RICHARD M. GOODWIN World trade multipliers The world matrix multiplier Keynes gave a clear and convincing analysis of how to ameliorate the endemic tendency of capitalist economies to engender periods of high unemployment. Partly as a result of his influence, postwar history has been one of unparalleled and nearly continuous growth in output and employment. For the past four or five years, how- ever, this trend has been broken: everywhere there is unemploy- ment and varying degrees of depressed trade. In spite of a (waning) commitment to Keynesianism, governments stand paralyzed before this situation. It is generally agreed that this inaction has two prin- cipal sources: the effort to decelerate inflation (discussed in the following section); and the fact that reflation of an economy leads directly to a worsening of the balance of payments. Because of the enormous growth of international trade, the scope for independent action by a nation has been seriously eroded. Not unconnected with this growth in trade has been a truly impressive accumulation of data, so that it becomes ever more feasible to attempt a quanti- tative analysis of the trade network of the many countries or regions of the world. An extensive amount of such work has been carried out; but the inaction of the world’s leaders in the face of a continuing and even deepening crisis is an indication of how little the complexity of The author is Reader in Economics at Cambridge University. This article has benefited substantially, both in form and in content, from suggestions by David Vines and from a number of critical comments by Professors James Meade, Lord Kahn, and Sir Richard Stone. It is taken from the author's forth- coming book, Essays in Linear Economic Structures (London: Macmillan, 1980), and is reprinted by permission of the publisher. Jeumel nf Pont Revnectan FrannmicstGnvina O80 Mat 14 Me. 210320 JOURNAL OF POST KEYNESIAN, ECONOMICS the whole problem is understood. The purpose of this article is to present a simple, largely static, prototypical analysis of the prob- lem, using statistically derived data to illustrate—but not to pro- pose—possible solutions. Because of its extreme simplicity, the method can be extended in a straightforward manner to a very large number of countries or regional groupings. Three topics are treated in the first section: (a) how the foreign trade matrix multiplier may be decomposed so as to separate out the internal effects of a demand expansion (or contraction) from the international feedback operating through the trade network; this analysis borrows a method developed by Pyatt and Stone for national accounting (Pyatt and Stone, 1977; Stone, 1978); (b) what will be the balance-of-payments consequences of such a mul- tiplier process consequent upon the expansion of demand in any one country; (c) following from (b), what coordinated demand ex- pansions will have no effect whatsoever on the various net balances of payments. The foreign trade multiplier A world payments matrix is required giving total money outlays for goods and services by country or region and specifying from whom to whom. The columns consist of the outlays of each coun- try on its own and on every other country’s outputs. The rows then yield own output and exports to every other country, since, just as every sale is a purchase, so every import is an export. Row / less column i is the current net foreign balance (NFB) of i. Sup- pose that we can extract marginal propensities to import from the data (assuming that such dubious parameters exist and remain con- stant for a time). Then demand for one country’s output comes from interindustry outlays, exogenous outlays, and imports by any other country." To better understand the nature of the problem, it is helpful to consider the way in which a change in the exogenous outlays in ‘Hence, a) Payya; + 21 = Pray, i= 1,2, -.-50 or, in matrix notation, (2) Bpq +z =pa, 90 that pq = (7—B) 2. Being linear, these relations hold for changes up or down. (An overbar indi- cates a constant, where it is helpful, and an underbar a diagonal matrix.)WORLD TRADE MULTIPLIERS = 321 any one region stimulate output both domestically and in other regions. A decomposition can be used to separate out the direct ef- fects of any such outlays, public or private, and the indirect effects through the trade network, both domestic and international. Coun- tries are divided into two groups according to any useful basis— for example, surplus and deficit, or more developed and less devel- oped, or one unit and rest of the world (ROW). As with any mul- tiplier, the inverse will show the extent to which outputs must rise to generate sufficient leakages to match the stimulus of the injec- tions, the exogenous outlays. The total effect consists of three parts: the direct result in the originating bloc, both primary and internally derived; the secondary indirect or international feed- back effect on the originating bloc; and a tertiary effect that is confined to the other bloc (both primary and internally derived ef- fects).? Thus exogenous outlay z, contributes to demand plus the secondary demand it generates as given by M, — J. The resulting activity leads to imports from group 2, where similar effects fol- low. This then leads to imports by group 2 from group 1, which will stimulate further activity, and so on.? As a prototype world matrix multiplier, I propose to use a freely adapted version of an elaborate analysis carried out by Thorbecke and Field (1974). They divide the world into eleven regions; for Jack of data, however, the Sino-Soviet countries as well as Australia, New Zealand, and South Africa, are omitted, which leaves nine regions. The omitted regions are treated as exogenous. The nine re- gions are: the European Economic Community (EEC), non-EEC Western Europe, Japan, the United States, Canada, Latin America, the Middle East (including parts of North Africa), Africa (includ- ing a few odd bits), and Southeast Asia. All quantities are in U.S. dollars at 1958 exchange rates. The estimations are based on an- nual data for 1953-67. Thorbecke and Field found price effects to ? Stone names these intra, M;, inter, My, and extra, M3 @) U8)" = [M3M,M4). It is more convenient, however, to have these in the additive form given by Stone (see Appendix C). (4) U8) = 0) + aM, — My) + (Ms Ma, — MMi, 5) =[ + [My —11] + 14M, — My) + [MMM — May). 