Short-Run Fluctuations in U.S. Imports of Raw Materials, 1928–39 and 1947–52

Gertrud Lovasy; H. K. Zassenhaus

doi:10.5089/9781451978872.024.A004

Jump to Content Jump to Main Navigation

Short-Run Fluctuations in U.S. Imports of Raw Materials, 1928–39 and 1947–52

Author: Ms. Gertrud Lovasy¹ and H. K. Zassenhaus

¹ 0000000404811396https://isni.org/isni/0000000404811396International Monetary Fund

Publication Date:: 01 Jan 1953

eISBN:: 9781451978872

Language:: English

Keywords:: SP; foreign company; country; raw material; sc; raw materials imports; company fall; consumption goods; stock-output ratio; Exchange restrictions; Imports; Income; Reserve assets; Consumption; Africa; Europe; import-output relation; raw materials output; raw materials production; Manufacturing

Download PDF (1.1 MB)

THE PURPOSE of this paper is to help to explain the movement of imports of raw materials into the United States. Though both the nature of the problem and the statistical material available make it impossible to offer more than a tentative contribution, the points analyzed below appear to throw some useful light upon the problem.

Abstract

Earlier studies¹ of the relevant annual statistical data have established two facts: (1) in the United States there is a close relationship between raw materials imports and domestic manufacturing production, with an elasticity between these two variables that is close to unity; (2) though the correlation between imports and domestic output can generally be improved by adding relative prices (i.e., import prices relative to domestic prices) as an explanatory variable, the elasticity of imports in respect to relative prices has been found to be small.²

It is clear, therefore, that variations in domestic output of manufacturing production—i.e., output which absorbs raw materials—are mainly responsible for variations in raw materials imports. This is true for quarterly as well as for annual variations. The treatment of possible supplementary explanatory price variables, however, runs into several difficulties. To the extent that import prices are independent, rather than dependent, variables in respect to import quantities (considering the volume of U.S. demand, this is far from certain for a number of import commodities), short-run price fluctuations are likely to induce speculative imports and then lead to inventory changes (in the same direction that prices change). Gradual, long-run price movements, on the other hand, may be expected to influence the “mix” of imported and domestic products in the volume of total raw materials absorbed by domestic production. This “mix” will affect imports unless it is offset by changes in the ratio of total raw materials imports to product output. But the short-term and long-term price effects will tend to counteract each other, and quarterly price and quantity data are unlikely to register them separately. From the point of view of testing whatever results there may be, however, quarterly observations are desirable.

Even a cursory examination of Charts 1 and 2 indicates that changes in domestic manufacturing production do not fully explain changes in the quantity of raw materials imports. Both in several interwar years and after 1947, the quantities of imports fluctuated more widely than domestic output. This suggests that, instead of some explicit price variable, one or both of two other explanatory variables should be tested in addition to domestic manufacturing output: domestic raw materials inventories, and domestic output of raw materials which may be considered competitive with imports. Since, unfortunately, no adequate inventory data are available for the years 1928–38, only the latter of these two variables can be tested; here it seems appropriate also to give consideration to a linear trend.

The relationships suggested between imports, inventories, and domestic output of competing materials may be thought of quite simply as follows. A significant rise of domestic manufacturing (i.e., raw materials absorbing) output will be made possible in the first instance by a corresponding increase in the domestic output of raw materials, provided that this increase is not checked by the fact that domestic raw materials output is already close to capacity; the absolute increase of imports of raw materials will be less than the increase in domestic output of such materials, and it possibly will occur later. In conditions of high domestic employment and near-capacity production, however, imports are likely to be drawn on to a greater extent, and possibly earlier, than domestic raw materials supplies. The upward movement of (real) imports will then be proportionally larger than that of domestic import-absorbing output. On the other hand, a drop in domestic manufacturing production is likely to lead to a reduction of imports rather than of domestic raw materials output, especially under conditions of unemployment and of low domestic output relative to capacity. Here again, imports will show a proportionally wider variation than domestic manufacturing production.

At the same time, upward (downward) movements of domestic manufacturing production will, unless foreseen by domestic producers, almost immediately lead to a decline (rise) in the ratio between manufacturers’ raw materials stocks and their current output of finished products. Raw materials imports are likely to be stimulated as soon as a fall in the stock-output ratio begins to carry the ratio below what domestic manufacturing producers may consider a “normal” level. Similarly, imports may be depressed by a rise in the stock-output ratio unless the increase is insufficient to raise the stock ratio above “normal”. Thus, the stock-output ratio could be expected to move inversely to imports in a succeeding period, and this inventory mechanism could exaggerate import movements compared with those of domestic manufacturing production.

