RBA Annual Conference – 1992 Discussion
1. Arthur Grimes
Doug McTaggart's survey on the costs of inflation has drawn together a quite extraordinary wealth of information on the subject. It represents the most complete survey that I have seen on this issue – both at a general level and with specific reference to Australia.
Coming from a central bank with a specific mandate to maintain price stability, I welcome such a listing and enumeration of the costs of inflation and inflation variability. The Reserve Bank of New Zealand has conducted a number of research projects on this subject and, in general, the findings of this research are in accordance with those of McTaggart.[1]
I find it difficult, therefore, to comment on a paper which has been well prepared and with which I am in significant agreement. Nevertheless, I wish to address four issues arising from McTaggart's paper. These relate to (i) the use of the production function as the organising principle of the paper; (ii) the paper's econometric work; (iii) some potentially spurious arguments; and (iv) some extra evidence not included in McTaggart's paper.
(a) The Production Function Approach
The paper is organised around a standard production function where output (Q) is a function of labour (L) and capital (K). The choices of both labour and capital inputs may be affected by the level of inflation (π) and its variability (σπ) Thus he specifies:
This much is unobjectionable. However, McTaggart, on occasions, appears to assert that all costs of inflation can be addressed within this framework.[2] But that is certainly not the case. There can be welfare costs of inflation even in the absence of any changes to the levels of output or productive inputs. Take the example, discussed by McTaggart, of the effect of inflation on labour supply decisions arising from the lack of perfect indexation of tax scales. It may well be that the income and substitution effects of an increase in the real income tax rate on labour supply perfectly offset each other. One will therefore observe no change in labour input or output as a result of this factor. But there is a welfare cost, since it is the substitution effect (i.e. the distortion of price signals) that is relevant in the micro-economic welfare analysis.
Even within the production function setting (i.e. ignoring this type of welfare cost) one must be careful in interpreting the costs of inflation. Following his specification of the production function, McTaggart contends that the costs of inflation arise when any of the following hold:[3]
But, in an otherwise undistorted economic environment, the sign of these partial derivatives is irrelevant. If any are positive or negative, there will be a cost of inflation. For instance, consider the Tobin effect (Tobin (1965)), in which an increase in inflation causes an increase in the capital stock at the expense of a decrease in real money balances.[4] The mechanism here is that an increase in inflation causes the real rate of return on (non-interest bearing) money balances to decline, so causing a fall in real money balances and a switch to other portfolio items, including physical capital. As a result, the capital stock and output increase.
This increase in the capital stock and output must be considered a cost of inflation. The increase in inflation causes people to hold less than the optimal level of transactions balances, so increasing the cost of conducting trade. Indeed, in this respect, it remains an open question as to whether zero inflation is costly relative to a small negative inflation rate for the reasons outlined by Friedman (1969) in considering the optimum growth rate of money.
McTaggart occasionally notes such costs,[5] and so is apparently aware that any deviation caused by inflation from the zero inflation allocation of resources is costly. The problem is that it is difficult to reconcile these types of costs with the use of the production function as the paper's organising principle.
(b) Econometric Work
A number of problems arise with the econometric and related work conducted in the paper. I will concentrate on two main areas.
(i) Inflation and Inflation Uncertainty
McTaggart derives a measure of inflation uncertainty from a forecasting equation for inflation (reported in Table 1 of Appendix B of his paper). The difficulty with this approach is that any mis-specification of the inflation equation will affect the measure of uncertainty. In this case, the inflation equation appears mis-specified. In the steady state (assuming constant growth rates of all variables), McTaggart's inflation equation yields:
where INF is the inflation rate, GDPG is real GDP growth, and M3G is the growth rate of the M3 money stock. The coefficient on M3G is substantially different from unity, while the coefficient on GDPG is of the wrong sign (in a quantity theory of money context). Durbin's h statistic, which is a more accurate measure of first-order autocorrelation in the presence of a lagged dependent variable than the reported Durbin-Watson statistic, cannot be calculated for the equation as it involves a complex root; while not conclusive, this suggests an autocorrelation problem as well. These problems make the paper's estimate of inflation uncertainty unreliable.
Even if it was reliable, care needs to be taken in interpreting this measure. McTaggart notes the estimated high inflation uncertainty in the 1970s, but this seems to ignore his own call for care when dealing with the effects of supply shocks – as shown also by Fischer (1981, 1982). It is quite likely that the high inflation ‘uncertainty’ in the mid-1970s is purely an artifact of ex post inflation forecasting errors in the face of the oil shocks. They do not necessarily reflect any increase in ex ante uncertainty.
