George Selgin, 1997.
The Case for Zero Inflation
Several obstacles stand in the way of instantly-equilibrating general price changes. First among them are fixed money contracts that cannot easily be ‘indexed’ to general price movements. Such contracts include both wage contracts and nominal debt contracts, the most notorious of which is the government’s ‘contract’ offering holders of high-powered money balances a fixed, zero nominal rate of interest. Second, ‘menu costs’ and other expenses involved in posting and sometimes negotiating new money prices can make the price level ‘sticky’ in the short run. 14 Finally, sellers may be reluctant to change, and especially to lower, their prices in response to monetary disequilibrium even when the fixed costs of doing so are very small. Some analysts (e.g. Okun, 1980, pp. 145ff.) link this reluctance to the inelastic demand for products of firms whose customers face high shopping costs. Yeager (1986, p. 377) attributes it, in part at least, to the fact that money, ‘unlike other goods, lacks a price and a market of its own’. This fact makes any equilibrating price level change something of a public good:
‘Money’s value (strictly, the reciprocal of its value) is the average of individual prices and wages determined on myriads of distinct though interconnecting markets for individual goods and services. Adjustment of money’s value has to occur through supply and demand changes on these individual markets.’
New Keynesian writings also assign a crucial role to what they call ‘aggregate demand externalities’ as a source of sluggish price adjustment. According to Ball and Mankiw (1994, p. 18),
‘The private and social gains from price adjustment [following a negative money shock] are very different. If a single firm adjusts its price, it does not change the position of its demand curve; it simply moves to a new point on the curve. This adjustment raises profits [not taking menu costs into account], but the gain is second order. In contrast, if all firms adjusted to the monetary shock, the aggregate price level would fall, real balances would return to their original level, and each firm’s demand curve would shift back out. … Unfortunately, an individual firm does not take this effect into account because, as a small part of the economy, it takes aggregate spending and hence the position of its demand curve as given. Thus firms may not bother to make price adjustments that, taken together, would end a recession.’ 15
Sluggish price adjustments are also likely to be uneven, with some prices adjusting ahead of others, so that equilibrating price-level movements typically involve temporary alterations of relative prices. Monetary theorists going as far back as Richard Cantillon and David Hume have understood that the relative price effects of any money supply shock depend on the monetary ‘transmission mechanism’ – that is, on the precise way in which nominal money balances are added to or subtracted from the economy. In fact, both money supply and demand shocks first make their presence felt, not in all markets at once, but in particular markets from which their effects slowly spread to the rest of the economy (Yeager, 1996a). Clark Warburton (an ‘Old Monetarist’) discusses the case of a positive money supply shock:
‘The first change occurs at the point where the additional money is introduced into or taken out of the economy and is expressed in an increased or decreased demand for the goods and services desired by the persons directly affected by the change in the quantity of money.’ ( 1951, pp. 298-99)
Consider an unexpected round of central bank open-market purchases. The purchases ‘inject’ new high-powered money directly into the bond market, raising the value of government securities. The high-powered money quickly makes its way into commercial banks, who use it to make more loans, at lower rates. 16 Borrowers use the loans to purchase labour, capital goods, and durable consumer goods. Eventually an overall rise in spending raises the general price level, eliminating what had been a surplus of money balances. In principle, short-run monetary ‘injection’ effects can temporarily alter relative prices even if all money prices are quite flexible.
Temporary, relative price changes connected to bouts of monetary disequilibrium introduce ‘noise’ into money price signals, and thus ‘degrade the information conveyed by individual prices’ (ibid., p. 374). Businessmen, workers and consumers rely on this degraded information (because it is better than nothing), and end up wasting resources. The quote from The Economist (page 14 above) makes this very point. Monetary disturbances have real effects, not just because of the time it takes for the price level to adjust, but also because of the devious path taken by individual prices during the adjustment process.
Finally, changes in the overall price level of the sort needed to eliminate monetary disequilibrium can themselves promote ‘unnatural’ changes in real economic activity: economic actors may confuse general price changes with relative price changes, either because they suffer from ‘money illusion’ (a genuine failure to consider the meaning of general price changes) or because they only observe local price movements and infer (imperfectly) what is happening to prices in more far-removed markets. One frequently offered scenario of monetary expansion has workers reacting to higher money wage-rates while overlooking changes in the ‘cost of living’, so that employment rises (temporarily) above its natural or full-information level. Implicit in such scenarios is the assumption that changes in real money demand or nominal money supply, and consequent changes in the price level, are not perfectly anticipated by economic agents: while workers or consumers might easily anticipate steady, long-term trends in the equilibrium price level, they are likely to be surprised by, and fail to recognise, random changes. Nor would complete knowledge of the schedule of changes in the nominal money-stock (assuming such knowledge could be had) be sufficient to avoid price-level surprises, unless the public could also make precise forecasts of future changes in real money demand. It follows, then (according to zero inflationists), that the surest way to avoid money illusion is to avoid changes in the price level altogether.
