9 Design, Measurement, and Communication: Chile’s Experience with Inflation Targeting
- Carol Carson, Claudia Dziobek, and Charles Enoch
- Published Date:
- September 2002
The Central Bank of Chile was granted independence at the end of 1989. Not unlike other such experiences elsewhere, granting independence to the central bank was a political act but different from the situation in Anglo-Saxon economies. Independence is established by a constitutional act, requiring a high quorum in Parliament for any change in the bank’s status. Moreover, the scope of the Bank’s independence is very wide, encompassing both instrument and target. Indeed, the charter states that the Bank should pursue “price stability, as well as the normal functioning of domestic and external payments” (Chile, 2000). Also, unlike other economies, the central bank is free to intervene not only in money and credit markets, but also in the market for foreign exchange.
Some argue that Chile was an early adopter of inflation targets, because in September 1990 the central bank explicitly announced an objective for year-on-year inflation for 1991 (Mishkin and Schmidt-Hebbel, 2002). This de facto definition of a target, as well as the tough stance on inflation, was prompted from the start by a complex macroeconomic and political situation. A high degree of uncertainty surrounded the economic implications of the political transition, and annual inflation had been accelerating from 1987 onward.
Without entering into a discussion of whether Chile was or wasn’t an early adopter of inflation targeting,1 it is clear that the use of a forecasted target for inflation was deemed relevant to the accountability of monetary policymakers. For almost a decade (until September 1998) in its annual hearing in the Senate, the Central Bank of Chile announced an inflation target for the following year. Disinflation proceeded quite within the range of these targets.
In September 1999, in part due to the aftermath of the Asian crisis, but also because inflation had reached levels close to the average of Chile’s main trading partners, a new framework for monetary policy was announced. Included were the following: (1) the suspension of the exchange rate band and a regime of floating exchange rates unless “exceptional circumstances” were to occur; (2) greater transparency and accountability, through a public calendar for the dates of the monetary policy meetings and the publication, three times a year, of a Monetary Policy Report; and (3) a redefinition of the inflation target, set at 3 percent, but within a range of 2 percent to 4 percent.
This new framework indeed radically changed the way monetary policy was formulated in Chile, and the importance of the change in the exchange rate regime should not be minimized. Several aspects of the new inflation target are immediately apparent. First, the target changed from a point—the bank’s practice over most of the 1990s—to a range. Second, the period over which the target was to be achieved was left unspecified, apart from a transition year, 2000, that had an interim year-on-year target of 3.5 percent. However, the official inflation forecasts were constructed using a 24-month horizon. Third, no particular emphasis was put on the difference between headline and core inflation.
Structure of the Consumer Price Index
Although a wide variety of measures of core inflation exist, in practical terms some of the credibility problems associated with the choice of an index have been avoided by the central bank. In 1998, the National Statistics Office, which is in charge of tracking price statistics, determined a measure of core inflation, consumer price index excluding fuel prices (CPIX), mainly gasoline, heating oil, and kerosene (CPIfuels in the following equation), as well as some seasonal fruits and vegetables (CPIfruits).
Although the CPIX is a standard indicator, and has the advantage that the Central Bank of Chile was not involved in its definition, it is convenient, for the purposes of forecasting, to consider explicitly a number of prices contained in this core index. For one thing, the tariffs on public utilities, CPIutilities, follow known adjustment clauses and have a sizable weight on the consumer price index (CPI). Some of the different variables that enter into the determination of each of these tariffs include wages, the exchange rate, wholesale prices, and taxes. As these are the prototypical cost-push variables that enter into traditional models of price determination, econometric estimation of those models without separately considering these tariffs will undoubtedly bias the estimation of key parameters such as the size of the exchange rate pass-through or the weight of labor unit costs.
Apart from these tariffs, other prices also follow well-known adjustment rules. Rents, mortgages, property taxes, highway tolls, and others (CPIindex) are closely indexed to past inflation. Thus, including them in the core measures would increase the persistence of the inflation process. Other prices are excluded because they relate directly to monetary policy actions (CPIfinancial). Finally, a number of foods are also taken out (CPImeats). A measure of core inflation is left that considers 70 percent of the CPI, called CPIX1, which is used in the econometric estimation of macroeconometric models within the Bank.
