1. Introduction. The old 'cost push' versus 'demand pull' debate on inflation is being reargued in the context of the current world inflation. [A full version of this article, with graphs and equations, is at Our Current Inflationary Experience]
In the best of the current debate, which is carried on in a world context, the contrast is between 'special factors' and a 'demand pull' explanation that focuses on the build up of price expectations, initiated and buttressed by the lack of any mechanism to control the rate of growth of the world money supply. The key determinant of the world money supply since the last war has been the United States money supply and, since the mid 1960s, this has been increasing rapidly. The flavour of the debate is captured in a recent panel discussion on 'world inflation', which includes contributions by Fritz Machlup, Steven Marris, Assar Lindbeck and Harry Johnson.
Marris notes that 'there are a large number of countries involved, and if you look at them each individually you can find reasons for recent inflationary tendencies which seem quite specific to that country and goes on to emphasise various socio-economic factors. These include a general 'loss of fear' of economic insecurity, and 'growth of frustration or anger', for example about slow economic growth, or based on increased awareness of income differentials.
Johnson argues that the essential analytic problem is to explain the more or less simultaneous rise in the price of various factors, goods and services and that most of the special factors can at best explain movements in relative prices. He notes that the major discipline imposed for the control of inflation is the balance of payments, although the United States has been in a position to treat its external position with so called 'benign neglect'. If the United States macroeconomic policy is inflationary, as it has been for almost all of the period since the mid 1960s, other countries will import inflation via the well established channels for the international transmission of fluctuations in prices and activity, unless they are prepared to float their exchange rates. And of course, until very recently, hardly any countries have been prepared to do this. Johnson notes the 'nonsense' of an incomes policy in individual countries with fixed exchange rates during periods of world wide inflation. Machlup and Lindbeck fill in some gaps by spelling out channels for the international transmission of changes in prices and activity. They mention
(i) direct effects via the prices of traded goods, (ii) income effects from the trade account, (iii) monetary effects of the balance of payments, (iv) particularly large wage settlements in more profitable export sectors which flow on to other wage earners and (v) the understandable propensity of governments to avoid acting against inflation until the balance of payments deteriorates.
There was pretty general agreement among all panel members about the importance of these transmission mechanisms, most of which have long been recognised. Also well established is the proposition that individual countries can inflate at a faster (slower) rate than the world average, for example in response to a higher (lower) rate of domestic credit expansion
Individual nations can choose their own inflation if they are prepared to frequently devalue (revalue) to maintain external balance. The problem is to devise tests of the primary initiating factor(s) in recent and earlier experience. This paper attempts to shed some light on this issue by examining some recent empirical evidence. By way of further introduction, however, the state of the debate in Australia is briefly summarized first.
2. The Debate in Australia. In the Australian discussion of inflation the 'special fac tor' has, almost invariably, been the Arbitration Commission. For example, a recent article on the state of the Australian economy argues:
'The increase in inflation in recent years can be largely explained by the relatively low level of unemployment and the relatively large increases in award wage rates'.6
A later article by the same author is even more emphatic, pointing out that from January 1970 to January 1971, award wage rates increased by 11 per cent, 'the biggest increase in any year since 1951', 7 although admitting that allowance for this left a significant residual amount of inflation to be explained.
In both articles the writer concedes that award wage increases may be based on factors such as the rate of inflation, but this does not prevent him from registering a strong plea for an incomes policy. There have been four attempts to explain quantitatively award wage determination in Australia, 8 and although the most recent is discussed in more detail below, the clear message is that award wages can reasonably be regarded as endogenous and dependent on price and productivity movements in particular. But the timing of some of the adjustments may be somewhat exogenous, which allows the possibility of some 'cost push' in the short run.
A recent survey of a number of articles on wage and price determination in Australia (and New Zealand) concludes:
' . . . the available econometric evidence points to the important role of cost pressures, mainly domestic, in our inflationary processes, but it also recognises a subordinate role for demand pressures, again mainly domestic'.9
This is puzzling. The main evidence for the dominance of 'cost push' elements comes from a simulation of an award wage increase in the Treasury model,10 despite a very important caveat which Higgins makes and Simkin acknowledges:
'These results depend on an equation for award wages which takes major award decisions as exogenous, although [quoting Higgins] 'the decisions of tribunals are not independent of short term economic conditions, particularly price changes and capacity to pay'.11
If awards respond to price and productivity changes one cannot infer from simulation analysis of increases in award wages that 'cost push' is an important source of inflation.