II. PRODUCTIVITY AND RELATIVE PRICES
Superfluous and Meaningful Changes in the Price Level
Consider first an example of a genuinely superfluous change in the price level. Imagine an economy where both the supply of various factors of production and the productivity of those factors (and hence, real output or income) are unchanging. Imagine also that the real demand for various goods and services, apart from money, is unchanging. In such an economy, a change in the general level of output prices can occur only as a result of some change in the nominal quantity or velocity of money, leading to a change in the overall demand for final goods and services, that is, in aggregate spending or ‘nominal income’. A central bank might, in principle at least, manage the stock of money so as to prevent such changes in nominal income, thereby keeping the price level constant. By assumption, consumer preferences and technology are not changing, so that the only information conveyed by any price level movement is information concerning the central bank’s failure to maintain a stable value of nominal spending. […]
If an economy with constant productivity is like a Baroque organ fugue, an economy with changing productivity is more like a Romantic symphony. In the latter sort of economy, movements in the general price level may form a meaningful component of the ‘tune’ being played by money price signals: higher, ‘louder’ price signals can convey a message of fallen productivity and greater all-around scarcity (a higher price of output relative to inputs), while lower, ‘softer’ ones can convey a message of greater abundance (a lower price of output relative to inputs). Trying to improve an economy’s performance by stabilising the price level in the face of changes in productivity is – I plan to argue – like trying to improve a symphony by adjusting the volume knob so that the majestic finale plays as softly as the sombre adagio.
The Productivity Norm and ‘Menu’ Costs
Let us first consider whether the overall burden of money price adjustments would be greater or smaller under a productivity-norm regime than under a zero-inflation regime. The regime that faces higher overall price adjustment or ‘menu’ costs will, presumably, be more prone to temporary relative price distortions. 19 […]
Imagine an extreme case where a change in productivity affects the output of only one good. … Suppose, for example, that 1,000 final goods are produced using three distinct factors of production. A technological improvement causes an outward shift in the supply schedule for good x, so that the quantity of good x producers would be willing to supply at any given price is twice the previous quantity. Suppose also that x formerly had a price (included in the price index) of one dollar per unit. Under a productivity norm policy, the monetary authorities do not adjust the quantity of money in response to a productivity shock, so that, with an unchanged velocity of money, nominal spending stays constant. Assuming (1) that x has a unitary price elasticity of demand; and (2) that demand for goods other than x is independent of real purchases of x (thus abstracting from the need for any ‘secondary’ relative-price adjustments), the price of x falls to 50 cents. This implies some (perhaps very slight) decline in the price level. Prices of all other goods remain unchanged, including the prices of the three factors of production whose marginal value productivity is also unchanged. The new equilibrium price structure requires one price adjustment only.
Now suppose, instead, that the price level is kept stable under identical circumstances. To accomplish this, the authorities expand the supply of money to achieve a uniform, though very slight, increase in the prices of 999 goods and of the three factors of production. The sole exception is good x, the price of which must (as in the previous case) still be allowed to fall, only less than in proportion with the improvement in its rate of output. Only in this way can the price index remain stable after allowing needed adjustments in relative prices. 21
21 Some zero-inflationists might protest that their ideal policy would not require any monetary response to a single productivity-based price change, since such a change would typically have only a minuscule effect on the price level (cf. Dowd, 1995, p. 725n). But this stance begs the question: how many prices must be affected by underlying productivity shocks (or, alternatively, how great must be the overall impact of these shocks on a given price index) before price-stabilising policies come into play? Anyway, the argument being made here does not ultimately hinge on the assumption that output in one market only is altered by a change in productivity.
Going the next step, it is easy to generalise our conclusion by noting that it will hold for any possible set of productivity disturbances affecting less than all 1,000 goods. Thus, if the productivity of 999 of the 1,000 industries changes, then a productivity norm requires 999 individual money price adjustments, as opposed to 1,003 for a zero inflation norm.
