Keywords

JEL Classifications

‘Inflation is a process of continuously rising prices, or equivalently, of a continuously falling value of money’ (Laidler and Parkin 1975, p. 741). Because there are several ways of measuring prices, there are also several different measures of inflation. The most commonly used measures in the modern world are the percentage rate of change in a country’s Consumer Price Index or in its Gross Domestic Product deflator. Measures of inflation in earlier periods are based on fragmentary samples of prices, such as those of corn and other staple commodities, or of labour.

Inflation has been a feature of human history for as long as money has been used as a means of payment, and as Milton Friedman (1970, p. 24) famously wrote, ‘inflation is always and everywhere a monetary phenomenon, in the sense that it cannot occur without a more rapid increase in the quantity of money than in output’.

Anna J. Schwartz (1973) provides a compact account of the history of inflation from antiquity to modern times. One of the earliest documented inflations in the ancient world occurred following Alexander the Great’s conquest of the Persian Kingdom (330 BC); the Roman Empire experienced rapid inflation under Diocletian at the end of the third century AD. We have no knowledge of inflation for the thousand years that followed the fall of the Roman Empire. But we do have data from the Middle Ages onwards. The inflation episodes during the Middle Ages were modest, and during those years there was a tendency for periods of rising prices to be interspersed by periods of falling prices. This pattern of intermittent inflation and deflation persisted all the way through to the Great Depression of the 1930s. Since the Great Depression, there has been a general tendency for prices to rise every year (with trivial exceptions). In the 1970s and early 1980s, serious inflations – of more than ten per cent a year – gripped most of the industrial world. But this ‘double-digit’ inflation era was short-lived, and by the mid-1980s inflation rates had returned to the more modest levels experienced in the late 1960s. In the early 2000s, there was little sign of high inflation returning in the major economies. Individual inflations of spectacular dimensions occurred in inter-war Europe, during the fall of Nationalist China (1948–9), and in modern times in some Latin American nations, Israel, and Zimbabwe. Some of these were episodes were hyperinflations – inflation rates that exceeded 50 per cent per month.

It is the fact that inflation has been so variable over time and across countries that gives rise to the question: what are the causes and the consequences of inflation? It is the enormously rich variation in inflationary experience that also provides the data which makes progress in answering those questions possible.

The literature on inflation is large, and several comprehensive, if dated, surveys of it are available (see Bronfenbrenner and Holzman 1963; Johnson 1963; Laidler and Parkin 1975). No up-to-date survey of the literature on inflation was available as of 2006.

Attempts to understand inflation have been aided by the insight that anticipated inflation has different effects from unanticipated inflation. It is convenient to use that distinction in organizing this article. But it must be borne in mind that the distinction between anticipated and unanticipated inflation is analytical. It is not a distinction that has an immediate or direct correspondence with actual historical inflations.

Anticipated Inflation

Anticipated inflation is an idealized situation in which prices are rising at a rate at which all economic agents expect them to rise. No one is caught by surprise. What are the effects of a fully anticipated inflation?

There is little disagreement on the answer to this question concerning the effects on nominal variables – on such things as nominal interest rates, wages and foreign-exchange rates. Other things equal, the higher the expected rate of inflation, the higher the level of nominal interest rates, the higher is the rate at which wages rise, and the faster the rate of currency depreciation. Furthermore, these effects are one for one. An x per cent higher anticipated inflation raises nominal interest rates by x per cent, makes wage rates rise x per cent faster, and makes the currency depreciate x per cent faster.

There is less than complete agreement about the effects of anticipated inflation on real economic variables. Abstracting from transitory adjustment paths, all economic theories predict monetary neutrality: a one-shot change in the quantity of money leads to a proportionate change in the levels of all prices (and wages) and has no real effects. But not all economic theories predict monetary superneutrality – that real variables are neutral with respect to changes in the growth rate of the quantity of money.

There are three alternative views in the literature concerning money’s superneutrality. One view is that money is superneutral – a change in the anticipated inflation rate has no effects on output (or economic welfare). A second view is that in increase in the anticipated inflation rate increases output (and economic welfare). Yet a third view is that a higher anticipated inflation rate lowers output (and economic welfare).