Several points should be noted by way of qualification of this mechanism. First, to the extent that the movements of both the manufacturers’ stock-output ratio and the domestic competing raw materials output reflect the effects of changes in (relative or absolute) import prices, these changes are in effect implicitly allowed for in the explanation attempted here. This—apart from statistical difficulties involved in computing an adequate price index corrected for import duties—diminishes the force of a possible objection to the procedure suggested here, on the ground that price effects have been neglected. Second, raw materials imports are subject to long-run influences—such as the general improvement of efficiency in the use of raw materials resulting from technical advances. Especially in the interwar years, these factors appear to have caused the declining trend of imports when compared with domestic manufacturing production. This trend has been explicitly allowed for in the explanation attempted here and will be discussed further below. Third, no attempt has been made here to separate the influence of tariff changes. Except for a few raw materials (wool, copper), this influence was probably minor. It could be investigated with some hope of useful results only by examining individual commodities separately. The same applies to the effect of relative price movements on the substitution of imported for domestic raw materials. Finally, since quarterly data are used throughout this study, the question of possible seasonal variations in imports arises. For the immediate postwar years, but not for 1928-38, the existence of such seasonal variations can be established.

Statistical Material

The hypotheses described above have been applied to the time series for the interwar and postwar periods, shown in Charts 1 and 2 and in Tables 1 and 2.

^{Table 1.}

Indices of the Volume of U.S. Imports of Crude and Semimanufactured Materials (M), Domestic Output of Manufactured Goods (P), and Domestic Output of Import-Competing Crude and Semimanufactured Materials (C), Quarterly, 1928–39¹

(1928–35 = 100)

^¹For description of series, see text, section on Statistical Material.

^{Table 1.}

(1928–35 = 100)

		M	P	C	M/P	C/P
1928	1	128.1	112	110	114.4	98.2
	2	117.8	119	112	99.0	94.1
	3	112.3	121	115	92.8	95.0
	4	116.4	122	121	95.4	99.2
1929	1	140.5	129	119	108.9	92.2
	2	141.4	141	124	100.3	87.9
	3	131.9	135	123	97.7	91.1
	4	126.9	126	125	100.7	99.2
1930	1	119.6	118	112	101.4	94.9
	2	112.7	119	111	94.7	93.3
	3	100.9	103	101	98.0	98.1
	4	107.1	95	101	112.7	106.3
1931	1	101.9	94	92	108.4	97.9
	2	95.0	100	94	95.0	94.0
	3	97.1	88	92	110.3	104.5
	4	102.1	77	94	132.6	122.1
1932	1	84.2	74	82	113.8	110.8
	2	76.8	69	78	111.3	113.0
	3	65.6	64	75	102.5	117.2
	4	75.1	69	80	108.8	115.9
1933	1	67.5	64	80	105.5	125.0
	2	80.6	79	89	102.0	112.6
	3	112.0	94	98	119.1	104.3
	4	87.6	83	93	105.5	112.0
1934	1	82.9	89	90	93.1	101.1
	2	83.9	97	92	86.5	94.8
	3	75.0	89	91	84.3	102.2
	4	71.8	88	93	81.6	105.7
1935	1	94.2	104	96	90.6	92.3
	2	94.3	107	97	88.1	90.7
	3	100.8	110	105	91.6	95.5
	4	103.4	118	116	87.6	98.3
1936	1	105.8	115	111	92.0	96.5
	2	104.4	127	115	82.2	90.6
	3	110.1	128	120	86.0	93.8
	4	113.4	140	127	81.0	90.7
1937	1	132.7	143	128	92.8	89.5
	2	130.7	152	138	86.0	90.8
	3	118.0	141	140	83.7	99.3
	4	106.4	123	137	86.5	111.4
1938	1	82.8	102	125	81.2	122.5
	2	73.7	101	116	73.0	114.8
	3	88.4	108	123	81.9	113.9
	4	92.5	122	131	75.8	107.4
1939	1	95.2	123	129	77.4	104.9
	2	97.0	126	131	77.0	104.0

^¹For description of series, see text, section on Statistical Material.

^{Table 1.}

(1928–35 = 100)

		M	P	C	M/P	C/P
1928	1	128.1	112	110	114.4	98.2
	2	117.8	119	112	99.0	94.1
	3	112.3	121	115	92.8	95.0
	4	116.4	122	121	95.4	99.2
1929	1	140.5	129	119	108.9	92.2
	2	141.4	141	124	100.3	87.9
	3	131.9	135	123	97.7	91.1
	4	126.9	126	125	100.7	99.2
1930	1	119.6	118	112	101.4	94.9
	2	112.7	119	111	94.7	93.3
	3	100.9	103	101	98.0	98.1
	4	107.1	95	101	112.7	106.3
1931	1	101.9	94	92	108.4	97.9
	2	95.0	100	94	95.0	94.0
	3	97.1	88	92	110.3	104.5
	4	102.1	77	94	132.6	122.1
1932	1	84.2	74	82	113.8	110.8
	2	76.8	69	78	111.3	113.0
	3	65.6	64	75	102.5	117.2
	4	75.1	69	80	108.8	115.9
1933	1	67.5	64	80	105.5	125.0
	2	80.6	79	89	102.0	112.6
	3	112.0	94	98	119.1	104.3
	4	87.6	83	93	105.5	112.0
1934	1	82.9	89	90	93.1	101.1
	2	83.9	97	92	86.5	94.8
	3	75.0	89	91	84.3	102.2
	4	71.8	88	93	81.6	105.7
1935	1	94.2	104	96	90.6	92.3
	2	94.3	107	97	88.1	90.7
	3	100.8	110	105	91.6	95.5
	4	103.4	118	116	87.6	98.3
1936	1	105.8	115	111	92.0	96.5
	2	104.4	127	115	82.2	90.6
	3	110.1	128	120	86.0	93.8
	4	113.4	140	127	81.0	90.7
1937	1	132.7	143	128	92.8	89.5
	2	130.7	152	138	86.0	90.8
	3	118.0	141	140	83.7	99.3
	4	106.4	123	137	86.5	111.4
1938	1	82.8	102	125	81.2	122.5
	2	73.7	101	116	73.0	114.8
	3	88.4	108	123	81.9	113.9
	4	92.5	122	131	75.8	107.4
1939	1	95.2	123	129	77.4	104.9
	2	97.0	126	131	77.0	104.0