So does a productivity norm always involve fewer money price adjustments? The answer is no: retaining the same basic assumptions used above, it is possible to construct examples in which the number of money price adjustments required under a price-level stabilisation scheme is less than the number that would be required under the productivity norm. All of them would, however, involve some perfectly uniform percentage increase in productivity of all final-goods industries, such as would leave relative goods prices unchanged, requiring money price changes for factors of production only. Even here zero inflation would ‘win’ only provided that the number of distinct factors of production continued to be less than the number of distinct final goods. 22 In every other case, including ones in which all-around changes in productivity are combined with idiosyncratic changes involving one industry or group of industries, the total number of price changes required under zero inflation will always exceed the number required under a productivity norm, because a productivity norm generally requires fewer changes in nominal factor prices. Elsewhere I used the following example:
‘Suppose that ten goods and three factors of production are initially priced at $8 each. Weighing all goods equally, let the initial price index have a value of 10(8) = 80. Now suppose that output per unit input for one good quadruples, while output per unit of input for the rest doubles. Under the productivity norm, the price of the first good falls to 2; other goods prices fall to 4. [Factor prices don’t change.] Ten money price changes are required in all, and the price index will assume a value of 9(4) + 2 = 38. To achieve zero inflation, the money stock and input prices must increase by the factor 2.105; also, other prices must adjust to satisfy the formula 9(x) + x/2 = 80, which implies x = 8.421. Therefore, the prices of nine goods must be increased from $8 to $8.421, while the price of the tenth good must fall to $4.21. The total number of price changes required under zero inflation thus exceeds the number required under a productivity norm by the number of distinctly-priced factors of production.’ (Selgin, 1995a)
Because productivity, while constantly changing, never seems to advance uniformly in every sector of an economy (Kendrick and Grossman, 1980), it seems reasonable to conclude that, in practice, a productivity norm tends to involve fewer money-price adjustments than zero inflation. The ‘menu’ costs of price adjustment would therefore also be higher under zero inflation, assuming that they are lump-sum costs only. (As the example suggests, it makes no difference after all if the lump sum differs from one price to another.) 23
23 Allowing for variable as well as lump-sum costs of price adjustment could make a difference, since a productivity norm policy tends to involve fewer but larger price adjustments than its zero-inflation counterpart. It is, however, hard to see why costs of price adjustment should vary with the size of the adjustment to be made, especially in the case of output prices (the only ones that are likely to have to adjust substantially under a productivity norm).
Some readers may question the assumption that factor prices need not change under a productivity norm following idiosyncratic (for example, industry-specific) changes in productivity. They should bear in mind, though, that the supply of factors, and of labour especially, to any specific industry is highly elastic – a point recognised by at least one prominent zero-inflationist, the late Arthur Okun (1980, p. 98):
‘Productivity is the key to real wage gains in the economy as a whole, but the differential growth of productivity across industries over time has only a limited effect on the wage structure, for obvious reasons. Workers in industries that, for technological reasons, have low productivity growth … will quit in droves if they keep receiving [lower than average] wage gains. Conversely, firms in industries with rapid productivity growth do not need to pledge or deliver more rapid wage gains than others in order to hold on to their workers. Understandably, the differential growth of productivity across industries mainly changes relative prices over time … rather than significantly altering the pattern of relative wages.’ 24
Okun’s reasoning suggests that a productivity norm may have lower price-adjustment costs than zero inflation even if some of the ‘heroic’ assumptions made above are relaxed, that is, even allowing for the presence of secondary (income- and substitution-effect related) changes in relative output prices.
Suppose, for example, that a productivity shock leaves equilibrium relative wage rates unchanged but has ‘secondary’ relative price effects so widespread as to require a change in the equilibrium relative price of every good. A price-level stability rule will require some adjustment to every money price, including money wage-rates. A productivity norm, in contrast, requires a change in the money price of every good, but (taking Okun’s argument into account) does not require any change in money wage-rates. ‘Menu’ cost considerations therefore seem to offer clear grounds for preferring a productivity norm over zero inflation as a means for keeping the real economy on its ‘natural’ path.
Sellers’ Reluctance to Lower Prices
A decline in the selling price of some product for which demand is unit elastic, reflecting a drop in the product’s real unit cost of production and consequent outward shift in its supply schedule, leaves producers’ revenues and profits unaffected. Such a change need not place producers under any pressure to negotiate new wage-rates and salaries or even to change the size of their workforce. Because the reduction of prices required here is ‘painless’ – a mere result of having more to sell – there is no reason for producers to resist it or to act as if the benefits from not resisting it were mainly ‘public’ ones, external to themselves.
Likewise, for producers to increase prices in the face of shrunken productivity is relatively painless compared to what they must do if the monetary authorities insist on counteracting the rise in prices. Ralph Hawtrey (1930, p. 79) once offered the following illustration, where ‘consumers’ outlay’ is another name for total spending or nominal income:
‘Suppose…that a consumers’ outlay of £100,000,000 has been applied to 100,000,000 units of goods, and that producers who have hitherto received £20,000,000 for 20,000,000 units find their output reduced to 10,000,000 units, but the price of their product doubled. They still receive £20,000,000 and the other producers can continue to receive £80,000,000 for 80,000,000 units. But as £100,000,000 is now spent on 90,000,000 units the price level has risen by one-ninth. In order to counteract that rise, the consumer’s outlay must be reduced from £100,000,000 to £90,000,000. Every group of producers will find the total proceeds of its sales reduced by 10 per cent. Wages, profits and prices will be thrown out of proportion, and every industry will have to face the adverse effects of flagging demand and falling prices. The producers whose prices have been raised by scarcity will be no exception. Their total receipts are reduced in the same proportion, and they must reduce wages like their neighbours.’
