Versione Inglese Pdg Sabbatini - Volume Completo_03!10!2007

8/4/2019 Versione Inglese Pdg Sabbatini - Volume Completo_03!10!2007

1/194

The euro cash changeover, inflation and

consumers perceptions.

Lessons from the case of Italy.

Paolo Del Giovane and Roberto Sabbatini


2/194

V

To Lorenza, Luciano and Marco

To Pierluigi and Riccardo


3/194

VI


4/194

VII

Foreward [TO BE INCLUDED]


5/194

VIII


6/194

IX

Acknowledgments

This book is the result of a collective research effort conducted in theResearch Area of the Bank of Italy between 2003 and 2007. We wouldlike to thank all the authors of the chapters collected in this volume andthe other colleagues of the Bank who provided stimulating comments atvarious stages of the project, in particular..... Special thanks go to FrancescoLippi, who, as the co-editor of a previous publication on the same topic,shared with us the challenge of describing and interpreting the complex

phenomena treated in this volume. We have also benefited greatly from thecomments received by the participants in various seminars and conferenceswhere the material reported in the book was presented, in particular twoseminars at the European Central Bank and the conferences organized by theOECD (June 2005), the Western Economic Association International (June2006), the National Euro Changeover Committee of Malta (May 2007), theGovernment of Malta in cooperation with the European Central Bank andthe European Commission (October 2007). Finally, we would like to thankDaniel Dichter, Carla Lucidi, Roger Meservey and Christine Stone for theirexcellent linguistic assistance and the Bank of Italys printers for their helpin producing the book to their usual high standards. We acknowledge Il

Mulino publishing house for the authorization to use part of the material

published in Italian in the volumeLeuro e linflazione edited by Paolo DelGiovane, Francesco Lippi and Roberto Sabbatini (Bologna, 2005).The opinions expressed in the volume are those of the editors and the

other authors, and they do not necessarily reflect the views of the institutionsthey belong to.


7/194

X


8/194

XI

Contents

Foreword [TO BE INCLUDED]Acknowledgments

Contributors

List of abbreviations

IntroductionPaolo Del Giovane and Roberto Sabbatini

1 Perceived and measured inflation after the launch of the euro: explainingthe gap in ItalyPaolo Del Giovane and Roberto Sabbatini

2 Rounding and anomalous changes in Italian consumer prices in 2002Franco Mostacci and Roberto Sabbatini

3 Pricing behaviour and the introduction of the euro: evidence from a panelof restaurantsEugenio Gaiotti and Francesco Lippi

4 Did prices really soar after the euro cash changeover? Evidence fromATM withdrawals

Paolo Angelini and Francesco Lippi5 Memory for prices and the euro cash changeover: an analysis for cinema

prices in ItalyVincenzo Cestari, Paolo Del Giovane and Clelia Rossi-Arnaud

6 What are inflation perceptions about? A survey-based analysis onItalian consumers [TO BE INCLUDED]Paolo Del Giovane, Silvia Fabiani and Roberto Sabbatini

7 Impact of the cash changeover on prices and inflation perceptions in theeuro area: a survey of the empirical evidencePaolo Del Giovane and Roberto Sabbatini

List of figures

List of tablesAuthors index [TO BE INCLUDED]

Subject index [TO BE INCLUDED]


9/194

XII


10/194

XIII

Contributors

Paolo Angelini Director in staff, Department of Economic Outlookand Monetary Policy Studies, Bank of Italy.

Vincenzo Cestari Professor of Psychology, Lumsa University, Facultyof Educational Science, and National ResearchCouncil (CNR), Institute of Neuroscience.

Paolo Del Giovane Head of Monetary Analysis Division, Departmentof Economic Outlook and Monetary Policy Studies,Bank of Italy.

Silvia Fabiani Senior Economist, Department of Economic Outlookand Monetary Policy Studies, Bank of Italy.

Eugenio Gaiotti Deputy Head of the Department of EconomicOutlook and Monetary Policy Studies, Bank ofItaly.

Francesco Lippi Professor of Economics, University of Sassari, and

CEPR.Franco Mostacci Senior Researcher, Price Statistics Department,

Italian National Statistical Institute (ISTAT).

Clelia Rossi-Arnaud Professor of Psychology, University of Rome LaSapienza, Psychology Department.

Roberto Sabbatini Deputy Head of Economic Outlook Division,Department of Economic Outlook and MonetaryPolicy Studies, Bank of Italy.


11/194

XIV


12/194

XV

List of abbreviations

ATM Automated Teller Machine

CNR Consiglio Nazionale delle Ricerche (National Research Council)

CPI Consumer Price Index

EC European Commission

ECB European Central Bank

EU European Union

GR Il Gambero Rosso

HICP Harmonized Index of Consumer Prices

INSEE Institut National de la Statistique et des tudes conomiques

ISAE Institute for Economic Research and Analysis

ISTAT Italian National Institute of Statistics

NSI National Statistical InstituteOLS Ordinary Least Squares

SD Standard Deviation

SHIW Survey of Households Income and Wealth

UIC Ufficio Italiano Cambi

ULC Unit Labour Costs


13/194

XVI


14/194


15/194


16/194

1. Perceived and measured inflation after the launchof the euro: explaining the gap in Italy

Paolo Del Giovane and Roberto Sabbatini

Bank of Italy, Department of Economic Outlook and Monetary Policy Studies

1.1 Introduction1

Following the euro cash changeover, residents in the area perceived a muchlarger rise in prices than was recorded in official statistics. The gap betweenperceived and measured inflation reached an unprecedented size in thecourse of 2002, then narrowed or closed in most countries. In Italy, it wasextremely large and persistent, only beginning to diminish after two years.The perceived surge in prices was popularly associated with the introductionof the euro, even though national statistical institutes and central banksestimated the impact to be generally modest.

In the ensuing debate between experts, consumer organizations andmembers of the public conducted under the spotlight of the media, thecredibility of official data on prices was called sharply into question. Onthe basis of their everyday experience, households became convinced thatinflation rates were much higher than official figures. In Italy there was awidespread notion that retailers had converted the prices of many goods andservices at a rate of 1,000 lire to the euro (rather than at the official rate of1,936.27 lire), which would have implied an almost 100 per cent price hike.Private research centres published alternative estimates of inflation that weremuch higher than those of ISTAT, the National Institute of Statistics, andtended to bear out, if not actually reinforce, the publics impressions.

The gap between perceived and measured inflation is an importantphenomenon: it impairs firms and consumers ability to judge individual

1 The authors thank Luigi Cannari, Vincenzo Cestari, Lorenzo Forni, Luigi Guiso,Francesco Lippi, Marco Magnani, Stefano Siviero, Daniele Terlizzese, andparticipants in ECB, OECD and WEA workshops, for useful comments. Theyacknowledge Giornale degli Economisti e Annali di Economia for the authorizationto reprint this article, with small changes, from Volume 65, No. 2 (November 2006;copyright Universit Commerciale Luigi Bocconi).


17/194

4 P. Del Giovane and R. Sabbatini

prices, eroding the allocative efficiency of the price system; by alteringexpectations of future price trends, it can distort decisions on prices andwages; the credibility of monetary policy may be undermined if the qualityof the underlying price indices is called into question; and, in the case ofthe euro area, the publics acceptance of the common currency and itsinstitutional framework may be shaken.

In this chapter we first define the subject of inquiry and then examinewhether the observations based on personal experience are truly irreconcilablewith the official statistics. We systematically review the various argumentsoffered for the sudden worsening of inflation perceptions, starting from thehypothesis that a combination of factors was at work. In the case of Italy,we explore the possibility that a major cause was the interaction betweenhow individual perceptions are formed (e.g. attributing more weight toprice increases than to price cuts) and the observed behaviour of pricesand other economic variables during and after the cash changeover (e.g.larger increases in the prices of frequently purchased products). We alsoinvestigate the collective dimension of the phenomenon by considering therole of the media.

The chapter is organized as follows. Section 1.2 describes the indicatorcommonly used to measure inflation perceptions. Section 1.3 examineswhether there are grounds for doubting the accuracy of the ISTAT indexand highlights a few factors that might be of some importance in explainingthe discrepancy with respect to perceptions. On the hypothesis that the gaphad numerous causes, section 1.4 analyzes and assesses other explanations,including the interaction between the psychology of individuals inflationperceptions and price developments following the changeover. Section 1.5provides an overview of the correlations between inflation perceptions andthe factors we identified. Section 1.6 summarizes and concludes.