(ii) Inflation, Real Wages, Investment and Output
The econometric approach adopted in the paper to judge the impact of the level and variability of inflation on real wages is suspect. It is noted that the real wage (RWAGE) and unemployment (UNEMP) variables are integrated processes, while presumably the level of inflation and the variability of both inflation and unemployment are stationary. Immediately this suggests that McTaggart should test whether RWAGE and UNEMP are cointegrated; if they are not, then his estimated levels equation is not statistically valid. If they are cointegrated, then the differenced equations suffer from omitted variables bias and should be ignored, and the levels equation (which in that case would only include RWAGE and UNEMP) should be supplemented by an error correction equation. In addition, when dealing with integrated variables (even if they are contemporaneous and jointly determined), OLS is preferable to instrumental variables (which McTaggart uses) as an estimation technique.
Perhaps as a result of these issues, some of the results in this area are odd. For instance, the equations indicate that the change in unemployment has a positive influence on the real wage, which is counter-intuitive.
Similar econometric problems (relating to the use of integrated and stationary variables) exist with the work examining the effects of the level and variability of inflation on investment and output in the paper. These may account for the surprising positive estimates for the impact of the variability of inflation on both investment and output.[6]
(c) Some Potentially Spurious Arguments
McTaggart's list of the costs of inflation is so comprehensive that it is not surprising that one will not agree with his judgements on all issues. I note four such issues.
Firstly, McTaggart refers to the effects of a non-uniform inflation in which the prices of some variables rise faster (or slower) than the aggregate price index. I would characterise this phenomenon as the joint occurrence of a uniform inflation and of relative price changes. In the absence of inflation, these relative price changes would still be occurring and so their effects should not be interpreted as a cost of inflation. The only exception would be if the process of inflation caused the relative price changes to occur.[7]
Secondly, reference is made to a negative correlation between inflation and the stock market value of firms. As McTaggart notes, this effect is well documented and is not surprising if inflation does indeed impose real economic costs. However, there is some counter-evidence which could be considered. Kaul (1987) finds that this negative correlation exists in the post-War era in a number of countries, but that this is not the case prior to World War Two. In the earlier era, the correlation is either positive or insignificant. Kaul hypothesises that the differing signs reflect different monetary regimes, rather than anything more fundamental. In particular, if the monetary authorities adopt a counter-cyclical monetary policy, expected activity, and hence stock market prices, may well fall when inflation is rising; whereas a pro-cyclical policy response will have the opposite effect. Thus differing monetary policy regimes before and after the war may be driving the relevant correlations. This does not necessarily deny that inflation is costly and that these costs are reflected in stock prices, but it does counsel caution in interpreting the correlations.
Thirdly, McTaggart twice notes the possibility that the real exchange rate may be affected by the rate of inflation. Inflation can undoubtedly affect the real exchange rate in the short run, and popular commentaries frequently stress this issue. However, it is not clear whether competitiveness will initially be improved or worsened by an increase in inflation; this will depend also on the behaviour of the exchange rate. In the longer term, it is difficult to ascertain why inflation may be detrimental to competitiveness, except through indirect channels such as the effect of inflation on productivity.
Fourthly, there is one area in which I consider McTaggart has underestimated the costs of inflation. He downplays the costs associated with the redistributive effects of inflation. Unlike explicit tax policies, which generally involve a conscious decision regarding distributional considerations by policy-makers (reflecting their social welfare function), inflation causes unforeseen and unintended redistributions. The costs of such redistributions do not arise solely as a result of individuals spending resources directly to counter the effects (which is the criterion McTaggart adopts to assess whether costs exist in this case). They also arise because other government policies may have to change as a result. For instance, to the extent that inflation imposes relatively greater costs on poorer members of society, the income tax system may have to be made more progressive to offset this effect. The result is greater labour supply distortions which impact even on people who may be unaffected by the initial redistributive effects of inflation.
At a more general level, the redistributive costs of inflation can be likened to the redistributive effects of crime – recalling that debasing the coinage was once a capital offence![8] If no-one were to direct resources to preventing crime (perhaps because of a lack of crime-preventing technology), would crime impose costs on society? All theft in that case would be a pure redistribution. But, almost inevitably, people would argue that this form of redistribution was costly for society, and I see little difference in the case of inflation.
(d) Additional Evidence
McTaggart cites an extraordinary compilation of evidence on the costs of inflation. To this can be added the recent work of Gylfason(1989, 1991a, 1991b). He demonstrates (in 1991b), both theoretically and within a simulation model, how inflation can impact on capital accumulation so as to adversely affect the economic growth rate. Normally, in a neo-classical growth model (exhibiting constant returns to scale) the equilibrium growth rate is determined by the rate of population growth and the rate of technological progress; both factors being exogenous. Hence, the inflation rate cannot affect the long-term growth rate in such a model. But Gylfason shows in an endogenous growth model exhibiting increasing returns to scale – arising from the technology being a function of the size of the capital stock – that if inflation impacts negatively on capital accumulation, then it will also impact negatively on the long-term growth rate.