^¹For description of series, see text, section on Statistical Material.

^{Table 2.}

Indices of the Volume of U.S. Imports of Crude and Semimanufactured Materials (M), Domestic Output of Manufactured Goods (P), Domestic Output of Import-Competing Crude and Semimanufactured Materials (C), and Manufacturers– Stocks of Purchased Materials (S), Quarterly, 1947–52^¹

(1948 = 100)

^¹For description of series, see text, section on Statistical Material.

^{Table 2.}

(1948 = 100)

		M	P	C	S	S/P	M/P	C/P
1947	1	93.8	98	91	98.8	101	96	93
	2	91.6	97	92	98.0	101	95	95
	3	85.6	95	93	96.5	102	91	98
	4	95.7	100	97	101.6	102	96	97
1948	1	107.8	100	98	98.3	98	108	98
	2	94.7	100	99	100.1	100	95	99
	3	99.3	99	100	100.4	101	100	101
	4	98.3	101	103	101.2	100	97	102
1949	1	93.0	96	96	96.4	100	97	100
	2	85.6	89	90	89.4	100	97	101
	3	84.9	89	89	87.5	98	96	100
	4	104.0	93	98	88.6	95	112	105
1950	1	116.5	95	95	86.1	91	123	100
	2	118.3	102	100	85.6	84	116	98
	3	127.6	108	107	93.5	87	118	99
	4	136.9	115	112	106.7	93	119	98
1951	1	130.9	116	112	110.6	95	113	97
	2	119.5	118	116	115.0	97	102	99
	3	108.1	115	114	114.5	100	94	99
	4	102.8	114	117.5	121.1	106	90	103
1952	1	118.0	116	115	117.5	101	102	99
	2	107.9	116	108	114.6	99	94	94

^¹For description of series, see text, section on Statistical Material.

^{Table 2.}

(1948 = 100)

		M	P	C	S	S/P	M/P	C/P
1947	1	93.8	98	91	98.8	101	96	93
	2	91.6	97	92	98.0	101	95	95
	3	85.6	95	93	96.5	102	91	98
	4	95.7	100	97	101.6	102	96	97
1948	1	107.8	100	98	98.3	98	108	98
	2	94.7	100	99	100.1	100	95	99
	3	99.3	99	100	100.4	101	100	101
	4	98.3	101	103	101.2	100	97	102
1949	1	93.0	96	96	96.4	100	97	100
	2	85.6	89	90	89.4	100	97	101
	3	84.9	89	89	87.5	98	96	100
	4	104.0	93	98	88.6	95	112	105
1950	1	116.5	95	95	86.1	91	123	100
	2	118.3	102	100	85.6	84	116	98
	3	127.6	108	107	93.5	87	118	99
	4	136.9	115	112	106.7	93	119	98
1951	1	130.9	116	112	110.6	95	113	97
	2	119.5	118	116	115.0	97	102	99
	3	108.1	115	114	114.5	100	94	99
	4	102.8	114	117.5	121.1	106	90	103
1952	1	118.0	116	115	117.5	101	102	99
	2	107.9	116	108	114.6	99	94	94

^¹For description of series, see text, section on Statistical Material.

Imports (M)

The import series is the sum of recorded imports for consumption of crude materials and semimanufactures as defined by the U.S. Department of Commerce. To obtain import quantities, current value imports have been deflated by the corresponding unit value indices, as officially published, and reduced to an index with the quarterly average for 1928–35 as the base for the 1928–39 data, and the quarterly average for 1948 as the base for the 1947–52 data. The series is not seasonally adjusted.

Domestic production of manufactures (P)

The index of manufacturing production of the Board of Governors of the Federal Reserve System was first adjusted to exclude three component series which were thought to attract no, or very few, nonfood raw materials imports, and then calculated for the two periods on the basis of the quarterly averages for 1928–35 and 1948, respectively.

The excluded series are iron and steel, cotton consumption, and manufactured food products.

