COMMODITY AND DEVELOPING COUNTRY TERMS OF TRADE, WHAT DOES THE LONG RUN SHOW?

Recently, the debate surrounding the commodity net barter terms of trade and the terms of trade of developing countries has entered a new phase. In this JOURNAL, Spraos (I980) presents evidence of a 'stable' declining commodity, terms of trade and Sapsford (i985), although fl'nding 'instability' in the work of Spraos, argues in support of the 'stable' declining terms of trade hypothesis. Grilli and Yang (I988), henceforth GY, also provide evidence supporting this view and state the implications of this finding for developing-country terms of trade. A major contribution of GY is a published set of data on the main price indices with a high degree of consistency for the whole century.1 However, Cuddington and Urzua (I989) in thisJOURNAL, illustrate that the hypothesis of a 'secular deterioration' in the commodity terms of trade is not robust to the alternative claim that there was a once-and-for-all shift in commodity prices in the period I920-I. After taking this shift into account these authors find little evidence to support the view of a 'secular decline'. 2 This debate has important implications for the 'Prebisch-Singer' hypothesis of a 'stable declining commodity terms of trade' (see for instance Prebisch, I 950) . It also has implications for both development and stabilisation strategies adopted by commodity-dependent developing countries. A central issue in this analysis is the order of integration of the commodity terms of trade. A time series is said to be integrated of order o, I(o), if its mean and variance are constant, or roughly speaking if the series is stationary. A time series is said to be I(n) if the series requires first differencing n times before a stationary series is obtained. Alternatively, if adding a time trend is sufficient to induce stationarity, the series is termed 'trend stationary'. A 'trend stationary' series implies that the variable in question simply adjusts around a constant growth path, but an I(i) series is truly non-stationary and displays very different properties. A commonly cited example of an 1(i) series is the random walk. The order of integration of a single time series is related to the idea of cointegration of a number of time series. Consider two variables, X and Y, that are not themselves I (o) but are of the same order of integration. If the residuals from regressing X on Y, or vice versa, are I(o), or roughly speaking if the residuals are stationary, it can be concluded that the two variables are cointegrated. In turn this implies that there is some form of equilibrium