Gylfason (1989) presents cross-sectional evidence (using data from the World Bank's 1988 World Development Report) that annual growth in per capita GDP in economies with high inflation (i.e. inflation of 20 percent or more per year) was, on average, 3.4 per cent lower than in economies with low inflation (i.e. those with inflation below 4 per cent per year) during 1980–86 (the period covered by the World Development Report). This difference was statistically significant. Reddell (1990) has extended Gylfason's analysis. He demonstrated that the middle group of countries (i.e. those with average inflation rates between 4.0 and 19.0 per cent) had average growth rates almost exactly halfway between the growth rates of the high- and low-inflation countries. And when the countries were divided into four groups, based on their state of economic development, the same pattern emerged.
These results are in keeping with my own findings (Grimes (1991)) that price stability appears to be beneficial for economic growth. That paper analysed the impact of inflation on growth after taking into account the effects of disinflation and supply shocks. It utilised a pooled cross-section, time-series approach covering 21 industrial countries over 27 years. Intriguingly, it estimated that the inflation rate which maximised the growth rate was 0.98 per cent, with a confidence interval of 0.06–1.90 per cent, almost identical to the target inflation range of 0–2 per cent set for the Reserve Bank of New Zealand!
Footnotes
See Carey (1989), Reddell (1990), Greville and Reddell (1991) and Grimes (1991). [1]
For instance, in the Introduction, he states: ‘any real costs associated with inflation must, one way or another, derive from behaviour where one or more of the fundamental inputs into the aggregate supply equation, or production function, has been adversely affected.’ [2]
He states: ‘The various rigidities, structural non-adaptations and directly unproductive activities discussed above are all manifest in the negative first partial derivatives.’ [3]
Given the minuscule proportion of wealth that outside money constitutes, the Tobin effect is probably empirically irrelevant – indeed, Gale (1983) demonstrates that it may even be of the wrong sign. But at a theoretical level, the Tobin effect is a useful demonstration that an increase in the capital stock, and hence output, as a result of inflation may be costly in welfare terms. [4]
For instance, he states: ‘In the absence of macro-economic externalities, the increase in output should itself be regarded as a cost.’ [5]
In the latter area, this problem may also be due to the omission of secular technological progress from the production function which, despite McTaggart's claims, is not accounted for by the first order autoregressive error process. The use of a Cobb-Douglas specification may also be problematical, as a number of papers find the aggregate elasticity of substitution between capital and labour to be significantly less than unity (see, for instance, Grimes (1983)). [6]
This could potentially occur in the circumstances outlined by Archibald and Lipsey (1958) or by Hartley (1990). [7]
Macauley (1989) quotes a seventeenth century English writer on the consequences of (the then prevalent) private debasement of the coinage for the culprits: ‘Hurdles, with four, five, six wretches convicted of counterfeiting or mutilating the money of the realm, were dragged month after month up Holborn Hill. One morning, seven men were hanged and a woman burned for clipping.’ Hirst (1933) quotes Professor Thorold Rogers on the fate of kings who debased their coinage: ‘In his judgement the Kings of France were the worst offenders. Sometimes they diminished the weight of their silver coins; at others, they debased them by reducing their fineness. For this fraud upon his subjects, Philip the Fair was threatened with excommunication by Boniface VIII, and branded eternally by Dante.’ Hirst notes that Rogers also ‘attributes the weakness of France in the fourteenth century largely to the economic mischief caused by the monetary frauds of its kings’; while Macauley argues that ‘it may well be doubted whether all the misery which had been inflicted on the English nation in a quarter of a century by bad Kings, bad Ministers, bad Parliaments, and bad Judges, was equal to the misery caused in a single year by bad crowns and bad shillings’. [8]
References
Archibald, G.C. and R.G. Lipsey (1958), ‘Monetary and Value Theory: A Critique of Lange and Patinkin’, Review of Economic Studies, 26, pp. 1–22.
Carey, D. (1989), ‘Inflation and the Tax System’, Reserve Bank of New Zealand Bulletin, 52, pp. 18–26.
Fischer, S. (1981), ‘Relative Shocks, Relative Price Variability, and Inflation’, Brookings Papers on Economic Activity, No. 2, pp. 381–431.
Fischer, S. (1982), ‘Relative Price Variability and Inflation in the United States and Germany’, European Economic Review, 18, pp. 171–196.