Domestic output of nonfood raw materials which compete with imports of nonfood raw materials (C)

This index is based on the output and prices of 12 commodities.³ The quantity produced of each commodity was valued at the 1937 prices, or at the prices for the first production date after 1937. An index of the sum of the values at 1937 prices was then calculated, with the quarterly averages for 1928-35 and 1948 as bases for the prewar and postwar periods, respectively.⁴

Inventories (S) (1947–52 only)

Manufacturers’ inventories of purchased materials, as reported by the U.S. Department of Commerce for the end of each quarter, have been adjusted for price variations. Consultation with the Office of Business Economics of the Department of Commerce showed that a reasonably reliable series of inventories in real terms can be obtained only by deflating the inventories of each industry separately and then aggregating the results. This is so because inventory valuation practice varies from industry to industry. Such a procedure would involve considerable labor, however, since the corresponding wholesale price indices for each industry, which ought to be used as deflators, would have to be reweighted in proportion to the importance of the various materials in the inventories concerned. Moreover, such data as have been collected on purchased inventories, by industry, appear to be the results of imperfect sampling of the industries concerned and have therefore not been released.

Therefore, an approximation to a reliable inventory deflation method was adopted. It assumes that, on the average, one third of the reported (book value) inventories of purchased materials were valued at constant prices (i.e., on the last in, first out method). The remaining two thirds are assumed to have been valued at prices equal to the average wholesale prices of the five months preceding the inventory reporting date. Accordingly, one third of the reported purchased materials inventories have not been adjusted for price changes, and two thirds have been divided by the average total wholesale price index⁵ for the five preceding months.

Tests for individual industries (textiles) showed that this deflation method yielded significantly different results from the use of current wholesale prices.

Results of Statistical Tests

The explanatory hypotheses described above have been tested by computing a number of regressions. Since the available statistical material is more adequate for postwar years, the results for interwar (1928–38) and postwar (1947–52) periods are discussed separately.

Period 1928–88

Imports and domestic manufacturing production. As might have been expected, the correlation between imports and domestic manufacturing production that is based on quarterly data is not so high as that based on annual data. For annual data, a regression of the form M_t = aP_t + c gave a correlation coefficient in excess of .90, while the present study yields a coefficient of .81. Practically no improvement resulted from the use of a regression of the form M_t = aP_t bP_t–1 c; the regression coefficient of P_t–1 turned out to be clearly insignificant.

Linear trend and domestic competing output of raw materials. The correlation of the forms shown above was greatly improved by introducing a linear trend (T). The resultant regression was as follows:⁶

\begin{matrix} M_{t}^{*} = \underset{(.047)}{.759 P_{t}} - \underset{(.083)}{.679 T} + 34.708; & R = .929 & (1) \end{matrix}

A strong declining trend thus becomes clearly visible. The components of this regression are shown in Chart 3. The introduction of the trend results in a very acceptable fit of the regression to the original data. The residuals (M—M^*) show no systematic pattern, especially no noticeable difference in fit between the years previous to the Smoot-Hawley tariff and those following it. The large swing of (M—M*) in 1933–34 (which also shows up prominently in Chart 1) is apparently due to the dollar devaluation in those years.

That there is no significant lag of M behind P is again shown when P_t–1 is admitted to the regression. The resulting equation is:

\begin{matrix} M_{t}^{*} = \underset{(.119)}{.719 P} + \underset{(.119)}{.044 P_{t - 1}} - \underset{(.083)}{.677 T} + 34.286; & R = .930 & (1 a) \end{matrix}

The coefficient of P_t–1 is clearly insignificant.

A comparison of Charts 1 and 3 suggests that the trend (T) can, at least to some extent, be accounted for by the movement of the domestic output of import-competing raw materials (C): in the years 1928–32, the similarity of movement between C and P is less close than that between M and P, but in 1933–38, the opposite is true. The burden of adjusting raw materials input to P variations was clearly carried predominantly by M in the earlier years, and by C in the later years. Thus, C was “underadjusted” in 1928–32, and M was “overadjusted” in 1933–38. In other words, M varies throughout by at least as much as P (and by more in the later years), while C varies no more than P (and less in the earlier years).

In order to test the role of C, C has been substituted for T in the M-regression. This gives the following two equations:

\begin{matrix} M_{t}^{*} = \underset{(.166)}{1.087 P_{t}} - \underset{(.213)}{.557 C} + 44.086; & R = .840 & (2) \end{matrix}

\begin{matrix} M_{t}^{*} = \underset{(.191)}{.894 P_{t}} + \underset{(.184)}{.339 P_{t - 1}} - \underset{(.223)}{.721 C} + 46.138; & R = .853 & (2 a) \end{matrix}

It is thus seen that the substitution of C for the trend yields considerably less satisfactory results: the movement of import-competing domestic raw materials production explains, at best, only part of the trend found earlier. The correlation coefficients of (2) and (2a) are much smaller than those of (1) and (la); and the regression coefficients for C are less significant than those of T. T must therefore be accounted for, in part, by factors other than C, factors such as the trend in technical import-output relations referred to above.