Hawtrey also showed that his argument does not depend on the assumption of a unitary elasticity of demand:
‘If the shortage is in a product of which the elasticity is greater than unity, the adverse effect on the producers of that product is greater and on the other producers less. If elasticity is less than unity the adverse effect on the former is less and may be more than counteracted, but what they gain their neighbours lose. Whatever the circumstances, the stabilisation of the commodity price level in face of scarcity 26 will always tend to cause depression.’
The claim that it is relatively easy for producers to adjust prices in response to supply shocks agrees with many theories of output price rigidity. These theories suggest that product prices will be rigid only to the extent that factor prices are rigid, because product prices are often set according to ‘implicit contracts’ promising some fixed percentage mark-up of prices above unit costs (Okun, 1980, p. 170). Although this view accounts for a sluggish adjustment of product prices in response to changes in nominal income, it does not predict any ill-adjustment in situations of changing productivity. In the latter case, unit costs of production are themselves changing, so that adjustments in product prices tend to take place, even as factor prices and the total outlay for factors stay the same, to preserve a constant mark-up. Empirical studies broadly support this conclusion, by revealing that output prices are in fact ‘much more responsive to changes in costs than to shifts in demand’ (ibid., p. 169). It follows, as at least one zero-inflationist (Arthur Okun again) has admitted, that where ‘implicit contracts … are especially important, there may be a case for a horizontal wage trend (and a corresponding negative trend in prices)’ (ibid., p. 280). 27
27 Okun’s reasons for ultimately advocating zero inflation rather than a productivity norm are worth noting, especially in light of his own reliance upon an implicit-contracts model of aggregate unemployment. His reasons are (1) that a shift from zero inflation to deflation ‘would sacrifice some output for a period of time’ and (2) that a ‘modest upward trend in wage rates’ would allow for occasional changes in relative wages without requiring as many cuts in nominal wages as a productivity norm would require. Okun’s stand illustrates the difficulty proponents of zero inflation have in rejecting a productivity norm without implying that some positive inflation rate would be advantageous. Why assume that the transition costs of going from zero inflation to, say, 2 per cent deflation will be any greater than those of going from 12 per cent (the approximate US rate when Okun’s book appeared) to zero? And, if a ‘modest’ upward trend in wages (consistent with zero inflation) requires fewer nominal wage cuts, then a less modest trend, consistent with positive inflation, requires still fewer.
Up to now we have granted zero inflationists’ assumption that random changes in equilibrium money prices are entirely unanticipated by economic agents. This assumption is, however, not really appropriate in the case of downward price adjustments associated with changes in productivity. In truth such adjustments are likely to be perfectly anticipated by price-setting agents in the directly affected markets. The reason is simple: improvements in productivity are often (if not always) consciously aimed at by producers, who seek them precisely because they want to sell more than their rivals by charging less, without sacrificing profits (Haberler, 1931, p. 20). 28 […]
28 Naturally this cannot be said concerning setbacks to productivity, which are generally unexpected.
Monetary Injection Effects
Yet another difference between price adjustments made necessary by unaccommodated changes in productivity and adjustments made necessary by changes in the flow of nominal income (as must occur if the price level is to be kept stable in the face of productivity changes) is that the former come about in a relatively direct manner.
A productivity change implies an immediate shift in output supply schedules and market-clearing prices (with no necessary change in input supply schedules) for those products being produced more or less efficiently than before. In contrast, as we have seen, a less-than-perfectly anticipated change in the money stock, such as would be needed to maintain a stable price level in the face of some unanticipated but persistent change in aggregate productivity, affects most prices only indirectly, through a sequence of shifts in nominal demand schedules beginning with schedules in a few markets only – bond markets, usually – and eventually spreading through the rest. Relative prices, including real interest rates, are thus displaced from their natural or full-information values. It follows that, instead of avoiding monetary ‘injection effects’, a consistent policy of price-level stabilisation is likely to be a source of such effects whenever aggregate productivity changes unpredictably.
Yeager (1996a) disagrees with this view. He argues that, because any increase in productivity will typically be accompanied by an increased demand for real money balances, a monetary expansion aimed at stabilising the price level as productivity advances only serves to accommodate the public’s demand for ‘increased intermediation services’, avoiding a temporary excess demand for money and associated break in the flow of spending. This supposedly helps to avoid loan-market ‘liquidity effects’, keeping real interest rates at their natural levels.
But Yeager overlooks the rapid, if not immediate, tendency of output prices to respond to productivity (that is, unit cost) changes. He overlooks, in other words, how changes in the demand for real money balances based on innovations to aggregate productivity are accommodated by falling prices automatically and well ahead of any possible monetary policy response.
Because nominal prices do not adjust sluggishly to productivity (as opposed to aggregate spending) shocks, no excess demand for money arises. The flows of spending and intermediation continue unimpeded. Attempts by a monetary authority to ‘accommodate’ an increased demand for real balances based on some concurrent change in productivity do not, therefore, actually serve to offset prior shortages of money at all. Instead, such attempts disturb established states of monetary equilibrium by reversing or ‘rolling back’ prior, equilibrating changes in money prices. The process of ‘rolling back’ the price level itself introduces excess liquidity into the economy, pushing real interest rates temporarily below their natural levels.