The Classical Approach
It is well recognized that three key issues surround the operation of an inflation-targeting regime.2 One set of issues asks whether a central bank should target a point or a range. Many countries that have chosen inflation-targeting regimes have selected a range instead of a point as a target. The economic rationale for this choice is easy to understand: given the shocks that continuously hit the economy, it is easier to maintain inflation within a range than to continuously hit a single point. Moreover, the central bank needs some flexibility in accommodating certain types of shocks that have a one-off effect on price levels, and therefore only a transient impact on inflation. Another set of issues deals with the choice of the particular price index that will play the role of target. Although in some cases, such as the Bank of England’s use of a retail prices index excluding mortgage interest payments, particular prices are excluded because of their mechanical link to monetary policy, the relevant question is how to consider supply shocks and volatile prices, such as foods.
These issues are discussed below, using the Chilean experience as a backdrop. A key consideration is the credibility of monetary policy in terms of its effectiveness and the reputation of the Central Bank of Chile as the guardian of price stability. Therefore, more than reflecting a mechanical choice, the answers to these questions must be consistent with the central bank’s broad strategies for communicating monetary policy.
The discussion briefly presented above is clouded by the fact that policymakers as well as academics usually look at annual inflation as the key gauge of the achievement of the inflation target. It is revealing to think of the implications of this choice, and how the discussion of ranges, horizons, and measures is much starker if one looks at the pros and cons of using annualized monthly or quarterly inflation. Indeed, the common dimensions of this discussion are the persistence of the inflation process, the correct identification of the shocks, and finally the credibility of monetary policy, which is constrained by the lags with which it affects inflation.
The Persistence of Inflation Innovations
The use of monthly inflation can be immediately questioned. First, there are seasonal patterns in inflation in Chile, mostly due to the changing supplies of perishable foods. The importance of these prices should not be overlooked, because they represent a sizable portion of the consumption basket, as argued above. Thus, the movements in prices from one month to the next should not trigger a monetary policy response. An appropriate seasonal filter, such as the ones used for GDP, could easily solve this problem, but it can still be argued that focusing on seasonally adjusted monthly inflation is not informative. Why? Because of the lags with which monetary policy affects inflation, the policymaker must first assess the persistence of inflation surprises. Does unexpectedly high inflation one month signal accelerating inflation over the coming months? Although this is particularly relevant for inflation-targeting regimes, it also appears in price-level targeting. The magnitude of the monetary policy response to a particular shock depends on the size of the shift in the price level it generates.
At one extreme, if the inflation process is a random walk, then monetary policy should react aggressively to bring inflation down to target. At the other extreme, if the price level is a random walk, then the shock is fully transient. Note, however, that annual inflation tends to reflect for a while the increase in the price level. This is quite an obvious point, but it is usually clouded by the use of annual measures of inflation.
Noting that inflationary shocks are persistent seems to be enough to make a case for the use of forecasts and a target range for inflation. Measuring persistence requires some forecasting ability. Thus, if monetary policy affects inflation with a lag, then using a horizon for assessing the stance of monetary policy seems appropriate. Note, though, that for the case of annual inflation, this horizon should be longer than a year because, mechanically, annual inflation will reflect the shock to inflation until the baseline for comparison shifts as time goes on.
On the other hand, given that some shocks are fully transient, maintaining inflation at a fixed point is not the mandated response. This allows us to make the case for using a range, instead of a point.
Persistence and the Sources of Shocks
Although from a statistical viewpoint it is relevant to quantify the persistence of the shocks, from an economic perspective the sources of the shocks are equally or more important in understanding the inflation process. A first approach is to focus on a simple open-economy Phillips curve:
where π represents quarterly inflation, π* is external inflation in U.S. dollars, e is the logarithm of the nominal exchange rate (Chilean pesos per dollar), a hat over the variable represents a time derivative, y −
Here, a positive shock to inflation can be driven by four factors: (1) a shift in inflationary expectations; (2) an increase in imported inflation, through either a nominal depreciation or external dollar inflation; (3) a positive output gap; and (4) own innovations to inflation itself. The persistent effects of these shocks on inflation depend on a host of factors, related not only to the parameters of the model but also to the persistence of the innovations themselves and their interaction. For example, whether the depreciation of the exchange rate is persistent or transient not only directly affects the path of inflation but also shifts expectations. The same can be said of the likely long-lasting swings in aggregate demand.