3 The limiting value for both groups from this secondary feedback is given by M,M, — M,. The remaining effects that do not return to group 1 are given by the third term.322 JOURNAL OF POST KEYNESIAN ECONOMICS be generally less important than quantity effects; in this article. prices are taken as constant in the first section, and quantities as constant in the second section. For the purposes of this exercise, | shall simply take Thorbecke and Field’s estimates of marginal pro- Pensities without question, assuming that aggregated marginal pro- Pensities exist, are constant for a time, and may be estimated with sufficient accuracy from time series data. These researchers found, plausibly, that for the less developed regions imports depend on exports, so that in the first section pq for these regions means ex- ports. Also, for lack of data, they ignored the structure of trade between the less developed countries, which has the effect of re- ducing the estimated multipliers. To investigate the properties of the matrix, I have split it into the EEC, on the one hand, and the other eight regions (ROW) on the other. Though it could—and ideally should—be made dynamic, this is a static analysis. There- fore, the results of a change in exogenous demand are to be in- terpreted either as the sum of all effects over time of a single burst of spending, or as the steady state that is approached as a result of a steady rate of spending. If the EEC increases spending by 1000, the result within the EEC will be a derived increase in demand of a further 1591, [/- [M, —1]], a total multiplier increase in GNP of 2591. As a consequence imports rise by 534. These imports, being at the same time exports, lead after a time in each region separately to primary and secondary rises in the respective GNPs. In turn, these rises bring about an increase in the imports of these regions, i.e., exports from the EEC. A vital question, which can be answered only by detailed quan- titative analysis, is: to what extent will the feedback exports match the resulting imports? The ultimate increases in the GNPs are as follows: non-EEC, 886; Japan, 86; United States, 557; Canada, 54; and in exports: to Latin America, 41; to the Middle East, 62; to Africa, 30; and to Southeast Asia, 56. These increases yield an in- crease of 224 in the EEC GNP, [M,M, — M, ], which makes the final result 2815. The 224 adds another 46 to the import bill, making a total of 580 for imports. The induced increase in activity in the ROW regions increases EEC exports by only 87, leaving a massive deficit of 493. The internal expansion in the EEC leads to the increase in ex- ports shown in column (1) of Table 1. These increased demands on domestic production result in further imports; in the limit, theWORLD TRADE MULTIPLIERS = 323. Table 1 (3) (5) i) (2) (+a @) (3) + (4) EEC 3448 580 +87 493 Non-EEC +218 +19 +237 12 +185 Japan +H +3 +3 —2 +32 United States +140 +12 +182 +30 +182 Caneda +o +1 +9 +23 +32 Latin America +31 2) +34 32 a2 Middle East +47 +4 +51 —13 +38 Africa +18 +2 +20 —5 +15 Southeast Asia +420 «+4 +46 18 + 28 effect of the EEC expansion on the other payments balances is as shown in column (2). The direct and derived changes in NFBs are given in column (3) = (1) + (2). Finally, these regions will be im- porting from and exporting to each other on an increased scale, which will further alter the payments structure. The EEC receives a lift of 87, which does little to moderate its deficit. The remain- ing effects of ROW trading are given in column (4). Column (5) = (3) + (4) gives the complete result. The EEC deficit must be matched by surpluses elsewhere—in this case, mainly in the United States and the non-EEC. Balance of trade consequences The same analysis can be carried out for an expansion of 1000 by each region in turn. Since [{ — B]—! is nonnegative, all other regions are bound to gain in GNP from any exogenous spending by one region; because worldwide the NFBs sum to zero, the greater the deficit incurred by the originating region, the greater the gains of the others. In a clearer recognition of this mutual dependence lies a hope for better public policies in the future. The final effect on the NFBs, du, can quite easily be calculated.* The results, along with the effect on pq, are given in Table 2 in Appendix A. The “ By setting 5z; = 1000 successively in (6) 6u = (U-NU-s)* — Ne, where A is a diagonal matrix of the column sums of B, and 5u is final effect on the NFBs.324 JOURNAL OF POST KEYNESIAN ECONOMICS extent of the scatter of results over the world economy caused by individual reactions is an impressive indicator of how far economic integration has progressed. The foregoing analysis holds equally for negative as well as pos- itive shifts in exogenous outlays. Thus, suppose that, for whatever Teason, a moderate decline in world trade has occurred with the re- sult that the non-EEC region has experienced a deficit in its NFB of 352. If it then reduced its exogenous outlays by 1000, it would erase the deficit; but by the same token it would increase the def- icits of the other regions by the same amount, as is indicated by reversing the signs in column (4) of Table 2, thus extending and distributing a further decline of the same magnitude to the other regions, forcing them to take similar actions. The crisis is deepened, with outputs and employment falling pari passu. Whenever a de- cline occurs in world trade, there is necessarily a change in the structure of payments, which produces surpluses for some and def- icits for others. The result has a vicious asymmetry in that the sur- plus countries do nothing and the deficit countries, or at least some of them, apply to the International Monetary Fund (IMF) for help. That help commonly takes the form of a requirement for a sufficient reduction in outlays, either by fiscal or monetary means, to balance the negative NFB. By the above argument, how- ever, this only spreads an equal quantity of deficits in a rather arbitrary way among the other countries. In each case there is a determinate fall in output and employment. Thus, by the very na- ture of its design and purposes, the IMF becomes the agent of a sequential intensification and prolongation of the