The superneutrality result has been most elegantly and clearly stated by Sidrauski (1967). The result also is present in some modern theories of money that pay detailed attention to the physical environment in which monetary exchange arises (see, for example, Townsend 1980). The essential feature of models that generate superneutrality is that the real rate of interest is imposed by the structure of preferences (intertemporally additive with a constant rate of time preference). In equilibrium, the marginal product of capital is equal to this fixed rate of time preference so that, regardless of what happens to money, the capital stock and output rate are unaffected.

The natural rate hypothesis is a variant of the superneutrality proposition. This hypothesis, advanced by Friedman (1968) and Phelps (1968), states that money is superneutral in the particular sense that there is a unique natural unemployment rate that is independent of the anticipated rate of inflation. Any trade-off between inflation and unemployment is temporary and best thought of as a trade-off between unanticipated inflation and unemployment.

The second view that a higher anticipated rate of inflation increases output and improves economic welfare arises in two classes of models. The first is the so-called Mundell–Tobin effect (Mundell 1963, 1965; Tobin 1965). A higher anticipated inflation rate results in an increase in the opportunity cost of holding real money balances. According to the Mundell–Tobin view, this higher opportunity cost of holding money leads to a portfolio reallocation away from money and towards physical capital. The higher holdings of physical capital result in a higher stock of capital and therefore in a higher capital–labour ratio, which in turn leads to a higher level of output. A rise in the anticipated rate of inflation would put the economy on an adjustment path towards the new higher capital stock that would be associated with a transitory rise in the growth rate and a permanent rise in the level of output. A restatement of the Mundell–Tobin position couched in a modern rational expectations terms has been provided by Fischer (1979). The second type of model is one in which an asymmetry in price and wage adjustment – a downward rigidity – creates a long-run trade-off between inflation and the level of economic activity – a downward-sloping long-run Phillips curve.

The third view that a higher anticipated rate of inflation lowers output and economic welfare also arises in two classes of models. First in an overlapping-generations framework (Samuelson 1958; Wallace 1980) a rise in the anticipated rate of inflation leads agents to economize on their holdings of money which, in turn, leads them to save less and transact on a lower scale with the succeeding generation. Second, Clower’s (1967) suggested technological basis for money – the cash-in-advance constraint – generates super-non-neutrality. Using Clower’s assumption, Stockman (1981) shows that, because a higher anticipated inflation rate raises the opportunity cost of holding money, this, in effect, raises the opportunity cost of undertaking all transactions and, therefore, in equilibrium lowers the scale of transactions undertaken. In Stockman’s model, this results in a lower investment rate and lower capital stock. Thus a higher expected inflation rate leads to a lower level of output. A rise in the anticipated inflation rate will place the economy on an adjustment path that would result in a lower transitory growth rate and a lower permanent level of income.

Some of the above results can be thought of in terms of the substitute/complement relation between money and capital. If money and capital are substitutes in portfolios, then the Mundell–Tobin result arises. If money and capital are complements, as they implicitly are in the overlapping generations and cash-in-advance models, then higher anticipated inflation leads to lower output.

There is an abundance of empirical evidence on the alternative hypotheses about the effects of fully anticipated inflation. But the evidence is not entirely unambiguous. Because the very concept of anticipated inflation is analytical and not historical, in examining inflationary experience assumptions must be made concerning the extent to which inflations have been anticipated.

Comprehensive and systematic attempts that have addressed the question in the context of economic growth are those by Kormendi and Meguire (1985), Barro (1997), and Sala-i-Martin et al. (2004).

Using post-war data for 47 countries, Kormendi and Meguire analyse the effects of a change in the anticipated rate of inflation on output growth in a multivariate regression framework. Anticipated inflation was measured as simply the mean growth rate of inflation over the sample period (which went from the late 1940s to 1977). The finding of that study solidly rejects the Tobin–Mundell hypothesis and, in some formulations, fails to reject the opposite view.

Using data for about 100 countries between 1960 and 1990, Robert Barro finds that inflation has a negative effect on growth. The effect is small but significant and implies that maintained for a number of years, in inflation rate that exceeds ten per cent per year has a large cumulative effect on output. Barro is careful in his analysis of the endogeneity of inflation and growth to establish that causation runs from inflation to growth.