The relegation of 'domestic demand pressures' to a subordinate role, despite considerable emphasis on such factors in the papers surveyed is also somewhat puzzling. And the article discusses 'domestic' demand pressures, although Australian (and, one suspects, New Zealand) demand pressure has been considerably influenced by external factors. This possibility is recognised in principle in an earlier section of the survey, which states:
'It is, of course, possible that ['Quantity Theory'] influences may emerge as indirectly important through the workings of the complete macro-model, but that cannot be decided until simulation experiments have been conducted and published'.12
But the available macro-models clearly allow for some of the usual channels for the international transmission of price and activity fluctuations. For instance, in these models domestic demand pressures rise in response to an increase in exports, or decrease in imports that could be caused by a greater rate of inflation in the United States. There is also the more direct influence of the price of traded goods on the price level. Moreover, increased capital inflow, in response to tight Australian monetary policy and speculation on the possibility of a revaluation of the Australian dollar, which could result from a strengthening of the current account, also has an expansionary influence.
It is worth pointing out that the authors of none of the studies surveyed stress the possibility of their 'demand pressure' variables being dependent on external factors. By contrast, in the more limited business cycle literature in Australia, the role of the rest of the world is stressed far more. A recent study lists among its conclusions:
'The behaviour of the world economy was highly influential upon that of Australia throughout the whole of this period'.13
And in a discussion of 'Options for External Economic Policy' in May 1972, the speaker said that he would be surprised if 'at least some of our inflation could not be attributed to external effects,'14 since monetary policy had been less restrictive than it could have been because of the possibility of encouraging even more capital inflow, and because inflation and exchange rate changes abroad had raised the price of imports.
With a fixed exchange rate, a small open economy can hardly hope to avoid being directly influenced by economic events in the rest of the world. Superficially, the importance of award wages in the determination of average earnings15 suggests that award wage decisions play an important role in the domestic inflation process. But, as noted above, award wages do respond to current economic conditions. More importantly, even if the usual awards relationship is subjected to an exogenous shock, as some would argue occurred in 1971, which produces an increase in Australian prices greater than those in the rest of the world, this would be ultimately self-correcting via the balance of payments constraint, given fixed exchange rates.
3. Single Equation Evidence
(a) A Seven Country Study of Wages
A recent important article16 on the 'world wide wage explosion' provides some empirical support for the Johnson argument. Nordhaus constructed alternative wage equations for seven countries: Canada, France, West Germany, Japan, Sweden, the United Kingdom and the United States. For each country he tested equations constructed according to a monetarist argument, a frustration theory, an export constraint model and naive and 'expectations' Phillips curves. The most successful equation for the United States, and possibly Canada, was the expectations Phillips curve; for the remaining countries the single most successful equation was that constructed according to the export constraint model. He notes that the wage explosion first occurred in the North American countries and was followed by inflation elsewhere and observes that if the United States follows a passive balance of payments policy and other countries use demand management to maintain external balance, 'the United States will determine the world inflation rate'.17
Nordhaus is painting with a broad brush, and concedes that more detailed studies for each country would possibly produce 'better' equations and reveal other influences at work. In particular, the role of domestic prices in wage determination in the United Kingdom is well documented.18 At the level of generality by Nordhaus, more detailed studies seen as attempts to spell out the links in the transmission mechanism.
(b) A 'World' Phillips Curve
An alternative approach is even more aggregative than that of Nordhaus. Noting the difficulties of explaining inflation in open economies and that an increasingly integrated world is fast becoming the only perfectly closed economy,19 four British economists seek to develop a world Phillips curve.20 Their basic theoretical model is the expectations Phillips curve which we have already seen explains United State inflation best in Nordhaus' study. Their price, wage and unemployment data are basically weighted averages of those variables for the Group of Ten, and 'price expectations' based on a weighted average of past price changes, calculated according to a generalized error learning model based on optimal forecasting behaviour.21 The resulting wage equations are highly promising, with unemployment and 'price expectations' highly significant and together explaining about two thirds of variance in the rate of change of wages.
These results are highly suggestive and lend strong support to the general 'world expectation. plus demand pull' theory set out in the introduction.
(c) The Transmission Mechanism in Australia
A recent more detailed Australian study22 lends support to the above conclusions by finding the rate of change of average earnings to depend on labour market conditions which, in turn, are partly a function of import prices, foreign reserves, productivity changes and award wages. The rate of change of reserves may represent a liquidity effect but more likely, is important because both employees and employers regard the balance of payments as an important indicator of the state of the economy.