In contrast, if the monetary authorities prevent the price level from changing along with a change in productivity (for example, by making more units of money available just as expanded outputs reach retailers’ shelves), their actions will add ‘static’ to the price system, by causing a general change in aggregate spending. To be sure, agents will not be ‘surprised’ in this case by any change in the overall level of output prices; but they will be surprised by a general outward shift in both output and input demand schedules. Although the price level does not change, agents may confuse this general, nominal increase in demand with changes in the real demand for particular goods and factors of production.
Formally, the argument here is essentially the same one found in many recent proposals and assessments of nominal income (GNP or GDP) targeting. 30 The argument can be illustrated using the aggregate supply-demand framework shown in Figure 2a. The illustration includes both a long-run (LAS) and a short-run (SAS) aggregate supply schedule, where the former is vertical and the latter allows for the possibility of short-run monetary misperceptions and is therefore upward-sloping. 31 The rectangular-hyperbola, unit-elastic aggregate demand (AD) schedule shows all combinations of the price level (P) and real output (y) consistent with some given level of spending, which is assumed to be controllable by the monetary authorities. Real output starts out at some ‘natural’ level y(n), consistent with the intersection, at point a(n), of the short-run aggregate supply, long-run aggregate supply and aggregate demand schedules.
31 Although zero-inflationists will generally accept the assumption of a vertical long-run supply schedule (and associated vertical Phillips Curve), others reject it. For example, in a recent, influential article George Akerlof, William Dickins, and George Perry (1996) appeal to downward nominal wage rigidities to argue for a curving Phillips Curve. Here, a positive rate of inflation is supposedly needed to achieve maximum employment. The argument, in essence, is that, even assuming a non-negative trend for the average level of money wage-rates (as would exist under a productivity norm), changes in the distribution of the demand for labour across industries would necessitate downward money wage adjustments in adversely affected industries to allow them to maintain their workforce. If money wages are rigid downwards, workers in these industries will become unemployed.
This framework appears to exaggerate the extent to which money wage adjustments are needed to achieve an efficient allocation of labour in response to both temporary and permanent shifts in the distribution of the demand for labour. In the case of merely temporary shifts, employers may continue to employ the same number of workers, at their original wage-rates, knowing (or believing) that better days are ahead, and wanting to preserve good-will. In the case of permanent shifts in demand, lay-offs can perform the same allocative role as money wage-rate cuts – inducing workers to seek employment in industries where demand has risen. In the former case, inflation is not needed to avoid unemployment; in the latter, inflation could at best avoid unemployment only by perpetuating an inefficient allocation of labour.
Figure 2b is the corresponding labour-market diagram, where w is the money wage-rate, and N stands for man-hours of employment. The nominal demand for labour (LD) is assumed (for simplicity’s sake) to reflect the state of aggregate demand, while long- and short-run labour supply schedules (LLS and SLS, respectively) hold up their aggregate supply counterparts. Allowing that productivity is subject to change, the vertical LLS schedule implies that labour supply is inelastic in the long run with respect to changes in real wage-rates. In the short run, however, workers may engage in some ‘intertemporal substitution’ of labour for leisure or vice-versa, for example, by working less today with the intention of working more tomorrow in response to a perceived decline in their real wage-rates that they believe might be temporary. The upward-sloping SLS schedule allows for such an intertemporal substitution effect based on monetary misperceptions: workers perceive changes in their money wage-rates at once, while perceiving changes in the price level only after some delay. Workers therefore temporarily misperceive their real wage-rates.
The framework here, unlike the one implicit in the earlier discussion, does not invoke ‘menu’ costs of price adjustment. In reality, of course, menu costs and monetary misperception effects may simultaneously provide the basis for non-neutral effects of changes in the supply of or demand for money. At the moment, however, I wish to allow for monetary misperception effects only, abstracting from menu costs. The price-level policy best suited for avoiding monetary misperception effects may, after all, differ from the policy best suited for minimising menu costs.
Now consider the effect of a decline in spending, from AD to AD1, due, say, to an unexpected fall in the velocity of money. The natural rate of output has not changed, but with less being spent, the nominal demand for labour declines. Because workers are unaware of an ensuing drop in prices, the economy moves along the short-run aggregate and labour supply schedules to point b, involving a below-natural level of employment and output and lowered wage and price levels. Eventually, the misperception effect wears off – an event signified by a downward shift in the short-run labour and aggregate supply schedules, from SLS to SLS1 and from SAS to SAS1. The new short-run aggregate supply schedule crosses the new aggregate demand schedule at a point, c(n), that is once again consistent with natural levels of output and employment.
The policy implication of the above example ought to be straightforward: assuming they have the power to do so, the monetary authorities should make sure that aggregate demand does not fall, by offsetting any tendency for velocity to shrink with some appropriate increase in the money stock.
