Promoting Economic Development Through Digitalisation: Impacts on Human Development, Economic Complexity, and Gross National Income

Abstract

The advancement of digital technologies has become a transformative driver of economic development. Digitalisation is central to the global economy, enhances productivity, drives innovation, and promotes inclusive growth. Despite this potential, South Africa faces persistent challenges such as skills shortages, unemployment, poverty, and socioeconomic inequality. This study investigates the role of digitalisation in advancing economic complexity, human capital development, and gross national income in South Africa. A digitalisation index, constructed through Principal Component Analysis, ARDL models, and Granger causality analysis, provides insights into the short- and long-term impacts and causal relationship. The findings reveal that digitalisation and education significantly enhance human capital development in the long run, with digital infrastructure also driving immediate gains. For the gross national income model, digitalisation and education pose short-term pressures due to development expenditures, while institutional quality plays an important role in sustaining income. Economic complexity benefits positively from digitalisation over the long term, though short-term impacts stress the role of governance quality and infrastructure. Causality analysis further shows the interconnectedness of these variables, with digitalisation advancing economic complexity and human capital driving national income, reinforcing digitalisation. The results call for policies that align short-term developmental priorities with long-term sustainability. Investments in digital infrastructure, accessible education, and institutional frameworks are critical for building a skilled labour force while enhancing economic complexity and maintaining financial stability.

1. Introduction

The rapid advancement of digital technologies has contributed to economic development in recent decades (Elfaki & Ahmed, 2024; Sithole et al., 2024). It has promoted transformative changes across sectors and created new economic and social progress forms. The shift towards digitalisation has become a central factor in the global economy, mainly because of its ability to increase productivity, facilitate innovation, and drive inclusive growth (Magoutas et al., 2024; Shi & Wei, 2024). Digitalisation can enhance human development and economic progress for emerging and developed economies by improving economic structures and potentially increasing national income (Abbas & Zaman, 2024; Abu Alfoul et al., 2024). Likewise, from the proliferation of internet access to the rise in artificial intelligence and big data, digitalisation is creating new phenomena in traditional industries, new business models, and enhancing connectivity (Ciasullo & Lim, 2022; Javaid et al., 2024). Digitalisation now stands as a force in the global economy by increasing productivity improvements and technological innovation (Rahman & Malik, 2023).

Several challenges hinder South Africa’s economic development, including a skills shortage, high unemployment, pervasive poverty, inadequate infrastructure, and socioeconomic inequality (Bernstein, 2014; Plagerson, 2023; Zizzamia, 2020). Since the end of apartheid in 1994, the country has faced slow economic progress, as shown by economic indicators. Various indicators measure economic development, such as GDP per capita, Gross National Income (GNI), the Human Development Index (HDI), and economic complexity (Ali et al., 2024; Dahliah & Nur, 2021; Ncanywa & Stuurman, 2018; Ogujiuba et al., 2024; Rivera et al., 2023). Between 1994 and 2021, the average South African HDI increased by a mere 0.4, rising from 0.64 in 1994 to 0.68 in 2021 (Ogujiuba et al., 2024; United Nations Development Programme [UNDP], 1995). Although economic reforms were expected to enhance living standards, most South Africans have not experienced improvements. Additionally, South Africa’s GNI per capita has fallen from 3.44 in 1994 to −5.67 in 2020 (Ncanywa & Masoga, 2018; World Bank, 2024). The persistent lack of skills contributes to high unemployment and low wages, further affecting the economic growth. In today’s digital age, digitalisation is increasingly recognised as vital for economic advancement, even amid sluggish growth rates in developing economies (Asaleye & Garidzirai, 2024; Shukla et al., 2023). Real-time information sharing can lead to cost income and minimise the risk of information loss (Sahin & Robinson, 2002). Therefore, it is essential to investigate the relationship between digitalisation and economic development in South Africa.

Consequently, organisations increasingly view digitalisation as a convenient and cost-effective solution, reflecting a shift from face-to-face interactions to virtual engagements (Jain, 2021; Mohamed Hashim et al., 2022). Digitalisation is the integration of digital technologies into the daily lives of millions, a trend that has gained prominence in recent decades (Asif et al., 2024). The literature identifies vital components of digitalisation, including broadband users, internet users, and mobile users (Charfeddine & Umlai, 2023; Jamil, 2021; Karacuka et al., 2024). A digital economy can enhance firm productivity through process automation (Karacuka et al., 2024), boosting competitiveness, and positively impacting employment (Ogunwole et al., 2024). Increased employment is often considered a crucial first step toward economic development (Ordeñana et al., 2024). However, some scholars caution that while digitalisation facilitates economic growth, it poses risks for unskilled workers who may face job displacement (Ernst et al., 2019; Jain, 2021). Despite this, digitalisation’s role in promoting overall welfare is recognised in the empirical literature (Alexopoulou, 2024; Felbermayr et al., 2017; Torres & Augusto, 2020).

The relevance and motivation of this study arise from three key points: its alignment with the United Nations Sustainable Development Goals (SDGs), the complexities of measuring economic development in the digital age, and a notable gap in the empirical literature on digitalisation’s wider socioeconomic impacts. Firstly, digitalisation aligns with several United Nations Sustainable Development Goals (SDGs), particularly Goal 8: Decent Work and Economic Growth, Goal 9: Industry, Innovation, and Infrastructure, and Goal 10: Reduced Inequalities (Mondejar et al., 2021; Vyas-Doorgapersad, 2022). Access to information creates new job opportunities and supports the growth of innovative industries; this digitalisation may contribute to more inclusive and sustainable economic progress. As countries increasingly prioritise the SDGs, understanding digitalisation’s role in achieving these objectives is crucial. This study contributes to the literature on how digitalisation can drive sustainable development by enhancing productivity, expanding access to services, and developing economic participation.

Secondly, measuring economic development in empirical studies has been a challenging issue (Slemrod et al., 1995; Wu & Davis, 1999). Traditional metrics such as GDP provide a limited view concerning digitalisation, often overlooking the qualitative improvements digitalisation brings to healthcare, education, and overall quality of life, which was used in the study by (Aleksandrova et al., 2022). Studies also argue that the increase in growth in African economies has not promoted employment or welfare, referred to as jobless growth (Butkus et al., 2024; Léautier & Hanson, 2013). Furthermore, conventional indicators may not capture shifts in economic complexity, innovation capacity, or the inclusivity of economic growth—key areas where digital transformation has substantial influence. This study recognises the need to examine digitalisation’s impact using multidimensional indicators, such as the gross national saving, human development Index and economic complexity, to provide a more comprehensive understanding of its role in promoting economic development (Shulin, 1999).

Finally, despite the recognised importance of digitalisation, there is a notable gap in the empirical literature regarding its effects on socioeconomic metrics, particularly HDI, economic complexity, and gross national income. Existing research often focuses on digitalisation’s role in enhancing productivity or GDP (Ahmad & Schreyer, 2016; Szabo et al., 2024); yet, few studies have examined its influence on human development (Dimian et al., 2023; Grigorescu et al., 2021; Thite, 2022; Vasilev et al., 2023). However, investigating how digitalisation can promote economic development using indicators such as economic complexity, human capital development index, and gross national income is still growing. This study addresses this gap by examining how digitalisation contributes to various dimensions of economic and social progress, suggesting new insights for policymakers aiming to maximise the benefit of digitalisation and promote economic development. Based on the preceding, this study aims to investigate the relationships between digitalisation, human development, economic complexity, and gross national income using an Autoregressive Distributed Lag approach and Granger causality analysis. Specifically, the specific objectives are as follows:

• To examine digitalisation’s short-run and long-run effects on human development, economic complexity, and gross national income.
• To analyse the causal relationships between digitalisation, human development, economic complexity, and GNI.

