Bridging the digital divide: the impact of technological innovation on income inequality and human interactions

The paper addresses how technological innovation influences income inequality (INE) and how this relationship is shaped by economic growth, globalisation and export trade across countries at different development levels. Motivated by rising global inequality, the COVID-19 shock, and persistent digital divides in access and use, the authors examine whether innovation narrows or widens income gaps, and whether macro-structural forces buffer or amplify innovation’s impact on inequality. The study highlights gaps in prior work: limited analysis of feedback mechanisms between innovation and inequality, scarce evidence on moderating roles of growth, globalisation and exports, and insufficient heterogeneity analyses by development level and period. The research questions are operationalised via hypotheses: H1a/H1b (innovation reduces vs. exacerbates inequality) and H2a–H2c (economic growth, globalisation and export trade significantly moderate the innovation–inequality nexus).

The literature presents mixed evidence on the innovation–inequality nexus. Some studies, grounded in Schumpeterian and evolutionary perspectives, find that innovation can reduce inequality by raising labour income shares, promoting creative destruction that erodes incumbents’ rents, and through ICT diffusion that narrows gaps in the long run (e.g., Antonelli & Gehringer, 2017; Nguyen et al., 2020; Adams & Akobeng, 2021). Policy environments and human capital development can reinforce inequality-reducing effects (Yeo et al., 2021). Conversely, many studies highlight the ‘dark side’ of innovation: skill-biased and factor-biased technological change raises skill premiums, displaces low-skilled labour, and increases rent concentration, heightening inequality (Acemoglu, 1999; Antonelli & Scellato, 2019; Law et al., 2020). Sectoral and country heterogeneity, especially finance-driven wage premia, further complicate outcomes. Beyond direct effects, globalisation and trade openness can shape wage structures, skill premia and sectoral exposure, with evidence pointing in both directions depending on measurement (trade vs. financial globalisation), development stage, and sectoral composition (Helpman et al., 2010; Asteriou et al., 2014; Huang et al., 2022). The authors formalise hypotheses: H1a (innovation reduces inequality) vs. H1b (innovation exacerbates inequality); H2a (economic growth moderates innovation’s effect on inequality); H2b (globalisation moderates the relationship); H2c (export trade moderates the relationship).

Data: Unbalanced panel of 59 countries (31 developed, 28 developing) from 1995–2020. Key variables (all in natural logs): - Income inequality (INE): top 10% income share (World Inequality Database). - Technological innovation (INNO): number of patents (proxy for national innovation capacity). - Economic growth (GDP): GDP per capita (World Bank). - Globalisation (GLO): KOF Globalisation Index (economic, social, political dimensions). - Exports (EGS/ESG): Exports of goods and services (% of GDP) from World Bank. - Control: Human capital (HC), proxied by labour force (World Bank). All variables log-transformed to stabilise variance and mitigate heteroskedasticity. Model specification: - Model I (direct effect): ln(INE)_{it} = θ_i + β1 ln(INNO)_{it} + Σ ζ_i ln(Control)_{it} + μ_{it}. - Interaction models: - Model II: ln(INE)_{it} = θ_i + β1 ln(INNO)_{it} + β2 ln(GDP)_{it} + β3 [ln(INNO)_{it}×ln(GDP)_{it}] + Σ ζ_i ln(Control)_{it} + μ_{it}. - Model III: replace GDP with GLO and interaction with ln(INNO)×ln(GLO). - Model IV: replace GDP with EGS and interaction with ln(INNO)×ln(EGS). Interpretation follows Brambor et al. (2006): marginal effects depend on moderator levels. Econometric strategy: - Pre-tests: Cross-sectional dependence (CD) via Friedman (1937), Frees (1995), Pesaran (2004); slope heterogeneity via Pesaran & Yamagata (2008) and Blomquist & Westerlund (2013); second-generation unit root tests CADF/CIPS (Pesaran, 2007). - Estimation: Common Correlated Effects Mean Group (CCEMG; Pesaran, 2006) to address CD, unobserved common factors, slope heterogeneity, nonstationarity, and serial correlation; outcomes reported separately for developed vs. developing groups. - Robustness: Augmented Mean Group (AMG; Eberhardt & Bond, 2009) to account for unobserved common dynamic processes; results compared with CCEMG. - Causality: Dumitrescu–Hurlin heterogeneous panel Granger causality tests to probe directional links among INE, INNO, GDP, GLO, EGS, and HC. - Heterogeneity over time: Split-sample analysis for two subperiods (1995–2007; 2008–2020) to capture chronological shifts in effects.