1.2 Inflation perceptions

Public debate on the perception of a much sharper increase in consumerprices than that measured by official statistics has generally relied on

anecdotal evidence, personal impressions very often limited to specificproducts or classes of goods and services. Yet a survey is available thatoffers information on the perceptions of a representative sample of theentire population, based on questions regarding inflation in general: it ispart of the harmonized monthly consumer surveys conducted in all thecountries of the European Union, and in Italy by the Institute for EconomicResearch and Analysis (ISAE). The European Commission aggregatesthe results of the national surveys and publishes indicators for the whole


18/194

Perceived and measured inflation after the launch of the euro 5

European Union and for the euro area in the monthly press releaseBusinessand Consumer Survey Results.

On the subject of prices, respondents are asked to state their viewsregarding price developments in the last and the next twelve months.Regarding the past trend, consumers are asked: How do you think thatconsumer prices have developed over the last 12 months?. Answers mustbe chosen from the following list: risen a lot (N1), risen moderately(N2), risen slightly (N3), stayed about the same (N4), fallen (N5), anddont know. A composite measure of respondents opinions is obtainedas the balance between the weighted frequencies of the different responses.The extreme responses (risen a lot, fallen) are assigned twice the weight

of the intermediate responses, while the central response (stayed about thesame) is not counted. The balance statistic (S) is therefore obtained as S =

N1+ 0.5 N2 0.5 N4 N5, whereNidenotes the relative frequency of the

i-th response.Fig. 1.1 traces the inflation rate calculated by the statistical institutes

and the balance of perceptions calculated as just described for Italy andfor the euro area. The indicator of perceived inflation is purely qualitative the respondents are not asked to make a numerical estimate and can betranslated into a quantitative measure only with further assumptions (seebelow); in interpreting the figure, we can therefore compare the behaviourof the two curves but not their level. The behaviour of inflation measuredby the statistical institutes and that of consumers perceptions were closely

correlated between 1991 and 2001 both in Italy and in the euro area, althoughvariations in perceptions were more marked, especially in Italy. Betweenthe end of 2001 and the beginning of 2003, while official inflation remainedbroadly stable, perceived inflation increased sharply:2 the percentage ofItalian consumers who thought prices had risen a lot jumped from 10to 49 per cent; also counting those who answered risen moderately, theproportion almost doubled, from 48 to 89 per cent.

Inflation perceptions increased in 2002 in all the euro-area countries.In most countries, however, the trend reversed in the course of 2003; theturnaround was clear-cut in particular in Germany, where by the end of theyear the balance was lower than before the changeover,3 less marked inother countries, where perceptions settled at a fairly high level. In Italy andGreece, the deterioration was exceptionally large and persistent; in Italy

2 See European Central Bank (2002a, 2002b, 2003b, 2007) and Aucremanne andCollin (2005) for a description of the phenomenon for the euro area as a whole,and chapter 7 of this book for a survey of the studies available for individualcountries.

3 For an analysis of the evolution of prices and perceptions in Germany, seeDeutsche Bundesbank (2004).


19/194


(a) Italy

0.0

1.0

2.0

3.0

4.0

5.0

6.0

7.0

0

10

20

30

40

50

60

70

Perceptions

(qualitative indicator) (1)

Inflation as measured (2)

1991 1992 1993 199919981997199619951994 2000 2001 2002 2003 04

(b) Euro area

0.0

1.0

2.0

3.0

4.0

5.0

6.0

7.0

8.0

-10

0

10

20

30

40

50

60

70

Inflation as measured

Perceptions (qualitative indicator) (1)

1991 1992 1993 199919981997199619951994 2000 2001 2002 2003 04

Fig. 1.1. Inflation measured by the statistical institutes and inflation perceptions(twelve-month percentage change in the harmonized consumer price index;percentage balances of response).

Source: European Commission and Eurostat.

Notes: (1) Qualitative indicator obtained as the percentage balance of responses tothe monthly consumer survey (right-hand scale). (2) Twelve-month rate of increasein the harmonized index of consumer prices (left-hand scale).


20/194


peak values were recorded throughout 2003, only beginning to decline inthe first half of 2004.

The populations perception of pronounced price inflation went hand inhand with the belief that the introduction of the euro was the direct cause.Fig. 1.2, which is based on the European Commission survey The uro,Two Years Later, carried out between October and November 2003, showsthat the change in euro-area residents perceptions was positively related tothe conviction that prices were converted into euros to the disadvantage ofconsumers (panel a). This conviction was most widespread in Italy, beingheld by 96 per cent of respondents, against 89 per cent for the euro area asa whole. Fig. 1.2 also shows that higher inflation perceptions were usuallyaccompanied by greater dissatisfaction with the common currency (panelb). The Italians and Greeks, who had originally been very much in favourof European unification, were the most dissatisfied after the Germans. Bythe end of 2003, 44 per cent of Italians declared they were quite/veryunhappy that the euro had become their currency, 13 percentage pointsmore than in the survey conducted a year earlier. A comparable increasewas recorded only in Greece (14 percentage points), against an area averageof 7 points. It is worth noting that the statistical significance of the relationis diminished by Germanys outlier position; in fact, unhappiness aboutthe euro was greater there than elsewhere in the euro area, even thoughinflation perceptions diminished during 2003. In all likelihood, Germandissatisfaction was due more to the performance of other macroeconomicvariables, notably economic growth and employment, than to the perceivedtrend of inflation. It cannot be ruled out that the same factors affectedconsumer sentiment towards the euro in other countries as well.

The survey responses, which blame the euro for most of the perceived risein inflation, contrast sharply with the fairly low inflationary impact of thecash changeover estimated by the central banks and statistical institutes. 4

Although, as we pointed out earlier, there are no systematic surveysasking respondents to put a number on the inflation rate they perceive,5 itis possible to transform the European Commissions qualitative indicatorinto a quantitative measure for comparison with the official statistics. Anumber of different methods are available. One is to regress measured

4 See European Central Bank (2003a) and Mostacci and Sabbatini in chapter 2 ofthis book, respectively, for an analysis of the impact in the euro area and in Italy.

5 Two new questions aimed at obtaining quantitative estimates of inflationperceptions and expectations were introduced into the harmonized consumersurveys, on an experimental basis, in 2003 (Lindn 2006). The resulting dataset isnot public. Quantitative inflation perceptions are investigated in depth in chapter6, for different time horizons and different consumption bundles.


21/194


(a) Relation between change in inflation perceptions and conviction that priceconversion into euros was to the detriment of consumers (1)

0.0

0.5

1.0

1.5

2.0

2.5

3.0

3.5

4.0

4.5

70 75 80 85 90 95 100

0.0

0.5

1.0

1.5

2.0

2.5

3.0

3.5

4.0

4.5

BE

FI

PT

IE

FR

ES

AT

DE

NL

IT

EL

(b) Relation between change in inflation perceptions and unhappiness about the euro (2)

0.0

0.5

1.0

1.5

2.0

2.5

3.0

3.5

4.0

4.5

10 20 30 40 50 60 70

0.0

0.5

1.0

1.5

2.0

2.5

3.0

3.5

4.0

4.5

IE

BE

FI

PT

ES

FR

NL

AT

IT

EL

DE

Fig. 1.2. Inflation perceptions and sentiment about the euro.

Source: European Commission (2003).

Notes: (1) The y-axis plots the change in the qualitative indicator of inflation perceptions(ratio of the average balance in 2002-03 to the figure for December 2001). The x-axisplots the percentage of those who responded to the detriment of consumers to thequestion Did you personally notice that in your country, when converted into euros,prices have been: rather in favour of consumers; rather to the detriment of consumers;or one way or another the rises and falls balanced out?. (2) As in panel (a), the y-axisshows the change in inflation perceptions. The x-axis plots the percentage of those whoanswered quite/very unhappy to the question Are you personally: very happy, quitehappy, quite unhappy, or very unhappy that the euro has become your currency?.


22/194


inflation (the twelve-month increase in the general consumer price index)on the qualitative balance of perceptions. This method, in which the fittedvalue of the regression is interpreted as a quantitative measure of perceivedinflation, is based on the assumption that in the long run perceived inflationis equal to officially measured inflation, i.e. that consumers do not makesystematic errors over a sufficiently long time span.6 By estimating theequation for Italy in the period 1992-2001, the perception was quantified ataround 6 per cent at the end of 2003, with a 95 per cent confidence intervalof 1.5 percentage points. The results of the quantification proved to be verysensitive to the choice of the estimation period; for example, referring to ashorter time horizon (since 1997), estimated perceived inflation amountedto around 3.5 per cent.Given the uncertainty surrounding these estimates, inthe rest of this paper we prefer to use the qualitative balance of perceptionswithout further elaboration.

1.3 The official index

In the debate on the gap between perceived and measured inflation,the official consumer price indices have often been criticized for usinginappropriate methods of calculation and underestimating the inflationactually experienced by households. In Italy these assertions have been

upheld, among others, by a number of consumer associations and appearto be borne out by alternative estimates of inflation that are much higherthan those of ISTAT. In many cases, though, the methods underlying theseestimates are not well-documented, particularly the size and nature of thesample, the criteria for recording prices, and the weights assigned to thesingle products for the calculation of inflation (see below).

