Developments in external and internal balances - a selective and eclective review

BIS Economic Papers No 29
October 1990


"If Americans saved just two more pennies out of every dollar they earn, they would close the entire (balance of payments) gap and eliminate additional borrowing from foreigners." M. Darby, Newsweek, January 8, 1990, p.43.

In a nutshell, the above quotation captures the essence of this paper, which attempts to analyse balance of payments developments from an external as well as a domestic perspective, with private sector saving playing a major role within the latter. However, the quotation is also slightly misleading as the relationship referred to exists only in an ex post accounting sense or under very restrictive ex ante conditions. Indeed, as the paper will show, when various interactions are taken into account, the trade-off between saving and the external account is well below unity.

There are essentially four ways of analysing balance of payments developments and they may be illustrated using the accounting identity:

EX-IM S-I = Y-D = F

(i) the trade approach, which looks at EX and IM directly or in the form of net exports and takes as the primary determinants relative income and prices, including the real exchange rate;
(ii) the saving-investment approach, which analyses developments in domestic saving and investment (and in their components) and uses income, real interest rates, inflation, wealth and - in a few cases - the real exchange rate as the key determinants;
(iii) the absorption approach, which views EX-IM as the difference between output and total domestic demand and is implemented using a large-scale macroeconomic model. It can also take the form of analysing reduced-form aggregate supply and demand equations or sectoral developments with particular weight on the distribution between tradable and non-tradable goods;
(iv) the capital flow approach, which regards F as an exogenous variable or relates it to changes in the size and composition of international portfolios and then analyses how EX and IM adjust to a given change in F.

In this paper, the absorption and capital flow approaches will not be pursued but various versions of the trade and the saving-investment approaches will be tested and then evaluated within a broad framework which attempts to reconcile the empirical results and the transmission channels implied.

The paper is divided into three main parts, under the headings "External approach", "Domestic approach" and "Reconciliation and conclusions" and each part contains several sections and sub-sections. Section A of Part I analyses some simple time series properties of the current external account (measured in % of GNP), using a sample of sixteen countries over the period 1960-89. This analysis serves essentially two purposes: (i) to determine the order of integration of the variable which the two approaches attempt to explain and (ii) to identify trends and/or possible structural shifts which need to be further explored in structural equations.

Section B turns to export and import equations and reconsiders a result already presented some twenty years ago (see Houthakker and Magee (1969)); viz. that the relative size of export and import demand elasticities differs significantly between countries. The extent to which such differences still exist and have affected countries' real growth performance is further explored in Annex 1, whereas in sub-section (b) the trade equations and the corresponding empirical estimates are used in evaluating net exports with a view to explaining changes in the external accounts of the sixteen countries. The principal determinants in this sub-section, which constitutes the core of the external approach, are domestic and foreign demand growth and relative prices, measured by changes in real effective exchange rates and in the terms of trade.

Part II of the paper first reviews the main trends in domestic saving and investment along with various ways of linking current-account developments to changes in domestic saving and investment components and then focuses on two of them:

  • a model proposed by Roubini (1988) which relates the current account directly to the public sector deficit and total investment. In the extreme case where the underlying assumptions of the model hold, balance-of-payments developments can be explained entirely by the two determinants, and the extent to which this has actually been the case is tested and discussed in sub-section (b);
  • a second model (Turner (1986)) which looks at the behaviour of various saving and investment components and then derives balance of payments changes via the national accounting identity. The key determinants in this model are measures of the business cycle, income growth, inflation, interest rates, real effective exchange rates and the public sector borrowing requirement, with the latter serving both as a determinant of private saving and as a component of domestic financial balances. The implementation and evaluation of this model and the evidence can be found in sub-section (c).

Of the two 'domestic' models tested, the second proves to be the more satisfactory one and in Part III of the paper the key parameters and the associated changes in variables are compared with those obtained from the net export equations described above. The evaluation is based on contributions to balance of payments changes during 1980-89, with section (A) analysing developments in each of the sixteen countries separately. Section (B) then attempts to summarise the results and draw some general conclusions.

The empirical methodology applied throughout the paper is the two-stage approach recently proposed by Engle and Granger (1987). Essentially this involves estimating a long-run equation based on an underlying hypothesis, supplemented by an adjustment equation which determines the dynamic structure and the extent to which deviations from the long-run path are being corrected. To save space and facilitate the reading of the paper, the detailed estimation results are presented in Annex II, with only the key parameters in the text. For the same reason, the tests undertaken to determine the order of integration of the variables are not given, except for the balance of payments. With a few exceptions, however, all the variables used in estimating the long-run equations were integrated of order and thus satisfied the necessary condition for being co-integrated.