Annual average of the difference between real export growth and real GDP growth grouped by world bank income classification (high, upper middle, lower middle, low income). Crisis years (2008,2018,2020) show synchronized contractions across groups.
Source: Author’s calculation based on World Development indicators (World Bank, 2024)

This global trend uncovers substantial heterogeneity across development stages. Figure 4 decomposes the trade-GDP growth differential by World Bank income groups and reveals that in 2024 upper middle-income economies are now the primary drivers of global trade growth. Upper middle-income economies still undergoing industrial upgrading (e g., Vietnam, Mexico, Turkey) sustain a robust trade premium averaging 3.83pp (Table 9). Low-income countries show renewed engagement(1.35pp) possibly reflecting digital trade opportunities or regional integration.

In contrast, high income countries hover near parity(0.04pp), while lower middle-income nations actually exhibit trade underperformance(-0.58pp), likely due to infrastructural bottlenecks and premature deindustrialization (Rodrik,2008). The sharp post 2011 decline in the high-income trade premium (Figure 4) underscores a structural pivot toward non tradable services and in ward looking policy preferences

Table 9: Trade-GDP Growth Differential by Income Group (2024)

Cross sectional average of Trade-GDP differential in 2024, stratified by World Bank income groups. Positive values indicate export growth outpacing gdp growth; negative values indicate the reverse: sample;7,455 country year observations.
Source: Author calculations from WDI data

The divergent paths are further seen in the changing returns to trade openness. Figure 5 plots the relationship between trade openness and the GDP growth in two regimes pre and post 2009. Before the GFC, the slope is steep and positive aligning with Sachs and Warner’s (1995) hypothesis that trade openness spurs growth through efficiency and competition. After 2009, the slope flattens and the correlation weakens from -0.40 to -0.06 in dynamic terms. This suggests that openness alone no longer guarantees growth dividends, without complementary institutions, diversification or stock buffers, high openness may even intensify vulnerabilities (WTO,2023). Indeed, our panel regressions confirm that while trade openness remains statistically significant(p<0.05p), its marginal effect on export growth has diminished in magnitude and stability post crisis.

Scatter plot of trade openness against the GDP growth with OLS lines fitted for pre 2009 vs post 2009 subsamples. Regime split based on structural break identified via Chow Test (F = 15.4, p=0.03):
Source: Author calculations using WDI data

4.2. Services Trade Transformation

Service trade was once dismissed as a non-tradable residual in international trade models (Baumol,1967), it has since emerged as a central pillar of global commerce reshaped by digitalization, shifting comparative advantage and skill accumulation. Far from being passive, the service sector now exhibits distinct cyclical properties, structural momentum and crisis resilience that rival those of goods.

Using a comprehensive panel of 213 countries over 1990-2024, this analysis reveals that services are no longer a passive residual but an active and strategic component of global trade. The results in Table 10 reveal that only 50 countries (23%) are service dominant, while 163 (77%) remain goods dominant, however, economies with service dominant export structures demonstrated superior resilience with milder GDP declines and notably more stable export performance. As shown in figure 10, service dominance is concentrated in high income economies which consistently maintain service shares above 32% of total exports.

Table 10: Export composition and Crisis Resilience

Countries classified as service dominant (>50% services in pre- crisis total exports) or goods dominant (<50%) during crisis period (2008-2009) based on 213 countries.
Source: Author calculations using WDI data
Figure 6 shows the global share of services in total exports from 1990 to 2024. The series reveal a secular upward trend defined by sharp volatility, notable contractions in 1995(post Mexican peso crisis), 2000(dot com bust), 2011(Eurozone debt turmoil) and the 2021 supply chain distortions.

Figure 6: Global Service Share in Total Exports (1990-2024)

Annual cross country means of services share (service exports/total exports 100). Unweighted average across countries. Sample: N = 7,455 country year observations.
Source: Author calculations using WDI data

The 2009 surge amid the deepest global recession in decades is particularly telling, while goods trade collapsed by over 12% globally, digitally enabled services like IT outsourcing, business process services proved more resilient, even countercyclical in some economies. This pattern contradicts the traditional view of services as inherently non tradable and highlights the role of digital tradability in decoupling service exports from physical logistics constraints (Lopez & Jouanjean, 2017).

This transformation is not uniform across development stages. Figure 7 decomposes the service share by World Bank income group and reveals a pattern of ordered progression that aligns with the predictions of structural change theory. High income economies consistently exhibit the highest service export shares, reflecting their specialization in finance, digital platforms and professional services. In contrast low-income economies remain at the bottom, with service shares typically below 30% of total exports constrained by limited digital infrastructure, human capital and institutional capacity. Upper middle-income economies show moderate but rising shares, suggesting gradual upgrading into knowledge intensive services. This pattern aligns with the skill and technology intensity of modern tradeable services (Hoekman & Mattoo,2021).

Unlike goods which can be can be produced by unskilled labor and basic logistics, high value service exports demand advanced education, regulatory quality and connectivity, which are advantages concentrated in advanced economies. The persistent gap between service share to total exports amongst income groups points to the fact that digital globalization has not yet leveled the playing field, it may reinforce the existing divides unless accompanied by targeted investment in human and digital capital.

Annual cross country mean of services share (service exports/total exports 100) computed separately for each World Bank Income Group. Unweighted average across countries. Sample: N= 7,455 country year observations.
Source: Author calculations using WDI data

4.2.1 ICT Service Contribution

This divergence across income groups is largely driven by differential adoption and access to digital technologies, with ICT enabled services emerging as the primary source of modern service trade expansion. Yet the impact of ICT is not uniform over time. Figure 8 quantifies ICT’s contribution to annual growth in total service exports capturing not only its size but also its dynamic role as a growth engine. The series reveals striking volatility, ICT acted as a powerful accelerator in 1995 (3.12 pp) and 2022 (2.36 pp), periods aligned with digital commercialization and remote work diffusion.

Conversely, it became a significant drag during crisis years most notably in 2020 (-2.39 pp), when global digital service demand contracted amid recessionary pressures, and again in 2009 (-0.34 pp) and 2015 (-0.31 pp). this pattern reflects two forces; first, the skill biased nature of ICT services which concentrates gains in human capital rich economies (Autor et la.,1998); second, the high-income elasticity of digitally delivered services, which renders them procyclical despite low trade costs (Hoekman & Mattoo,2021). Far from being a stabilizing force, ICT enabled trade thus amplifies both booms and busts, reinforcing the advantage of economies that can smooth these cycles through diversified service portfolios.

