Trade Elasticities and the Ringgit

This post ought to make happy those who think that the Ringgit should appreciate to force Malaysian firms to be more competitive, as well as those who think the currency is being manipulated for export competitiveness.

Some background is called for here. What’s an elasticity? In simple terms, it’s the response of one variable to a change in another. In the present post, I’m calculating the response of exports and imports to a change in demand and supply variables, specifically the exchange rate (both nominal and real trade-weighted indexes), Malaysian GDP, as well as external GDP.

In the literature, for a change in the exchange rate to affect the trade balance, then the sum of the absolute values of the exchange elasticity of exports and imports have to be greater than unity (1). This is known as the Marshall-Lerner condition.

Since my formulation of exchange rates is different (for simplification purposes I use the reciprocal of the normal definition), then for the Marshall-Lerner condition to hold my calculated sum of exchange elasticities should be less than -1.

Some disclaimers before proceeding: I’m not using any lag structure (past values of variables), even though I really ought to – economic variables do not respond instantaneously to changes in other economic variables.

Also by rights, the variables should be used within a vector autoregression model rather than the single equation estimates I’m using here, to capture feedback effects and dynamics – but the models I’ve tried so far were unstable and generated some perverse results (Malaysian GDP granger causing World GDP for instance). I’m going to continue working on that, as using a VAR is the more technically correct approach (you run up straight against the Lucas critique otherwise), but the current estimates will have to do for now.

The demand model specifications are:

Log(exports) = log(exchange rate) + log(World GDP) + monthly seasonal dummies

Log(imports) = log(exchange rate) + log(Malaysian GDP) + log(exports) + monthly seasonal dummies

Data is at monthly frequencies (with interpolated GDP data), and the sample period is 2000:1 to 2007:12. World GDP was proxied by using a trade-weighted geometric average of US, Japan, EU, China and Singapore GDP (Malaysia’s five biggest trading partners). For the import demand functions, I had to use an ARCH model rather than standard OLS.

So what did I get?

For nominal exchange rates:

Exports
A 1% appreciation in the exchange rate results in a 1.55% fall in exports

A 1% rise in world GDP results in a 1.50% rise in exports

Imports
A 1% appreciation in the exchange rate results in a 0.28% rise in imports

A 1% rise in exports results in a 0.97% rise in imports

Changes in Malaysian GDP had no significant effects on imports

For real exchange rates:

Exports
A 1% appreciation in the exchange rate results in a 1.57% fall in exports

A 1% rise in world GDP results in a 1.49% rise in exports

Imports
A 1% appreciation in the exchange rate results in a 0.34% rise in imports

A 1% rise in exports results in a 0.99% rise in imports

Changes in Malaysian GDP had no significant effects on imports

So a depreciation in the exchange rate (“weaker” if you will) does result in an increase in the trade balance, as the Marshall-Lerner condition holds for both nominal and real exchange rates. Adding the export elasticity with the import elasticity yields:

Nominal
-1.55 - 0.28 = -1.83

Real
-1.57 – 0.34 = -1.91

…so both sums are less than -1.

Equally, an appreciation of the exchange rate will result in a deterioration of the trade balance, as exports fall faster than imports. At least, that’s what these figures suggest.

However, because I haven’t modeled feedback effects (except through the inclusion of the export series in the import demand equation), you should probably take these results with a bucket of salt. The fact that exports cause a one-to-one change in imports suggests that adding dynamics might change the picture dramatically, but I haven’t yet come up with a model I’m happy with.

I’m also working on breaking down the components of both exports and imports, which might yield further insights. One preliminary result that seems to be holding up: the exchange rate has no impact on imports of consumption or intermediate goods, which also suggests that the import demand functions I’ve estimated are questionable.

Technical Notes
1. Trade data from DOS
2. GDP data from IMF International Financial Statistics