesse, J., 2008. R&D, innovation, and productivity: New evidence Van Leeuwen, G., Farooqui, S., 2008. ICT investment and productivity. In: Eurostat Final
from Italian manufacturing microdata. Ind. Corp. Chang. 17, 813–839. Report Information Society: ICT Impact Assessment by Linking Data From Different
Hatzichronoglou, T., 1997. Revision of the high-technology sector and product Sources, pp. 163–189. Working Paper.
classification. In: OECD Science, Technology and Industry Working Papers 1997/02. Wagner, J., 2006. International Firm Activities and Innovation: Evidence From
Hitchen, E.L., Nylund, P.A., Ferràs, X., Mussons, S., 2017. Social media: open innovation Knowledge Production Functions for German Firms. University of Lüneburg,
in SMEs finds new support. J. Bus. Strateg. 38 (2), 21–29. Working Paper Series in Economics.
Jensen, M.B., Johnson, B., Lorenz, E., Lundvall, B.Å., 2007. Forms of knowledge and World Bank, 2019a. Africa’s Pulse, No. 19: Analysis of Issues Shaping Africa’s Economic
modes of innovation. Res. Policy 36 (5), 680–693. Future. World Bank, Washington, DC.
Kraemer-Mbula, E., Lorenz, E., Takala-Greenish, L., Jegede, O.O., Garba, T., World Bank, 2019b. South Africa Digital Economy Diagnostic. The World Bank Group,
Mutambala, M., Esemu, T., 2019. Are African micro-and small enterprises Washington D.C, 2010s-28.
misunderstood? Unpacking the relationship between work organisation, capability Wunsch-Vincent, S., Kraemer-Mbula, E., 2016. The Informal Economy in Developing
development and innovation. Int. J. Technol. Learn. Innov. Dev. 11 (1), 1–30. Nations: Hidden Engine of Innovation? Cambridge University Press, Cambridge, UK.
Kupfer, D., Ferraz, J.C., Torracca, J., 2019. In: A comparative Analysis on Digitalization Zulu-Chisanga, S., Chabala, M., Mandawa-Bray, B., 2021. The differential effects of
in Manufacturing Industries in Selected Developing Countries: Firm-level Data on government support, inter-firm collaboration and firm resources on SME
Industry 4.0. Department of Policy, Research and Statistics, UNIDO, Vienna, p. 16. performance in a developing economy. J. Entrep. Emerg. Econ. 13 (2), 175–195. In
Working Paper. press.
Lööf, H., Heshmati, A., 2006. On the relationship between innovation and performance: a
sensitivity analysis. Econ. Innov. New Technol. 15 (4–5), 317–344.
Cyrielle Gaglio is a postdoctoral researcher in Economics. She holds the position of
Lööf, H., Mairesse, J., Mohnen, P., 2017. CDM 20 years after. Econ. Innov. New Technol.
Visiting Researcher at the University of Helsinki (Finland), Department of Economics and
26 (1–2), 1–5.
Management. She is also Affiliated Researcher at Sciences Po-OFCE (France), Department
Lundvall, B.Å., 1988. Innovation as an interactive process: from user-producer
of Innovation and Competition. She was awarded a PhD in Economics from the University
interaction to the national system of innovation. In: Dosi, G., Freeman, C., Nelson, R.,
of Côte d’Azur, in 2017. She completed a postdoctoral fellowship at Sciences Po-OFCE
Silverberg, G., Soete, L. (Eds.), Technical Change and Economic Theory. Pinter
(2018-2019), a French research institute whose activities focus on economic research,
Publisher, London, and New York, 2010s-28.
forecasting and the evaluation of public policies. She earned a research fellow contract at
Lundvall, B.Å., 2010. National Systems of Innovation: Toward a Theory of Innovation
the University of Johannesburg (South Africa) under the Trilateral Chair in Transformative
and Interactive Learning, Vol. 2. Anthem press.
Innovation and two research fellow contracts at University of Helsinki under EIT Food
Mairesse, J., Robin, S., 2017. Assessing measurement errors in the CDM research-
projects (INNOFOOD - Innovating Food Value Chain Innovations, Increasing consumer
innovation-productivity relationships. Econ. Innov. New Technol. 26 (1–2), 93–107.
trust and support for the food supply chain and for food companies). Her main research
Mairesse, J., Mohnen, P., Kremp, E., 2005. The importance of R&D and innovation for
interests include international economics/trade, digital economy, and industrial
productivity: a reexamination in light of the 2000 French Innovation Survey. In:
economics.
Mairesse, J., Trajtenberg, M. (Eds.), Special issue of Annales d’Economie et de
Statistiques, 79-80 in Memory of Zvi Griliches, pp. 487–528.
Mazibuko-Makena, Z., Kraemer-Mbula, E., 2020. Leap 4.0 African Perspectives on the Erika Kraemer-Mbula is a Professor of Economics and is currently the Chair holder of the
