Beyond trade statistics: how much do exports actually contribute to domestic value added?

Global production has become fragmented and interconnected, giving rise to value chains where production stages span countries. A growing literature measures the national value-added embodied in exports, including accounting-based approaches that decompose gross exports into value-added components. Notable contributions include Johnson and Noguera’s (2012) measurement of bilateral trade in value added, applications to automotive production (Timmer et al., 2015), and analyses of international fragmentation (Los et al., 2015). Koopman, Wang, and Wei (2014, KWW) and Los, Timmer, and de Vries (2016, LTV) provided IO-based accounting and hypothetical extraction methods, respectively, to identify value-added components in exports. However, these standard IO approaches treat final demand as exogenous, neglecting that production inflows generate income, which then fuels household consumption and further production rounds. This paper aims to determine the economic impact of trade by incorporating the circular flow between production, income, and consumption into the IO framework. By endogenizing private consumption following Miyazawa (1968, 1976) and Sonis and Hewings (1993), the model reveals interdependencies between income and output generation processes. It implies that the value added of exports is not constrained by gross exports but can approach total value added, offering a complementary perspective that goes beyond trade statistics. The rest of the paper presents the conceptual framework, an empirical application, and conclusions.

The paper situates itself in the literature measuring domestic value added in trade. Johnson and Noguera (2012, 2017) introduced measures of value-added content of bilateral trade, with applications to global automotive production (Timmer et al., 2015) and analyses of fragmentation within and across regions (Los et al., 2015). KWW (2014) developed an accounting method to identify value-added components in gross exports using world IO tables, while LTV (2016) and Los and Timmer (2018) used the hypothetical extraction method (HEM) to quantify value added in exports and provided a unified framework for bilateral and aggregate measures. Further related work includes Borin and Mancini (2017), Wang et al. (2018), Arto et al. (2019), and Miroudot and Ye (2020, 2021). In contrast to these accounting-focused approaches, this paper adopts a multiplier perspective by closing the IO model with respect to households à la Miyazawa (1968, 1976) and Sonis and Hewings (1993), capturing income–consumption feedbacks that amplify trade impacts.

The paper extends the conventional multi-country, multi-sector input–output (IO) model by endogenizing household consumption, following Miyazawa (1968, 1976) and Sonis and Hewings (1993). In a two-country baseline, gross output in country s includes intermediate uses (domestic and imported), other final demand, and household consumption that is modeled as a fixed propensity to consume out of private income. Private income is assumed to be a fraction of value added (via value-added-to-output ratios), linking production to income and back to consumption. This yields an augmented system where the technical coefficient matrix is partitioned into production coefficients (A¹) and private consumption coefficients (A²), and output is solved as x = (I − A¹ − A²)⁻¹ Y, where Y is final demand excluding private consumption. The model captures larger multipliers than the conventional Leontief system because it internalizes the income–consumption circuit. To quantify the contribution of exports to domestic value added, the paper employs the Hypothetical Extraction Method (HEM). It computes: (i) Domestic value added of exports (VAXD) as the difference between actual value added and the hypothetical value added when all exports (intermediate and final) are set to zero for the considered country; (ii) Value added consumed abroad by final users (VAXC) by hypothetically canceling foreign final demand (including private consumption components abroad); and (iii) Value added of final exports (VAXF) by canceling the country’s foreign final demand only. Parallel definitions are provided in the standard IO framework (with exogenous household consumption), allowing computation of feedback measures—FD, FC, FF—defined as differences between the extended and conventional results, thereby isolating the contribution of the income–consumption linkages. The approach is generalized to M countries and N sectors: X = (I − A − A^M)⁻¹ Y, where A collects intermediate input coefficients and A^M the private consumption coefficients across countries and sectors. Value-added coefficients v (diagonal matrices of VA/output ratios) convert gross output to value added. All required components—intermediate transactions, private consumption shares, final demand excluding private consumption, and value-added ratios—are constructed from global IO databases such as WIOD (2014 edition). The paper details how to obtain A¹, A², Y, and v from WIOD use, final demand, and value-added blocks. Using HEM within this extended structure yields aggregate and bilateral measures of VAXD, VAXC, and VAXF. An illustrative two-country numerical example demonstrates how endogenizing consumption raises multipliers and increases VAXD, VAXC, and VAXF relative to the standard model, with feedback magnitudes depending on consumption propensities and the share of private consumption in final demand.