Therefore, understanding the sources of shocks within the broader context of the macroeconomy, in particular expectations formation, is key for designing an appropriate monetary policy response. The intensive use of forecasts is required, as it allows the central bank to take a stand on whether inflationary surprises are transitory.
A somewhat different approach consists of formulating the price / costs relationship directly:
Here, all variables are in logs or rates of change. Apart from the ones defined above, p is the log of the CPI, w is the log of wages, q is the log of labor productivity, p* is the log of external prices in dollars, u − ū is the gap between unemployment and the natural rate, and Δ is a first difference operator.
In this case, prices are determined as a variable markup over costs (both unit labor costs and imported components). Nominal wages follow full backward-looking indexation, adjusting to unemployment deviations and productivity growth.
The interpretation of an inflation innovation requires more structural analysis. Such an innovation could be due simply to a shock to markups, which will naturally fade over time, or to a broad increase in costs. As in the previous case, the latter situation can be the result of either an increase in the imported component or a persistent deviation of unit labor costs and/or the output gap from the trend.
A simple exercise consists in tracing out over time the path of inflation that results from the same shock in both models. However, the key difference is that the Phillips curve model cannot identify the source of this inflationary shock. To state the case in stark terms, in the price/cost model the structural shock is identified as a markup shock. The parameters of the model were calibrated using evidence from the Chilean economy (see Figure 9.1).3
Figure 9.1.Chile: Quarterly Inflation Following a Shock
Source: Central Bank of Chile.
The different dynamic responses of inflation to a similar shock are readily apparent. The Phillips curve model, because of the large backward-looking coefficient, indicates a long-lasting inflationary effect that dies out only gradually. The price/cost model, on the other hand, shows a markedly different response, given the error correction nature of the model, after expanding markups begin to contract. The deflationary impact is not completely dampened by the indexation clauses on wages, which still increase unit labor cost over time.
Given the uncertain nature of the structure of the inflation process, it is difficult to achieve fixed inflation level forever. Again, this is an argument for using a range instead of a point target. Moreover, the modeling of the pricing process involves very different sources of inflationary pressures. The Phillips curve model or the price/cost model described above are best suited to the understanding of “demand-pull” and “cost-push” factors that determine inflation. But the types of goods and services that follow this market are only a subset of those included in the CPI basket. Hence, for modeling purposes, some type of core inflation measure must be fitted using the traditional macroeconometric approach.
Credibility and the Price Stability Mandate
Typically, the cases for targeting a range for forecasted headline inflation are tempered if a “credibility gap” exists. The more complex and convoluted the design of inflation targeting, the less credible it will be to the general public, and therefore the less effective monetary policy will be—or so the argument goes.4 Indeed, it can be claimed that a lack of response to an inflation shock on the grounds that it is a transient phenomenon, or that it doesn’t change the medium-term outlook, can damage the anti-inflationary reputation of the central bank. Adopting a range instead of a point likewise can be interpreted as a sign of going soft on inflation or, worse, might work as a coordinating device toward the top of the range and not the center. Core measures of inflation can be discredited on the grounds that the goods and services excluded from the target were precisely chosen for the particular time and place, and respond in any case to an arbitrary choice. Simplicity then is better: use a point target for current, headline inflation.
These discussions are probably typical in most central banks that introduce an inflation-targeting framework, and they have been present in Chile as well. To deal with them, the Central Bank of Chile has distinguished between what can be called the “price stability mandate,” directly linked to the Constitutional Act of 1989, and the methodological target, used as the day-to-day benchmark for the evaluation of the stance of monetary policy.