Fourth Industrial Revolution. Jacana Publishers, Johannesburg. DSI/NRF/Newton Fund Trilateral Chair in Transformative Innovation, the Fourth Indus
Mohnen, P., Polder, M., Van Leeuwen, G., 2018. ICT, R&D and Organizational trial Revolution and Sustainable Development, based at the College of Business and Eco
Innovation: Exploring Complementarities in Investment and Production. No. nomics (University of Johannesburg). Trained as an Economist, she holds a Masters in
w25044. National Bureau of Economic Research. Science and Technology Policy by the Science and Policy Research Unit (University of
Muninger, M.I., Hammedi, W., Mahr, D., 2019. The value of social media for innovation: Sussex), and a doctorate in Development Studies from the University of Oxford. She spe
a capability perspective. J. Bus. Res. 95, 116–127. cialises in the analysis of innovation systems in connection to equitable and inclusive
OECD, 2019. Digitalisation and productivity: a story of complementarities. In: OECD development, and has done pioneering work on innovation in the African informal sector.
Economic Outlook 2019. OECD Publishing, Paris, France. She is a member of several research networks on innovation and development – such as
Oyelaran-Oyeyinka, B., McCormick, D., 2007. Industrial Clusters and Innovation Systems President of Globelics, advisory board member of Africalics and Steering Committee
in Africa. United Nations University Press, Tokyo. member of OpenAIR. Prof. Kraemer-Mbula is also a Visiting Professor at the UCL Institute
Polder, M., van Leeuwen, G., Mohnen, P., Raymond, W., 2010. Product, Process and for Innovation and Public Purpose (IIPP).
Organizational Innovation: Drivers, Complementarity and Productivity Effects.
CIRANO - Scientific Publications, 2010s-28.
Edward Lorenz is Emeritus Professor of Economics at the University of Côte d’Azur in
SEDA, 2016. The small, medium and micro enterprise sector of South Africa. In: Bureau
France and he holds the position of Distinguished Visiting Professor at the University of
for Economic Research, 1, pp. 1–35.
Johannesburg in South Africa. He is also Part-time Professor at the Aalborg University
South African Department of Telecommunications, 2017. National E-Strategy and ICT
Business School. He was awarded a BS in Economics from MIT, a MA in Economics from
SMME Support Strategy. Cape Town, South Africa.
the University of California, Berkeley, and a PhD in Economics from the University of
South African Department of Telecommunications, 2020. National Digital and Future
Cambridge in England. He is a member of the Scientific Board of the Globelics innovation
Skills Strategy. Cape Town, South Africa.
network and interim President of Eurolics, the European Branch of Globelics. Edward has
Srholec, M., 2011. A multilevel analysis of innovation in developing countries. Ind. Corp.
regularly provided expert advice to the OECD, EUROFOUND and the European Commis
Chang. 20 (6), 1539–1569.
sion and has contributed to numerous expert committees including the Measurement
Stiroh, K.J., 2005. Reassessing the impact of IT in the production function: a meta-
Working Group of the Directorate for Science, Technology and Industry at the OECD, the
analysis and sensitivity tests. Ann. Econ. Stat. 529–561.
OECD Working Party on SMEs and Entrepreneurship, and the expert group invited to
United Nations Conference on Trade and Development (UNCTAD), 2019. Digital
discuss the “Future of Europe’s innovation policies” by the DG Enterprise and Industry of
Economy Report 2019: Value Creation and Capture, Implications for Developing
the European Commission.
Countries. New York, UN.
12

1 s2.0 S0040162522003092 Main

Uploaded by

Copyright:

Available Formats

1 s2.0 S0040162522003092 Main

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

1 s2.0 S0040162522003092 Main

Uploaded by

Copyright:

Available Formats

Technological Forecasting & Social Change 182 (2022) 121785

Contents lists available at ScienceDirect

Technological Forecasting & Social Change

The effects of digital transformation on innovation and productivity:

development and stakeholder involvement in the adoption and use of

Fig. 3. The extended CDM model with digital communication technologies.

Productivityi = c+β1 Innovationi +β2 Log(sizei )+β3 Log(agei ) Table 1

definition of innovation can help account for what may appear to be an

Product innovation Process innovation

Social media 0.177* 0.172* 0.160*** 0.0391* 0.0325 0.0198

Product innovation 0.234** 0.221* 0.208* 0.205*

Note: Standard errors are given in parentheses.

You might also like