University of Surrey develops free software that brings complex trade analysis to ordinary laptops
Complex economic models used to analyse international trade could become more accessible to researchers with limited computing resources, thanks to free software developed at the University of Surrey.
The study, published in Econometrics: Computer Programs & Software, found that Capybara 2.0 substantially reduced the computer memory needed to run demanding statistical models, while remaining competitive with established alternatives on speed.
These models help economists examine trade between countries while accounting for differences between exporters, importers and trading partners. As those comparisons become more detailed, conventional calculations can exhaust a computer’s available memory, making an analysis impossible to complete.
Capybara tackles this problem by accounting for those differences without building the enormous tables of additional variables that conventional methods require. Developed by Mauricio Vargas Sepulveda, the redesigned software combines memory-saving calculations with checks intended to prevent misleading statistical estimates.
The study compared Capybara with Base R and two specialist statistical packages, Fixest and Alpaca, using international trade models. Tests were repeated 100 times on a standard Lenovo laptop. The largest benchmark contained 6,006 observations covering 1994 to 2006.
If a model needs more memory than your computer has, waiting longer will not solve the problem. You simply cannot run it. Reducing that demand can make the difference between being able to answer an economic question and having to leave it unanswered.Mauricio Vargas Sepulveda, PhD researcher in Economics at the University of Surrey
In the largest benchmark, Capybara used less than one megabyte of memory, compared with approximately 897 megabytes for the conventional base R approach. It also used substantially less memory than the specialist alternatives, which required approximately six megabytes for fixest and 54 megabytes for alpaca. These figures describe the particular model tested, rather than every possible application.
The new version also identifies situations in which a model cannot produce valid finite estimates, helping prevent distorted results. Another feature allows researchers to examine patterns towards the lower and upper ends of trade flows rather than focusing solely on the average.
Access to sophisticated economic analysis should not depend on having expensive software or powerful computing equipment. Bringing these tools together in a free package helps remove practical barriers, particularly for researchers working with limited resources.Mauricio Vargas Sepulveda, PhD Researcher in Economics
The software, documentation and code needed to reproduce the analysis are freely available, enabling other researchers to inspect the methods and apply them to their own data.
ENDS
Notes to editors
- For interviews with Mauricio Vargas Sepulveda, please contact: mediarelations@surrey.ac.uk
- The full study has been published in: Econometrics: Computer Programs & Software
Media Contacts
External Communications and PR team
Phone: +44 (0)1483 684380 / 688914 / 684378
Email: mediarelations@surrey.ac.uk
Out of hours: +44 (0)7773 479911