Navigating the nexus: unraveling technological innovation, economic growth, trade openness, ICT, and CO2 emissions through symmetric and asymmetric analysis

The study investigates how economic growth (EG), trade openness (TON), technological innovation (TIN), and information and communication technology (ICT) jointly affect CO2 emissions in Malaysia. Against a backdrop of accelerating climate change and Malaysia’s commitments to reduce greenhouse gas intensity by 45% by 2030 and target carbon neutrality by 2050, the research addresses the tension between growth, industrialization, and environmental sustainability. Prior work often examined these drivers in isolation, relied on linear methods, or lacked country-specific asymmetric analysis. This paper aims to fill that gap by using both symmetric (ARDL) and asymmetric (NARDL) approaches on Malaysian time-series data (1985–2021) to understand short- and long-run dynamics and potential asymmetries. The research is important for informing policy on aligning growth and digitalization with emissions mitigation, and clarifying the role of technological innovation in achieving Malaysia’s climate goals.

- Technological innovation (TIN) and CO2: The literature shows mixed effects. Some studies find innovation lowers emissions via efficiency gains and environmental technologies (e.g., Mensah et al., Lin & Zhu, Ganda; Wang et al., Li et al.), while others report positive associations or sectoral/industry-specific heterogeneity (Amin et al.; Erdogan; Shahbaz et al.; Long et al.). Effects can differ by development level and innovation type (e.g., patents/trademarks). Energy-focused R&D tends to reduce emissions; reduced patenting may correlate with sustainability in some contexts. - Economic growth (EG) and CO2: Growth generally raises emissions through higher energy demand and carbon-intensive production, particularly in early development stages. The literature highlights links between energy use, environmental degradation, and growth, with potential decoupling influenced by trade and policy (Ang; Khan et al.; Mahmood et al.; Yang et al.). Balancing growth with environmental constraints is emphasized. - Trade openness (TON) and CO2: Evidence is mixed and potentially asymmetric. Trade can raise emissions via scale and composition effects, but can also reduce them via technology transfer and efficiency. FDI plays a mediating role. Studies report positive, negative, and asymmetric impacts across countries and time horizons (Usman et al.; Wang & Zhang; Azam et al.; Irfan et al.; Jakada et al.; Mahmood et al.). - ICT and CO2: ICT can enable energy efficiency, monitoring, and green transitions, but also increases energy demand from device use, data, and infrastructure. Empirical findings are mixed, with both emission-reducing and emission-increasing effects reported across regions and sectors (Asongu et al.; Danish; Haini; Abdollahbeigi & Salehi; Tzeremes et al.; Pan & Dong; Weili et al.; Ebaidalla & Abusin). The net effect may depend on ICT maturity, policy, and complementary innovation.

- Data and variables: Annual Malaysian time-series data from 1985–2021 sourced from World Development Indicators (World Bank). Variables: CO2 emissions per capita (CO2), GDP per capita (EG), trade openness (TON, sum of exports and imports), ICT (mobile cellular subscriptions per 100 people), and technological innovation (TIN, total number of patent applications). All variables are log-transformed where applicable. - Empirical strategy: Employed Auto-Regressive Distributed Lag (ARDL) and Nonlinear ARDL (NARDL) to assess short- and long-run relationships and asymmetries among variables. • Preliminary tests: Chow structural break test (no significant breaks), unit root tests using ADF, DF-GLS, and Phillips–Perron to confirm integration orders I(0)/I(1) suitable for ARDL. BDS test indicates non-linear dependence, motivating NARDL. • ARDL bounds testing (Pesaran et al., 2001; Narayan critical values): Tests for cointegration among CO2, EG, TON, ICT, and TIN. Error-correction model (ECM) used to capture speed of adjustment. • NARDL (Shin et al., 2014): Decomposes EG, ICT, and TIN into positive and negative partial sums to test long-run and short-run asymmetries. Constructs series for positive (X_POS) and negative (X_NEG) shocks and estimates asymmetric cointegration and dynamic multipliers. • Diagnostics: Serial correlation (Breusch–Godfrey LM), heteroscedasticity (Breusch–Pagan–Godfrey, ARCH), normality (Jarque–Bera), and stability (CUSUM, CUSUMQ) tests. • Causality: VAR-based Granger causality in the symmetric framework to infer directional links among variables. • Robustness: Long-run coefficients cross-validated with FMOLS, DOLS, and CCR estimators. - Model specification: Baseline functional form CO2 = f(EG, TON, ICT, TIN). ARDL includes lag selection via information criteria; ECM captures long-run equilibrium correction. NARDL incorporates partial sums for asymmetric effects and estimates long-run (lambda) and short-run (phi) elasticities.