The study is structured as follows: after the introductory section, Section 2 presents the empirical literature, and Section 3 discusses the Empirical Models. Section 4 is the discussion and presentation of results. Section 5 concludes the study.

2. Literature Review

2.1. Theoretical and Empirical Review

2.1.1. Literature Review on Digitalisation and Human Capital

Theoretically, the connection between digitalisation and human capital has implications for productivity, economic growth, and inequality. Digitalisation changes the demand and supply structure for human capital by increasing the value of cognitive and digital skills while reducing the demand for manual tasks (Acemoglu & Autor, 2011). According to Human capital theory, investments in education and skills enhance individual productivity (Becker, 2009); however, rapid digital innovation has outpaced traditional education, creating a skills gap (Brynjolfsson & McAfee, 2014). The Skill-Biased Technological Change hypothesis posits that digital technologies favour high-skill workers, widening wage disparities as middle-skill jobs decline (Goldin & Katz, 2010; Mondolo, 2020). Digitalisation has caused a shift towards lifelong learning models to keep pace with evolving skill demands (Schmidt & Tang, 2020; Schwab, 2017). While theoretical perspectives have presented digitalisation’s negative and positive impacts on the economy, the consensus lies in its ability to promote long-run aggregate welfare.

Empirical studies stress the transformative impact of digitalisation on human capital and the labour market. For instance, (Acemoglu & Restrepo, 2018) found that automation and digital technologies have a dual effect, boosting productivity while simultaneously displacing routine jobs in sectors such as manufacturing and administration. This phenomenon has led to “employment inequality”, where demand for high-skill roles in technology-intensive sectors grows while middle-skill jobs decline (Autor & Dorn, 2013). (Balsmeier & Woerter, 2019) reported digitalisation increases the wage premium for workers with digital and cognitive skills, creating a more competitive and unequal labour market. Studies focusing on developing economies similarly reveal that digitalisation widens skill disparities, with limited digital infrastructure and education systems struggling to meet the demand for specialised digital skills (Dahlman et al., 2016; Jamil, 2021).

2.1.2. Literature Review on Digitalisation and Economic Complexity

Theoretically, the economic complexity theory documented that the diversity and knowledge intensity of a country’s exports suggest that nations with more complex economies tend to have greater innovation and growth potential (Balland et al., 2022; Hausmann & Hidalgo, 2011). Digitalisation contributes to economic complexity through knowledge transfer, enhancing productivity, and enabling the development of more products and services (Ordieres-Meré et al., 2020). Through digital platforms, data analytics, and automation, countries can improve production efficiencies and expand into knowledge-intensive industries, which increases the variety and complexity of goods produced. Digitalisation reduces entry barriers for new sectors by lowering costs and creating a scalable business environment; this may enable developing economies to diversify their exports and increase complexity. In addition, digitalisation improves global value chains and offers more interconnectedness between industries (Szalavetz, 2019).

Empirical studies reported the positive impact of digitalisation on economic complexity through the production processes and diversification of exports (Ha, 2023; J. Li et al., 2024; Oumbé et al., 2023). Matthess and Kunkel (Matthess & Kunkel, 2020) showed that increased internet penetration and digital adoption correlate strongly with shifts toward more complex, knowledge-intensive exports, particularly in developing nations. Similarly, (Götz, 2020) found that countries with digital infrastructure produce more developed goods and attract greater foreign direct investment in high-tech industries. In a study on Latin America, Sanchez-Riofrio et al. (Sanchez-Riofrio et al., 2022) observed that countries with higher digital penetration rates diversified more rapidly into complex industries.

Other empirical research documented that the influence of digitalisation on economic complexity may depend on factors such as education, infrastructure, and institutional quality (C. Li et al., 2023; Zaborovskaia et al., 2020). (Grigorescu et al., 2021) observed that while digitalisation supports economic complexity, its impact is significantly enhanced in countries with higher human capital levels, as skilled employment is critical for adopting and integrating new technologies. Additionally, Gruber (Gruber, 2019) pointed out that investments in digital infrastructure alone may be insufficient if not complemented by policies supporting innovation and skill development.

2.1.3. Literature Review on Digitalisation and National Income

Theoretically, digitalisation is recognised as a driver of national income growth through productivity, innovation, and economic efficiency. Theory suggests that digitalisation enhances national income by lowering transaction costs, enabling faster information flows, and facilitating the development of high-productivity industries (Mayer, 2021). As economies adopt digital technologies, they experience productivity gains, which enhances output and economic growth (Dedrick et al., 2003). The Solow growth model traditionally centred on capital and labour, has been extended to include technological progress as a key determinant of long-term growth, with digitalisation as a component (Solow, 1956). Furthermore, digitalisation can contribute to increased income through innovation, as access to digital facilities enables firms to cover more markets and innovate more effectively (Rachinger et al., 2019). Consequently, as digitalisation spreads, it promotes higher national income levels by supporting more efficient production and encouraging knowledge-intensive, high-value economic activities (Myovella et al., 2020).

Empirical studies demonstrate a positive relationship between digitalisation and national income. For example, (Hernandez et al., 2016) documented that increased internet penetration and mobile technology usage significantly correlated with higher GDP per capita, as digitalisation enhanced productivity and economic diversification. Similarly, Otarinia (Amin Otarinia, 2024) concludes that digital infrastructure investments substantially impact national income, with developing nations benefiting from these gains due to the relatively low baseline of digital integration. Likewise, (Sarangi & Pradhan, 2020) provided evidence that broadband adoption has led to a measurable increase in GDP growth rates, mainly by promoting efficiency in business operations and enabling rapid access to information essential for high-value economic activities.

Numerous studies showed that the impact of digitalisation on national income depends on supporting factors, such as education, regulatory frameworks, and infrastructure quality (Billon et al., 2010; Boikova et al., 2021; Paul et al., 2020). For instance, (Vasilescu et al., 2020) showed that higher educational levels lead to more significant income benefits from digitalisation, as a skilled worker can better use digital tools. Additionally, studies showed that countries with stable regulatory environments maximise digitalisation’s income-enhancing effects, as these conditions promote innovation and attract foreign investment (Vitkovics, 2023). Further evidence from developing economies suggests that digitalisation’s impact on national income is enhanced when combined with investments in complementary infrastructure, such as reliable electricity and transportation systems (Matthess & Kunkel, 2020).

2.2. Gaps Identified in Empirical and Methodological Approaches

While significant research has investigated the impact of digitalisation on various aspects of the economy, the majority have concentrated on the effects of digital technologies on growth and productivity (Amin Otarinia, 2024; Boikova et al., 2021; Myovella et al., 2020; Rahman & Malik, 2023; Sarangi & Pradhan, 2020; Szabo et al., 2024). Another major research focus has been the effect of digitalisation on employment (Acemoglu & Autor, 2011; Balsmeier & Woerter, 2019; Fietz & Lay, 2023; Karacuka et al., 2024). However, (Oumbé et al., 2023) extended this scope by analysing the long-term relationships among ICT, economic complexity, GDP per capita, government spending, and natural resources. Despite the study’s comprehensive approach, it was limited to examining economic complexity in a panel data framework without assessing the simultaneous impacts on human development and gross national income, essential for an all-inclusive view of digitalisation’s role in socioeconomic transformation.

