Can Debtor Countries Service Their Debts? Income and Price Elasticities for Exports of Developing Countries

International Finance Discussion Papers Number 277

April 1986

CAN DEBTOR COUNTRIES SERVICE THEIR DEBTS? INCOME AND PRICE ELASTICITIES FOR EXPORTS OF DEVELOPING. COUNTRIES

by

Jaime Mar quez Caryl McNeilly

NOTE: International Finance Discussion Papers are preliminary materials circulated to stimulate discussion and critical comment. References in publications to International Finance Discussion Papers (other than an acknowledgment by a writer that he has had access to unpublished material) should be cleared with the author or authors.

ABSTRACT

Interest in income and price elasticities for international trade has increased recently because of the debt crisis that many developing countries are experiencing. Estimates of income elasticities of import demand, however, range from a low of 1.3 to a high of 4.7. Such differerces have important implications for debtor and creditor countries alike. Using quarterly data for the period 1973-1981, this paper esitimates income and price elasticities for non-oil imports of five major industrial countries from non-OPEC developing countries. The

empirical results suggest that the income elasticity is closer to 1 than

to 4,

Can Debtor Countries Service Their Debts? Income and Price Elasticities for Exports of Developing Countries

Jaime Marquez International Finance Division Federal Reserve Board Washington, D.C. 20551

and

Caryl McNeilly Woodrow Wilson School of Public and International Affairs Princeton University Princeton, N.J. 08544

April 1986 (Revised October 1985)

Interest in income and price elasticities for international trade has increased recently because of the debt crisis that many developing countries are experiencing. Estimates of income elasticities of import demand, however, range from a low of 1.3 to a high of 4.7. Such differences have important implications for debtor and creditor countries, alike. Using quarterly data for the period 1973-1981, this paper estimates income and price elasticities for non-oil imports of five major industrial countries from non-OPEC developing countries. The empirical results suggest that the income elasticity is closer to 1 than to 4,

This paper is part of a larger modeling effort of trade and financial relations between major industrial countries and developing countries for the Multi-Country Model of the Federal Reserve Board. The work of McNeilly was completed while she was at the Board. We have benefited from the comments of F. Gerard Adams, William Cline, Russell Cooper, Vittorio Corbo, Jonathan Eaton, Neil Ericsson, Mike Gavin, Peter Hooper, Mohsin Khan, Cathy Mann, Mare Noland, Charles Siegman, and Ralph Tryon. Woodrow Hellamy provided helpful research assistance. This paper has been presented in seminars at the Federal Reserve Board and the World Bank. Tris paper represents the views of the authors and should not be interpreted as reflecting the views of the Board of Governors of the Federal Reserve System or other members of its staff.

Introduction This paper estimates income and price elasticities of non-oil imports of major industrial countries from developing countries. Besides their traditional role in analyses of international linkages and trade policies, knowledge of these elasticities is crucial to designing policy responses to the existing debt crisis.! From the debtors! standpoint, both stabilization policies and debt rescheduling agreements hinge on balance of payment projections that are crucially dependent on the choice of elasticity estimates. Moreover, whether these countries will, in general, be able to service their external debt depends on the response of their exports to growth in industrial countries, a response that is ultimately determined by income and price elasticities.. Knowledge of these elasticities is also relevant for policymaking in industrial countries. A restrictive monetary policy stance in the United States makes it more difficult for debtor countries to service their debt. Both the increase in interest rates and the induced contraction in U.S. imports reduce the debt-servicing capacity of debtor countries, a reduction that might feed back to the United States as declines in exports and disruptions in financial markets. An important component in determining the magnitude of these feedback effects is the income elasticity for industrial-country imports from

developing countries.

lsee the studies of Cline (1984, 1985), Dooley et. al. (1983), Adams et. al. (1983), Dornbusch (1985), Riedel (1984), Goldstein and Khan (1982), arid Bond (1985). For the debate on the policy implications of different elasticity estimates for the United States, see the testimonies of William Cline and Rudiger Dornbusch before the Joint Economic Committee, Mareh 28, 1984. See also the exchange between Cline and Dornbusch in the Brookings Papers No.2, 1985.

Despite the increased attention given to these elasticities, an examination of the more influential studies reveals a lack of consensus regarding their values. Dornbusch (1985) states thal:

",..the elasticity estimate of LDC export growth with respect

to OECD growth cannot be pinned down. ... At this stage there

is certainly no firm finding, here or in the literature,

regarding the elasticity." (p. 336-37)

Unsatisfactory as it may be, Dornbusch's characterization is quite accurate. Existing estimates of the income elasticity for non-oil exports of developing countries range from 1.3 (Goldstein and Khan 1982) to 4.7 (Dornbusch 1985). Differences this large in elasticity

estimates need to be addressed because of their importance for policy design in both debtor and creditor countries.

To place the analysis in perspective, section 2 reviews the more influential papers in this area. This review reveals that existing elasticity estimates are subject to biases arising from the omission of relevant variables, country aggregation, and simultaneity, all 2f which

“Limit the applicability of existing estimates to practical problems, Section 3 describes the import demand model used to estimate ineome and price elasticities of non-oil imports of Canada, Germany, Japan, the United Kingdom, and the United States from non-OPEC developing countries. Econometric estimation of these elasticities requires a fair amount of Sensitivity analysis coupled with hypothesis testing. This paper tests the maintained assumptions for the error term (normality, serial independence and homoskedasticity), the structural stability of

parameter estimates, the homogeneity of degree zero in prices, and the

dynamic specification.

A second contribution of this paper is the construction of quarterly time series for non-oil imports for the above five countries from ron-OPEC developing countries. To our knowledge, this kind of information is not publicly available and thus section 4 describes the methodology used to construct the data. Section 5 presents our empirical estimates for income and price elasticities for each of the five countries and uses them to examine the relation between OECD growth and interest rates consistent with sustainable debt servicing by developing countries. The main findings are that the OECD-aggregate income elasticity varies between 1.3 and 1.6, and that sustainable debt servicing requires a 1/2 percent increase in OECD growth for every 1 percent increase in interest rates. Section 6 summarizes the paper and

Suggests avenues for further research,

2. Review of Existing Studies Table 1 lists the more salient features of several studies of developing count:ies' exports. An examination of these features reveals important limitations that limit the applicability of their findings to practical questions. First, the own-price elasticity of import demand is commonly assumed to be zero. In addition to being unrealistic, this assumption might induce a bias in the estimate of the income elasticity and thus be respoisible for much of the controversy about its magnitude.

Second, importers are generally treated as if they were small econonies, a tenable assumption for countries with little participation in international trade, but an untenable one for most of the countries

considered here, either separately or when they are treated as a bloc.

Higher growth in industrial countries is certain to increase non-oil imports from developing countries with a corresponding impact on their price, Although the small country assumption might be relaxed by using two-stage least squares, many studies in this area rely on ordinary least squares for parameter estimation.2

Third, an examination of the nature of the data used in previous studies suggests that available elasticity estimates are not strictly comparable to each other. On the one hand, Cline (1984) focuses on multilateral non-oil imports of the bloc of industrial countries, which include non-oil imports of OECD countries from themselves.3 On the other hand, Dornbusch examines multilateral non-oil exports of developing countries, which include non-oil exports of developing countries to themselves. It is immediately clear that neither of these two measures of multilateral trade are suitable for estimating the income response of imports of industrial countries rrom developing countries. Furthermore, changes in the country composition

of multilateral trade flows might be responsible for the erratic

2Grossman's (1982) analysis does not suffer from the limitations mentioned above, but it only applies to 0.7 percent of U.S. imports. (The 1978 value of U.S. imports from developing countries in Grossnan's paper is $1.2 billion, whereas U. S. total imports in 1978 were $183 billion, $42 billion of which were imports from non-oil developing countries (IMF, Direction of Trade, Yearbook 1981, p. 380).) This study makes it possible to establish the extent to which Grossman's findings extend to a non-oil commodity aggregate and the degree to which they apply to other industrial countries.

3This point has already been observed by Dornbusch (1985). Note that according to IMF data (Supplement on Trade Statistics, 1982, p.80), only 16 percent of total imports of the OECD region come from non-oil developing countries. This share is even lower for non-oil imports. Similarly, according to IMF data only 57 percent of exports of developing countries flow to industrial countries.

behavior of the aggregate elasticity estimates noted by Dornbusch (1985).4

Finally, almost all of these studies rely on import relations between country blocs. From a policy standpoint, aggregation across countries allows us to study how exports of developing countries fluctuate in response to the business cycle of all industrial countries. But unless all importing countries are identical, the resulting elasticity estimates are not helpful for analyzing the effects on trade of developing countries of stabilization policies of a particular importing country. From an econometric standpoint, the chief drawback of country aggregation is the potential for aggregation biases.

