Technical Summary: The World Trade Web: A Multiple-Network Perspective

Problem Statement

International trade (IT) is a fundamental driver of global economic development, yet existing literature largely treats the World Trade Web (WTW) as an isolated phenomenon. While the topological properties of the WTW and its evolution have been extensively analyzed, there is a significant gap in understanding its interplay with other complex macroeconomic and social networks. Specifically, the dissertation addresses the lack of empirical research on the structural and causal relationships between the WTW and two other critical global phenomena: International Migration (IM) and Foreign Direct Investments (FDI). The core problem is to determine whether these networks share topological properties, whether they are correlated, and to what extent they influence one another (e.g., whether migration or FDI acts as a complement or substitute for trade).

Methodology

The research employs a dual approach combining complex network analysis with advanced spatial and selection econometrics.

1. Network Analysis and Filtering

• Data Sources: The study utilizes the World Bank's Global Bilateral Migration dataset (1960–2000), the NBER-UN trade dataset (SITC-4), and a proprietary dataset on Transnational Corporations (TNC) control networks (representing FDI stocks) for the year 2010.
• Hypergeometric Filtering: To make weighted networks (trade flows, migration stocks) comparable and to identify statistically significant links, the author applies a hypergeometric benchmark. This stochastic filter normalizes flows based on the total trade/migration volume of origin and destination countries, retaining only links that represent a significant departure from random co-occurrence (p-value < 0.01). This converts weighted networks into binary, tractable networks for topological comparison.
• Topological Metrics: The study compares degree distributions, clustering coefficients, reciprocity, and assortativity (specifically disassortativity patterns) across the WTW, International Migration Network (IMN), and FDI Network (FDIN).
• Jaccard Index & RCA: To measure the overlap between networks, the Jaccard index is used. A new product classification is derived by ranking products based on the overlap between WTW and IMN, which is then compared against Rauch's (2002) classification of homogeneous vs. differentiated goods.

2. Econometric Approaches

• Spatial Econometrics (Chapters 2 & 3): To address network interdependencies and Multilateral Resistance Terms (MTR), the study moves beyond standard gravity models. It employs Spatial Durbin Models (SDM) and Spatial Error Models (SEM). Crucially, the spatial weight matrix ($W$) is not based on geographical distance but on topological distance derived from the migration network (IMN). This allows for the estimation of direct effects (bilateral) and indirect effects (third-country spillovers).
• Heckman 2-Step Selection Model (Chapter 4): To analyze the relationship between Trade and FDI while accounting for the high frequency of zero flows (extensive margin), the study uses the Heckman 2-step (H2S) selection model. This involves a probit model to estimate the probability of a positive flow, followed by a log-linear gravity equation on positive flows with a selection correction.
• Instrumental Variables: To address endogeneity (reverse causality between trade and migration), the study uses lagged migration stocks as instruments.

Key Contributions

1. A New Methodology for Network Comparison

The dissertation proposes a novel method to compare two distinct economic phenomena by filtering them through a hypergeometric benchmark and analyzing their topological overlap. This leads to a new classification of goods based on the similarity between trade and migration network topologies, offering a robust alternative to existing classifications like Rauch's.

2. Global Perspective on Migration-Trade Links

Unlike previous literature that often focuses on specific ethnic networks or single country pairs, this work analyzes the migration-trade relationship from a global, network-wide perspective. It explicitly quantifies both direct and indirect (network) effects of migration on trade.

3. Comparative Analysis of Trade and FDI

This is the first work to propose a broad-scale comparison between the WTW and the global FDI network (represented by TNC control structures). It moves beyond treating them as separate layers to investigate their joint topological properties and causal interplay across different economic sectors.

Key Results

Migration and Trade (Chapters 2 & 3)

• Topological Correlation: The WTW and IMN are disassortative and highly clustered. There is a significant overlap between the two networks, particularly for specific product categories.
• Product Classification: The new "overlapping" product classification (based on WTW-IMN similarity) identifies goods where migration significantly boosts trade. This classification confirms that migration has a stronger effect on "overlapping" goods than on non-overlapping ones, validating the approach against Rauch's differentiated goods classification.
• Direct vs. Indirect Effects:
  • Migration significantly boosts trade (direct effect).
  • Indirect (Network) Effects: The study finds a negative indirect effect (substitution/competition) for homogeneous goods. Migrants in third countries compete with the origin country's exports of homogeneous goods.
  • Conversely, for differentiated goods, the negative indirect effect disappears, and the total impact of migration remains positive and significant.
• Spatial Econometrics: Using the IMN as a weight matrix in a Spatial Durbin Model successfully filters out residual autocorrelation (MTR), confirming that migration networks are a valid proxy for multilateral resistance terms.

Trade and FDI (Chapter 4)

• Topological Similarities and Differences: Both WTW and FDIN exhibit power-law strength distributions and disassortativity. However, the WTW is significantly denser and more clustered than the FDIN.
• Complementarity vs. Substitution:
  • Distance: Trade and FDI are complements, and this positive correlation strengthens as geographical distance increases.
  • Upstreamness: The relationship is non-linear. For industries closer to final use (downstream), trade and FDI are complements. For upstream industries (raw materials), they tend to be substitutes.
  • Regional Trade Agreements (RTA): The presence of RTAs increases trade but reduces FDI, suggesting that trade agreements act as a substitute for investment.
  • Sectoral Analysis:
    • Manufacturing (Secondary): Strong complementarity between trade and FDI.
    • Services (Tertiary): Substitution effect; firms choose one channel over the other.
    • Primary Sector: Substitution when flows are in the same direction (export-export); no significant relationship when flows are in opposite directions (export-import).
• Asian Context: Asian countries (ASEAN + China) show larger trade volumes and a stronger complementarity with FDI, reinforcing their role as a global manufacturing hub.

Significance and Claims

The author claims that this work provides a "multiple-network perspective" that reveals complex, non-linear interactions between trade, migration, and investment that are invisible when these phenomena are studied in isolation.

• Theoretical Contribution: The dissertation demonstrates that trade is not an independent layer but is deeply intertwined with migration and investment networks. It validates the use of network topology to define economic "neighbors" and spatial weights, offering a more accurate way to model multilateral resistance than traditional geographical distance.
• Empirical Robustness: By introducing a new product classification based on network overlap and confirming it through spatial econometrics, the work offers a more robust tool for identifying which goods are most sensitive to migrant networks, particularly intermediate goods which traditional classifications often miss.
• Policy Relevance: The findings suggest that migration acts as a trade facilitator, especially for differentiated goods, while FDI and trade act as complements in manufacturing but substitutes in services. The study highlights that Regional Trade Agreements may inadvertently reduce FDI flows, and that distance plays a crucial role in determining whether trade and investment reinforce or replace each other.

The dissertation concludes that understanding the "World Trade Web" requires a multi-layered approach, as the structural properties and dynamics of trade are fundamentally shaped by its correlation with migration and foreign investment networks.