Furthermore, studies by (J. Li et al., 2024) and (C. Li et al., 2023) have investigated the relationship between natural resources, economic complexity, digitalisation, and growth, while others have examined links between digitalisation, economic complexity, and energy (Ali et al., 2024; Charfeddine & Umlai, 2023; Elfaki & Ahmed, 2024; Ha, 2023). However, most of these studies tend to isolate one dimension—economic complexity, gross national income, or human capital development—without including these factors to understand better how digitalisation can drive economic transformation and promote economic development. An approach that compares these dimensions could yield new insights into how digitalisation supports development. Studies examining the economic channels of digitalisation and technological innovation have analysed short-run and long-run effects as well as causal relationships (Asaleye & Garidzirai, 2024; Ncanywa & Stuurman, 2018; Ogunwole et al., 2024; Rivera et al., 2023). Based on this, the study uses the autoregressive distributed lag (ARDL) model and Granger causality analysis to achieve the objectives. The study captures both short- and long-term dynamics, as well as causal relationships between digitalisation and socioeconomic variables; this methodological choice is supported by the preliminary analysis, which reveals mixed integration orders in the data. The importance of ARDL is due to its flexibility to handle such series (Obadiaru et al., 2018).

This study includes education, infrastructure, and institutional quality as control variables due to their important roles in human development index, economic complexity, and gross national income. Education builds human capital and raises productivity, enhancing human development, driving innovation, and promoting economic growth (Arisukwu et al., 2019; Asaleye et al., 2021; Barro & Lee, 2013). The infrastructure supports trade, connectivity, and service delivery, facilitating higher productivity and market access and boosting the human development index, economic complexity, and gross national income (Calderón & Servén, 2010). Lastly, institutional quality promotes a stable environment, reduces corruption, and supports efficient governance, crucial for sustained economic development and complex economic activities (Acemoglu et al., 2001).

We construct a digitalisation index by applying the Principal Component Analysis (PCA) using three indicators: mobile cellular subscriptions, individuals using the internet, and fixed telephone subscriptions. Standardisation of each indicator was conducted to ensure comparability, as suggested by Jolliffe and Cadima (Jollife & Cadima, 2016). We then conducted PCA, reviewed the eigenvalues, and explained the variance of each component, where the first two components (PC1 and PC2) captured a significant portion of the total variance, indicating the nature of digitalisation (Jollife & Cadima, 2016). Previous studies show the PCA’s effectiveness in summarising correlated variables into principal components improves interpretability and reduces multicollinearity (Abdi & Williams, 2010). Accordingly, we saved PC1 and PC2 as composite digitalisation variables and used both components in our model to represent digitalisation (Abdi & Williams, 2010). More information about the series used in this study is provided in

Table 1. Information about the series used in the study.

Variables	Definition	Source
ECP	Economic Complexity (Economic Complexity Index computed using HS product classification)	Atlas of Economic Complexity—Country Complexity Rankings, Harvard University.
HDI	Human Development Index	UNDP Human Development report.
GNI	Gross National Income	World Bank Data (WB) and OECD
PC1	First Indicator of Digitalisation (Digitalisation index computed using principal component analysis)	Authors computed using individual using the internet, mobile cellular Subscription and fixed telephone subscription—data from WB.
PC2	Second Indicator of Digitalisation (Digitalisation index computed using principal component analysis)	Authors computed using individual using the internet, mobile cellular Subscription and fixed telephone subscription—data from WB.
EDU	Education (Government expenditure on education, percentage of GDP).	UNESCO Institute for Statistics.
INF	Infrastructure (Air transport, registered carrier departures worldwide)	International Civil Aviation Organisation, Civil Aviation Statistics.
INS	Institution Quality (Control of corruption)	Worldwide Governance Indicator

Source: Authors’ Computation.

.

3. Empirical Models

Three models were adopted to estimate economic development measures to meet the study’s objectives as follows:

$$ECP=f(PC1,PC2,EDU,INF,INS)$$

(1)

$$HDI=f(PC1,PC2,EDU,INF,INS)$$

(2)

$$GNI=f(PC1,PC2,EDU,INF,INS)$$

(3)

In Equations (1)–(3), $ECP$ is economic complexity, $HDI$ is human capital development index and $GNI$ is gross national income. The digitalisation indicators are $PC1$ and $PC2$, which are computed using Principal Component Analysis from three indicators as follows: mobile cellular subscriptions, individuals using the internet, and fixed broadband subscriptions (Abdi & Williams, 2010; Jollife & Cadima, 2016). The control variables are education ($EDU$), infrastructure ($INF$), and institutional quality ($INS$), which are selected after a thorough investigation of the existing literature on the determinants of economic complexity, human capital development, and gross national income (Barro & Lee, 2013; Calderón & Servén, 2010). Likewise, (Ncanywa & Stuurman, 2018) argue that gross national income (GNI) provides a more comprehensive measure of economic development than GDP.

Equations (1)–(3) are explicitly written as follows:

$$EC{P}_t={\alpha }_0+{\alpha }_{1}PC1+{\alpha }_{2}PC2+{\alpha }_{3}EDU+{\alpha }_{4}INF+{\alpha }_{5}INS+{\epsilon }_t$$

(4)

$$HD{I}_t={\beta }_0+{\beta }_{1}PC1+{\beta }_{2}PC2+{\beta }_{3}EDU+{\beta }_{4}INF+{\beta }_{5}INS+v_t$$

(5)

$$GN{I}_t={\chi }_0+{\chi }_{1}PC1+{\chi }_{2}PC2+{\chi }_{3}EDU+{\chi }_{4}INF+{\chi }_{5}INS+e_t$$

(6)

In Equations (4)–(6), the respective intercepts are ${\alpha }_0$, ${\beta }_0$ and ${\chi }_0$ where respective parameters are ${\alpha }_1,{\alpha }_2,\dots ,{\alpha }_5$ for Equation (4), ${\beta }_1,{\beta }_2,\dots ,{\beta }_5$ for Equation (5) and ${\chi }_1,{\chi }_2,\dots ,{\chi }_5$ for Equation (6). The period of observation is denoted by ‘$t$’. The respective error terms are ${\epsilon }_t$, $v_t$ and $e_t$. The study uses ARDL to achieve the first objective. The ARDL (p, q) models are given as follows:

$$EC{P}_t=\alpha +{\displaystyle \sum _{i=1}^p{\varphi }_{1}EC{P}_{t-i}+{\displaystyle \sum _{j=0}^{q_1}{\varphi }_{2}PC{1}_{t-j}}}+{\displaystyle \sum _{k=0}^{q_2}{\varphi }_3}PC{2}_{t-k}+{\displaystyle \sum _{l=0}^{q_3}{\varphi }_4}ED{U}_{t-l}+{\displaystyle \sum _{m=0}^{q_4}{\varphi }_5}IN{F}_{t-m}+{\displaystyle \sum _{n=0}^{q_5}{\varphi }_6}IN{S}_{t-n}+{\epsilon }_t$$