To emphasize these biases, table 2 presents income elasticities estimated using data at different levels of aggregation: data for our five countries (Canada, Germany, Japan, the United Kingdom, the United States) considered as one bloc, data at the country level, and a pooled sample of our five countries. The elasticity estimates are based on a linear regression of the growth rate of non-oil imports on the GNP growth rate, which is the specification used by Cline (1984). Tre elasticity estimated with data aggregated across countries is 2.9, which is relatively close to the estimate of Cline (1984). However, the elasticity estimates using individual country data range from -1.2 for

Japan to 5.9 for the United States. To determine whether these

4uhe use of multilateral trade flows might bias the price elasticity. For example, Bergsten and Cline (1983) do not allow for price effects in their estimating equation. However, empirical tests using their published data reveal that prices exert a negative, but insignificant, influence on imports. This apparent absence of price effects might be the result of a negative price elasticity for imports from non-OECD countries being offset by a positive price elasticity for OECD exports to themselves, which Bergsten and Cline include in their import data.

differences in country elasticities are significant, they are estimated with a pooled sample of our five countries (fixed effects). With an F-test for the hypothesis of equal elasticity across countries, the evidence suggests that such differences are indeed significant. The results of table 2 also suggest that the behavior of U.S. imports might be the dominant force in explaining the behavior of aggregate imports and thus largely responsible for the differences in estimates due to country aggregation. This last possibility is empirically supported by F-tests rejecting the null hypothesis of equality between the U.5. income elasticity and any other country's income elasticity.

Several conclusions emerge from this review and from the results of table 2. First, growth in non-oil exports of developing countries is inadequately explained by growth of importers alone, Failure to account for price effects and dynamic adjustments could result in biased and erratic income elasticities. Second, estimation methods need to recognize that ordinary least squares do not capture general equilibrium effects, which might be responsible for a simultaneit}y bias. Finally, reliance on multilateral trade flows aggregated across 2ountries might impart a certain instability to parameter estimates. Chanzes in the country composition of an aggregate trade flow could lead to changes in the aggregate elasticity estimate even if the elasticities at the

country level remain unchanged.

3. Empirical modeling of import demand

3.1 The imperfect substitute model

This analysis assumes that domestically produced goods are imperfect substitutes for non-oil imports from developing countries, and that these imports are separable with respect to oil purchases. As a result, the demand for these imports can be expressed as

(1) Me = F(Yt, Pe, QM),

>

where Fy?0, Fp<o, Fq,9,

Mt = quantity of non-oil imports from developing countries,

Yt = importer's real income,

Pt = price of imports relative to the price of domestic goods;

Pyt = dollar price for non-oil exports of developing countries;

Et = exchange rate, dollar/foreign currency;

Pyt = GDP deflator for the importer;

Q_ = price of imports relative to price of developed countries;

= Pxt/Qxti

Qyt = dollar price for exports from developed countries. According to (1), increases in either domestic income or domestic prices raises non-oil imports from developing countries. In addition, (1) “allows imports from developed countries to be either complements (Fq>0)

or substitutes (Fg<0) to non-oil imports from developing countries.

3,2 Econometric Specification and Hypothesis Testing

For empirical applications, economic theory provides generally little more than the selection of relevant variables and their anticipated effects. AS a result, pinning down elasticity estimates requires a significant amount of hypothesis testing. This paper tests for the functional form of (1), for the maintained assumptions of the error term (normality, homoskedasticity, and serial independence), for the parameter

restrictions imposed in the dynamic specifications, for parameter

constancy, and finally, for price homogeneity. Appendix 1 tests these hypotheses when the income elasticity is allowed to vary over. the business eycle.®

The results from Box-Cox tests (presented in appendix 1), provide partial support for a log-linear formulation: 7 (2) InMe = ao + Jga ghtinye + Yea ghtinpy + Ypa, gh&lnd, + S1mMy—4

+ )idiSi + OD + ug,

where L is the lag operator, S; iS a seasonal dummy for the ith quarter, D is a dummy for one-time events (e.g., dock strikes), and u is a white noise random term. Besides being the more commonly used formulation, the choice of (2) allows the parameters to be-interpreted as elasticities, which facilitates the computation of the test-statistic

for the hypothesis of price homogeneity.

The above tests are performed with the GIVE computer software developed by David Hendry. With the exception of normality and homoskedasticity, Thursby and Thursby (1984) test for these hypotheses as well as sor the independence between residuals and predetermined variables. Because this paper uses two-stage least squares for parameter estimation, such independence is not being assumed here.

6This possibility has been examined by Marston (1971), Khan and Ross (1975), and Haynes and Stone (1983), among others. In the present context, such distinction might be useful to determine the long-run prospects for the balance of payments of developing countries.

7The Box-Cox tests of Khan and Ross (1977) for total multilatera:. imports Suggest a log-linear formulation. For non-oil imports from non-oil developing countries, the Box-Cox tests provide only partial support for a log-linear formulation because the log-likelihood function is very flat. In turn, this flatness is due to a relatively constant import-GDP ratio throughout the sample (see data appendix). Although this constancy suggests that the elasticity estimates are invariant to a choice between a logarithmic or a linear specification, this invariance does not apply to the properties of the error term, a point stressed by Seaks and Layson (1983). The Box-Cox tests performed here check for serial correlation, homoskedasticity, and normality, the last of which is the fundamental basis for standard Box-Cox tests.

The error term is generally assumed to satisfy the assumptions needed for classical inference -- namely, serial independence, normality, and homoskedasticity. While serial correlation is usually tested, the other two assumptions are taken as valid. This paper follows the suggestion of Thursby and Thursby (1984) to examine the possibility of serial correlation of order greater than one, in addition to the Durbin-Watson statistic. To this end, we test the null hypothesis that all the coefficients of an AR(4) for the residual are equal to zero.

The hypothesis that the residuals behave according to the normal. distribution is tested with the Jarque-Bera statistic (Jarque and

Bera 1980), denoted here as JB: (3) JB = T(y2/(6u2) + (1/24) (u,/n2 - 3)2)~ x2(2),

where T is sample size and Wy is the jth central moment of the distribution of the estimated residuals, G. The first term of (3) measu*es the skewness of the distribution of 0 whereas the second term measures the departure of the estimated kurtosis from the kurtosis associated with the normal distribution. The test for homoskedasticity

is based on the work of Engle (1982) on autoregressive conditional

heteroskedasticity (ARCH). Based on the following model a nw az (4) E(U?;U¢-7) = Yo + Yi Ut-41 >

the null hypothesis of homoskedasticity cannot be rejected if Y,=0, which is tested with a t-statistic. The response of imports to changes in either income or prices

is generally subject to some delay. Contracts, delivery lags, and

10

uncertainty over future supply conditions may account for such a celayed response. Failure to recognize lagged responses could potentially bias the income elasticity estimates with implications for the design cf policy responses to the debt crisis. From an empirical standpoint, the key issues are the length of time that it takes for imports to adjust fully and the nature of the adjustment process. To maintain the analysis at a manageable scale, it is assumed that the full resporise takes place in one year (Grossman 1982 and Goldstein and Khan 1985) and that the dynamic response follows either a Koyck adjustment lag or an Almon distributed lag.

Testing for the validity of the restrictions imposed by these two dynamic specifications involves comparing the sum of squared residuals for both the original and an "unrestricted" dynamic specification. The latter is constructed by eliminating the parameter restrictions associated with the original dynamic specification arid adding, as regressors, all predetermined variables lagged one period.9 The associated test-statistic under the null hypothesis is (5) F=((SSRy - SSR,)/n)/(SSR,/(T-K)) ~F(n, T-K), where SSR ,= sum of squared residuals under the null hypothesis,

SSR,= sum of squared residuals under the alternative hypothesis,

n = number of additional parameters, K = number of regressors.

8For the Koyck distributed lag, ajg is set to zero for 2>1 and all j. For the Almon distributed lag, 6 is set to zero. The Almon lag is assumed to follow a second degree polynomial as in Thursby and Thursby (1984). Endpoint restrictions in this polynomial were needed to avoid the multicollinearity problems, which are also indicated by Grossman (1982). As Klein et. al. (1985) point out, multicollinearity problems are more serious for two-stage least squares than for OLS.

QNote that the construction of the unrestricted specification will not include the lagged dependent variable if this variable already exists in the original specification.

11

Although this test is powerless to discriminate among various specifications for which the null hypothesis is not rejected, it indicates whether a given dynamic specification is not valid.

Constancy in structural parameter estimates is needed to forecast accurately the consequences of either policy decisions or changes in the exogenous variables. However, structural parameters may be changing because of changes in trade patterns during the process of economic development or in government policies. Furthermore, one may question the hypothesis of parameter constancy in light of the pronounced changes in oil prices and exchange rates of the seventies. To test for this hypothesis, (2) is first estimated with data through 1979 and then used to forecast non-oil imports through 1981. Under the null hypothesis of parameter constancy the expected forecast error is

zero, and the associated test-statistic is (Chow 1960, p. 560)

“a(t)2/ (7, ~ K)) ~ F(T,, Ty- K).