Next, consider the effects of a positive productivity shock, starting with the same initial equilibrium as in the previous example. This case is illustrated in Figure 3. Assume that the monetary authority sticks to a productivity norm, and so does nothing (assuming a fixed ‘natural’ rate of factor input) other than maintain a stable level of aggregate spending. In this case, unit production costs fall, meaning that more output is produced by the same quantity of labour and capital. Both the long-run and the short-run aggregate supply schedules shift to the right, from SAS to SAS1 and from LAS to LAS1, and so does the natural rate of output. The resulting ‘natural’ equilibrium, d(n), involves the same lowering of the price level as the previous case, but no change in money wage-rates (since neither the supply schedule nor the demand schedule for labour shifts), hence, no monetary misperception effects. Although workers may still fail to perceive or respond to the general decline in prices, the ‘failure’ turns out to be optimal: the short-run increase in real wage-rates is consistent with long-run equilibrium. Output moves directly to its new natural rate, y(n)1.
What happens in the case just described if the authorities, instead of stabilising spending, attempt to stabilise the price level? Then, rather than let the economy come to rest at its ‘natural’ equilibrium, d(n), the authorities expand the money stock to generate a higher aggregate demand schedule (AD1) that intersects the new long-run supply schedule at a point consistent with the old price level. This expansion of spending raises the demand for labour to LD1, and so causes the economy to ‘ride up’ its new, short-run labour and aggregate supply schedules to equilibrium points (e) involving higher-than-natural levels of employment and output. As in the case of a pure spending shock, things return to normal once the short-run aggregate supply schedule adjusts. Thus, attempts to stabilise the price level in the face of productivity shocks themselves become a source of disequilibrating monetary misperception effects that would be avoided if the price level were simply allowed to adjust along with changing unit production costs.
Figure 4 shows what happens if the monetary authorities take steps to prevent an increase in the price level following a set-back to productivity. The adjustments are opposite to those just described. To combat the tendency of prices to rise, the authorities must reduce the money stock and aggregate demand. As was the case in the first illustration (where demand fell but productivity was unchanged), the decline in spending diverts the economy to a set of equilibrium points (g) involving below-natural levels of employment and output. Indeed, from the point of view of workers, who initially perceive a nominal shift in the demand for labour only without noticing any similar shift in the demand for output, the two situations are identical. Evidently, it is shifts in aggregate demand, and not changes in the price level per se, that sponsor monetary misperceptions and consequent, ‘unnatural’ changes in output and employment.
III. DEBTORS AND CREDITORS
Price Movements and ‘Windfalls’
For separate consideration is the effect of a productivity norm on contracts between debtors and creditors, where debtors have committed themselves to making fixed-money payments in the future, and creditors have agreed to receive these fixed-money payments.32 It is generally assumed that fixed nominal debt contracts are easier to write and execute than other kinds, including contracts in which payments are indexed to some measure of the general price level. Proponents of zero inflation claim the absence of unexpected price-level changes to be a requirement for the successful employment of such fixed-debt contracts, and especially for avoiding ‘windfall’ transfers of wealth from creditors to debtors or vice-versa. So long as the price-level is kept constant, the argument goes, neither debtors nor creditors will (on the whole) have any reason to regret their reliance upon fixed-debt contracts. A constant price level is also supposed to promote long-term investment by eliminating a source of uncertainty that would otherwise discourage such investment (e.g. Hoskins, 1990, p.35).
32 Although the discussion that follows refers explicitly to loan contracts, most of the same considerations apply to other fixed-money obligations, including explicit or implicit fixed-money wage contracts.
The argument, like most arguments for a constant price level, is perfectly valid so long as aggregate productivity is unchanging. But if productivity is subject to random changes, the argument no longer applies. Imagine, for example, that everyone expects both the price level and productivity to remain unchanged. 33 Then, if the price level is kept constant in the face of unexpected improvements in productivity, readily adjusted money incomes, including profits, dividends, and some wage payments, will increase; and recipients of these flexible money payments will benefit from the improvements in real output. Creditors, however, will not be allowed to reap any gains from the same improvements, as debtors’ real interest payments will not increase despite a general improvement in real earnings. Although an unchanged price level does fulfil creditors’ price-level expectations, creditors may still regret having engaged in fixed nominal contracts, rightly sensing that they have missed out on their share of an all-around advance of real earnings, which share they might have been able to insist upon had they (and debtors also) known about the improvement in productivity in advance.
33 The argument that follows still holds if we allow that agents accurately anticipate some changes in productivity, while also anticipating how the monetary authorities will respond to these changes.
