Guardians of Growth: Can Supply Chain Pressure, Artificial Intelligence, and Economic Inequality Ensure Economic Sustainability

Abstract

This study examines the effects of supply chain pressure, smart AI, and socio-economic fairness on long-term economic sustainability. To this end, this study uses quarterly data from 1999 Q1 through 2024 Q4 for the United States and employs the recently introduced Wavelet Cross-Quantile Regression (WCQR) to analyze this relationship. This study finds that smart AI, supply chain pressure (SC), and renewable energy consumption (REC) significantly drive U.S. economic growth, with the strongest short-term effects appearing when adoption and output are in the lower quantiles, reflecting threshold and diffusion dynamics. SC enhances growth once supply chain networks reach a critical level of connectivity, while REC generates substantial gains at low penetration levels, illustrating a “catch-up” effect. In contrast, economic inequality (EI) generally dampens growth, especially at moderate to high inequality levels; however, long-term reductions in EI yield positive returns in high-growth states by improving social cohesion and workforce productivity. Based on these findings, this study proposes funding low-adoption AI now, scaling to mid-adoption users mid-term, and entrenching long-term gains through economy-wide upskilling.

1. Introduction

Economic sustainability—a core pillar of sustainable development—emphasizes not only the efficient use of resources but also the resilience and inclusivity of growth trajectories. In the United States, this concept has gained prominence amid concerns over environmental degradation, social disparities, and infrastructure challenges [1]. According to the 2024 Sustainable Governance Indicators Network, the U.S. currently ranks 17th in economic sustainability, reflecting its advanced economy but also signaling structural vulnerabilities in areas such as emissions intensity and social equity [2]. Even before the COVID-19 pandemic, strains in the U.S. growth model—ranging from aging infrastructure to widening income gaps—highlighted the need for integrative frameworks that marry economic dynamism with long-term ecological and social well-being. This study addresses the central research problem: can the synergistic application of transparent logistics, smart AI, and socio-economic fairness serve as “guardians of growth,” ensuring a sustainable economic future for the United States?

Supply chain pressure—a paradigm that encompasses the challenges of demand volatility, regulatory constraints, and operational complexity—has demonstrated tangible impacts on both efficiency and sustainability. Enhancing blockchains and Smart Contracts illustrates how blockchain and IoT technologies can help firms manage supply chain pressure by improving transparency, reducing transaction costs, and curbing resource waste across supply networks [3]. Studies [4] have further shown that in developed economies, mitigating supply chain pressure through improved visibility in air transport and freight operations causally drives GDP growth by optimizing capacity utilization and reducing logistical bottlenecks [5]. Yet, despite these advances, research remains sparse on how strategies for managing supply chain pressure can be embedded within circular economic systems—where resource loops are closed and value is retained—highlighting the gap this study seeks to fill.

Smart AI technologies, encompassing machine learning, predictive analytics, and optimization algorithms, hold transformative promise for aligning growth with sustainability goals. Reference [6] demonstrates that AI-enhanced decision-making frameworks can reduce carbon footprints in U.S. supply chains by up to 15% through superior route optimization and demand forecasting [6]. Foundational work by [7] frames intelligent infrastructure as a lynchpin for resilient, adaptive economic systems capable of self-monitoring and continuous performance improvement [8]. Nonetheless, challenges related to data governance, algorithmic bias, and the energy demands of large-scale AI deployments must be addressed before AI’s full potential as a guardian of sustainable growth can be realized.

Economic inequality—which refers to disparities in income, education, and access to resources—is a critical factor affecting long-term sustainability. The United Nations’ 2030 Agenda highlights the importance of addressing inequality through the principle of “leaving no one behind,” directly linking equitable outcomes with economic resilience and the Sustainable Development Goals [9,10]. In the U.S., [11] has found that persistent economic disparities threaten progress toward sustainability targets, suggesting that without targeted interventions to reduce inequality, benefits from circular practices, logistics transparency, and AI innovation may remain uneven and unsustainable ([12,13]). Addressing economic inequality in the design and implementation of growth strategies is therefore essential for achieving a holistic and durable model of economic sustainability.

Drawing on the foregoing discussion, this study addresses the following research questions aligned with its objectives:

• RQ1: How does supply chain pressure support economic sustainability?
• RQ2: What is the impact of smart AI on the sustainability of economic systems?
• RQ3: In what ways does socio-economic fairness impact economic sustainability?
• RQ4: What is the effect of renewable energy consumption on economic sustainability?

This study contributes to the ongoing literature in the following ways: First, this research undertakes the inaugural empirical analysis of supply chain pressure mechanisms and their influence on sustainable economic growth in the United States. By integrating detailed supply chain visibility metrics with growth models rooted in energy and development economics, it addresses a pivotal gap in the existing literature. This study’s insights will inform targeted policy interventions designed to enhance logistical transparency, curb resource inefficiencies, and promote resilient, low-carbon development pathways. In doing so, it offers actionable guidance for policymakers seeking to align logistical reforms with overarching sustainability objectives.

Second, this study critically evaluates the capacity of smart AI–driven decision-making frameworks to catalyze green growth within the United States, the world’s largest economy. By harnessing advanced predictive analytics and machine learning algorithms on high-frequency energy and economic data, it quantifies AI’s potential to optimize resource allocation, curb carbon intensity, and accelerate the transition to a low-carbon industrial base [14]. The findings from this U.S. case study provide a rigorous benchmark for both emerging and advanced economies, demonstrating how digital innovation can be strategically leveraged to meet sustainable development objectives.