T T, (6) ( ¥ Get)? - JF Gct)2)/t, 7f = = 1

1 t=1 t

ie oe er!

where T, = number of observations in the estimation period, number of observations in the forecast period.

lar] is} u

Price homogeneity may not hold in empirical analysis with macercdata such as ours. Aggregation across commodities or use of unit value indexes as a proxy for transaction prices may account for an empirical rejection of price homogeneity. Therefore, it might be of interest to examine whether price homogeneity holds. Given the log-linear specification, testing for price homogeneity amounts to

expressing (2) in terms of price levels and then testing whether the sum

12

of price coefficients is zero. The test-statistic for this hypothesis

is distributed as F(1, T-K).!9

4, Data construction Central to this analysis is the availability of quarterly time series of non-oil imports for Canada, Germany, Japan, the United Kingdom, and the United States from non-OPEC developing countries. Since, as far as we are aware, data of this nature are not publicly available, a deta:.led explanation of the methodology for its construction and the assoc:.ated sources appears in the data appendix.!1

Non-oil imports from developing countries are derived as the difference between total and oil imports, both from developing countries. Data for each country's total imports from developing countries, measured in dollars, are obtained from the Direction of Trade compiled by the International Monetary Fund.!@ Data for each country's oil. imports from non-OPEC developing countries, in value terms, are not readily available. However, data on bilateral oil imports (crude plus

products) for each of the five countries from non-OPEC developing

10The validity of price homogeneity is tested at each point in time and not just in the long-run. For the Koyck lag, the null hypothesis is ep, + Ep + Eqx=0 where e; is the compensated price elasticity for the ith price; for the Almon lag, the null hypothesis is p x,2 + Vp,2 +

Vqx, 2=0, where Vi,2 is the quadratic term in the Almon polynomial for the ith price. If there are no cross- price effects, then Vqx, 2=€ 3x70.

11The Yearbook of International Trade Statistics, published by the United Nations, presents bilateral trade flows with commodity disaggregation, but on an annual basis.

l2rhe Direction of Trade does not include bilateral trade flows disaggregated by commodity. In addition, data for Ecuador and Gadon need to be excluded from the group of non-oil developing countries and included in the group of oil-exporting countries, which is the basis for our calculations.

countries, measured in barrels per day, can be obtained from the Quarterly Oil Statistics compiled by the Organization for Economic Cooperation and Development. !3

With data on oil import prices, one can estimate the value of oil imports from developing countries. Note, however, that neither OPEC's: official price nor the spot price are ideally suited to value oil imports from non=OPEC developing countries. Differences in the sulphur content of oil imports from alternative suppliers, in the crude«product composition of imports, and in transportation costs limit considerably the usefulness of the above prices in our study. To bypass this problem, the paper computes first a price of oil for each country as the ratio between the value of total oil imports and the total volume of oil imports. This average price, which is influenced by oil purchases from OPEC, is then adjusted to take into account the existing country-~differences in the crude-product composition of oil imports, the different regional sources for imports, and the different gravities of crude oil.

With data on the value of oil imports from developing countries, non-oil imports are computed as the difference between total and oil imports, both from developing countries. As a percentage in 1981, non-oil imports from developing countries account for 5.2 percent of total imports for Canada, 12.3 percent for Germany, 13.9 percent for

Japan, 13.6 percent for the United Kingdom, and 18.9 percent for the

13tne earliest date covered by this source, on a quarterly basis, is 1973, which determines the starting period of our analysis. This data source measures oil in metric tons; the data in this study are converted to barrels using standard conversion factors.

United States. The aggregate of these imports account for 56.6 percent of total exports of developing countries to industrial countries.

The price for non-oil exports of developing countries, P;, is a weighted geometric average of the prices of non-oil raw materials and of manufactures. The weights vary over time and are equal to the (normalized) shares of these two classes of goods in merchandise exports of non-OPEC developing countries. Note that this export price is based on multilateral trade weights, which is a limitation of our analysis. However, we found no better alternative. Each country's real non-oil imports from non-OPEC, M, are obtained by deflating their nominal non-oil imports by the non-oil dollar export price of developing countries.

The export price of developed countries, Q,, is a weighted geometric average of their dollar export prices, with weights changing over time and equal to bilateral trade shares. Finally, the exchange rate, the GDP deflator and real GNP for each country are conventionally defined. The data appendix contains the official data sources, the details regarding the constructions of our data set, and time-series

plots of the more relevant relations.

5. Empirical Results

5.1 Elasticity Estimates

The parameters associated with (2) are estimated using two-stage least squares with quarterly data for the period 19741 to 1981IV. In adcition to considering two lag adjustments, we test for the importance of

cross-price effects, which results in a total of four estimating

15

equations per country. Table 3 displays the long-run estimates of income and price elasticities and the significance levels associated with each of the six hypotheses described in section 3.14

For Canada, the long-run income elasticity ranges between 1.0 and 1.5. The own-price elasticity ranges between -0.3 and -0.9 whereas the cross-price elasticity varies between -0.4 and -0.9. Based on the test-results, the best specification for Canada allows cross-price effects with a Koyck distributed lag. For this specification, it is not possible to reject the hypotheses of normality, serial independence, and homoskedasticity. Furthermore, the data support the hypotheses of price homogeneity and parameter constancy.

For Germany, the income elasticity ranges between 1.9 and 2.0, which is a narrow range of variation. The own-price elasticity ranges from -1.0 to -1.3, whereas the cross-price elasticity varies from -0.1 to -0.5. The only specification fully supported by the data has a Koyck distributed lag with no cross-price effects. For this specification, all three hypotheses about the residual are accepted; similarly, the hypotheses of price homogeneity, and parameter constancy

cannot be rejected by the data.

14predicted values for the export price are generated according to (t-statistics in parentheses):

Pxyt = 0.82 + 0.000757Y¢-1 ~ 0.0173EXt=1 + 0.0174PDOME=3, R2=0.95, (0.9) (1.8) (-4.6) (3.0) 19711-1981IV

where Y is our five-country GDP aggregate, in 1972 dollars and exchange rates; EX is the weighted average value of the dollar, and PDOM is the weigited average value of consumer price indexes for the five countries with 1972 exchange rates. The sample means of these variables are 1.885 for Py, 2313 for Y, 100.841 for EX, and 60.69 for PDOM.

For Japan, the own-price elasticity ranges from -1.3 to -I.4 whereas the cross-price elasticity turns out to be not statistically significant. The results also point to a negative income elasticity that ranges from -0.2 to -0.4, being rather significant for the ‘Almon specification. A negative income elasticity for imports is not necessarily in contradiction with economic theory if domestic goods are perfect substitutes for imports. Under this assumption, Magee (1975) establishes that imports will have a negative income elasticity when the income response of the domestic demand for importables is sufficiently small.15 In this case, an increase in income is associated with greater production of domestic importables, which reduces the import gap. Because the composition of developing countries' exports has been shifting towards the kind of goods that Japan produces, an inerease in real income in Japan means greater availability of domestic importables, which reduces Japan's imports from developing countries.16 Note that this negative income elasticity is not due to a violation of the maintained hypotheses for the error term. Furthermore, the data support the hypotheses of price homogeneity, lag structure, and parameter

constancy.

15~his condition can be stated as (SI/DI)ng y>na,y» where ng,y and ng,y are the domestic income elasticity of supply and demand respective -y, and SI and DI are the domestic supply and demand for importables.

16Japan's negative income elasticity for these imports might be the result of a sustained shift away from developing countries and towards

developed countries as a source of inputs. Lower quality and unreiiable supply of the former countries may account for such a shift. Khan and Ross (1975) find a negative elasticity for potential income in total, multilateral imports of Japan. Appendix 1 allows for cyclical and

secular income effects, but the results still point to a negative income elasticity.

17

For the United Kingdom, the income elasticity estimate varies from 1.5 to 2.1. The own-price elasticity ranges from -0.1 to -0.2, whereas the cross-price elasticity is not significantly different from zero. The data support the hypotheses of normality, homoskedasticity, price homogeneity, parameter constancy, and dynamic specification. However, the residuals have serial correlation of order greater than one, a problem not detected by the Durbin-Watson statistic. In view of this serial correlation, it is difficult to give full credit to the test resu..ts for the United Kingdom, which suggests that the associated estinates should be seen as tentative.17

For the United States, the income elasticity ranges from 1.8 to 2,2. The estimate of the own-price elasticity ranges from -0.5 to -2.9, whereas the cross-price elasticity varies from -0.4 to -2.9. The test results reveal that, of the three hypotheses for the error term, only homoskedasticity is rejected by the data with any consistency. The Almon distributed lag with cross-price effects produces the more reasonable estimates, even though this lag structure is not fully supported by the data. The alternative specifications, while consistent with the statistical criteria, do not have significant price elasticities which is in direct contradiction to the available evidence (Grossman 1982, Warner and Kreinin 1983).

The evidence of table 3 suggests several general propositions. First, income elasticities estimated at the country level do not display

any tendency towards the estimates of either Cline (1984) or Dornbusch

17wrnen the serial correlation problem is corrected, the estimates for inecme, own- and cross price elasticities remain virtually unchanged. The reason to present the specification with the serially correlated errors is to maintain uniformity of specification across countries.