Annual cross country means of ICT’s contribution to service export growth, computed as (ICT share of service exports) (annual service export growth rate. Values represent percentage points contributions. Sample: N = 7,455
Source: Author calculations using WDI data

Collectively, these findings challenge the prevailing theoretical framework. Services are no longer the residual sector but an active driver of trade dynamics, shaped by digital technology, human capital and evolving demand. Their rise redefines the concept of comparative advantage and challenges policy makers to invest in digital infrastructure, skills and regulatory frameworks that enable service led growth

4.3 Regional Divergence in Openness

The post-Cold War era was widely interpreted as one of converging trade policies, countries across regions liberalized, integrated into global value chains and steadily raised their trade-to-GDP ratios (Baldwin, 2016). This narrative was the key reason for optimism that globalization would be broadly inclusive.

Yet our analysis of 213 countries over 1990-2024 reveals that this convergence has not also stalled but also reversed. Instead of a unified path toward efficiency driven openness, the world now exhibits regionally stratified trade regimes shaped by institutional capacity, geopolitical alignment and risk resilience. This shift reflects a major adjustment in how trade and institution interact, consistent with endogenous trade policy models and the constraints of the globalisation trilemma.

A striking indicator of this shift is the rising dispersion in trade openness across nations. As shown in Table 11, the standard deviation of trade openness (% of GDP) across 213 countries surged from 64.3 in 2007 to 77.9 in 2008, the peak year of financial globalization and remained elevated thereafter (69.7 in 2011, 64.3 in 2010). This contrasts sharply with the 1990s, when dispersion gradually narrowed.

The Chow test confirms a structural break at 2009 (F = 15.4, p < 0.01), marking the end of the convergence regime. Critically, this divergence is not random, it aligns with endogenous trade policy models (Maggi & Rodriguez,2007). The analysis further shows that global average openness leveled off around 60% after 2008, but this masks a divergence, a small set of hyper-globalized hubs pulled the mean upward, while the majority of countries either stagnated or retreated.

Table 11: Top 5 years with the highest trade openness dispersion

Top 5 years ranked by standard deviation of trade openness. Only years with > 50 countries observed included. Dispersion reflects systemic divergence in integration levels. Sample: N=7,455 country year observations.
Source: Author calculations using WDI data

This divergence is fundamentally regional. Figure 9 plots the evolution of average trade openness by World Bank region and reveals divergent paths and distinct regional models of integration each shaped by structural characteristics, policy choices, and institutional endowments. East Asia & Pacific, Europe & Central Asia, Middle East & North Africa.

East Asia’s trajectory aligns with the ‘developmental state model (Wade, 1990; Amsden, 1989). The governments in these regions actively promote export led growth through infrastructure investment, credit allocation, and strategic tariff protection transitioning into GVC hubs. The region’s openness is not passive liberalization but state managed globalization, reinforced by strong contract enforcement and logistics efficiency (Levchenko, 2007).

Europe & Central Asia benefits from single market integration (EU) and post-Soviet reorientation toward Western markets. The EU’s regulatory harmonization drastically reduced non-tariff barriers, enabling high openness without proportional vulnerability (Baldwin, 2016).

Middle East & North Africa show high openness driven largely by commodity exports (oil, gas), which inflate trade to GDP ratios mechanically. However, non-resource sectors remain relatively closed, a pattern consistent with the ‘rentier state model’ (Lucas, 1985), where resource wealth reduces pressure for broad based trade reform.

On the contrary South Asia, Sub-Saharan Africa, North America persistently operate below 70% openness, but for fundamentally different reasons: Sub-Saharan Africa remains constrained by institutional and infrastructural bottlenecks, weak property rights, port inefficiencies, and fragmented regional markets amplify trade costs (Hoekman & Olarreaga, 2007). While regional agreements like African Continental Free Trade Area (AfCFTA) aim to reverse this, SSA’s openness reflects ‘premature deindustrialization’, an ability to sustain manufacturing led export growth, leaving economies reliant on volatile primary commodities. Rodrik (2018).

South Asia exhibits policy induced restraints. Despite growth, India and Pakistan have maintained relatively closed regimes due to import substitution strategies, strategic self-reliance doctrines (“Atmanirbhar Bharat”), and political economy resistance to liberalization (Panagariya, 2008). This reflects the endogenous trade policy model which argues that when domestic lobbies favor protection, openness remains suboptimal even amid growth (Maggi & Rodríguez-Clare, 2007).

North America presents a paradox: despite deep integration via United States Mexico Canada Agreement (USMCA), its aggregate openness stays modest (60–65%) because of its large domestic market size. As predicted by the Linder hypothesis (1961) and gravity models, large economies trade less relative to GDP simply because they can satisfy more demand internally (Head & Mayer, 2014). Thus, North America’s low openness points scale driven self-sufficiency.

The persistence of these clusters unchanged even after 30 years of globalization suggests that trade openness is not converging toward a global norm, but diverging along institutional and geopolitical lines. Institutions tend to determine whether openness yields gains or volatility (Levchenko,2007). economic scale inversely affects trade to GDP ratios and state strategy mediates global pressures and domestic priorities (Rodrik,2011)

Annual regional trade averages of trade openness (% of GDP), computed as unweighted cross country means within each region. Sample: N=7,455 country year observations.
Source: Author calculations using WDI data

4.4. Crisis Resilience Optimal Threshold

Contrary to both protectionist and hyper-globalization theories, we find a non-linear relationship between trade openness and crisis resilience. Countries with trade openness below 60% of GDP suffered significantly milder GDP contractions during the 2008-2009 GFC (-0.77%) compared to more open economies (-2.11%, p<0.05). This pattern holds across all tested thresholds (Table 12), with the worst outcomes concentrated among hyper open economies (>100%), such nations experienced both deeper recessions and higher cross-country volatility (σ = 3.94)

Table 12: Crisis Impact by Pre-Crisis Trade Openness Threshold (2008-2009)

Countries grouped by pre crisis trade openness (% of GDP) using bins (<30%, 30-60%,60-100%,100-200%,>200%). openness is measured as average over pre crisis years. Crisis impact measured over 2008-2009. Sample: N=7,455 country year observations
Source: Author calculations using WDI data

The above findings align with Rodrik’s (2018) concept of smart globalization where he argues that maximal openness is not universally beneficial, instead countries should regulate integration to their institutional capacity. In the absence of strong buffers such as deep financial markets, diversified financial markets, social safety nets or diversified production, high trade openness leads to extreme external shocks (Levchenko,2007). Without robust institutions, hyper integration undermines rather than enhances stability (Rodrik,2011).