Third, recognizing that prior studies [8,9,15] have largely overlooked the joint distributional and spectral heterogeneity of key economic variables, this study adopts the recently introduced Wavelet Cross-Quantile Regression (WCQR) approach suggested by [16]. WCQR uniquely combines multi-resolution wavelet decomposition with cross-quantile estimation to capture dynamic, scale-specific relationships across the entire conditional distribution, without imposing restrictive linearity or homoskedasticity assumptions. By disentangling how variables interact at different frequencies and quantiles, this methodology uncovers subtle, policy-relevant heterogeneities—revealing, for instance, whether low-quantile shocks in AI adoption or supply chain pressure reforms have distinct short- versus long-term impacts on growth. These nuanced insights will inform the design of differentiated, frequency-aware policy interventions that accommodate both temporal dynamics and distributional disparities in the pursuit of economic sustainability.

The subsequent sections are as follows: Section 2 reviews the literature, Section 3 outlines the data and methodology, Section 4 presents the results, and Section 5 concludes the study.

2. Theoretical Framework and Literature Review

2.1. Theoretical Framework

Economic growth (EG) is fundamentally driven by the efficient allocation of resources, technological innovation, and equitable distribution of economic gains. Social Economic Fairness (SEF) enhances human capital formation and consumer demand by ensuring access to education, healthcare, and credit, thereby expanding the productive capacity of the workforce [17,18]. When households perceive that growth dividends are shared fairly, savings and investment behavior improve, supporting capital accumulation and fostering a virtuous cycle of demand-led expansion [9]. When growth dividends are shared and savings are matched, households save more, banks lend more, and demand-led growth accelerates. Artificial intelligence (AI) contributes to EG through productivity gains; by automating routine tasks, optimizing production processes, and enabling data-driven decision-making, AI raises total factor productivity and spurs innovation across sectors [19,20]. Moreover, AI’s ability to process vast datasets in real time enhances allocative efficiency in labor and capital markets, further propelling long-run growth trajectories.

Parallel to these distributional and technological drivers, the transition to a low-carbon economy critically depends on renewable energy consumption (REC) and supply chain pressure (SC). Investment in renewable infrastructure (solar, wind, and hydro) reduces energy import dependence and mitigates volatility in production costs, while generating positive spillovers in green technology R&D—outcomes that translate to higher output and employment [21,22]. Supply chain pressure reduces transaction costs, minimizes operational frictions, and accelerates market integration, thereby enhancing the growth benefits of renewable energy deployment by ensuring that clean energy inputs are delivered more efficiently and reliably [3]. When considered alongside complementary factors—including fair institutions, advanced AI, renewable energy transition, and streamlined logistics—supply chain pressure forms part of a cohesive framework in which robust social policies support technology adoption, AI-driven innovations improve energy efficiency and logistics transparency, and these combined productivity gains reinforce broader, more inclusive economic growth [23]. Based on this framework, the following null hypotheses are structured as follows:

H1: 

There is no significant relationship between supply chain pressure and economic sustainability.

H2: 

The adoption of smart AI has no significant impact on economic sustainability.

H3: 

Economic inequality does not significantly affect economic sustainability.

H4: 

The proportion of renewable energy consumption has no significant effect on economic sustainability.

2.2. Literature Review

This section of the study presents a summary of past studies regarding the factors affecting economic growth. A growing body of literature highlights the significant influence of supply chain pressure (SC) on economic growth (EG). Studies using Granger causality, GMM, and variance decomposition methods consistently show that increased supply chain efficiency or pressures often stimulate growth across different contexts. For instance, global analyses [3] and research across OECD countries [24] have found that tighter or more responsive supply chains positively affect the GDP, with panel analyses for the EU [25] and G-20 nations [23] confirming both one-way and bidirectional linkages. Similarly, country-specific evidence from Tunisia [4] and multi-country studies [26] demonstrates that SC improvements or pressures enhance growth by optimizing production flows, reducing bottlenecks, and improving logistics reliability. Collectively, these studies underscore the findings of [27,28,29,30] which states that efficient and well-managed supply chains serve as a critical structural driver of economic expansion.

Renewable energy consumption (REC) has also emerged as a key driver of economic growth in multiple empirical contexts. Panel threshold regressions on OECD countries [22], ARDL analyses for India [21], and CSARDL studies in Latin America [31] uniformly indicate that increased renewable energy consumption fosters growth. Similarly, systematic reviews [32] and dynamic ARDL studies for G-20 nations [33] corroborate the positive association between renewable energy deployment and economic performance. These findings suggest that the shift toward low-carbon energy not only mitigates environmental risks but also generates tangible economic benefits, likely through energy cost savings, innovation stimulation, and the creation of green jobs. Hence, REC appears to be both a sustainable and growth-enhancing policy lever.

The relationship between economic inequality (EI) and growth, however, appears more nuanced. In certain contexts, higher inequality stimulates growth, particularly in developing countries and certain high-income nations, where capital accumulation or targeted investments can generate economic returns [10,17,18]. Conversely, studies in Tunisia [13] and analyses of 43 major economies [9] have found that higher inequality reduces growth, indicating that disparities in income and resource access can constrain consumption, innovation, and social stability. These divergent outcomes underscore that the effect of inequality on economic performance is context-dependent, mediated by institutional quality, human capital distribution, and policy frameworks. Policymakers thus face the challenge of balancing growth objectives with the equitable distribution of economic gains.