The interpretation of the price stability mandate was announced publicly on September 2, 1999, establishing the range of 2 percent to 4 percent as the target for CPI inflation over the indefinite future. Moreover, it was specifically claimed that this target was not a “thick-point” and that it was symmetric: deviations on the downside were to be considered as damaging to price stability as deviations on the upside. These factors allow the following, more formal, statement of the price stability mandate that the Central Bank of Chile has established:
The unconditional expectation of annual inflation should be 3 percent; that is, agents should consider the midpoint of the target range as an unbiased estimate of inflation in the future.
The unconditional variance of inflation is bounded from above. The announcement of a range of 2 percent to 4 percent means that the central bank will allow some degree of variation. However, this does not imply that 95 percent of the time (or 100 percent!) annual inflation will be within the bounds of the range. The magnitude is left open.
Notwithstanding these details, the key point is that the inflation target is a permanent commitment to price stability: over the long run, the Bank will be held accountable for achieving, on average, annual CPI inflation of 3 percent.
So where does this leave the discussion between headline and core, forecast versus current, and target versus range? These operational aspects belong precisely there: to the operation of monetary policy. Indeed, the price stability mandate is vague enough to need a more specific implementation for the practical formulation of monetary policy.
Rules Versus Preferences
Another way of framing this distinction is by distinguishing between the choice of a monetary policy rule and the definition of the preferences of the central bank. In other words, the appropriate question is: Given that the central bank wants to uphold its price stability mandate, should it react to headline or core inflation, forecast or current inflation? Should it be more aggressive at the edges of the range than at the center?
This reframing of the issue is helpful, in that it turns attention away from what the central bank’s preferences should be toward how it can conduct an efficient monetary policy. Instead, the relevant issue shifts to the proper parametrization of a reaction function. This point is not minor. Indeed, the first question is a normative one, while the second one relates to efficiency considerations. While the first revolves around choosing a particular point on a downward-sloping trade-off between inflation volatility and output volatility, the second one consists of discarding those policy rules that entail an upward-sloping trade-off between inflation volatility and output volatility.
Although the example above refers to the role of output stabilization in monetary policy, it can easily encompass other issues, such as headline versus core, or forecast versus current. García, Herrera, and Valdés (2002) use a small macro model, including an expanded version of the Phillips curve mentioned above, that considers a direct pass-through from headline to core inflation due to widespread cost of living adjustment mechanisms in the Chilean economy. They find that more efficient trade-offs (that is, trade-offs that lead to both lower inflation and output volatility) are achieved by reacting to headline forecasted inflation. These better trade-offs are a result of the high persistence of the inflation process induced by that model. Targeting current core inflation is not efficient, given that it prevents monetary policy reaction to supply shocks that feed the underlying inflation process.
However, this is a model-dependent answer. Different models will give different trade-offs, and different magnitudes in terms of volatility.
Inflation Targeting in Practice: Facing Supply Shocks
The previous sections dealt with some general questions that have arisen in Chile in earlier years. This section will give a more precise description of how the inflation-targeting framework has been implemented and communicated in Chile since 1999. The period from 1999 to the present is quite interesting, because since the inception of this monetary policy framework the Chilean economy has been subject to two large “stress tests.” The sharp oil price increase that began in late 1999 directly affected the domestic prices of fuels in Chile, introducing a large wedge between core measures of inflation and headline inflation. Then, on top of a substantial depreciation of the exchange rate from September 1999 to the end of 2000, the deepening of the Argentine financial crisis and the subsequent contagion through the region induced an additional sharp depreciation. The discussion surrounding these events is structured in the following way: first, an assessment of the magnitude of the shock as seen at the time; second, its impact on inflation forecasts; and third, a description of the monetary policy response and communication strategy.