(1985). Rather, they reveal a tendency towards a value of 1.5, which is consistent with the evidence of Thursby and Thursby (1984) and Goldstein and Khan (1985). To emphasize these differences, table 4 uses the results of table 3 to compute income and price elasticities for an OECD-aggregate. The results indicate that the aggregate income elasticity varies between 1.3 and 1.6, which is not only. a narrow range of variation, but one that is far below the elasticity estimates of Cline and Dornbusch. '8

Second, with the exception of Canada and the United Kingdon, the estimates of the own-price elasticity are smaller than both ~-1 and the cross-price elasticity. This evidence suggests that non-oil imports from developing countries face greater competition from domestic gcods than from exports of developed countries, which is consistent with the findings of Grossman (1982). From a commercial policy standpoint, large own-price elasticities suggest that protectionist actions might be effective in eliminating competition from developing countries. Bit by the same token, this effectiveness might exacerbate debt-servicing difficulties of the latter countries with a potentially adverse feedback effect on OECD countries. Third, the Koyck lag adjustment seems tc receive greater support from the data than the Almon specification, a finding already noted by Thursby and Thursby (1984).

Overall, the elasticity estimates shown in table 3 are

consistent with both economic theory and estimates from independent.

'18note that Riedel estimates an income elasticity close to ours, despite his assumption of no price effects. However, his analysis is valid

until 1978 and refers to manufactured goods only, which because of their relatively high price elasticity, tend to display relatively small price variations that might lead to statistically insignificant price effects.

studies, possess a narrow range of variation, and satisfy a number of statistical criteria. All of these considerations give some credence to

our estimates.

5.2 Implications of Our Findings for Debt Servicing

One of the most important questions currently faced by policy makers is whether exports of debtor countries can grow faster than their interest payment:s--that is, whether their debt-servicing requirements are sustainable. Because these exports are tied to growth in industrial countries, it might be of interest to estimate the OECD growth rate consis:zent with sustainable debt servicing. To this end, total export revenues are modeled as

(7) X = Py(Y¥) M(Py(¥)/(E Py), Y)/B,

where 8 represents the share of exports of developing countries to industrial countries. Under the assumption of constant 8,

differentiation of (7) with respect to Y yields

(8) K = [(Py/Y)(1 + €) + ny, where X = growth rate of total exports in nominal terms, Py = growth rate of the dollar price of exports, Y = growth rate of industrial countries, € = Own-price elasticity <0, n = income elasticity >0.

According to (8), higher growth in industrial countries translates into higher export revenues of developing countries through two channels: an income effect, represented by n, and a price effect which is decomposed into a terms-of-trade effect, P,/Y, and a direct price effect, €. Both equation (8) and the sustainability condition

(growth in nominal exports2nominal interest rate) might be used to

20

derive a relation between OECD growth and interest rates consistent with sustainable debt servicing: (9) Te/r 2 1/((Py/Y) 1 + ©) +n), where ¥* is the threshold OECD growth rate andr is the nominal interest rate. According to (9), Y¥ is inversely related to both the terms-of-trade effect and the income elasticity, but directly related to both the own-price elasticity (e<0) and nominal interest rates. Substituting n=1.6, e=-0.7, and Py/Y=0.97 into equation "9), we find that an increase in interest rates of 1 percent must be accompanied by a 1/2 percent increase in OECD growth for debtor countries to service their debts on a sustainable basis.19 Thus if interest rates are equal to 8 percent (as in Cline 1984, p.60), then servicing sustainability requires an OECD growth rate of at least 4,3 percent, which is above Cline's (1984, p.67) estimate of 3 percent. 20 The difference between these two growth estimates can be traced to Cline's assumption of a zero price elasticity and to his relatively high income elasticity estimate, since both of these considerations tend to

lower the critical growth rate associated with a given interest

19These parameter estimates are found in table 4, and they provide the lowest threshold growth rate. According to the regression results of footnote 14, a one percent increase in Y raises Px by 0.97 percent. Chu and Morrison (1984) estimate an income-price elasticity of 2, but their analysis excludes manufactured goods, which tend to have large price elasticities for demand and supply, and thus relatively small price changes in response to demand-supply shifts.

20The critical growth rate is derived as 4.3= 8/(0.97(1-0.7)+1.6). With simulations of the LINK model, Klein (1984) finds a critical growth rate of 4.66 percent per year.

21

rate.<! To the extent that OECD countries, as a whole, will not sustain a 4.3 growth rate, these findings suggest that OECD growth, important as it may be, will not be the decisive factor in enabling

borrowing countries to service their debt on a sustainable basis.

6. Conclusions The objective of this paper has been to estimate income and price elasticities of non-oil imports of major industrial countries from non-OPEC developing countries. These elasticities have been the subject of increased attention recently in view of their importance for designing policy responses to the debt crisis. Despite their importance, a review of the literature reveals sharply divided views regarding their magnitudes. This lack of consensus stems from three sources: use of multilateral trade flows aggregated across countries, omission of price effects, and reliance on ordinary least squares for parameter estimation. The elasticity estimates derived in this study eliminate each of these limitations.

| The empirical results at the country level yield an aggregate

income elasticity that varies from 1.3 to 1.6, a relatively narrow range

21cline has pointed out to us that his average elasticity is approximately 2, below his marginal elasticity estimate of 3. The lower average elasticity is due to both a subjective interpretation of what the average OECD growth rate is and a negative intercept in his estimating equat‘:.ons (see Cline 1985). A negative intercept implies a declining trend in imports of industrial countries from developing countries, a trend not borne out by the data. More likely, a negative intercept is the result of omitted variables and measurement errors that impart an upward bias to the marginal income elasticity. More importantly, Cline's (1984) long-term projection model relies on the marginal and not the average elasticity. In any event, substitution of his average elast:.city and zero price elasticity into (9) yields a critical growth rate of 2.7 percent, well below our growth estimate.

22

of variation, and one which suggests an upward bias for estimates of previous studies. With respect to the question of debt-servicing sustainability, our elasticity estimate suggests that reliance on growth of industrial countries to pull out the developing countries from the present debt crisis is not warranted.

This conclusion is subject to a number of qualifications. First, the analysis has dealt with the bloc of non-OPEC developing countries that is a larger aggregate than just debtor countries. It is conceivable that bilateral trade between the United States and major debtors is characterized by higher income elasticities than those obtained here. We are not aware of any available evidence on the magnitude of these elasticities, and it seems that their estimation is a natural avenue for future research. Second, interest rates and OECD growth are not independent of each other, as the analysis of this paper has assumed; a more general equilibrium analysis is needed. Finally, some of the elasticity estimates need further refinement and further testing. Though this analysis has not exactly pinned down the elasticities, it has narrowed down considerably the range of disagreement. Given these limitations, and the importance of this

subject, it seems that further research will yield positive returns.

Non-oil Exports of Developing Countries

Table 1

Country Aggregation Aut-hor Exporter

Bond (1985)2 Non-Oil LDC

Goldstein Non-Oil

Khan (1982)> Lpc

Cline (1984)° World

Dornbusch? Non-0il

(1985) LDC Grossman4@ Non-Oil (1982) LDC Riedel Non-Oil

(1984) LDC

Importer

Industrial

Industrial

Industrial

World

Industrial

Data

Annual 1967-81

Annual 1963-80

Annual 1961-81

Annual 1960-83

Quarterly 1968-78

Annual 1960-78

Comparison of Selected Studies

All Goods

Non-0il

Non-0il

Non-Oil

7T-digit SITC

Non-Oil

23

Elasticity

Commodity Income

2.4

1.3

Price

0.0

0.0

-1.2

=4.5, -0.5

i cre A

a. The estimation method is indirect least squares.

b. The estimation method uses the average of the percentage changes of the ratio between LDC exports to OECD countries and the OECD real GDP.

c. The estimation method is ordinary least squares.

d. The estimation method is two-stages least squares.

24

Table 2 Non-oil Imports from Non-OPEC Developing Countries Income Elasticities and Country Aggregation Biases

1974-1981 Income Hoi NMi=Nus Hor ni=n.

Level of Elasticity - F(2,145) F(5,145) Aggregation n R? (sig.level) (sig.level) Canada 1.265* -0.03 6.162

(0.003) Germany 1.013% -0.01 6.036

(0.003) Japan -1.209* -0.03 6.112

(0.003) United 0.239* -0.03 6.108 Kingdom (0.003) United 5.931 0.27 States Pooled data: 1.595* 0.01 2.546 fixed effects (0.03) Aggregate of 2.858 0.16

5 Countries

* Coefficient not significant at the 5 percent significance level.

The dependent variable is the growth rate of non-oil imports from. non- OPEC developing countries, and the explanatory variable is the ir.come growth rate for the associated country bloc.

Column 3 presents test-results for the hypothesis that the income elasticity of the ith country equals the U.S. income elasticity.

Column 4 presents the test-result for the hypothesis of a common income elasticity across countries.

25

66°0 66°0 €g°0 go°o 06°0 gl°0

9° O-

Of *i-.