(7)

$$HD{I}_t=\beta +{\displaystyle \sum _{i=1}^p{\psi }_{1}EC{P}_{t-i}+{\displaystyle \sum _{j=0}^{q_1}{\psi }_{2}PC{1}_{t-j}}}+{\displaystyle \sum _{k=0}^{q_2}{\psi }_3}PC{2}_{t-k}+{\displaystyle \sum _{l=0}^{q_3}{\psi }_4}ED{U}_{t-l}+{\displaystyle \sum _{m=0}^{q_4}{\psi }_5}IN{F}_{t-m}+{\displaystyle \sum _{n=0}^{q_5}{\psi }_6}IN{S}_{t-n}+v_t$$

(8)

$$GN{I}_t=\chi +{\displaystyle \sum _{i=1}^p{\lambda }_{1}EC{P}_{t-i}+{\displaystyle \sum _{j=0}^{q_1}{\lambda }_{2}PC{1}_{t-j}}}+{\displaystyle \sum _{k=0}^{q_2}{\lambda }_3}PC{2}_{t-k}+{\displaystyle \sum _{l=0}^{q_3}{\lambda }_4}ED{U}_{t-l}+{\displaystyle \sum _{m=0}^{q_4}{\lambda }_5}IN{F}_{t-m}+{\displaystyle \sum _{n=0}^{q_5}{\lambda }_6}IN{S}_{t-n}+e_t$$

(9)

From Equations (7)–(9), $p,q,q_1,q_{2,}\dots ,q_5$ represent the lag lengths for the dependent and each of the independent variables, respectively. The parameters for respective variables are ${\varphi }_1,{\varphi }_{2,}\dots ,{\varphi }_6$, ${\psi }_1,{\psi }_{2,}\dots ,{\psi }_6$ and ${\lambda }_1,{\lambda }_{2,}\dots ,{\lambda }_6$ in Equations (7), (8), and (9), respectively.

In an ARDL framework, the short-run model with an Error Correction Model (ECM) represents the immediate, short-term relationship among variables while incorporating an error correction term to adjust for deviations from the long-run equilibrium. This ECM term indicates the speed at which the dependent variable returns to equilibrium after a shock to one of the independent variables. For a general model where $Y$ is the dependent variable and $X_i,X_2,\dots ,X_6$ are the independent variables, the short-run ECM equation can be expressed as follows:

$$\Delta Y_t=a+{\displaystyle \sum _{i=1}^p{b}_i\Delta Y_{t-i}}+{\displaystyle \sum _{j=0}^q{c}_j\Delta X_{1,t-j}}+{\displaystyle \sum _{k=0}^r{f}_k\Delta X_{2,t-k}}+\dots +{\displaystyle \sum _{m=0}^s{g}_m}\Delta X_{6,t-m}+\Phi EC{M}_{t-1}+v_t$$

(10)

where $\Delta$ represents first differences, capturing short-run changes in the variables. $a$ is the constant term, $b_i,c_j,f_k$ and $g_m$ represent short-run coefficients for the lagged differences in the dependent variable and independent variables. $EC{M}_{t-1}$ is the s the lagged error correction term derived from the long-run cointegrating relationship. $\Phi$ is the adjustment coefficient, which measures the speed at which the system corrects itself back to the long-run equilibrium, and it must be negative and significant to show adjustment toward long-run equilibrium. $v_t$ is the error term.

In selecting the optimal lag length for an ARDL model with 112 quarterly observations, three primary information criteria such, the Akaike Information Criterion (AIC), Schwarz Bayesian Information Criterion (BIC), and Hannan–Quinn Criterion (HQIC), are mostly considered due to the balances between model complexity and fit. AIC, often used in smaller samples, favours additional parameters to enhance model accuracy by capturing more dynamics (Akaike, 1974). BIC, in contrast, imposes a stricter approach to the model, resulting in a simpler model that helps avoid overfitting (Schwarz, 1978), while HQIC compromises the two (Hannan & Quinn, 1979). AIC was chosen for this study because of its suitability in capturing detailed data dynamics in quarterly observations.

The second objective is achieved by employing the Granger non-causality approach. In the presence of mixed integration orders, we utilise the Toda–Yamamoto Granger non-causality approach, which is an alternative to traditional Granger causality tests, particularly when variables may exhibit different integration properties (Toda & Yamamoto, 1995). This method avoids the potential issues of spurious regression and pre-testing biases by augmenting the lag order of a VAR model by the maximum integration order. Therefore, this model can be expressed as follows:

$$A_t={\alpha }_0+{\displaystyle \sum _{i=1}^{k+d_{\max }}{\alpha }_i{A}_{t-i}}+{\displaystyle \sum _{j=1}^{k+d_{\max }}{\beta }_j{C}_{t-j}}+v_t$$

(11)

$$C_t={\beta }_0+{\displaystyle \sum _{i=1}^{k+d_{\max }}{\delta }_i{A}_{t-i}}+{\displaystyle \sum _{j=1}^{k+d_{\max }}{\theta }_j{C}_{t-j}}+e_t$$

(12)

We then conduct a modified Wald test on the coefficients of the first (k) lags of the independent variable. The null hypothesis posits that these coefficients are jointly zero ${\beta }_1={\beta }_2=\dots ={\beta }_k=0$, suggesting that does not Granger-cause ‘C’ does not Granger-cause ‘A’ (Zapata et al., 1996). Rejection of the null hypothesis indicates causality, while non-rejection implies that ‘C’ does not Granger-cause ‘A’. Preliminary analyses, such as correlation analysis, descriptive statistics, and unit root tests, were carried out. The outcome of the unit root test motivates the techniques used to achieve the two objectives in this study.

Table 1 provides information about the series used in this study. As mentioned in Section 2, we created a digitalisation index using Principal Component Analysis on three indicators: mobile cellular subscriptions, internet usage, and fixed telephone subscriptions. Each indicator was standardised for comparability. PCA results showed that the first two components (PC1 and PC2) captured a significant portion of the variance, effectively representing digitalisation. Based on PCA’s ability to summarise correlated variables, improving interpretability and reducing multicollinearity, PC1 and PC2 were used for digitalisation variables. The data are from 1996 to 2023. The period selected for analysis is based on data availability. Although sample data are provided in yearly frequencies, the limited observations pose a small-sample issue. The original time series’ short duration (with annual data) may not effectively capture potential shifts in long-run relationships (Johansen, 2002). To address this, we converted the yearly data into quarterly frequencies using the quadratic match-sum method, enhancing the analysis of cointegration among economic complexity, the human development index, gross national income, and digitalisation; this method has been widely applied in empirical studies (Asaleye et al., 2023; Kisswani & Harraf, 2021; Türsoy & Faisal, 2018).

4. Presentation of Result and Discussion

4.1. Preliminary Analyses Results

Table 2. Correlation Analyses Result.

