EXCHANGE RATE, NONTRADABLE GOODS AND THE TERMS OF TRADE

Abstract

The paper seeks to formulate and tests for Singapore a model of exchange rate determination focusing on non-traded goods and terms-of-trade shifts. The model in mind is a commodity-exporting economy where world terms-of-trade play an important role in relative domestic demand between tradeable and non-tradeable goods and creates fluctuations in real domestic income. The application of cointegration and error-correction techniques to the empirical models is a vast improvement over the traditional ordinary least square (OLS) regressions. Estimating the model in an error-correction framework allows the incorporation of both short and long run dynamics to be presented within the empirical model.

Often, studies have concentrated mainly on the tradable (manufacturing) sector while sidelining the service sectors. As Singapore moves towards becoming a knowledge-base economy with strong emphasis on service-related industries, the growing influence that the supply of non-traded goods will have on exchange rates cannot be ignored. As such, an important innovation to this article is the formulation of a new measure for non­ traded goods using detailed sectoral data. Furthermore, for a small, open economy like Singapore, variations in the terms of trade, can have devastating effects on the stability of our economy given our high dependence on imports.

Using both the one-step error correction technique and the two-step Engle-Granger methodology, the model estimates are quite plausible, largely in line with theoretical predictions ioth the exception of the significant absence of the monetary variable.

Justification for its insignificance lies in the possible endogeneity of the money supply. Both models have desirable empirical characteristics, including plausible error correction forms, evidence of strong support for cointegration and rapid convergence to the long-run equilibrium. A barrage of diagnostic tests, inclusive of parameter stability and residual tests, indicates that the estimated error correction models are parsimonious. Empirical findings that are unique to this paper include: an estimated adjustment factor of -0.98 meaning that any short discrepancy from the equilibrium exchange rate is expected to be rectified within the quarter which is not often obtained empirically.

We find a significant dominant role for non-tradeable supply and terms of trade adjusted real income in exchange rate determination for both error correction models. Although the structural models fail miserably in contrast to the nominal random walk model for out-sample forecasting performance, two important conclusions can be reached: 1. Structural models have far better forecast performances for longer horizons (Especially for the in-sample dynamic simulations). 2. Error-Correction models estimated using the two-step Engle-Granger methodology produced superior forecasting results than the one-step technique.

This study was borne out of frustration given the severe lack of empirical work in this area. It is rather surprising to find so few work done given the significant influence that fluctuations of the terms of trade and non-traded goods supply have in affecting Singapore's exchange rates and economic growth. In conclusion, the results obtained provide considerable support for non-traded goods supply and (terms-of-trade adjusted) real income as 'real' determinants of nominal exchange rates.