66°t

uow ly

30 Jud-860u9

&

. ooo0o0o0°o

phous

geunqonuys 3e7

p47 yauasowoH 0 Jud oHOdY 97 Sepa SOWoH quoyqeTausoo04ny ear AQT TEWUON peysal sesey ods}

aaowoom NM DONO oe

“mG was zu

39 Tud-660uU9 80 Jud-UMO awooul

80 Jud-S60U9 ON

Auewuay

T861-PL6T soyeuTysy AQTOTAseT_ un1-bhuo]

mew ww neers

80 Jud-S60uU9

seme eRe

“aTqe? au; JO pue ey 7e sejoU seg

phoua

geunqonuys 3e4 pAaTeuaZowon

oHOuV 37 4Sepsey SoWoH quoyyeTauucs04qny ear AAT TeWUON peysel sosey ,odAH

“ad ugs zu

80 Jud-Ss80uD 30 Jud -uMO awooul

80 Jud-660uU9 ON

epeue)

svtaquno) DULGo [PART GidO-USj] WOaZ S4AACGUIT [TTO-UON

€ eToeL

26

80 Jud-660u9

wop3uyy peatun

80 Jud-66049 ON

yhoud

geunyonuys 8a]

p43 TeuesowoH 80 Tad oHOUV 079Sepex GowoH quotqe Tesu0004ny ear AAT Tewson peqsel secey ods

*n'd ¥as ey

890 Jud-860U9 90 Tud-UMO e@wooul

80 Tud-660uU9

ueder

(penutzUCO) € eTqeL

‘eTqe; JO pus ey 7e saqou 39S

yhous

9eunyonays 8e1 p*a}eueBowoH 80 Tad oHOUY 9796epe% SOwOH quOTqe Teus000 {ny ear Aap Tewson peysel secoyyodsH

“a'a ¥gas 7

80 Jud-S60u9 80 Jud-UMO awooul

@0 Tud-86019 ON

27

“wopBuTy paqyTUN ayy 4oJ TOG6l pue AIGL6I UT pue 62989S pagyun ayy dos IGL61 UT | JO anTea e& Gaxeq G aTqTesea Auunp ayL

*sysauqodséy TInu ayy 4oafau 04 pauynbau sy 96° JO anTea-4 e& ‘4694

HOYV 84Q 4od [aAaT aouKoTJyUs{S paquodau ayy Jy ‘9699 HOYV 3uy

“*sysauqodAy AQT TTqeqs uaqoweued ayy YUAtmM paqyeyoosse

Oyweukp payoyuysauun ayy UyIM payeyoosse

*sysayqodky AyyauaZowoy adyud ayy yuyta payeyoosse

*paqjoafau sy sysayyoddy payeyoosse ayy uayy ‘G6°O Spaaoxsa

JO uoyydaoxa ayy UITM TaaaT aoueoyJyuBTS *J

"uot ect jJyoeds TAAST adUeDTJTUBTS °e

T@aAeT soUuRedTJTUBTS °p

*stsayuyodsy AyyoTysepayxsowoy ayy UAta payeyoosse OF46TIeIS-7 °O

*sysaujodsy uoTyeTauuoo0yne OU ayy UIJM paqeTooSsse

*stsayuqodsAy AqyTewuou ayy yAyM paqeyoosse

TasaT aouRoTJyuByTS *q

[aeaaT aduedTJJusTS °e

*TaAaT aouRedTJTUSTS YUsduad G ayy 4e JUROTJTUBYS Jou yuayoOTJJa0D »

nS°O 96°0O 8S°0 66°0 1€°0 0S°0 Sh‘O 6L°0 hho 0s°€ Go°e 09°h 0g*z 26°O Sh‘O 16°0 Sh°O S6°O 20°0 9n°O S0°0 hi°?e Lz*2 65°14 6L°2 90°O 90°0 10°0 90°0 16°0O £6°0 68°0 n6°O 68°t #ch’O 0°O 0°0 06 *e- #46 °0- #£9°0- #0S°0- 4e°t 18° 02°z 96°1

uouTty yohoy

80 Tud-G60u9 80 Tdd-660U9 ON

684e9S peqTuN

(penuT3U00)

€ STqeL

ghouo geunqonuys 8e4 p{aTeuasowoH 80 Tud oHOUV 0736epay SOWOH qu0Tqe Teus0004ny e’a°r AqTTewu0N peqseL sesey ody

yas eu

80 Jud-660U9 80 Jud-UMO guoouy

28

Table 4 Income and Own-Price Elasticities Country Level and Aggregate Level

OECD4 Shares Income elasticity Own-Price elasticities smallest. largest smallest largest

Canada 4.7 0.95 1.52 -0.85 -0.25 Germany 16.0 1.93 2.02 -1.30 -0.95 Japan 21.0 -0. 38 -0.13 -1.39 -1.25 U.K. 8.6 1.35 2.06 -0.22 -0.11 U.S. 49.7 1.81 2.20 -2.90 -0.50 Aggregate 100 1.29 1.64 -2.00 -0.68

a. Normalized shares in OECD GDP. For the actual shares in the OECD aggregate, multiply by 0.70; these shares are obtained from the OECD Economic Outlook, 1981, p. 16.

29

Appendix 1

Estimates for Alternative Specifications

This appendix presents elasticity estimates which allow for both business cycle and quantum effects. In addition, it examines the extent

to which the functional form of (2) is supported by the data.

A1.1 Cyclical and Secular Responses to Output

A number of empirical studies have shown that the income elasticity might vary over the business cycle. From an empirical standpoint, the key question is to decompose a given income path into its secular and its cyelical components. Khan and Ross (1975) and Marston (1971) estimaze secular (or potential) income as a trend of actual income, which permits computing income fluctuations as the difference between potential and actual income. This approach has been criticized by Haynes and Stone (1983), who argue that a spectral decomposition of output provides a more reliable estimation procedure.

Given that the Multi-Country Model operates in time domain only, and given that Haynes and Stone acknowledge that their approach has certain limitations, this paper estimates potential output as a trend of actual output. Under this assumption, equation (2) becomes (A1) in Mg = a9 + a, In YPOT, + ap In (GDP/YPOT)¢ +

+ a3 (L) In Pe + ay (L) In Qe + 6 In Me-y, where YPOT = EXP(Bo + 8, TIME) with the 8's estimated by OLS. Note that (A1) implicitly assumes that a decomposition of output into its cyclical and secular components eliminates lagged effects of income on imports

(Marston 1971). This observation suggests that the value of § should be

30

zero in estimation. But, for the sake of sensitivity, the results presented here include the case where 6 is non-zero.

Table Al presents long-run elasticity estimates following the format of Table 3. On the whole, a comparison of the results between these two tables reveals that the use of potential output (as constructed here) produces no major changes in the elasticity estimates or in the various test-statistics. The two exceptions are the United Kingdom (where there the income elasticy increases and the own-privxe elasticity declines) and Canada (where there is an increase in the significance levels of the various hypotheses). It must be emphasized that this robustness of parameter estimates rests on the assumption that potential output can be estimated as trend of actual output. A more complete sensitivty analysis should address the issues raised by Haynes

and Stone (1983).

A1l.2 Quantum Effects

A second limitation of equation (2) is the assumption of no quantum effects -- that is, the response of imports to changes in either prices or income is the same regardless of the magnitude of the changes in these variables. The hypothesis of quantum effects is examined here by

postulating the following parameter behavior:

a,(L) = a1 = a19 + ay, AlnYy

ao(L)

ao = G29 + ao, AlnP,

a3(L) = ag = a39 + a3, AlnQy.

With the above formulation, testing for the hypothesis of' no-quantum effects amounts to testing whether @11, G24 and a3, are

zero. Once again, it is assumed that the lagged response of import:s is

31

due to the omission of quantum effects, and that once these effects are recognized, there is no need for additional lagged variables. Elasticity estimates presented in table A2 reveal a relatively

robust income elasticity but a good deal of sensitivity in the own- and cross-price elasticities. Several reasons may account for this sensitivity. First, dynamic adjustments are not solely due to quantum effects, as it is implicitly assumed in the results of table A2. Second, already high levels of collinearity might be exacerbated, which

prevents isolation of coefficient estimates.

A1.3 Choice of Functional Form The general functional form of equation (2) is

h d (a2) (Mp - 1)/A= og + (a, (L)76(L)) (XE - 1)/d) +

d + (a, (L)/6(L)) (PE - 1)/x) d + (a, (L)/6(L)) (Cp - 197A) + ut

where A is the Box-Cox parameter, 6(L)

1 - 6b, aj(L) = ) ojgh®,

(for j=1,2,3), up~ NID(O, g2), and L is the lag operator. The value of i is determined by maximizing £, + (a-1) }inYe, where £, is the concentrated log-l.ikelihood function for h=1. The maximizing value of ) is found through a grid-search on which 4h is allowed to vary from 0 to 1 witha 0.1 step size. This procedure is applied to a total of eight specifications for each country: with and without cross-price effects, with and without potential output, and two distributed lags (Koyck and

Almon).

32

Figure Al displays the log-likelihood function for each value of 4 for all specifications for all countries. The results reveal a log-likelihood function almost invariant to the choice of Box-Cox parameter. One possible reason for this invariance is the constancy of the import-GDP ratio (see Appendix 2), which would give the same income elasticity estimate for either A=0 or A=1.

However, to find that elasticity estimates might be invariant to the choice of \ is not equivalent to saying that the properties of the error term are invariant to the choice of \, a point stressed 3y Seaks and Layson (1983). They argue that the optimal i is influenced by the presence of heteroskedasticity. To recognize this influence, they derive the log-likelihood function with (proportional ) heteroskedasticity, under the assumption of normality. The approach followed here is to test the assumptions behind the error term before altering the likelihood function. After all, if the residuals do not behave according to the normal distribution, then there is no point in deriving its log-likelihood function with or without heteroskedastic errors.