• From WIOD 2014 aggregates, intermediate exports dominate gross exports in most regions (e.g., Australia 84.4%, Brazil 74.2%, ROW 71.8%), while China’s intermediate and final exports are roughly balanced (49.9% vs. 50.1%). Intermediate imports exceed final imports in all regions.
• Standard IO results (ratios relative to gross exports E and value added GVA): VAXD/E ranges 74.4% (Japan) to 86.5% (United States); VAXD/GVA ranges 9.6% (United States) to 19.3% (China). VAXC/E and VAXF/E are lower, with VAXF/E led by China at 27.4% (6.5% of GVA), reflecting China’s specialization near final stages of global value chains.
• Extended model results (with income–consumption feedbacks) substantially increase contributions: VAXD/E exceeds 100% in all cases, ranging from 115.6% (China) to 231.9% (Brazil); VAXD/GVA ranges from 23.8% (United States) to 37.3% (EU27), indicating exports generate domestic value added beyond the accounting value embedded in exports.
• Extended VAXC/E also rises markedly (e.g., 210.3% United States; 231.2% Brazil; 112.5% China). Extended VAXF/E increases as well, with notable values for Brazil (115.5%), United States (56.0%), EU27 (53.1%).
• Feedback effects (extended minus standard) highlight the role of household demand: FD/E is highest in Brazil (154.8%) and the United States (127.9%), and lowest in China (33.9%); similar patterns hold for FC and FF. These asymmetries align with the ratio of private consumption to final demand/value added: China shows relatively low C/D (37.1%) and C/GVA (35.4%), while the United States shows high C/D (66.5%) and C/GVA (68.6%).
• Conceptually, VAXD in the extended model is no longer bounded by gross exports but by total value added, since domestic income–consumption loops induced by exports generate additional domestic value added not captured by export accounting alone.

The study addresses how much exports actually contribute to domestic value added by embedding the household income–consumption channel into the IO structure. Findings show that once private consumption is endogenized, exports trigger additional rounds of domestic production via income generation and induced consumption, so the effective contribution of exports (VAXD) can exceed gross exports and approaches total value added as private consumption intensifies. Empirically, countries with higher private consumption relative to final demand and value added (e.g., United States, Brazil) exhibit larger feedback effects, while economies like China—with lower private consumption shares—show smaller feedback. This reframes the interpretation of trade statistics: accounting measures of value added embedded in exports understate the economic impact of exports on domestic production. The heterogeneity across countries indicates that structural features—consumption propensities, positions in global value chains, and import content of production—mediate the transmission from exports to domestic value added. Thus, policy analysis that incorporates consumption–income feedbacks may reach different conclusions about the sensitivity of domestic economies to external shocks and the benefits of trade liberalization.

The paper introduces a tractable framework that extends multi-country IO analysis by endogenizing household consumption, capturing the circular flow between production, income, and consumption. This resolves the standard IO limitation of exogenous consumption in trade impact studies and yields quantitative and qualitative insights: exports’ effective contribution to domestic value added is larger than measured by standard accounting decompositions, potentially exceeding gross exports and approaching total value added. Empirical results using WIOD 2014 show substantial country differences in feedback magnitudes, linked to private consumption shares, altering the perceived relative importance of economies in the global trade landscape. The approach is directly implementable with existing world IO databases and can inform trade, industrial, and economic planning policies. Future research should relax assumptions on fixed consumption propensities and explore substitution between domestic and foreign uses, integrate welfare analysis, and examine dynamic adjustments to better capture the breadth of trade impacts.

• The fixed relationship between income and consumption (rigid propensities) may overstate impacts, placing results at the upper bound of possible effects.
• The IO framework does not allow substitution between foreign and domestic final consumption or intermediate inputs.
• Results are limited to production indicators (value added) and do not capture welfare impacts.
• Model closure with respect to households abstracts from behavioral changes and price effects present in general equilibrium settings.