Why I Built No Model

Three modelling approaches were available for the AGOA question and I reviewed all of them before choosing none.

A gravity model, of the kind described here, would have needed a specification prone to spurious results on a question about one country exporting to one destination. The model is built for many pairs, and this was not that.

A partial equilibrium model simulates one market at a time. Its strength is detail at product level. Its weakness is that it rests on a handful of behavioural elasticities, and badly estimated elasticities produce confident nonsense.

A computable general equilibrium model simulates the whole economy at once and is the most powerful of the three. It also takes two to three years to build, demands complete and consistent national data of a kind Southern Africa rarely has, and is usually out of date by the time it is finished.

So the study used trade and tariff data directly. See descriptive trade analysis.

Why this belongs here

Choosing a simpler tool is a methodological decision and not an absence of one. It should be argued for in the open, because the alternative is dressing up straightforward work in borrowed authority, and that is harder to check.

Further reading

• WTO and UNCTAD (2012). A Practical Guide to Trade Policy Analysis. free Read it.Sets out descriptive analysis, partial equilibrium and computable general equilibrium side by side, which is exactly the choice this note describes.

From An analysis of South Africa exports to the United States under the African Growth Opportunity Act, MCom in Management Practice (Trade Law and Policy), University of Cape Town, 2015. Read it in OpenUCT.