- Baseline (Model I, CCEMG): Innovation increases inequality in both groups, with stronger effects in developed economies. - Developed: 1% ↑ INNO → 0.041% ↑ INE (SE 0.019; Table 2). - Developing: 1% ↑ INNO → 0.017% ↑ INE (SE 0.011; Table 2). - Moderation (Interaction models, CCEMG): - Economic growth (Model II): - Developed: ln(INNO)×ln(GDP) coefficient = +0.212 (SE 0.106), implying growth amplifies innovation’s inequality-increasing effect (Table 3). - Developing: ln(INNO)×ln(GDP) = −0.221 (SE 0.082), indicating growth attenuates the inequality effect (Table 4). - Globalisation (Model III): - Developed: ln(INNO)×ln(GLO) = −0.909 (SE 0.516), buffering the inequality effect (Table 3). - Developing: ln(INNO)×ln(GLO) = −0.361 (SE 0.146), also buffering (Table 4). - Export trade (Model IV): - Developed: ln(INNO)×ln(EGS) = −0.380 (SE 0.216), buffering (Table 3). - Developing: ln(INNO)×ln(EGS) = +0.102 (SE 0.044), exacerbating (Table 4). - Robustness (AMG): Signs and significance broadly consistent with CCEMG for baseline and interaction effects across both groups (Tables 5–7). - Temporal heterogeneity: In developed countries, innovation’s inequality-increasing effect is significant in both subperiods and stronger in 2008–2020 (e.g., 1% ↑ INNO → 0.286% ↑ INE in subperiod II). In developing countries, the effect shifts from inequality-increasing in 1995–2007 to inequality-diminishing in 2008–2020, consistent with maturation of technology adoption and policy reforms (Tables 8–9). - Causality (Dumitrescu–Hurlin): Bidirectional causality between INNO and INE in both groups. INE and GDP/EGS show bidirectional causality in both groups. For globalisation, no significant causality with INE in developed countries, while bidirectional causality exists between GLO and INE in developing countries (Table 10). - Overall: H1b supported (innovation exacerbates inequality on average). H2a–H2c supported: GDP, GLO, and EGS significantly moderate the innovation–inequality relationship, with directions differing by development level.

The findings resolve key debates by showing that innovation’s average effect tends to increase income concentration, particularly in developed economies with abundant high-skill endowments and strong intellectual property regimes. Mechanistically, skill- and factor-biased technological change raises the premium for high-skilled labour and capital owners while displacing or stagnating low-skilled workers, amplifying wage dispersion. At the sectoral level, finance and high-tech sectors capture disproportionate rents, contributing to top income growth. Internationally, production shifts and fragmented value chains elevate R&D and high-skill demand in developed economies, further widening gaps. Moderators play distinct roles: economic growth in developed economies intensifies the inequality effect (possibly via concentration of rents and technological barriers), while in developing economies growth can diffuse gains through industrialisation, job creation, and improved public services, mitigating inequality. Globalisation generally buffers the innovation–inequality link by facilitating knowledge diffusion, skill upgrading, and mobility, with stronger effects in developed economies. Export trade reduces the inequality effect in developed countries (via upgrading, broader employment, and competitive pressures) but exacerbates it in developing countries where export baskets remain low value-added and gains concentrate among scarce high-skill workers. Temporal analyses indicate that as technologies and policies mature, innovation’s inequality effects can attenuate or reverse in developing contexts, highlighting the importance of policy environments and stages of technological diffusion. These results underscore the need for inclusive innovation policies that expand access, skills, and sectoral upgrading to distribute technological gains more equitably.

Using CCEMG/AMG on panel data from 59 countries (1995–2020), the study shows that technological innovation generally heightens income inequality, with stronger effects in developed economies. Economic growth, globalisation, and export trade significantly moderate this relationship, but the direction of moderation depends on development level: growth amplifies the inequality effect in developed economies and dampens it in developing ones; globalisation buffers the effect in both; exports buffer in developed but exacerbate in developing economies. Temporal heterogeneity reveals strengthening inequality effects in developed economies post-2008 and a transition toward inequality reduction from innovation in developing economies as diffusion and policy supports mature. Policy recommendations include: investing in skills training and vocational education, strengthening labour protections and employment services, fostering diversified and inclusive innovation (lowering technological barriers, R&D infrastructure, cross-sector collaboration, start-up support, balanced IP protection), and advancing inclusive digitalisation (broadband and public digital infrastructure, user-friendly technologies, digital literacy, data security and privacy). These policies aim to bridge digital divides and ensure wider participation in and benefits from technological progress. Future research should probe feedback loops between inequality and innovation, mediating channels, and the roles of education, fiscal/monetary, and competition policies in shaping the distribution of innovation gains.

- Sample coverage is limited to 31 developed and 28 developing countries; broader global samples would improve generalisability. - Inequality is proxied by the top 10% income share; alternative measures (e.g., Gini, top 1%, bottom 50%) should be examined for robustness. - The dynamic nature of technological change implies results reflect a snapshot; effects may evolve over time and across policy regimes. - Potential endogeneity and feedback mechanisms, while explored via causality tests, warrant deeper causal identification and mediation analyses. - Generalisability may be constrained by country-specific institutions, sectoral structures, and data limitations.