The IMF's Statistical Systems in Context of Revision of the United Nations' A System of National Accounts
Chapter

10 Measurement of a Nation's Terms of Trade Effect and Real National Disposable Income Within a National Accounting Framework

Author(s):
Vicente Galbis
Published Date:
September 1991
Share
  • ShareShare
Show Summary Details
Author(s)
Mick Silver and Khashayar Mahdavy 

The United Nations' 1968 A System of National Accounts (SNA) and subsequent guidelines and manual1 include no recommendations for the measurement of both the terms of trade effect and real national income. However, many countries have in practice introduced measures of the terms of trade effect into their accounts.2 The current revision of the SNA is to incorporate changes in the terms of trade and real national disposable income (NDI) into the national accounts. This paper first outlines the rationale for their present exclusion and, second, proposes a framework for their inclusion in the accounts. Third, alternative formulae for measuring the terms of trade effect are surveyed. An interpretative framework is proposed with a case also being given for the use of a Tornqvist (translog) formula if a single trade-based index is desired. Empirical results are finally provided that support the use of the proposed interpretative framework and clearly show how present practice, to a large extent based on the Nicholson formula, can give quite distorted results.

I. System of National Accounts: Terms of Trade Effects and Real National (Disposable) Income

The SNA defines gross domestic product (GDP) in three ways:

  • In terms of domestic expenditure

  • As domestic production

  • As the income accruing to factors from that production.

GDP at purchasers' values or producers' values (SNA, page 233) may be defined as the producers' value of the gross outputs of resident producers, including the distributive trades and transport, less the purchasers' values of their intermediate consumption (that is, the producers' values of the value added of the resident producers) plus import duties. It is also equal to the total of the gross expenditure on the final use of the domestic supply of goods and services valued at purchasers' value less imports of goods and services valued to include insurance and freight, or the sum of the compensation to employees, consumption of fixed capital operating surplus and indirect taxes, net of resident producers and import duties.

Measured at current prices, all three estimates should be equal. However, in economic analysis it is often of interest to separate price changes from quantity changes to establish changes in the volume of production (at constant prices) or the purchasing power of income (real income). Changes in domestic production at constant prices may differ from changes in the real income arising from that production because of changes in the (net barter) terms of trade of a country. A country may produce, for example, the same physical amount of output utilizing the same physical amount of inputs. Yet an increase in the relative price of exports to imports will lead to an increase in the physical volume of goods and services that can be purchased from the factor income arising from domestic production, while leaving the volume of domestic production unchanged.

National income may differ at current period prices from domestic production at current prices in that the country will benefit (or lose) from flows of net receipts from abroad of incomes from employment, entrepreneurship, and property as well as net current transfers from abroad. The SNA defines national income (at market prices) to include employee compensation and the net income from property and entrepreneurship, irrespective of whether it is from abroad or arises from domestic production (SNA, page 120, paragraph 7.4). Thus, national income is equivalent to domestic product (that is, GDP) plus net factor income from abroad, less capital consumption. Gross national product (GNP) (at market prices) is commonly used in national accounting and is equal to national income plus capital consumption. The SNA defines net national disposable income (NNDI) as national income plus net redistributive (current) transfers from abroad. Gross national disposable income (GNDI) is equivalent to NNDI but before allowing for capital consumption. Thus, the real income accruing to an economy differs from domestic production at constant prices not only because of the inclusion of the terms of trade effect but also because of the inclusion of real net current transfers and factor income from abroad.

A consolidated system of accounts at both current and constant prices is given in Table 1. The terms of trade effect t is introduced in the income and outlay account to differentiate domestic product at constant prices from real domestic income, the latter including t. The term t is also introduced into the external account to represent the real income accruing from trade as opposed to the volume of goods and services. This precludes the appearance of a deficit at constant prices and a surplus at current prices due to terms of trade changes.

Table 1.Consolidated Accounts at Current and Constant Prices
AccountCurrent PricesConstant Prices/Real Terms
ProductionY = P = C + K + X − Mp = c + k + x − m
Income and outlayY + F + A = C + Sy + f + a = c + s
y = p + t
AccumulationS = K + Ns = k + n
ExternalF + A + X − M=Nt + f + a + x − m = n
Note: Y is domestic income; P, domestic product; C, consumption; K, investment; S, savings; X, exports; M, imports; N, net lending abroad; A, net current transfers abroad; F, net factor income from abroad; and t, the terms of trade effect. The lowercase letters denote the flows at constant prices or in real terms; that is, divided by an appropriate price deflator.
Note: Y is domestic income; P, domestic product; C, consumption; K, investment; S, savings; X, exports; M, imports; N, net lending abroad; A, net current transfers abroad; F, net factor income from abroad; and t, the terms of trade effect. The lowercase letters denote the flows at constant prices or in real terms; that is, divided by an appropriate price deflator.

Consider the deflation of the value of N at current prices in the external account. Each element of N is deflated by a price index of goods and services. For F and A the price indexes py and pa are the goods and services on which F and A will be utilized (if positive) or forgone (if negative). For X and M it will be three price indexes of exported and imported goods and services, px and pm. However, this will yield measures of the physical volume of exports and imports and not the surplus (deficit) deflated by a price index of goods and services for which it will be utilized (forgone), denoted by p. The measure n is concerned with flows of purchasing power, yet the terms x and m in the production and external accounts are concerned with flows in the volume of goods and services. Thus, the external account defined purely in terms of purchasing power is given by

and in terms of purchasing power (f and a) and volumes of exports and imports (x and m) by
An alternative approach has been advocated by Stuvel3 and Bjerke4 based on an additive, as opposed to a multiplicative, decomposition of value changes into their volume, price level, and terms of trade effects; that is,
the first term on the right-hand side of equation (3) being the terms of trade effect, the second term the price level effect, and the whole of the right-hand side the “gain from terms of trade” (see Bjerke, pp. 186–87). Ramussen5 advocated this measure of the gain from terms of trade, more recent support being given by Hamada and Iwata,6 who showed a similar formula to measure gains in welfare in terms of consumption dominance. However, this paper explores alternative measures of terms of trade effects under the approach given by equation (1), since, first, this is compatible with the multiplicative formulation of the SNA and, second, the welfare links derived by Hamada and Iwata, although an important contribution, rely on an assumption of X = M in the current period, which is too restrictive to justify its application in national accounting. The original assumption in the paper is that the accumulated value of discounted surpluses and deficits in the balance of payments is zero, though this is (implicitly) later relaxed when considering the gain or loss in the current period.

To convert equation (2) into equation (1) the terms of trade effect is introduced, t = nn′; that is,

The SNA recognized the need to measure real income but, while two solutions were mentioned as being in common use for the construction of real GNP, no recommendations were made (SNA, pp. 52–53).

The United Nations' 1977 Guidelines were intended to provide a comprehensive framework for price and quantity statistics within the SNA yet paid little attention to the measurement of national income in real terms. Real national income was recognized as being a more difficult concept than GDP at constant prices, since “… it does not in fact relate to any identifiable set of goods and services.” As such the guidelines did not include recommendations in this area. However, they continued to note that “To the extent that a need is felt for an estimate of national income and/or its components in constant prices, the only procedure that now appears defensible is the use of a prices, the only procedure that now appears defensible is the use of a single general deflator, such as for instance the implicit deflator of the gross domestic product, for all components.”7

The United Nations' 1979 Manual provided an elaboration of issues concerned with price and quantity statistics given by the United Nations (SNA and Guidelines). However, it noted that the choice of deflator for the measurement of real income was inevitably to some degree arbitrary and subjective. As such the term “accounts at constant prices” in the SNA was recommended to apply only to identifiable flows of goods and services which could be directly factored into price and quantity components. The gains or losses to national income from terms of trade movements, while recognized as being useful for certain types of economic analysis, were held to “… not be recorded in the accounts for purposes of international reporting.”8 It was argued that such estimates were for the users of statistics, not the compilers, since first the choice of deflator was held to depend on the kind of analysis undertaken and, second, the judgment involved in the choice of deflator was believed to introduce a level of subjectivity beyond that acceptable in a statistical publication. This has not stopped many countries from introducing measures of the terms of trade effect into their accounts. (The United Kingdom's national accounts include an adjustment for terms of trade effects, as do those for the United States. GNP adjusted for terms of trade in the latter case is referred to as a “command (over goods and services) series.” For details of U.K. and U.S. practices, respectively, along with an outline of alternative methods, see Hibbert and Denison.)9 Indeed a large variety of formulae have been proposed over the years for the measurement of terms of trade effects (these are discussed later), though none has provided a satisfactory solution to the problems discussed earlier.

