( J) Evaluation of Sources. I appreciate the examiners'

Transcription

1 THE SUPPLEMENT TO MY "THE IMPORT DEPENDENCE IN THE AUSTRALIAN ECONOMY" I have heard from Professor H. W. Arndt that my three examiners have unanimously recommended that my thesis be accepted for the degree of Doctor of Philosophy "subject to the University being satisfied that the candidate has made such minor corrections or amendments as are specified" in the examiners' reports. In particular I have been required a supplementary report to defend myself against criticisms under three headings : (1) Theory; (2) Statistical Tests; ( J) Evaluation of Sources. I appreciate the examiners' kindness in giving me an opportunity to defend myself as well as in making many helpful suggestions at the oral. 1 Theory It is reported that "when questioned about the omission (of three important variables - the effect of export growth, shifts in the terms of trade and capital inflow) at the oral, Mr. Ikema regarded these variables as broadly neutral in the long run" without evidence to support the assertion. Because of my short explanation, there was a misunderstanding between the examiners and myself. When I said that the growth of exports is broadly "neutral" in determining the import-gnp ratio, I meant that it is neutral in the sense that in the long run the balance of trade is brought into equilibrium

2 2 through the market mechanism in the foreign exchange market or policies affecting the demand and supply of foreign exchange, and hence exports and imports generally move together. However, the neutrality of export growth in this sense does not imply that total export receipts, the terms of trade and/or net capital inflow have no effect on the import-gnp ratio. Let me consider briefly the ways in which these three variables affect the import-gnp ratio. First of all, there are really only two variables which need to be considered, namely, total export receipts and net capital inflow. The third variable, the commodity terms of trade, is redundant because the import prices included in the index of the terms of trade have already been considered through the import prices relative to general price level. The export prices which are the other element of the index can be considered through the movements of total export sales. Then, the total export receipts affect the import-gnp ratio in two primary ways. First, exports are a part of the national product and, as explained in the thesis, the growth of GNP has a decreasing effect on the import-gnp ratio in the process of economic development. I have used the GNP as one of the explanatory variables, so that the direct effect of exports on the ratio in this way has already been reflected in the estimates of the coefficients of this explanatory variable.

3 The second primary effect of a growth of exports on the import-gnp ratio is through the supply J of foreign exchange. Other things being equal, it is certain that a sustained increase in export receipts will increase the supply of foreign exchange and lead to an increase in total capacity to import, resulting in an increase in the import-gnp ratio. Such an increase in exports is mainly determined by circumstances abroad, or exogenously, and is independent of the import-gnp ratio. In this sense, the growth of exports has an "export-pull" effect on the import-gnp ratio. Now turn to the effect of a net capital inflow. A net capital inflow has the same effect on the import GNP ratio through an increase in the supply of foreign exchange, provided that the measured capital inflow represents new investment by foreign residents or companies and not a reinvestment of profits accruing from past investment in Australia. As far as this effect is concerned, a net capital inflow increases the import-gnp ratio. However, the capital inflow, particularly of new direct foreign investment, has another (i. e., decreasing) effect on the ratio when this investment occurs in import-competing industries. In the first part of the second paragraph on page 45, I intended to stress the important role played by net capital inflow in reducing the import-gnp ratio from the pre- World War I period ( ) to the 1920's. What was meant on this page was that the fall of the

4 Liimport-GNP ratio from the pre-world War I period to the s was in part due to the effect of the war itself (whi ch caused unavailability of many import commodities and inspired domestic import substitution) but it was also due to the growth of domestic markets and the investment by foreign firms which began manufa cturing in Australia or expanded their operations during the s. By increasing the output of importcompeting goods, the net foreign capital inflow over this period tended to reduce the import-gnp ratio. The new direct foreign investment after World War II has had a similar effect on the import-gnp ratio. Thus the total supply of foreign exchange is determined by the total export receipts and capital inflow. The foreign exchange earnings so determined are available for imports of commodities and invisibles and net transfers out, although the latter two are relatively small items, and in such ways export receipts and capital inflow affect directly the import-gnp ratio. In writing my thesis I had in mind chiefly the twentieth century period of Australian history, although I presented some discussion in Chapters III and IV of the nineteenth century. As far as the twentieth century is concerned, the capital inflow has been significantly directed to the growth of import replacement industries and this effect of net capital inflow has offset the increasing effect on the import GNP ratio through the supply of foreign exchange. Nevertheless, I should have considered explicitly the role of exports and capital inflow in the chapters on the aggregate import-gnp ratio.