The Oil Shock, 1999 to 2000
After a prolonged slump, oil prices bottomed in January and February 1999, when Brent crude averaged between US$9 and US$10 per barrel. From then on, prices began a sharp upward drift, reaching between US$26 and US$27 per barrel one year later (a 180 percent increase). Chile imports a high (90 percent) and increasing portion of the petroleum it consumes, and this oil price hike had a sizable effect on a host of domestic prices. It passed through to domestic gasoline and heating oil prices, which as of March 2000 had increased 27 percent over the previous year, having a direct impact on inflation of between 1 and 2 percentage points. This adjustment was partial, though, because subsidies from the Petroleum Stabilization Fund kept the domestic parity well below international prices for a prolonged period, at a cost of several hundred million dollars. Still, this adjustment left a gap of about 20 percent by the end of the year 2000. Moreover, the prices of regulated services as well as public transportation were also affected. All these led to sizable inflationary pressures, introducing a growing wedge between CPI and CPIX inflation that reached more than 3 percentage points by the end of 2000 (Table 9.1).
|Year-on-Year Percentage Change|
|Crude Oil Prices||Utilities|
|Year||Brent (US$ per barrel)||Domestic equivalent1||Fuels||Public transport||Telecommunications||Water||Electricity||CPI-CPIX Inflation gap|
Corresponds to the domestic parity for Gasoil after subsidies from the Petroleum Stabilization Fund.
Corresponds to the domestic parity for Gasoil after subsidies from the Petroleum Stabilization Fund.
The magnitude of these effects was obviously worrisome, given the recent revamping of the inflation-targeting framework. There were two main concerns. First, the widespread indexation clauses in the Chilean economy could induce a large degree of persistence in inflation, even after the oil price shock faded away. A permanent increase in inflation, without a corresponding policy response, could undermine not only the credibility of monetary policy, but also the anti-inflationary reputation of the Central Bank of Chile. This second concern, more related to the public standing of the institution, was viewed as the more critical. Indeed, only a change in private sector inflationary expectations would undermine the nominal anchor of the economy, putting at risk the monetary policy framework at the very moment it was being reformulated.
The policy and communication response to this problem was twofold. The central bank tightened its policy twice, in January and March 2000, in 25-basis-point increments, publicly reacting to inflation risks. It must be noted that it was not only oil prices that played a role here, but also the swift growth of late 1999, which could indicate that a faster-than-expected recovery was under way. However, by May 2000, when the first Monetary Policy Report (MPR) was due to be released, prospects for rapid demand-driven growth had diminished, so only the oil price shock remained.
The strategy for communicating how the Central Bank of Chile viewed inflation risks was mainly reflected in the inflation forecast. Given that the oil price shock was a one-off event in the baseline scenario, the increase in inflation also, mechanically, had a one-off nature. This was clearly reflected in the fan charts in May 2000, which showed that the acceleration in annual inflation would peak between the fourth quarter of 2000 and the first quarter of 2001, but then would tend toward 3 percent at the end of the forecast horizon.
By September 2000, there had been no major development in oil prices. However, gloomy news on the aggregate demand side and the labor market was beginning to indicate a substantial decrease in medium-term inflationary pressures. The central bank reacted to this set of developments by easing monetary policy, indicating in the September MPR that risks were weighted on the downside in the medium term, and that without a proper response inflation would drift to 2 percent over the following two years. Thus, monetary policy was eased even though it was public knowledge that the Bank expected annual inflation to keep accelerating for at least one quarter.
Annual CPI inflation actually peaked at 4.7 percent in January 2001. Then prices fell in February, partly because of seasonal variations but also because of a sharp slowdown in core inflation. Given that annual CPI inflation fell to within the 2 percent to 4 percent range a quarter ahead of forecasts, monetary policy was eased several times over the course of the first half of 2001. However, the Argentine financial problems kept getting worse, and by mid-2001 had reached proportions that led to worries about medium-term inflation prospects.
The Argentine Crisis, 2001
The crisis was of course reflected in the sharp depreciation of the exchange rate. Even though, since flotation, the exchange rate had depreciated by a substantial amount, in the view of the central bank most of that reflected a set of developments in macroeconomic fundamentals—the general lack of capital inflows to emerging economies, the slow growth of domestic demand, and the monetary policy responses in 2000 and early 2001—that had no out-of-the-ordinary impact on inflation forecasts. Overall, in real terms the exchange rate had depreciated by 8 percent from September 1999 to May 2001. Although the Argentine problems were widely known, the baseline scenario for monetary policy was that macroeconomic and structural policies in Argentina would be able to cope with financial turbulence, avoiding a collapse of confidence and widespread contagion.