depression. An objective and operable compensation principle exists in such a system. Table 2 allows one to calculate easily the total effect of any combination whatsoever of expansionary policies. Those who gain reserves can always compensate the losers and still remain bet- ter off, because of the general rise in output. Each column of du adds to zero, so that, if the other regions all return their gains to the expanding region or regions, compensation is made and can be deducted from the increased outputs, leaving a net gain of output. Thus, suppose that by prior agreement the following exogenous outlays are undertaken: the EEC, 1000; Japan, 500; and the United States, 2000. After allowance is made for the transfer of reserves, the net gains will be distributed as shown in Table 3a in Appendix A, with no loss of reserves by anyone. It appears that to realize such a scheme technically nothing more is required than an IMF- type clearinghouse.WORLD TRADE MULTIPLIERS = 325 Coordinated international expansion The general nature of the problem of coordinated international ex- pansion has long been familiar. In the 1930s a pioneering proposal was put forward by Frisch for national “circulation planning,” a scheme equally relevant to international planning. The latter tension was made after the war by Frisch in the form of a bargaining matrix for cutting imports in such a manner as to minimize the cuts. The subtle, complex procedure unfortunately ignored the effects through GNP and was purely restrictive in aim. Although unfortunately the problem was explicitly excluded from The Gen- eral Theory, it is from Keynes that the most attractive solution derives. The guiding principle is clear: some sort of general expan- sion would ensure that no one incurs a serious deficit. From data not yet available at the end of the war, it would have been necessary to determine the relative magnitudes that would produce such a result. But the fatal lack is that there is no sovereign fiscal and monetary authority to monitor and control the expansion. Who is to expand and by how much, and when is deceleration to occur? For this one needs not only an agency but also operationally reliable parameters. Fortunately, the growth in the relative importance of world trade has been overmatched by the rich accumulation of statistics. With this information it is conceivable that, in the man- ner of indicative planning, national governments can be cajoled into a rough cooperation without sovereign power to force them. Thus, given a set of target changes, 57q*, in GNPs, the required changes in aggregate outlays can be calculated.’ As in national planning, one requires a set of expenditures, not simply a lump sum: it is no good to pump more demand into an economy al- ready fully employed in several crucial sectors. Aggregation has been one of the failures of Keynesian demand management. A ma- trix analysis, with n? multipliers, allows the choice of that combi- nation of different exogenous outlays which will bring the economy to any particular desired level of each of the main economic sectors. Internationally, however, net foreign balances are a separate and more pressing problem. Any move toward fuller employment will bring an assortment of positive and negative NFBs. No single set of exogenous outlays can deliver both desired outputs and zero net balances. To achieve two targets, one requires two instruments; two sets of targets require two sets of instruments. Suppose that, *(7) 5z* = [1—B] bpq°.326 JOURNAL OF POST KEYNESIAN ECONOMICS with unemployed resources, we are given the task of finding what changes in outputs in the various countries would not alter their net foreign balances, i.e., neither improve nor worsen them.* The exports from country i induced by changes in output in country j must exactly equal imports to country / from all other countries, the imports being the result of the changes in the output of coun- try i. Equations (8) and (9) detect that change in the vector of ex- ogenous outlays (z), which will increase (or decrease) the outputs of the economies together to the degree required in order to yield no change in the current net foreign balance. Given reliable estimates of an unchanged set of marginal pro- pensities, the solution consists in finding the vector 5pq that nulli- fies the variations in the NFBs. Such a solution always exists, since 6 is simply degenerate. No fall in reserves of any region will occur if output and exogenous injections are expanded in the proportions shown in Table 3b of Appendix A. There are, however, two separate but interconnected problems: the first is how, as above, to increase output without running into deficits; the second is the more pressing one of how to reduce pre- existing deficits and surpluses. The latter is a dynamic problem of reallocating production among regions, those with a surplus ex- panding and those with a deficit contracting. This can be formu- lated in terms of a simple homeostatic feedback routine.’ It has often been found that, once such a control routine is in operation, provided changes of the control variables are frequent, such changes For each country, we require that (8) E ay5jq)* pei Eaxideiai® nei. Adding a4;5p;q;* to each side, one has in matrix form 9) [8-2] 6pq* = 18] 5pq* = 0. ‘7 Since an increase in z; brings a fall in uj, and a fall in zj brings a rise, let (0) 62 = e[u—0} an = dba (12) 8pq = (1B) 162. Hence 3) pq = elI—B1~* fa, which can be simplified to a4) pq = e({/—B1"1 UA) — 1 ba, yielding a system of discrete first-order difference equations, straightforwardly soluble.WORLD TRADE MULTIPLIERS 327 need not be large. One attractive feature of such a scheme is that, in principle, one need know only the size of € to operate it. There are two aspects to any set of outputs: the relative levels of outputs, and the general level of outputs. The feedback routine sorts out the relative outputs but leaves the general level substantially unal- tered. The difficulty is that those with deficits resist contraction and those with surpluses feel no compulsion to expand. If, how- ever, in a situation of general depression, we join the two princi- ples—a balanced expansion and a zeroing exchange feedback— the situation is transformed. No region need actually contract; some simply grow more slowly than the others. Such a solution re- quires a delicate calculation, but in principle it is possible to revive world trade in such a way that all outputs grow and that they grow so as to eliminate gradually all exchange imbalances.