The rest of this section examines whether there is cause to doubt theaccuracy of the official index, and highlights some elements that might helpto explain the difference with respect to perceived inflation.

In the course of the 1990s the methods used in the EU countries to calculatethe consumer price indices were harmonized in several essential respects.The definition of the index, the items in the basket, the classification, the

formulae for aggregating elementary data, the frequency of re-basing, and thetreatment of some specific items were all standardized (Eurostat 2001). As a

6 ISAE (2002) adopts an alternative approach (the latent variables method)to quantify perceptions; the results are comparable to ISTATs measure ofinflation.


23/194


result of this process, the methods employed today by the national statisticalinstitutes, including ISTAT, to calculate the harmonized index of consumerprices (HICP) can be said to correspond to the best practices availableinternationally.7 However, there is undoubtedly room for improvement. InEurope, efforts are being made to improve quality adjustment and to assesshow to treat owner-occupied housing, currently not included in the basket (seeEuropean Central Bank 2005). As regards Italy in particular, the treatmentof specific items may need to be revised. An example is rent, for which thecorresponding deflator in national accounts data (see Banca dItalia 2005)and the data collected in the Banca dItalias Survey of Household Incomeand Wealth (Banca dItalia 2004, 2006) suggest higher increases in recent

years than measured by the consumer price index. 8As to the operating procedures, in Italy, as in other countries, the index

is calculated from a sample survey of the prices of the goods and servicesthat account for the largest share of household consumption; in Italy thereare 1,043 such items. The survey is carried out centrally by ISTAT for justover a fifth of the goods and services in the basket, while for all the restit is conducted locally by the municipal statistical offices in 87 provincialcapitals and 39,000-odd survey points. The duties of these offices thereforeinclude the most delicate part of the calculation of a price index, namelythe acquisition of the basic data. In carrying out these tasks they haveconsiderable leeway, albeit within the general guidelines set by ISTAT. Wehave no direct evidence of the quality of the local price surveys, and in

particular of how the list of goods and services to be surveyed is drawn upor the sample of shops selected. We can only rely on the indirect indicationsobtained by comparing price changes recorded in the different cities. There islimited dispersion of the distribution of the month-on-month and the twelve-month inflation rates, measured by the coefficient of variation, and thisdispersion has diminished in recent years for both indicators (Del Giovaneand Sabbatini 2005). The absence of extreme and systematic differencesbetween cities suggests that no deep-rooted anomalies exist.

The public and the press have repeatedly asserted that the ISTAT index doesnot capture exceptionally large movements in the prices of some products. Thisargument is usually raised after particularly sharp price increases for very specific

7 Del Giovane and Sabbatini (2005) take a closer look at the methods used tocompute the price indices, considering specific aspects that have been the subjectof some debate the reference population, and the effects of substituting andintroducing new products in the basket and conclude that there are no featuressuggesting that they systematically generate an underestimation of inflation; ifanything, the substitution effects and the impact of introducing new productscould lead to an overestimation.

8 Concerning this particular issue, see also Sect. 1.4.1(c).


24/194


9 See Del Giovane and Sabbatini (2005) for details. The answers to a specificquestion in the survey carried out by Del Giovane, Fabiani and Sabbatini(Chap. 6) suggest that the consumers behaviour is in fact consistent with thishypothesis.

10 The weight refers to the 2003 basket. It was lower in previous surveys, as shownin Table 1.1.

goods or services, such as a particular vegetable sold in a given market, and itfails to take into consideration the average nature of price indices in general, aswell as specific methodological aspects of the treatment of some items.

The breakdown into 207 product items that ISTAT publishes everymonth on its Internet site only occasionally refers to the prices of singlegoods or services. Most are aggregations of similar goods or services,obtained by weighting the single products in proportion to their shares inconsumption. The prices of these products can move in different directionsin the short term: for example, the price of zucchini, which are included inthe item fresh vegetables, may diverge from that of other componentsof the same item. As the weight of single products tends to be small, their

impact on changes in the price of the aggregate item and the general index iscorrespondingly limited. Moreover, even for single products the publishedindex is an average of prices recorded throughout Italy. Averaging bothproducts and prices results in the smoothing of larger movements.

For the items fresh fruit and fresh vegetables, consisting of a varietyof products, the comparison between individual experience and ISTATsobservations can also be affected by the method of calculation employed tocompute the price indices; this excludes varieties whose high prices meanthey are likely to be discarded in favour of similar products a plausiblehypothesis, given the range of alternatives.9 In practice, this procedurelessens the impact of supply shocks on the prices of specific varieties, but itdoes not tally with the everyday experience of individuals, who observe the

extreme prices even if they decide not to pay them.Another much-discussed item is motor-vehicle insurance; here again, thecomparison between individual impressions and the official statistics is affectedby the methodological approach. In Italy, as in the rest of Europe, the costof insurance is included in the calculation of inflation net of claims paid andtherefore has a small weight in the average basket (just over 1 per cent).10 Therationale is that the household sector as a whole pays out premiums to insurancecompanies and receives payments for claims; but from the point of view ofthe individual consumer who has not had any accidents and not received anypayments, the incidence of this expenditure item is well above the average.

Although some individual price changes are smoothed as we describeabove, a careful examination of the ISTAT data shows that large movements


25/194


are present, especially when looking at the monthly peaks instead of theannual average. Table 1.1 shows the ten items of the 207-item breakdownthat recorded the largest twelve-month price increases in 2002 and 2003,regardless of the month in which they occurred, and the ten with the largestdecreases. In 2002 there were extremely large increases in the prices of somefood products (30 per cent for potatoes and 20 per cent for fresh vegetables)and services (between 14 and 22 per cent for air transport, insurance,bathing establishments and maritime transport). Some of the same itemsalso registered substantial price reductions in other months, confirmingtheir pronounced variability.11 In section 1.4.1 we discuss the possibilitythat increases and decreases may be perceived asymmetrically and that the

weight of individual items in the formation of perceptions may not be thesame as their actual weight in the basket.

All told, the information available does not suggest that the indicescalculated by ISTAT systematically miscalculate and underestimateinflation,12 although the methodological treatment of specific items is opento improvement. Moreover, we cannot rule out the possibility of flaws in thedata collection procedures, especially regarding the choice of retail outletsand the sample of goods and services whose prices are recorded, for whichthe municipal statistical offices are responsible.13

As to the alternative estimates with respect to ISTATs, the impressionis that they are partial, based on a much more limited basket of goods andservices and a much smaller sample of sales points than those monitored by

ISTAT. The estimates released by Eurispes, a private research institute, are acase in point owing to the attention they received in the media. In the summerof 2002, the director of Eurispes announced that inflation in Italy was runningat a rate of 8-9 per cent, more than three times the figure calculated by ISTAT.The following year, Eurispes reported that food prices had risen by 29 percent between December 2001 and December 2002, compared with ISTATs

11 The presence of exceptionally large price movements becomes even more evidentif we look at more disaggregated data. Del Giovane and Sabbatini (2005) show thatfor a 550-item breakdown of ISTAT data price changes range between +58 and-18 per cent. Moreover, these changes were not unusual with respect to previousyears. In every year between 1998 and 2001 the largest variations (both upwards

and downwards) were comparable to those in the two years under review.12 This conclusion is supported by the results of Angelini and Lippi (see Chap. 4), wholook at trends in ATM cash withdrawals. They find no substantial breaks betweenthe period preceding and that following the cash changeover; this is the oppositeof what should have been observed (with flat consumption) if, as suggested byperceptions, actual inflation had been much higher than that measured by theofficial statistics.

13 See Buzzigoli, Di Iorio and Marliani (2002) for a detailed examination of the processused to produce ISTATs price indices, particularly the gathering of basic data.