Barro’s finding is challenged by Sala-i-Martin, Doppelhofer and Miller. Using data from 1960 to 1996 for 88 countries and 67 variables considered candidates for influencing the rate of economic growth, and using a Bayesian averaging of classical estimates approach, they find that neither average inflation rate nor the square of the inflation rate has a significant effect on the growth rate.

The work just summarized takes a reduced form and linear approach, and these features limit its utility. Future work on the effects of inflation on growth should be directed toward looking at structural accounts of the linkages and seeking highly nonlinear and perhaps nonparametric relationships between these two variables.

Investigations of the neutrality of unemployment (and output) with respect to anticipated inflation has been the subject of innumerable studies, and Laidler and Parkin (1975) review the state of this literature up to the mid-1970s. The conclusions that emerged from this work were mixed, and most of the results generated on data-sets that ended around 1970 showed the existence of a trade-off. But as the data for the 1970s (with its high inflation rate) were added, the picture changed and Laidler and Parkin concluded that it was not possible to reject the view that the unemployment rate is neutral with respect to anticipated inflation.

This conclusion is challenged in three different ways. First, the classic Sargent (1976) shows that reduced-form equations estimated for a given sampling interval over a given sampling period cannot distinguish among alternative theories, even though the theories have radically different policy implications. The implication of this result for Phillips curve trade-offs is that useful inferences can be made but only by estimating reduced forms over different sub-periods or countries across which policy rules differed systematically. As of 1976, Sargent thought that not much of this type of work had been done, so that little was known.

Second, further empirical work seemed to be consistent with the view that a permanent trade-off exists. King and Watson (1994) study the US Phillips correlations and Phillips trade-offs in a bivariate time-series analysis. They use the unit root (I(1)) inflation process to get around the Sargent (1976) problem (see Fisher and Seater 1993; King and Watson 1997, for details), and estimate structural models to interpret the data and compute the long-run trade-offs and sacrifice ratios (cost of lowering inflation) associated with each model. Except for the extreme case of a real business cycle model, they find long-run trade-offs between inflation and unemployment.

The same conclusion is reached by Akerlof et al. (1996), but for a different reason. They report evidence of permanent downward wage stickiness, which implies a long-run trade-off. This evidence comes from four sources: ethnographic surveys, Bureau of Labor Studies data on the distribution of wage changes in manufacturing establishments, union settlements (in both the United States and Canada), and the authors’ own survey of individuals in the Washington DC area. The authors’ were aware that Panel Study of Income Dynamics (PSID) data showed evidence of extensive downward wage flexibility, but argue that individual reporting errors are large, and when corrected for using data from the Current Population Survey, downward rigidity is present. The presence of downward wage rigidity would constitute a serious challenge to the natural rate hypothesis – the neutrality of the unemployment rate with respect to the anticipated inflation rate. And not surprisingly, much work has been done to check the conclusion reached by Akerlof, Dickens and Perry. Parkin (2000) summarizes this work, which concludes that the money wage rate is not downwardly rigid and that the appearance of downward rigidity results from three sources of bias; measurement error, rounding error and long-term contracts. Controlling and correcting for these sources of bias points towards wage flexibility. Clearly more work is needed to settle this issue.

The third challenge to monetary neutrality comes from a series of papers by Barro (1977, 1978) and Mishkin (1982a, b). Decomposing money growth into anticipated and unanticipated components, Barro reports that only unanticipated money growth influences unemployment and real GDP and (as predicted) both anticipated and unanticipated money growth influences the price level. Mishkin shows that Barro’s estimation procedure, while providing consistent parameter estimates, delivers incorrect standard errors. When Mishkin replicates Barro’s exercises with valid tests, he rejects the restrictions implied by neutrality. (He does not reject the restrictions implied by rationality.)

The literature just reviewed deals with the consequences of anticipated inflation and not its causes. Questions concerning causality are more naturally addressed in the context of an investigation of unanticipated inflation.

Unanticipated Inflation

It is not possible to analyse unanticipated inflation in isolation, independently of other aspects of aggregate economic performance. Fluctuations (at the business cycle frequency) in the general level of economic activity and in inflation, though far from perfectly correlated, share some common features. There is, for example, a general positive correlation between inflation and real income (or equivalently, a negative correlation between inflation and unemployment). There is also a positive correlation between money and income as well as between the velocity of circulation of money and income.