* When one or more countries reach the limit of resources, then the growth tate must be decelerated. The others will then have to pursue the same objectives by other means—for example, by structural changes such as modernization, subsidies, tariffs, quotas, and ex- change rates— in order to achieve anything like simultaneous full employment of resources. To illustrate such a solution, I propose a simplification and al- teration of the data. Since Canada is both small and asymmetrically linked with the United States, these two sectors are amalgamated into North America. Because the payments structure of the less developed regions is lacking, they are aggregated into a single sec- tor, ROW, and a fictitious internal expenditure propensity is added. Furthermore, their import propensities are scaled up so as to sum approximately to unity, in line with the finding that their imports are determined primarily by their export revenue; thus, any in- crease in their exports is fully distributed over their imports in the previous proportions. The system is given in Table 4 in Appendix A. By manipulating the system as suggested following Table 4, we obtain Table 5, which indicates that all outputs increase except that of the EEC, which has the worst deficit; its output drops only slightly and then joins the general upsurge. Every NFB is reduced and would eventually disappear in the absence of new disturbances. ® The required solution takes the form as) p4 = yKpa(l + yt + el— B81 Al — Na, where K is to be chosen correctly and ‘y is any desired and feasible rate of growth,328 JOURNAL OF POST KEYNESIAN ECONOMICS Such a policy is relatively short term, with the specifically Keynesian purpose of shifting the world economy from a depressed to a buoyant state. It does not and cannot deal with more persistent and deep-seated imbalances, just as pushing a national economy to “full employment” does not and cannot bring full employment to all regions and to all industries. It is. however, much more feasible to make the necessary structural changes in exchange rates, tech- nology, real wages, and such, in a buoyant state of trade. The con- verse is that effecting the necessary structural changes will not of itself bring full employment. It should be added, of course, that the two types of policy, though logically distinct, should not be kept separate in practice: the structural changes should be antici- pated and introduced gradually. Unfortunately it is impractical to achieve a coordinated world fiscal policy; but that is no reason for not investigating its proper- ties fully, if only because such examination clarifies a basic aspect of every individual nation’s problem. Any region or country that pursues an expansive policy gains but suffers a usually unbearable loss of reserves. With adequate quantitative information, it should be possible for regional economic blocs to engage in bilateral or multilateral expansion. Table 6 in Appendix A shows the separate effects of each region’s adding, in turn, 1000 to its (domestic) ex- ogenous outlay. Column (1) shows the results for GNP and reserves if the EEC spends 1000 additional. Column (2) shows the same for non-EEC and similarly for the others. At least partial offsetting of reserve losses is possible, but no simple solutions emerge from the table. This may be because of the special behavior of North Amer- ica as provider of reserve currencies, or because of the special role of the less developed regions. It is a generally agreed policy objective for the more developed regions (MDRs) to assist less developed regions to develop, but the implementation of such policies has been notably feeble. The fact that the foreign balance is a dominant constraint for many LDRs may provide a more potent basis for such policies. The question is: under conditions of underemployment in the MDRs, would a pol- icy of subsidizing the LDRs be beneficial to the MDRs as well? Thus, suppose ROW is given a nonrepayable open book account for additional imports from the MDRs. The final resting place of the credit is given in column (5) of Table 6. The initial “loan” can be extinguished by the return of the gains in reserves. i.e., 13 from the EEC, 251 from the non-EEC. 113 from Japan, and 606 fromWORLD TRADE MULTIPLIERS = 329. the United States. When these unrequited exports are subtracted from the increased outputs, there are left the following net gains in goods and services that would otherwise not have been pro- duced: 1248 1100 184 1241 Such a policy could be executed so as to lead to a permanent in- crease in world reserves, with the result that the 5u’s would be net increases in reserves with possible further consequences. The 1000 could be a non-interest-bearing, nonrepayable loan to ROW and could lead to increases in cash reserves (assumed acceptable to the consortium) held by the various MDRs. These results are to be con- trasted with the quite different ones arising from an increase by any one region in its internal exogenous spending. It is illuminating to view the international payments system in terms of Say’s law of markets.’ In a sense, the aggregate demand for foreign exchange is always identically equal to the supply forth- coming. But any level of world output satisfies, with no relation or relevance to full employment; nor is there any bar to a rising or falling level. In agreement with Say, no one market will necessarily be in equilibrium: when some are in surplus, others must be to an equal and opposite extent in deficit. But also, as we have seen, there always exists a particular solution in which Say’s law is satis- fied not only in aggregate but also in detail for each sector. The present fiscal stalemate of the world economy is a tragic failure of policy. Governments have been consistently ignoring the essential Keynesian message. If inflation were brought down to tolerable levels, there would still be wasted labor and capacity. If Britain and the United States were made competitive with Germany and Japan, that alone would not solve the problem. The simple consequence of Say’s law