The same conclusion may be drawn from a regression of C on P and a trend T. It yields a very high correlation coefficient:

\begin{matrix} C * = \underset{(.042)}{.684 P} + \underset{(.074)}{.291 T} + 27.86; & R = .935 & (3) \end{matrix}

The simple correlation between C and P is in fact higher than that between M and P (r_c.p = .912; r_M.P = .813), while C shows a positive trend, compared with the negative one for M. This suggests that in 1928–38 M was gradually replaced by C. While the covariation of C with P was stronger than that of M, a given change in P produced a proportionally larger change in M than in C. This is shown by the fact that the (partial) elasticity of M with respect to P, as computed from the regression (1) above, is larger than the elasticity of C with respect to P. The two elasticities obtained are as follows:

\begin{matrix} E_{M . P} = 0.81; & E_{C . P} = 0.68 & (4) \end{matrix}

This may be taken to measure the “marginality” of raw materials imports, if the explanation just given of the relationships between M, P, and C is kept in mind. The P-elasticity of imports is less than unity (when the effect of T, and thus of C, is neglected); domestic raw materials input appears to be considerably less affected than imported input by a variation of domestic manufacturing production. When allowance is made for the effect of the falling trend on the M-P relationship, M must be expected to show, in response to a recession of P, a (downward) variation proportionally larger not only than that of C, but also than the (upward) M variation which would occur in response to an equal but upward movement of P; M suffers more from a fall in P than it benefits from a rise in P.⁷ The converse is true for C, though less emphatically so, since both the P-elasticity of C and C’s trend are quantitatively smaller.

Period 1947–52

For the postwar years (for which adequate inventory observations are available), there was, in fact, no noticeable trend for M; none could have been expected in view of the small number of years and the disturbing influence on imports—at least through 1948—of supply difficulties. The P-effect on M apart, inventory variations appear to have been pronounced enough to outweigh the influence on imports of C, except perhaps for the latter half of the postwar period.

General results. In accordance with the hypotheses described earlier, quarterly imports were correlated with domestic manufacturing production of the same quarter and the stock-output ratio in domestic manufacturing of the preceding quarter. The main regression obtained is as follows:

\begin{matrix} M_{t}^{*} = \underset{(.149)}{.938 P_{t}} - \underset{(.276)}{1.487} {(S / P)}_{t - 1} + 153.873; & R = .903 & (5) \end{matrix}

The components of this regression are shown in Chart 4. Again, though the correlation coefficient is somewhat smaller than that for the interwar years, the fit to the original data is satisfactory. There is, however, room in the (M—M^*) residuals for the operation of an additional systematic influence, and it will be seen later that these residuals contain what appears to be a significant seasonal component.

Inventories and imports. Prima facie, the form which the influence of inventory changes on imports takes is debatable. There are two major possibilities. One is that changes in inventories of raw materials are, by and large, “voluntary”; that is, the inventory level at any time is due predominantly to advance planning by manufacturing producers in anticipation of finished output changes and price movements but independent of the current level of output. Therefore, the inventory level is, at any given time, in accord with manufacturers’ desires at that time, except for its imported part which lags behind its domestic part. In this case, inventory movements are also autonomous relative to domestic raw materials production, and the sequence of changes would run from inventories to production as well as to imports of raw materials. A positive (though lagged) correlation would be expected between inventory changes (ΔS) and import residuals which are not explained by the movements of domestic manufacturing production.⁸

Alternatively, changes in inventories of raw materials are assumed to be predominantly “involuntary”; that is, the level of inventories at any time is typically not in accord with manufacturers’ desires at that time. It is reasonable to assume that manufacturers have a reasonably clearly defined notion of a “normal” level of inventories of raw materials relative to current production. When—as a result, say, of an unexpected rise in demand—the volume of finished goods production increases, raw materials for the enlarged output will, in the first instance, have to come from existing raw materials inventories. Their depletion, together with the higher product output, reduces them to a level which is deficient when compared with their desired “normal” level. Increased raw materials requirements are met from both domestic and imported sources. But imports can be adjusted only with some time lag (larger, at any rate, than whatever lag there may be in the similar adjustment of domestic raw materials output). Hence, if inventory changes are “involuntary”, a negative correlation would be expected between current imports and the immediately preceding ratio of inventories to the volume of raw materials absorbing production (i.e., S/P).

This “involuntary” inventory hypothesis for the relationship between raw materials imports and stocks is clearly no more than a very much abbreviated summary of a more complicated mechanism.⁹ It does, however, lead to relatively simple relationships, and the statistical test, as presented by equation (5), gives a satisfactory result. The improvement of the correlation of the form (5) over a simple correlation of M on P is, as in the case of the interwar data, substantial. For the latter, the correlation coefficient was only .749, against .903 for regression (5).

The M-residuals which are still unexplained by this relation are relatively small (see Chart 4). Moreover, they appear to be largely seasonally determined.¹⁰ The exception is 1949, explainable probably by the import-retarding effect of expectations of European devaluations. A fairly complete explanation of M thus appears to be given by three factors: current domestic manufacturing production, preceding quarter inventory-output ratio, and seasonal changes.