The ‘stylized facts’ about the business cycle (shared by all economies) raise difficult questions about cause and effect. Of the four variables – the price level, real output, the money supply and the velocity of circulation – which, if any, is the prime mover? Do fluctuations in the growth rate of the money supply cause fluctuations in the other variables? Do autonomous movements in the price level, perhaps stemming from wage-push pressure, initiate the fluctuations in money, velocity and output? Does the business cycle have its origin in real factors that initiate fluctuations in output, which in turn lead to induced fluctuations in money supply growth, inflation and velocity?

At one level questions such as these are statistical and are capable of being investigated using econometric methods that detect causality, such as those proposed by Granger (1969). Studies based on such methods have not, however, delivered decisive results.

Most investigations of the possible causes of inflation have sought to understand the phenomenon by identifying the sources of inflation and studying the transmission mechanism whereby those sources are translated into variations in the rate of inflation and in other economic aggregates. This approach is one which seeks to understand both inflation and the business cycle as an integrated phenomenon.

There are three broad classes of theories that have been proposed for understanding the unanticipated and cyclical aspects of inflation. The first of these stems from the work of Keynes (1936) and emphasizes both price stickiness and the potential for autonomous movements in prices. On this view, the normal state of affairs would be one in which wages and prices are relatively sticky, responding only gradually to aggregate demand shocks. Shocks to aggregate demand arise from a variety of sources. One possibility is that autonomous fluctuations in investment produce fluctuations in aggregate demand. Other possible sources of aggregate demand fluctuations are fluctuations in wealth and interest rates which in turn are induced by fluctuations in the growth rate of the money supply. Fluctuations in wealth and interest rates can induce fluctuations in investment and consumption. All of these potential sources of variation in aggregate demand lead to cycles in both output and the price level. Initially, a change in demand will have bigger output effects than price-level effects, but eventually prices and wages will adjust to reflect fully the change in aggregate demand. The resulting co-movements in output and prices will be positively, though not strongly, correlated.

From time to time this normal state of affairs is disturbed by autonomous price shocks. The most commonly hypothesized source of price shocks is wage-push. It is suggested that, at times of substantial industrial or social unrest, movements in the level of money wages will act as a type of social safety mechanism. The idea that wage-push results from sociological phenomena was particularly popular amongst economists in the UK in the early 1970s (see, in particular, Balogh 1970; Jones 1972; Wiles 1973; Hicks 1974. By the time the first oil shock occurred (late 1973), ‘wage-push’ gave way to ‘oil-push’ as the most commonly identified source of autonomous movement in inflation.

When autonomous movements in the price level occur, the phenomenon that came to be known as ‘stagflation’ quickly follows. The autonomous price rise raises the inflation rate and lowers output (raising unemployment). If the higher unemployment and lower output induces an increase in the growth rate of the money supply, then even further price-level rises occur.

This traditional version of the Keynesian theory of inflation and the business cycle, together with some of the sociological embellishments that have been briefly reviewed above, is very thoroughly explained and elaborated in Laidler and Parkin (1975).

More recent and sophisticated versions of the Keynesian theory of cycles and inflation may be found in papers by Fischer (1977), Phelps and Taylor (1977), and Taylor (1979, 1980). The essence of these ‘New Keynesian’ theories is the existence of long-term contractual arrangements in labour markets. Such arrangements result in wages, the major element of costs, being predetermined. This stickiness of wages and costs results in a stickiness of prices, even if the expectations of prices that form the basis for the long-term labour market contracts are formed rationally.

A second approach to understanding cyclical fluctuations is one based on incomplete contemporaneous information about aggregate demand. This approach, sometimes called the ‘New Classical Theory’, was first suggested in the early 1970s by Lucas (1972, 1973). The approach is broadly consistent with the Keynesian mechanism of aggregate demand determination but proposes an alternative theory of aggregate supply. Individual economic agents are assumed to operate in informationally isolated ‘islands’ and to be incapable of distinguishing relative from absolute price level changes. The resulting confusion causes them to respond to absolute price changes as if they were relative price changes. This response results in positive co-movements in output and the price level.