Critically, the most resilient group isn’t the least open, but the moderately open (30-60%), this suggests that partial integration allows countries to capture efficiency gains while retaining policy space to absorb shocks. This supports the emerging consensus that the future of trade lies not in blanket liberalization but in prudent institutionally grounded openness (WTO,2023).

The era of convergence assumed a single rules-based trading system. Today we see emerging geoeconomic blocs (WTO,2023), a US aligned bloc prioritizing friend shoring, a China centered bloc emphasizing infrastructure led integration and a Global South navigating between them. The results in the regional divergence and openness thresholds are early evidence of this realignment, where trade policy is no longer about efficiency alone, but about strategic autonomy and shock absorption.

4.5. The Fragmentation of Global Trade Policy

The retreat from convergence is not merely regional or threshold-based, it is systemic. The analysis reveals a fundamental decoupling between macroeconomic synchronization and trade policy divergence, captured in the evolving cross-country dispersion of GDP growth versus trade openness (Figure 10). From 1990 to 1998, GDP growth was more dispersed than trade openness, income trajectories diverged but trade policy converged under the Washington Consensus and post-Cold War liberalization wave (Williamson, 1990). However, this relationship inverted after 1998, trade openness became consistently more dispersed than GDP growth through 2019. This represents a move away from globalization driven uniformity in trade policy toward an era of strategic diversity

Annual cross country Std.dev. of GDP and Interquartile range (IQR) of trade openness. Dispersion measures computed only for years with >50 countries observed. Sample: N=7,455 country year observations.
Source: Author calculations using WDI data

This pattern aligns with endogenous trade policy models (Maggi & Rodríguez, 2007), which predict that as countries develop distinct institutional capacities and geopolitical alignments, their optimal openness levels diverge. The spike in trade dispersion during 2020 to 2021 while GDP dispersion fell further underscores this, the pandemic exposed asymmetric supply chain vulnerabilities, prompting nations to recalibrate openness based on resilience, not just efficiency (Baldwin, 2022). By 2022 through 2024, trade dispersion remained elevated, confirming that the shocks accelerated and not reversed fragmentation.

This pattern aligns with WTO (2023) diagnosis of re-globalization along bloc lines, rather than a single system, we now see global trade structured around distinct spheres (US aligned, China-centered, Global South), each with trade norms. This dispersion is a systematic feature if the current order.

Traditional trade theory assumed that openness would converge as information costs fell (Helpman &Krugman, 1985). But institutions matter, Levchenko (2007) argues that only countries with strong property rights and contract enforcement benefit from openness, others face volatility without gains. Our dispersion results confirm this, globalization has strained the world tiers of integration, each with its own risk return trade off.

Critically fragmentation isn’t random it correlated with institutional quality where high-income OECD stabilizes, geopolitical alignment especially East Asia deepening China links, commodity dependence in MENA fluctuates with oil prices. Thus rising dispersion isn’t a failure of globalization it’s a logical evolution in a multipolar world.

4.6 Extreme Trade Hubs as Structural Outliers

While most countries conform to a broad global relationship between income dynamics and trade openness, a small set of economies consistently defy this pattern. Using a global OLS regression of trade openness (% of GDP) on GDP growth, we compute residuals, the gap between actual and predicted openness for each country and year. Then, for nations with at least 10 years of data, we calculate the mean absolute residual, which measures how persistently a country deviates from the global norm, irrespective of direction.

The results in Table 13 reveal ten extreme outliers. Singapore leads with an average deviation of 260.9 percentage points, followed by Djibouti (242.5), San Marino (237.8), and Hong Kong SAR (237.0). These economies operate outside the standard trade-growth linkages that govern most of the world.

These deviations reflect structural exceptionalism. For Singapore and Hong Kong, high residuals come from their roles as global logistics and financial intermediaries, their trade flows include re-exports, intra-firm transfers, and digital routing that bear little relation to domestic GDP. Djibouti’s deviation arises from its strategic position as a trans-shipment hub for East Africa, while Luxembourg and Ireland reflect Base Erosion and Profit Shifting (BEPS) driven distortions i.e. intellectual property shifting that inflate trade statistics without corresponding reals sector activity (Wier, 2019).

Table 13: Countries with Consistently Atypical Trade GDP Dynamics

Countries ranked by mean absolute residuals from global OLS regression of trade openness on GDP growth. Residuals = actual – predicted trade openness. Only countries with >10 years of data included. Sample: N = 7,455 country year observations.
Source: Author calculations using WDI data

Critically, this behavior is institutionally embedded. As Rodrik (2018) observes, such jurisdictions succeed not through comparative advantage in production, but through comparative advantage in arbitrage exploiting regulatory, geographic, or fiscal gaps in the global system. Their institutions are optimized for global network efficiency, not domestic inclusivity (Acemoglu & Robinson, 2012).

Even these outliers show signs of strain. When plotted over time (Figure 11), their residuals once stably above 200 pp begin declining after 2014 and drop sharply between 2020 and 2021. This suggests that geoeconomic fragmentation is undermining the neutrality that once made these hubs indispensable. As supply chains reconfigure along bloc lines (U.S., EU, China), pure intermediation models face pressure to align politically eroding their value as impartial nodes (WTO, 2023).

These outliers have profound implications for how we measure globalization. Because their trade flows are decoupled from domestic production, including them in global aggregates creates a misleading picture of integration. Their exclusion of them in earlier analyses (Figure 3) was essential to reveal the true trajectory of the global majority where openness has plateaued or retreated. Moreover, their existence warns against policy imitation. As Baldwin (2022) cautions, Singapore is not a model it’s an exception. Most countries lack the institutional, geographic or political conditions to replicate enclave based hyper openness. Attempting to do so without complementary capacities risks amplifying vulnerability without capturing gains.

