THE PURPOSE OF APPENDIX • Analyze a few of the more technical points underlying the econometric formulation in the text and to outline the rationale behind the specific choice we’ve made among the various competing procedures. THE FUNCTIONAL FORM
• X, the dependent variable, is measured as a value share in GDP in every
case where it is appropriate. Expressing the dependent variables as shares provides estimates of structural change as this term is commonly understood. When production increases, for example, value added in all sectors usually expands too, but it is the relative rates of expansion which determine the changes in the structure of production. THE FUNCTIONAL FORM • In a semilog equation, income and size elasticities are no longer constant for all values of income and size (as they are in double-log equations). When the quadratic log terms are included, the growth and size elasticities of any variable X are given by:
• where ,31, 12 , zyi, and 72 are the estimated coefficients of the
regression given above, and X is the predicted value corresponding to the level of income or size at which the elasticity is being computed. NONLINEARITIES AND MULTICOLLINEARIT • Theoretical considerations and empirical results have suggested the existence of nonlinearities in the relationships analyzed, even after the usual first step of transforming all or part of the data into logarithms. On the theoretical level, for example, growth elasticities for the shares of any aggregate cannot continue indefinitely to exceed or fall short of one, since the share has an upper level of 100 percent and is bounded from below by zero. • These nonlinearities can be taken into account either by retaining a linear formulation for subsets of data ranked by increasing values of per capita income and population or by fitting nonlinear functional forms. NONLINEARITIES AND MULTICOLLINEARIT • he log quadratic formulation is one way to take nonlinearities into account, and it has proven particularly useful in this case to represent the transitional range. However, extrapolations of the parabolic equations to either extreme of the income or size scales are subject to large margins of error. For this reason, in the figures of predicted values in Chapter 2 the estimated equations were not used to predict below an income level of $100 and above a $1000 level but were replaced by the actual observed averages of countries in these ranges. REAL INCOME, RELATIVE PRICES, AND EXCHANGE RATE CONVERSIONS • Every comparative study based on international cross-section data has to face the problem of how to make commensurable the value figures expressed in the various local currencies. The usual practice is to convert all domestic values into a common currency (U.S. dollars as a rule) through official exchange rates. The resulting international spread in per capita incomes undoubtedly exaggerates the true differences in real income. In our sample, for example, income per capita in the group of countries with lowest income (less than $100) represented in 1965 no more than 3 to 4 percent of per capita income in the richest nations (over $2000). 2 TYPES OF DIFFICULTIES IN THE INCOME COMPARISON
The Relatively Poor
Statistical Coverage In Relative Price Effect Low-Income REAL INCOME, RELATIVE PRICES, AND EXCHANGE RATE CONVERSIONS • In addition to the problems of collecting the basic data needed to estimate national product and income, the researcher has to confront the existence of a substantial nonmonetized sector in the economies of less developed countries (LDCs). A comparatively large share of economic activity in the poorer countries does not cross the market and has to be estimated to construct the national accounts. This estimation is not only subject to a wide margin for error but also often underestimates the extent of the nonmarket activities. REAL INCOME, RELATIVE PRICES, AND EXCHANGE RATE CONVERSIONS • The second problem related to exchange rate conversions has two distinct aspects. First, official exchange rates are not equivalent to equilibrium exchange rates. The heavy reliance of LDCs-particularly since the Second World War-on high tariffs, import quotas, export subsidies, and other trade interferences as substitutes for open devaluations, and the selective application of these instruments among different industries, gave rise to a price structure with little or no resemblance to relative international prices.