Are inflation expectations differently formed when countries are part of a Monetary Union?

Transcription

1 Are inflation expectations differently formed when countries are part of a Monetary Union? Amina Kaplan Master Thesis, Department of Economics, Uppsala University January 15, 13 Supervisor: Nils Gottfries

2 Abstract I study to what extent consumers expectations of inflation are formed differently in the short run for countries that belong to a monetary union, which implies a common inflation-targeting central bank, in contrast to countries with national inflationtargeting central banks. I measure if there are differences in consumers reactions to inflation deviation from target, persistency of expected inflation and the ability to predict accurate inflation in the respective considered countries. I also measure average deviation and average absolute deviation of actual and expected inflation from the inflation target. The results suggest that the respective country s average reaction to inflation deviation from target, degree of persistent expectation and the ability to predict accurate inflation rates are in the same range as well as the results for the average and average absolute deviations. Therefore, I conclude that there are no substantial differences in the formation of consumers expectations in countries belonging to a monetary union and countries with national inflation-targeting central banks, in the short run. Keywords: Actual Inflation, Inflation Expectations, Inflation Target, Monetary Union, Persistency, Prediction, Rational Expectations.

3 Contents 1. Introduction The Hypothesis Approach Taken In This Study Review of Previous Research Approach in Previous Research Results from Previous Research Theory and Method Theory Inflation-Targeting The Central Bank s Model of the Economy Rational Expectations Method Actual and Expected Inflation Deviation from Target Expectation Formation Persistency of Expected Inflation Predicting Inflation Data Overview of Data The European Consumer Survey The Probabilistic Approach Quantifying the Qualitative European Consumer Survey Modifications and Drawbacks of the Probabilistic Approach Results and Implications Results Statistics for Actual Inflation Statistics for Expected Inflation How Does Expected Inflation React to Actual Inflation? Persistency of Inflation Expectations Prediction of Actual Inflation Implications Conclusion References Appendix A: Database Appendix B: Summary Statistics... 37

4 Appendix C: Figures of Actual, Expected and target Inflation... 4 Appendix D: Figures of Expectation Uncertainty and Time-Varying Response Thresholds Appendix E: The Augmented Dickey-Fuller Test... 5

5 1. Introduction For a long time central banks have had an interest in monitoring the behavior of inflation expectations in the economy and in understanding the nature of the process by which expectations are formed. It is very important to have reliable measures of expected inflation for different reasons. For example, if they provide a useful or relatively accurate predictor of future inflation, measures of expected inflation can reveal important information in forward-looking analyses of price developments. Expectations of higher inflation rates can give rise to cost-push effects where employees demand higher wage settlements. This means that in an environment where overall inflation is expected to be high, employers may be willing to pay higher wages since they assume that changes in costs can be more easily passed on to the customers in the form of higher prices (Christiano and Gust, ). From the perspective of a central bank expectations of inflation are also relevant for other reasons. Inflation expectations can be viewed as a determinant of money demand and play an important role in affecting real interest rates (Forsells and Kenny, ). According to Taylor (1999), inflation expectations are among the key arguments in a central bank s reaction function, i.e. the relationship relating interest rates to inflation and the output gap. Over the medium-term horizon (i.e. the central bank target horizon), inflation expectations are seen as a measure of how credible a central bank's commitment to achieve price stability is, as discussed by Fromlet (13). Macroeconomic theory shows the importance of inflation expectations when trying to understand how the economy is functioning. From the perspective of aggregate price formation, the Phelps-Friedman Expectations-Augmented-Phillips-Curve relies on adaptive expectations when trying to explain why firms and workers are slow to adjust their expectations to business cycle conditions in the short run. Since expectations play an important role in the model, suggestions of consumers having more rational expectations emerged. This resulted in the ideas behind the rational expectations models that say: given that all available information is utilized, 1

6 expectations should not lead to systematic mistakes. Individuals should be able to learn from past experiences and only unexpected changes in monetary policy should affect real variables. 1.1 The Hypothesis In order to stabilize inflation expectations and achieve low and stable inflation in the long run, recent research stresses the importance of credible and transparent information about future monetary policy (Friedman, ). The focus is on the implications of having an inflation-targeting central bank or not, to what extent volatility in actual and expected inflation and the formation of expectations differs. There is however no previous research of how expectations of inflation form in countries within a monetary union that has a common inflation-targeting central bank, and the extent of significant differences in volatility of actual and expected inflation, in contrast to countries with own national inflation-targeting central banks. Therefore, in this study I test the hypothesis that: assuming a central bank of a monetary union or of a single country could achieve low and stable inflation by having an explicit inflation target and be credible, inflation rates for countries within a monetary union may still deviate from the target. This is because a central bank in a monetary union, such as the Economic and Monetary Union (EMU) which I study, considers the average inflation rate in the whole EMU and not the EMU countries specific inflation rates. This is in contrast to countries with national inflation-targeting central banks (non-emu) with the objective to keep the national inflation rate in line with the target. This means that counterbalancing national inflation rates in the EMU that on average are below but close to %, which is the target rate considered by the European Central Bank (ECB), would be considered in line with the target. 1 The inflation history of the EMU confirms this, as shown in figure 1 and in the charts in Appendix C. The results suggests that the ECB has managed to keep the average inflation rate of the EMU in line with the inflation target while inflation rates for the EMU individual countries has diverged from the target to different extents. 1 The ECB is the central bank of the EMU.