- Descriptive and pre-estimation: • ICT shows highest volatility; variables generally suitable for estimation. Unit root tests indicate mixed I(0)/I(1); ARDL applicable. • ARDL bounds test: F-statistic = 7.939 (> upper bound 5.914 at 1%), indicating cointegration among variables. • BDS test confirms non-linear dependence, justifying NARDL. - Linear ARDL (Table 6): • Long run: EG (LEG) 0.671*** (t=3.798), LTON 0.755*** (t=8.078), ICT 0.005*** (t=7.535) all positively and significantly related to CO2; LTIN 0.042 (ns). • Short run: ΔLEG 0.615*** (t=4.328), ΔLTON 0.154* (t=2.011), ΔICT 0.002*** (t=3.819) increase CO2; ΔLTIN −0.042*** (t=−3.917) reduces CO2, indicating short-run emission-reducing effect of innovation. • ECM(−1) = −0.480***: about 48% annual speed of adjustment toward long-run equilibrium. • Diagnostics: High R2 (0.998), no serial correlation, homoscedasticity, normal residuals, and correct specification (RESET). - Nonlinear ARDL (Table 7): • Long run asymmetries: EG_POS 1.278***; EG_NEG ns. TON 0.363**. ICT_POS 0.002*; ICT_NEG 0.010** (both positive), indicating ICT shocks raise CO2 with stronger effect for negative ICT shocks. TIN_POS −0.123** reduces CO2; TIN_NEG ns. • Short run: D(EG_POS) 0.920*** raises CO2; D(ICT_POS) 0.001* and D(ICT_NEG) 0.007** raise CO2; D(LTIN_POS(−1)) −0.029** indicates lagged innovation-driven reductions. D(LTON) ns in short run. • ECM(−1) = −0.720***: about 72% speed of adjustment, indicating rapid convergence. • Diagnostics: No serial correlation or heteroscedasticity; normal residuals; model stability supported (CUSUM/CUSUMQ noted in text). - Causality (Tables 8–9): • Bidirectional: TIN ⇄ CO2 (5%); TON ⇄ CO2 (1–5%); TON ⇄ ICT (10% and 1%). • Unidirectional: CO2 → ICT (10%); GDP → TON (5%); GDP → TIN (1–5%); ICT → TIN (10%). No causality between GDP and CO2 in either direction. - Robustness (Table 10): FMOLS, DOLS, CCR confirm positive and significant long-run links of GDP, TON, and ICT with CO2; TIN not significant, aligning with ARDL long-run and NARDL findings on asymmetry and short-run effects. - Policy-relevant quantitative highlights: Positive growth and trade effects on emissions are economically large (e.g., long-run elasticities ~0.67 and ~0.76 in linear ARDL), while innovation reduces emissions in the short run and under positive innovation shocks in the long run (NARDL LTIN_POS −0.123**).

The findings directly address the research question by demonstrating that Malaysia’s economic growth, trade openness, and ICT expansion are associated with higher CO2 emissions in both short and long horizons under linear assumptions, while technological innovation contributes to emission reductions in the short run and under positive innovation shocks in the long run. The asymmetric NARDL results refine this picture: increases in GDP have strong positive effects while decreases do not symmetrically reduce emissions; ICT shocks raise emissions in both directions, with negative ICT shocks exhibiting a larger effect; and positive innovation shocks reduce emissions whereas negative shocks are not significant. These asymmetries underscore the importance of sustaining and amplifying innovation progress rather than allowing regressions in technological advancement. The evidence of cointegration and rapid error correction suggests a stable long-run equilibrium among the variables, while diagnostic and robustness checks support the reliability of estimates. Causality patterns reveal two-way links between trade openness and emissions and between innovation and emissions, highlighting feedback loops where greater emissions may stimulate innovation and trade dynamics that, in turn, affect emissions. The unidirectional effect from CO2 to ICT implies that rising emissions may drive ICT adoption/usage changes or measurement/management responses. Overall, the results emphasize that without targeted policies, continued growth, trade expansion, and ICT proliferation risk elevating emissions. However, fostering environmental technological innovation can mitigate these effects. For Malaysia, aligning economic and digital development with emission reduction requires policy instruments that accelerate clean innovation, enhance energy efficiency, guide ICT toward greener trajectories, and internalize carbon costs in production and trade.

This study contributes by jointly analyzing EG, TON, ICT, and TIN impacts on CO2 emissions in Malaysia using both symmetric (ARDL) and asymmetric (NARDL) approaches over 1985–2021. Key contributions include documenting: (i) positive effects of growth, trade openness, and ICT on emissions under linear modeling; (ii) short-run and asymmetric long-run emission-reducing roles of technological innovation; and (iii) asymmetric responses to positive versus negative shocks in GDP, ICT, and TIN. Robustness checks (FMOLS, DOLS, CCR) corroborate the main long-run associations. Policy implications include prioritizing clean technology R&D and diffusion; leveraging ICT for efficiency while managing its energy footprint; integrating environmental criteria into trade policy; aligning growth strategies with low-carbon transition; employing carbon pricing; upgrading energy infrastructure; and supporting awareness campaigns that encourage behavioral change. Coordinated, cross-sectoral strategies and public–private partnerships can expedite the shift toward renewable energy and energy-efficient technologies. Future research could broaden scope via cross-country comparisons, incorporate additional variables (e.g., renewable energy penetration, sectoral composition, energy prices, financial development), extend the time horizon, and apply alternative or complementary methods (e.g., quantile methods, structural breaks, machine learning-based causal discovery) to capture heterogeneity and evolving dynamics.

- Methodological focus on ARDL/NARDL; results may differ under alternative econometric frameworks. - Single-country (Malaysia) analysis limits generalizability; cross-country heterogeneity not captured. - Variable set excludes potentially relevant factors (e.g., renewable energy share, sectoral energy mix, energy prices, institutional quality) that could mediate effects. - Annual data may mask intra-year dynamics; data limitations could affect precision of innovation and ICT measures. - Asymmetric causality not explored due to model restrictions; structural shifts beyond tested breaks may still exist.