Comparing international consumption patterns

Apart from Selvanathan and Selvanathan (2003), other studies on international comparisons of consumption patterns include Clark (1940, 1957), Chen (1999), Clements and Chen (1996), Clements and Ye (2003), Clements and Selvanathan (1994), Houthakker (1957), Kravis et al. (1982), Lluch and Powell (1975), Lluch et al (1977), S. Selvanathan (1993), Theil (1987), (1996), Theil et al. (1989) and Theil and Suhm (1981).

PPP rates are either produced by the International Comparisons Project (see, e.g., Summers and Heston 1991) or by some other body such as the famous Big Mac Index published by The Economist magazine (see Ong 2003). A related approach is to use a measure of the “equilibrium exchange rate,” computed from either econometrically estimated trade/current account equations, or a computable general equilibrium model (see, e.g., Clark and MacDonald 1998; Driver and Westaway 2001; MacDonald 2000; Montiel and Hinkle 1999; Williamson 1994, and Wren-Lewis and Driver 1998). A recent innovation in this area is by Lan (2003) who relates the equilibrium exchange rate to the steady-state value associated with a simple time-series model of the deviations from PPP. More recently, Chua (2003) compares cross-country incomes on the basis of food consumption patterns.

Of course, it could be argued that the budget shares themselves reflect international differences in the price of traded goods relative to nontraded. In view of the degree to which the food budget share (the largest share in most instances) is related to income per capita across countries (Chua 2003; Theil 1987; Theil et al 1989), it seems that this problem is relatively minor when compared to the problems of using prevailing exchange rates when making international comparisons. Nevertheless, it should be acknowledged that no measure is perfect and this qualification should be kept in mind.

For a detailed description of the data, see Clements et al. (2004).

A referee has pointed out that the cross-country average of the food budget share of 32% is somewhat higher than that reported in Theil et al. (1989) using the dataset developed by Kravis et al. (1982). While the countries and time periods involved are not identical, it seems that a major reason for the difference is that the commodity “food” in this study includes beverages and tobacco (Selvanathan and Selvanathan 2003, pages 72 and 127), while in Theil et al. food excludes beverages and tobacco.

Another way of making the same point is in the two-good case where q _i *=M/p _i is the maximum quantity the consumer can purchase of i (i=1, 2) with income M. Suppose the consumer is indifferent between the two specialised baskets (q ₁* , 0) and (0, q ₂*), so that u(q ₁*, 0)=u(0, q ₂*), where u(.,.) is the utility function. For any positive fraction λ, we can define the diversified basket (λq ₁*,(1−λ)q ₂*), which is feasible as p ₁ λq ₁*+p ₂(1−λ)q ₂*=M. Convexity of indifference curves implies that the diversified basket is always preferred to specialisation, that is, u(λq ₁*, (1−λ)q ₂*)>u(q ₁*,0)=u(0, q ₂*). See Dixit and Stiglitz (1977).

The previously mentioned “problems” associated with food consumption in Zimbabwe and Sri Lanka are also manifest in columns 4 and 5 of Table 3. As the food budget share, w ₁, for Zimbabwe is lower than that expected on the basis of its GDP, (1−w ₁)=W ₂ is higher, causing W ₂ H ₂ to also be higher, as indicated by the entry in row 45 of column 4 of Table 3. This in turn leads to the corresponding entry of column 5 to be lower. A similar argument, but in reverse, applies to Sri Lanka.

For a survey, see Clements and Selvanathan (1994, Sec. 9).

Here and subsequently, when we refer to prices elasticities averaged over countries, it is understood to mean the elasticities obtained from Eq. 5.2 for i, j=1,..., 8 evaluated with ϕ=−0.41 and the relevant cross-country averages of the income elasticities and budget shares.

We follow the previous approach of defining countries 1–25 as “rich” and the remaining 18 as “poor.”

It is worth mentioning that Theil and Moss (2003) consider the related problem of how a country qualifies for admission to the Group of Seven. This does not seem to be on the basis of population nor GDP per capita alone; e.g., the world’s most populous country, China, is not a member, while some rich countries, in GDP per capita terms, such as Norway, Switzerland and Australia, are also excluded. Accordingly, Theil and Moss postulate that eligibility for membership (E) is “produced” with inputs population (P) and GDP per capita (G) with constant returns to scale Cobb–Douglas technology, \(E{\text{ = }}P^{\delta } G^{{{\text{1 - }}\delta }} \), where δ is a parameter lying between zero and one. Theil and Moss seem to favour the value δ≈.2 and note that δ≈.5, whereby population and GDP are equally weighted, is too high as this amounts to eligibility being \(E{\text{ = }}{\sqrt {P{\text{ $ \times $ }}G} }\), which is increasing in total GDP, P×G.