5 5 The thesis as a whole clearly suggests that the effect of net capital inflow (or new direct foreign investment) can be more explicitly analysed in the study of individual industries. The industry studies showed thatdirect foreign investment was very important in reducing the import shares in particular commodity market. An illuminating example was found in the case of sewing machine industry (see, in particular, pages , and 180). The examiners also found it difficult to understand my statement, "this decline (in the import- GNP ratio between the s and 19JO 's) has been qualitatively attributed mainly to tariffs and depression". In the oontext, the statement should be read as meaning "other people have said that... ' It is clear from the regression equations that tariffs are regarded as an exogenous variable and that their effects are assessed quantitatively, not qualitatively. As suggested by the examiners, it is reasonable to introduce into the analysis the rate of immigration as an exogenous variable, because it affects the growth '' of GNP. However, in Chapter II where the cross-country variation in the import-gnp ratio is studied, the size of nation as measured by GNP or population, as well as geographi cal area (or natural resources), are all tested as independent variables and found to be significant. The study shows that immigration certainly increases population and hence GNP, so that it decreases the import-gnp ratio. Similarly geographical area can affect the rate of growth of GNP through the development of natural resources.

6 6 2 Statistical Tests It is reported in the examiners' reports that; "The statistical tests that are used are also applied uncritically. A high regression coefficient is accepted as evidence of a causal relationship without further and more sophisticated statistical analysis. " This criticism is difficult for me to follow. By "a high regression coefficient", I think, the examiners mean "a high correlation coefficient," or a "high coefficient of determination". May I stress that I have never automatically accepted the regression equation with the highest correlation coefficient as the best estimate. The reader would also be able to find in the thesis statements such as that "every regression coefficient in the above equation is significant at the 1 per cent (or 5 per cent) level". This means that every regression coefficient has been critically examined for its significance by the t-test. Therefore I have not uncritically used any regression equation in the thesis. On this point it should be added that the A.N.U. provided one of the best programs which enabled me to apply the t-test, the F-test, the Durbin-Watson d test, etc. at the same time. Such a program is not available in the University where I work now.

7 7 Concerning the footnote on page 47 in which imports have been regressed on income and t-tests have been made, the examiners have required to report further statistical tests - for example the Durbin-Watson test. To test whether there are positive (or negative) autocorrelated disturbances, the following Durbin Watson ratios were computed for each period; D-W Ratios 1 ) For the period ( 3 1 ) ) For the period ( 1 3) ) For the 1920/21-'29/ 30 period ( 10) ) For the 1930/31-'39/ 40 period ( 10) ) For the 1946/ /68 period (22) where figures in brackets show the number of observations for the corresponding period. On the other hand the Durbin-Watson d statistic gives the following values of dl and du*; Significant Point No. of Observations dl and ~ : 1 % dl du oo Comparing our computed d values with values of dl and~ for corresponding observation numbers, first, it is clear that there is no negative autocorrelation. Nor, * Durbin and Watson do not tabulate dl and~ for the observation number less than 15, so that dl and~ for 15 observations are used here for 10 and 1j observations.

8 secondly, is there any positive autocorrelation except for the period of 19JO/J 1-1 J9/4o where the Durbin-Watson 8 test is inconclusive. ** It is concluded that the equations under consideration are statistically acceptable, or significant. J Evaluation of Sources I have stated clearly that there exist some inconsistencies in the data used. For example, "imports are valued in c. i. f. and, after 190J, including duties as well" (page J4) and "the commodity classification is not comparable between the periods before 1901 and after 19 1 J 11 (pa ge 7 1 ). This kind of inconsistency in data is so often, if not inevitably, associated with an analysis involving a very long time span. about some inconsistent data, I impossible for us to go further. If we a re too careful think, it will be As the examiners pointed out, however, it is important to g ive some discussion of the implication of such an inconsistency for the analysis. First, the import-gnp ratio at constant prices is used because, as the examiners rightly pointed out, we are concerned with the quantum of imports in relation to the quantum of national inc ome. In p a rticular the ** The Durbin-Watson test is conducted as follows: ( 1 ) If d < d (or 4-d < d ), reject the hypothesis of random dis~urbances in ~avour of that of positive (or negative) auto-correlation; (2) If d > ~(or 4-d > du), do not reject the hypothesis; (J) If d < d < d (or 4-d < d < 4-~), the test is inconclustve and flirther ob~ervations should ideally be obtained.