Table A3 indicates whether a given specification passes (P) or fails (F) the assumptions of normality, serial independence, arid homoskedasticity for each value of \. Out of 440 cases (11x8x5), the hypothesis of normality is rejected in 14 instances (Canada 9, Japan 1, United States 4), the hypothesis of serial independence is rejected in 81 instances (Canada 1, United Kingdom 80), and the hypothesis of homoskedasticity is rejected in 57 instances, all of them for the United States. The results for the United Kingdom indicate serial correlation

of order four, which does not necessarily invalidate the choice o? A,

33

since there is no evidence of first-order serial correlation. The results for the United States indicate a fairly well grounded heteroskedasticity problem for the Almon distributed lag. In contrast, the data supports the three hypotheses for the Koyck distributed lag for 4>0.4, for which the hypothesis of A=O cannot be rejected. These results, when combined with the ease of interpretation of a logarithmic formulation, provides some basis for the choice of functional form of

(2).

34

n9°O- € 96°14

uow ty

80 Jud-660U9

Auewusy

30 Jud-S60U9 ON

yhoud geunyonuys 3e7 pA3}auaBowoH ed Jud oHOUV 979Sepa7 SOWOH quot qe lasu0904ny edr AqtTewuoN payselL seseyyodséyH

"m'G was zu

80 Jud-S60u9 39 Jud-UMO awoouy Te} Used

T86T-PLé6T SZOOF JA TROT TOAD-TeTNDeS sojyeuTysy AYTOTAseTT unrz~hHuoT setarqunop butdoTeAeq DadO-UON “UbaAzF sjzazodur [TO-UuON TY eTqeh

39 Jud-ssoug

8d Jud-S60U ON

“oTqe} FO pus syA We Ss7O0U 8285

yhoun

geunqonuys Be]

p43 }euaBowoHn 89 Jud oHOUV 37 4Sepaey sowoH quote Tauu0s04ny e'd°r AqyTewu0N peqysel seseyqodséH

*m"d was zu

80 Jud-SS0u9 30 Jud -uMO awoodul Te} UWI Od

98 °0 25°0 n6°O h9°O 2L°0 91°0 2h‘ 00°0 €n°O Sg *0 €£0°0 €o°t 10°0 86 °0 66°0 g6°0 66°0 Sh °O 10°0 Sh°O 20°0 09°41 26°1 19°L €6°1 90°0 S0°0 S0°0 S0°O 2L°0 n’0 €L°0 SL°0 *h0*0O- *50°O- 0°0 0°0 *60°0- abb'O- *80°0- «21 °Og2°2 69°1 hE 2 eL*t

o0h0y uow Ty yofoy

80 Jud-S60uU9 0 Jud-S60u9 ON

wop3uty payyun

j” a2eunyonuy4s

p43 }auaBowoH Co) oHOUY 9746epa 60!

quot ye Ta44u090

ear Aqttew

peysel sesey,0dAH

@ Jud-660u9 80 Jud-uUMO awoouy Te} Ua4od

oud 8e1 Tad WOH qny JON

(penut3UC9) TY eT4ei

80 Jud-S60u9

ueder

“aTqeq JO pues ayy je sajou oes

30 Jud-660u9 ON

nous

geunyonuys 8e7

p44} auasowoH 89 Jud oHOUY O746epax soWwoH quo} ye Tauuos04ny e'd'°f AyyTeWuON peqsel sacay zodséH

80 Tud-680u9 89 Tud-UMO awooul Teyquayod

36

“wop8uyy F2dTUN ayy 4OJ I0G6l PUe AISLOL uy pue sayqeqs paysun au y wos IS/61 ut | JO anTeaA e sayey g aTqyeuea Auwnp aul

“stsayyodky [Inu ayy 4afau 04 Pauytnbau st 96° JO anTea-4y e@ ‘4804

HOUV 844 4od ‘“paqoefau sy} syTsauyodAy payeyoosse ayy uay. 'S6°O Spaaoxa TaasT aouedTJTUBTS paquodau ayy Jt ‘34884 HOW auq JO UOT4daoxe ayy UITM “syeseujodsy AAT 1qQeys uaqyoweued ayy UYTM pajeyoosse TaAaT BOURDTJTUBIS “J

“uoTyeotjyoads OyTweukp payyuysauun ayy UATM PayeTOOSse TeAaeT |adURdTJTUBTS ‘a

“sTsayjedAy AyrauaBowoy adqud ayy YU4TM payeyToosse TaraT @oueot JTUBTS *p “syTsaujodky Ayo TISepaxsowoy ayy UYTM PayepToosse O74STIe4S-4 °O *sysayjodsY uoyyeTausooo04Ne OU ayy UTM PayeyOOSSEe TaraT aouedyJTUBTS "q “spTsayjodAy AATTewsoUu ayy UATM PayeToOSSE TardT BdueOTJTUBTS *e

*TaAaT BdURdTJTUBTE quaduad G ayy ye 4UROTJTUByS you yuayoTJyaoy ,

16°0 £9°0 66°0 s9°0 66°0 h6°O £6°0 S6°0O 02°€ HLre L6°E c8°e 08°0 OL°O 28 °0 69°0 99°0 00°0 St°0 10°O 20°2 02°2 Ont E12 90°0 90°0 40°0 90°0 26°0 £6°0 06°0 46°0 6L°1 «SE °O 0°0 0°0 09°2- #00°L- «ft °O- #59 °Oeet eel 90°2 nO"

yo foy uow ty 9OLOY

a0 Tud-ss0u9 0 Jud-S60U9 ON

6a3e4S paytun

(penutquoo) TV eTqeL

Roun geunyonuys 8e4 pAaTaueZowoH 39 Jud oHOWV OT9SepsySowoH quot ye Tass0904ny e’a°r AqtTewuon peqsel saseyqodéH

°M"d ¥as zu

89 Jud-660u9 89 JUd-UMO awoout [1 4Ua4Od

37

Table A2 Non-oil imports from Non-OPEC Developing Countries Income and Price Elasticities Quantum Effects

Long Run Elasticities Canada U.K. Germany Japan U.S. Income

010 1.17 0.89% 2.28 -0.17% 2.07

O44 -0.04% -0.04% 0.54 0.06% -0.09* Own--Price

a20 -0.77* -0.16* -1.16 -0.70 -O.71*%

a4 -1.68% . 0. 38* “4,39 -2.78% 0.25* Cro3s-Price

230 -0.06* 0.25* -0.49 -0.63* 0.25%

034 -4, 35% -0.95* 8.30 -0.89* -2.00* R? (adj.) 0.39 0.47 0.94 0.22 0.90 Normality: JB@ 0.68 0.59 0.10 0.21 0.99 Homoskedasticity ARCH? -0.54 -0.70 0.69 - 1.36 4.75 Serial Indep.@ 0.23 0.99 0. 80 0.98 0.99

eer

* Coefficient estimate is not significant at the 5 percent significance level.

a. See notes in Table 3.

Table A3 Non-Oil Imports from Non-OPEC Developing Countries

Box-Cox Tests and Testing of Error Properties (Notes at the End of Table)

Koycek

Almon

Koyck

Almon Serial

Serial

Serial

Serial

Indep. Homosked.

Normality

Homosked.

Indep.

Normality

Indep. Homosked.

Normality

Indep. Homosked.

Normality

a@aaaaaaaaaada

QOaaaaaaaaaiaa

Q&aaaaaamamaaaa

0.0

Germany 1

Oaaaadnaaa

Qaaaaqannhaowaa

Qaaaaaadaaaa

NMNMTNOM OHO ee eo 6 oe ee oooooo0o0oer

Qaaadmhaaaaaamas

QOaaaaaiaaaaa tm

Oaaaadaaankhkaaaas

0.0

6 ao a it oO Lo)

Qaaaaaa

aaaaqaaa

Vt O&O 0. a,

MH NMI NO oe ee ewe ooooo°o

aaaa

aaaa

aaa. a

~- ONO oe ee oo0oe

aOaaaaamaaaaa

aOaaaaaaaaaa

aOaaaadamaaaaa

0.0

Germany 2

a.

fe.

a.

a,

a.

a.

0.0

Canada 2

Oaaadaaankaa

aaandnandnadqdaa

QOaaaaaadaaad

HKNMTNOMDANHNO oe ee ew ew ee eooo0o0o0o0oc0ooOe

Qamaaaannaan

Oaaaaaaaaa

Qaaaaanhaanada

QaOaaaaqaaaaaa

Qaaqaaaaaaaa

Qaaadmaaqaaaaa

KHNMTNOE-DHO Ce ee oooo000000-

aaaaadadaaadaad

Oaaaaaaaaaa

aaadaanaaaaa

0.0

Germany 3

aaaaaaaaa

Oamajandnanknadna

Qaaaadaaa

NMeTNO™ONHAO oe oe oe eo oe ee oooooo0oo-

Qa maaankmhanhaa

Oaaaaaadkhaaa

aaadaadnaaaand

0.0 0.1

Canada 3

QOaagqaadaiaaa

Qaaaaqadaaadka,

G&miaaaaaaa

NMTNOM ODO ooooo0o0o0o0co-

a.

a.

a.