Artificial intelligence (AI) consistently emerges as a positive driver of economic growth across global and national studies. Regression-based and ANN analyses [8,15,19,20,34] confirm that AI adoption enhances productivity, innovation, and efficiency, which collectively translate into stronger economic performance. The evidence suggests that AI not only accelerates technological diffusion but also complements renewable energy and supply chain management by improving operational efficiency and decision-making. Integrating AI with policies on renewable energy, supply chain optimization, and inequality mitigation can therefore create synergistic effects, fostering sustainable and inclusive economic growth. These findings collectively highlight that technology-driven interventions, when combined with structural and social policies, are pivotal for long-term economic resilience (

Table 1. Summary of past studies.

Author(s)	Nation(s)	Period	Method	Findings
Supply Chain Pressure (SC) and Economic Growth
[3]	Global	Undefined	Granger Causality	SC → EG
[24]	OECD countries	1870–2009	FGLS	SC ↑ EG
[4]	Tunisia	1971–2012	Variance Decomposition	SC→ EG
[26]	75 countries	2000–2014	GMM	SC → EG
[23]	G-20 countries	1961–2016	Panel VEC	SC ↔ EG
[25]	European Union (EU)	1995–2019	GMM	SC ↑ EG
Renewable Energy and Economic Growth
[22]	OECD	1997–2015	Panel threshold regression	REC ↑ EG
[32]	Undefined	2010–2021.	systematically reviews	REC ↑ EG
[21]	India	1985–2021	ARDL	REC ↑ EG
[31]	Latin America	2003–2020	CSARDL	REC ↑ EG
[35]	Egypt	1990 to 2021	ARDL	REC ↑ EG
[33]	G-20 countries	1990–2018	dynamic ARDL	REC ↑ EG
Economic Inequality (EI) and Economic Growth
[13]	Tunisia	1984–2011	ARDL	EI ↓ EG
[17]	HIDC and LIDC	1960–2010	GMM	EI ↑ EG
[9]	43 major economies	1995–2019	Multi-Regional Input-Output model	EI ↓ EG
[10]	Singapore	1978–2019	wavelet techniques	EI ↑ EG
[18]	Developing Countries	1990–2017	AMG	EI ↑ EG
Artificial Intelligence (AI) and Economic Growth
[19]	Global	Not defined	static analysis	AI ↑ EG
[20]	United States	2010–2019	Regression	AI ↑ EG
[34]	Global	Not Defined	Regression	AI ↑ EG
[8]	Global	Not Defined	ANN	AI → EG
[15]	Global	2010–2022	ANN	AI → EG

Note: AI: artificial intelligence; ANN: artificial neural network; REC: renewable energy consumption; EG: economic growth. ↑: increase, ↓: decrease; ↔ feedback causality; →: one-way causality.

).

2.3. Gap in the Literature

Despite a substantial body of research showing that supply chain pressure, renewable energy consumption, economic inequality, and artificial intelligence each have significant, and often nonlinear, effects on economic growth, existing studies remain largely siloed by factor and methodology. Research on supply chain pressure has predominantly focused on mean-based unidirectional or bidirectional effects, renewable energy studies emphasize threshold effects and institutional heterogeneity, and economic inequality analyses reveal context-dependent outcomes. However, no prior work integrates these drivers within a unified, distribution-sensitive framework. Similarly, the impact of AI’s growth has been examined through static regressions and machine learning causality tests, but without precise temporal adoption measures or consideration of scale-specific feedback across the growth distribution. As a result, there is no comprehensive investigation of how AI, supply chain pressure, and economic inequality jointly shape economic sustainability—an effect that may vary across different quantiles and time horizons. This study addresses this gap by employing the Wavelet Cross-Quantile Regression (WCQR) approach to capture asymmetric, scale-dependent spillovers and tail effects among these critical drivers.

3. Data and Methods

3.1. Data

In this study, we examine the guardians of economic growth (EG) in the United States. The factors considered include supply chain pressure (SC), smart AI (AI), and economic inequality (EI). The data for EG has been obtained from https://fred.stlouisfed.org/series/A939RX0Q048SBEA (accessed on 8 August 2025), and it is measured as Chained 2017 Dollars, using the Seasonally Adjusted Annual Rate. The data for supply chain pressure has been obtained from https://www.newyorkfed.org/research/policy/gscpi#/overview (accessed on 8 August 2025), and it is measured as an index. Smart AI data has been obtained from https://www.investing.com/etfs/spdr-select-sector---technology-historical-data (accessed on 8 August 2025), and EI data has been gathered from https://ourworldindata.org/grapher/economic-inequality-gini-index (accessed on 8 August 2025), with income inequality used as a proxy for SEF. The Gini coefficient measures inequality on a scale from 0 to 1. Lastly, the data for REC has been gathered from https://www.eia.gov/totalenergy/data/monthly/index.php (accessed on 8 August 2025) and is measured in (Trillion Btu). Higher values indicate higher inequality. This study used the quadratic-match sum approach described by [36,37] to convert SC and EI data into quarterly data. The data for this study spans from 1999 Q1 through 2024 Q4.