Friedman, M. (1969), ‘The Optimum Quantity of Money’ in M. Friedman (ed.), The Optimum Quantity of Money and Other Essays, Aldine Publishing Co, Chicago.
Gale, D. (1983), Money: In Disequilibrium, Cambridge University Press Cambridge.
Greville, R. and M. Reddell (1992), ‘The Costs of Inflation’, in Monetary Policy and the New Zealand Financial System (3rd ed.), Reserve Bank of New Zealand, Wellington.
Grimes, A. (1983), ‘The Reserve Bank Real Wage-Employment Relationship: A Reply and Further Results’, New Zealand Economic Papers, 17, pp. 69–78.
Grimes, A. (1991), ‘The Effects of Inflation on Growth: Some International Evidence’, Weltwirtschaftliches Archiv, 127, pp. 631–644.
Gylfason, T. (1989), ‘Inflation and Economic Decline: A Coincidence?’, Skandinaviska Enskilda Banken Quarterly Review, 2, pp. 35–40.
Gylfason, T. (1991a), ‘Inflation, Growth and External Debt: A View of the Landscape’, World Economy, 14, pp. 279–297.
Gylfason, T. (1991b), ‘Endogenous Growth and Inflation’, Institute for International Economic Studies, Seminar Paper No. 502, Stockholm.
Hartley, P.R. (1990), ‘Hoarding in General Equilibrium’, Economica, 57, pp. 467–484.
Hirst, F.W. (1933), Money: Gold, Silver and Paper, Charles Scribner's Sons, London.
Kaul, G. (1987), ‘Stock Returns and Inflation: The Role of the Monetary Sector’, Journal of Financial Economics, 18, pp. 253–276.
Macauley, T.B. (1989), The History of England, Popular Edition, Vol. 2, Longmans, London.
McTaggart, D. (1992), ‘The Cost of Inflation in Australia’, this volume.
Reddell, M. (1990), ‘Some Evidence on the Costs of Inflation’, Discussion Note, Reserve Bank of New Zealand, Wellington.
Tobin, J. (1965), ‘Money and Economic Growth’, Econometrica, 33, pp. 671–684.
2. General Discussion
The discussion of McTaggart's paper focused on three aspects of the cost of inflation:
- the ‘moral’ dimension of inflation;
- the effect of inflation on the accumulation of both physical and human capital;
- the argument that the principal cost of inflation is the cost of actually disinflating.
With regard to the moral dimension, a few participants argued that inflation is akin to theft and, therefore, should be minimised. This interpretation did not, however, receive unanimous support. If inflation is fully anticipated there need not be any costs, let alone theft. Unanticipated inflation, like any unanticipated shock, necessarily implies ‘winners’ and ‘losers’. But there was a feeling that unanticipated inflation changed the essential nature of contracts and, therefore, should be minimised. Some participants felt that a distinction should be made between ‘theft’ and a tax. Inflation was more akin to a tax on money balances and not theft. However, most calculations of this ‘tax’ typically showed that it was relatively small compared to the size of taxes in the government's budget.
It was more widely agreed that the main costs of inflation were likely to concern its impact on economic activity. Here discussion focused mainly on investment and capital accumulation. As depreciation allowances are in nominal historical terms, inflation causes a deterioration of company cash flow and investment. However, a number of participants took issue with the ‘back-of-the-envelope’ calculation provided by McTaggart, which used estimates of this negative cash flow effect on investment in a simple production function to show that a sustained 1 per cent rise in inflation in 1970 would have resulted in a 5 per cent loss of a year's real output by 1991. According to this methodology, the average 7 per cent inflation over this period resulted in a 35 per cent output loss, compared to a situation of zero inflation. Some felt that the partial nature of the exercise, where other things were held constant, was inappropriate because it would exaggerate the cost of inflation. In particular, the cost of capital was not included in McTaggart's investment equation. Since inflation leads to a preference for debt finance, and financial deregulation may have reduced the relative importance of internal cash flow in funding investment, failure to take account of these factors may have led to an over estimation of the effect of inflation on the capital stock.
Nevertheless, there was widespread support for the proposition that inflation had adverse consequences for business investment. Its effects on the structure of the capital stock (rather than its level) were felt to be particularly pronounced. For example, inflation was seen to encourage property investment rather than investment in plant and equipment. It was recognised, however, that these effects are difficult to quantify.
Finally, it was argued by some participants that the principle cost of high inflation comes as a result of the eventual desire by policy makers to disinflate. This disinflation necessarily requires movement along the short-run Phillips curve, which can be quite costly in terms of lost output and employment. The point here is that some of these losses may be permanent. With regard to the labour market, there is the loss of human capital in the recession. With regard to investment, higher real interest rates and lower investment might permanently reduce the capital stock and, thus, potential output per capita.