7 Figure 1: Euro Area Inflation: Jan 1997 Nov jan-97 mar-98 maj-99 Actual Inflation Target Inflation Table 1: Inflation Jan 1997 Nov 13 Average Inflation 1,94 jul- sep-1 nov- jan-4 mar-5 maj-6 jul-7 sep-8 nov-9 jan-11 mar-1 maj-13 Inflation Target -,56 Standard Deviation,764 The results does not reveal to what extent volatility in inflation rates in the EMU countries are different from non-emu countries inflation rates, therefore the results need to be put into contrast. 1. Approach Taken In This Study In order to measure to what extent there are differences in average deviation of actual and expected inflation, volatility and how consumers form expectations of inflation in the EMU countries, in contrast to non-emu countries in the short run, I try to answer three questions: 1. Does the divergence in EMU national inflation rates indicate that consumers in EMU countries react differently to inflation than consumers in non-emu countries? To answer this question I first measure the deviation of actual and expected inflation from the inflation target in the respective EMU and non-emu countries by estimating average deviation and average absolute average deviation. As a second step I estimate to what extent inflation deviation from the target in one period affect expectations of inflation in the next period. These measures are important since they show if a central bank s inability to, in the short run, keep inflation in line with the target is leading consumers to expect different inflation than the inflation target.. Will inflation expectations be more persistent in EMU countries? 3

8 To answer this question, I measure how persistent expectations of inflation are in EMU countries and non-emu countries. As discussed above, inflation rates in EMU countries may be more long-lasting since the ECB is unable to adjust inflations on country levels. 3. Are consumers in the EMU and non-emu rational, i.e. do consumers inflation expectations accurately predict inflation despite how expectations are formed? To answer this question, I measure whether changes in expectations of inflation on average predicts changes in the actual inflation rate. Outline The rest of the study is structured as follows. Chapter offers a brief and selective summary of previous research along with similarities, differences and contribution from this study. Chapter 3 considers what the theory of rational expectations, inflation targeting and the central bank reaction function might say about the extent of differences in actual and expected inflation. The differences are estimated by using measures presented in the Method section 3.. Chapter 4 discusses the data of the actual and expected inflation series and the process to obtain a quantitative estimate of expected inflation from a qualitative survey. Chapter 5 presents the estimated results and the implications that can be drawn. Chapter 6 concludes. 4

9 . Review of Previous Research In this chapter I present the approach and results from previous research on inflation-targeting, actual and expected inflation volatility and the formation of inflation expectations. The purpose is to relate and present differences in the approach taken in this study and in previous research and to present the results that could be expected from this study based on previous research..1 Approach in Previous Research The approach taken in previous research compares differences between inflationtargeting (IT) countries and non-inflation-targeting (non-it) countries. This differs to some extent from my study where I compare countries in a monetary union that have a common inflation targeting central bank with countries that have national inflationtargeting central banks. It is however possible to compare my approach with previous approach because I consider the ECB as an IT central bank in a monetary union and countries within the EMU as non-it. This is because the EMU countries respective inflation levels are not being considered separately by the ECB but weighted into an average value for the whole of EMU. In the following discussion and in the rest of the study I will therefore do the same partition in IT and non-it countries. Hence, EMU countries are being considered as non-it countries and non-emu countries are considered as IT countries. The weights are based on the size of a country s GPD level. 5

10 . Results from Previous Research Expected and actual inflation volatility Levin, Natalucci and Piger (4) measure to what extent there are differences in the volatility of inflation expectations in IT and non-it countries. 3 The authors could not find any significant difference but suggest that expected inflation could still be more anchored in IT countries since they may react differently in certain cases. For example, they may be less responsive to macroeconomic developments and instead show uncertainty of institutional character due to, in case of the UK, the possibility of adopting the euro. Ball and Sheridan (3) come to the same conclusion of insignificant difference between IT and non-it countries when measuring a broad range of macroeconomic indicators for the OECD countries. Expectation Formation Fromlet (13) evaluates how well ten inflation-targeting central banks have achieved their objective to on average keep inflation on target by using a conditional measure. Assuming the central bank is credible and has the main objective to keep inflation on the announced target it should not be possible to predict future deviation from the target today. The author evaluates whether actual inflation deviation from the central bank inflation target in one period predicts actual inflation deviation from the target in the next period. The null hypothesis (α=β=) could be rejected for six of ten countries. Levin, Natalucci and Piger (4), measure to what extent inflation targeting countries have a measurable influence on consumers inflation expectation formations. They measure differences in short-and long run expectations for IT and non-it countries and found that inflation targeting have played a significant role in anchoring long-run inflation expectations. 3 IT countries: Australia, Canada, New Zealand, Sweden and the United Kingdom). Non-IT countries: US, Japan, Denmark, and four of the largest euro area members - France, Germany, Italy and the Netherlands. 6