3.2. Justification of Variables

By exerting supply chain pressure (SC), U.S. firms can streamline operations across every stage—from raw material procurement to final delivery—identifying inefficiencies, minimizing inventory buffers, and adapting more rapidly to shocks. This reduces unit costs and preserves output even in volatile markets, thereby supporting stronger economic growth. Smart AI (AI) further amplifies these gains by leveraging large-scale operational data to forecast demand accurately, optimize production schedules, and automate predictive maintenance, thereby boosting factor productivity and reducing downtime that would otherwise constrain quarterly output. Simultaneously, addressing economic inequality (EI) through progressive taxation, equitable wage policies, and inclusive access to education and capital broadens consumer spending, stabilizes aggregate demand, and strengthens human capital, all of which underpin more resilient and evenly distributed economic expansion. Lastly, increasing the share of renewable energy consumption complements these drivers by insulating the economy from fossil fuel price volatility, reducing environmental externalities, and stimulating investment in advanced industries, collectively securing sustainable per capita growth while safeguarding natural resources for future prosperity.

3.3. Empirical Method

This study employed the recently introduced Wavelet Cross-Quantile Regression method introduced by [38] and modified by [16]. This approach reveals distribution-specific effects across time–frequency scales, handling non-stationarity and lead–lag structure that standard quantile or wavelet methods miss.

In accordance with [16], we decomposed the original series into frequency-localized components, isolating short- and long-term fluctuations. $\mathrm{Let} \left\{X_t\right\} \mathrm{and} \left\{Y_t\right\}$ be two zero-mean, covariance-stationary series. Apply the Maximal Overlap Discrete Wavelet Transform (MODWT) suggested by [39] to obtain the following equations:

$$X_t=A_{X,t}^J+\sum _{j=1}^J D_{X,t}^j,$$

(1)

$$Y_t=A_{Y,t}^J+\sum _{j=1}^J D_{Y,t}^j$$

(2)

where $D_{\cdot ,t}^j$ are the detail (high-frequency) coefficients at scale $j,$ capturing fluctuations in the order of $2^{j-1} \mathrm{to} 2^j$ observations. $A_{\cdot ,t}^J$ is the smooth (low-frequency) approximation at the coarsest scale, $J$.

The recursive filtering operations are as follows:

$$A_i^1=\sum _k h_{i-k}^1{s}_k,D_i^1=\sum _k g_{i-k}^1{s}_k,$$

(3)

$$A_i^{j+1}=\sum _k h_{i-k}^{j+1}{A}_k^j,D_i^{j+1}=\sum _k g_{i-k}^{j+1}{A}_k^j,$$

(4)

where $h^j \mathrm{and} g^j \mathrm{are} \mathrm{the} \mathrm{level}\text{-} j$ low- and high-pass filters, respectively. This decomposition allows us to analyze relationships at multiple temporal resolutions.

Next, each wavelet component is transformed into its quantile function to focus on distributional extremes rather than mean behavior, using the quantile estimation suggested by [40].

We defined the empirical cumulative distribution function (ECDF) of $\left\{D_{X,t}^j\right\} \mathrm{as} F_{X,j}\left(x\right)$. The nth quantile is as follows:

$$Q_{X,j}\left(n\right)=inf\left\{x:F_{X,j}\left(x\right)\ge n\right\},$$

(5)

and similarly, for $\left\{D_{Y,t}^j\},\right.$

$$Q_{Y,j}\left(t\right)=inf\left\{y:F_{Y,j}\left(y\right)\ge t\right\}.$$

(6)

By mapping raw coefficients into these quantile scores, we isolate, for example, the upper-tail movements, which are critical when extreme events (e.g., market crashes or spikes) drive the relationship.

We then link the quantiles of $X \mathrm{to} \mathrm{those} \mathrm{of} Y$ within each scale to capture asymmetric and nonlinear dependencies.

At a fixed scale, $j,$ the conditional t-quantile of $D_Y^j$ is modeled given the $n\text{-}\mathrm{quantile}$ of $D_X^j$ as follows:

$$Q_{Y,j}\left(t\mid Q_{X,j}\left(n\right)\right)={\beta }_0(t,n,j)+{\beta }_1(t,n,j)Q_{X,j}\left(n\right).$$

(7)

Equivalently, letting $f_{n,j}\left(X\right)=Q_{X,j}\left(n\right) \mathrm{and} f_{t,j}\left(Y\right)=Q_{Y,j}\left(t\right),$

$$f_{t,j}\left(Y\mid f_{n,j}\left(X\right)\right)={\beta }_0(t,n,j)+{\beta }_1(t,n,j)f_{n,j}\left(X\right).$$

(8)

This multiscale, quantile-focused framework thus yields a nuanced view of the dynamic dependence structure, informing targeted policy or risk-management strategies that hinge on specific time frames and distributional tails. Figure 1 presents the flow of analysis.

4. Results

4.1. Descriptive Statistics

Figure 2 presents a concise summary of key univariate distributional statistics for our five-core series—AI, GDP growth (EG), renewable consumption (REC), economic inequality (EI), and supply chain pressure (SC). The bottom three panels report minima, maxima, means and medians: AI hovers around 3.7 on average (median = 3.56), with a range from 2.60 to 5.44, EG clusters tightly near 11 percent (10.78–11.14), REC averages around 6.17 (median = 6.25), SEF sits near 0.41 with virtually no spread (std = 0.01), and SC oscillates around zero (mean = 0.03) but spans −1.32 to 4.26. The middle three rows show skewness and kurtosis: SC is the most skewed (2.11) and heavy-tailed (kurtosis = 7.85), EI also displays high kurtosis (8.18), despite modest skew (0.51), while AI, EG, and REC exhibit only mild asymmetry (|skew| ≤ 0.71) and near-normal peakedness (kurtosis ≈ 1.6–2.4). Finally, the Jarque–Bera statistics at the top quantify overall departures from Gaussianity: EI (120.82) and SC (179.09) are extreme outliers, AI (10.50) and REC (9.62) also reject normality, and EG (4.61) barely satisfies the normality assumption at conventional levels. Color-coding by magnitude (red → blue scale) makes it immediately clear that only EG approaches normality, while fairness and logistics stand out as the most non-Gaussian series in our dataset.