One facet of the current review of the SNA is the inclusion of measures of terms of trade effects and real income in the accounts. This paper will next consider alternative frameworks for these measures, followed by the related issue of choice of deflators. Some proposals for formulae are then made along with some evidence that supports the proposals and shows the quite misleading results that would arise from present practice.

II. Accounting Frameworks

We consider two frameworks for incorporating measures of real national income into the national accounts. Both take as the starting point the GDP at constant prices, the measurement of which is well established. The first is given in Table 2 and identifies explicitly the terms of trade effect in real terms to arrive at the gross domestic income (GDI) in real terms. Net factor income from abroad in real terms and net current transfers in real terms are then added to arrive at the GNDI in real terms. In this presentation, the terms of trade effect in real terms is given by equation (1); that is, it distinguishes between the trade balance, with exports and imports deflated by their own respective price change, and the trade balance deflated by the price movement of the goods and services on which the income arising from the trade balance will be utilized or forgone.

Table 2.Proposed Accounting Framework for Presenting Measures of Real Income: Version 1
1GDP at constant prices
2Plus terms of trade effect in real terms
3Equals GDI in real terms
4Plus net factor income from abroad in real terms
5Equals GNI in real terms
6Plus net current transfers from abroad in real terms
7Equals GNDI in real terms
8Less consumption of fixed capital at constant prices
9Equals NNDI in real terms
Note: GDP is gross domestic product; GDI, gross domestic income; GNI, gross national income; GNDI, gross national disposable income; NNDI, net national disposable income.
Note: GDP is gross domestic product; GDI, gross domestic income; GNI, gross national income; GNDI, gross national disposable income; NNDI, net national disposable income.

An alternative accounting framework (given by Blades, but see also Stuvel)10 shows the derivation of NDI without explicit identification of the terms of trade effect in real terms. This is shown in Table 3. In this presentation, the terms of trade effect in real terms is incorporated within the framework as the sum of imports of goods and services at constant prices, less exports of goods and services at constant prices, plus net exports of goods and services in real terms. A single deflator, that reflects the prices of goods and services comprising domestic expenditure, is used to deflate net current receipts from abroad. Choice of deflators apart, the final estimate of NDI in Table 3 is the same as that given in Table 2. Items 2, 3, and 7(a) in Table 3 comprise the terms of trade effect; that is, item 2 of Table 2 and equation (1).

Table 3.Proposed Accounting Framework for Presenting Measures of Real Income: Version 2
1GDP at constant prices
2Less exports of goods and services at constant prices
3Plus imports of goods and services at constant prices
4Equals gross domestic expenditures at constant prices
5Less consumption of fixed capital at constant prices
6Equals net domestic expenditure at constant prices
7Plus net current receipts from abroad in real terms
(a) Net exports of goods and services
(b) Net factor income received
(c) Net current transfers received
8Equals NNDI in real terms

III. Review of Accounting Framework

The framework outlined in Table 3 provides for changes in real income to be measured with reference to a single set of goods and services (that is, those constituting net domestic expenditure), as opposed to providing a hybrid measure of production and income concepts as shown in Table 2. While, on the face of it, this is an advantage, this approach suffers from several weaknesses.

Table 2, unlike Table 3, is based on a clear analytical path that starts with changes in the volume of a country's domestic production and asks how these changes may differ from changes in its real income. The first reason lies with the effect of changes in its terms of trade. A country may produce the same but earn more because of favorable changes in its terms of trade; the extent of these benefits is estimated in real terms by the terms of trade effect. The second reason is due to changes in real net factor income from abroad; the third, real net current transfers from abroad; the final, capital consumption. A primary purpose of national accounts is to aid users in their analyses of economic flows, and this framework draws attention to, and quantifies, the major flows determining changes in a country's real income, changes in production being one of the determining factors. Table 2 may constitute a hybrid of concepts, but this is quite meaningful within the analytical framework.

•The approach given in Table 3 does not make explicit the terms of trade effect. Indeed Blades proposed in a discussion document that the framework in Table 3 be accompanied by separate estimates of the terms of trade effect using a trade-based deflator.11 However, in using a different deflator, different results would arise from that explicitly shown for the terms of trade effect and that implicit in the estimate of real NDI. The calculation of a terms of trade effect using a trade-based deflator was considered by Blades to be separate from, and inappropriate for, the measurement of real NDI, and regarded as a separate measurement problem. The basis for this position was the case given by the United Nations Statistical Office, which was held to argue convincingly that the index used for deflating the terms of trade effect must be trade based.12 Yet, the same paper notes that the choice between deflators can only be judged in terms of how appropriate they are for measuring “… how much more was earned (from production and other sources) in terms of quantities of goods and services for which this income is utilized.” There is no reason to confine the goods and services for which this income is utilized to traded ones. Real production differs from real income, in part, owing to the terms of trade effect. A case has yet to be presented for taking the terms of trade effect out of this framework. It is, in any event, implicit in any comprehensive measure of real NDI.

•The presentation of a single measure of NDI gives the distinct impression that the nation's real income is being estimated in some objective sense, whereas what is being estimated is real income contingent on a particular utilization of the income flows. The fact that a deflator is used consistently throughout the framework does not give any further credence to the use of that deflator.

IV. Alternative Deflators

Table 4 lists alternative deflators and their corresponding terms of trade effect. These deflators can be classified into two categories:

Table 4.Different Price Deflators and Their Terms of Trade Effects
Terms of Trade Effect
T = (X − M)/P −(X − m),
FormulaaPrice Deflator Pwhere x= X/Px and m = M/Pm
Trade-based deflators
1Nicholsonb/P = Pm
2Anti-NicholsonP = Px
3Burge-GearyP = Pm for M > X
P = Px for X > M
4Geary
5United Nations Statistical Office
6Courbis and Kurabayashic
Non-trade-based deflatorsd
7Stuvel, Norwaye
8ScottP= CPI
9Godley and Crippsf
10Second SNAg

See footnotes 13 and 14 in the text.

Also proposed by the Organization for European Economic Cooperation (Stuvel, Statistics of National Product) and the SNA as one of two possibilities.

Also advocated by Stuvel, National Accounts Analysis.

C, I, and G denote private consumption, investment, and government expenditure, respectively; the lowercase equivalents being these values at constant prices. X, M, Px, and Pm are as defined earlier.

At market prices, the proposal was for the measure to be net of capital consumption, with an appropriately deflated value of net investment included in the denominator.

At factor cost.

The first SNA is Nicholson's formula.

See footnotes 13 and 14 in the text.

Also proposed by the Organization for European Economic Cooperation (Stuvel, Statistics of National Product) and the SNA as one of two possibilities.

Also advocated by Stuvel, National Accounts Analysis.

C, I, and G denote private consumption, investment, and government expenditure, respectively; the lowercase equivalents being these values at constant prices. X, M, Px, and Pm are as defined earlier.

At market prices, the proposal was for the measure to be net of capital consumption, with an appropriately deflated value of net investment included in the denominator.