11 Demertzis and van der Cruijsen (11) measure whether there are differences in EMU consumers expectations about the EMU average inflation rate and the national inflation rate. The authors concluded that the ECB has managed to keep the average EMU inflation expectations in line with the inflation target, contrary to expectations of national inflation levels. The authors suggest that it may be because consumers have an inflation target in mind that is country-specific and have anchored the target to a different degree from country to country. Persistency Erceg and Leving (3) show that, when using a DGE (Dynamic General Economic) model with optimizing agents and staggered nominal contracts, inflation persistence and output costs of disinflation will rise if private agents have limited information about the central bank s objective. Orphanides and Williams (3) also stress the importance of clear policy objectives for formation of expectations. The authors use a model where economic agents are assumed to have imperfect knowledge and which results in agents having some degree of adaptive expectations. Prediction of inflation Mankiw, Reis and Wolfers (4) measure the extent of disagreement between and in-between the consumers, economists and professional economists expectations of inflation. The authors conclude that all three expectation series yield relatively accurate inflation forecasts and, despite measuring different populations, they seem to agree on the development of inflation. The survey of consumers however reveals greater disagreement (i.e. more extreme observations) compared to the survey of economists and professionals. The authors also conclude that the disagreement among professional economist rise and fall with disagreement among consumers and economists. 7

12 3. Theory and Method In this chapter I begin by explaining what inflation-targeting implies and the assumptions made to derive an inflation-targeting central bank s reaction function. I then explain the implications of having rational expectations. The chapter ends with an explanation of the method used to answer the three questions stated in the introduction. 3.1 Theory Inflation-Targeting In recent decades many central banks have adopted an inflation target since it is assumed to lead to low and stable inflation which, contrary to high and instable inflation, results in a more favorable economic development. One reason is that low and stable inflation reduces the uncertainty about future price developments. 4 Inflation-targeting is a policy aiming at price stability of a specific price index within a specified range. The chosen index in most inflation-targeting countries is the consumer price index (CPI). CPI is a cost-of-living index illustrating the development of prices and consists of average goods and services used by consumers, such as food and clothes. Having a stable CPI that is well known and published regularly facilitates consumers decisions in a market economy. The primary objective of the ECB s monetary policy is to keep the average inflation rate in the EMU below, but close to %, as measured by the Harmonized Index of Consumer Prices (HICP) over the medium term. The Swedish central bank s, Riksbanken, primary objective is an annual inflation of % over a two year period as measured by the Consumer Price Index. The primary objective of the Bank of England is to have % annual inflation at all times and is measured by the 4 The level of the inflation target is normally positive and above. However, I will not get into greater detail of why or how the inflation target is chosen. More explanation can be found in the study by Akerlof, Dickens and Perry (1996). 8

13 Consumer Price Index. The US inflation target, implicitly for a long time but explicitly stated in January 1, is an annual % increase in the price index for Personal Consumption Expenditures over the medium term. 5 There are some drawbacks when using the CPI as an operational index for monetary policy. For example, the CPI can be affected by short-run transitory effects such as fluctuations in energy and food prices which the central bank may want to disregard when implementing monetary policy. The bank can avoid transitory effects by using a CPI that excludes food and oil prices and/or be explicit in the communication of when it will not counteract what they view as transitory effects on inflation. For example, by presenting a measure excluding the current transitory shock. (Berg, 5) 3.1. The Central Bank s Model of the Economy The inflation-targeting central bank can achieve its objective by setting the refinancing rate on the overnight-market. This is a market where institutions can borrow funds from each other and from the central bank on a daily basis, with the condition to lend or sell the funds back at an agreed price. The refinancing rate is expressed as an annualized percentage and is the difference between the selling and lending price that the central bank charge for lending or buying assets from an institution. It is usually constant for at least one week. 6 A rise or lowering of the refinancing rate affects the inflationary pressure in the economy through at least three different channels in the monetary policy s transmission mechanism: 7 the short- and long-term interest rates, the volume of credit and the exchange rate. Borrowing from Fromlet (13), I assume that an inflation-targeting central bank s model of the economy looks like this: π = F(i, z, η ) (1) 5 I assume that a medium-term horizon suggest a two-year-period. 6 The instrument used to change the short-term nominal interest rate on the overnight market is called the refinancing rate by the ECB, repurchase rate or repo rate by the Bank of England and Riksbanken and the discount rate by the Federal Reserve. 7 The transmission mechanism is the process by which a central bank, through its monetary operations, affects the economy and the price level. I will not discuss the implication of a lowering or a rise in the refinancing rate but Information can be found on the respective central bank s webpage. 9