In both the Keynesian and New Classical approaches, the key driving variable generating the cycle – fluctuations in both real output and the inflation rate – is a fluctuating growth rate in the money supply. This is not to deny that other things might, from time to time, shock the economy. Rather, it is a proposition about the major ongoing source of cyclical variation. Within both of the theories, positive co-movements of velocity are explained by appealing to the idea that to some degree the cycle itself is forecastable. To the extent that it is, higher rates of inflation at the cyclical peak will in part be anticipated and, therefore, reacted to. It is always efficient to reduce money holdings when the opportunity cost of holding money increases. Higher expected inflation rates, leading to higher nominal interest rates, induce such economizing and are, therefore, the major source of procyclical fluctuations in velocity.

A third approach to understanding aggregate fluctuations denies the primacy of variations in the money supply growth rate, or in any other sources of aggregate demand fluctuation in generating the cycle. This approach, known as ‘real business cycle theory’, has yet to gain a major following but has, in recent years, begun to spawn a growing and important literature (see, in particular, King and Plosser 1984; Kydland and Prescott 1982; Long and Plosser 1983; Nelson and Plosser 1982). Though differing in details, the essential proposition of the new real business cycle theories is that aggregate fluctuations emanate from technological shocks to the aggregate production function or, in some versions, from sector-specific shocks and from the interactions between sectors of the economy – although a large literature has now incorporated Calvo (1983) price stickiness or monopolistic competition.

Technological shocks that generate fluctuations in full-employment output would, other things equal, generate negative co-movements in prices, and, presumably, to the extent that such movements were forecastable, countercyclical movements in velocity. Since such co-movements do not occur, it seems as if the real cycle theories are in substantial trouble. King and Plosser (1984) address this problem directly by proposing that technological shocks which affect real output induce responses in money and credit that accommodate – indeed over-accommodate – the real fluctuations. Thus, when there is a positive shock to aggregate supply, this induces an even bigger rise in the total volume of money and credit and, therefore, induces procyclical co-movements in money, prices and output. To the extent that these are forecastable, economizing on real balances generates procyclical velocity.

There is not, at the present time, any definitive and systematic evidence capable of disposing convincingly of any of these three alternative approaches; nor is there any overwhelming evidence suggesting that any of them is clearly in the lead.

Inflation in Open Economies

The alternative approaches to understanding inflation that have been reviewed so far have (implicitly) examined inflation in a closed economy. Most practical concerns about inflation arise in individual countries which are open economies. The international trade and international capital market transactions undertaken by such countries have an important bearing on their inflation performance. Also, the foreign-exchange rate regime – fixed or flexible – has an important influence upon a country’s inflation performance. It was during the period of rapidly accelerating inflation in the 1970s that open economy theories and the international transmission mechanism gained in prominence (see Parkin and Zis 1976a, b).

The main feature of the analysis of inflation in an open economy is the emphasis on the limited potency of domestic monetary policy under fixed exchange rates. In a country, or more interestingly in a world, operating on fixed exchange rates, individual countries’ monetary policies have no effect on the country’s rate of inflation. Instead, monetary policy influences the country’s balance of payments. In such a world, inflation is a world phenomenon, not a national phenomenon. It is the growth rate of the world money supply that determines the world average rate of inflation. Theorizing along this line had, in fact, made good progress even as early as the middle of the 18th century at the hands of David Hume (1752). It was rediscovered and popularized in the 1960s and early 1970s by Mundell (1971) and Johnson (1973).

The rediscovery of David Hume’s analysis provided interesting insights into the resurgence of world inflation at the end of the 1960s. An attempt on the part of the United States to finance its Great Society programme and the Vietnam War with limited tax increases and with an increase in the growth rate of the money supply – with an increase in the inflation tax – became the engine of an inflation that engulfed the entire fixed exchange-rate world.

Understanding the international generation and transmission of inflation in a flexible exchange rate world, such as that which had emerged by the mid-1970s, is still far from settled. At the centre of the problem of understanding inflation is the problem of understanding the determination of foreign exchange rates. Large and rapid movements in foreign-exchange rates are seen as having a potentially powerful and rapid effect on domestic price levels. The forces that determine exchange rates are still, however, far from well understood. Viewing the foreign exchange rate as following a random walk is as precise as any structural theories of the exchange rate that have so far been proposed and tested.