In summary, exceptions like Ireland and Luxembourg are better understood as features of divergent global economic order than as leaders of convergence. Their specific vulnerabilities highlight how even the highly efficient, specialized hubs are exposed to the pressures of strategic decoupling and shift toward resilience based national policies.

Annual residuals (actual – predicted trade openness) from pooled OLS benchmark. Horizontal dashed line at zero denotes global norm. Sample: N=7,455 country year observations.
Source: Author calculations using WDI data

Together, these patterns confirm that global trade isn’t converging toward a single trade model. Instead, it is diversifying into several distinct categories; a large group of moderately open economies operating with in established institutional constraints; a set of nations with more variable economic performance due to commodity dependence or limited policy space; and a small number of highly specialized jurisdictions whose distinct operational model faces specific vulnerabilities to systemic shifts.

4.7: Volatility Clustering and Univariate Dynamics

Before modelling time varying correlations, we first established whether volatility itself was time varying and persistent in the individual series. this was done using the Engle’s (1982) ARCH-LM test on residuals of constant mean for goods export log returns. The test yields a p value of 0.03 rejecting the null hypothesis of homoscedasticity at 5% level. This provided statistical evidence of volatility clustering. Large changes in trade flows tend to be followed by large changes of either sign, a defining characteristic of financial and macroeconomic time series.

Volatility reflects underlying economic frictions, institutional arrangements, and structural rigidities. The univariate GARCH analysis reveals three distinct volatility regimes across goods exports, services exports, and GDP as shown in Table 14.

4.7.1 Goods Exports: Supply Chain Disruptions and Inventory Cycles

The GARCH (1,1) model for goods export log returns yields α + β = 0.93, indicating high but stationary persistence. This reflects the physical nature of goods trade disruptions like port closures, input shortages and logistics bottlenecks which spread through global value chains (GVCs) and take time to resolve. Firms respond with inventory adjustments and supplier diversification, processes that dampen but do not eliminate volatility (Carvalho & Salehi, 2019). Critically, the non-zero α = 0.182 implies sensitivity to recent shocks consistent with news driven inventory cycles (Bloom et al., 2018). When firms observe a demand shock, they immediately cut orders (high α), but full supply chain rebalancing takes years (high β). The result is a volatility half-life of nearly a decade (9.1 years) and evidence of deep structural inertia in physical trade networks.

4.7.2 Services Exports in a Digital Era.

In stark contrast, services exports exhibit IGARCH behavior (α + β = 1.0), a rare finding in macroeconomic flows. This suggests permanent shifts in volatility, consistent with the economics of digital service platforms. Unlike trade in goods, digitally delivered services such as software, cloud computing and consulting are subject to network effects that reinforce market leadership, low marginal costs of scaling that can amplify growth surges and Irreversible investments in specialised capabilities e.g. in where once a country specializes in IT services, it rarely reverts.

These features generate non-mean reverting volatility, a shock e.g., pandemic driven remote work boom permanently elevates the variance of service exports. There is no return to normal only adaptation to a new equilibrium. This aligns with theories of digital globalization as a step function process (Jensen & Kletzer, 2010), rather than a smooth trend.

4.7.3 GDP: Macroeconomic Stabilization and Domestic Buffering

GDP log returns show moderate persistence (α + β = 0.80) and negligible news impact (α ≈ 0). This reflects the buffering role of domestic institutions which have automatic stabilizers such as unemployment insurance and progressive taxation, diversified non-tradable sectors e.g. healthcare, education, construction and the Central bank credibility in anchoring expectations.

Unlike trade which is exposed to global shocks, GDP benefits from spatial diversification within borders. A collapse in exports may hurt manufacturing, but services and public sectors absorb labor and income. Hence, GDP volatility decays faster (half-life ≈ 3.1 years) and shows little reaction to daily news consistent with the ‘great moderation hypothesis’ (Stock & Watson, 2003), now extended into the post-Global Financial Crisis era.

We validate the GARCH specification using Autocorrelation Functions (ACF) plots (Figure 12). Before GARCH filtering, the squared log returns of goods exports show strong significant autocorrelation up to lag 1(~0.5), decaying slowly to ~0.01 by lag 5, a clear evidence of volatility clustering. After standardising the series by GARCH estimated conditional volatility, the ACF of squared standardized residuals become flat, indicating that the model successfully captures the time varying variance structure. This diagnostic confirms that GARCH (1,1) is sufficient.

Figure 13 plots annual goods export goods log return alongside GARCH estimated conditional volatility. Returns surge dramatically in 2009(GFC) and 2015(commodity bust, early trade tensions). Critically, while returns swing negative during crises, volatility remain non negative measuring the magnitude of uncertainty not direction. The sustained elevation of the volatility confirms that the world has not returned to the low uncertainty regime of hyper globalization. Consequently, global trade is now characterized by structurally higher volatility, driven by geo political fragmentation, supply chain reconfiguration and financial instability.

Returns computed as annual cross country mean of country pf goods export log returns conditional volatility derived from estimated parameters. Sample: N=7,455 country year observations.
Source: Author calculations using WDI data

In Summary, the univariate analysis confirms three critical facts: Volatility is time varying and persistent in trade and gdp flows especially in goods exports where shocks exhibit long memory; Trade has become structurally more volatile since the global financial crisis while GDP volatility has declined creating an asymmetry in risk exposure; The GARCH (1,1) model is well specified as validated by ACF diagnostics and provide reliable conditional volatility estimates for each series.

4.8: Time Varying Elasticities and Dynamic Shock Linkages.

Having established that trade has become significantly more volatile than GDP since 2008 (Section 4.6), the next question is, how do shocks to trade and output co-evolve over time? To answer this, we move beyond static correlations and estimate time varying linkages using a transparent, non-parametric approach suited to annual macro data.

With fewer annual observations, parametric multivariate volatility models (like DCC-GARCH) are prone to overfitting and convergence issues. Instead, we follow the approach of Rancière et al. (2008) in macro finance. We estimate time-varying volatility as the 5-year rolling standard deviation of global average log returns, we further construct standardized shocks by dividing returns by their local volatility and compute a 5-year rolling correlation between these standardized series. This method isolates changes in co-movement strength from changes in individual volatility levels, revealing how the relationship itself evolves.