At factor cost.

The first SNA is Nicholson's formula.

  • Those based on the price of exports or imports or some average of the two13

  • Those based on the price of other economic variables in addition to or instead of export and import prices.14

Trade-Based Deflators

Trade-based deflators include price changes of imports (Table 4, formula 1), exports (formula 2), imports or exports depending on the sign of the trade balance (formula 3), and unweighted (formulae 4 and 5) or weighted (formula 6) averages of imports and exports. Nicholson's formula 1 shows the change in the quantity of imports that can be obtained for the same quantity of exports as a result of a change in the terms of trade. It values the gains (losses) from terms of trade changes as purchases of (forgone) imports. The anti-Nicholson formula 2 assumes in its valuation that net imports add to (if positive) or subtract from (if negative) a country's ability to provide future exports to pay for present imports. The Geary-Burge formula attempts to improve on Nicholson's by including the realized gain (where X > M), though since

… an accumulation of assets can be used either to increase future imports or reduce future exports, and an accumulation of liabilities can be liquidated by either reducing future imports or increasing exports, it is not clear why the deflator should depend upon the sign of net exports.15

Averages of Px and Pm have been proposed; the Geary formula is an unweighted arithmetic formula, and the United Nations' an unweighted harmonic mean. The terms of trade effects can be seen from Table 4 to be generally identified in terms of what has been called a “projection basis” multiplied by the difference in relative import and export prices; that is, 1/Pm −1/Px. The United Nations Statistical Office claims that its formula can be justified in an economic sense. In its example, if X = 240 and M = 200, it is claimed that the economic justification for using (X + M)/2 = M + (XM)/2 = 200 +40/2 = 220 as a projection basis is “one can assume that for 200 the whole external trade cycle was completed, while for 40 only half of the cycle (it was already exported but not yet imported; similarly … if it had already been imported but not yet exported).”16 Yet, this is not an economic justification since nowhere in economic theory are reserves considered to be halved simply because they have not been spent. It might be argued that the formula has an economic justification if export-import surpluses cancel in the long run. However, a terms of trade effect is evaluated for a current period and should be meaningful in these terms. It has also been claimed that the results can be clearly interpreted. That the projection basis is the arithmetic mean, or the deflator a harmonic mean, of export and import prices is a clear interpretation of the basis of the calculation. We know that the unit terms of trade effect is applied to an arithmetic mean of X and M, or “… the whole of what has been completed entirely plus we take half of what has been completed at 50 percent.”17 This is hardly a clear interpretation of its effect.

The United Nations cites as a justification for its formula that terms of trade should only have a redistributive effect. However, as Nicholson has argued:18

It is wrong to assume, as is sometimes done, that the adjustments to income (or the adjustments to product) in the two-country case should be equal and opposite. The desire for articulation should not lose touch with economics. The gain (or loss) from changes in the terms of trade in the product of the one country is necessarily equal to the loss (or gain) in the income of the other country [our emphasis].

Finally, Courbis and, later, Kurabayashi proposed a linear combination of export and import prices, weighted respectively by the relative shares of real exports and real imports in real total trade. If a weighted index is required, there is much to commend the Tornqvist, or trans-log, formula. This is given by

where the subscripts c and b respectively denote current and base periods.

The formula has much to commend it. First, on intuitive grounds, the formula improves on the Courbis and Kurabayashi formula in that it uses an average of current- and base-period weights, and not just base-period weights. Given that very few additional resources are necessary to compile this formula, compared with even Nicholson's simple formula, there is no justification for using a simple base-period-weighted index as proposed by Courbis and Kurabayashi. The Tornqvist formula may be chained to ensure that the weights used remain representative. Second, Diewert and Morrison have shown that it is suitable for ascertaining the effects on welfare (indicated by short-run output changes) of changes in a country's terms of trade in a manner that fits into an analytical framework for measuring technical change and total factor productivity.19 Third, since the Tornqvist index utilizes an average of base- and current-period weights it will fall within the bounds defined by Konus20 and Samuelson and Swamy21 for a “true” constant utility (or production) index defined in economic theory. Finally, Diewert has related the formula used for an index number to the functional form of the underlying aggregator function.22 For example, in price index number work there is an underlying utility function by which a theoretical price index can be defined with regard to maintaining a constant utility or cost of living. The Tornqvist index is exact for a translog functional form and, more importantly, is also defined as being superlative since it is exact for a flexible functional form. A flexible functional form is one that is capable of providing a second-order differential approximation to an arbitrary twice continuously differentiable linearly homogeneous aggregator (in this case utility) function. That the index provides a good approximation to a class of other functions is thus an advantage. However, it is one shared by Fisher's “ideal” index, the geometric mean of Laspeyres and Paasche, which also lies within the Konus bounds. In practice, Tornqvist and Fisher's indexes provide similar results.23

In summary, we note the following for trade-based deflators.

  • The Nicholson and anti-Nicholson formulae both provide results that can be quite clearly interpreted, yet the interpretation does not provide an indicator of the objective change in real income since there is no reason to believe, for example, in the case of the Nicholson formula that a surplus will be used to buy imports at the time that it arises or in the future. It is difficult to justify one over the other since both formulae are simply different ways of viewing the same phenomena. Indeed, the Geary-Burge formula is based on the idea that Pm provides a more illuminating view when there is a deficit and Px when there is a surplus. However, the lack of consistency in the interpretation of the result, along with other difficulties, precludes the use of this formula.

  • The Nicholson and anti-Nicholson formulae for the terms of trade effect will diverge as X and M diverge. It will be shown in the empirical work that the differences between the results are quite substantial for developing countries. That two formulae give substantially different results, while both being equally justified, is not unknown in the SNA (for example, Laspeyres versus Paasche). However, it is not a state of affairs that we would wish to encourage.

  • An average of the two formulae is neither conceptually sound nor justifiable in economic theory. It is purely a compromise. If a weighted average is to be adopted (and the justification for this is not clear) a Tornqvist (or translog) formulation is preferable to the Courbis and Kurabayashi formula.

Non-Trade-Based Deflators

The non-trade-based deflators (Stuvel, Scott, Godley and Cripps, and the second SNA) given in Table 4 use one or a combination of the implicit deflators of the domestic economy for deflation of the surplus or deficit. Stuvel's formula is based on the implied deflator for net domestic product at market prices, as an expression of changes in the purchasing power of money. Scott argues that the ultimate purpose of both foreign and domestic investment and trade is to meet present and future consumption. The consumer price index (CPI) is therefore advocated as the appropriate deflator. Godley and Cripps chose the implied price index for total domestic expenditure at factor cost as a deflator. The SNA, while never mentioning the effects of terms of trade, discusses two approaches. The first yields the Nicholson formula; the second

is to deflate the gross national product by a price index number of current and capital final expenditure in the domestic market. The two methods will give different results only if the balance of payments is different from zero and then only if the price index number for imports differs from the price index number of final purchases in the domestic market (SNA, paragraph 4.8, p. 53).

Choice Between Deflators

Before drawing some conclusions on alternative approaches it is worth mentioning a criterion for choice between deflators that is often cited.24 This is that the terms of trade effects should be symmetrical when viewed from the domestic earner or the rest of the world; that is, the gains in one sector should be equal to the losses in the complementary sector. The Geary-Burge, Geary, Courbis-Kurabayashi, and Tornqvist formulae appear to possess this property.25 The Nicholson and anti-Nicholson formulae do not produce symmetrical terms of trade effects; however, the Nicholson formula applied to one country and the anti-Nicholson formula to the trading partner will be symmetrical.