Award wages in turn depend on price and productivity changes. Award wages are influenced by the national wage decision, metal trades margins decisions and agreements, other tribunals' decisions and other consent agreements. The national wage cases provide key decisions, and appear to have adjusted for price changes and (less fully) productivity changes in the normally annual increases. There also appears to be a less frequent ¡°major productivity adjustment¡± every 5 years or so. More specifically, there were significantly larger than normal increases in the 1959, 1964 and 1970 decisions and there is some evidence in the Commission judgements to suggest that these are longer run productivity adjustments.23 The last of these increases may also have been influenced by the strong reserves position in that year.
The awards equation presented by Jonson and Maher is:
J4P.WA = 1.057 + 0.345 J4P.CPI
+ 0.224 JXP.CPI + 0.560 JYP.CPI
+ 0.365JXP.PRD + 0.1 74JZP.PRD
- 4.64 QCB + 0.159 J4P.WA-1
S.E.E. = 0.830 R 2 = 0.918 D = 3.79
To explain the notation briefly: J4P is an operator notation denoting an annual percentage change, WA denotes award wages, CPI the consumer price index, and PRD a measure of aggregate productivity. JXP is an operator notation indicating that the variable is the change in prices (or productivity) since the last national wage increase, and JYP.CPI denotes the annual percentage change in the consumer price index over the year prior to a metal trades margins increase. JZP.PRD denotes the percentage increase in aggregate productivity since the last major productivity adjustment in the national wage; QCB is a dummy variable with the value unity in the quarters influenced by the 1965 national wage case and zero otherwise, to represent the influence of the composition of the bench in that year. Chart A below illustrates the actual values of changes in award wages over the equation's sample period, (1960(l) - 1972(2)), and also the values estimated by the equation. The line of dashes shows the increase that could have been expected in 1971 if the 'major productivity adjustment' had not been made. (The value of JZP.PRD in each quarter of 1971 is multiplied by the coefficient on JZP.PRD and deducted from the estimated value of award wage changes, i.e. JZP.PRD is set equal to zero in 1971).
To the extent that the timing of the larger than normal award increases are exogenous, those economists who attribute our recent inflation in part to 'wage push' have a point. As illustrated in Chart A, the awards equation suggests that the rate of increase of awards was almost four percentage points higher than 'normal' in 1971.
click to enlarge
4. A Whole Model Test
Single equation results are often suggestive, but are rarely as convincing as a 'whole systems' test using a well specified theoretical or empirical macroeconomic model. Thus some simulations of the RBA1 model of the Australian economy24 have been constructed to try to indicate the relative importance of imported and wage inflation in our recent experience.
The type of simulation analysis used here examines the influence of changes in some exogenous var iables on the endogenous variables of the model. The model is first solved using actual values of the exogenous variables and generated values for the lagged endogenous variables (technically, with a non-stochastic, dynamic simulation). Comparing the 'control solutions' for the endogenous variables with their historical values provides one measure of the performance of the model, and the revised version of the model used in the following tests, like its predecessor, performs well in this respect.25 This 'control solution' provides the benchmark with which to compare the solution from other simulations which use different values for some exogenous variables.
(a) Imported Inflation
Among the variables that transmit overseas inflation (or more generally, variations in activity) to Australia are private capital inflow, import prices and exports. These are exogenous in the model used and increases in them have the expected expansionary (inflationary) impact partly 'monetary', partly 'real'. Chart B illustrates the actual values of the volume of exports from 1958(3) to 1973(2). The dotted line is the trend plus seasonal component estimated for the period to 1969(2) which, when compared with the actuals, illustrates the considerable upward divergence from 1969(3). There is also a very similar rise in the value of exports, and similar, although sharper, rises in private capital inflow and import prices with an abrupt halt in the rise in capital inflow in the final quarter of 1972. It is argued here that a large part of these increases can be attributed to overseas inflation, although empirical evidence is not over-abundant. click to enlarge
In May 1972 it was pointed out that: 'There can be few people in the world who are anticipating a decrease in the value of the $A relative to ¡êSterling or $US in the forseeable future'.26 Snape judged that the expectation of a revaluation of the SA more than compensated for any interest rate differential, and the resulting increase in capital inflow was the most important factor in the large external surplus.
In a recent paper27, M. G. Porter presents econometric evidence on the relative importance for capital inflow of speculation on a revaluation of the $A, the stance of monetary policy, the current account balance and changes in income. He stresses the inconsistency of attempting both to peg the nominal exchange rate and to assign monetary policy to domestic stabilization,28 which 'served both to trigger speculative capital inflow and to destabilize the money supply .'29. The trigger for the speculative capital inflows in his equation is the state of the balance of payments, and this was built up by the improvement in the current account as well as the tight monetary policy in 1970 designed to offset the expansionary effects of the strengthening current (and then capital) account. The recent exchange rate changes and controls on capital inflows have had an important short run influence on capital inflow, and by March quarter 1973 the reduction of capital inflow well below trend has begun to somewhat offset the continued growth of exports.