The effects of voluntary inventory changes (resulting from price expectations and government foreign purchasing—stockpiling) and of domestic import-competing raw materials output are scarcely, if at all, observable.

Import-competing domestic raw materials output and imports. The preceding analysis has carried the explanation of imports so far that little room is left for the operation of C. In fact, if C is included explicitly in the M-regression, the result is quite unsatisfactory:

\begin{matrix} M * = \underset{(.489)}{.887 P_{t}} - \underset{(.275)}{1.486} (S / P)_{t - 1} + \underset{(.521)}{.0.57 C} + 153.276; & R = .904 & (5 a) \end{matrix}

The correlation of equation (5) is thus not improved by the inclusion of C. The regression coefficient of C is clearly insignificant and has the wrong sign for a C-M replacement hypothesis.

The main reason for this inconclusive result is the high correlation between C and P, as shown in the following equation:¹¹

\begin{matrix} C * = .894 P - .005 (S / P)_{t - 1} + 10.42; & R = .956 & (6) \end{matrix}

At the same time, this equation yields a (partial) P-elasticity of C of 0.90—a value which is only insignificantly smaller than the (partial) P-elasticity of M, as computed from equation (5), of 0.92.

These results seem to indicate two conclusions. First, during the postwar period, the domestic output of import-competing raw materials made no noticeable inroads on the input share of imported materials; there is no evidence for a significant substitution of this sort. Second, compared with prewar years, domestic output of import-competing materials was able to keep in step with manufacturing production considerably better. The significant comparison is that between the movements of C and M relative to P with those before the war, rather than that between C and M after the war. Production capacity for C appears to have been built up during the war to an extent sufficient for C to play a much stronger part relative to M than before the war.

These considerations suggest a general remark on the differences between the two elasticities in the prewar and the postwar years. The higher P-elasticity of M for 1947–52 (0.92 instead of the prewar 0.81) seems to be explained by the fact that in 1947–52 account was taken of the inventory effect, which had to be neglected for 1928–38. This appears to be additional evidence for the significance of inventory movements. The difference between the two elasticities for C (prewar, 0.68; postwar, 0.90) probably may be explained by the advances made in providing production facilities for C, as just mentioned. While, therefore, no C-M substitution effect is visible for 1947–52, it does appear when a comparison is made between the postwar and the prewar periods.

Apart from the probability that the C-M substitution is a long-run rather than a short-run phenomenon, its record during the postwar years is likely to have been obscured by the series of abnormal events from the immediate prewar years through the devaluations of 1949 to the outbreak of the Korean war. The shift of U.S. demand toward wool, the synthetic rubber admixture regulations, the rapid relative exhaustion of domestic oil supplies, and the over-all fact that the response of domestic substitutes to increasingly tight domestic supplies of several nonferrous metals as well as of natural fibers was still at an experimental stage, may explain the C and M record. On the other hand, there is increasing evidence that the Korean war has provided the final impetus to the adoption of a series of domestic substitutes (rubber, aluminum, textiles) whose role in C had, in 1952, just begun to make itself felt.

All this, of course, does not mean that there has been no replacement of imported by domestic raw materials in the postwar period as a whole, in comparison with 1928-38. On the contrary, a comparison of the levels of M, P, and C in two base periods—1948 for the postwar period and 1928–35 for the interwar period—shows that, while manufacturing production approximately doubled, output of import-competing domestic raw materials increased by about 170 per cent, and raw materials imports by only about 30 per cent. World War II gave a strong impetus to this substitution.

Appendix: “Voluntary” Inventory Changes and Imports

The hypothesis that voluntary inventory movements (in the sense defined above) largely explain postwar short-run (quarterly) import movements has been set forth by A. C. Harberger in an unpublished paper. Statistical evidence for 1946 and the following years was produced to support this hypothesis. Quarterly changes in real inventories were compared with quarterly deviations of real imports either from a linear trend (for total imports and imports of semimanufactures) or from the quarterly average for the period of years covered (crude materials). All series were unadjusted for seasonal variations. Wholesalers’ and manufacturers’ inventories of purchased materials, deflated by current wholesale prices and their raw material component, respectively, were used for the comparison with imports. Manufacturers’ inventories of purchased materials, deflated in the same way, were used for the comparisons with imports of semimanufactured and crude materials. The exclusion of inventories of goods in process and of finished products is justified on the ground that the former, though voluntary, are not autonomous but depend closely on the level of output rather than on speculative anticipations, while finished product inventories are equally little sensitive to these anticipations. The emphasis is thus even more on autonomous inventory changes than on voluntary changes. Harberger admits the possible importance of involuntary inventories, but only insofar as they occur in the case of finished products (as a result of final demand fluctuations) and lead to voluntary adjustments of wholesalers’ and manufacturers’ inventories of raw materials.

When charted, these series show an almost constant lead of inventory changes over the import deviations of one quarter (Chart 5).¹² This was regarded as significant evidence in support of the hypothesis that voluntary inventory movements were an important part of the explanation of postwar short-run import movements.