ECP Model
	ECP	PC1	PC2	EDU	INF	INS
ECP	1
PC1	−0.9069	1
PC2	0.3548	−0.4362	1
EDU	−0.7262	0.8541	−0.4214	1
INF	−0.1548	0.1652	0.5795	0.0195	1
INS	0.838	−0.9146	0.2638	−0.773	−0.2179	1
HDI Model
	HDI	PC1	PC2	EDU	INF	INS
HDI	1
PC1	0.9289	1
PC2	−0.1782	−0.4362	1
EDU	0.8704	0.8541	−0.4214	1
INF	0.3983	0.1652	0.5795	0.0195	1
INS	−0.8847	−0.9146	0.2638	−0.773	−0.2179	1
GNI Model
	GNI	PC1	PC2	EDU	INF	INS
GNI	1
PC1	−0.5544	1
PC2	0.0981	−0.4362	1
EDU	−0.6967	0.8541	−0.4214	1
INF	−0.2029	0.165	0.5795	0.0195	1
INS	0.5618	−0.9146	0.2638	−0.773	−0.2179	1

Source: Authors’ Computation.

presents the correlation matrix for the variables in this study. Economic Complexity (ECI) demonstrates a strong inverse correlation with the first digitalisation index (PC1) at −0.9069 and education (EDU) at −0.7262 while exhibiting positive correlations of 0.3548 with the second digitalisation index (PC2), −0.1548 with infrastructure (INF), and 0.838 with institutional quality (INS). The first digitalisation index shows correlations of −0.4362, 0.8541, 0.1652, and −0.9146, with the second digitalisation index, education, infrastructure, and institutional quality, respectively. The second digitalisation index is positively correlated with education (0.8541) and infrastructure (0.1652) but is inversely related to institutional quality (−0.9146). Education and infrastructure are weakly correlated (0.0195), while education and institutional quality are inversely correlated at −0.773. The infrastructure and institutional quality exhibit a modest inverse relationship (−0.2179). The Human Development Index (HDI) correlates strongly with the first digitalisation index (0.9289), education (0.8704), and moderately with infrastructure (0.3983) but has inverse associations with the second digitalisation index (−0.1782) and institutional quality (−0.8847). Finally, gross national income (GNI) has notable correlations with the first digitalisation index (−0.5544), education (−0.6967), infrastructure (0.2029), and institutional quality (0.5618), while showing a weak positive relationship with the second digitalisation index (0.0981).

Table 3. Descriptive Statistics.

	ECP	HDI	GNI	PC1	PC2	EDU	INF	INS
Mean	0.2975	0.6786	1.1963	0.3742	0.178	0.7154	5.1732	1.7726
Median	0.3073	0.6812	1.184	−0.1732	0.2742	0.7241	5.1728	1.7761
Maximum	0.5756	0.7424	1.2787	3.469	0.9508	0.8178	5.4600	1.8763
Minimum	−0.0009	0.6285	1.1248	−1.91	−0.7574	0.6367	4.8811	1.6322
Std. Dev.	0.1591	0.0403	0.046	1.6689	0.428	0.0521	0.1359	0.0591
Obs.	104	104	104	104	104	104	104	104

Source: Authors’ Computation.

summarises the descriptive statistics for variables used in this analysis. Mean values are as follows: economic complexity at 0.2975, human development index at 0.6786, gross national income at 1.1963, first digitalisation index at 0.3742, second digitalisation index at 0.178, education at 0.7154, infrastructure at 5.1732, and institutional quality at 1.7726. Standard deviations indicate variability across these measures, with economic complexity at (0.1591), human development index (0.0403), gross national income (0.0460), first digitalisation index (1.6689), second digitalisation index (0.4280), education (0.0521), infrastructure (0.1359), and institutional quality (0.0591). Median values show central tendencies close to mean values, with economic complexity at 0.3073, human development index at 0.6812, gross national income at 1.184, first digitalisation index at −0.1732, second digitalisation index at 0.2742, education at 0.7241, infrastructure at 5.1728, and institutional quality at 1.7761. Due to missing data, observations range from 112 to 104.

Table 4. Stationarity Test Result.

Level
	Augmented Dickey–Fuller			Phillips–Perron
	Intercept	Trend and Intercept	None	Intercept	Trend and Intercept	None
ECP	−0.1617	−3.9520 **	−1.4969	−0.4447	−2.5419	−1.4879
HDI	−0.4144	−2.5777	1.2524	0.0586	−2.2926	1.5653
GNI	−1.8647	−2.6206	−0.3894	−1.6938	−2.0716	−0.5805
PC1	0.4456	−2.5535	0.0724	1.3152	−2.5021	1.0463
PC2	−1.944	−1.9124	−2.2136 **	−1.9532	−2.0069	−2.4062 **
EDU	−0.7053	−2.5659	0.5208	−0.421	−1.5926	0.9518
INF	−1.1645	0.4268	−0.2638	−1.6067	−0.0286	−0.0834
INS	−0.9508	−3.6817 **	−1.2982	−1.3019	−2.7371	−1.4464
First Difference
	Augmented Dickey–Fuller			Phillips–Perron
	Intercept	Trend and Intercept	None	Intercept	Trend and Intercept	None
ECP	−5.8290 ***	-	−4.0995 ***	−4.9847 ***	−4.9902 ***	−5.0169 ***
HDI	−5.2194 ***	−5.3219 ***	−5.0177 ***	−5.5147 ***	−5.5942 ***	−5.6551 ***
GNI	−5.2250 ***	−5.2465 ***	−5.2316 ***	−4.7400 ***	−4.7154 ***	−4.7499 ***
PC1	−3.1305 **	−3.2948 **	−2.1457 **	−4.3809 ***	−4.6174 ***	−3.7566 ***
PC2	−4.5813 ***	−4.6047 ***	-	−4.5075 ***	−4.5080 ***	-
EDU	−3.1351 **	−3.3893 **	−3.0939 ***	−4.9624 ***	−4.8581 ***	−4.9501 ***
INF	−2.9571 **	−3.7782 **	−2.9857 ***	−5.8847 ***	−7.6414 ***	−5.9071 ***
INS	−4.2921 ***	-	−4.1185 ***	−3.7369 ***	−3.7226 **	−4.0036 ***

Note: ** and *** display significance @ 5%, and 1%, respectively. Source: Authors’ Computation.

reports the stationarity test results using the Augmented Dickey–Fuller (ADF) and Phillips–Perron (PP) tests across three specifications: intercept, trend and intercept, and none. Results indicate that economic complexity and institutional quality are stationary at the 5% significance level under the trend and intercept specification in the ADF test. The second digitalisation index is stationary in the “none” specification at the 5% level in ADF and PP tests. All other variables achieve stationarity at first difference. These results confirm that the series are integrated in order one, with select variables stationary at the same level.

Table 5. F-Bound Test Result.

F-Bounds Test
Null Hypothesis: No Cointegration				Significance Levels
Model	Test Statistic	Value	k	Signif.	I (0)	I (1)
ECP	F-statistic	5.9470	5	10%	2.75	3.79
HDI	F-statistic	4.6057	5	5%	3.12	4.25
GNI	F-statistic	4.4562	5	2.5%	3.49	4.67
				1%	3.93	5.23

Source: Authors’ Computation.

presents the F-bound test results for cointegration across the three models: economic complexity, human development index, and gross national income. The null hypothesis of no cointegration is rejected when the F-statistic exceeds the upper bound I (1), indicating cointegration. Conversely, no cointegration is present if the F-statistic is below the lower bound I (0). An inconclusive result occurs when the F-statistic falls between the bounds. Significance levels are reported at 1%, 2.5%, 5%, and 10%. The F-statistics for economic complexity, human development index, and gross national income are 5.9470, 4.6057, and 4.4562, respectively, indicating cointegration for these models at various significance levels. The model selection criteria are presented in Figure A1, Figure A3 and Figure A5 of Appendix A.

4.2. ARDL and Causality Results

4.2.1. Autoregressive Distributed Lags (ARDL) Results

Economic Complexity Model

The ARDL analysis results in

Table 6. ARDL Result for Economic Complexity.