9 9 wellknown controversy over the "law of declining importance of foreign trade", i.e., the hypothesis that import-gnp ratio will decline as a nation develops, is usually stated in real terms. Therefore to make what I have done comparable with other people's works on the same subject required me to state the Australian import-gnp ratio in real terms. In this context relative price movements should be taken into account. One of the examiners is not convinced, however, that available price indices are sufficiently sensitive to reflect changes in relative prices and in the composition of imports and national product. This is again a problem which arises in historical data. I would agree that we should use the best available price indices, even though they are not entirely sufficient theoretically, if we are economists but not statisticians. Incidentally, I also presented the series of import-gnp ratio at current prices for the whole period, so that those who are in favour of this series can use them (see Column 5 in Table A3.1 on pages ). Secondly, as already mentioned above, my import series is not consistent, since there is a break in 1903 after which date duties collected areadded to imports c.i.f. The question arises: Does not this inconsistency account for most of the post-1900 rise in the import-gnp ratio? Let me estimate roughly what level of tariffs is required for the pre-federation period to reverse the trend in the import-gnp ratio as presented in the thesis. To do this, simply divide the import-gnp ratio for the period by the

10 import-gnp ratio (where imports are c. i.f. excluding 10 duties) for the pre-federation period. This gives In other words, if the average tariff for the pre- Federation period were higher than 17 per cent, the import-gnp ratio (where imports are c. i. f. including duties) for the pre-federation period would be higher than the import-gnp ratio for the period. This level of the average tariff would be equivalent to about 90 per cent of the average tariff level of the 1920/2 1-r24/25 period. Now the problem is whether or not we have any evidence to suggest that the tariff level for the pre Federation period is higher than 90 per cent of the level of the early 1920s. If there is, then Australia's import-gnp ratio declined constantly since the pre- Federation period; if not, it first rose and then declined after the period as stated in Chapter III. It seems to me very likely that the latter was the case. Thirdly, it is pointed out in the examiners' reports that "the main rise in the (import-gnp) ratio probably occurred between 18JO and instead of between 1870 and On this point, I need more careful study. But may I present the following hypothesis : In a country such as Australia which was then a colony of settlement the import-gnp ratio is l ikely to be nearly one hundred per cent at its beginning, then gradually to decline until the economy is organised "normally"; thereafter the ratio will tend to increase because of factors presented in

11 the thesis and it will then decline again but this decline is associated industrialisation or import 1 1 s u bstitu tion. The question is when the Australi an economy was organised "normally". Did it settle down by 1830? Or by 1860? Fourthly, I used the concept of average tariffs in my analysis of the effec t of protection on the import-gnp ratio. The defi c i e n c y of the measure of avera g e tariffs used is give n some c onsideration on pages Despit e the defi ciency of this me asure it was employed because of the difficulties of obtaining an alternative. A. T. Carmody 1 s weight e d tariff index is available as an alternative only for the / 40 period. As far as the p eriod / 40 (excluding 1914/ / 20) is c onc erned, the avera g e tariff index and Carmody 1 s index are c losely correlat e d as fo l lows : (S- 1 ) T = C, ( ) R 2 = , where Tis the avera g e tariff index I used and C Carmody 1 s, both using the base of 1920/ / 25. (See foo t not e on pa g e 52). This statistically significant rel ation betwe en the two indices must b e viewed in the light of the absence of a significant trend and in view of the considerabl e amplitude of the flu ctuations. Thus, as stated on page 52, the us e of the avera g e tariff i ndex as an indication of changes in tariffs over time is far less suspect than it would appear to be on purely theoretica l grounds.

12 12 It is true, as the examiners pointed out, that my average tariff index bears no relationship to the movement of Carmody's tariff index for the period /29 (excluding 1914/ /20) as shown in the following estimated equation: (s-2) T = C, (0.1125) R 2 = This is due to a smaller number of observations. The proper relationship between the two indices should be justified in the light of Equation (S-1) or the equation on page 52 which includes more observations. Finally, there is an inconsistency in commodity classification between the periods before 1901 and after 191J. This is again due to the available data. The examiners have suggested in their reports that I should at least have attempted to check these figures with my own break-down from the Overseas Trade Bulletin. At one stage of writing the thesis, I actually attempted to do so. But picking up items from Overseas Trade Bulletins according to the list of the classification borrowed from the Commonwealth Bureau of Statistics, I found that it was impossible for me to check item by item within the limited time available for me at that stage. I therefore, after all, gave up checking. Had I kept checkin g, I would not have been able to submit the thesis. On the treatment of motor vehicles as capital goods I make this comment. The item "motor vehicles" has been recorded under "producers' materials" since

13 1955 and prior to that year under "capital equipment" 13 (or capital goods). To maintain comparability this item is reclassified under "capital goods". ( See page 194). It is rather difficult to decide into which item "motor vehicles" should go. But whichever way this item is treated, the trends shown in Fig. J-1 are not affected considerably. Makoto Ikema August 1971