Technically, a first objection to this procedure may be that the elimination from imports of a trend (or average) is meaningless unless the purpose is to eliminate the influence of such basic factors as domestic manufacturing production.¹³ If, in order to allow more adequately for the influence of P on M, the procedure is repeated, this time plotting the quarterly changes of stocks (ΔS) against the deviations of imports (M—M^*) from an import-output regression $(M_{t}^{*} = 1.152 P_{t} - 13.496)$ , the lead-lag pattern is seen to hold for 1947–48, and for the three quarters beginning with the third quarter of 1951. For the intervening period, when the influence of inventories might have been expected to have been especially strong, the pattern is not visible (Chart 6). At best, therefore, the statistical evidence merely supports the view that the movement of imports in the two subperiods requires different explanations, a “voluntary” inventory movement pattern in the one, and an “involuntary” pattern in the other. From the movement of the (M—M*) residual in Chart 4, however, it may be seen that the (“involuntary”) pattern suggested in the text fits the whole 1947–52 period uniformly well, and, therefore, on purely statistical grounds this hypothesis seems preferable.

Harberger’s hypothesis rests on the assumption that the change of raw materials inventories, at a given time, is in accord with entrepreneurs’ plans at that time—except for the imported part of the inventories. Since the time needed to make a change in import plans effective (i.e., the lag between import orders and import receipts) must be known to entrepreneurs, there is really no reason why the imported part of inventory changes should form an exception to his “voluntary” pattern. On his assumption, therefore, the lag which has been observed for at least part of the 1947–52 period would not be expected. It would seem, therefore, that the alternative “involuntary” pattern here suggested also has the advantage of consistency.

Miss Lovasy, economist in the Special Studies Division, Research Department, was formerly with the Economic and Financial Department of the League of Nations. She is the author of a memorandum on “International Cartels” published by the United Nations, and of several articles in economic journals.

Mr. Zassenhaus, economist in the British Commonwealth Division, European Department, is a graduate of the University of Bonn and the University of Bern. He was formerly Professor of Economics at Colgate University and research associate at the Twentieth Century Fund.

The latest is J. H. Adler, E. R. Schlesinger, and E. von Westerborg, The Pattern of United States Import Trade Since 1928: Some New Index Series and Their Application (Federal Reserve Bank of New York, 1952).

This is in contrast to the findings for U.S. imports of finished manufactures, which indicate that elasticities with respect to appropriate price variables are larger than unity.

The commodities and prices used in constructing this index are as follows:

Commodity	Price
Aluminum (primary production)	$0.2008 per lb.
Copper (mine production)	0.13 per lb.
Lead (receipts by smelters, domestic ore)	0.0601 per lb.
Zinc (mine production)	0.0690 per lb.
Crude petroleum (total production)	1.39 per barrel
Residual fuel oil (total production)	1.03 per barrel
Vegetable oils (total production, crude)	0.10 per lb.
Lumber (total production)	24.25 per 1,000 board feet
Woodpulp (total production, all grades)	60.00 per short ton
Rayon (domestic production, all types)	0.62 per lb.
Wool (total production, greasy basis)	0.437 per lb.
Synthetic rubber (domestic production, all types)	0.185 per lb. (for postwar years only)

Commodity	Price
Aluminum (primary production)	$0.2008 per lb.
Copper (mine production)	0.13 per lb.
Lead (receipts by smelters, domestic ore)	0.0601 per lb.
Zinc (mine production)	0.0690 per lb.
Crude petroleum (total production)	1.39 per barrel
Residual fuel oil (total production)	1.03 per barrel
Vegetable oils (total production, crude)	0.10 per lb.
Lumber (total production)	24.25 per 1,000 board feet
Woodpulp (total production, all grades)	60.00 per short ton
Rayon (domestic production, all types)	0.62 per lb.
Wool (total production, greasy basis)	0.437 per lb.
Synthetic rubber (domestic production, all types)	0.185 per lb. (for postwar years only)

Data are from International Monetary Fund, International Financial Statistics, publications of the American Bureau of Metal Statistics, and U.S. Bureau of the Census, Statistical Abstract of the United States.

In order to test whether the use of 1937 prices in the construction of this index for the years 1947–52 was defensible, the same series was computed with postwar prices. No significant difference from the series with 1937 prices resulted

A test showed that the use of wholesale prices of nonfarm products alone did not yield significantly different results.

In this and subsequent equations, the figures given in parentheses under the regression coefficients are the standard errors of the respective coefficients.

In the recession of 1937–38, this effect was strong enough to lead to a reduction, between peak and trough of the respective series, of the average import-input in P— i.e., M/P—by almost 25 per cent, in response to the reduction of P by about one third. In the recovery from 1932 to 1937, however, M/P actually fell by 10 per cent (see Table 1).

This hypothesis has been elaborated in an unpublished paper by A. C. Harbergen and is further discussed below, in the Appendix.