26/194


Table 1.1. Extreme price changes: 207-item breakdown

2002 2003

Product item Weight%

12-monthchange

Annualaverage

%change

Product item Weight%

12-monthchange

Annualaverage

%change% month % month

Upward Upward

Potatoes 0.28 30.5 Feb. 10.4Other lodging

services 0.31 32.5 Aug. 15.6

Air transport 0.67 22.1 Aug. 6.1

Postal bank

services 0.15 26.7 Jan. 26.7Fresh vegetables 1.22 19.6 Apr. 13.6 Potatoes 0.25 23.6 Nov. 2.6

Insuranceservices 0.31 18.7 Mar. 11.6

Inland watertransport 0.00 20.8 June 12.8

Bathingestablishments 0.17 15.9 July 6.6 Italian cigarettes 0.49 15.6 Nov. 12.1

Newspapers 0.28 14.9 July 13.0 Lubrificants 0.11 13.6 Mar. 10.6

Maritimetransport 0.14 13.8 June 8.9 Other fuels 0.28 12.6 Mar. 3.4

Fresh shellfish 0.28 12.6 Aug. 11.0 Banking services 0.64 11.9 Jan. 8.9

Fresh fruit 0.96 11.0 May 8.9 Liquid fuels 0.73 11.7 Mar. 2.9

Amusement

parks 0.09 9.2 Sept. 6.0 Fresh shellfish 0.34 11.1 Jan. 5.9Total 4.40 Total 3.29

Downward Downward

Air transport 0.67 -3.9 Dec. 6.1 Other fuels 0.28 -2.0 Oct. 3.4

Potatoes 0.28 -4.4 Nov. 10.4Audiovisual

equipment 0.44 -2.3 Nov. -1.3

Liquid fuels 0.68 -4.5 Jan. -0.2 Amusement parks 0.09 -2.7 Sept. 0.3

Electricity 1.22 -5.2 Mar. -1.5 Petrol 1.97 -2.8 Jan. 1.5

Other meats 0.18 -6.7 Apr. -2.4 Skilift 0.09 -3.5 Dec. 2.5

Gas 1.87 -7.3 Apr. 4.8 Medical Products 2.92 -6.5 Mar. -3.8

Petrol 1.97 -7.5 Jan. -2.7 Maritime transport 0.13 -8.1 Sept. -0.5

Poultry 0.65 -7.5 Mar. -2.4Telephones

and accessories 0.90 -13.6 Dec. -3.9

Other fuels 0.28 -9.6 Jan. -2.7 Potatoes 0.25 -13.7 Apr. 2.6

IT equipment 0.24 -14.7 Jan. -11.5 IT equipment 0.21 -17.0 May -14.5

Total 8.05 Total 7.28

Source: Based on ISTAT data (2003 basket).


27/194


estimate of 3.8 per cent. When examining the estimates made by Eurispes wemust bear in mind that they were largely based on an ex post reconstruction ofprices in December 2001, the reference month for calculating inflation, whichthe institute had not recorded at the time. As Eurispes itself explained (2003):

The prices in 2001 had to be derived from: printed price lists and those still storedon computers, shopping receipts of the previous year, newspaper advertisementsand advertising circulars listing products and prices, discussion groups (composedof customers and sellers) that met in front of stalls in open-air markets or withinshops (p. 2). Very often, for both shops and market stalls, the only way to obtaininformation on past prices was to involve the customers present at the time and the

seller in a discussion (p 8).14

Apart from the methodological issues raised by the construction of thesample and the indices and the definition of the set of products, Eurispessuse of respondents (households and retailers) retrospective evaluationsto derive price changes makes this estimate better suited to measuringperceived than actual inflation, since actual data recorded in 2002 arecompared with data based partly on individuals recollections.

1.4 Why the gap between official and perceived inflation?

We have seen how some aspects of the methods used to compute theofficial indices may have helped to create a gap with respect to perceptionsof inflation, particularly for specific items. However, these aspects cannotexplain either the widespread nature or the magnitude of the divergence. Inthis section we look at other possible explanations to see whether they canreconcile individual inflation perceptions with the official statistics.

Section 1.4.1 examines how some factors attributable to asymmetries in theway individuals inflation perceptions are formed and in the actual behaviourof prices might have become accentuated during or after the euro cashchangeover.15 It can be hypothesized that individuals inflation perceptions are(a) more sensitive to increases in the prices of frequently purchased products;(b) more sensitive to price increases than to price decreases or over-influencedby extreme changes; (c) formed or expressed differently depending on the

14 Translations of quotations from Italian are ours; emphasis added.15 Examples of asymmetry in subjective evaluation can be found in Kahneman and

Tversky (1979). Although their analysis relates to a different issue (decision-makingunder uncertainty) it provides some interesting results for an understanding of thephenomenon considered in this paper. See also Tversky and Kahneman (1974) foran analysis of the heuristics employed in subjective assessments.


28/194


stratum of the population. These hypotheses regarding the psychology ofinflation perceptions imply that perceptions are sharper when the followingprice developments occur: (a) larger price rises for frequently purchasedproducts than for other goods; (b) a change in the distribution of price changes,with a larger proportion of upward and downward movements and a greaternumber of extreme changes; (c) greater differences in inflation rates for theconsumption baskets of individuals or specific categories of persons. Suchinteractions may have been amplified if there was a link between individualsexpectations concerning the impact of the changeover to the euro and theirex-post perceptions.16 Research in experimental psychology (Traut-Mattauschet al. 2004) has found that the more pessimistic their expectations, the more

individuals tended to overestimate the changeovers inflationary effects. Thisrelation is attributed to a mechanism called selective output correction, thepropensity to check data confirming ones expectations less accurately thandata disproving them.

In section 1.4.2 we investigate the possibility that consumers blamedinflation for losses of purchasing power not directly attributable to it, suchas variations in income or price rises for goods not included in the officialbasket. Section 1.4.3 considers the degree to which the medias exceptionalcoverage of inflation following the cash changeover may have amplifiedindividual perceptions. Lastly, section 1.4.4 examines whether perceptionsmay have reflected consumers propensity to gauge the new prices in eurosby mentally converting them into lire at approximately 2,000 lire per euro

(as against the official locked-in rate of 1,936.27 lire) or on the basis of animprecise recollection of the old prices.

1.4.1 Asymmetries in individual perceptions and actual price

movements

(a) Frequency of purchases

After the cash changeover consumers had to memorize a large numberof prices re-denominated in euros. A learning process of this scale cantake a long time and its duration may vary from one product to another.

16 The publics concerns on the eve of the changeover and in the first days of 2002emerge from the accounts of daily newspapers (to cite one article among many,see The impact of rounding, in Il Sole 24 Ore of 28 December 2001) andfrom the survey conducted by the European Commission in November 2001,according to which 78 per cent of the Italian respondents were afraid of abusesand cheating on prices, against 70 per cent in the euro area.


29/194


It is plausible that the process will be swifter and easier for the pricesof goods and services that are purchased frequently, in some caseseven several times a day.17 We can assume, therefore, that perceptionsafter the changeover reflected movements in the prices of frequentlypurchased goods and services to a greater degree than in the past anddisproportionately with respect to their weight in the basket. However,simply paying more attention to these products is not sufficient to inducea higher perception of inflation than the official rate. For this to happen,an additional condition must be satisfied, i.e. the prices of frequentlypurchased goods and services must rise faster than those of the otherproducts in the basket.

We assess whether this condition was met in the data by classing the207 items published by ISTAT for the consumer price index (CPI) into twogroups: those purchased at least once a month (high frequency) and theremainder (low frequency). In the first group (representing around 40per cent of the ISTAT basket) we placed food products, tobacco, everydayhousehold products (detergents, soaps, etc.), newspapers, some services(local transport and railways, postal and banking services, restaurants andcoffee shops, recreational and cultural services) and fuels.18

There were marked divergences in the twelve-month changes in the twosub-indices as early as 2001 (Fig. 1.3) and this gap remained wide for thenext two years. The prices of frequently purchased products rose sharplyin the second half of 2002 and again in mid-2003, while in the other set of

products inflation was broadly stable or decreasing.According to these results part of the divergence between perceived andofficially measured inflation can be attributed to the gap between the inflationrate for frequently purchased goods and services, to which consumersplausibly attached more weight in formulating their opinions of average

17 Marques and Dehaene (2004) examine the psychological mechanismsthrough which individuals form their estimates of prices and the possiblerelation between frequency of purchase and rapidity of the learning process.

An experiment presented in the work, conducted on a sample of Austrianand Portuguese individuals, confirms the hypothesis that learning is fasterfor frequently consumed goods. On the importance of this aspect, see alsoGuiso (2003), Dziuda and Mastrobuoni (2005) and Ehrmann (2006), aswell as Brachinger (2006) and the related discussion by Hoffmann, Leiferand Lorenz (2006).

18 An alternative approach to a priori selection of the items is to identify them witha correlation analysis. This method has been used for Italy by ISAE (2002) andfor France by INSEE (2003) and Crdit Agricole (2004).


30/194


inflation, and the inflation rate for less frequently purchased items.19 However,it should be noted that the inflation rate for the high-frequency products wasless than 4 per cent, well below many consumers personal estimates ofgeneral inflation. It may be that the effect on inflation perceptions of pricerises for frequently purchased products was amplified by the consumerstendency to seek confirmation for their pessimistic expectations regardingthe inflationary effect of the changeover (Traut-Mattausch et al. 2004).