Long Run Result
Variable	Coeff.	Std. Error	t-Statistic	Prob.
PC1	0.2034 ***	0.0556	3.6578	0.0006
PC2	0.0947	0.0744	1.2734	0.2079
EDU	−1.1584 ***	0.3958	−2.9265	0.0049
INF	−0.084	0.2251	−0.3731	0.7104
INS	−0.6998	0.6156	−1.1366	0.2603
C	1.0915 ***	0.1759	6.2054	0.0000
Short Run Result
Variable	Coeff.	Std. Error	t-Statistic	Prob.
D(PC1)	−0.0906 *	0.0496	−1.8269	0.0728
D(PC2)	−0.1718 ***	0.0385	−4.4568	0.0000
D(EDU)	0.1170	0.4973	0.2353	0.8148
D(INF)	0.0087	0.0664	0.1321	0.8953
D(INS)	−0.7183 ***	0.2419	−2.9693	0.0043
CointEq (−1)	−0.3026 ***	0.0486	−6.2256	0.0000
R-squared	0.8289	F-statistic	8.0368 ***
Adj. R-squared	0.7258	Prob	0.0000
Durbin-Watson	2.2266
Diagnostic Checks
Histogram Normality Test			19.527 ***
Breusch-Godfrey Serial Correlation LM Test:			1.7752
Heteroskedasticity Test: ARCH			0.1175
Stability			Stable

Note: * and *** display significance @ 10% and 1%, respectively. Source: Authors’ Computation.

show economic complexity’s long- and short-run determinants. In the long-run analysis, the first digitalisation index (PC1) and education (EDU) emerged as significant predictors of economic complexity. Specifically, the first digitalisation index demonstrates a positive and significant coefficient of 0.2034, implying that a 1% increase in digitalisation (as captured by PC1) is associated with a 0.2034% rise in economic complexity. This finding aligns with the theoretical expectations that enhanced digital infrastructure and connectivity promote economic sophistication by facilitating access to advanced technologies and enabling complex production processes, a relationship corroborated by recent studies emphasising digitalisation as a driver of economic diversification and productivity growth (Ha, 2023; J. Li et al., 2024; Oumbé et al., 2023).

Contrarily, education (EDU) has a statistically significant negative coefficient of −1.1584 in the long run, indicating that a 1% increase in education levels is associated with a 1.1584% reduction in economic complexity, which is an unexpected inverse relationship because higher educational attainment contributes to economic growth. The result may indicate that promoting education does not directly translate to the specialised skills required to enhance economic complexity. A similar paradox has been observed in economies where education systems are geared towards general knowledge but may lack the technical rigour necessary for development sectors demanding high economic complexity (Snyder, 2013).

In contrast, in short-run dynamics, first and second digitalisation indices and institutional quality (INS) are statistically significant but have inverse relationships. A 1% increase in the first and second digitalisation indexes corresponds to a 0.0906% and 0.1718% reduction in economic complexity, respectively. These results may reflect the disruptive but temporary adjustments associated with adopting digital technologies, where early-stage digitalisation can initially displace traditional sectors or labour structures before contributing positively in the long run (Grigorescu et al., 2021). Institutional quality, measured through a corruption index, also has a short-run negative effect on economic complexity (−0.7183), suggesting that weaker institutional environments may hinder the swift deployment of resources toward complex economic activities. This aligns with evidence that improved institutions are essential for innovation and directing resources toward sectors that elevate economic complexity (C. Li et al., 2023).

The model’s diagnostic statistics affirm its robustness and stability. The error correction term is −0.3026, indicating that approximately 30.26% of any disequilibrium in economic complexity is corrected each period, indicating the model’s capacity to return to long-run equilibrium. The R-squared and adjusted R-squared values (0.8289 and 0.7258, respectively) confirm a high degree of explanatory power, while the Durbin–Watson statistic (2.2266) indicates no evidence of autocorrelation. The F-statistic (8.0368, p < 0.05) further corroborates the model’s overall significance. The results from the Breusch–Godfrey Serial Correlation LM Test and the Heteroskedasticity Test: ARCH (with values of 1.7752 and 0.1175, respectively) show no significant issues of serial correlation or heteroskedasticity, enhancing confidence in the model’s reliability. The stability model result is presented in Figure A2 in Appendix A.

The normality tests conducted for this model and the other two models below in this study indicate deviation from normality in the residuals. However, this does not compromise the validity of the models, as their reliability is safeguarded by the central limit theorem (CLT). The CLT asserts that with sufficiently large sample sizes, the distribution of sample mean approximates normality, irrespective of the underlying population distribution. This property ensures the robustness of statistical inferences that depend on asymptotic normality, particularly in large samples. While deviations from normality may lead to marginal reductions in statistical power, they do not invalidate the results when the models are well-specified and the sample size is robust (Du et al., 2017). The literature supports that for larger samples, asymptotic properties mitigate the need for strict normality in residuals, allowing for reliable and valid model interpretation even under such deviations (Lumley et al., 2002; Pek et al., 2018).

Human Development Index Model

The ARDL results in

Table 7. ARDL Results for the Human Development Index.

Long Run Result
Variable	Coeff.	Std. Error	t-Statistic	Prob.
PC1	0.0184 **	0.0081	2.2774	0.0267
PC2	0.0272 **	0.0109	2.4959	0.0156
EDU	0.3526 ***	0.0556	6.3401	0.0000
INF	0.0379	0.0297	1.2767	0.2072
INS	−0.028	0.0725	−0.3861	0.7009
C	0.0426 ***	0.0078	5.4348	0.0000
Short Run Result
Variable	Coeff.	Std. Error	t-Statistic	Prob.
D (PC1)	0.0177 ***	0.0038	4.6203	0.0000
D (PC2)	0.0103 ***	0.0027	3.8068	0.0004
D (EDU)	0.2476 ***	0.0476	5.2019	0.0000
D (INF)	0.0317 ***	0.0059	5.3550	0.0000
D (INS)	−0.0276	0.0198	−1.3941	0.1690
CointEq (−1)	−0.1506 ***	0.0274	−5.4948	0.0000
R-squared	0.9424	F-statistic	25.4026 ***
Adj R-squared	0.9053	Prob	0.0000
Durbin-Watson	2.2620
Diagnostic Checks
Histogram Normality Test			8.687 **
Breusch-Godfrey Serial Correlation LM Test:			1.7340
Heteroskedasticity Test: ARCH			0.0041
Stability			Stable

Note: ** and *** display significance @ 5% and 1%, respectively. Source: Authors’ Computation.

provide the human development index (HDI) over the long and short term. In the long run, digitalisation indices (PC1 and PC2) and education (EDU) exhibit significant positive relationships with the human capital development index. Specifically, the first and second digitalisation indices have coefficients of 0.0184 and 0.0272, respectively, indicating that a 1% increase in the first and second digitalisation indices results in a 0.0184% and 0.0272% increase in human development indices. This aligns with the growing body of research stressing the transformative role of digitalisation in human capital development, as it enhances access to information improves learning opportunities, and connects individuals to global networks, encouraging skill acquisition and knowledge transfer (Elfaki & Ahmed, 2024; Sithole et al., 2024).