0.0 0.1

Germany 4

a.

On

a

a

a

t,

0.0

Canada 4

38

oh OO OO om Oo a

GQwmwaaaaqjaasaaa

a.aaaaawmaa

ao r.aoaaackraaa

Oa irwtdaaadnanad

Qawaaaaaaad

NMaTNCHODO ee 8 ee eo ew aoo0oocoocoe

QOaaaagqaaaaaa

aoaaiaaadaana

Qaaaanhmaaaa

Qaaaaaadaaaa

aaaawaaadnaa

hRaadaaaaaaa

-“NMFTNOEDANHO Cae eo ew we we cooo0oocoooe

Table A3 (continued)

Almon

Koyck

Almon

Koyck

Serial

Serial Indep.

Serial

Serial

Indep. Homosked.

Normality

Homosked.

Normality

Indep. Homosked.

Normality

Indep. Homosked.

Normality

a

a.

fu

a.

1

U.K.

a.

a

a.

0.0

Japan 1

Qagaagaaaaaa

Cae Cee fe Con Cau few Cru Cy Coy foe

aaaanananaaaa

Qaaaaqaaaaaaa

fina Cos Cou fru Con Cru Cry Cry Cor Cre

aaaaaandmaaa

HK NMTNOLeDAHAO oe 6 eo ee ew we oooooo0o0o0c0o-

Qaaaaaaaaaa

aaaaaaAaaaa

aaaaaaadaiad

aaadaaadkhaaa

oaaaaaanand

aoajaaaaeoaoaa a

HKNMTNOODHO oe oe oe eo ew ew ee ooooocooo-

Qaaadaaaaaaana

Cay Gan fee Cru Cru Con Cre Cru Cre Cor Cre

Oaaaaaaaaaa

aaaadaaadkmaaoas

fina Cou Con fe Cra Cae Con Crs Exe Crs Cre

Qaqaaaaiandnhanand

OK NMTNOLS DAS ee ee ew we we ew ooooo0o00co004+

U.K

Qaqaaqaaaaanaaa

QOaaaaandamaaaaa

Qaaaaadmanhaadada

aaaaaaaaaaa

aaaaaaadanda

oaaaaaaaaa

OK NMTANOMDHO ee © © © oe ee ee oooo0ooooo0oor-

Japan 2

aaaadaadaaaa

OQ. A. A. OL Oy fey Cee Con Cae Ce Ce

aOaaaaakmadawaa

QOagaaaadmaana

Coy Cana Ce Ce Coe Cen Cae Cee Gn Cee Cae

QOaaaaadaaadaaa

OK NMATANOLYDAHAOS oo © © © oe ew we ee oooo00o00o0c0c0oO-—

U.K.

Qaaaaqaaaaanwiaa

aaanaadamanaand

Qaadaaaaadaaa

Qaaaaaaaadnaa

aaaaanmaaaan

aaaaaaamaaaa

OK NMTMNOK”-DAHO Ce ee ee ooooo00o0000-

Japan 3

39

QOaaqaaaauwadad

Qn OC. Cy fee Cer fey Ge Cen Cry fe Cy

Oaaaaaaaaaa

aaaaaankaaaas

Oi Oe eS

Qaagdaaaaaaa

CK NMTNWErDAHO oe 6 eo we ew we ooooo0oo0o0o0004e

U.K. 4

aaaaaaaqaaaaa

aaaaaaqaaaaaa

&aadgaadaaaadaa

Oaaadaaankhaankna

Qaanaanananan

aOoaaaaaaadada

OK NMTMNOMODAHAO ee 6 eo oe oe te ew ee ooo00o0o000004¢

Japan 4

Table A3 (continued)

Almon

Koyck

no cross-price e€

ffects

ffects

potential output and

cross~price e

Case 1 Case 2

Serial

Serial

Indep. Homosked.

Normality

Indep. Homosked.

Normality

n

E

MER 3 8 OW bat OB e 228

Case 3 Case 4

tu

a

a

1

U.S.

Ces Cou Cou Cou Eee Gee Ceo Eee Cou fos

Oadaaaaaoaoaa

Oaaaaaanand

B&meamaooaaoaa a

Qaaaaankhankaad

Qaaaaaaaaaa

rKHNMTNOr OHNO eo ee eo oe oe ee eoo°oc0c°0ce0c°cdoe

Cee Ces Ces Cou Cou fee Coe fee Cor fer A,

2aqaaaaaamaaa

Boe meoaawwmaaaa

Binh OooaaAaa a

oa aaaaaaaa

Qaaaadkiaanaaaa

SOCK NMTNOMDADHO Cn ee ee oeooooo0oooo°0o°or

U.S.

Gee Cou Cee Cee fy Coe Cee fee Cow Cow Cow

aaaaaaaaaaa

Qaaaaadmaaaada

Bhi nheoaaaaaaa

Qaaadadaaaankhaaaa

Qaaaaankaaaa

OK NMTNOMDAHAOS ee eo oe te te te we ooooo0o0oo0o°00co0-

U.S.

fe

a,

a.

te

a

a

fo}

U.S.

40

Coe Cee Se Cee Eee Coe Cee Cte Cee Cow

Qaaaaandtaaand

Oaaaaddaanhaaa

GBiaeaAacncanwad

Qaaanhaaaanw

aaa aaaaawaa

KNMTFNOMrMOANO oe ee ew eo we oe ooooooco0oce

Al

$}0ajja voTAd-sso1z9 s]0ajyja voTAd-sso1o ou

£* 7° T° T L° 7°

(POOUTTSYTT 80T :stxe-A ‘a1ajaueied XO9-xXog :STXe-x)

s3sal xo9j-xog

setijquno) sutdojTeaeq 9yq0-UON wory szaoduy TTo-uon TV eansty

epeue)

wopsuty peztun

Aueuiay

ueder

so02e1S peqtug

42

3ndjno [eT Uuajo0d $]09FFJa soTAad-ssoaod

3ndjno Tetquajod S}08sja eoTad-sso1zd ou

T L° 7°

(penut uo.) TV ean3ty

wWopsuTy pejtuyg

Aueuiay

ueder

so}7e1g peitun

43

Appendix 2

Definitions of Variables and Data Used

All data are in billions of U.S. dollars, quarterly at annual rates, unless otherwise noted. Where source data differs, it has been scaled to conform to this convention. A complete explanation of all source

abbreviations follows the data definitions.

Mi = real non-oil imports of the jth developed country from non-OPEC developing countries.

= [MiLV - (MiLO * (MiPETV * iER/MiTO) )]/LPXNO.

MiLV goods imports of country i from non-OPEC developing countries (millions of $US, FOB), DOT line 71.

MiLO

oil (erude and refined) imports of country i from non-OPEC developing countries (thousand metric tons), OECD, divided by 0.136 (metric tons/barrel).

MiPETV

ut

Petroleum imports of country i (local currency), IFS line 71a [for Germany, Japan, U.K., U.S.]; BOC Cansim series B43154 (crude) and B4¥3157 (refined) [for Canada].

iER = U.S. dollar exchange rate ($U.S./local currency), MDL: quarterly average of series SXMBCD (Canada), SXMBDM (Germany), SCDBJY (Japan), and SXDBUKP (U.K.).

MiTO

Total oil imports of country i (thousand metric tons), OECD divided by 0.136 (metric tons/barrels).

LPXNO

i]

Non-OPEC developing countries non-oil export price index: Weighted average of external trade deflators for manufacturers

and non-oil primary products (WB), rebased to 1972 = 1,

LPXNO

iPGNP

44

converted from annual to quarterly data according to seasonal

pattern of non-OPEC LDC export unit value index, IFS line 74D.

Real income of ith developed country.

iGNP (1972 prices).

Canada: nominal GNP, Cansim series D40551 (millions of

Canadian $), divided by constructed absorption deflator, CSR

Table 1.2.

Germany: constructed from components (billions of DM), DIw.

Japan : constructed from components (billions of Yen), BCJ.

U.K. : nominal GDP (billions of pounds sterling), ET, Table 2, multiplied by 0.62113 (implicit GDP deflator), ET, Table 4,

U.S. : NIA Table 1.2, line 1.

Relative* price (foreign prices/domestic prices).

(LPXNO/iEL)/iPGNP.

Non-OPEC developing country non-oil export price index (WB,

defined above), divided by 1972 average value.

Spot Exchange Rate Index --$US/local currency exchange rate

(MDL, defined above), divided by 1972 average value.

GNP Deflator.

Canada: Nominal GNP (millions of Canadian $), CSR Table 1,2,

series D40551, Germany: Nominal GNP (billions of DM), DIW,

defined above).

Japan: Nominal GNP constructed from components (billions of

Yen), BOJ, divided by Real GNP (BOJ, defined above).

U.K.: Nominal GNP (billion of pounds sterling), ET Table 2,

divided by real GNP (ET, defined above).

U.S.: NIA Table 7.1, line 1.

Sources BOC BOJ

CSR

DIW

DOT

IFS

MDL

NCA

OECD

l&

45

Bank of Canada .