Despite the absence of a convincing theory of inflation in an open economy, the effects of policy coordination (or its absence) have been studied. A central question addressed by Oudiz and Sachs (1984) and Obstfeld and Rogoff (2002) is whether unilateral national monetary policy rules are inferior to international monetary coordination. The answer is that they are not.

Positive Theories of Central Bank Behaviour

Recent developments in understanding inflation have been dominated by the rational expectations revolution and the related and more far-reaching revolution that has uses rigorous dynamic general equilibrium analysis. Some of the implications of that revolution have been discussed above and have been to strengthen and refine the theories of inflation that emphasize fluctuations in the growth rate of the money supply as the principal source of fluctuations in inflation and other economic aggregates.

The rational expectations hypothesis holds that expectations are formed by making predictions of future inflation on the basis of the mechanisms that generate actual inflation. If inflation is indeed caused by rapid monetary expansion, then forecasting future inflation is the same thing as forecasting future monetary policy. But monetary policy itself emerges from an ill-understood political process. In most countries the task of formulating monetary policy has been delegated to a central bank. Yet, in determining monetary policy, central banks are often influenced by the economic and political environment in which they operate and must also take account of the consequences of their actions for the behaviour of the economy as a whole.

In order to understand the inflationary process, with people forming expectations rationally, it becomes necessary to understand the policymaking mechanisms and the forces that generate varying monetary growth rates. The first serious analysis of this problem was that by Kydland and Prescott (1977) and the problem has been investigated more recently by Barro and Gordon (1983a, b) and Cukierman (1992). In the models proposed by these writers, a central bank’s goal is to achieve an optimal combination of inflation and unemployment. Lower inflation and lower unemployment are seen by the central bank as desirable objectives. The bank is constrained, however, by a short-run trade-off between inflation and unemployment – a trade-off arising from the considerations described above. A surprise rise in inflation would produce a cut in unemployment while a surprise drop in inflation would produce a rise in unemployment. The precise way in which the short-run trade-off between inflation and unemployment constrains the central bank depends on the expectations of private agents concerning the bank’s behaviour. A central bank that can credibly precommit to a particular rule about inflation – perhaps a zero-inflation rule – would be a bank that could engender rational expectations of zero inflation. It would be optimal for such a bank to in fact precommit to a zero rate of inflation and then deliver that rate.

The ability to precommit and with credibility seems to require some mechanism for binding the central bank that does not have a readily identifiable counterpart in the real world. Central banks are, in fact, free to pursue whatever policies they wish at their discretion. Since this fact is known to all private economic agents, it will be rational for them to take it into account when forming expectations about central bank behaviour. The equilibrium that results in this case will be such as to ensure that the actual inflation rate chosen by the bank is one that removes any temptation for the bank to depart from that rate and further exploit the short-run trade-off. Put differently, the inflation rate chosen will be the best available at the natural rate of unemployment. Only in such a situation would the central bank have no further temptation to attempt to exploit the short-run trade-off. Thus without the ability to precommit to a fixed (and presumably zero) rate of inflation, a central bank will end up delivering a higher rate of inflation than that which is socially desirable.

One feature of the positive theories of inflation developed by Kydland–Prescott and Barro–Gordon that some people find disquieting is the time inconsistency. (In game theory language, the equilibrium concept is Nash rather than sub-game perfection.) Attempts to develop positive analyses that do not have this feature have been based on reputation. One such approach, in Barro and Gordon (1983a), uses the so-called ‘trigger strategy’ model of reputation suggested by James Friedman (1971). A model is proposed in which the central bank would be punished if it delivered too high a rate of inflation and in which it takes time to restore the bank’s reputation. In equilibrium, the bank never does inflate at a rate that requires the punishment to be inflicted.