14 where π is the inflation rate in month t + τ, i is the central bank s refinancing rate in month t, z is a vector of state variables in month t and η is a set of exogenous shocks between month t and t + τ. The vector of state variables z includes variables known by the central bank, such as the inflation rate, the output gap and the inflation target. τ is the central bank s inflation target horizon, i.e. the lag in the effect of monetary policy. The aim of the inflation-targeting central bank is to minimize the expected deviation of the inflation rate τ months ahead from the inflation target π, i.e. min E [ (F(i, z, η ) π ) z ] () To get the first-order condition from equation (), I take the derivative with respect to i : If F is linear the first-order condition becomes: E [(π π ) (,, ) z ] = (3) E [π π ) z ] = (4) Given all the information available today, the expected inflation target deviation in month t + τ should be equal to zero and should have no correlation with, for example, the inflation rate in month t. A positive correlation means that the central bank has not reacted enough to the inflation in some periods, and deliberately conducted a monetary policy that was not sufficient to keep the inflation on target. This is also true for the central bank of a monetary union. Assuming that the main objective of the central bank is the inflation rate and the only information variable in month t used is the inflation deviation from target, the firstorder condition looks like this: E [(π π ) π π ] =. (5) The equation states: given today s relation between the inflation and the inflation target, there should not be any expected inflation deviation from the target. Hence, assuming that the central bank is clear in its communication of the target and 1

15 committed to it, average consumer expectations should be equal to the inflation target in the long-run, i.e. π = π Rational Expectations Rational expectations mean that agents utilize all available information in order to avoid systematic mistakes. If agents are rational when forming expectations of inflation they should take all available public and private information into consideration. As Mankiw, Reis and Wolfers (4) discusses, rational expectations assume that agents have the same information set and form expectations conditionally on that information which imply absence of disagreement. With the most explicit proof being the survey answers, it is natural to suggest that agents do not incorporate information in the same way and therefor will disagree on the future development of prices. It is also possible to assume that agents do not have the same information set. The assumption in the study of agents having rational expectations therefore relies on the premises that it is the average expectations that are measured. This allows for disagreement in expectations of inflation. For example, the inflation-targeting central bank s explicitly stated objective is a long-run objective that applies to the time-horizon that the bank s trade-off between real outcomes and inflation is met. Inflation expectations that are formed in the short run (i.e. over a 1 month period) may therefore behave differently because of the uncertainty and lagged effects of monetary policy. Shocks that affect the economy in month t (or even some months before) and cause the inflation rate to deviate from the target rate will usually not be brought back to target immediately by the central bank. This is because constant attempts to counteract short-run deviations would lead to unstable inflation levels and affect consumers arbitrarily. 8 8 As suggested by the short-term aggregate supply curve, monetary policy can affect real variables in the short run if it generates unexpected inflation. As implied by the time-inconsistency theory, such policy will lead to undesirable rates of inflation and little or no gain in real variables in the long-run. The implication is however not a guarantee that monetary policy makers will avoid such a policy. Therefore, central banks that are not expected to act on its announcements of keeping inflation on target would not be considered credible by rational consumers. 11

16 3. Method I use different equations in order to answer the questions to what extent consumers in the EMU and non-emu react differently to inflation rates, have more persistent inflation expectations and if these expectations are rational and lead to accurate predictions. I regress the equations using OLS with Newey-West standard errors after conducting the Augmented Dickey-Fuller unit root test. The Augmented Dickey- Fuller tests non-stationarity of the series because stationarity is required in order to get useful results. Statistically significant results suggest that the series are stationary Actual and Expected Inflation Deviation from Target First, I measure the average deviation and the absolute average deviation of actual and expected inflation deviation from the inflation target in EMU and non-emu countries. These are two standard statistics that provide quantitative evaluations of the actual and expected inflation series. Counterbalancing errors will lead to smaller average deviation than average absolute deviation Expectation Formation Assuming that consumers are rational and the central bank is credible 11 and has inflation stabilization as an explicit objective, the consumers inflation expectations in an inflation-targeting country should be equal or close to the target on average. Hence, a deviation from the target in month t should lead people to expect insignificant deviation in month t To test this hypothesis I use the following equation: π, π = α + β(π π ) + ε, (6) 9 The null hypothesis of a unit root (i.e. non-stationarity) could be rejected for all the series. Hence, all of the series, π,, π, π, π and π π are stationary. The results are presented in Appendix E. 1 Additional estimates of the standard deviation of the actual and expected inflation series are presented in table B3 and in figures B3 and B4 in Appendix B. 11 A credible inflation-targeting central bank is the one that is believed to be firmly committed to low inflation and which is time-consistent. A transparent policy means one that the public understands to be credibly committed to low inflation. (Friedman, ) 1 The actual/expected inflation rate is the actual/expected year-on-year inflation rate: π /π, =. 1