Bank of Japan.

Canadian Statistical Review, published quarterly by Statistics Canada.

"Lange Reihen der vierteljahrlichen volkswirtschaftlichen Gesamtrechnung fur die Bundesrepublik Deutschland", published quarterly by Deutsches Institut fur Wirtschaftsforshung, Berlin.

Direction of Trade Statistics, published monthly by the International Monetary Fund.

Economie Trends, published monthly by U.K. Central Statistical Office.

International Financial Statistics, published by the International Monetary Fund.

Macro Data Library of the Federal Reserve Board of Governors. National Income Accounts in Survey of Current Business, published monthly by U.S. Department of Commerce, Bureau of

Economic Analysis.

Quarterly Oil Statistics, Tables B1 (Crude + Natural Gas

Liquids + Refinery Feed Stocks) and B2 (Total Products), published quarterly by OECD International Energy Agency. World Bank, Economic Analysis and Projections Department,

Division of Global Analysis & Projections.

Figure A2 Non-Oil Imports from Non-OPEC Developing Countries 46 Relative Prices

United States

1978 1977 1979 1981

Japan

1978 1977 1979 1981

Germany

1975 1977 1979 1981

United Kingdom

1975 1977 1979 1981

Canada

1975 1977 1979 1981

Figure A3 47 Non-Oil Imports from Non-OPEC Developing Countries Import-GDP Ratio

A) : 2 United States 1 oO

1975 1979 1981 Ms mae 3 Japan 2 1

1975 1981 4 3

Germany

2 1

1975 1979 1981 6 United Kingdom a 4 2

1975 1979 1981

16 Canada _ |, ; - wont 0.8 0

1981

48

REFERENCES

Adams, F.G., E. Sanchez, and M. Adams, 1983, Can Latin America earry its international debt? A prospective analysis using the Wharton Latin American debt simulation model, Journal of Policy Modeling, 5, 41¢-41.

Bergsten, F. and W. Cline, 1983, Trade policy in the 1980's: An overview, in Cline (ed.), Trade Policy in the 1980's (MIT Press, Cambridge).

Bond, M., 1985, Export demand and supply for groups of non-oil developing countries, IMF Staff Papers, 32, 56-77.

Chow, G., 1960, Tests of equality between sets of coefficients in two linear regressions, Econometrica, 28, 591-605.

Chu, K. and T. Morrison, 1984, The 1981-82 recession and non-oil primary commodity prices, IMF Staff Papers, 31, 93-140.

Cline, W., 1984, International debt (MIT Press, Cambridge).

Cline, W., 1985, International debt: Analysis, experience and prospects, Journal of Development Planning, No. 16, 25-55.

Dooley, M., W. Helkie, R. Tryon, and J. Underwood, 1983, An analysis of external debt positions of eight developing countries through 1990, International Finance Discussion Paper No. 227, Federal Reserve Board.

Dornbusch, R., 1985, Policy performance links between LDC debtors and industrial nations, Brookings Papers on Economic Activity, No. 2, 303-366.

Engle, R., 1982, Autoregressive conditional heteroskedasticity wit:h estimates of the variance of the United Kingdom inflation, Econometrica, 50, 997-1008.

Goldstein, M. and M. Khan, 1982, Effects of slowdown in industrial countries on growth in non-oil developing countries, Occasional Paper No. 12, International Monetary Fund.

Goldstein, M. and M. Khan, 1985, Income and price effects in foreign trade, in R. Jones and P. Kenen (eds.), Handbook of International Economics, vol. II (North-Holland, Amsterdam).

Grossman, G., 1982, Import competition from developed and developing countries, Review of Economics and Statistics, 64, 271-281.

Haynes, S. and J. Stone, 1983, Secular and cyclical responses of 1J.S. trade to income: An evaluation of traditional models, Review of Economies and Statistics, 65, 87-95.

49

Jarque, C. and A. Bera, 1980, Efficient tests for normality, homoscedasticity, and serial independence of regression residuals, Economic Letters, 6, 255-259.

Khan, M. and K. Ross, 1975, Cyclical and secular income elasticity of the denand for imports, Review of Economics and Statistics, 57, 357-61.

Khan, M. and K. Ross, 1977, The functional form of aggregate import demand, Journal of International Economics, 7, 149-169.

Klein, L., 1984, World recovery and debt prospects, in B. Hickman (ed.) International monetary stabilization and the foreign debt problem, Proceedings of a Conference Co-sponsored by Project Link and the Federal Reserve Bank of San Francisco.

Klein, L. and M. Nakamura, 1985, Singularity in the equation systems of econometrics: Some aspects of the multicollinearity problem, in J. Marquez (ed.), Economic theory and econometrics (Basil Blackwell, Oxford).

Magee, S., 1975, Prices, income, and foreign trade, in P. Kenen (ed.), International trade and finance: Frontiers for research (Cambridge University Press, Cambridge).

Marston, R., 1971, Income effects and delivery lags in British import demand: 1955-67, Journal of International Economics, 1, 375-399.

Riedel, J., 1984, Trade as the engine of growth in developing countries, revisited, Economic Journal, 94, 56-73.

Seaks, T. and S. Layson, 1983, Box-Cox estimation with standard econometric problems, Review of Economics and Statistics, 65, 160-164.

Thursby, J. and M. Thursby, 1984, How reliable are simple, single equation specifications of import demand? Review of Economics and Statistics, 66, 120-128.

United States Congress, Joint Economic Committee, 1984, Hearings on International Debt, March 28.

Warner, D. and M. Kreinin, 1983, Determinants of international trade flows, Review of Economics and Statistics, 65, 96-104,

50

International Finance Discussion Papers

IFDP NUMBER TITLES AUTHOR(s) 1986 280 Taxation of Capital Gains on Foreign Garry J. Schinasi Exchange Transactions and the Non-neutrality of Changes in Anticipated Inflation 279 The Prospect of a Depreciating Dollar Jacques Melitz and Possible Tension Inside the EMS 278 The Stock Market and Exchange Rate Dynamics Michael K. Gavin 2TT Can Debtor Countries Service Their Debts? Jaime Marquez Income and Price Elasticities for Exports Caryl McNeilly of Developing Countries 276 Post-simulation Analysis of Monte Carlo Neil 8. Ericsson Experiments: Interpreting Pesaran's (1974) Study of Non-nested Hypothesis Test Statistics 275 A Method for Solving Systems of First Order Robert A. Johnson Linear Homogeneous Differential Equations when the Elements of the Forcing Vector are Modelled as Step Functions 274 International Comparisons of Fiscal Policy: Garry Schinasi The OECD and the IMF Measures of Fiscal Impulse 273 An Analysis of the Welfare Implications of Andrew Feltenstein Alternative Exchange Rate Regimes: An David Lebow Intertemporal Model with an Application Anne Sibert 1985 (partial listing) 272 Expected Fiscal Policy and the Recession William H. Branson of 1982 Arminio Fraga Robert A. Johnson 271 Elections and Macroeconomic Policy Cycles Kenneth Rogoff Anne Sibert 270 Assertion Without Empirical Basis: An David F. Hendry

Econometric Appraisal of Monetary Trends

in ... the United Kingdom by Milton Friedman

and Anna J. Schwartz

ER

Neil R. Ericsson

Please address requests for copies to International Finance Discussion

Papers, Division of International Finance, Stop 24, Board of Governors of the Federal Reserve System, Washington, D.C. 20551.

IFDP NUMBER 269

268

267

265

264

263

262

261

260

259

258

257

256

255

254

International Finance Discussion Papers

TITLES

Canadian Financial Markets: Proposal for Reform

The Government's

Was It Real? The Exchange Rate Interest Differential Relation, 1973-1984

The U.K. Sector of the Federal Reserve's Multicountry Model: The Effects of Monetary and Fiscal Policies

Optimal Currency Basket in a World of Generalized Floating: An Application to the Nordic Countries

Money Demand in Open Economies: Substitution Model for Venezuela

A Currency

Comparing Costs of Note Issuance Facilities and Eurocredits

Some Implications of the President's Tax Proposals for U.S. Banks with Claims on Developing Countries

Monetary Stabilization Policy in an Open Economy

Anticipatory Capital Flows and the Behaviour of the Dollar

Simulating Exchange Rate Shocks in the MPS and MCM Models: An Evaluation

Trade Policy for the Multiple Product Declining Industry

Long Memory Models of Interest Rates, the Term Structure, and Variance Bounds Tests

Currency Substitution and the New Divisia Monetary Aggregates: The U.S. Case

The International Transmission of Oil Price Effects and OPEC's Pricing Policy

U.S. Banks' Lending to Developing Countries: A Longer-Term View

Conditional Econometric Modelling: An Application to New House Prices in the United Kingdom

51 AUTHOR(s) Garry J. Schinasi

Richard Meese

Kenneth Rogoff

Hali J. Edison

Hali J. Edison Erling Vardal

Jaime Marquez

Rodney H. Mills

Allen B. Frankel

Marcus H. Miller Arnold Kling Arnold Kling Catherine Mann Gary S. Shea Jaime Marquez Jaime Marquez Henry S. Terrell

Rod Mills

Neil R. Ericsson David F. Hendry