شناسایی تاثیر قوانین و مقررات تجارت در حرکات نرخ ارز واقعی: شواهد از آسیا

Empirical examinations of the relationship between the real exchange rate and the terms of trade are hampered by the need to control for other factors. In some studies these controls are not considered, and in others they are imposed in an ad hoc manner. The use of a latent factor model for real exchange rate changes overcomes this obstacle. However, latent factor models have the disadvantage that it is not possible to identify the role of a particular observed variable in the model. This paper proposes a method of incorporating observed information on changes in the terms of trade into a latent factor model of real exchange rate changes, relying on the observed covariance structure of the data to identify and calibrate some of the parameters. The method is applied to annual real exchange rate data for six Asian economies and shows that the contribution of terms of trade volatility ranges up to 24% of real exchange rate volatility. These terms of trade effects are offset in a number of countries by the covariation between the latent factors and the terms of trade.

Two important sources of relative price change in any economy are the real exchange rate and the terms of trade. There are many theoretical linkages between these two prices, although the direction of the relationship is not theoretically established in all models. In addition there is an observed empirical regularity between changes in the terms of trade and changes in the real exchange rate for many small open economies. However, the empirical literature produces very mixed results on the relationship between the two variables, ranging from little empirical relationship to strong and significant correlations; see Devereux and Connolloy (1996) and Broda (2002), for example. The theoretical relationship between the terms of trade and the real exchange rate is complicated by potential endogeneity. Consider assuming the terms of trade as exogenously determined, which is a frequently made and often not unreasonable assumption. For a commodity producing country, an increase in the terms of trade, driven by an increase in export prices, results in an increased demand for and subsequent appreciation of the domestic currency. An increase in the terms of trade driven by import price falls may similarly result in a depreciation of the international currency. This is Edwards and van Wijnbergen (1987) ‘Proposition 2’ linkage between terms of trade shocks and the real exchange rate (‘Proposition 1’ deals with the terms of trade and tariffs). Broda (2002) illustrates a formal model of these linkages, and Lane (1999) claims a similar relationship as ‘mechanical’ in the presence of home bias in consumption. Both Edwards and van Wijnbergen (1987) and Long and Pitchford (1992) detail the importance of the source of the terms of trade shock in assessing its subsequent impact, and whether in fact ‘Proposition 2’ will hold in general. The introduction of other complexities in theoretical models often provides less clear-cut relationships between terms of trade shifts and the direction of real exchange rate movements. Edwards and Ostry (1992) analyse the case of capital controls, and find that the direction of the relationship depends on the relative size of the elasticities of intertemporal substitution for consumption and between importables versus non-tradeable goods. With an increasing degree of endogeneity in the model between terms of trade and real exchange rates the correlation between the direction of terms of trade and real exchange rates becomes less clear, as illustrated in the three cases examined in Benigno and Thoenissen (2003). For particular countries, such as Australia and Canada, the relationship between the terms of trade and the real exchange rate is generally observed to be very strong; for example, Karfakis and Phipps (1999) and the literature reviewed in Aruman and Dungey (2003) for Australia, and Amano and van-Norden (1995) for Canada. Wider evidence of significant relationships between these variables is found in Habermeier and Mesquita (1999), Mendoza (1995) and Broda (2002) who examine the relationship in developing countries. The results are not always consistent. Habermeier and Mesquita (1999) using a panel of 51 countries, including 26 developing countries, find that developing countries have smaller exchange rate effects from the terms of trade than developed markets, however, Mendoza (1995) using 30 countries, of which 23 are developing, finds that the terms of trade effects are slightly larger in developing economies than in developed markets. Further, Broda (2002) splits a panel of 75 developing economies into fixed and flexible exchange rate regimes and finds that terms of trade contributions to real exchange rate volatility are larger in developing countries with a more flexible exchange rate regime. The relative size of the contribution of terms of trade volatility to real exchange rate volatility in developing countries varies enormously. Mendoza (1995) finds up to 49% of total real exchange rate volatility due to terms of trade volatility; Broda (2002) reports an average of 13% contribution for fixed exchange rate countries, and up to 43% in floating regimes. Devereux and Connolloy (1996) report a range of between zero and 67% for Latin American economies.1 While some of these studies examine directly only the bivariate link between the terms of trade and the real exchange rate, for example, Karfakis and Phipps and Amando and van-Norden, most studies use the terms of trade as one control variable in explaining real exchange rate movements. There are good reasons to expect other factors to affect the real exchange rate; see Lane and Milessi-Ferretti (2000). The question addressed in this paper is what proportion of real exchange rate volatility can be attributed to volatility in the terms of terms of trade. The area of interest is the terms of trade effects themselves, however, it is desirable to control for all other aspects that may produce real exchange rate volatility. In a similar problem Broda (2002) identifies a VAR approach with macroeconomic control variables such as the degree of openness, financial development and fiscal policy. This choice of control variables is somewhat ad hoc, and in order to avoid it here a latent factor model approach is adopted. Latent factor models of exchange rate movements have been proposed by Diebold and Nerlove (1989) and Mahieu and Schotman (1994), for example. Dungey (1999) shows how the model of Mahieu and Schotman (1994) can be estimated in a three factor version by the recognition of arbitrage conditions. The current paper builds on that framework to incorporate the role of observed factors into the latent factor model. The usual means of incorporating terms of trade effects in a real exchange rate equation is to adopt a small open economy assumption and assume that the terms of trade are exogenously determined. In this paper the observed variables are allowed to covary with the unobserved factors. In particular, this allows us to recognize that the terms of trade for a particular country varies both in response to international economic conditions and events, and also to some extent in response to country-specific events. Despite the sample of countries being small open economies, we recognize that in these countries not all other macroeconomic conditions are completely independent of changes in the terms of trade for that country (for example, production may vary with terms of trade in export oriented economies such as East Asia). In this paper the covariation between the terms of trade and the unobserved factors is controlled using calibrations based on the statistical properties of the relationships between the terms of trade and real exchange rates over the sample data. Some experiments illustrate the degree of sensitivity to the calibration assumptions. The remainder of this paper is laid out as follows. The data is described in Section 2. Section 3 builds the latent factor model of real exchange rate changes, and explicitly develops the method for including the terms of trade as an observed variable potentially correlated with the unobserved factors. The empirical application of this model to a panel of five Asian economies is reported in Section 4. In particular, the results demonstrate that the independent contribution of the terms of trade varies quite widely over the sample, up to 24%, but that this is often offset by the effect of the correlation between the terms of trade and the latent factors. When the assumption of terms of trade exogeneity is maintained, such as in Broda (2002) and Mendoza (1995) the results generally understate the contribution of country-specific events—the extent of the understatement depends on the true level of correlation between the terms of trade and the country-specific latent factor. Concluding remarks follow in Section 5.