Too much attention should not be paid to the criterion of symmetry. As Nicholson has pointed out, symmetry is not a desirable property. Nicholson's formula gives the gain (loss) in income in terms of its ability to purchase (forgo) imports:26

… The other possible adjustment to terms of trade, symmetrical to the one just described, would consist in replacing imports by the value of exports at base year prices, needed to finance the actual level of imports. The result would indicate the level of domestic product needed to meet the current level of national expenditure [that is, the anti-Nicholson formula]. . . . It makes sense as economics. For the same goods and services which form part of the domestic and national product of one country form the imports and hence are part of the real income of the other country. These are simply two sides of the same coin.

Thus, symmetry would not be a desirable property for the Nicholson formula since it would not make economic sense. Indeed the symmetry properties of the other formulae cited do not justify their use. A substantial positive terms of trade effect would mean that a country no longer has to export so many goods to maintain import levels. Thus, the previously exported goods can now be used for domestic consumption. The terms of trade effect is thus used to purchase home-produced goods. The price movement of the bundle of goods produced by those resources will dictate the extent of the terms of trade effect. There is now no reason to expect symmetry. If the resources are used for domestic production of chocolate as opposed to biscuits, that should not be mirrored in any calculation for the country's trading partner.

A similar argument applies to a factory (or industrial sector). If the price increases of what it sells are greater than what it buys in, there is a terms of trade gain. What it spends this gain on will dictate the extent of the gain. If spent on machinery, whose price increases at a faster rate than the price of the output of the factory, the gain becomes a loss. However, machinery may be regarded as a third sector, and terms of trade gains may be aggregated across sectors. As should be apparent, intersectoral terms of trade can be calculated as long as sufficiently detailed information on input-output flows (volume and price) between sectors (including the rest of the world) are available. Many of the issues discussed in this paper apply to inter-sectoral terms of trade effects, though see Bjerke for more details.27

The choice between these deflators can only be judged in terms of how appropriate they are for measuring how much more was earned (from production and other sources) in terms of quantities of goods and services for which this income is utilized. A broader analytical framework is required for this. The effect on real income of a change in the terms of trade ultimately depends on what the surplus is spent on, or the nature of the response to the deficit, be it drawing on financial assets, cutting expenditure on inventories, and/or fixed capital and/or labor, exporting more or curtailing consumption of imported goods. In the case of imported goods Nicholson's formula would be the appropriate deflator. Three approaches are considered.

Approach A is to use a single deflator and to interpret it as implying one possible way of meeting a deficit or utilizing a surplus.

Approach B is to use a single deflator or set of deflators, but also to use several further deflators—for example, the CPI, implied deflator for current and/or fixed capital formation, import-export unit values. Thus, several measures are provided on the possible terms of trade effects arising, for example, from the different ways in which a trade surplus is utilized, or what is forgone when there is a deficit. If all give similar results, the single-measure formula provides a reasonable approximation to the terms of trade adjustment. If, for example, one indicator is widely divergent from the rest, the question must be asked whether this is a likely target of expenditure from the trade surplus and, if so, the bias borne in mind. Users would be warned by the very publication of several results that a reliable estimate of the objective economic change in real income may not be given in the table. The different estimates would show the real income changes under different spending scenarios. Little additional resources are required to deflate by the alternative deflators available.

Approach C is to calculate a deflator as a weighted average of the price movements of consumption, imports, investment, expenditure, and so on. Problems inherent in calculating such a deflator include, first, that the weights should reflect the additions to consumption expenditure, investment expenditure, imports, and so on that will arise from the surplus (or be forgone due to the deficit), as opposed to the existing pattern. Second, there may be a time lag between the occurrence of the trade gap and its effect on the economy.

Approach A, as argued earlier, may be misleading; approach C has severe measurement problems. Approach B is therefore proposed. There are some advantages in proposing a single deflator that should always be used, the effect from alternative deflators also being included or being relegated to footnotes. The Nicholson formula has some clear advantages in that, first, it has a plausible explanation (reserves or surplus may be used to finance more imports or imports may be curtailed when there is a deficit), second, it is already in use and, third, it is appropriate for the terms of trade effect because it is grounded in trade relations. However, the anti-Nicholson formula should be used along with the Nicholson formula. It may strike some readers that Nicholson's formula was rejected earlier but is being proposed here. As a trade-based deflator, the Nicholson and anti-Nicholson formulae can both yield very different results, yet neither can be considered theoretically superior. The use of Nicholson's formula in this context is highly problematic. However, the use of his formula within an interpretative framework, where it is clearly given as one of many options, is conceptually sound.

In summary, it is concluded that an appropriate framework should take the following points into account.

  • It should be based on the use of two identifiable standard selections of goods and services; the Nicholson and anti-Nicholson formulae would be suitable for this, though others might be considered.

  • Users should be clearly warned that these formulae do not provide indicators of objective economic flows, but ones contingent on particular scenarios. As such, estimates of alternative spending patterns by use of a range of further deflators are proposed as footnotes to the table of results.

  • The differences between using different deflators are shown by the empirical work to be substantial in many cases and, therefore, can mislead the economist if a single measure is published.

  • If a single trade-based deflator is to be used the Tornqvist index has many advantages.

Extension to Net Factor Income and Current Transfers from Abroad

It was noted in the second subsection above (“Non-Trade-Based Deflators”) that a measure of real income not only requires the adjustment for terms of trade effects but also adjustment first to include net factor income and current transfers from abroad and, second, to express these aggregates in real terms. Again, the stance taken is to ask what the net factor incomes and current transfers will be spent on or, if negative, what sacrifice will be made to meet the outflow. If net factor incomes are a surplus of migrants' earnings being sent home, the deflator for consumption expenditure is applicable. If the factor income is composed of a repatriated surplus from investment abroad, and this is to be used to finance fixed capital formation, the deflator for fixed capital formation is applicable. If the transfers are to finance particular activities, appropriate deflators may be found. If these sums are in deficit, the task is more difficult. We have to ask what has been forgone to meet the deficit and what was the price increase of these items. It is proposed that the Nicholson and anti-Nicholson approaches be extended to these items so that a systematic interpretation can be put on the GNDI. However, alternative deflators should also be applied to net factor incomes and current transfers to identify the effects from alternative scenarios. Thus, if, for example, there is a very large change in net factor income, which can be identified by the user as being from migrants' wages, then the user may use the results based on a consumer expenditure deflator. In conclusion, we note the following.

•Net factor income and current transfers may amount to quite substantial flows in proportion to the GDP; as such, guidelines are necessary on their deflation.

•It is proposed that the interpretive framework used for measuring the terms of trade effect be extended to these items; the Nicholson and anti-Nicholson formulae would be the main deflators with the effects of using alternative deflators also shown.

Empirical Work

It has been argued that the choice of deflator may be of little practical importance since different deflators give similar results and the terms of trade effect is generally small. This has been based on the results of empirical work undertaken by Gutmann and the United Nations Statistical Office.28 Gutmann only considered seven industrialized countries and Saudi Arabia. Little difference was found for the industrialized countries, yet for Saudi Arabia one formula led to a gain from terms of trade larger than Saudi Arabia's GDP, while a different formula showed a gain of only half that size. The United Nations' study had quite an extensive coverage of countries, yet only Nicholson's and the United Nations Statistical Office's formulae were considered.

Table 5 provides results on the effects of using different deflators for measuring terms of trade effects. The study covered 50 countries (20 industrial and 30 nonindustrial). The purpose of this study was to illustrate that terms of trade effects calculated by different formulae may, for some years, yield substantially different results. The period chosen varied between countries, being constrained by the availability of data for each country. At best, the data set included annual figures from 1948 to 1985. The terms of trade effects were calculated with respect to a reference period of 1980 equivalent to 0.00. It is noted that the levels as well as signs of the effect would differ if another reference period were chosen. In addition, it is stressed that 1980 weights are not being applied throughout the series; in many cases, the base period may be changed every five to ten years. Some of the differences may be substantial because of the time span used for the analysis. Yet in most cases, the maximum difference fell quite close to 1980, but the differences remained marked. Space constraints precluded the presentation of results for individual years. Because the study was concerned with differences in individual years, growth rates were not used as summary measures for comparisons. The following summary measures were used.