The strong positive effect of education on human development index, with a coefficient of 0.3526, suggests that a 1% increase in education levels yields a 0.3526% increase in human capital development. This finding supports the theoretical underpinnings of human capital theory, which posits that investments in education yield significant returns in terms of improved health, productivity, and overall well-being (Becker, 2009). The finding also aligns with empirical studies indicating that education is a foundational driver of human development and economic progress, particularly in developing economies (Arruanlinggi et al., 2024). However, infrastructure and institutional quality do not significantly impact the human development index in the long run. While infrastructure development is critical for economic growth, its impact on human capital may depend on other factors, such as the quality of governance and accessibility, which may limit its effectiveness in promoting the human development index as observed (Acemoglu & Autor, 2011).

In the short run, the first digitalisation index, second digitalisation index, education, and infrastructure significantly impact the human development index with coefficients of 0.0177, 0.0103, 0.2476, and 0.0317, respectively, reflecting their immediate influence on human capital development. The positive short-run effects of digitalisation indices (PC1 and PC2) suggest that rapid advances in digital access and infrastructure contribute incrementally to the human development index. However, these impacts may stabilise as digitalisation matures (Reynolds et al., 2021). Furthermore, the immediate influence of education on the human development index remains substantial in the short run, as reflected in the 0.2476 coefficient, showing the crucial role of education as a long-term investment and a catalyst for immediate improvements in human well-being and capabilities, corroborating findings in similar human capital development literature (Arruanlinggi et al., 2024).

Infrastructure measured through air transport also significantly impacts human capital development in the short run. A 1% increase in infrastructure leads to a 0.0317% increase in the human capital development index; this suggests that accessible infrastructure enhances human development by improving mobility and access to essential services crucial for health, education, and economic opportunities. However, institutional quality remains insignificant in the short run, suggesting that immediate improvements in human capital development may not directly depend on governance quality but may require a sustained governance framework for long-term developmental gains (Huo et al., 2024).

The model diagnostics confirm the robustness and stability of the ARDL estimates. The error correction term is −0.1506, indicating that approximately 15.06% of the disequilibrium in HDI is corrected each period, pointing to a moderate speed of adjustment back to long-term equilibrium. The high R-squared and adjusted R-squared values (0.9424 and 0.9053, respectively) indicate a strong explanatory power, while the Durbin–Watson statistic (2.2620) suggests no significant autocorrelation. Furthermore, the F-statistic (25.4026, p < 0.05) supports the overall significance of the model. Additional diagnostics, Breusch–Godfrey Serial Correlation LM Test (1.7340) and Heteroskedasticity Test: ARCH (0.0041), indicate that the model is free from issues of serial correlation and heteroskedasticity, ensuring the reliability of the results. Stability tests also confirm model stability, with additional results in Figure A4 in Appendix A.

Gross National Income Model

The ARDL results in

Table 8. ARDL Results for the Gross National Income.

Long Run Result
Variable	Coeff.	Std. Error	t-Statistic	Prob.
PC1	−0.0407	0.0383	−1.0612	0.2923
PC2	−0.0524 **	0.0245	−2.1398	0.0360
EDU	−0.4565 **	0.2219	−2.0566	0.0436
INF	−0.0009	0.0711	−0.0134	0.9893
INS	0.8086 **	0.3462	2.3354	0.0225
C	−0.0096 ***	0.0032	−2.9308	0.0046
Short Run Result
Variable	Coeff.	Std. Error	t-Statistic	Prob.
D (PC1)	0.0106	0.0172	0.6193	0.5378
D (PC2)	0.0016	0.0108	0.1529	0.8789
D (EDU)	0.1168	0.1221	0.957	0.3419
D (INF)	−0.0097	0.0187	−0.5187	0.6056
D (INS)	0.2134 **	0.0923	2.3114	0.0238
CointEq (−1)	−0.1446 ***	0.0328	−4.4053	0.0000
R-squared	0.8034	F-statistic	12.4305 ***
Adj R-squared	0.7351	Prob	0.0000
Durbin-Watson	2.1764
Diagnostic Checks
Histogram Normality Test			131.3471 **
Breusch-Godfrey Serial Correlation LM Test:			1.9465
Heteroskedasticity Test: ARCH			0.0093
Stability			Stable

Note: ** and *** display significance @ 5% and 1%, respectively. Source: Authors’ Computation.

present the effects of digitalisation, education, institutional quality, and infrastructure on gross national income in both long- and short-run perspectives. In the long run, the second digitalisation index (PC2), education (EDU), and institutional quality (INS) are significant determinants of gross national income, although their impacts vary in direction. Notably, second digitalisation index and education have negative coefficients of −0.0524 and −0.4565, respectively, implying that a 1% increase in the second digitalisation index and education reduces gross national income by 0.0524% and 0.4565%, respectively. These findings suggest a possible trade-off effect, where higher levels of digital access and education investments may initially redirect resources away from income towards immediate consumption or human capital enhancement, which is consistent with the idea that expanding digital and educational access requires substantial upfront investments (George, 2024). Such a shift aligns with findings in developing economies, where digital and educational expansions are associated with increased consumption-driven economic activity, reducing immediate income but potentially enhancing long-term economic capacity (Yang et al., 2023).

On the other hand, institutional quality positively impacts gross national income, with a coefficient of 0.8086, suggesting that a 1% improvement in institutional quality leads to a 0.8086% increase in national income; this relationship shows that stabilising and improved institutions is important in economic security and encouraging income among individuals and businesses. Quality institutions enhance trust and confidence in the financial system, leading to higher income rates as individuals feel more secure in the stability and reliability of the economic environment (Boikova et al., 2021; Paul et al., 2020). This finding is particularly relevant for policymakers in emerging economies, where institutional improvements can drive economic growth through enhanced capital formation and income mobilisation (Samariddin Bakhriddinovich & Khasan, 2024; Vitkovics, 2023). However, neither the first digitalisation index nor infrastructure is statistically significant in the long-run model, indicating that these factors do not have a meaningful long-term impact on national income.

In the short run, only institutional quality remains a significant factor positively related to gross national income, while digitalisation indices, education and infrastructure show no significant effects; this finding implies that short-term changes in digital access, education levels, or infrastructure improvements do not immediately influence income behaviour, whereas improvements in institutional quality create a conducive environment for income even in the near term. Institutional factors may create a more favourable condition for economic stability, potentially enhancing income propensity (Samariddin Bakhriddinovich & Khasan, 2024). The absence of short-run effects for other variables could reflect the temporal lag in translating digital or educational advancements into economic behaviours such as saving, a pattern observed in the developmental process where capital formation lags behind initial policy changes (Fietz & Lay, 2023).

The error correction term of −0.1446 indicates a moderate speed of adjustment, with approximately 14.46% of the disequilibrium in national income corrected each period. This suggests that although short-run deviations occur, there is a gradual realignment towards long-term equilibrium, which is typical in income behaviour as economic agents adjust over time to macroeconomic changes and policy shifts. The high R-squared and adjusted R-squared values of 0.8034 and 0.7351 demonstrate the model’s strong explanatory power, while the Durbin–Watson statistic of 2.1764 suggests no significant autocorrelation issues. Additionally, the F-statistic of 12.4305 (p < 0.05) confirms the overall significance of the model, validating the relationships among the variables. Diagnostic tests also support model reliability. The Breusch–Godfrey Serial Correlation LM Test (1.9465) and the Heteroskedasticity Test: ARCH (0.0093) indicate no serial correlation or heteroskedasticity issues. Stability tests in Figure A6 in Appendix A confirm the model’s reliability over time.