is that the payments balance is consistent with any state of trade, buoyant or depressed. In many ways the present international situation is analogous to the pre-Keynesian 1930s. Once the downswing had begun, each individual firm tried °The current variations in foreign exchanges always satisfy Say’s identity, since 16) [—B) pq =0 for all 2g.330 JOURNAL OF POST KEYNESIAN ECONOMICS to solve its own problem by reducing its purchases of labor and materials, which only transferred the problem to other firms. Sim- ilarly governments, faced with reduced tax revenue, cut expendi- tures, triggering only further decline. The error lay in partial equi- librium analysis, which considers the part instead of the whole— an error once again becoming fashionable. Thus now each govern- ment tries to solve its problem while taking the rest of the world as given, a practice that quite simply results in the wrong answers. Though this article is phrased mainly in numerical terms, its aim is not to formulate quantitative policy but rather to illuminate some kinds of much needed policies. What is required is a set of numbers—not just any numbers, but a set sharply pointed at the essential problems in a reasonably comprehensible form. Would it not be splendid if one did have such a set that gave usable answers? Would it not stop politicians and economists from being blown about by the winds of doctrine? It might even finally lay to rest Keynes’ devastating 1939 attack on Tinbergen’s pioneering econ- ometric model. Some years ago von Neumann made the claim that the weather could be predicted, given a large enough organization and a big enough computer—on the sensible grounds that all weather is somewhere, and that we know the physics of its motion. Analogously, all the world’s economic weather is somewhere, and all we need to know is how it flows. Our knowledge of the flow is faulty and incomplete but we can, to some extent, “make the weather.” World price-level multipliers This section considers the value dual of the world economy. Schemes for raising output and employment are commonly blocked by the preoccupation with inflation as caused by excessive demand. In any case, it is a fact that one of the chief reasons for the per- sistence of unemployment is the effort to decelerate inflation. It should therefore be useful to investigate a prototype model of the international transmission of inflation. In spite of some complica- tions, the model is kept simple enough to be expandable into a large-scale, multiregion model. As in the first section, three topics are considered: (a) The dual of the foreign trade multiplier registers the effects of an increase in cost components spreading over the whole trade system. The Pyatt-Stone decomposition proves equally illuminating here. (b) The inevitable consequences of such an in-WORLD TRADE MULTIPLIERS = 331 fectious inflation produce a complex pattern of disturbances to the NFBs. (c) Finally, a brief consideration of variable exchange rate control is introduced. The point of view adopted here is essentially the classical, Ricar- dian rent doctrine. The presumption is that the long, substantially uninterrupted growth of output led to an increased pressure on supplies of unproduced goods, resulting in a rise in their prices— their rents. Even if this is not the explanation of the recent rise in raw material prices, it will surely be so in the future. Given such a deterioration in the terms of trade, the standard of living, the real value of earned incomes (wages and salaries), should fall. For a variety of reasons, this fall (or failure to rise) in real income has been resisted with partial success. The hypothesis used here is that the conflict over the distribution of income is resolved by contin- uing inflation. Analytic convenience dictates the ignoring of tech- nical progress, which would complicate the analysis but would not alter its essence. To treat the. problem of prices, it is desirable to elaborate the model by distinguishing cost-determined prices from those deter- mined by relative scarcity (supply and demand). The latter, with special reference to raw materials, I shall call “rents” for short. The data used thus far do not make the distinction, so I have made arbitrary changes. Therefore, the numbers are consistent with the Thorbecke/Field data but must unfortunately be regarded as illus- trative, with no pretense to realism. The positions of North Amer- ica and non-EEC are interchanged in order to subdivide along MDR/LDR lines. This analysis assumes constant output and an underlying con- stant technology (including given and constant markups).'® The general level of wages is taken as ambiguous, partly cost determined and partly scarcity determined, and therefore treated on an ad hoc exogenous basis. All exchange rates are given and constant; all prices are in a common reserve currency—in this case, dollars. "©The basic matrix is an Ic} = (41 (Alla), from which earned income (wages for short) and rents have been extracted. lence (1g) = < + + >U—c], where w; is the general level of wages, py is the general level of rents, and F is everything else, treated as exogenous. (See Goodwin, 1949, 1953.)332 JOURNAL OF POST KEYNESIAN ECONOMICS The initial price levels are unity, so that GNP is real.'? Given the empirical values (see Appendix B), one can find the effects of any combination of changes in wages, rents, or exogenous outlays, for example, investment or government spending. Cost pricing and the fact that, for the balance of payments, it does not matter whether price or quantity varies, allow the dual analysis. Introducing prices and quantities makes the problem nonlinear, however, and hence, in order to handle large systems, it is necessary to reduce them to linearity by taking each in turn as constant. The crucial assump- tion is not the dual independence of the two sets of variables, but rather the constancy of the input coefficients, which effectively excludes price elasticity. Suppose that raw material prices rise by 75% in non-EEC and by 160% in ROW; the resulting increase in price levels would be distributed among EEC, North America, Japan, non-EEC, and ROW as follows: 12% 8% 17% 61% 146%. If the workers in regions 1 to 4 raised their earnings so as to main- tain their real values,'? while in region 5 wages remain constant, further price rises of 5% 4% 4% 27% 2% would ensue, giving a total set of price rises of 17% 12% 21% 88% 148%.'