Table 5.Comparison of Terms of Trade Effects Using Various Deflators: Mean Absolute Difference, Standard Deviation of Absolute Differences, and Maximum Difference Between Nicholson's Formula and Others
Anti-NicholsonCPI DeflatorGDP DeflatorGeary
MeanMeanMeanMean
(standardMaximum(standardMaximum(standardMaximum(standardMaximum
Countrydeviation)(year)deviation)(year)deviation)(year)deviation)(year)
Non-industrial countries
Brazil,30.9123.545.7161.432.0134.616.869.0
1961-84(32.1)(1974)(47.0)(1974)(33.1)(1984)(17.5)(1974)
(thousand million
Brazilian
cruzeiros)
Burkina Faso,3.28.26.117.88.020.01.53.9
1954-83(2.5)(1983)(5.7)(1976)(6.9)(1983)(1.2)(1983)
(thousand million CFA
francs)a
Colombia,5.820.11.78.71.57.42.58.8
1950-85(5.3)(1966)(2.0)(1976)(1.8)(1976)(2.3)(1977)
(thousand million
Colombian pesos)
Cyprus,4.416.03.810.75.322.22.38.8
1950-84 (million(4.0)(1973)(2.6)(1973)(6.4)(1984)(2.2)(1973)
Cyprus pounds)
El Salvador,41.4287.221.7155.725.9109.619.2112.5
1952-83 (million(62.7)(1983)(35.1)(1974)(27.2)(1974)(27.2)(1983)
Salvadoran
colones)
Ethiopia,25.672.044.2114.532.767.612.836.8
1961-80 (million(24.4)(1975)(34.4)(1975)(17.0)(1978)(12.7)(1978)
Ethiopian birr)
Fiji,9.838.04.314.74.415.44.216.7
1960-83 (million(10.2)(1981)(4.1)(1972)(4.0)(1981)(4.4)(1981)
Fiji dollars)
Greece,12.339.210.931.412.545.56.721.8
1951-85(9.5)(1973)(9.3)(1973)(11.7)(1973)(5.4)(1973)
(thousand million
Greek drachmas)
India,3.910.35.814.65.214.12.36.1
1960-80(3.2)(1966)(4.8)(1965)(4.7)(1966)(1.9)(1966)
(thousand million
Indian rupees)
Jordan,30.467.733.582.133.582.016.139.1
1969-84 (million(19.6)(1969)(26.8)(1975)(26.8)(1975)(11.4)(1969)
Jordanian dinars)
Kenya,302.71,080.1366.31,752.0504.72,213.2161.2618.6
1965-85 (million(304.3)(1971)(413.6)(1971)(505.2)(1971)(167.9)(1971)
Kenyan shillings)
Korea,178.1507.2208.9670.9167.6711.198.9289.2
1961-84(173.2)(1971)(204.3)(1971)(205.4)(1974)(98.7)(1971)
(thousand million
Korean won)
Liberia,7.432.112.346.913.452.93.717.3
1971-84 (million(8.8)(1971)(16.9)(1971)(16.2)(1971)(4.6)(1971)
Liberian dollars)
Malawi,30.365.516.654.120.864.018.439.2
1967-85 (million(21.4)(1978)(18.3)(1972)(20.4)(1972)(13.4)(1968)
Malawi kwacha)
Source: Compiled from the International Monetary Fund's database, similar databeing available from various issues of its International Financial Statistics.

Franc de la Communauté financière africaine.

United NationsCourbis and
Statistical OfficeKurabayashiTornqvistNicholson
MeanMeanMean
(standardMaximum(standardMaximum(standardMaximumValue at
deviation)(year)deviation)(year)deviation)(year)Meanspecified date
15.561.814.451.114.753.3145.61974167.6
(16.1)(1974)(14.6)(1984)(14.8)(1974)1984−194.6
1.64.10.92.20.92.12.11982−5.2
(1.3)(1983)(0.6)(1982)(0.6)(1982)19760.7
1983−3.0
2.910.12.89.02.68.640.0196682.2
(2.7)(1966)(2.4)(1966)(2.4)(1977)197610.5
197767.7
2.28.01.96.91.97.217.2197347.4
(2.0)(1973)(1.7)(1973)(1.8)(1973)198425.9
20.7143.618.9128.619.3121.9349.81983−1,229.8
(31.3)(1983)(27.7)(1983)(27.7)(1983)1979−310.8
12.836.011.331.111.030.0112.11975229.3
(12.2)(1975)(10.3)(1975)(10.5)(1977)1977254.5
1978146.4
4.919.04.516.34.316.172.51981113.5
(5.1)(1981)(4.4)(1981)(4.3)(1981)
6.119.64.414.15.216.817.3197350.8
(4.7)(1973)(3.3)(1973)(4.0)(1973)
2.05.21.74.01.84.516.3196617.4
(1.6)(1966)(1.2)(1966)(1.4)(1966)196514.1
15.233.88.118.49.522.219.0196918.1
(9.8)(1969)(5.2)(1974)(6.4)(1974)197443.8
197525.1
151.3540.0140.1481.3138.6506.53,139.119714,422.3
(152.2)(1971)(135.4)(1971)(139.2)(1971)
89.1253.669.1199.478.6223.8606.21971750.7
(86.6)(1971)(66.1)(1979)(76.6)(1971)19791,705.3
197497.6
3.716.04.018.43.917.954.61971125.5
(4.4)(1971)(5.0)(1971)(4.8)(1971)
15.232.712.123.413.527.368.0197877.1
(10.7)(1978)(8.0)(1972)(9.5)(1978)1972106.3
196868.8
Source: Compiled from the International Monetary Fund's database, similar databeing available from various issues of its International Financial Statistics.

Franc de la Communauté financière africaine.

Anti-NicholsonCPI DeflatorGDP DeflatorGeary
MeanMeanMeanMean
(standardMaximum(standardMaximum(standardMaximum(standardMaximum
Countrydeviation)(year)deviation)(year)deviation)(year)deviation)(year)
Malaysia,300.21,220.7330.01,005.6173.0536.9137.0541.3
1955-85 (million(339.4)(1982)(239.1)(1969)(174.0)(1979)(150.5)(1982)
Malaysian ringgit)
Malta,6.722.810.734.114.144.93.713.1
1965-84 (million(6.5)(1970)(11.5)(1970)(15.2)(1970)(3.8)(1970)
Maltese liri)
Morocco,0.32.00.43.30.43.10.10.9
1957-77(0.4)(1976)(0.7)(1976)(0.7)(1976)(2.7)(1976)
(thousand million
Moroccan
dirhams)
Pakistan,1.63.81.95.62.05.70.82.0
1970-84(1.4)(1982)(1.8)(1970)(1.8)(1970)(0.7)(1979)
(thousand million
Pakistani rupees)
Philippines,1.55.51.36.61.36.70.93.9
1948-85(1.3)(1949)(1.6)(1949)(1.5)(1949)(0.8)(1949)
(thousand million
Philippine pesos)
Seychelles,6.118.87.032.510.037.03.19.0
1976-82 (million(6.8)(1982)(11.6)(1982)(14.0)(1982)(3.4)(1982)
Seychelles
rupees)
South Africa,421.11,067.7409.91,278.7218.71,007.3249.5649.3
1948-85 (million(302.6)(1962)(356.1)(1971)(259.4)(1985)(182.1)(1962)
South African
rand)
Sri Lanka,1,312.73,635.52,326.06,953.72,233.26,474.5812.62,296.0
1960-84 (million(849.6)(1969)(1,581.8)(1969)(1,454.4)(1969)(564.2)(1969)
Sri Lanka rupees)
Tanzania,174.9626.7265.81,269.3252.51,142.890.3328.7
1965-80 (million(191.7)(1978)(339.3)(1974)(299.7)(1971)(99.3)(1978)
Tanzanian
shillings)
Thailand,2.39.02.310.62.310.51.25.5
1950-85(2.4)(1969)(2.8)(1969)(2.8)(1969)(1.4)(1969)
(thousand million Thai
baht)
Togo,6.818.91.74.04.111.04.012.0
1971-78 (million(6.8)(1974)(1.3)(1977)(3.2)(1974)(4.4)(1974)
CFA francs)a
Trinidad and285.5790.2235.0676.0125.1343.3120.8318.9
Tobago,
1966-82 (million(245.4)(1972)(194.9)(1972)(100.5)(1972)(101.8)(1974)
Trinidad and
Tobago dollars)
Tunisia,97.2244.011.532.39.025.031.168.6
1961-83 (million(71.2)(1965)(9.1)(1961)(7.1)(1976)(21.0)(1965)
Tunisian dinars)
Source: Compiled from the International Monetary Fund's database, similar databeing available from various issues of its International Financial Statistics.