(b) Distribution of price changes

The official consumer price indices are calculated by weighting the individualitems according to their shares in total household consumption. However, theindividual consumers perceptions may be more responsive to the upwardmovement of one price than to the downward movement of another, evenif the two changes are simultaneous and of the same order of magnitude;likewise, an exceptionally large price change for a product whose weightin the index basket is modest may have a stronger effect on perceptions

19 The relationship between perceptions and price changes for specific categories ofgoods and services, including the most frequently purchased, has been analyzedby European Central Bank (2003b) for the euro area as a whole and by variousstudies for single countries of the area (see De Nederlandsche Bank 2002, Santoset al. 2002, Walschots 2002, Banco de Espaa 2003, Buiten 2003, lvarezGonzlez et al. 2004, Fluch and Stix 2005, Aalto-Setl 2006, Kurri 2006, inaddition to the two French studies mentioned in the preceding footnote).

1.0

1.5

2.0

2.5

3.0

3.5

4.0

1.0

1.5

2.0

2.5

3.0

3.5

4.0

1997 1998 1999 2000 2001 2002 2003

high frequency

low frequency

Fig. 1.3. Inflation rates for goods and services in Italy according to frequency ofpurchase (quarterly data; twelve-month percentage changes in the sub-indices)

Source: Based on ISTAT data.


31/194


than a small price change for a product whose weight is considerable.20This asymmetry of perceptions could grow in periods of heightened publicdebate and media coverage of inflation; typically, large price rises receivemore attention than small increases or reductions.

At a given inflation rate, perceptions formed in the ways described areaffected by the dispersion of the individual price changes. Suppose, forexample, that there was an increase in the number of goods and servicesrecording large upward or downward movements but that the changescancelled each other out. If households assign more weight to increases thanto decreases and to large changes than to small ones, at a given inflation ratethere would be an increase in perceived inflation.

To assess the importance of these effects, we calculated the dispersionof the inflation rates of the individual goods and services included in theCPI, measured as the standard deviation of the cross-section distribution oftheir percentage changes with respect to the previous quarter, using data ata different level of disaggregation.

Fig. 1.4 shows the dispersion of the quarter-on-quarter percentage changesof consumer prices on the basis of the 207- and 550-item breakdowns

0.0

0.5

1.0

1.5

2.0

2.5

3.0

0.0

0.5

1.0

1.5

2.0

2.5

3.0

1997 1998 1999 2000 2001 2002 2003

CPI

standard deviation - 207 itemsstandard deviation - 550 items

Fig. 1.4. Average consumer price inflation in Italy and its dispersion (percentagepoints) (1)

Source: Based on ISTAT dataNotes: (1) Standard deviation calculated for the cross-section destribution of the quarter-on-quarter percentage changes of the series with the level of disaggregation indicated.

20 The survey-based results reported in chapter 7 are consistent with the asymmetryhypotesis


32/194


21 The first breakdown corresponds to the level of disaggregation of the datapublished on ISTATs website, while the second is provided by ISTAT onrequest. Since the number of items at this level of disaggregation changes fromyear to year and it was around 550 in the period we are looking at, we refer to thisset as the 550 items.

22

Here we refer to the elementary price quotes (e.g. 250 g of a specific brand ofcoffee sold in a certain shop in a given city) recorded each month for the CPI.Although the sample is partial and not representative of the composition of thebasket (the total weight of the 48 products corresponds to around 20 per cent ofthe total basket; see Veronese et al. 2005 for a detailed description of the dataset),it does allow us to examine price behaviour at the highest level of disaggregation;this makes it particularly useful for ascertaining whether broadly stable averageinflation in Italy at the beginning of 2002 co-existed with large movements inrelative prices that may have been perceived asymmetrically by consumers.

provided by ISTAT.21 The dispersion of inflation rates for the individualitems was similar at both levels of disaggregation between 1999 and 2003,with the notable exception of the first quarter of 2002, when it was markedlyhigher for the more disaggregated series (the spike recorded a year earlierwas of comparable magnitude for the two breakdowns). This supports theview that the changeover was accompanied by rather large movements inrelative prices that do not show up in the aggregate index.

This initial finding is confirmed by an examination of the dispersion of thequarter-on-quarter changes in the elementary prices of a subset of 48 goodsand services.22 Fig. 1.5 shows that the dispersion increased rapidly in the firstquarter of 2002 and, though less fast, also in the second half of 2003, when

average inflation was broadly stable. In 2002-03 the standard deviation was4.4 percentage points, compared with 3.3 points in the previous five years;the average of the price changes was in line with past levels, except for thefirst half of 2002.

In short, our findings validate the hypothesis that relative prices changedto an unusually large degree during the cash changeover and that, giventhe asymmetrical perceptions of consumers, this may have appreciablyinfluenced their assessment of inflation.

The disaggregated data on individual prices point to other significant changesin the distribution of price movements during and after the changeover. Fig.1.6 shows that in the first quarter of 2002 a much higher percentage of priceschanged than in the same period of the preceding years (around 17 per cent,

against an average of around 9 per cent). This difference is also evident insubsequent quarters, although to a smaller degree. The figure shows, in addition,that the proportion of large price increases (i.e. more than 5 per cent quarteron quarter) in 2002-03 was similar to that of the five previous years and only


33/194


slightly more pronounced in the first quarter. By contrast, there was a greaterincrease in the share of prices that rose by less than 5 per cent or decreased.These findings are consistent with the evidence presented by Gaiotti and Lippi(see Chap. 3) for the Italian catering sector. They showed that, compared withprevious years, in 2002 there was a sharp rise in the percentage of restaurantsthat changed prices rather than a peak in the size of the changes. Prices mayhave been changed more often in 2002, in catering and other sectors, as aresult of the decision by businesses to bring forward planned price changes tocoincide with re-pricing in euros and so avoid incurring the same costs twice(in the case of restaurants, for example, the cost of re-printing menus).

The figures reported in Table 1.2 show that in 2002-03 the increase in theproportion of prices raised or lowered involved products in every category.

They also highlight a specific development in the service sector, where therewas a substantial rise in both the share of quarterly price increases between5 and 10 per cent and that of increases exceeding 10 per cent. This finding isalso consistent with the results presented in chapter 3 for Italian restaurants,according to which in 2002-03 price increases were larger in the provinceswhere competition was low, and with studies of other euro-area countriesshowing large price rises for some services, particularly those with lesscompetition (e.g. see Deutsche Bundesbank 2004).

-0.5

0.0

0.5

1.0

1.0

3.0

5.0

7.0

1997 1998 1999 2000 2001 2002 2003

standard deviation (right-hand scale)

average (left-hand scale)

Fig. 1.5. Average consumer price inflation in Italy and its dispersion: subset of 48goods and services (percentage points) (1)

Source: Based on ISTAT data

Notes: (1) Average and standard deviation calculated for the cross-sectiondistribution of the percentage changes on the previous quarter of the elementaryseries of a subset of 48 goods and services included in the CPI basket.


34/194


0.0

2.0

4.0

6.0

8.0

0.0

2.0

4.0

6.0

8.0

decreases

increases between 0 and 5%

increases exceeding 5%

1998 1999 2000 2001 2002 2003

Fig. 1.6. Sign and size of price changes in Italy: subset of 48 goods and services(percentages) (1).

Source: Based on ISTAT data.

Notes: (1) The percentages shown are calculated on changes with respect to theprevious quarter.

To summarize, the disaggregated data support the hypothesis that theprice movements accompanying the changeover diverged somewhat from

the past a higher percentage of price changes in general and of verylarge increases in the service sector and may have altered householdsinflation perceptions, but are less visible from the aggregate indicesalone. The differences with respect to the past do not appear to have beenlarge enough on their own to justify such a surge in perceived inflation,and may have been amplified by the link between pre-changeoverexpectations and post-changeover perceptions.