Although little recent analytic work on import or export determination appears to have been done, the improvement in the current account must be largely attributed to the world inflation. General theoretical considerations may suffice for most, but some additional points are worth making. A casual examination of the annual value of exports divided into the four broad categories 'wool and sheepskins', 'other rural', 'ores and concentrates' and 'other' in Table 1, for each recent financial year is interesting. 'Other rural' shows the most consistent increase, followed by 'ores and concentrates' with 'other' close behind. 'Wool and sheepskins' fell over the period, an exogenous offset to the miner al boom, which in any case presumably did not greatly influence 'other rural'. click to enlarge
had an important short run influence on capital inflow, and by March quarter 1973 the reduction of capital inflow well below trend has begun to somewhat offset the continued growth of exports. Although little recent analytic work on import or export determination appears to have been done, the improvement in the current account must be largely attributed to the world inflation. General theoretical considerations may suffice for most, but some additional points are worth making. A casual examination of the annual value of exports divided into the four broad categories 'wool and sheepskins', 'other rural', 'ores and concentrates' and 'other' in Table 1, for each recent financial year is interesting. 'Other rural' shows the most consistent increase, followed by 'ores and concentrates' with 'other' close behind. 'Wool and sheepskins' fell over the period, an exogenous offset to the mineral boom, which in any case presumably did not greatly influence 'other rural'. The size of the shocks applied can perhaps be illustrated best as percentages of the actual values in 1964-68, every fourth quarter. This is done in Table 2, which also illustrates the assumed equivalence of the two time periods. click to enlarge
(b) Wage Inflation
As discussed above, there is some weight to the argument that the 1971 increase in award wages was 'larger than usual', and that this represented an inflationary shock to the system. Thus an exercise similar to that outlined in section 4(a) can be carried out by imposing an 'award wages' shock in 1965. Thus it was assumed that the increase in award wages in 1965-66 was 3 percent higher than the historical value, a slightly smaller absolute shock than the 'productivity adjustment' of 1971, but, relative to the historical values of award increases, larger than imported shocks.
(c) Increase in Cash Benefits
In revising this work it was noticed that cash benefits have also grown strongly above their previous trend from 1970. Arguments analogous to the above suggest that this is another exogenous influence on our recent inflationary take-off, and so the effect of equivalent increases in cash benefits has been tested. Cash benefits were raised above their historical trend in each quarter, e.g. by 9.3 per cent in quarter 4, 17.9 per cent in quarter 8, 25.5 per cent in quarter 12 and 32.3 percent in quarter 16. It can be argued that the rise in cash benefits was at least partly a response to the exceptionally strong balance of payments at the time.
(d) Other Recent Changes
The 1972 Budget included a sizeable reduction in personal income tax rates, and this is having some influence by the end of the sample period. Press comments have discussed extensively the likely spending increases as the new government introduces its policy changes but so far increased government spending has been small. Thus there is only a small above trend shock of approximately 2.5 per cent to current government spending imposed for the last two quarters of the sample period.
(e) An important Caveat
The simulation results are only useful if the model provides an adequate approximation to the workings of the Australian economy. While this will always be debatable, one particular problem must be mentioned. In the simulations reported, the macroeconomic policy instruments in the model, for example tax rates, interest rates and government spending are left at their historical values. There is however quantitative32 as well as literary33 evidence that monetary and fiscal authorities react in fairly predictable ways to changes in targets for stabilization such as unemployment, prices and international reserves. Hence the results have to be interpreted in the light of the present treatment of policy weapons as exogenous in the RBA1 model.
(f) The Model Used
The model is closely related to that described in Norton and Henderson34. The main change is the replacement of their equation 14, which explained the level of average weekly earnings, by the version developed by Jonson and Mahar35. As noted above, the new equation contains award wages as an explanatory variable and in view of the above discussion on the endogeneity of awards this required the inclusion of an equation to explain award wages36. Limitations on computer core space required the dropping of an equation, and we chose the equation for other direct capital inflow, a component of private capital inflow. Since the tests involved imposing exogenous changes in private capital inflow, this did not seem to be an important omission. More minor changes included the replacement of the equation for government securities by an alternative developed recently.