As confirmed by our Chow test (F = 15.4, p < 0.01), 2008 marks a decisive regime shift. In the post-crisis era (2009–2024), trade volatility surged by 87.4% from 0.060 to 0.112 while GDP volatility declined by 21.4%, from 0.098 to 0.077 (Table 15). This finding overturns the ‘Great Moderation’ narrative by Bernanke (2004). While domestic economies benefited from automatic stabilizers and service led growth, global trade absorbed systemic shocks aided by automatic stabilisers and supported by the growth of services. The result is a new macroeconomic reality where nations are internally stable but externally exposed.

Table 15: The Post-2008 Structural Break in Volatility

Figure 14 reveal this stark regime shift. Before 2008, GDP growth was more volatile than trade, a reflection of domestic business cycle fluctuations. After 2008, this relationship flipped decisively, trade volatility surged by 87.4% (from 0.06 to 0.112), while GDP volatility declined by 21.4% from 0.098 to 0.077. the most volatile years for trade were 2009(0.150), 2011(0.150) and 2012(0.140) all crisis or post crisis periods.

This divergence reflects a fundamental shift, domestic economies have stabilized through automatic stabilizers, service led growth and monetary credibility, while global trade has become the primary carriers of systemic risk. The year-to-year autocorrelation of trade volatility (0.63) confirms structural uncertainty driven by on-going supply chain restructuring of supply chain restructuring and geopolitical fragmentation.

Annual Std.dev of trade and gdp log returns computed using a 5 year rolling window. Volatility series derived from country level data aggregated to global annual means. Sample: N =7,455 country year observations:
Source: Author calculations using WDI data

These volatility irregularities reshaped the trade-GDP relationship. Figure 15 shows that the average correlation between standardized shocks weakened from -0.40 (pre-2008) to -0.06 post 2008, indicating near total decoupling in normal times. Yet this baseline decoupling masks dramatic crisis-specific dynamics: In 2009 (GFC) correlation plunged to -0.81, an extreme negative co-movement, both GDP and exports collapsed, but exports fell faster, creating strong negative correlation in growth rates. This reflects a synchronized global demand shock transmitted through financial and trade linkages (Levchenko et al., 2010).

In 2020 correlation was -0.58 significant but less extreme. The weaker correlation suggests partial insulation, likely due to digital services resilience, diversified supply chains, and unsynchronized policy responses. In 2010, correlation hit -0.85, GDP rebounded strongly but trade lagged, revealing true decoupling in recovery.

Critically, the negative sign during crises captures asymmetric adjustment speed as firms cut cross border orders faster than they reduce domestic activity (Kahn et al., 2007), this causing trade to overshoot GDP declines. The high absolute correlation confirms extreme synchronisation just with trade leading the fall.

Figure 15: Dynamic Conditional Correlation Between Standardized Trade and GDP Shocks

Rolling correlation of volatility standardized log returns(GDP and Trade), each return is divided by its 5 year rolling standard deviation. Sample: N =7,455 country year observations:
Source: Author calculations using WDI data

In summary, these findings overturn the two pillars of globalization theory: First, trade is no longer a stabilizing force but an amplifier of global shocks. Second, the trade GDP link is no longer stable but sporadic, dominant in calm periods but explosive in crises. This defines fragile integration, where openness delivers efficiency gains in good times but exposes economies to outsized losses in bad times.

4.9 Panel Regression Insights: The Macroeconomic Foundations of Export Performance.

This section examines the core macroeconomic determinants of export performance using a random effects panel model for 213 countries from 1990-2024. We estimate separate models for total exports, goods and services exports to undercover potential heterogeneity in their drivers, a critical distinction often masked in aggregate analyses (Lodefalk,2014).

The regression results are presented in Table 16. All models show strong explanatory power with R-squared values between 0.73 and 0.89.

Figure 16 :Panel Regression Results for Export Determinants

4.9.1 Income Elasticity Is Strong but Divergent

World demand proxied by GDP dominates global trade outcomes, while revealing a profound structural divergence. Services operate under a different economic logic than goods, one less tied to financial markets and more anchored in domestic stability. Across all three models, GDP growth exhibits a high, stable elasticity; goods elasticity (0.635) and services elasticity (0.701) both statistically significant at the 1% level. For instance, a 1% increase in national income raises service exports by 0.701%,

The slightly higher elasticity for services reflects their dependence on human capital, innovations ecosystems and domestic demand for digital infrastructure, all tightly related to national income. In contrast, goods face supply chain bottlenecks and input constraints that dampen responsiveness. This finding overturns the narrative of “decoupled globalization” and reaffirms that trade flows remain highly sensitive to domestic economic conditions reflecting the dual role of GDP as both a proxy for productive capacity and domestic demand for imported inputs used in export production (Goldstein & Khan, 1985).

4.9.2 Exchange rate and FX Volatility.

In the total exports model (Model 1), the coefficient on exchange rate is 0.010 with p values of 0.83, this indicates that changes in exchange rate level interpreted as currency depreciation have no statistically significant effect on total export volumes. In aggregate, depreciation no longer reliably boosts export volumes, likely due to widespread invoicing in dominant currencies (e.g. USD) and pricing to market behavior that decouple export volumes from bilateral exchange rate movements. This challenges the central principle of trade policy focused on competitive devaluations. It suggests that in an era of complex global value chains and competition, non-price factors such as reliability, quality and integration into production networks dominate price competitiveness for aggregate trade flows (IMF,2015).

The most revealing result lies in the opposite responsiveness to exchange rate volatility. In the Goods exports model (β =-7.837e-06, p =0.04) showing a statistically significant negative effect. Implying that higher forex volatility reduces goods exports, this is consistent with classic trade theory, currency uncertainty raises hedging costs, disrupts pricing and deters investment in export-oriented manufacturing (Goldstein &Khan,1985). In the service exports model (β =8.919e-06, p=0.06) showing a marginally insignificant positive association, while not significant at a 5% level, the direction and magnitude suggest a real economic phenomenon.

During currency turbulence, firms may shift toward digitally delivered, dollar invoiced services like cloud computing, software and consulting as a hedge against local currency depreciation. This aligns with Hoekman and Mattoo’s (2021) finding that digital platforms enable pricing to market in stable currencies, insulating service exporters from local forex swings.