4.2. Diagnostic Test Results

Figure 3 presents the p-values and significance levels for eight diagnostic tests—Difference Sign, Keenan, Mann–Kendall, Robust Jarque–Bera, SJ, Teräsvirta, Tsay, and White—applied to our five-core series (SC, EI, REC, EG, and AI). Each test is based on a specific null hypothesis (H₀) regarding the data’s structure, with an alternative hypothesis (H₁) asserting the opposite: Difference Sign (H₀: no serial dependence or trend in consecutive difference signs; H₁: dependence or trend), Keenan (H₀: symmetry around the median; H₁: asymmetry, often trend-driven), Mann–Kendall (H₀: no monotonic trend; H₁: significant trend), Robust Jarque–Bera and SJ (H₀: Gaussian distribution; H₁: non-Gaussian skewness or kurtosis), Teräsvirta (H₀: linear autoregressive dynamics; H₁: nonlinear dynamics), Tsay (H₀: linear structure; H₁: threshold/regime non-linearity), and White (H₀: homoskedastic residuals; H₁: heteroskedasticity). The heatmap’s white-to-deep-red gradient corresponds to rising statistical significance (lower p-values), and asterisks mark conventional thresholds (*** p < 0.01; ** p < 0.05; * p < 0.10). Virtually all series reject Difference Sign and Mann–Kendall at p < 0.10—indicating evidence of pervasive dependence or trends—while Teräsvirta and Tsay similarly register widespread nonlinearity, except in a few high-frequency cases. Normality diagnostics strongly reject Gaussianity for SC and EI (p < 0.01), moderately for REC and AI (p < 0.05), but barely or not at all for EG; the White test flags heteroskedasticity in REC, EG, and AI at high significance, marginally in SC, and not at all in EI.

Furthermore, SC is the most problematic parameter, rejecting normality (RobustJB, SJ), trend independence (Mann–Kendall), nonlinearity (Teräsvirta), and heteroskedasticity (White), all at p < 0.10. EI likewise fails across the board—only its variance remains stable—yet Keenan and Tsay confirm nonlinear asymmetries. Renewable energy consumption (REC) shows a strong trend and nonlinearity rejections, mixed normality results (RobustJB **; SJ nonsignificant), and clear heteroskedasticity. Artificial intelligence (AI) mirrors REC’s profile but adds SJ’s rejection of normality and White’s heteroskedasticity flag. By contrast, EG most closely meets classical assumptions—passing Keenan, SJ, and RobustJB tests for symmetry and normality—yet still exhibits dependence and nonlinearity via the Difference Sign, Mann–Kendall, Tsay, and White tests. Together, these diagnostics underscore the necessity of robust, nonlinear modeling techniques to capture the complex dynamics in each series.

4.3. Stationarity Test Results

Next, this study used the wavelet-based quantile PP (WQPP) suggested by [41] to decompose each series (EI, SC, AI, REC, and EG) into short- (1–4 quarters), medium- (8–16 quarters), and long-term (>16 quarters) components via MODWT, and then computed the Phillips–Perron statistic at each decile $\tau \in \{\mathrm{0.05,0.10},\dots ,0.95\}.$ Under the null $H_0 : “\mathrm{the} \tau \text{-}\mathrm{quantile}$ process has a unit root” (i.e., is non-stationary) vs. $H_1:$ “it is stationary,” we compared the estimated test statistic curve against critical values at the 10% (solid line), 5% (dashed), and 1% (dotted) levels. In the short-term panel, all series’ statistics lie below the 5% threshold, and often below the 1% threshold, across nearly every quantile; therefore, $H_0$ is strongly rejected (stationarity holds). In the medium band, most quantiles reject $H_0$ at 10% (with a few marginal cases), indicating weaker but still widespread stationarity. In the long-term panel, only some deciles—particularly for EG and REC—cross below the 10% line; therefore, non-stationarity cannot be ruled out at many quantiles for low-frequency fluctuations (Figure 4).

4.4. Wavelet Cross-Quantile Regression

Next, this study employed the Wavelet Cross-Quantile Regression. Figure 5 shows the effect of artificial intelligence (AI) on economic growth (EG) in the United States. The result shows that smart AI’s contribution to U.S. GDP growth is greatest when both AI adoption and output growth are in their lower tails—i.e., at the 10th–30th percentiles of AI and the 10th–30th percentiles of growth (Short-term CQR Heatmap). This indicates that when firms have only begun to integrate AI (low AI quantiles) and the economy is under-performing (low EG quantiles), each incremental improvement in AI yields outsized productivity gains by automating routine tasks and freeing up labor for higher-value activities (Acemoglu & Restrepo, 2019) [42]. Conversely, at high AI quantiles (τAI ≥ 0.7) and high growth quantiles (τEG ≥ 0.7), the slope remains positive but attenuates—which is consistent with diminishing marginal returns once AI capabilities and complementary skills are widespread [43].