17 where π, is the expected annual inflation rate 1 months ahead measured in month t, π is the inflation target, π is the annual inflation rate in month t, and ε represent unexpected factors that affect expectation of inflation in month t + 1. The result gives an indication to what extent inflation deviation from the inflation target in month t leads consumers to expect inflation deviation from the target in month t + 1. The null hypothesis is α = β =. If I cannot reject the null hypothesis, I conclude that inflation expectations are on average equal to the inflation target and that the central bank has been credible in its objective to keep the inflation on target. If β is positive and significant, this suggests that inflation expectations deviate from the target because the actual inflation deviates from the target. An insignificant β suggests that actual inflation deviation from the target does not explain deviation of expected inflation of from the target. That is, consumers expectations of inflation are not affected by the current inflation rate. A positive (negative) significant α suggest that the consumers underestimate (overestimate) actual inflation deviation from target Persistency of Expected Inflation Assuming that consumers are rational and hence update their expectations when presented with new information, previous expectations of inflation should not affect current expectations of inflation. If the central bank is credible, expectations equal to the inflation target may suggest high a degree of persistency. I use the following equation to measure persistency: π, = δ + ρπ, + ε, (7) where π, is the expected annual inflation rate in t + 1 month and measured in month t, π, is the expected inflation rate in month t and measured in month t 1 and ε contains unexpected factors that affect the expectation of inflation in month t A constant α > imply that when π = π then π, π >. 13

18 A significant ρ indicates the extent of persistency in consumers expectations of inflation. An insignificant ρ suggests that consumers are not having persistent (i.e. backward-looking) expectations. The δ is not interpreted as meaningful in this analysis Predicting Inflation To measure if EMU and non-emu consumers form expectations of inflation rationally, i.e. if the degree of persistent inflation expectation and reaction to inflation deviation can be used to estimate actual inflation, I us the following equation 14 : π = Φ + θπ, + ε, (8) where, π is the annual inflation rate in month t, π, is the expected inflation rate in month t and measured in month t 1 and ε is the unpredictable deviation from target in month t + 1. The equation gives an indication of the correlation between expectations and actual inflation. A significant θ close to 1 suggests that consumers expectations could be used as estimates of the actual inflation rate while an insignificant θ suggests that consumers expectations should not be used as estimates of actual inflation. A positive (negative) significant Φ suggest that consumers on average underestimating (overestimating) actual inflation. That is, changes in expectations For example, a significant positive θ > 1 (θ < 1) suggest that consumers expectations on average give rise to greater (smaller) change in actual inflation than the initial change in expectations. A significant θ = 1 predicts changes in average inflation expectations to equal changes in average actual inflation. 14 Other common measures of forecast accuracy are the average absolute deviation and the Root Mean Squared Error (RMSE) of inflation expectation deviation from the actual inflation. The results are presented in table B3 in Appendix B. 14

19 4. Data The data obtained for actual inflation and the data I use to estimate the expected inflation are presented in this chapter. The chapter ends with an explanation of the approach taken to convert answers from a qualitative survey into a quantitative estimate of expected inflation. 4.1 Overview of Data In this study I analyze direct survey measures of inflation expectations of 1 EMU countries and three non-emu countries. 15 I chose to analyze these countries because the data is publically available and conducted on a monthly basis. For example, all the countries in the European Union (EU) use the same standardized consumer survey, which is similar to the US consumer survey, to obtain a quantitative estimate of expected inflation. This is in contrast to other inflation targeting countries such as New Zealand, which collects answers from consumers concerning their perceptions of inflation on a quarterly basis and Australia, which obtains the expected inflation series from private market participants and for which past series are not publically available. The data of the expected inflation and annual inflation series for the EU countries are obtained from the European Commission database and the Eurostat statistical database respectively. The US inflation expectation series (The Surveys of consumer) are obtained from the Survey Research Center at the University of Michigan 16 and the annual inflation series from the OECD database. 17 I study two different time periods which I call the full-period and the sub-period. The full-period covers the period from January 1998 to September 13 and includes the 15 EMU: Austria, Belgium, Finland, France, Germany, Ireland, Italy, Netherlands, Portugal and Spain. Non-EMU: Sweden, United Kingdom (UK) and the United States (US). Sweden, UK and the EMU countries are part of the European Union (EU). 16 The questions asked are: 'During the next 1 months, do you think that prices in general will go up, or go down, or stay where they are now?' and 'By what percent do you expect prices to go up, on the average, during the next 1 months? 17 I chose to use the price index explicitly stated in the monetary objective of the EMU and non-emu central banks because it is the index consumers are assumed to base their expectations on. The items that are excluded or included in the target price index are publically available. 15