Such divergence is profound, financial markets constrain goods but may catalyze services. It explains why service dominant economies exhibited greater crisis resilience, they were not just less exposed but actively adaptive.

4.9.3 Price stability as the foundational determinant

Inflation exerts a consistent, negative effect across all export types (β ≈ -0.0047, p < 0.01). This aligns with menu cost and quality signaling models (Krugman, 1987), higher inflation increases price uncertainty, erodes brand credibility, and raises transaction costs for international buyers. For services where quality is intangible the cost of inflation is particularly severe. The result reinforces the importance of macroeconomic stability, countries seeking export success must prioritize price stability as much as trade liberalization.

4.9.4 Consistent returns to trade openness

Trade Openness remains a positive and significant predictor of exports, but effect is marginal (β = 0.15, 0.21) and declining in magnitude from goods to services reflecting a key insight from Section 4.3 that openness is necessary but not sufficient. Without complementary institutions skilled labor, digital infrastructure, regulatory quality, openness yields diminishing returns and heightened vulnerability.

The stronger effect for goods underscores their dependence on border measures i.e. tariffs and customs efficiency, while services rely more on behind the border enablers i.e. data laws and IP protection, factors not captured by aggregate openness metrics.

In summary, the results confirm we live in a dual track trade system, domestic stability i.e. GDP growth and low inflation is the foundation of export success. Digital integration not just openness determines service competitiveness, financial volatility i.e. FX and interest rates matters far less than traditionally assumed. This duality explains the paradox of modern globalization, aggregate trade may slow, but high value services accelerate. It also redefines policy priorities; for goods, stabilize currencies and streamline boarders; for services, invest in human capital and digital governance.

5. IMPLICATIONS FOR THEORY AND POLICY

5.1 Implications for Core Trade Theories: Accounting for Time Varying Dynamics

5.1.1 Limits of Constant Parameter Models

Our findings fundamentally challenge the constant parameter assumptions underpinning much of international trade theory. The extreme persistence in volatility (α+β=0.93 for goods) and the structural break in correlations (-0.40 to -0.06) suggest that trade elasticities are not stable over time but evolve with the global economic regime. Let’s consider three key failures uncovered in this study:

Elasticity instability, traditional trade models often relied on stable income elasticities as modelling conveniences particularly in policy applications where import demand was assigned benchmark income elasticities between 1.0 and 2.0 (Goldstein & khan,1985). We however find that such stability assumptions are untenable in the post 2008 era, the income elasticity of exports has not only declined to -0.65 but also varies systematically by sector and regime, reflecting deeper structural shifts in how trade responds to growth.

Non stationarity co-movement: The trade-GDP correlation shifted from -0.40 to near zero post-2008, with extreme episodic spikes during crises (-0.81 in 2009), this reveals state dependent linkages invisible to constant parameter models.

Volatility persistence: Trade volatility surged by 87.4% after 2008 and exhibits high autocorrelation (0.63), implying shocks have multi-year effects, this contradicts the homoscedastic and no serial assumptions underlying standard panel regressions (Wooldridge,2010).

These findings call for a paradigm shift toward time varying parameter models, regime switching frameworks, and state contingent policy rules. As Primiceri (2005) showed for monetary policy, allowing coefficients to drift over time reveals hidden structural breaks and improves forecasting. In trade, such an approach would treat openness as a dynamic strategy adjusted to global uncertainty and domestic capacity.

5.1.2 Reconciling Services Trade with Traditional Theory

Traditional trade theory from Ricardian’s comparative advantage, Heckscher-Ohlin’s factor endowments, and New Trade Theory’s increasing returns were built almost exclusively on the economics of tangible goods. Services were treated as a residual and non-tradable sector and assumed to be immobile, labor intensive, and insensitive to global integration. Our findings however demonstrate that this framework is misleading in the digital age. Three empirical facts from this analysis defy traditional models:

First, services exports exhibit higher GDP elasticity (0.701) than goods (0.635), contradicting the concept that services are domestically constrained. Instead, they scale with national income through human capital, digital infrastructure, and innovation ecosystems, factors absent from classical production functions.

Second, services are statistically insensitive to exchange rate volatility (β = 8.919e-6, p =0.64), while goods are significantly harmed by it (p =0.04). This overturns the assumption that all trade is equally exposed to financial markets. Digitally delivered services often invoiced in USD/EUR benefit from lower transaction costs and reduced direct exposure to bilateral exchange rate volatility, rendering traditional exchange rate pass through models obsolete.

Third, service dominant economies suffered far milder GDP contractions during the GFC (-1.11% vs. -3.94%), revealing services not as a passive buffer, but as an active stabilizer. This resilience stems from lower supply chain dependence, remote delivery capability, and pricing power, features incompatible with goods centered trade shocks.

Such patterns demand a new theoretical foundation for services trade, one that centers on digital platforms, intangible capital, and global pricing regimes rather than tariffs, transport costs, or factor proportions. Hoekman and Mattoo (2021) argue that services trade is governed by ‘behind-the-border’ regulations (data flows, IP rights, professional licensing), not ‘across the border’ barriers. Our results confirm this, services respond to domestic fundamentals i.e. GDP, institutions and not on external price signals i.e. exchange rates.

5.1.3 Limits of Conventional Trade Openness Models

Trade openness has been traditionally modeled as having a uniformly positive relationship with economic growth. More openness meant more growth, efficiency, and convergence, policy advice followed mechanically, liberalize borders, reduce tariffs and join global value chains. Our analysis however reveals a nonlinear relationship where openness can be a source of vulnerability as much as opportunity. Three findings redefine the economics of openness:

First, the resilience threshold: Countries with trade openness below 60% of GDP experienced significantly milder crisis contractions (-0.77% GDP impact) than those above 60% (-2.11%). This contradicts the more is better dogma and confirms Rodrik’s (2018) “smart globalization” hypothesis. Openness must be calibrated to institutional capacity, without strong domestic buffers, fiscal space, regulatory quality and social safety nets, high openness intensifies external shocks rather than smoothing them.

Second, systemic fragmentation: Cross country dispersion in trade openness has surpassed GDP growth dispersion since 1998, signaling the end of policy convergence. Since global trade is no longer converging toward a single model, it’s now instead characterized by distinct tiers; a small group of extreme hubs (e.g. Singapore, Djibouti); economies with volatile performance tied to commodity cycles; and a large group of moderately open nations. This stratification limits the usefulness of aggregate measures of global openness which are distorted by unrepresentative hubs and underscores the need for differentiated policy frameworks.