An alternative approach by Barro (1986) uses the reputation analysis developed by Kreps and Wilson (1982). In this model there are two potential ‘types’ of central banker, one that likes inflation and one that dislikes it. The inflationary central banker has an incentive to masquerade as a non-inflationary type in order to induce low inflation expectations. By inducing low inflation expectations, the inflationary central bank will, at some point, be able to exploit those low expectations and produce a surprise inflation; it will do this by following initially a strategy of inflating at exactly the same rate as would be chosen by a non-inflationary central bank. At some later point it will pursue a mixed strategy – a strategy analogous to choosing an inflation rate by drawing numbers from an urn. Once this mixed strategy has resulted in a high rate of inflation, the inflationary central banker is revealed, and expectations about inflation as well as actual inflation will rise.

Another feature of the Kydland–Prescott and Barro–Gordon models that is objectionable is that the central bank targets an unemployment rate below the natural rate. If it were to target the natural rate, there is no tension between its inflation and real goals. Cukierman overcomes this objection by replacing the symmetric loss function of the standard model with an asymmetric loss function: the central bank weighs positive deviations from the natural unemployment rate more heavily than deviations below the natural rate.

Backus and Driffill (1985) and Cukierman (see Cukierman 1992, ch. 3), have suggested another modification to the standard model: the possible interactions between labour unions (working as a unified wage-setting institution) and the central banks. In this case, inflation (and money supply growth) is determined as the outcome of a game between the central bank and the economy-wide labour union.

Empirical tests of the alternative positive theories of central bank behaviour have been conducted by Ruge-Murcia (2003) and by Cukierman and Gerlach (2003). Ruge-Murcia uses US time-series data and rejects the Barro–Gordon formulation but does not reject the Cukierman asymmetric loss function formulation. Cukierman and Gerlach use data for 22 OECD countries and reach a similar conclusion.

Other recent developments in understanding central bank behaviour arise from the normative analysis of monetary policy to achieve an inflation target, and it is convenient to discuss this topic in the context of inflation policy below. But, before that, it is convenient to consider the links between monetary policy and fiscal policy.

Inflation and Fiscal Policy

A further consequence of the rational expectations and dynamic general equilibrium revolutions has been to force attention back to the connection between fiscal and monetary policy. The simple accounting fact that government expenditure must be financed, either by taxation, by borrowing or by money creation, implies that any analysis of the determination of money growth must at the same time make consistent propositions about fiscal policy and deficit financing. Of course, variations in the growth rate of interest-bearing debt can provide a good deal of insulation of money growth from the deficit. Nevertheless, large and persistent deficits may give rise to rational expectations of future money growth, even in the face of currently firm monetary policies. Sargent and Wallace (1981) have shown that, if the fiscal authority is the prime mover and follows taxation and spending policies that are independent of monetary policy, then, essentially, inflation and, ultimately, money growth are fiscal phenomena. Whether these findings are of practical importance is a matter of some controversy. Sargent (1982), studying the ends of four big inflations, has argued that adjustments in fiscal policy have been crucial to ending inflation. By implication, the emergence of a large and apparently uncontrolled deficit would be seen as the origin of serious inflation. Work by Dornbusch and Fisher (1986) offers a different interpretation, however, placing major importance on the behaviour of the foreign exchange rate.

The link between fiscal policy and inflation is most complete in Woodford’s (1995) fiscal theory of the price level. Because the quantity of money demanded depends on the opportunity cost of holding money, which in turn depends on the rational expectation of the inflation rate, there is a large number (infinite) of equilibrium price level paths. The standard (mostly unstated) approach rules out all the purely speculative equilibria and selects the unique equilibrium based on the monetary fundamentals. In which the government’s choice of how to finance its debt determines the inflation rate. The fiscal theory of the price level rejects this approach and rules out equilibria by the government’s selection of its debt financing regime. As an example, Kocherlakota and Phelan (1999) show that, with a policy of constant taxes and constant money, the fiscal theory predicts that a one-time cut in the quantity of money generates a speculative hyperinflation (in contrast to the standard model prediction of a one-time fall in the price level).

Policy Towards Inflation

Analyses of policies towards inflation have changed over the years. Advocacy of gradually slowing down the growth rate of the money supply and advocacy of controls on wages and prices were the most commonly heard policy suggestions for controlling inflation in the 1960s and early 1970s. Those who saw autonomous wage and price movements as the principal source of inflation saw prices and incomes policies as the major weapon to control it. Those who saw money growth as the source of inflation embraced monetary gradualism as the most obvious cure. A prodigious amount of work attempting to evaluate alternative policies was undertaken, much of which is surveyed by Laidler and Parkin (1975).