? The initial effect is simply the originating rise in wages or rents. Then, as costs are up, internal prices must rise, which is a further increase in costs in the region itself, until a final, static, limiting 11 The arithmetic manipulations are given in Appendix B. ‘2 That this assumption will underestimate the force of cost-push has been emphasized to me by Professor Meade. If wages are determined on expected as well as realized inflation, then the inflation will tend to accelerate instead of settling down to a steady rate. To take account of this would unduly compli- cate the analysis, 13-The Pyatt-Stone decomposition illuminates the international transmission of inflation. ag) =<6v>[/—c}" (20) = <6v>[ [+ Ny —11] + INQ Ny M1 + IN3N2Ni —N2Ni)1-WORLD TRADE MULTIPLIERS 333 value is reached.'* The originating price rises lead to higher export costs, which become import costs in varying degrees in other re- gions; the latter then suffer a parallel inflation, which is reflected back to the originating region.'* There are clearly lags in any such process and, though a primitive dynamic analysis is possible, it would be complicated. These lags become important because of the tendency of induced earning to rise. A compromise solution is to assume that the price rises are substantially completed before wages rise. Then a fairly simple stepwise solution is possible. To clarify this complex situation, consider a single rise in Pys by 200%, where the world economy is divided into two regions, the MDRs (1, 2, 3) and the LDRs (4 and 5). The rise in prices in the LDRs is<1.6% 46%>. This impulse transmitted through ex- port prices leads to increases in the MDRs of <1.9% 2.3% 5.0%>. Making the simplifying assumption that these rises are substantially completed before any reaction in wages occurs, and that this reac- tion is a uniform attempt to maintain real wages in all regions ex- cept 5, we find further increases to <2.7% 3.0% 5.9% 2.8% 46.7%>. Then comes a second round: since world demand was such as to sustain originally a rise in py,,, presumably its price level will again rise so as to maintain its real value. The final result of the second rise is<12% 8% 17% 61% 146%>. A second ef- fort of wage earmers will further lift this to <19% 12% 21% 88% 148%>. The effect of a rise in wages that maintains their real value is to reduce the real value of rents back to their previous value. There- fore, the process may be considered as likely to continue indefi- nitely or until something else changes.'® The solution is a constant rate of inflation, different for each region because of differing cost structures. Any single price-level rise or any combination of them may be calculated. Thus, if py, rises by 50%, Py, by 70%, py, not at all, py, by 130%, and py, by 220%, the system will gradually 14 Given by [1+ [Ni — 11]. *5In the limit this additive effect is given by [N,N, —N, ]. Quite separately all other regions, reflecting back and forth their price rises, will engender the remaining extra effect [N3N2N — 2M]. *6It is shown in Appendix C that such a dynamic process can be formulated and transformed into = I+ e111. Pq334 JOURNAL OF POST KEYNESIAN ECONOMICS approach the following steady rates of inflation: <11.8% 28.0% 4.6% 20.1% 54.7%>. This analysis, far too crude and simple to represent the irretriev- able complexity of reality, nonetheless seems to me much more convincing than the simplicities of currently fashionable monetar- ism. The preconception that money is the root of all evil is very at- tractive, but it will not do, confusing, as it does, symptom and dis- ease. By contrast, it is clear that massive increases in output press upon fixed or inelastic supplies, leading to rises in both price and value. In turn, the consequent lowering of real earnings is resisted in the MDRs, which leads to successive bursts of price increases. Balance of payments consequences Just as inflation will be distributed over the regions in a determi- nate but irregular way, so also will be the balance-of-payments ef- fects. Consider a rise in py, by 75% and in Pys by 160%; the re- sulting NFBs will be <—50.4 —111.0 —25.7 +29.5 +157.4>. The MDRs, having to pay more for their imports, generate deficits matched by surpluses elsewhere; similar effects can be calculated for changes in wages, investments, etc. It is therefore easy to show how any change in the exogenous variables is bound to precipitate balance-of-payments difficulties in a regime of fixed exchange rates.'7 Variable exchange rates The special.aspect of an international payments matrix is the exis- tence of exchange rates. The explicit introduction of exchange rates with domestic prices presents no serious difficulties in the ac- counting as such. Far more than in the case of outputs, however, the analysis is not only of limited usefulness but may also be quite misleading, since there is the implicit assumption of zero price elasticity of demand. Thus, for a deficit it quite correctly recom- mends appreciation instead of devaluation: in effect it gives only the hook, not the whole of the J curve, thus ignoring the serious question of longer run substitution effects. Such complexities lie 17 The assumed initial values, pq(0), give rather large NFBs. The price levels that would balance payments are given by K[0—C] =0, where Gis the diagonal matrix of the row sums of C. With K = 567.7, bq = <2590 4945 1062 953 460>.