Third, the hub illusion: Extreme trade hubs exhibit average deviations from the global trade GDP norm exceeding 200 percentage points, but their model is institutionally exceptional and not generalizable. Singapore for example thrives on authoritarian efficiency; Luxembourg on tax arbitrage; Djibouti on geographic monopoly. Attempting to replicate these models without complementary institutions leads not to convergence, but to risky exposure.

Together, these insights demand a new economics of openness, one that treats it not as a policy target, but as an equilibrium outcome shaped by domestic institutions, geopolitical alignment, and risk tolerance. The goal is no longer maximal openness, but optimal resilience, the level of integration that maximizes gains while minimizing fragility. As our panel regressions confirm, GDP growth and macroeconomic stability, not border liberalization are the true drivers of export success.

5.2 Policy Implications Amidst Persistent Volatility

5.2.1 Rethinking Trade Stabilization Policies

The extreme persistence of trade volatility (9.1year) half-life for goods means that temporary stabilization measures are inadequate. Traditional countercyclical policies designed for business cycle fluctuations (typically 2-5 years) may be mismatched with trade volatility that persists nearly a decade. This therefore suggest:

Longer Policy Horizons: Trade stabilization policies need multi-year frameworks rather than annual appropriations. This could include longer-term trade credit guarantees, multi-year infrastructure investments for trade resilience, and extended adjustment assistance programs.

Volatility-Contingent Policies: Since volatility itself is persistent and predictable to some extent, policies could be explicitly conditioned on volatility regimes. For example, trade finance support could automatically scale up when volatility exceeds certain thresholds.

5.2.2 Services Trade as a Stabilization Tool

The superior crisis resilience of service-dominant economies (-1.20% GDP impact vs -1.81% for goods-dominant) suggests that services trade development can enhance macroeconomic stability. This represents a potential "double dividend": services contribute to growth whilst reducing volatility exposure. We therefore recommend:

Services export promotion: Targeted support for sectors with both growth potential and stability characteristics e.g. ICT, education and professional services.

Digital trade infrastructure: Investment in broadband, data protection frameworks, and digital payment systems to enable services trade.

Services Trade Specific Risk Management: Insurance and financing products tailored to services different risk profile.

5.2.3 Optimal Openness Strategies

The finding that moderate openness (30-60% of GDP) delivers the best crisis outcomes creates an optimal policy mx that many countries may wish to target. Low-income countries should prioritize regional integration over global exposure, this lowers logistical risk. Middle income countries should use openness selectively by liberalizing services and digital trade while maintaining safeguards on critical goods e.g. food, energy and health. Therefore, trade agreements could include provisions that enhance rather than undermine resilience, such as diversified sourcing requirements, crisis cooperation mechanisms, and adjustment support.

5.2.4 Exchange Rate Policy Reconsidered

The statistically insignificant effects of exchange rates on exports (β=0.010, p=0.83) challenges decades of trade policy focused on competitive devaluations. This suggests that, countries should prioritize exchange rate stability over depreciation, countries may gain more from stable exchange rates that facilitate long term trade relationships than from competitive devaluations that offer temporary advantages.

Additionally, since export success increasingly depends on non-price factors like quality and reliability, services components that exchange rate changes don’t affect. Policy should correspondingly shift toward quality upgrading and supply chain integration rather than price competition.

5.3 Implications for Business Strategy and Risk Management

5.3.1 Extended Risk Horizons

The persistence of trade volatility means businesses need to extend their risk management horizons. Traditional annual planning cycles may be inadequate when volatility shocks have effects lasting nearly a decade. This calls for: Multi-year rather than annual hedging of currency and commodity risks; Planning for different volatility environments rather than just point forecasts and supply chain resilience investments as volatility persistence makes supply chain disruptions more costly.

5.3.2 Differentiating Goods and Services Strategies

Our results show a stark divergence: goods exports are harmed by exchange rate volatility (p= 0.04), while services are unaffected (p = 0.64). This demands distinct strategies. Goods focused firms must invest in currency hedging, dual invoicing (USD and local currencies) to reduce exposure to financial turbulence. Services focused firms on the other hand should leverage their FX insulation as a competitive advantage expanding into volatile currency markets where competitors retreat, and pricing in hard currencies without volume loss.

5.3.3 The Value of Trade Openness Intelligence

The non-linear relationship between openness and resilience means that businesses need sophisticated measures of trading partner vulnerability. Simple measures of trade volume may be misleading; what matters is the composition and structure of trade relationships.

Companies may want to balance efficiency through concentrated sourcing from highly open economies with resilience by diversified sourcing including moderately open economies.

5.4 Limitations and Future Research Directions

While this study offers a comprehensive empirical portrait of post-hyper globalization trade dynamics, it is not without limitations. Acknowledging these boundaries not only strengthens the credibility of the findings but also creates a fertile ground for future research.

5.4.1 Data Limitations and Opportunities

While our analysis covers 213 countries over 35 years, several data limitations warrant consideration:

Services trade measurement: Services trade data remain less comprehensive than goods data, particularly for digitally delivered services and intra firm trade. Future work should leverage OECD and WTO data or firm level digital export records to isolate truly tradable, scalable services.

Disaggregated volatility: Our analysis at the goods-services level could be extended to sectoral or even product level volatility, which might reveal important heterogeneity masked by aggregation.

5.4.2 Methodological Extensions

This study uses rolling window and GARCH based approaches suited to annual macro data. However, three methodological advances could deepen insights: State space models with time varying parameters for instance Time Varying Parameter Vector Autoregressions (TVP-VARs) could formally test for structural breaks in elasticity and correlation regimes, moving beyond visual inspection; Machine learning clustering e.g. k-means on crisis response profiles could identify latent resilience archetypes beyond pre-defined groups i.e. goods vs. services; Network analysis of supply chains could map how volatility propagates through Global Value Chains (GVC), linking firm level fragility to national outcomes.

5.4.3 Unexplored Questions

Several critical questions remain open:

1. Does digital infrastructure mediate the services resilience effect? our results show service dominant economies are more resilient, but is this due to broadband penetration, cloud adoption, or regulatory frameworks?