(c) Average basket and individual baskets

While the consumer price index refers to the average consumption ofall resident households, the perceptions of an individual or category of

individuals mainly reflect movements in the prices of the goods and servicescomposing a personal basket of consumption.Typically, a personal basket differs from the average basket according

to the individuals social and economic situation (income, age, householdstatus, education, etc.). For example, both essential and non-essentialitems have a significant weight in the basket of wealthy households, whileessentials predominate among the less well-off. Given these differences,an individuals personal inflation rate and the average rate for the whole


35/194


Table 1.2. Sign and size of quarterly price changes in Italy by product categories:subset of 48 goods and services

(percentages) (1)

1996 1997 1998 1999 2000 2001 2002 2003

Energy products

Negative 23.3 20.0 36.9 13.0 23.4 49.3 40.1 41.10 44.0 61.8 56.2 30.3 23.0 23.5 14.7 8.0Between 0 and 5% 30.1 18.2 6.7 51.0 47.5 26.2 44.0 48.8Between 5 and 10% 2.5 0.1 0.2 5.6 4.9 0.9 1.0 1.9Exceeding 10% 0.1 0.0 0.0 0.1 1.1 0.2 0.1 0.2

Total 100.0 100.0 100.0 100.0 100.0 100.0 100.0 100.0

Fresh foods


Total 100.0 100.0 100.0 100.0 100.0 100.0 100.0 100.0

Processed foods


Total 100.0 100.0 100.0 100.0 100.0 100.0 100.0 100.0

Non-food and non-energy products

Negative 1.1 1.1 1.1 1.5 1.2 1.2 2.4 1.7

0 95.4 96.2 96.4 96.0 96.0 95.5 93.0 94.4Between 0 and 5% 1.2 1.3 1.0 1.1 1.3 1.5 2.8 2.1Between 5 and 10% 1.4 0.9 1.0 0.9 0.9 1.2 1.0 1.0Exceeding 10% 0.9 0.5 0.5 0.5 0.7 0.7 0.8 0.8

Total 100.0 100.0 100.0 100.0 100.0 100.0 100.0 100.0

Services


Total 100.0 100.0 100.0 100.0 100.0 100.0 100.0 100.0

Total

Negative 4.0 3.7 4.1 3.2 3.2 4.7 5.7 5.20 89.7 91.4 91.8 91.0 90.1 89.1 85.8 86.9

Between 0 and 5% 3.0 2.3 1.8 3.6 3.9 3.3 5.5 4.8Between 5 and 10% 1.7 1.3 1.1 1.3 1.5 1.6 1.5 1.6Exceeding 10% 1.6 1.3 1.1 0.9 1.3 1.4 1.6 1.5

Total 100.0 100.0 100.0 100.0 100.0 100.0 100.0 100.0

Source: Based on ISTAT data.Notes: (1) The percentages shown are derived by calculating: (i) the quarter-on-quarter percentage changes in the elementary prices considered; (ii) for eachquarter, the percentages for the different intervals; and (iii) the annual average ofthe percentages obtained in step (ii).


36/194


population may hardly correspond at all. Moreover, the relation between theinflation an individual experiences and the inflation he or she perceives mayvary with the consumer category to which that person belongs.23

The difference between the average and the individual basket may beconsiderable in the case of items with a large weight for their purchasers butthat are purchased by a fairly small percentage of households and so have amodest impact on average inflation. A very clear example of this is rent, whichis a major expense for tenants but has no impact on the monthly expenditureof home-owners. In Italy, where only around 20 per cent of households live inrented accommodation, rent is a small item in the average basket (3.1 per cent),

making a substantial difference between the average basket and the personalbasket of anyone who pays rent. Further divergence at the individual levelmay reflect the type of lease, geographical area, or city neighbourhood.24

Another reason for the gap between average and individually perceivedinflation is that the former is measured annually, while some products areusually purchased less than once a year. The weight of the price of durablegoods, such as cars, in the CPI depends on the average frequency of purchaseby Italian consumers. However, the items importance relative to totalexpenditure will be greater than average for people who actually made sucha purchase in the last year, and will be nil for those who did not.

To gauge the impact of the discrepancy between the individual basket ofa category of consumers and the average basket, we made some estimates of

inflation by class of expenditure. These estimates are rough measures of thephenomenon as the available data allow us to capture the weight effect (i.e.the distribution by type of consumption for the segment of population in thegroup) but not the price effect (i.e. the prices paid by the households in thegroup, which depend on their shopping habits, including their preferencesregarding retail outlets and types of products).25

23 The survey-based results reported in chapter 6 indicate in fact that individual inflationperceptions vary depending on the consumption bundle they refer to and that a largemajority of the consumers are not aware of the composition of official basket.

24 Avaible evidence suggests in fact that in recent years rents have increased muchmore in big cities than in smaller towns and for new leases more than for existing

ones (Banca dItalia 2004, 2005, respectively).25 Capturing this second effect would require surveys of the different prices paid byconsumers in each group, and these are not available. In this regard, it is to benoted that ISTAT observes the prices of several varieties of the same product atlarge distributors, but only the price of the best selling variety at traditional shops(ISTAT 2005). This is because by recording only the prices of best-selling varietiesthe institute can consider the consumer goods purchased by the majority of thepopulation while minimizing the quantity of information collected every month.On this issue, see Campiglio (2004) and Del Giovane and Sabbatini (2005).


37/194


The estimates of the inflation rate for each decile of equivalent expenditure showthat the differences were small in 2002 and virtually nil in 2003 (Table 1.3).26

In 2002 average inflation was slightly higher for the households with thehighest consumption. This is because the households that consumed less were

hardest hit by the rise in food prices, which far outpaced the general index(3.6 as against 2.5 per cent), but benefited from the more favourable pricedynamics for two items with a large weight in their basket: energy products(-1.9 per cent) and goods and services subject to price controls (2.2 per cent).In addition, the 3.9 per cent rise in the prices of unregulated services hadmuch less impact for the least-well-off households (around 15 per cent of theirshopping basket) than for the most affluent ones (around 35 per cent).

26 Baldini (2004a) and ISTAT (2003, 2007) obtain similar results using analogousdata and methods.

Table 1.3. Average inflation rate by decile of equivalent expenditure(year-on-year percentage change in the consumer price index; annual

averages) (1)

Decile of equivalentexpenditure 2002 2003

First 2.1 2.5Second 2.2 2.6Third 2.2 2.6Fourth 2.3 2.6Fifth 2.3 2.6Sixth 2.4 2.6Seventh 2.4 2.5Eight 2.5 2.6Ninth 2.5 2.6Tenth 2.6 2.4

Total 2.5 2.6

Source: Based on ISTAT data.Notes: (1) Variations calculated on the basis of price indices roundedto 2 decimal places; for the annual average for 2003, this entails aslight difference with respect to the ISTAT figure (2.7 per cent), whichis calculated on the index rounded to 1 decimal place.


38/194


These results, which, we repeat, do not take account of the price effectand are therefore only approximate, contrast with the widely held belief thatthe negative effects of the price movements of 2002 and 2003 were greaterfor low and middle-income households than for well-off ones.27

1.4.2 The individuals economic situation

Perception of a larger rise in inflation than shown by the official rate anddifferences of perception between one type of household and another could bethe result of assessments based not only on experienced inflation but also onother factors affecting the individuals economic situation, such as increases

in the prices of items not included in the index basket or income developmentsaffecting various segments of the population. In other words, consumers mightconfuse the loss of purchasing power caused by inflation with a more generalimpoverishment due to other factors. We shall consider two specific factors.

The first is the behaviour of house prices, which are not included in theconsumer price index basket because the purchase of a dwelling involvesboth investment and consumption choices and it is difficult to distinguishbetween the two motives.28 If this exclusion were not clear to consumersurvey respondents (who are unlikely to be familiar with the characteristics ofthe basket and the methodology of the ISTAT index), their responses wouldrefer not so much to their inflation perceptions as to their perceptions ofpurchasing power, which they would gauge taking into account the potential

purchase of a house. The steep rise in house prices in 2002-03 by morethan 28 per cent between the second half of 2001 and the second half of 2003 might have had an impact on perceptions of price trends in general andnot just property prices. Moreover, depending on the distribution of homeownership among different segments of the population and the differencesin the rate of increase in house prices in areas inhabited by different socialclasses, increases in house prices could have different effects on differentcategories of consumers.

27 For an attempt to evaluate the possible direction of the effect of different pricechanges in different varieties of a single product see Mostacci, Natale and

Pugliese (2004).28 In most EU countries, including Italy, the consumer price index basket includesrent and, for both owner-occupiers and tenants, some housing-related expenditures,such as maintenance; however, it does not include expenditure by the home-ownerfor the purchase of the dwelling, nor other user costs, such as interest payments onmortgages. This approach was adopted by Eurostat for the harmonized consumerprice index. Eurostat, in cooperation with national statistical institutes, is currentlystudying possible treatments of owner-occupied housing in the harmonized index.For more on the subject, see European Central Bank (2005).


39/194


As far as income developments are concerned, according to the Bank ofItalys Survey of Household Income and Wealth in 2002 (Banca dItalia 2004)the changes in households annual disposable income with respect to 2000varied considerably according to the work status of the head of the household(Table 1.4). Compared with a sample-wide increase of 6.8 per cent in nominalterms and 1.1 per cent in real terms, the income of households headed by self-employed workers increased more (10.1 per cent in nominal terms and 4.4per cent in real terms) than that of households headed by payroll employees(5.7 per cent in nominal terms, nil in real terms). The nominal income ofhouseholds headed by blue and white-collar payroll workers rose by only3.9 per cent, declining by 1.8 per cent in real terms. Households headed bypensioners recorded a gain of 6.4 per cent (0.7 per cent in real terms).