In the medium term, we observe that mid-quantile pairs (τAI ≈ 0.4–0.6 and τEG ≈ 0.4–0.6) register the steepest slopes, reflecting a diffusion phase in which early adopters pass the productivity frontier to laggards [44]. Here, firms at middle levels of AI use see sustained growth benefits in an economy that is itself growing at moderate rates. In the long term, slope differentials across all combinations of AI and EG quantiles narrow: high-AI/high-EG pairs no longer outpace low-AI/low-EG pairs by as wide a margin, as both technology spillovers and up-skilling have permeated across the entire distribution [43,45]. Our AvgCQR summary—which averages the CQR slopes over all AI quantiles for each EG quantile—closely mirrors the full CQR surface, confirming that the strongest effects occur at lower growth quantiles (τEG ≤ 0.3), regardless of AI quantile (AvgCQR vs. CQR). Likewise, standard QR estimates at each EG quantile (conditioning only on AI’s marginal distribution) reproduce the same ordering of slopes across quantiles, although they understate tail heterogeneity insofar as they do not account for the joint AI–EG distribution. Together, these robustness checks affirm that smart AI impacts EG positively across the entire joint distribution of AI adoption and economic performance, with the greatest leverage when both variables are at their lowest quantiles.

Figure 6 shows the effect of supply chain pressure (SC) on economic growth (EG) in the United States. In the short-term CQR surface, the impact of SC on U.S. economic growth is highly heterogeneous across the joint distribution of SC and growth quantiles. The steepest positive slopes appear at moderate SC adoption levels (τSC ≈ 0.3–0.5) when growth itself is in the upper tail (τEG ≈ 0.8–0.9), whereas very low SC quantiles (τSC ≤ 0.1) exhibit neutral or slightly negative effects on low growth quantiles (τEG ≤ 0.2) (Short-term CQR Heatmap). This “threshold” pattern reflects the idea that initial investments in supply chain infrastructure yield little payoff until a critical mass of connectivity and coordination is reached, after which efficiency gains—through reduced transaction and search costs—rapidly translate into output expansions [26].

As SC deepens beyond the threshold, diminishing marginal returns set in, explaining the tapering off of slopes at the highest SC and EG quantiles [3]. Moving to the medium term, we observe a pronounced positive effect of even low-quality supply chain networks (τSC = 0.1) on very high growth states (τEG = 0.9), while mid-quantile combinations (τSC ≈ 0.3–0.6 and τEG ≈ 0.3–0.6) register more muted or slightly negative slopes (Medium-term CQR Heatmap). This reversal arises because high-growth environments—characterized by robust demand and capital availability—can leverage even modest SC improvements to unlock further productivity gains, whereas slower-growing economies face adjustment costs and resource misallocation when supply chain networks are only partially optimized.

Over time, as networks stabilize, the heterogeneity across SC quantiles diminishes, and benefits become more evenly distributed, reflecting the diffusion of best practices and managerial know-how throughout the supply chain [4,46].

In the long run, the CQR surface converges toward uniformly positive slopes across most SC and EG quantile pairs, with the largest gains concentrated at moderate SC (τSC ≈ 0.2–0.4) and moderate growth quantiles (τEG ≈ 0.2–0.4) (Long-term CQR Heatmap). This flattening mirrors the lifecycle of supply chain innovations: once core connectivity and information systems are established, incremental upgrades yield broadly similar productivity dividends across the board, albeit with slightly lower returns at the extreme upper tail due to market saturation [26]. Crucially, both the average CQR (AvgCQR) summary and standard quantile regression (QR) estimates replicate these patterns—stronger effects at lower-to-mid EG quantiles and tapering at the top—confirming that our results are not driven by outliers or methodological artifacts. Together, these findings demonstrate that improvements in SC consistently promote U.S. economic growth, with the magnitude and distribution of benefits evolving from highly skewed in the short term to broadly uniform in the long term.

Figure 7 shows the effect of renewable energy consumption (REC) on economic growth (EG) in the United States. In the short run, our CQR surface shows that the marginal effect of REC on U.S. GDP growth is highest when both REC and output are at their lowest quantiles (τREC ≤ 0.1, τEG ≤ 0.1), with slopes tapering off as we move toward higher REC and EG quantiles (Short-term CQR Heatmap). This “catch-up” phenomenon occurs because initial investments in renewable capacity—when penetration is minimal—unlock substantial gains by diversifying the energy mix, stabilizing prices, and reducing import dependence [47]. By contrast, economies already at high REC shares experience diminishing returns to further renewables deployment, as the easiest efficiency gains have already been achieved [48,49]. Both the average CQR summary and standard QR estimates reproduce this pattern—although QR slightly underestimates the extreme-tail slopes—confirming the robustness of our CQR findings.

Over the medium horizon, the CQR heatmap reveals that positive REC–growth slopes persist across almost all quantile combinations; however, they are particularly elevated when REC sits in the lower-to-middle range (τREC ≈ 0.2–0.4) and growth is moderate to high (τEG ≈ 0.5–0.8) (Medium-term CQR Heatmap). This reflects a learning-by-doing effect: as renewable firms scale up, unit costs fall and complementary industries (e.g., smart grids, storage) mature, amplifying productivity spillovers for economies already demonstrating solid growth [50]. Economies at the very top of the growth distribution can leverage even modest REC increases to fuel further expansion, whereas those in the lower growth tail see smaller incremental payoffs until critical capacity thresholds are met. Again, AvgCQR and QR checks corroborate these medium-term dynamics, underscoring that our CQR estimates are not driven by outliers or specification biases. In the long run, the joint REC–EG CQR surface flattens: positive slopes appear nearly uniform across all quantiles, albeit with a residual superiority at low REC/high EG combinations (Long-term CQR Heatmap). By this stage, renewable infrastructure, regulatory frameworks, and market institutions have fully diffused, so incremental REC additions yield broadly similar growth dividends across the board.