(g) The Empirical Results
Each of the following tables presents the results of imposing the external shocks, the wages shock, the cash benefits, shock, then all shocks together37 on the model. For simplicity, only the results for five of the model's eighty endogenous variables are presented but these are among the more important macroeconomic magnitudes. It should be noted that the figures reported are the changes in the levels of these endogenous variables caused by the shocks imposed. An important difference to the results reported at the Conference of Economists is that the shocks are continued on to include events from the very recent past so that inferences can be made about our current inflation and not just about the historical experience of 1970-71. Table 3 sets out the influence of the shocks on selected endogenous variables, 4, 8, 12 and 16 quarters from the start of the shocks. click to enlarge
These results are quite devastating for the cost-push hypothesis; in quarter 8, the period of maximum impact from the wages shock, there is 0.52 per cent. added to the rate of inflation by that shock; this is less -than a quarter of the 2.25 per cent added by the imported shocks. Together with the additional 0.31 per cent from the above trend rise in cash benefits,38 the shocks add over 3 percent to the rate of inflation in quarter 8. With the small contribution from the other shocks discussed above, over 4 per cent is added by quarter 16. This seems to go a long way towards explaining by analogy recent and indeed current experience.
The relative influence of each shock on the rate of inflation is summarized for quarters 8 and 16 in Table 4, and this shows clearly that even by the end of the sample period (quarter no. 16 is equivalent to 1973(2)) when the various recent external policy changes were having some impact on imported inflation, the most important direct effect on our inflation rate is still the world inflation, even apart from possible indirect effects on the policy instruments. And of these, a larger than usual increase in government spending is not in any sense an important contributing factor, some recent press comment notwithstanding. click to enlarge
The results of course depend on the specification of the model, and can thus be used to illustrate features of that model. One interesting result is the rise in unemployment in the wages shock simulation, despite a rise in gross domestic product. This results from a rise in wages relative to prices. In the RBA1 model employment, and hence unemployment, depends on the relative price of labour as well as real gross domestic product, and the wages simulation illustrates the strengths of the former effect and provides empirical support for a proposition argued by Boxall and Stammer in a recent paper.39 This effect can perhaps help explain our recent experience of 'shifting Phillips curves', although Table 3 illustrates that when all the shocks are imposed together unemployment falls, although not dramatically until the end of the sample period.
A complete analysis of this issue would have to incorporate several refinements, which were not feasible for this paper, but two relevant points about the model's specification should be made. Firstly, the model does not incorporate policy reaction functions, and thus the simulations do not allow for the deflationary policy response that also occurred in recent experience. Allowance for this would produce less inflation and more unemployment in the results. A second point is that the price equations in the RBA1 model exhibit long adjustment lags, which may be partly a consequence of the relatively small change in prices in the estimation period (1959-69); whatever the reasons, the slow adjustments of prices mean that, in the short¨Crun, large increases in demand have rather more influence on employment than in raising prices than seems plausible. Thus the extra inflation in quarter 16 is only a little more than 4 per cent per annum, although the actual four quarter span rate of inflation in 1973(2) was just over 8 per cent, about 5 per cent above the average for the 1960s.
In the model, wages adjust more quickly than prices, which is one reason for a rise in real wages in all simulations. Another is the direct influence of changes in reserves in the earnings equation; the coefficient on reserves in earnings equation used is however only 0.023 which with (say) a 50 per cent annual increase in reserves adds 1.15 per cent to the annual rate of change of earnings in the short-run. As the short-run weights on earnings in the various price equations are small (and their adjustment lags long) changes in reserves provide an additional reason for a rise in real wages in the 'imported inflation' and 'all shocks' simulations. It is not, however, a major reason for the primacy of imported inflation in the results as there are several other more important channels for the impact of external factors on prices in the model.
5. Conclusion The whole model test reported above suggests that Australia's current inflation owes considerably more to foreign influences than to 'domestic cost pressures'. This reinforces the introductory theoretical points, and single equation evidence, which argue for a generalized 'world demand pressure' explanation of the recent world wide inflation, and its applicability to Australia. This does not entitle us to blame the foreigners for our inflation; we could always have floated, or successively revalued our exchange rate. And it does not establish the proposition that our inflation is independent of domestic wage and demand policies. A much 'larger than usual' national wage increase, for example, adds to inflation in the short-run, as does a more rapid expansion of domestic credit. What happens in the longer run depends crucially on what happens to the exchange rate: if it is kept fixed Australian inflation must come into line with the world average; if the exchange rate is devalued we will have an even greater rate of inflation, and if we appreciate we will have a lower rate. More specifically, exchange rate flexibility is a necessary but not sufficient condition for inflation control. If we accept a freely floating exchange rate, our exchange rate (and our inflation rate) becomes a function of our domestic policies, and in particular of our monetary policy, relative to those of other countries.
This article was published in Australian Economic Review, Second Quarter, 1973.