[/—C] will indicate the proportions in the changes in wages or rents that would cause no additional balance of pay- ments problems: <53>=<+2.972 +1485 +0.977 +0.583>.WORLD TRADE MULTIPLIERS 335 beyond the scope of this article; a brief indication of the analysis with fixed real trading patterns is given in Appendix C. Exchange rates convert domestic price levels to (dollar) units. Thus, from the pq that nullifies the NFBs can be derived the do- mestic price level which does the same; or, given domestic price levels, one can calculate the exchange rate which does the same. In the illustrative example of Appendix B, the initial exchange rates (in terms of dollars) of the five regions are: <0.25 1.0000 0.0667 2.00 0.050> The calculated exchange rates that reduce the NFBs to zero are: <0.3650 1.0000 0.1292 1.4490 0.0279> In the absence of such sophisticated precalculations, the regions could attempt homeostatic control, making small iterative changes in exchange rates on the basis of no wider information than their recent NFBs. Suppose that a country expects its own and foreign prices, incomes, and exogenous outlays to remain constant; it wishes to change its exchange rate, taking the others as constant, so as to balance its payments.'® Thus, given the initial values above, if the EEC wished to act independently to balance its pay- ments, it should set 8, = 0.4356 instead of 0.3650. This is a sim- ple example of the purchasing power parity doctrine. If all price levels are rising at different constant rates, then each exchange rate should fall at the same rate at which its price level is rising to main- tain a constant NFB. The reason for the loss of confidence in depreciation as a cure for a negative NFB is evident in a system like this. If a country or a region depreciates in the hope of improving its NFB, it will raise all import costs. If it is small in relation to the world economy, then the domestic price level goes up by the import price increase. This leads to a multiplier effect on other domestic prices. Fur- thermore, if exogenous domestic costs are raised so as to main- tain their real value, costs rise still further, and hence prices, and so on until prices have risen in the same proportion as the devalu- ation, thus negating the effect of the devaluation and making price elasticity irrelevant. Since costs and prices respond quickly and 18 The solution is 0; = {Chi4j1orak0X0; — Cu).336 JOURNAL OF POST KEYNESIAN ECONOMICS output substitution is slow and somewhat weak, the price inflation may be substantially completed before any substitution has time to be effective.'” The fact that output is kept constant in the analysis of price lev- els and that price is kept constant in dealing with output is not as inconsistent as it seems. It is merely a device for reducing a non- linear problem to the more tractable form of a pair of linear ones. Nearly simultaneous variations of the twin motions can be studied. First price is held constant and then output is held constant to find the change in price. The process is then repeated successively. In this manner the parallel but somewhat independent behavior of both is obtained. It remains true, however, that systematic inter- action between the two is neglected. In these systematic interac- tions lie deeper and more intractable problems beyond the scope of this article. By laying out here a simple, largely accounting an- alysis, it is hoped that a useful foundation will be provided upon which more ambitious (and fallible) superstructures can be erected. 19To illustrate, suppose that region 4 depreciates in relation to the dollar from 2.000 to 1.700 in order to improve its NFB. If the other regions are taken to be unaffected, the behavior of pi, is as follows: t 0 1 2 3 4 « ag 960 960 979 996 10120... 1128 The dollar value of 960 at 2.0 is the same as 1128 at 1.7, so the effect of the depreciation is entirely dissipated. REFERENCES ‘Goodwin, R. M. “The Multiplier as Matrix.” Economic Journal, 1949, 59, 537-55. . “Static and Dynamic Linear General Equilibrium Models.” In Input-Output Relations. Leiden: Netherlands Economic Institute, 1953. Pyatt, F.G., and Roe, A. R. Social Accounting for Development Planning. Cambridge University Press, 1977. Stone, R. “The Disaggregation of the Household Sector in the National Accounts.” World Bank SAM Conference, 1978. ‘Thorbecke, E., and Field, A. J. “A Ten-Region Model of World Trade.” In International Trade and Finance, Essays in Honour of Jan Tinbergen. ed. by Willy Sellekaerts. London, Macmillan, 1974.0 oO + e6r— 4 lot o 4+) 9 +01 BL6— 9LOL bo tb w+ ce Lel+ zy 0 ° 0 zx + 4 lt He BW BH B O+ of 61+ 46 SL+ O& 8 + Oy 6+ BY B+ WD L b+ @ t+ @ C+ Ww 9 giz+ We SE+ OF Ze+ 9 S OLit @S Gilt 96 zBl+ Lag oss oot vE+ 16 ZE+ 98 EC ty + €2% teeE- SESE SOl+ 988 Zz 4S+ OIE 96+ ves e6r— G1EZ IL 7QU— le—N Nl =9, [8—Ng="g QB, la—1) = 20g = 49 001 JO 1030ag Aq 10}9ag suoHoafuy zoey, V XIQNaddV338 JOURNAL OF POST KEYNESIAN ECONOMICS Table 3a 8 bu = 3329 —400 +3329 1410 +262 +1148 820 —193 + 820 59668 — 55 +5966 388 +227 + 161 23 +5 + 88 102 +63 + 9 58 +30 + 2 139 + 68 7 Table 36 oq & EEC 1000 +17 Non-EEC 1886. + 294 Japan 260 + 132 United States 3477 +1133 Canada 423 + 246 Latin America 46 +2 Middle East WwW — 90 Africa 64 + 33 Southeast Asia 121 + 60 Table 4 1 2 3 4 5 EEC 1 0.386 —0.049 —0.067 —0.015 —0.220 Non-EEC 2 0.084 0.302 —0011 —0014 —0.161 Japan 3 0.012 -0.016 0863 —0.014 —0.117 =[/—8] North America 4 0.086 —0.026 —0.325 0.408 —0.317 ROW 5 0.214 —0.025 —0.085 0.037 0.825 1 3513 0.728 OSt4 = 0.286 1.261 2 1.631 3.739 0430 0.326 1.351 3 0251 0.139 1.241 0.084 = 0.302 =[/—#] + 4 1.698 0.737 1.360 2.753 1.847 5 1.083 0.349 0.335 0.216 1.604 1 0.376 0.049 0.067 0.015 0.220 2 0.084 —0.116 0.011 0.014 0.161 3 0.012 0.016 —0.488 0.014 0.117 =(8 4 0.066 0.026 0.325 —0.080 0.317WORLD TRADE MULTIPLIERS 339 In 1967 the GNPs were roughly 2590 1920 Pq(0) = { 800 720 1150 yielding the NFBs +291 Abq(0) = u(0) = }— 92 +269 From fg = 0 1.544 3.667 0.659 9.106 1.000 Suppose that by general agreement the following regime were to be kept: &pG = 0.05(300)pq(1.05)r + 0.02[[7— 8] [1-0] — Nag The behavior of real output, for the first three periods, would be as shown in Table 5. Table 5 pati) pat2) aia) : u(3)* 1 (2586 2587 (2589 He 438 2 2019 2087 2155 . +275 3 807 B15 824 fa cu 4 7364 7523 7683 Ba +254 5 1159 1169 1180 . oO Table 6 8pq= (1-8) 182 Su = —AN VBI" — be _ — (2) _ 3) - 4 _ (5) Bo bu ig fg Bg 3513 968 728 + 7 «#9514 + 5 206 +3 1281 +13 1631 +303 3739-305 «= 430 «+ 80-326 «+61 «1361 +251 251 +94 139 +52 1241 —835 84 +32 302 +113 1608 «+557 «737 «+242 «1360 «+446 «2753-97 «1847 +606 1063 +11 349 + 4 335 + 3 216 +2 1694 —983 orwne340 JOURNAL OF POST KEYNESIAN ECONOMICS APPENDIX B World price multipliers For initial price levels of unity, [p] [A] [p]~' = [A] = [8], except that wages and rents are extracted from both A and C. 1 2 3 4 5 1 0.211 0.015 0.067 0.049 0.054 2 0.066 0.190 0.325 0.026 0.078 Japan 3 0.012 0.014 0.025 0.016 0.029 = [A] = [2] 4 0.084 0.014 0.011 0.255 0.040 5 0.053 0.037 0.085 0.255 0,043 =<0.250 0.220 0.103 0.321 0.525> ay = <0.153 0.182 0.089 0.122 0,220> These numbers are dimensionless and represent the proportions of GNP ex- pended domestically (columnwise) on imports, wages, and rents. 