20 period of the financial-crisis and the great uncertainty that occurred from mid-8 until recently. The sub-period is chosen so that the time of the financial-crisis is not covered; hence, the period is January 1998 to April The European Consumer Survey The European Consumer Survey (EC survey) of expected inflation is a direct source of information from the answers of approximately 5, consumers in the EU on their perception of how consumer prices will develop during the next twelve months in their own respective countries. 19 The survey is harmonized across all European countries and the standardized question, as expressed in Forsells and Kenny (), is as follows: By comparison with the past 1 months, how do you expect that consumer prices will develop in the next 1 months.. 1. there will be more rapid increase in prices,. prices will increase at the same rate, 3. prices will increase at a slower rate, 4. prices will stay about the same or 5. prices will fall slightly. As mentioned in Berk (), the use of survey data should be handled and interpreted with care since the answers given can be different from the ones that consumers act upon, are sensitive to sampling errors and how the questions are formulated. In the following analysis I assume that the wording of the questions in the EC survey and the US survey do not affect the survey respondents. Hence, I assume that they all interpret the questions in the same way. 4. The Probabilistic Approach In general the result of the survey is presented as a balance statistic. Balance is the difference between positive and negative answering options measured as percentage 18 From the period May 8 to May 9 there is no data of expectations available for Ireland. 19 All the covered countries except of the US. There s a sixth response option: don t know, but is not given any weight. 16

21 points of total answers. Hence, it only gives a qualitative indication of the expected change in inflation in the next 1 months. 1 To obtain a quantitative estimate of the expected rate of inflation from the EC survey, I will use the Probabilistic Approach. It is used by Forsells and Kenny () and is based on previous studies by Carlson and Parkin (1975), Batchelor and Orr (1988) and Berk (), among others Quantifying the Qualitative European Consumer Survey The probabilistic approach can be explained as giving the share of respondents replying to each answering option a probability, i.e. as a maximum likelihood estimate of areas under the density function of aggregate inflation expectations. This is pictured in figure. Figure : The Aggregate Probability Distribution for Expected Inflation S S S S S π, π, λ λ π, κ π, + κ There are five response thresholds that divide the distribution. For example, consumers expect the inflation to rise (fall) if the expected inflation rate, π,, is at least λ units above (below) zero. Consumers are also expected to reply conditionally on their perceptions of past inflation, π,. Hence, if the expected inflation rate is at 1 Denoting S (for i = 1,,3,4 and 5) as the sample proportions choosing one of the five response categories, the balance statistic is calculated as (S + S ) ( S +S ). For example, if 1,59% of the respondent answer S, using the normal distribution with (μ,σ)=(,1), the answer is approximately -1,5. By using equation (13) the answer is converted into a positive sign. 17

22 least κ units above (below) the perceived rate of inflation, they expect prices to increase at a more (less) rapid pace. I use equations (9) to (1) to estimate the expected inflation rate π,, its standard error (σ) and the two response thresholds (κ and λ) with the assumption of continuous expectations drawn from a normal distribution. I denote S, (for i = 1,,3,4 and 5) as the sample proportions choosing, in month t 1 for month t, one of the five answering options stated above. I use the following equations 3 : π, = π, [ (,, ),,,, ] (9) σ, = π, [,,,, ] (1) κ = π, λ = π, [ [ (,, ),,,, ( ),,,, ] (11) ] (1) N denote the inverse of the normal probability distribution function. Hence, Z, = N [1 S, ], (13) Z, = N [1 S, S, ], (14) Z, = N [1 S, S, S, ], and (15) Z, = N [S, ]. (16) There is no reason to expect that consumers, even if behaving rationally, have the same belief about inflation as the inflation rate represented by official reports. Therefore, I estimate the perceived rate of inflation by using the answers in the EC survey from the question of past price developments, as discussed in Berk (). The harmonized question asks how the respondent perceives that consumer prices have developed over the last 1 months. The five answer options stated above are 3 See the Appendix in Berk (1999) for a full derivation and further discussion of these equations. 18

23 boiled down to three conclusions: up (if respondents answer S, S and S ), unchanged (if respondents answer S ) and down (if respondents answer S ). By using the probability approach with only three thresholds and assuming continuous expectations drawn from a normal distribution, consumers expect the inflation to have increased or decreased if the expected perceived inflation is at least κ units above or below zero, respectively. I use equations (13) to (16) and the following two equations to estimate the perceived inflation rate. π, =, and (17) π, = 4 month moving average of (π, /π ) π, (18) The π, is the average expected perceived rate of inflation in month t 4 and measured in month t 1, the π, is the estimated perceived rate of inflation, the 4 month moving average of (π, /π ) is the 4 month moving average of the consumers expected perceived inflation as a ratio of actual inflation Modifications and Drawbacks of the Probabilistic Approach There are some modifications in the above equations compared to the original Carlson and Parkin (1975) results. For example, the scaling parameter (i.e. the perceived rate of inflation) was estimated by imposing unbiased expectations but is here derived directly from the EC survey. This modification has an important advantage since it helps facilitates empirical testing regarding possible bias in consumers inflation expectations. Another advantage is that the response thresholds are allowed to vary over time which means that no ad hoc assumptions need to be made and are estimated from the survey replies. For example, in the case of high aggregate uncertainty of inflation, the thresholds would also be high, as mentioned in Forsells and Kenny (). A disadvantage of the probability approach is that even if the share of respondents expects prices to increase faster, it will not lead to a higher quantitative estimate of expected inflation. 4 The expected perceived rate of inflation is the rate which is derived from the EC survey answers. It is not necessary to multiply it with a 4 month moving average of actual inflation or any other constant, but is done in order to take into account changes in actual inflation over the target-horizon. 19