Boeri and Brandolini (2004) show that differences in the trends in householdincome in recent years have affected the incidence of relative poverty the

Table 1.4. Italian household income by work status of the head of household,2000-2002

Annual net household income Annual net per capita income (2)

Work statusof the head ofhousehold (1)

EurosPercentage

change2002/2000

EurosPercentage

change2002/2000

2000 2002 nominal real (3) 2000 2002 nominal real (3)

Payrollemployee 28,651 30,293 5.7 0.0 9,005 9,696 7.7 2.0of which:

blue andwhite-collarworkers 26,477 27,505 3.9 -1.8 8,306 8,798 5.9 0.2

Self-employed 36,568 40,245 10.1 4.4 11,095 12,168 9.7 4.0

Not in work 19,761 20,981 6.2 0.5 9,738 10,460 7.4 1.7of which:

pensioners 20,596 21,911 6.4 0.7 10,341 11,081 7.2 1.5Total 26,098 27,868 6.8 1.1 9,597 10,363 8.0 2.3

Source: Banca dItalia (2004).

Notes: (1) Member with the highest income. (2) Annual net household incomedivided by the number of members. (3) Incomes were deflated using the householdconsumption deflator of the national accounts, which showed an increase inprices between 2000 and 2002 of 5.7 per cent.


40/194


share of low-income households among self-employed workers has diminishedbut it has increased among production and clerical workers although theindicators for the population as a whole have remained broadly stable.

All told, although there appear to be no appreciable differences in 2002-03 in the inflation experienced by the different classes of expenditure, thegreater inflation perceptions of some categories of households may havesprung from increased poverty due to factors other than inflation.29

1.4.3 The media

Price movements and, above all, the contrast between consumers perceptions

and the official statistics received enormous coverage in the media in the periodfollowing the cash changeover. This close coverage, especially when it gavegreater resonance to declarations and studies disputing the official estimates,may in turn have fuelled the perception of a generalized surge in prices.30

To obtain an idea of the extent of media coverage at the time, we selectedarticles published in two newspapers (Il Sole 24 Ore, Italys leading economicand financial daily, andLa Stampa, a newspaper with nationwide circulation)whose headlines included the expression inflazione or caro-vita (a popularsynonym for inflation that combines the words expensive and life) andwith the words associazioni or consumatori in the text.31

The annual data are reported in Table 1.5 and the monthly data in Fig. 1.7,which also show the figures on inflation perceptions in the same period.

ForIl Sole 24 Ore, the annual data show a substantial increase in such articlesin 2003 (94 articles, compared with 71 in 2002 and 65 in 2001) but not in thepreceding year. However, Fig. 1.7 shows that in both 2002 and 2003 therewere significant peaks in that newspapers coverage of the subject: there are

29 The results presented in chapter 6 are consistent with this hypothesis. On theseissues, see also Baldini (2004b), Golinelli and Parigi (2005) and Saraceno (2004).

30 Boeri (2004) stressed the exceptional media coverage and suggested that it mayhave contributed to the gap between perceived and official inflation. He alsonoted that many newspapers lent credence to alternative, non-official estimatesof inflation without checking their rigorousness.

31

We chose these two newspapers because they were available in the same onlinedatabase, allowing us to use the same search procedure to select the articles. Itshould be kept in mind that our articles came only from these two newspapers,one targeting a specialist readership, the other aimed at readers with above-average levels of education and income, and that we did not include radioor television. Similar results to those presented below were obtained with analternative search, in which we selected articles with headlines including at leastone of the following three expressions for inflation: caro-vita, caro-prezzi,costo della vita (see Del Giovane and Sabbatini 2005).


41/194


five months in which at least 10 articles appeared, compared with just a singleinstance in the seven previous years.32 In the case ofLa Stampa, the pronouncedincrease in coverage of the subject is very evident from the annual figures. Thenumber of articles more than doubled between 2001 and 2002 (from 21 to 48)and rose again slightly in 2003. Moreover, Fig. 1.7 shows that the monthly peaksin the number of articles were easily the highest in the ten years considered,

especially those recorded in the summer of 2002, when there was also an upsurgein inflation perceptions. Many Italian newspapers published fairly sensationalistarticles in that period, particularly those that gave ample space to criticismsof ISTATs methods and to alternative estimates of inflation. There was alsoheavy media coverage between the end of 2002 and the beginning of 2003,when inflation perceptions hit a post-changeover peak, and again in the autumnof 2003, when, after a brief decline, they rose back to the previous high.33

Table 1.5. Number of articles selected fromIl Sole 24 Ore andLa Stampa

(Selection criteria: Inflazione or caro-vita in the headline,associazioni or consumatori in the text)

Il Sole 24 Ore La Stampa Total

1995 30 15 451996 41 38 791997 32 13 451998 22 5 27

1999 33 19 522000 60 26 862001 65 21 862002 71 48 1192003 94 52 146

32 In March 2000, the attention of the press was justified by a rise in inflation to a ratethat was not particularly high in absolute terms (2.6 per cent) but that was roughlydouble the rate of a year earlier and the highest figure in more than three years.

33 Among the topics covered in these two periods much space was given to the argumentbetween Confindustria and Confcommercio (the confederations of Italian Industry andItalian Commerce, respectively) on the responsibility for the rise in prices and to the disputebetween ISTAT and Eurispes on the accuracy of official statistics. Of special interest isthe storm over ISTATs admission of a measurement error in February 2002. The errorwas very small (a reduction in the prices of pharmaceuticals was attributed to January2003 instead of to February, leading to a 0.1 percentage point underestimation of inflationin January) and would have passed unnoticed in a less highly charged environment. Onthe episode, see various contributions on the website www.lavoce.info.


42/194


(a)Il Sole 24 Ore

0

10

20

30

40

50

60

70

0

2

4

6

8

10

12

14Inflation perceptions (left-hand scale)

Number of articles (right-hand scale)

1995 200120001999199819971996 2002 2003

(b)La Stampa

0

10

20

30

40

50

60

70

0

2

4

6

8

10

12

14

Inflation perceptions (balance, left-hand scale)

Number of articles (right-hand scale)

1995 200120001999199819971996 2002 2003

Fig. 1.7. Inflation perceptions and articles selected inIl Sole 24 Ore andLa Stampa(Selection criteria: Inflazione or caro-vita in the headline, associazioni orconsumatori in the text)

Thus, in the period under review, we see that some abrupt rises inperceptions coincided with peaks in the number of articles containing key

words bearing closely on the issue of the cash changeover and inflation.Over the medium-term, a comparison of Fig. 1.6 and Fig. 1.1 shows thatthe previous phases of sharply deteriorating inflation perceptions (1995-96and 1999-2000), which were phases of rising official inflation, were indeedaccompanied by an increase in media coverage, but not on the scale of themost recent period, when official inflation remained instead broadly stable.

Our findings suggest that there may be a relationship between the trendof inflation perceptions and the attention the press pays to the phenomenon.


43/194


However, a priori, this causal link may work in both directions: a sharpdeterioration in inflation perceptions is newsworthy, and extensive mediacoverage may in turn validate and reinforce individual perceptions. Thistwo-way link is confirmed by the Granger-causality tests we performed onthe data presented in this section.

1.4.4 Rounding and recollection of lira prices

According to ISAE (2003), consumers recourse to rounding may haveinfluenced their inflation perceptions in 2002. Italian consumers supposedly

converted prices mentally into euros at 2,000 lire to the euro instead of thecorrect rate of 1,936.27 lire. This meant an upward rounding of 3.2 per cent,which added to official inflation would have implied perceived inflation ofalmost double the official rate in 2002. This argument received considerableattention in the press when it was put forward by the president of ISTATduring Senate hearings in October 2003. Extending its study to the approximateconversion rates that were probably used in the other euro-area countries, ISAEcalculated that the potential effect of rounding by consumers was positive by 2to 3 per cent in Germany and Greece and 1 to 2 per cent in Austria and Ireland,negative by around 1 per cent in Belgium, and negligible in the remainingcountries. The study also showed that in 2002 the gap between perceived andofficial inflation was generally higher in the countries where the potential

effect of rounding was greatest.34

As ISAE itself points out, this argument may help to explain the gapbetween perceived and official inflation in 2002, when consumers werecomparing prices denominated in euros (and mentally converted intolire) with lira prices in the previous year, but not the gap in 2003, whenestimating the twelve-month change in prices involved comparing twoprices expressed in euros. However, some consumers may have beenmislead because in their mind they continued to refer to the last lira pricesthey had observed before the introduction of the euro. If this were so,then two types of approximation would have been at work. The roughand ready mental conversion of euro prices into lira would have causeda first overestimation, followed by a second, equally important onecaused by making a comparison, say, at the end of 2003 on the basis of a

price from two years earlier. If these consumers did not realize that their

34 See also Ehrmann (2006) for a cross-country analysis of the relationship betweenthe complexity of the currency conversion rates and the changeover effects onmeasured and perceived inflation.