4.2.2. Causality Test Result

The causality results summarised in

Table 9. Causality Results.

Dependent Variable	Independent Variable	MWALD Stat.	p-Value	Causality
ECP	GNI	2.7597	0.2516	No
ECP	HDI	0.3891	0.8232	No
ECP	PC1	6.0231 **	0.0492	Yes
ECP	PC2	0.1807	0.9136	No
GNI	ECP	7.8933 **	0.0193	Yes
GNI	HDI	6.1064 **	0.0472	Yes
GNI	PC1	2.0400	0.3606	No
GNI	PC2	1.5955	0.4503	No
HDI	ECP	1.4078	0.4946	No
HDI	GNI	3.4868	0.1749	No
HDI	PC1	3.1769	0.2042	No
HDI	PC2	9.0221 **	0.011	Yes
PC1	ECP	2.5448	0.2802	No
PC1	GNI	6.0022 **	0.0497	Yes
PC1	HDI	0.8204	0.6635	No
PC2	ECP	1.6575	0.4366	No
PC2	GNI	0.9020	0.637	No
PC2	HDI	0.4789	0.7871	No

Note: ** displays significance @ 5%. Source: Authors’ Computation.

show the interrelationships among digitalisation (represented by PC1 and PC2), human capital development (HDI), gross national income (GNI), and economic complexity (ECP). The unidirectional causality from the first digitalisation index to economic complexity suggests that advancements in digitalisation, as captured by PC1, drive improvements in economic complexity. This finding aligns with recent research emphasising digitalisation’s transformative role in modern economies by enabling complex, technology-driven activities, and diversified economic structures (Magoutas et al., 2024; Shi & Wei, 2024). Digitalisation facilitates knowledge transfer, enhances productivity, and supports innovation, essential for achieving higher economic complexity (J. Li et al., 2024; Szalavetz, 2019).

Another notable causation flows from economic complexity to gross national income, indicating that as economies become more complex, they contribute positively to gross national income. This result is consistent with the argument that more complex economies exhibit higher productivity levels and innovation capacities, leading to more income growth (Sarangi & Pradhan, 2020). Economies with diverse, knowledge-intensive sectors are better equipped to sustain high levels of output and income, even given global economic volatility. These findings have policy implications for developing countries seeking sustainable income growth; prioritising policies that diversify economic structures through digitalisation and innovation can lead to higher national income levels.

A third unidirectional causation from HDI to GNI shows the importance of human capital in driving national income. Human capital directly influences economic performance by enhancing productivity and enabling labour mobility (Becker, 2009). This result indicates much needed investment in human capital as a means of improving economic growth, a perspective supported by numerous studies that connect human development with higher levels of national income (Abbas & Zaman, 2024). Additionally, the causality from the second digitalisation index to the human capital development index indicates that digitalisation also positively impacts human capital development. The second digitalisation index, which reflects the specific side of digitalisation, contributes to improved human capital through enhanced access to information, digital learning resources, and communication tools that empower individuals and improve education (Abu Alfoul et al., 2024). Digital tools have been shown to play a role in augmenting educational infrastructure and supporting skill development, particularly in remote or underserved regions (Liu et al., 2024). This result suggests that integrating digital resources into educational systems can be a strategy for human capital development, especially in low- and middle-income countries.

The unidirectional relationship from gross national income to the first digitalisation index illustrates that national income levels can influence the advancement of digital infrastructure. Higher income levels provide governments and businesses with the resources needed to invest in digital technologies, which can reinforce economic growth by creating a more connected and efficient economic environment (Hao et al., 2024). The absence of significant causality in other variable pairs suggests independence, implying that certain variables do not directly influence each other within the framework of this study. These non-significant relationships indicate the specificity of digital, human capital, and economic complexity linkages to income, as opposed to more comprehensive, undifferentiated causal interrelations.

5. Conclusion and Policy Recommendations

The rapid advancement of digital technologies has increasingly driven economic development, fundamentally transforming sectors, and generating new pathways for economic and social development. Digitalisation is now a cornerstone of the global economy due to its potential to boost productivity, encourage innovation, and promote inclusive growth. However, South Africa continues to deal with persistent developmental challenges, including a skills shortage, high unemployment, poverty, infrastructure deficits, and socioeconomic inequality. Since the end of apartheid in 1994, economic progress has been gradual, as reflected in key economic indicators like GDP per capita, gross national income/income, the human development index and economic complexity.

Despite the centrality of digitalisation, research on its role in promoting economic development through diverse indicators, such as economic complexity, human capital development, and gross national income, remains emergent. In addressing this gap, our study constructs a digitalisation index using Principal Component Analysis based on mobile cellular subscriptions, internet usage, and fixed telephone subscriptions. Through the ARDL model and Granger causality analysis, we examine the relationships between digitalisation, human development, economic complexity, and gross national income. Our objectives are to investigate digitalisation’s short- and long-term impacts on these dimensions and analyse their causal relationships.

Key findings reveal that digitalisation and education contribute positively to human capital development in the long term. In contrast, digitalisation and infrastructure improvements significantly drive immediate gains in human development. These results show that digitalisation is critical in promoting adaptable, skilled labour that supports socioeconomic growth. For the gross national income model, the findings indicate that digitalisation and education exert downward pressure on income due to development-related expenditures, while strong institutional quality significantly boosts income. This balance suggests that enhancing governance frameworks is essential for maintaining financial growth alongside digital and educational investments. Our results in the economic complexity model show that digitalisation has a positive long-term effect, though education exhibits a negative relationship. Short-term impacts show that both digitalisation indices and institutional quality negatively influence economic complexity; underlining the importance of balancing immediate educational and governance investments with long-term economic goals. The causality analysis further shows the relationship between these factors. Our findings indicate that digitalisation promotes economic complexity, while human capital development is crucial in driving national income. Higher national income strengthens digitalisation efforts, creating a self-sustaining technological advancement and economic growth cycle.

These insights indicate the need for policy frameworks aligning short-term development priorities with sustainable long-term objectives. Policymakers should prioritise investments in digital infrastructure and accessible educational resources to build a skilled labour force. Additionally, policies should integrate digital initiatives with education systems to enhance economic complexity and promote skills suited to an evolving economic change. Strengthening institutional quality is crucial to sustaining income rates amid increasing digital and educational advancement expenditures for gross national income. Given the result, recognising the unidirectional causality between digitalisation, human capital, and economic complexity, policies should aim for incorporation by addressing both immediate and enduring dimensions of socioeconomic development.

While providing valuable insights into the role of digitalisation in economic development, this study is limited by its reliance on aggregate indicators, which may obscure regional or sectoral variations within South Africa. Also, while our findings show the importance of digitalisation, economic complexity, and human capital development in South Africa, the implementation of policy recommendations for a country with a similar structure should consider country-specific economic conditions, including GDP levels and priority investment areas. Policymakers should implement digitalisation strategies based on national economic structures, resource availability, and developmental priorities to ensure effectiveness and sustainability. Future research could enhance these insights by conducting country-specific case studies or incorporating a more detailed analysis that accounts for variations in GDP levels and investment priorities. Additionally, the study focuses on a specific set of digitalisation and economic development indicators, potentially overlooking other critical dimensions such as environmental sustainability and social inclusion. Finally, future studies should investigate the role of cultural characteristics and social barriers in technology adoption to provide a more comprehensive understanding of digitalisation’s impact.