A full test of the latter is not attempted here. It would, however, be necessary to make such a test if the results presented later in this paper should prove acceptable. To test the postulated sequence between such statistical evidence of demand for finished output as manufacturers’ new orders and the corresponding output should not pose any real difficulty. That producers’ anticipations are typically in error, at least to the extent of throwing (imported) inventories off “normal”, is not implausible, certainly not for such years as 1950–52, but clearly this hypothesis needs more detailed examination. And finally, there is reason to believe that the “norm” itself changed when the composition of output changed toward considerably larger military, compared with civilian, end-use components in 1951–52. After a change of this kind, it need not be assumed that a higher S/P ratio would discourage imports. To this extent, and perhaps also because of speculative imports in the two years, there were voluntary inventory changes which induced import movements in the same direction. These last two points can be dealt with, with some hope of success, only by examining the relationship between M, P, and S, either for individual raw materials or/and for individual industries—an undertaking which is beyond the limited scope of this paper.

By a test of the homogeneity of the quarterly means (means of residuals for the first, second, etc., quarter) on the basis of the F distribution, the assumption of the absence of seasonality is rejected at the 5 per cent significance level.

Since the coefficient of (S/P)_t–1 is here obviously insignificant, this regression shows also that C—in contradistinction to M—does not lag behind the stock-output ratio, a result which gives some statistical support to what was above postulated as likely: Domestic output of import-competing raw materials is adjusted to changes in the raw materials stock ratio without the lag found for the import component of raw materials input.

The series originally prepared by Harberger have been recomputed here to take account of the U.S. Department of Commerce revised inventory and wholesale price series.

Except for the period from the first quarter of 1950 through the second quarter of 1951, Harberger’s results still hold when the first differences of inventories, deflated as in the text above, are plotted against the deviations of real imports of semimanufactured and crude materials from a linear trend. During the first six quarters of the Korean war, the statistical support, at least for his explanation, fails. Harberger was aware of this objection, and adopted the relatively simple procedure of eliminating a linear trend from the import side—to allow for the growing supply in the rest of the world of exportable materials—as a computational short cut rather than on analytical grounds.

Other Publishers

Save
Cite
Email this content

Share Link

Copy this link, or click below to email it to a friend
Email this content
or copy the link directly:

https://www.elibrary.imf.org/view/journals/024/1953/002/article-A004-en.xml
The link was not copied. Your current browser may not support copying via this button.

Link copied successfully

IMF Staff papers: Volume 3 No. 2

Author: International Monetary Fund. Research Dept.

Volume/Issue:: Volume 1953: Issue 002

Publisher:: International Monetary Fund

ISBN:: 9781451978872

ISSN:: 1020-7635

Pages:: 140

DOI:: https://doi.org/10.5089/9781451978872.024

Keywords
Statistical Material
Results of Statistical Tests

View raw image

(1928–35 = 100)

View raw image

Indices of the Volume of U.S. Imports of Crude and Semimanufactured Materials (M), Domestic Output of Manufactured Goods (P), Manufacturers’ Stocks of Purchased Materials (S), and Domestic Output of Import-Competing Crude and Semimanufactured Materials (C), Quarterly, 1947–52¹

(1948 = 100)

View raw image

Indices of the Volume of U.S. Imports of Crude and Semimanufactured Materials, Quarterly, 1928–39¹

(1928–35 = 100)

View raw image

Indices of the Volume of U.S. Imports of Crude and Semimanufactured Materials, Quarterly, 1947–52¹

(1948 = 100)

View raw image

U.S. Inventory and Import Cycles¹

(In millions of 1926 U.S. dollars)

View raw image

U.S. Inventory Changes (ΔS) and Deviations of Imports from Import Regression on Output (M—M^*)¹

Commodity	Price
Aluminum (primary production)	$0.2008 per lb.
Copper (mine production)	0.13 per lb.
Lead (receipts by smelters, domestic ore)	0.0601 per lb.
Zinc (mine production)	0.0690 per lb.
Crude petroleum (total production)	1.39 per barrel
Residual fuel oil (total production)	1.03 per barrel
Vegetable oils (total production, crude)	0.10 per lb.
Lumber (total production)	24.25 per 1,000 board feet
Woodpulp (total production, all grades)	60.00 per short ton
Rayon (domestic production, all types)	0.62 per lb.
Wool (total production, greasy basis)	0.437 per lb.
Synthetic rubber (domestic production, all types)	0.185 per lb. (for postwar years only)

Topics

Countries

Series

About

Resources

Abstract

Statistical Material

Imports (M)

Domestic production of manufactures (P)

Domestic output of nonfood raw materials which compete with imports of nonfood raw materials (C)

Inventories (S) (1947–52 only)

Results of Statistical Tests

Period 1928–88

Period 1947–52

Appendix: “Voluntary” Inventory Changes and Imports

Same Series

Other IMF Content

Other Publishers

Asian Development Bank

Inter-American Development Bank

Nordic Council of Ministers

The World Bank

Share Link

Table of Contents

Headings

Figures

Metrics