Franc de la Communauté financière africaine.

United NationsCourbis and
Statistical OfficeKurabayashiTornqvistNicholson
MeanMeanMean
(standardMaximum(standardMaximum(standardMaximumValue at
deviation)(year)deviation)(year)deviation)(year)Meanspecified date
150.1610.4156.0562.2148.0557.82,933.31982−7,127.9
(169.7)(1982)(171.8)(1982)(162.6)(1982)1979731.5
1969−940.8
3.411.42.88.83.210.73.9197037.8
(3.3)(1970)(2.1)(1970)(3.1)(1970)
0.11.00.10.70.10.70.31976−2.1
(0.2)(1976)(0.1)(1976)(0.1)(1976)
0.81.90.61.30.61.42.61982−3.5
(0.7)(1982)(0.5)(1975)(0.5)(1979)19793.9
19703.4
1975−4.8
0.82.70.72.20.82.912.0194911.5
(0.7)(1949)(0.6)(1949)(0.7)(1949)
3.09.42.77.72.88.070.71982−41.9
(3.4)(1982)(2.9)(1982)(3.0)(1982)
210.5533.9223.2631.1251.0676.13,245.519623,966.2
(151.3)(1962)(167.8)(1962)(184.1)(1962)19714,518.0
1985−2,498.1
656.31,817.8598.91,554.9622.71,687.811,101.6196910,754.6
(424.8)(1969)(376.6)(1969)(417.2)(1969)
87.4313.370.0206.265.3213.4699.81978603.9
(95.8)(1978)(69.9)(1978)(68.0)(1978)1974438.4
1971955.9
1.14.51.04.01.14.518.2196936.0
(1.2)(1969)(1.1)(1969)(1.2)(1969)
3.49.43.211.93.311.113.6197445.4
(3.4)(1974)(3.9)(1974)(3.8)(1974)197712.1
142.7395.1152.8467.1145.4415.11.71972−3,209.1
(122.7)(1972)(136.6)(1974)(126.1)(1974)1974−1,520.3
48.6122.240.790.534.375.9357.01961−250.5
(35.6)(1965)(27.5)1965)(23.0)(1965)1965−350.4
1976−431.1
Source: Compiled from the International Monetary Fund's database, similar databeing available from various issues of its International Financial Statistics.

Franc de la Communauté financière africaine.

Anti-NicholsonCPI DeflatorGDP DeflatorGeary
MeanMeanMeanMean
(standardMaximum(standardMaximum(standardMaximum(standardMaximum
Countrydeviation)(year)deviation)(year)deviation)(year)deviation)(year)
Turkey,79.1183.8114.4282.4107.5260.647.0107.7
1968-81(56.0)(1977)(82.0)(1977)(80.9)(1977)(33.2)(1977)
(thousand million
Turkish liras)
Zambia,389.31,673.7232.6951.6300.21,197.0298.61,337.6
1964-80 (million(452.4)(1969)(254.7)(1969)(359.5)(1969)(363.7)(1969)
Zambian kwacha)
Zimbabwe,10.343.721.564.321.573.45.725.9
1964-83 (million(12.2)(1965)(21.6)(1982)(20.8)(1982)(6.9)(1965)
Zimbabwean
dollars)
Industrial countries
Australia,0.52.30.42.10.32.00.31.5
1949-85 (million(0.6)(1952)(0.5)(1973)(0.4)(1973)(0.4)(1952)
Australian
dollars)
Austria,0.31.51.38.61.410.50.20.8
1948-85(0.4)(1977)(1.8)(1951)(2.2)(1951)(0.2)(1977)
(thousand million
Austrian
schillings)
Belgium-1.68.02.910.43.514.00.84.2
Luxembourg,(2.0)(1972)(2.5)(1972)(3.3)(1972)(1.0)(1972)
1953-84
(thousand
million francs)
Canada,0.21.00.42.40.32.00.10.5
1948-85(0.2)(1984)(0.6)(1984)(0.5)(1984)(0.1)(1984)
(thousand million
Canadian dollars)
Denmark,0.41.70.52.00.62.10.20.9
1948-85 (million(0.5)(1976)(0.4)(1950)(0.5)(1950)(0.2)(1976)
Danish kroner)
Finland,0.21.60.31.10.31.70.10.9
1951-84(0.3)(1975)(0.3)(1975)(0.4)(1975)(0.1)(1975)
(thousand million
Finnish markkaa)
France,0.62.31.44.81.55.20.31.2
1951-85(0.7)(1982)(1.4)(1960)(1.5)(1960)(0.4)(1972)
(thousand million
French francs)
Germany,2.68.63.211.13.811.01.34.7
1952-85(2.1)(1973)(2.6)(1973)(3.0)(1973)(1.1)(1973)
(thousand million
deutsche mark)
Iceland,0.20.90.62.30.83.00.10.5
1950-85 (million(0.2)(1973)(0.6)(1952)(0.8)(1950)(0.1)(1973)
Icelandic kronur)
Ireland,33.4152.441.3193.053.8303.817.686.5
1948-85 (million(37.1)(1973)(42.7)(1951)(57.9)(1951)(20.2)(1973)
Irish pounds)
Source: Compiled from the International Monetary Fund's database, similar databeing available from various issues of its International Financial Statistics.

Franc de la Communauté financière africaine.

United NationsCourbis and
Statistical OfficeKurabayashiTornqvistNicholson
MeanMeanMean
(standardMaximum(standardMaximum(standardMaximumValue at
deviation)(year)deviation)(year)deviation)(year)Meanspecified date
39.591.927.652.229.962.1111.21977117.9
(28.0)(1977)(16.3)(1974)(19.4)(1977)1979171.0
194.6836.9233.11,120.6265.11,224.81,446.819693,305.7
(226.2)(1969)(297.7)(1969)(330.0)(1969)
5.221.85.020.35.222.7125.21965298.7
(6.1)(1965)(5.7)(1965)(6.2)(1965)198288.2
0.21.20.21.20.31.20.8197312.3
(0.3)(1952)(0.3)(1973)(0.3)(1973)19523.6
0.10.80.10.70.10.88.1197718.5
(0.2)(1977)(0.1)(1977)(0.2)(1977)1951−0.6
0.84.00.84.20.84.158.41972109.6
(1.0)(1972)(1.0)(1972)(1.0)(1972)
0.10.50.10.50.44.02.81984−7.7
(0.1)(1984)(0.1)(1984)(0.9)(1984)
0.20.80.20.80.20.86.7197610.5
(0.2)(1976)(0.2)(1976)(0.2)(1976)19502.0
0.10.80.10.70.10.82.319756.5
(0.1)(1975)(0.1)(1975)(0.1)(1975)
0.31.10.31.20.31.114.819601.1
(0.4)(1982)(0.4)(1972)(0.4)(1972)197243.4
1982−17.6
1.34.31.44.614.638.924.0197365.0
(1.0)(1973)(1.1)(1973)(11.3)(1974)197432.5
0.10.50.10.50.10.55.61950−395.1
(0.1)(1973)(0.1)(1973)(0.1)(1973)1952−521.4
1973890.3
16.776.215.069.915.373.6177.61973895.3
(18.6)(1973)(16.7)(1973)(17.3)(1973)1951−4.2
Source: Compiled from the International Monetary Fund's database, similar databeing available from various issues of its International Financial Statistics.