2. How will climate shocks interact with trade volatility? The 9.1 years half-life of trade shocks suggests climate induced disruptions e.g. port closures and crop failures, these could have decade long trade consequences.

3. Can resilience optimal openness be quantified dynamically? A policy rule that adjusts openness thresholds based on real time volatility indicators could offer a new stabilization tool for developing economies.

6. CONCLUSION

6.1 Summary of Key Findings

This study has provided comprehensive evidence that global trade has entered a new volatility regime characterized by four fundamental shifts:

1. Increased and persistent volatility: Trade volatility has increased 87.4% since 2008 and exhibits extreme persistence, with half-lives approaching a decade for goods trade. This represents a transition from temporary fluctuations to persistent risk factors that require longer-term management strategies.

2. Structural decoupling: The correlation between trade and GDP growth has weakened from -0.40 pre-2008 to -0.06 post-2008, indicating changing transmission mechanisms that challenge standard open economy models.

3. Heterogeneous integration paths: We find an inverted U-shaped relationship between trade openness and crisis resilience, with optimal outcomes at 30-60% of GDP. This suggests that both hyper-globalization and protectionism may be suboptimal compared to strategic, moderate openness.

4.Structural resilience divide: service dominant economies suffered 72% smaller GDP contractions during the GFC (-1.20 vs -1.81) and exhibited greater recovery strength. services are also insensitive to exchange rate volatility (p=0.64) unlike goods (p=0.04) confirming a dual track trade system.

6.2 Theoretical Contributions

This study makes three core contributions that challenge and extend the theoretical foundations of international trade economics.

First, it overturns the assumption of stable trade growth linkages. Traditional models from the elasticities approach (Goldstein & Khan, 1985) to New Trade Theory (Krugman, 1979) treat income elasticity as a structural parameter or assume smooth co-movement between trade and output. Our evidence of declining elasticity (≈0.65), regime dependent correlation, and persistent volatility clustering demonstrates that these relationships are not static but state contingent and time varying. This demands a shift toward adaptive trade theory one that incorporates regime switches, uncertainty persistence, and crisis-driven recoupling as central features.

Second, it establishes services as a distinct trade regime. Classical trade theory was built on goods: their transport costs, tariff sensitivity, and factor intensity. Our finding that services exhibit higher GDP elasticity, insensitivity to exchange rate volatility, and superior crisis resilience confirms they operate under different economic laws. This validates emerging frameworks like the “digital trade paradigm” (Hoekman & Mattoo, 2021) but goes further by providing global macro evidence that services are not merely “less tradable goods,” but a separate channel of globalization governed by digital platforms, human capital, and institutional quality not borders or logistics.

Third, it redefines openness as a conditional, not universal, good. The canonical view holds that openness drives convergence (Sachs & Warner, 1995). Our identification of a resilience optimal openness threshold (~60%) and the systemic divergence between hubs and mainstream economies shows that openness is institutionally mediated. This supports Rodrik’s (2018) “smart globalization” hypothesis but adds empirical rigor by quantifying the trade-off between efficiency and fragility. The result is a new theoretical proposition that openness is an equilibrium outcome of domestic capacity rather than a policy lever in isolation.

Together, these contributions move trade theory beyond the hyper-globalization paradigm toward a framework that embraces heterogeneity, uncertainty, and adaptation as defining features of 21^st century globalization.

6.3 Policy Recommendations

Strategic rather than maximal openness: Export success is driven first and foremost by domestic factors. GDP growth and low inflation significantly boost exports across all sectors. In contrast, extreme openness without institutional backing increases vulnerability. Countries with trade openness below 60% of GDP consistently experienced milder crisis contractions, while hyper open economies suffered deeper downturns unless they possessed exceptional institutions (e.g., Singapore). Governments should therefore prioritize macroeconomic stability, human capital, and regulatory quality and regulate their degree of integration to that foundation.

Services trade promotion: Service dominant economies were far more resilient during the GFC and are insulated from exchange rate volatility. Governments should actively foster digitally tradable services e.g. fintech, software and education through investments in broadband, digital skills and data governance.

Longer term stabilization frameworks: With trade volatility shocks persisting for over nine years, traditional countercyclical tools are inadequate. Instead, governments should establish automatic trade stabilizers for instance contingent credit lines triggered by volatility thresholds and build strategic reserves of critical inputs whilst mandating supply chain stress testing for key industries.

De-emphasize exchange rate targeting for trade purposes: The insignificance of exchange rate level for total exports (p = 0.83) and the sectoral divergence in FX volatility effects show that currency depreciation is no longer a reliable export tool. Central banks should prioritize price stability over competitive devaluation and support private hedging markets instead, recognizing that currency policy alone cannot drive export success.

Together, these recommendations shift the goal of trade policy from maximizing flows to managing fragility

6.4 The Future of Global Trade

The findings of this study indicate that global trade is undergoing a structural transformation, with three key features likely to define its future trajectory.

First, trade openness will become increasingly stratified. The strong performance of extreme hubs (e.g., Singapore, Djibouti) is not replicable for most countries due to their unique institutional or geographic advantages. Meanwhile, countries with moderate openness (<60% of GDP) consistently exhibited greater crisis resilience. This suggests that future trade integration will not follow a single global path but will instead diverge into distinct tiers based on institutional capacity and geopolitical alignment.

Second, services will play an expanding role in trade resilience. Service dominant economies experienced significantly smaller GDP contractions during the GFC (-1.20% vs. -1.81%) and showed no sensitivity to exchange rate volatility (p = 0.64). As digital infrastructure expands, services are likely to account for a growing share of stable, high value exports particularly in countries with strong human capital and regulatory frameworks.

Third, volatility persistence will reshape policy horizons. The estimated half-life of goods export volatility shocks is 9.1 years, indicating that trade uncertainty is not transitory but long lasting. This implies that both governments and firms must shift from short term stabilization tools to multi-year risk management strategies, including strategic diversification, supply chain redundancy, and automatic stabilizers triggered by volatility thresholds.

These trends point not to a reversal of globalization, but to its reconfiguration around resilience, differentiation, and domestic institutional strength. Openness will remain important, but its benefits will be conditional on complementary domestic capabilities not guaranteed by integration alone.