Complexity package on SSC. Update and example files.

[Charlie Joyez]

In 2019 I first launch a Complexity package on SSC which I detailed in this post.
The aim was to provide Stata users with an easy way to compute economic complexity indexes a la Hidalgo & Hausmann.

Since, the package greatly evolved, so I'll review here the major changes, and post some real-world example.
You can update it typing

ssc install complexity,replace

• The package now offers two computational methods for Economic Complexity (Method of Reflections, and eigenvector method), and one derived metrics in the literature (Fitness index).
• The input doesn't need to be expressed in Revelead Comparative Advantage Tables, but can be written in any "performance" indicator across a set of activities (exports value, employment, ...)
• The table input can now be expressed as a varlist or using mata or Stata matrices.
• The package is now autonomous and no longer requires additional packages such as moremata to run
• The package offers additional metrics of relatedness to be computed such as coherence (index of intra-relatedness) or complexity outlook index (potential of complexity).
• The package also returns the adjacency matrices of the feature space (e.g. product space)

To provide you with an example I took 2023 exports value by country/product (HS07 4 digits from BACI) in file Exports2023HS4d.dta attached
To compute the Economic Complexity Index, just run the following code (from the do-file also attached)

reshape wide Xij, i(iso3) j(HS) /*first transform the list (long format) to a matrix of specialization (wide format)*/

*Then compute complexity:
complexity, varlist(Xij*)

The lines of code below show some options

complexity, varlist(Xij*) diversity outlook /*To compute diversity and Complexity Outlook Indices in addition to ECI*/

*To compute the same ECI index but using the method of reflection
complexity, varlist(Xij*)  method(mr) iterations(40)

*To compute the fitness index
complexity, varlist(Xij*)  method(fitness)

The ranking obtained is in line with the various rankings available online (here or there) with most complex economies being Switzerland, Japan, Korea and Germany.


I hope you'll have fun computing your own complexity scores from any specialization (exports, employment, innovation,...) ,at any scale or for any actor (countries, counties, firms,..)

Read the help-file for more info. A full article may follow.

Best,
Charlie