In June and July uncertainty about the feasibility of this baseline increased sharply, and with it the exchange rate in Chile began to fluctuate closely with the sovereign spread of Argentine bonds.
The pressure on the foreign exchange market induced more than an 8 percent real depreciation between June and July. The board of the Central Bank of Chile deemed this to constitute an “extraordinary circumstance,” thus allowing intervention in the foreign exchange market. A set of guidelines for intervention was announced in mid-August. The bank would use up to a maximum amount of intervention, US$2 billion, for direct sales to the interbank and stock markets. An increasing amount of dollar-linked debt was to be issued, all within the year. The initial impact was substantial, and the exchange rate appreciated about 3 percent in a few days.
The inflation outlook was presented in the September MPR, and the baseline scenario included arguments similar to the ones that were used to justify a transient deviation of forecasted inflation after the oil shock. Basically, as long as medium-term inflationary expectations were aligned with the target, the increase in costs associated with the exchange rate depreciation would induce a one-off reaction of markups, thus avoiding a persistent effect on the path of inflation.
Then came the terrorist attacks of September 11 and the deepening of Argentina’s financial woes. The third quarter also saw an unusual acceleration of core inflation, which was interpreted as the first sign of sizable pass-through from depreciation. However, these phenomena proved to be transient, and core as well as headline inflation has slumped quite dramatically over the course of the fourth quarter of last year and the first quarter of this year. Also, the deflationary impact of global developments as well as the Argentine crises appears now more substantial. Thus, in the recent MPR, the inflation forecast was kept similar to the one presented in September, including some residual pass-through over the course of the first half of this year.
The most recent developments have deepened this shift in perception. Private sector inflationary expectations for the medium term are between 2 percent and 3 percent, and the board eased monetary policy in January and in February. More easing is expected by the markets.
The Chilean experience with inflation targeting over the past few years offers some interesting lessons. First, the discussion of whether to target a range or a point, or use headline versus core, or focus on forecasted versus current inflation, is important, but only within the narrow confines of the operational implementation of monetary policy. Over time, the Central Bank of Chile has made a distinction between those questions and the “deep mandate” to keep price stability. Thus, there are no general recipes; factors such as the speed of the different transmission mechanisms, the amount of credibility achieved by the monetary authorities, and the particular institutional price-setting mechanisms are all considerations that must be taken into account.
Second, the public communication of the difference between these operational aspects of inflation targeting is key. Unlike the previous stage of inflation targeting, in which the annual target coincided with the inflation forecasts, the Central Bank of Chile now publishes medium-term inflation forecasts that in general differ from the 3 percent target, and even, as in 2000, deviate from the floor or ceiling of the range. The ability to publicly commit to price stability, but at the same time recognize deviations of inflation that are believed transient, is a cornerstone of inflation targeting.
Third, the considerations that surround the conduct of monetary policy under inflation targeting stem from the exchange rate arrangement. The role of the exchange rate as the key adjustment mechanism has both advantages and challenges. The advantages lie in an autonomous monetary policy, which is the key tool in conducting a countercyclical macroeconomic policy. The challenges are the need to deal carefully with the risks of pass-through to prices from depreciation and to maintain the credibility that keeps inflation expectations anchored in the target.
A proper understanding by the public of the monetary transmission mechanism appears to be a sensible communication goal. Such an understanding can be achieved through the publication of the macroeconometric models used for forecasting and simulation, as well as through public speeches by the governor and the members of the board. In addition, while the development of new statistics directly aids the estimation and simulation of the macroeconometric models, it can also assist the communication and formulation of monetary policy in an uncertain environment.
The views expressed in this chapter are solely those of the author and should not be attributed to the Central Bank of Chile.
It can be argued that Chile’s policymakers targeted a variety of variables (including the exchange rate) using a number of instruments not completely consistent with full-fledged inflation targeting, such as an exchange rate band, capital controls, and large sterilized interventions of these inflows.
Various quarterly macroeconomic models have been estimated for the Chilean economy. Simple Phillips curves can be found in Medina and Valdés (2002a and 2002b), as well as in García, Herrera, and Valdés (2002). A more complex price/cost block in the line of the one shown in the text is estimated in García and Restrepo (2001).