As a consequence of the rational expectations and dynamic general equilibrium revolutions, the focus of the policy debate has shifted markedly from that of seeking to manipulate variables such as key wage settlements (the prices-and-incomes policy solution) or the growth rate of the quantity of money (the monetarist solution). Instead, attention has turned to thinking about the way in which different institutional arrangements interact to produce different inflation rates. And the emphasis has shifted from policy as an action to policy as a process or set of rules.

One line of research has examined the consequences of alternative monetary systems, including the adoption of alternative forms of commodity money (see, in particular, ‘Conference on Alternative Monetary Standards’ 1983). Another research direction has been the investigation of targeting nominal income growth as a means of conquering and avoiding inflation (Tobin 1983; Taylor 1985).

But the idea that has attracted most attention both in the research community and among central banks is the use of a monetary policy rule that seeks to achieve either an inflation rate target or a price level target. The study of inflation or price level targeting has both a positive and a normative dimension and sometimes the two are not explicitly distinguished.

Svensson (1999) has provided a nice distinction between what he calls ‘instrument rules’ and ‘targeting rules’ for monetary policy. In the context of inflation targeting (and that is the context of most of the recent literature on monetary policy) an instrument rule specifies how the policy instrument responds to the current state of the economy. The current state can include current forecasts of future variables. A targeting rule, in contrast, states that the policy instrument shall be set at the level that makes for forecast inflation rate equal the inflation target.

The policy instrument that features in instrument rules is either the overnight interest rate on inter-bank loans or the monetary base. Woodford (2003) provides the authoritative account and discussion of the interest rate instrument rule and shows that such a rule can, in principle, deliver low and stable inflation provided that it incorporates the ‘Taylor principle’, which states that the interest rate must change in the same direction as a change in the inflation rate but by more than the change in the inflation rate (Taylor 1993, 1999).

McCallum (1988) has explored the use of a monetary base rule and compared the robustness of interest rate and monetary base rules.

It is a curious fact about the models that explore the use of an interest rate rule that money plays either no role or no essential role in the inflation process. The models in which money plays no role are typically specified as reduced forms in which inflation is generated by expected inflation and the output gap; the output gap responds to the real interest rate, which equals the nominal interest rate set by the central bank minus the inflation rate; and expectations are rational. Other models are specified at a deeper structural level with consumers maximizing intertemporal utility of consumption and leisure and monopolistically competitive firms setting prices according to a Calvo (1983) formula.

In some models, money enters through a ‘shopping time’ function (King and Wolman 1996). But whether present in the model or not, money plays no essential role in the inflation process. This fact is emphasized in Woodford (2003) by his exploration of the cashless economy and is seen as a virtue because it might provide insights on inflation in a future economy when technological change has driven money, as we know it, out of existence.

It is also a curious fact that inflation targeting amounts to targeting a variable whose value cannot be influenced by a central bank’s current actions until well beyond the bank’s forecast horizon. It is the long and variable lags in the response of inflation (and output) to monetary policy that led Friedman to his original advocacy of a money stock growth rate target.

The evolution of inflation over the coming years will provide valuable evidence on both the inflation process, the currently out-of-favour monetarist ideas, and the wisdom of the current policy regimes.

Conclusion

Macroeconomics in general, and the theory of inflation in particular, is in a fluid state. The foregoing has attempted to review that state and provide a picture of the path that we have taken in getting to it. We have broad agreement on the facts to be explained and broad agreement on the behaviour of nominal variables (for given real variables) in an inflationary economy in which the path of inflation is anticipated. We also have broad agreement that fully anticipated inflations, though in many theoretical models capable of generating non-neutralities, are nevertheless to a good approximation neutral. Beyond that there is little in the way of firm knowledge. We have a variety of models of macroeconomics and inflation, and many clear theoretical results. We do not have much, however, in the way of solidly based rejections of any of the available models. Uncertainty surrounds both the issue of the impulse (or impulses) that generate inflation and other fluctuations and the issue of the propagation mechanisms that translate those impulses into movements in output and the price level.

See Also