1 2 3 4 5 0.211 0.042 0,021 0.036 0,025 0.024 0.190 0.036 0.007 0.013 0.039 0.126 0.025 0,039 0.044 =[C]=([@]" [4] [7] 0.114 0.053 0.005 0.255 0.025 0.114 0.223 0.057 0.040 0.043 wbwone These numbers likewise are dimensionless and are to be read row-wise as pro- portions of GNP systematically going to domestic use and to exports. 1 2 3 4 5 1 1.345 0.087 0.033 0.067 0.038 2 0.044 1.251 0.049 0.018 = 0.021 3 0.070 0.184 1.037 0.062 0.054 =[l-c"* 4 0.203 (0.114 0.018 1356 0,043 5 0.094 = 0.312 0.076 0.068 1,057 = <»> [/—-c]* = <<96w, 159w, 25w, 104w, = 172w,> +<158y, 903Py, SPy, 177Py, 278Py5> +<443 2368 380 357. 13>>WORLD TRADE MULTIPLIERS 341. The international transmission of inflation Given w(0) = <6.75 1000 3.28 592 3.66> Py(0) = <2.506 1.455 1440 1.322 0.950> then (0) = <1487 5270 389 1207 907> pq(0) = <2590 7220 800 1920 1200> APPENDIX C Given a transactions matrix, X, in a common unit of account—for example, Bancor—let it be divided into two: X = H+Z, where H contains the parts systematically related to GNP, and Z contains the remainder, treated as exog- enous, Division of the respective columns of H by their respective GNPs pro- duces A. For each country or region, q is the real GNP and p is the average price level. Let [p] [4] [P] “' = [B], where the columns of B are dimension- Jess numbers that give the propensities to spend out of GNP. Bpq = [P] [A] [a] is the systematic part of the transactions matrix of the dollar value of real out- lays for any level of real output, g, with p held constant. In equilibrium, bra} + zfs} fab. vetsnafs} =2 ft}, we nave fa} = U8] frp. The variations in the NFBs are given by bu = B {Sea} — {Bee} = Bea}, with A as the diagonal matrix of the column sums of B. Row sums are total re- ceipts and column sums total outlays, so that the diagonal domestic transac- tions cancel, leaving the NFBs. fh =e-xift and {a} = era fi ~ tafe} pute}. In terms of the exogenous injections, {ee} = B-Al WB) {52} = B—A U+B+B? +...1 {82} = (U-N U—a)* — 1 {62}. The derivation of an additive decomposition of the inverse can be accom-342 JOURNAL OF POST KEYNESIAN ECONOMICS plished as follows: Separate the inverse into diagonal and off-diagonal blocks: cr bey Bg ove gio q i = [M,M,] + (M,M,M, —M,M,). Divide B similarly into diagonal and off-diagonal blocks: [8] = [D} + (8-2). Let pp)" = B). This yields all the elements necessary for —B\ = + iM, — 11) + 1, —M,1 + MMM, MM) Let [C] = [7] [4] [@] and

= <1>[Z].. These tell us, given the quan- tities being traded, what the effects of price changes will be on the resulting transactions: =H+<1>Z = [q) [A] [a] + <»> =[pAq] + <> =<>[I-¢c]", all in world reserve monetary units. Variations in p alter B but not C; sym- metrically, variations in q alter C but not B. In both, the basic coefficients ay are taken to be constant throughout, representing a technology assumed to re- main unchanged over a short period. For the value dual, wages and rents are extracted from A and treated quasi-exogenously. These two categories are added to to form an enlarged , thus = C+ + + =<>[/-c]". Since they are linear, the equations hold for increments as well as for any one item separately, e.g., 5p = 87474; where G; is row i of [/—C]*. The Pyatt-Stone decomposition for prices is accomplished in exactly the same way as with the output inverse. U-c}" =N,WN, = U+1N,-11] + 1N,N,-N,] + INN,N,—-N,N, ]. A rise in any term of v will entail a worsening of the terms of trade of otherWORLD TRADE MULTIPLIERS 343 regions. If this worsening is resisted, there results a distributional conflict that is resolved by continuing inflation. A simple aggregative example illustrates the point, p=(a,w+aypyX1—a)", giving distributive shares as 1 =(a, w/p+ ayPy/pX1 —a)". If one value is raised, the other must decline. If it refuses to do so, the inflationary process is represented by (t+ 1) = (a, (2) eee (1~a)"t where 0) w(t) = pete p=" +O (©) (0, from which pt+1)= (: sa) (1-2) p(s). ‘The consequence is a constant rate of inflation. For an interdependent sys- tem, the evolution will be much more complicated, with both transient and steady-state differences in the inflation rates. To see the eventual steady state, we may proceed as follows, with initial values indicated by an overbar: = <+ + >Ul—C]* =<>I-q" = [pq] [pal Uc. Let

/)"*. Then =(pal lc} =

@-cl". If all wages, rents, and exogenous outlays attempt to maintain their real values in the face of an initiating increase of 8x9 WA) = (HIBDP ADs Py hd =(1 + Ry MPP AD: FQ) = GIP). sparse =lia—cr + Yager ey] Fai = Spy sy 9; ~|+ Minit wa}: 7 i {G;] isa diagonal matrix consisting of the elements of the ith row of [/—C]~!. ‘The effect of a given change on the NFBs is = [0] —IC].344 JOURNAL OF POST KEYNESIAN ECONOMICS Consequently = [—C]" [o—C] = [l+C+C? +...][o—C] = [— [-C]" [a]. [g—C] = 0 has a nontrivial solution, KAq, since [a — C] is simply de- generate, This relation can be used to find what changes in w or py will re- sult in no change in the NFBs, for example, <8Kpq>[—C] = , of 5. Given rigid domestic factor prices, homeostatic control would be as fol- lows: <60> =-s, 0 = [56] [/—-C]* <épq> = -s [pq] [9-1 [/-C]* =—s[0(0)] - /—o] /—C]*]. An equilibrium solution, fq, will nullify the right-hand side. This means that such an adaptive flexible exchange rate can change domestic prices (with fixed domestic wages, rents, etc.) into international prices in such a way as to avoid deficits and surpluses. Thus, setting the sensitivity parameter s at 107" to guar- antee stability, and with a constant = <5948 5270 5895 604 18,109>, we get the following time series: pq(0) <2590 7220 800 1920 1200> q(t) <2808 7840 1049 2225 425> Pq(2) <2873 7948 1154 2410 196> Pq(3)<2891 7910 1210 2563 129> It is convergent, and even rather rapidly, to pq(n)<3781 7220 1551 1391 671>. It remains doubtful that this control routine could be made to converge fast enough to avoid the exhaustion of reserves, not to mention the difficulties inherent in the tendency of domestic wages, rents, investments, etc., to alter frequently, both in response to variations in the price level and for a variety of other reasons.Copyright © 2002 EBSCO Publishing