Franc de la Communauté financière africaine.

Anti-NicholsonCPI DeflatorGDP DeflatorGeary
MeanMeanMeanMean
(standardMaximum(standardMaximum(standardMaximum(standardMaximum
Countrydeviation)(year)deviation)(year)deviation)(year)deviation)(year)
Italy,485.92,254.0558.43,703.0531.63,291.1269.61,250.1
1951-84(523.9)(1973)(762.4)(1973)(640.5)(1973)(294.1)(1963)
(thousand million
Italian lire)
Japan,540.53,013.4462.62,925.6485.02,678.2328.01,953.5
1952-85(732.7)(1972)(746.2)(1985)(771.8)(1978)(457.0)(1972)
(thousand million
Japanese yen)
Netherlands,0.20.90.62.30.83.00.10.5
1950-85(0.2)(1973)(0.6)(1952)(0.8)(1950)(0.1)(1973)
(thousand million
Netherlands
guilders)
New Zealand,73.7593.453.5537.454.0455.941.1325.6
1950-84(10.63)(1974)(95.7)(1974)(98.4)(1974)(59.0)(1974)
(million New
Zealand dollars)
Norway,1.36.31.06.30.95.60.62.8
1950-85(1.5)(1977)(1.5)(1984)(1.3)(1984)(0.8)(1977)
(thousand million
Norwegian
kroner)
Spain,48.2226.826.394.527.489.527.8138.7
1954-84(65.7)(1967)(26.0)(1967)(23.5)(1984)(39.9)(1967)
(thousand million
Spanish pesetas)
Sweden,0.21.10.42.60.42.30.10.5
1951-85(0.3)(1973)(0.5)(1973)(0.5)(1973)(0.1)(1973)
(thousand million
Swedish kronor)
Switzerland,0.10.50.30.90.41.40.10.3
1948-85(0.1)(1978)(0.2)(1948)(0.4)(1948)(0.0)(1978)
(thousand million
Swiss francs)
United Kingdom,0.22.00.22.10.32.50.10.9
1949-85 (million(0.4)(1974)(0.4)(1974)(0.5)(1974)(0.1)(1974)
pounds sterling)
United States,3.714.65.129.65.229.32.17.8
1948-85(3.4)(1985)(5.9)(1985)(5.9)(1985)(1.9)(1985)
(thousand million U.S.
dollars)
Source: Compiled from the International Monetary Fund's database, similar databeing available from various issues of its International Financial Statistics.

Franc de la Communauté financière africaine.

United NationsCourbis and
Statistical OfficeKurabayashiTornqvistNicholson
MeanMeanMean
(standardMaximum(standardMaximum(standardMaximumValue at
deviation)(year)deviation)(year)deviation)(year)Meanspecified date
242.91,127.0228.61,030.4237.01,079.56,186.7197312,015.8
(261.9)(1973)(235.2)(1973)(251.2)(1973)19636,983.3
270.31,506.7289.11,680.0467.51,630.95,150.8197212,653.3
(366.3)(1972)(407.4)(1972)(459.5)(1978)197812,982.3
19855,801.4
0.11.00.11.00.11.05.6197313.6
(0.1)(1973)(0.1)(1973)(0.1)(1973)19521.4
19502.2
36.8296.735.0230.037.7274.7738.619741,023.9
(53.1)(1974)(44.5)(1974)(51.1)(1974)
0.63.20.62.80.62.914.5197723.6
(0.8)(1977)(0.8)(1984)(0.8)(1977)198426.1
24.1113.418.477.222.2101.7210.51967214.0
(32.9)(1967)(21.3)(1968)(29.1)(1967)1968280.8
1984−337.8
0.10.50.10.60.10.54.6197310.4
(0.1)(1973)(0.1)(1973)(0.1)(1973)
0.10.30.10.30.00.32.319787.5
(0.0)(1978)(0.0)(1978)(0.0)(1978)1948−0.9
0.11.00.10.90.10.92.51974−10.9
(0.2)(1974)(0.1)(1974)(0.1)(1974)n
1.87.31.86.01.96.736.0194828.0
(1.7)(1985)n(1.7)(1948)n(1.8)(1985)198533.9
Source: Compiled from the International Monetary Fund's database, similar databeing available from various issues of its International Financial Statistics.

Franc de la Communauté financière africaine.

Source: Compiled from the International Monetary Fund's database, similar databeing available from various issues of its International Financial Statistics.

Franc de la Communauté financière africaine.

The absolute difference between the terms of trade effects as measured by Nicholson's formula and, for example, Courbis and Kurabayashi's formula was derived for each year and the mean, standard deviation and maximum (absolute) difference were calculated. This was repeated for each of the formulae given in Table 4. Since Nicholson's formula was used as a point of reference, its mean (absolute) value is given, as are its values for those years when the differences between Nicholson's formula and any other formula were at a maximum for the period. The results for Brazil can be used to illustrate the use of these measures. The Nicholson and anti-Nicholson measures differed most, for the period 1961–84, in 1974 by 123.5 million cruzeiros (at notional 1980 prices). Yet, as the final column shows in 1974, the terms of trade effect calculated by Nicholson's formula itself was only 167.6 million cruzeiros, the difference between the two formulae amounting to nearly three fourths of Nicholson's estimate. On average, the Nicholson formula gave an (absolute) value of 145.6 million cruzeiros. The average (absolute) difference between the Nicholson and anti-Nicholson formulae was 30.9 million cruzeiros, around one fifth of the average value for Nicholson's formula, but still substantial. The high standard deviation shows that in many years the results might be much higher than this average. The overall results show the following.

•The difference between using import and export unit value indices (Nicholson and anti-Nicholson formulae) as deflators can be substantial. The results show that for nonindustrial countries not only are the maximum differences substantial but also the means and variances of these differences are high relative to the mean (using Nicholson's formula) with high standard deviations. Thus, high differences are not isolated events in any one year. A mere glance at the values will show this to be the case for almost all nonindustrial countries. For many industrial countries, the differences are less marked on average; yet, for particular years, the maximum differences remain high. The results are likely to be similar for countries with small trade balances. It may be argued that large imbalances are unlikely to hold for a given country over a long period. However, the SNA cannot be useful if it gives a good estimate most of the time, especially if “most of the time” is when the phenomenon is having little impact on the economy.

•If a trade-based deflator is to be used, a weighted or an unweighted average must fall between the results from using an import and export price index. The difference between the Geary and the United Nations Statistical Office's formulae can be seen to be generally small. Both use unweighted averages (arithmetic and harmonic mean, respectively) of import and export unit value indices as deflators. The United Nations formula differs only slightly from the weighted Courbis and Kurabayashi formula, the use of a weighted, as opposed to unweighted, formula not being crucial, though differences can be quite high in certain circumstances (for example, for Greece in 1976 and Morocco in 1973). The differences between the weighted formulae, Tornqvist's and Courbis and Kurabayashi's, can again be seen to be generally small, but there are cases (for example, Canada in 1984) when it is quite significant. If a weighted trade-based formula is to be used, the question whether the base period or any average of base- and current-period weights needs to be addressed. Since the use of a chained formula would involve little extra resources, this demands consideration, though empirical work on this has not been carried out.

