Trade Networks

DSTA

Ch. 2: Trade networks

Important concepts

  • directed networks

  • weighted networks

  • sorts, and their quantitites

  • time

discover multiplex networks

Towards Clustering

  • reciprocity

  • assorativity

  • discover hidden structures

The directed network model

Theme: discover non-trivial relationships among countries

look at how they trade and what they trade

Weighted networks

The BACI-CEPII dataset:

Bipartite networks

The country-to-product network induces country-to-country and product-to-product relationships.

Reconstruction

\[ C = M_{cp}\cdot M_{cp}^T \]

\[ P = M_{cp}^T\cdot M_{cp} \]

Analysis of neighbours

For a node i, let \(k_i\) be its degree.

For directed networks: \(k_i = k_i^{in} + k_i^{out}\).

The distribution of degree \(P(k)\) provides a signature of the network.

The average degree is denoted \(\langle k \rangle.\)

Reciprocity

For a given directed network, reciprocity is the probability that of having links in both directions between two vertices.

R measures how the economies of two countries become interconnected (or interdependent).

\[ r = \frac{L^\leftrightarrow}{L} \]

\(L^\leftrightarrow\): number of reciprocal links

\(L\): total number of links.

Assortativity

Do vertices tend to connect with those with similar/dissimilar degree? Compute

  • the avg. degree of node \(i\)’s neighbors:

\[ K_{nn}(i) = \frac{\sum_{\langle ij\rangle} k_j}{k_i} \]

  • Next, the avg. \(K_{nn}\) for the \(n_d\) nodes which have degree \(d\)

\[ K_{nn}(d) = \frac{\sum_{i:k_i=d} K_{nn}(i)}{n_d} \]

\[ K_{nn}(d) = \frac{\sum_{i:k_i=d} K_{nn}(i)}{n_d} \]

Are \(d\) and \(K_{nn}(d)\) close?

Does assortativity grow over time?

Balassa’s RCA

The export matrix \(M\) is an adjacency matrix which reprents a bipartite graph.

Each scalar value \(M_{cp}\) corresponds to the aggregated export of product \(p\) by country \(c.\)

We can compute fractional ownership of export, product by product.

Many countries export coffee, so none really controls it.

Italy exports 100% of Bergamot oil (cfr. Prince of Wales tea).

The Revealed Comparative Advantage (RCA) is in controlling a high fraction of some product.

\(\sum_{p^\prime} M_{cp^\prime}\): total value of export by country \(c\).

\(\textrm{RCA}_{cp} = \frac{\frac{M_{cp}}{\sum_{p^\prime} M_{cp^\prime}}}{\frac{\sum_{c^\prime} M_{c^\prime p}}{\sum_{c^\prime} \sum_{p^\prime} M_{c^\prime p^\prime}}}\)