Now imagine instead that the price level is allowed to fall in response to improvements in productivity. Creditors will automatically enjoy a share of the improvements, while debtors will have no reason to complain: although the real value of the debtors’ obligations does rise, so does their real income, while the nominal payments burden borne by debtors is unchanged. Debtors can, in other words, afford to pay higher real rates of interest; they might therefore, for all we know, have been quite happy to agree to the same fixed nominal interest rate had both they and creditors been equipped with perfect foresight. 34 Therefore the debtors’ only possible cause for regretting the (unexpected) drop in prices is their missed opportunity to benefit from an alternative (zero inflation) that would in this case have given them an artificial advantage over creditors. The debtors ‘loss’ is, as Haberler (1931, p. 21) put it, only lucrum cessans [i.e., lost gains], not damnum emergens [i.e., loss].
Many years ago Samuel Bailey (1837, pp. 115-18) made much the same point. Suppose, he said, that A lends £100 to B for one year, and that prices in the meantime unexpectedly fall 50 per cent. If the fall in prices is due to a decline in spending, A obtains a real advantage, while B suffers an equivalent loss. But if the fall in prices is due to a general improvement in productivity, A’s gain is not matched by any absolute diminution of B’s wealth, because the enhanced real value of B’s repayment corresponds with the enhanced ease with which B and other members of the community are able to produce a given amount of real wealth. Likewise, if the price level were allowed to rise unexpectedly because of a halving of productivity, ‘both A and B would lose nearly half the efficiency of their incomes’, but ‘this loss would arise from the diminution of productive power, and not from the transfer of any advantage from one to the other’. Bailey concluded from this that productivity-based price level changes offer ‘no pretext for interfering with the literal construction of [a fixedmoney] contract, as a contract for quantity without reference to value’ (ibid., p. 121). By the same token, the price change would not justify an attempt by the monetary authorities to interfere with the course of prices.
Still another way to think of the argument is in terms of optimal indexation. The usual view is that, absent costs of doing so, debtors and creditors would be inclined to index money rates of interest to the rate of inflation or deflation, so that more inflation means higher (nominal) interest rates ex post, and more deflation means lower (nominal) interest rates. But if the growth rate of productivity (hence, real income) is also subject to shocks, debtors and creditors might be just as anxious to index money rates of interest to the rate of productivity growth, so that slower productivity growth leads to lower (real) interest rates ex post and more rapid productivity growth leads to higher (real) interest rates. 35 Under a productivity norm, the price level and productivity move opposite to one another, so that the two forms of indexation would have offsetting effects, making both redundant. Under zero inflation, in contrast, productivity indexing would require an upward adjustment of nominal interest rates proportional to the higher growth rate of real (and, in this case, nominal) income.
If the debtor-creditor advantages of price-level stability are not obvious in situations where productivity is advancing, they are still less obvious in situations where productivity suffers a setback. Francis Edgeworth (1925 , p. 222) once observed that those who plead for stabilising the money value of nominal debts in times of increasing prosperity ‘might be embarrassed if the principle were extended to the case of declining prosperity’. […]
Indeed, if productivity unexpectedly falls – as it may during wartime or when a harvest fails or when a cartel manages to restrict output of some basic raw material – the unfortunate consequences, both ethical and practical, of a price-level stabilisation rule cannot easily be denied, for the rule here requires a contraction of all non-fixed money incomes. Besides leading to a further depression of real activity (if prices and wages are sticky), such a rule might well result in certain debts not being paid at all. Some creditors might, in other words, escape the consequences of fallen productivity, by letting others bear a disproportional burden. Is such an outcome more equitable than one that causes all creditors to suffer some loss? Does it enhance the performance of fixed contracts, or otherwise encourage long-term investment? Surely not.
The Productivity Norm and the Optimum Quantity of Money
Imagine an economy with a capital stock made up entirely of maintenance-free machines, each producing £500 of output annually and initially selling for £10,000 (implying a discount rate of 5 per cent). In equilibrium, an investment in fixed-value bonds earns the same real rate of return as an investment in machines. Suppose that money incomes, the price level, and productivity in this economy are, initially, constant. Bonds then earn both a money and a real rate of return of 5 per cent, while money earns a rate of return of zero.
Next, imagine that, holding the stock of machines constant, regular design changes cause their physical productivity to increase at an annual rate of 4 per cent. 39 Under a productivity norm, the output price level declines at a rate of 4 per cent, and money earns an equivalent real rate of return. Although both the monetary value of output and the rental price of machines remain unchanged, the real return on machines also increases by 4 percentage points. An investment in machines therefore earns a real return of 9·2 per cent. 40 It follows that the money rate of interest on fixed-nominal-value bonds will continue to be 5 per cent, making their real return the same as that of a machine. There is still a 5 percentage-point gap between the real return on bonds and the real return on money.
39 Although the supply of anyone kind of machine is likely to be highly elastic with respect to a change in that machine’s relative productivity, the supply of machines-in-general – that is, the supply of capital – may be quite inelastic with respect to a change in machines’ overall productivity.