Measuring the contribution of volatility in the terms of trade to volatility in real exchange rates is a relatively uncluttered area of the existing literature. Previous estimates of this contribution in developing countries range widely, from almost negligible effects reported in Devereux and Connolloy (1996) to the almost half of total volatility reported in Mendoza (1995). To gain these estimates, a small open economy assumption is usually justified with Granger causality tests to demonstrate the exogeneity of the terms of trade from domestic economy conditions. The contribution of this paper is to develop an alternative model which explicitly takes into account the potential for correlation between domestic and international conditions and the terms of trade index of a particular economy. Real exchange rate changes are modeled as a linear combination of three independent latent factors, a fixed (numeraire) factor, a country-specific factor and a common world factor. In addition terms of trade changes are identified as an effect, and these are allowed to vary with both the country-specific and common world factor (the variation with the numeraire factor is set to zero). Calibration experiments show that the effects of varying the correlation between the terms of trade changes and the country-specific factor is not in the contribution of the terms of trade to real exchange rate volatility, but in the contribution of the country-specific factor. In general for the sample here, if the correlation of the terms of trade and the country-specific latent factor is understated, then the contribution of domestic country conditions (the country-specific factor) will be understated. An advantage of this approach is that it incorporates information from an integrated system of exchange rates, but produces results on the contribution of terms of trade volatility for each individual country’s real exchange rate. The application of this model to the six Asian economies of Thailand, Indonesia, Malaysia, Philippines, Pakistan and Sri Lanka, with bilateral exchange rates expressed against the US dollar, produces contributions of terms of trade volatility to real exchange rate volatility ranging up to 24%. The contributions are higher in the East Asian economies, and almost negligible in Pakistan and Sri Lanka.