24 5. Results and Implications In this chapter I present the results for actual and expected inflation deviation, consumers formation of inflation expectations, persistency of expectations and prediction accuracy in EMU and non-emu countries. The chapter ends with a discussion of the implications that the results suggests. 5.1 Results Statistics for Actual Inflation The estimated results for average deviation of actual inflation from the inflation target during the full-period and the sub-period are presented in figure 3 to 6. 5 Figure 3: Average Deviation of Actual Inflation from Target Inflation: Full-Period Jan Sep 13 1,5 1,5 -,5-1 The results in figure 3 suggest that four EMU countries have average deviations above the inflation target and six countries have average deviations below the target during the full-period. The average deviations for the non-emu countries during the full-period are below target for Sweden while it is above the target for the UK and 5 The average deviation and average absolute deviation rates are presented in Appendix B.

25 US. The results suggest that the EMU and the non-emu average deviation rates are approximately in the same deviation range. Figure 4: Average Actual Inflation Deviation from Target Inflation: Sub-Period 1,5 Jan Apr 8 1,5 -,5-1 During the sub-period the EMU countries average inflation deviation are below the inflation target. The non-emu countries Sweden and UK have average deviation rates below the target while the US has an average deviation rate above the target. The EMU and non-emu average actual inflation deviation rates are roughly in the same deviation range. Comparing the average inflation deviation of the respective EMU countries during the full-period and the sub-period, the results suggest that the deviations are smaller for all EMU countries during the full-period, with the exception of Belgium and France. The same is true for the UK and the US which have smaller deviations during the fullperiod compared to the sub-period. The EMU and non-emu average actual inflation rates are also to some extent smaller during the full-period relative to the sub-period. Average Absolute Deviation of Actual Inflation from the Inflation Target The results presented in figure 5 of average absolute deviation for the EMU and non- EMU countries during the full-period suggest that the deviations are approximately in the same deviation range, with Ireland having a relative extreme value. The respective EMU and non-emu average absolute deviations are similar. 1

26 Figure 5: Average Absolute Deviation of Actual Inflation from Target Inflation: Full-Period 1,5 Jan Sep 13 1,5 The estimated rates for the sub-period average absolute deviations presented in figure 6 suggest that the deviations for the EMU and non-emu countries are about the same, with Italy having a relative small deviation. The EMU and non-emu average absolute deviation rates are also similar. Figure 6: Average Absolute Deviation of Actual Inflation from Target Inflation: Sub-Period 1,5 Jan Apr 8 1,5 The average deviation rates are smaller than the average absolute deviations rates during both periods for all respective countries suggesting that the countries inflation rates have been counterbalancing. Both the average deviation and the average absolute deviations during the full-period have been larger relative to the sub-period.

27 5.1. Statistics for Expected Inflation The results for average deviation and average absolute deviation of expected inflation deviation from the inflation target, in the EMU and non-emu countries during the full-period and sub-period, are presented in Figure 7 to 1. Figure 7: Average Deviation of Inflation Expectation from Target Inflation: Full-Period Jan Sep 13 1,5 1,5 -,5-1 The estimated results presented in figure 7 suggest that Ireland, Portugal and Spain, contrary to the other 7 EMU countries, have average deviations above the inflation target. The average deviations for the non-emu countries are below the target for Sweden and UK and well above the target for the US. The average deviation rates for the EMU and non-emu countries are roughly in the same deviation range. Figure 8: Average Deviation of Inflation Expectation from Target Inflation: Sub-Period Jan Apr 8 1,5 1,5 -,5-1 3

28 Figure 8 suggest that the result for the EMU and non-emu average deviations are to some extent different. The EMU countries average deviations are in general below the target with the exception of Ireland, Portugal and Spain. The non-emu average deviations are both below and above the inflation target. The results in figure 7 and 8 suggest that the average deviations for the EMU and non-emu countries are roughly in the same range during both the full-period and the sub-period, with US having the largest average deviations. The most notable difference between the average deviations in the full-period compared to the subperiod is for UK with an above average deviation during the full-period and a below average deviation in the sub-period. Average Absolute Deviation of Inflation Expectations from the Inflation Target The results in figure 9 suggest that the EMU and non-emu average absolute deviations are in the same range during the full period, with the exception of Ireland and Portugal. The EMU and non-emu average deviations are approximately in the same deviation range. Figure 9: Average Absolute Deviation of Inflation Expectations from Target Inflation: Full-Period 1,5 Jan Sep 13 1,5 Comparing the EMU countries average absolute deviations in the full-period with the sub-period deviations presented in figure 1, the results suggest that the deviations are larger during the full-period, with the exception for the Netherlands. The average 4