44/194


perceptions referred to a change over 24 months and not over 12, whichis the standard lag for computing inflation, this effect alone would haveproduced a perceived inflation rate roughly twice as high as the actualrate (since annual inflation in Italy was very similar in 2002 and 2003).The overestimation would be even greater if the last purchase to whichthe consumer mentally referred occurred well before December 2001 orif he had in mind not the last lira price paid for a good or service but anaverage of the prices observed over a certain span of time, such as the twoor three years preceding the cash changeover. In both cases, the consumerwould attribute to the post-changeover period a change that had actuallyaccumulated over a much longer period of time. This type of effect can

be particularly relevant in the case of the prices of rarely purchased goodsand services (such as durable consumer goods) and of those that increasedsharply over several consecutive years (restaurant meals).

Evidence on the accuracy of consumers memory for prices is provided inchapters 5 and 6 of this book. In the former, Cestari, Del Giovane and Rossi-Arnaud investigate Italian movie-goers recollection of pre-euro cinemaprices, finding that only a small percentage of respondents recalled the correctprice, and that the average prices recalled were much lower than the actualpre-euro prices and dated back to years before the changeover. In the latter,Del Giovane, Fabiani and Sabbatini obtain similar results for a different item(the newspaper) and a representative sample of italian consumers.

Concerning the difficulties consumers may have encountered in mentally

converting prices, the European Commission survey, The uro, Two YearsLater, conducted at the end of 2003, offers some interesting findings. Italywas the euro-area country with the highest percentage of respondentsexperiencing a lot of difficulty with the euro (29 per cent, against anarea average of 14 per cent) and, together with Greece, the only countryin which this percentage was higher than at the end of 2002 (when it hadstood at 14 per cent). Italy also had the highest percentage of respondentswho reckoned mentally in the old national currency for everyday purchases(46 per cent, compared with an area average of 30 per cent) and was theonly country in which they outnumbered the respondents who thought ineuros (33 per cent, compared with an area average of 46 per cent).

1.5 An overview of the correlations

Our analysis so far suggests that a multiplicity of factors probably contributedto the large discrepancy between perceptions and official measurements ofinflation. In this section we present the results of a descriptive analysis of thecorrelations between perceptions and the factors considered. We use a linearregression in which the dependent variable is the perception of inflation


45/194


(balance of the responses in the European Commission survey; see Sect.1.2) and the regressors include the factors in the preceding section for whichwe can construct a monthly indicator. The exercise is designed to measurethe correlations between perceptions and the variables considered, and notnecessarily to attribute a causal link to them, especially in the case of thevariable referring to the media (see below).

Table 1.6 contains the value of the estimated coefficients and a summarystatistic of the correlation (the adjusted R2) for each regression.35 The term ofcomparison is the regression in which the only explanatory variable is inflationmeasured by the consumer price index (CPI), whoseR2 is equal to 0.68 forthe estimation period 1997-2001 (regression 5 in the table) and drops to 0.56

when the period is extended to include the two years 2002-03 (regressions1), exhibiting a discontinuity at the time of the cash changeover.

The first exercise (regression 2 in the Table) considers variables referringto the factors we discussed in section 1.4.1, attributable to the presence ofasymmetries in the formation of perceptions and in the behaviour of prices:36

the twelve-month rate of increase in the prices of the most frequently purchasedgoods and services (CPI_high) and in the remaining ones (CPI_low);

the standard deviation calculated on the cross-section distribution of thetwelve-month percentage changes in the prices of a subset of 48 productsin the ISTAT basket (STD).37

The coefficients have the expected positive sign and are statisticallysignificant for CPI_high and (marginally) for STD.38 The explanatory power

is significantly greater than when CPIalone is considered.

35 The estimation period we consider begins in 1997 in view of the availability ofthe indicator for the dispersion of price changes. Excluding this regressor, we alsomade the estimations starting from 1992, obtaining similar results. Coefficientsfor lagged variables were reported when statistically significant.

36 Indicators of the different composition of the shopping baskets by category ofconsumer were not included in the regression, as our analysis of inflation by classof expenditure did not find significant differences.

37 The variable that captures the importance of extreme price increases did not provesignificant, consistently with the evidence of Sect. 1.4.1(b) that at the time of the

changeover their share in total price changes only increased in some sectors andmarginally for the index as a whole.38 The coefficient ofSTD is probably of scant statistical significance because some

of the phenomena captured by the variable are similar to those captured withthe breakdown of inflation by frequency of purchase (the available evidencesuggests that the dispersion was higher for frequently purchased goods). Thevariable has greater statistical significance in the regression (not reported in thetable) in which it is considered separately, together with the variable CPI. Onlythe two-period lagged value ofSTD proved to be significant in regression 2.


46/194


Table 1.6. Summary of the correlations- dependent variable: balance of inflation perceptions (1)

Regressors

Estimation period: 1997.4 2003.12

1 CPIt

0.56

Coefficients(2):

CPI36.27 (10.06)

Asymmetries

2 CPI_hight, CPI_low

t, STD

t-20.68

CPI_high 19.26 (10.95)

CPI_low 2.30 (0.80)STD

t-23.90 (1.86)

Media

3 CPIt, NEWSP

t-1, NEWSP

t-2, NEWSP

t-3, NEWSP

t-40.77

CPIt

14.65 (4.00)

NEWSPt-1

1.34 (4.83)

NEWSPt-2

0.78 (2.59)

NEWSPt-3

0.68 (2.21)

NEWSPt-4

0.73 (2.51)

Asymmetries and media


t, STD

t-2,NEWSP

t-1, NEWSP

t-2, NEWSP

t-3, NEWSP

t-40.80

CPI_high 9.12 (4.64)

CPI_low 1.68 (0.57)STD

t-23.75 (2.26)

NEWSPt-1

1.14 (4.29)

NEWSPt-2

0.71 (2.53)

NEWSPt-3

0.50 (1.73)

NEWSPt-4

0.61 (2.27)

For comparison: estimation period 1997,4 2001,12

5 CPIt

0.68


t, STD

t-2,NEWSP

t-1, NEWSP

t-2, NEWSP

t-3, NEWSP

t-40.79

Legend: R2 adjusted R-squared, CPI 12-month changes in the general index ofconsumer prices; CPI_high 12-month changes in the price index for frequently

purchased goods and services; CPI_low 12-month changes in the price index forless frequently purchased goods and services; STD standard deviation calculatedon the cross-section distribution of the 12-month percentage changes in the pricesof a subset of 48 goods in the Istat basket;NEWSP number of articles published in

Il Sole 24 Ore orLa Stampa with the key words inflazione or caro-vita in theheadline and associazioni or consumatori in the text.

Notes: (1) Regression of monthly inflation perceptions (balance of responses to the ISAEsconsumer survey; see section 1.2) on the variables indicated. (2) T-test in brackets.


47/194


In the second exercise (regression 3) we include, along with CPI, thevariable that measures the media coverage of the debate on inflation (seeSect. 1.4.3), given by the number of articles published in Il Sole 24 Oreor La Stampa containing the key words inflazione or caro-vita in theheadline and associazioni or consumatori in the text (NEWSP).39

Again, the coefficient has the expected sign and is significant; R2 rises to0.77. Given the two-way causal link between perceptions and the variable inquestion (see Sect. 1.4.3), this increase should be read as an indication of thepossible importance of the reciprocal link between the collective formationof perceptions and the medias treatment of the subject.

Lastly, when all the preceding factors are considered together (regression

4), the correlation rises to 0.80, coinciding with the figure we obtain withthe same variables when we limit the estimation to the period precedingthe changeover (regression 6). When the alternative regressors are included,the break found for the regression with only officially measured inflationdisappears.

The same evidence is offered by Fig. 1.8, which shows the behaviourof perceptions and of the fitted values obtained, respectively, with theregression of perceptions on officially measured inflation alone and with theregression including all the alternative regressors.

39 We excluded the contemporary value of the variable NEWSP, even though itproved significant when included, because some of the articles surveyed eachmonth could have appeared after that months consumer survey.

Fig. 1.8. Inflation perceptions in Italy: balance resulting from the consumer surveyand estimates

Source: European Commission and authors estimates.

-10

0

10

20

30

40

50

60

70

80

-10

0

10

20

30

40

50

60

70

80Inf

Versione Inglese Pdg Sabbatini - Volume Completo_03!10!2007

Documents

Transcript of Versione Inglese Pdg Sabbatini - Volume Completo_03!10!2007