Figure 8 shows the effect of economic inequality (EI) on economic growth (EG) in the United States. We observed that in the short run, CQR reveals that higher economic inequality (EI) generally dampens U.S. GDP growth across most combinations of inequality and output quantiles (Short-term CQR Heatmap). The steepest negative slopes occur at mid to high EI quantiles (τEI ≈ 0.3–0.7) when the economy is also growing at moderate rates (τEG ≈ 0.3–0.7), suggesting that elevated inequality—reflected in uneven income distribution or limited access to capital—can suppress private investment and increase social friction before the benefits of more equitable resource allocation emerge [11]. Notably, at the extreme lower tail of both EI and EG (τEI = 0.1, τEG = 0.1), the slope approaches zero, indicating that when both inequality and output are already low, incremental changes in inequality exert little additional drag. Both the average CQR summary and standard QR line estimates reproduce these sharply negative short-term effects—although QR underestimates the depth of the mid-quantile penalties—confirming robustness.

Over the medium term, the CQR surface still shows predominantly negative EI–growth relationships, but the intensity of the drag eases for economies at the higher end of the growth distribution (Medium-term CQR Heatmap). In particular, when output is strong (τEG ≥ 0.7), even moderate increases in EI (τEI ≈ 0.2–0.4) result in only a mild negative slope, and in a few high-growth/high-inequality cells, the effect briefly turns positive. This pattern reflects that once initial adjustment costs are absorbed, structural reforms, progressive taxation, and targeted social investments can mitigate the adverse effects of inequality, fostering human capital accumulation and stabilizing aggregate demand [12,51]. Again, AvgCQR and QR robustness checks closely mirror the CQR surface, underscoring that medium-term inequality effects are weaker yet remain largely adverse for slower-growing segments [11].

In the long run, CQR reveals a bifurcated outcome: economies achieving sustained high growth (τEG ≥ 0.8) actually see positive marginal gains from reduced inequality when starting from high inequality levels (τEI ≥ 0.7), whereas lower growth states still register a modest negative slope (Long-term CQR Heatmap). This convergence and partial reversal align with evidence that in mature, high-performance economies, narrowing economic disparities can enhance social cohesion, improve workforce productivity, and ultimately support broader growth [10].

The uniform flattening of slopes elsewhere indicates that, over time, economies learn to integrate inequality reduction measures without sacrificing dynamism. Both the AvgCQR metric and standard QR estimates corroborate these long-run dynamics, demonstrating that the initial trade-off between inequality and growth attenuates and can even flip positive under favorable growth conditions [52].

4.5. Wavelet ANN Granger Causality Result

The study employed the Wavelet ANN Granger causality (WANNGC), which merges the Wavelet Granger causality suggested by Adebayo et al. (2025) [53] with ANN. Figure 9 presents the results of the WANNGC. On the original (undecimated) series, only REC significantly improves GDP forecasts (p ≈ 0.03), while economic inequality (EI), supply chain pressure (SC), and artificial intelligence adoption (AI) show no reduction in out-of-sample error (p ≈ 1.00 for each) when added to a pure GDP network (ANN Granger: Original). This indicates that, absent any scale filtering, only broad-spectrum movements in renewables carry enough nonlinear signal to predict aggregate output beyond GDP’s own lagged dynamics.

At the D1 (short-term) scale (quarter-to-quarter fluctuations), none of the four drivers—EI, REC, AI, or SC—yields a significant drop in forecast error (all p >> 0.10). High-frequency “noise” in inequality, clean energy deployment, supply chain improvements, or AI investment does not materialize rapidly enough to influence next-period GDP once its own recent history is taken into account.

Moving to D2 (short-to-medium term) and D3 (medium term), the signals begin to emerge. In D2 (roughly 1- to 2-year cycles), adding any single driver still fails to significantly improve predictions (p > 0.10), suggesting that multi-annual adjustments remain masked by transitional dynamics. By contrast, at the D3 horizon (2–4 years), REC and EI become marginally significant (p ≈ 0.18 and p ≈ 0.11, respectively), showing that sustained shifts in renewable capacity and inequality measures start to leave a detectable nonlinear imprint on growth once cyclical volatility is filtered out.

Finally, at the D4 (long-term) scale (>4 years), all four variables approach or cross conventional significance thresholds: EI → EG (p ≈ 0.07) and REC → EG (p ≈ 0.10) become marginally significant, and SC and AI also near p ≈ 0.25–0.35. This demonstrates that the fundamental causal drivers of U.S. growth—particularly inequality and the diffusion of renewables—unfold over multi-year horizons. Wavelet-based ANN Granger thus reveals that only when short- and medium-term noise is stripped away do these deep, nonlinear relationships emerge.