40 (1·05)(1·04) =1·092
Now suppose that the authorities decide to stabilise the price level. To do this they must engineer a 4 per cent annual growth rate of money earnings. The prices of factors of production will then increase at the same rate, so that capital continues to earn a real return of 9·2 per cent. The equilibrium money rate of interest on bonds will then rise to 9·2 per cent, making for a 9·2 percentage-point gap between the rate of return on bonds and that on money. Equilibrium money holdings therefore decline, moving the economy further from Friedman’s ideal.
V. THE PRODUCTIVITY NORM IN PRACTICE
Which Productivity Norm?
Until now the implications of ‘a productivity norm’ have been considered without bothering to distinguish between labour productivity and total factor productivity. As noted previously, the distinction is irrelevant in a world where the ratio of capital to labour input is not changing. In the real, industrialised world, however, the capital-labour ratio does change, mainly by growing over time.
Because improvements in labour productivity reflect both improvements in total factor productivity and more capital-intensive production, a labour productivity norm would tend to be more deflationary than a total factor productivity norm: implementing such a norm means setting the growth rate of nominal income equal to the expected growth rate of (quality-adjusted) labour input. 53 As the capital-labour ratio changes, holding the quality and composition of the labour stock constant, money wage-rates remain unchanged, and real wage-rates are kept in line with an improving marginal product of labour entirely by means of falling output prices. The rental price of capital must, in contrast, decline in proportion to the decline in capital’s marginal product as production becomes more capital intensive. To the extent that labour input is less subject to measurement errors than the input of capital services, a labour productivity norm might be put into effect with greater accuracy than its total-factor productivity counterpart. Finally, because it is more deflationary, a labour productivity norm would come closer than would a total factor productivity norm to achieving an ‘optimum’ money stock.
53 See the Appendix (below, pp.72-3) for a formal demonstration of this and other statements made in this section.
A total factor productivity norm involves setting the growth rate of nominal income equal to an average of expected labour and capital input growth rates, where the growth rate of each factor is weighted by its share of producers’ expenses. Such a norm would therefore stabilise, not money wage-rates, but an index of factor prices, so that money wages increase somewhat as production becomes more capital-intensive and decline on those more rare occasions in which production becomes less capital intensive. The rental price of capital goods would, consequently, not have to adjust as much in response to any given change in capital’s marginal product as it would under a labour productivity norm. Moreover, price-level movements would be more closely related to changes in real unit production costs. Finally, although the amount of real capital input is more subject to measurement error than the amount of labour input, popular measurements all suggest a relatively stable growth rate of capital input. This means that a total factor productivity norm will be less subject to input forecast (as opposed to measurement) errors than a labour productivity norm.
So which option is more consistent with overall macroeconomic stability? The answer is far from obvious, and the question warrants further research. For the moment, I am inclined to favour the total factor productivity option on pragmatic grounds: as long as the capital-labour ratio does not change, a total factor productivity norm is equivalent to a labour productivity norm; when the capital-labour ratio does change but total factor productivity does not, a total factor productivity norm is equivalent to zero inflation. A total factor productivity norm therefore represents something of a compromise between a labour productivity norm and a zero inflation norm, making it the less controversial and politically more attractive option, as well as a useful stepping-stone from zero inflation to a ‘labour standard’, should the latter ultimately prove better in theory.
It is also relevant to observe that, regardless of its precise form, a real-world productivity norm is bound to be far from perfect. This has to be so, not only because we often face a choice between keeping wages stable on one hand and keeping prices in line with real unit costs on the other, but also because of the great difficulties involved in measuring and forecasting the growth rates of labour and capital input. To be sure, measurements of real output growth are themselves fraught with problems (Morgenstern, 1963, Chap. 14); while the extreme volatility of productivity itself makes forecasting real output growth far more difficult (measurement errors aside) than forecasting real input growth. 54 The real choice we face is, therefore, not really a choice between a true productivity norm or a truly constant price level, but between some crude approximation of a productivity norm and some equally crude approximation of a constant price level.
54 In the United States between 1948 and 1981, the annualised peak-to-peak growth rate of real output varied from 6·59 per cent to minus 0·0024 per cent – a standard deviation of 2·697 (Bureau of Labor Statistics, 1983). This mainly reflects the underlying volatility of productivity growth. The standard deviation of the peak-to-peak growth rate of labour input during the same period was only 1·03; the standard deviation of the capital input growth rate was still smaller.
A Free-Banking Alternative
Next, consider the implications of a change in the velocity of money, starting from a situation of supply-demand equilibrium in the market for bank reserves. Suppose velocity falls. Then nominal income also falls, giving rise to an excess supply of bank reserves. Given a fixed stock of reserves, individual banks try to lend out their share of the excess which, in hot-potato fashion, merely gets tossed around a bit by the banks without ever actually leaving the banking system. The tossing-around process is, however, one that expands the quantity of bank money until the old level of aggregate spending is restored. The excess supply of reserves thus gets eliminated, since the demand for reserves rises again, in effect ‘cooling’ the hot potato. An increase in velocity has similar consequences, except that the money stock shrinks instead of expanding.