29 absolute deviations for the non-emu countries are also to some extent larger during the full-period compared to the sub-period. Figure 1: Average Absolute Deviation of Inflation Expectations from Target Inflation: Sub-Period 1,5 Jan Apr 8 1,5 The average deviations are in general smaller and the average absolute deviations larger during the full-period compared to the sub-period for all countries suggesting that consumers expectations are counterbalancing and associated with greater uncertainty How Does Expected Inflation React to Actual Inflation? Table 4 and 5 presents the regression results of the correlation between actual inflation deviation and inflation expectations deviation during the full-period and the sub-period respectively. The results give an indication of how credible the inflation target is when consumers form their expectations of inflation in the short run. The null hypothesis of α = β = is rejected for all countries on high significant levels during the full-period as suggested by the results in table 4. The results imply that consumers take into consideration current inflation levels when forming expectations of the actual inflation rate essentially the same. This is true for all the countries where β is both highly significant and around the same magnitude. 5

30 Table 4: Test of the impact of inflation deviation from target on expected inflation, Full-period: Jan 1998 Sep 13 Country α β α = β = Austria -.37***.86*** *** (.43) (.5) Belgium -.441***.484*** *** (.45) (.43) Finland -.44***.719*** *** (.4) (.46) France *** 44.54*** (.34) (.4) Germany -.415***.78*** 41.8*** (.5) (.46) Ireland.54.57*** *** (.51) (.) Italy -.683***.639*** 3.64*** (.39) (.53) Netherlands -.419***.79*** 5.86*** (.41) (.4) Portugal.7.737*** 6.6*** (.8) (.69) Spain -.466***.667*** *** (.58) (.41) Sweden -.316***.71*** *** (.59) (.61) UK -.475***.738*** 347.8*** (.4) (.54) US 3.44*** (.58).334*** (.51) *** Used equation: π, π = α + β(π π ) + ε A variable marked by *, **, *** indicates significance at the 1, 5 and 1% levels, respectively. Mean errors in parentheses The constant α is significant and negative for all countries suggesting that consumers systematically underestimate the actual inflation rate and expect it to be below the target. The exceptions are France, Ireland, Portugal with insignificant constants and the US with a systematic overestimation of the actual inflation rate. The regression results for the sub-period presented in in table 5 suggest that the UK and Italy have smaller β and the US, Portugal and France higher β compared to the full-period. That is, consumers react differently to inflation both in EMU and non- EMU countries during different time periods. The constant α is significant and negative for most countries during the sub-period but slightly different from the rate in the full-period. For example, the US seems to overestimate the actual inflation rate to a larger extent during the full-period compare to the sub-period, contrary to most of the other EMU and non-emu countries which have similar constants during the full-period and the sub-period. 6

31 Table 5: Test of the impact of inflation deviation from target on expected inflation, Sub-period: Jan 1998 Apr 8 Country α β α = β = Austria -.15***.933*** (.6) (.86) *** Belgium -.473***.567*** (.48) (.58) *** Finland -.478***.678*** (.4) (.47) *** France -.55***.656*** (.37) (.33) 547.1*** Germany -.465***.69*** (.68) (.61) 3.58*** Ireland.188**.548*** (.99) (.54) 15.5*** Italy -.599***.436*** (.55) (.94) 6.8*** Netherlands -.41***.855*** (.49) (.49) 184.5*** Portugal *** (.19) (.171) 19.*** Spain -.47***.745*** (.89) (.8) 45.7*** Sweden -.37***.79*** (.73) (.7) 43.89*** UK -.55***.615*** (.77) (.16) 48.6*** US 1.79***.531*** (.54) (.53) *** Used equation: π, π = α + β(π π ) + ε A variable marked by *, **, *** indicates significance at the 1, 5 and 1% levels, respectively. Mean errors in parentheses Overall, the results in table 4 and 5 suggest that inflation deviation from target affect consumers expectations of inflation and are not equal to the inflation target on average. The actual, expected and target inflation rate for the respective EMU and non-emu countries during the full-period is presented graphically in the figures in Appendix C Persistency of Inflation Expectations Table 6 and 7 presents the regression results of consumers expectations and to what extent they are persistent during the full-period and the sub-period respectively. The results give an indication to what extent previous expectations of inflation affect inflation expectations. There is, as suggested by the highly significant results in table 6, a high degree of persistency in expectations of inflation in both EMU and non-emu countries during 7