5. Conclusions and Policy Directions

5.1. Conclusions

By uniting supply chain pressure, smart AI, and economic inequality as the “Guardians of Growth,” we forge a powerhouse alliance that transforms waste into wealth, data into dynamism, and inclusion into innovation. Together, these four pillars ignite a resilient, equitable, and future-ready economy—proving that sustainable prosperity is not a choice but our collective imperative. In doing so, this study employed data from 1999/Q1 to 2024/Q4 for the case of the United States. Furthermore, this study employed Wavelet Cross-Quantile Regression (WCQR). This approach captures heterogeneous, scale-dependent relationships across the entire joint distribution of variables, revealing dynamic effects that traditional single-scale or mean-based methods obscure. Using Wavelet Cross-Quantile Regression, this study finds that smart AI, supply chain pressure (SC), and renewable energy consumption (REC) significantly drive U.S. economic growth, with the strongest short-term effects appearing when adoption and output are in the lower quantiles, reflecting threshold and diffusion dynamics. SC enhances growth once supply chain networks reach a critical level of connectivity, while REC generates substantial gains at low penetration levels, illustrating a “catch-up” effect. In contrast, economic inequality (EI) generally dampens growth, especially at moderate to high inequality levels; however, long-term reductions in EI yield positive returns in high-growth states by improving social cohesion and workforce productivity.

5.2. Policy Remarks

This study sets forth precise policy guidelines tailored to specific periods and quantiles as follows:

(a)

Short-term (10th–30th quantiles): Launch a Targeted AI and Renewable Acceleration Fund for SMEs in lagging regions, combining matching grants and zero-interest loans to deploy off-the-shelf AI and basic blockchain-enabled supply chain tracking (e.g., digital waybills, chain-of-custody verification). Pair this with mandatory upskilling programs and a Micro-Grid and Logistics Connectivity Initiative to integrate small-scale renewables (rooftop solar + batteries) at logistics hubs, thereby lowering operating costs and stabilizing power for new digital tools. Complementarity: AI automates routine tasks, blockchain ensures data integrity, and micro-renewables reduce energy-related constraints.

(b)

Short- to medium-term (mid quantiles ~0.4–0.6 for AI–EG; 0.3–0.6 for SC–EG): Introduce a Diffusion Accelerator Tax Credit for firms that push integrated AI–blockchain–renewables stacks above median adoption, phasing out incentives as adoption rises. Provide interoperability and data-sharing grants to support open APIs, smart-contract templates for freight and energy settlement, and IoT grid-ready standards. Expand Regional Green Grid Partnerships to co-finance community solar, storage, and EV logistics fleets in moderate-growth states. Complementarity: AI predicts demand and renewable output; blockchain executes auditable contracts; and clean power reduces OPEX, thereby amplifying diffusion benefits.

(c)

Medium- to long-term (upper tail ≥ 0.7): Shift to R&D-driven frontier grants funding university–industry consortia on next-gen AI (foundation models, autonomous agents), blockchain-enabled supply chains (privacy-preserving provenance), and advanced renewables (solid-state storage, green hydrogen, power-to-X). Include integrated testbeds at ports, warehouses, and microgrids. Grants should prioritize spillover to mid-quantile adopters via open toolkits, reference designs, and licensing requirements, ensuring frontier innovations diffuse across the distribution and sustain high-end growth.

(d)

Cross-cutting Fairness and Dynamic Evaluation: Implement a Phased Fairness Adjustment Mechanism linking incremental increases in minimum wages and social spending to GDP per capita quantile milestones, smoothing out short-term crowding-out effects while enhancing long-term human capital and morale. Establish a Quantile Adaptive Policy Dashboard that continuously monitors AI, SC, REC, inequality (EI), and growth metrics through CQR, AvgCQR, and QR measures, enabling the real-time recalibration of incentives and supports to target the quantiles and horizons yielding the highest marginal growth returns.

5.3. Practical Implications

Policymakers should transition from uniform incentives to quantile- and horizon-specific packages that evolve as adoption deepens. In the short term (10th–30th quantiles), targeted matching grants and zero-interest loans for SMEs in lagging regions—combined with mandatory upskilling, micro-logistics hubs, and pilot rooftop-plus-battery systems—can generate rapid efficiency gains by integrating off-the-shelf AI, basic blockchain traceability, and distributed renewables. As firms move toward the middle of the adoption distribution, a Diffusion Accelerator Tax Credit and interoperability/data-sharing grants should incentivize integrated stacks (AI + blockchain logistics + electrification), while Regional Green Grid Partnerships co-finance community solar, storage, and EV fleets to reduce digital OPEX and amplify network effects. At the technological frontier (≥70th quantile), governments should shift to R&D-driven frontier grants, incorporating real-world testbeds, shared compute resources, open datasets, milestone-based prizes, and tech-transfer mandates to ensure that breakthroughs diffuse back to mid-quantile adopters and sustain broad-based growth.

5.4. Limitations and Future Direction

While our CQR-based assessment and quantile-adaptive policy framework offer significant insights into how AI, logistics, renewables, and fairness drive growth across different economic states and time horizons, several limitations warrant consideration. First, the analysis relies on annual aggregate data, which may mask important intra-year dynamics and sectoral heterogeneity; future work should incorporate higher-frequency and firm-level observations to capture more granular adoption patterns. Second, our causal inferences rely on the assumption of weak exogeneity and network-wide diffusion. Extending the framework to allow for spatial spillovers and time-varying parameter quantile models would deepen understanding of regional interdependencies. Third, measurement error in variables like “transparent logistics” and socio-economic fairness could bias slope estimates, suggesting a need for richer composite indicators or instrumental-variable quantile techniques. Finally, while the proposed dashboard enables real-time quantile monitoring, piloting it within policy agencies and evaluating its performance in live decision-making contexts will be critical—both to refine the automated triggers and to ensure that quantile-targeted interventions translate into measurable welfare gains over time.