•The use of a range of deflators shows that the measure of terms of trade effect can be quite dependent on the selection of goods and services chosen. The deflators used represent a small selection of those available, and their choice was somewhat arbitrary. The CPI and GDP deflator were selected to be analyzed in conjunction with the Nicholson and anti-Nicholson deflators. It has already been shown that the Nicholson and anti-Nicholson deflators can diverge substantially, especially for nonindustrial countries. There are often quite marked differences between Nicholson's and the CPI and GDP deflators (for example, Trinidad and Tobago, Belgium, Luxembourg, and Malaysia). The differences between the Nicholson and the implicit GDP deflator and the CPI are not as pronounced as between the Nicholson and anti-Nicholson deflators owing to the interrelationship between import price changes, consumer price changes, and the implicit GDP deflator. Yet, what is apparent is that the four results in many cases can be quite different but become more meaningful when viewed together.

V. Conclusions

This paper has argued for an analytical framework for the measurement of terms of trade effects and NNDI that was based on separately identifying the GDP at constant prices and separately ascertaining the effects of terms of trade movements, net factor income, and net current transfers from abroad. Several possible deflators were surveyed, and empirical work has been provided to show that the measurement of terms of trade effects are very sensitive to the choice of deflator. Since there is no single deflator that is appropriate to all countries in all circumstances, an interpretive framework has been proposed. Under this framework the economist can ascertain changes in real income based on different spending scenarios. This may be unique in national accounting, but so is the problem of measuring real income flows. Countries that continue to use a single measure may, as shown by the empirical work, provide economists with seriously misleading results if the assumptions implicit in a different formula are more likely to apply at that time. The United Nations Expert Group on the SNA Review was sympathetic to the case presented for the adoption of the framework given in Table 2 but has, at the time of writing, yet to propose a recommended choice of deflator.

Note: This paper originally appeared in the Journal of the Royal Statistical Society, Series A (Statistics in Society), Volume 152 (Part 1, 1989), pp. 87–107, and is reprinted here, with only minor editorial modifications, by permission. It is based on a discussion document by the authors that was presented under the aegis of the IMF to the United Nations Expert Group on the SNA review (Luxembourg, 1986). The research was conducted while the authors were consultant and summer intern, respectively, at the Fund. The authors are grateful for comments received from the meetings and from the journal’s referees and to Chandrakant Patel and Simon Quinn of the Fund’s Bureau of Statistics and John Muelbauer of Oxford University for their help. The views expressed are those of the authors and should not be attributed to the Fund.

United Nations, Guidelines on Principles of a System of Price and Quantity Statistics, Statistical Papers, Series M, No. 59 (New York, 1977); United Nations, Manual on National Accounts at Constant Prices, Statistical Papers, Series M, No. 64 (New York, 1979).

See, for example, Hibbert and Denison for an account of U.K. and U.S. practice: J. Hibbert, “Measuring Changes in a Nation’s Real Income,” Economic Trends, No. 255 (1975), pp. 28–35; E.F. Denison, “International Transactions in Measures of the Nation’s Production,” Survey of Current Business, May 17–28, 1981.

G. Stuvel, “A New Approach to the Measurement of Terms of Trade Effects,” Review of Economics and Statistics (August 1956), pp. 294–307; Statistics of National Product and Expenditure, No. 2, 1938 and 1947–55. (Paris: Organization for European Economic Cooperation, 1957),” Asset Revaluation and Terms-of-Trade Effects in the Framework of the National Accounts,” Economic Journal Vol. 69 (1959), pp. 275–92; and National Accounts Analysis (London: Macmillan, 1986).

K. Bjerke, “Some Reflections on the Terms of Trade,” Review of Income and Wealth, Series 14 (No. 2, 1968), pp. 183–98.

P.N. Ramussen, Studies in Inter-Sectoral Relations (Amsterdam: North-Holland, 1956).

K. Hamada and K. Iwata, “National Income, Terms of Trade and Economic Welfare,” Economic Journal, Vol. 94(1984), pp. 752–71.

United Nations, Guidelines, p. 6, paragraph 28.

United Nations, Manual, p. 7, paragraph 1.6.

See note 2, above.

D.W. Blades, “Real National and Household Disposable Income: Discussion Document for the United Nations Expert Group on SNA Review Devoted to Price and Quantity Comparisons,” Report ESD/STAT/220(86)892 (Luxembourg: Organization for Economic Cooperation and Development, 1986); Stuvel, National Accounts Analysis, Chapter 6.

Ibid.

United Nations Statistical Office (UNSO), “The Treatment of Terms-of-Trade Effect in Measuring Economic Growth,” Report ESA/SAT/AD.2714 (New York, 1986).

J.L. Nicholson, “The Effects of International Trade on the Measurement of Real National Income,” Economic Journal, Vol. 70 (1960), pp. 608–12; anti-Nicholson; R.W. Burge and R.C. Geary, “Balancing of a System of National Accounts in Real Terms,” in Meeting of the International Association for Research on Income and Wealth, 1957; R.C. Geary, “Problems in the Deflation of National Accounts: Introduction,” Review of Income and Wealth, Series 9 (1961); UNSO, “Treatment of Terms-of-Trade Effect”; R. Courbis, “Comptabilité nationale à prix constants et à productivité constante,” Review of Income and Wealth, Series 15 (No. 1, 1969), pp. 33–76; and Y. Kurabayashi, “The Impact of Changes in Terms of Trade on a System of National Accounts: An Attempted Synthesis,” Review of Income and Wealth, Series 17 (No. 3, 1971), pp. 285–97.

Stuvel, “A New Approach”; M.F.G. Scott, “What Price the National Income?” in Economics and Human Welfare: Essays in Honor of Tibor Scitovsky, ed. by M.J. Boskin (New York: Academic, 1979); W. Godley and F. Cripps, “London and Cambridge Economic Bulletin II,” The Times (1974); and the present SNA.

Denison, “International Transactions,” p. 27, citing a comment by Walter Salant.

UNSO, “Treatment of Terms-of-Trade Effect”

Ibid.

Nicholson, “Effectsof International Trade.”

W.E. Diewert and CJ. Morrison, “Adjusting Output and Productivity Indexes for Changes in the Terms of Trade” Economic Journal, Vol. 96 (1986), pp. 659–79.

A.A. Konus, “The Problems of the True Index of the Cost of Living,” Econometrica, Vol. 7 (1939), pp. 10–29 (English translation).

P.A. Samuelson and S. Swamy, “Invariant Economic Index Numbers and Canonical Duality: Survey and Synthesis,” American Economic Review, Vol. 64 (1974), pp. 566–93.

W.E. Diewert, “Exact and Superlative Index Numbers,” journal of Econometrics, Vol. 4 (1976), pp. 115–45; “The Theory of the Cost-of-Living Index and the Measurement of Welfare Change,” in Price Level Measurement, ed. by W.G. Diewert and C. Montmarquette (Ottawa: Statistics Canada), pp. 163–233.

B. Hansen and E.F. Lucas, “On the Accuracy of Index Numbers,” Review of Income and Wealth, Series 30 (No. 1, 1984), pp. 25–38.

For example, Hibbert, “Measuring Changes” and P. Gutmann, “The Measurement of Terms of Trade Effects” Review of Income and Wealth, Series 27 (No. 4, 1981), pp. 443–53.

Gutmann, “Measurement of Terms of Trade Effects.”

Nicholson, “Effects of International Trade.”

Bjerke, “Some Reflections.”

Gutmann, “Measurement of Terms of Trade Effects”; UNSO, “Growth Indices Adjusted for Terms-of-Trade Effect for Seventy-Nine Countries,” Report ESA/STAT/AC.27/4, Annex 1 (New York, 1986).

    Other Resources Citing This Publication