Consumers' Attitudes and Their Inflation Expectations

Finance and Economics Discussion Series Divisions of Research & Statistics and Monetary Affairs Federal Reserve Board, Washington, D.C. Consumers’ Attitudes and Their Inflation Expectations Michael Ehrmann, Damjan Pfajfar, and Emiliano Santoro 2015-015 Please cite this paper as: Ehrmann, Michael , Damjan Pfajfar, and Emiliano Santoro (2015). “Consumers’ Attitudes and Their Inflation Expectations,” Finance and Economics Discussion Series 2015-015. Washington: Board of Governors of the Federal Reserve System, http://dx.doi.org/10.17016/FEDS.2015.015. NOTE: Staff working papers in the Finance and Economics Discussion Series (FEDS) are preliminary materials circulated to stimulate discussion and critical comment. The analysis and conclusions set forth are those of the authors and do not indicate concurrence by other members of the research staff or the Board of Governors. References in publications to the Finance and Economics Discussion Series (other than acknowledgement) should be cleared with the author(s) to protect the tentative character of these papers.

Consumers(cid:146)Attitudes and Their In(cid:135)ation Expectations (cid:3) Michael Ehrmann Damjan Pfajfar y z Bank of Canada Federal Reserve Board Emiliano Santoro x University of Copenhagen March 10, 2015 Abstract Thispaperstudiesconsumers(cid:146)in(cid:135)ationexpectationsusingmicro-leveldatafrom the Surveys of Consumers conducted by University of Michigan. It shows that beyond the well-established socio-economic factors such as income, age or gender, other characteristics such as the households(cid:146)(cid:133)nancial situation and their purchasing attitudes are important determinants of their forecast accuracy. Respondents withcurrentorexpected(cid:133)nancialdi¢ culties,pessimisticattitudesaboutmajorpurchases, or expectations that income will go down in the future have a stronger upward bias in their expectations than other households. However, their bias shrinks by more than that of the average household in response to increasing media reporting about in(cid:135)ation. Equivalent results are found during recessions. JEL classi(cid:133)cation: C53, D84, E31 Keywords: In(cid:135)ation Expectations; News on In(cid:135)ation; Consumer Attitudes. WewouldliketothankJohnRobertsandseminarparticipantsatDIWBerlin,NewcastleUniversity, (cid:3) Lund University, the Bank of Canada, the 2013 CEF meetings in Vancouver and the Workshop on Price Dynamics, In(cid:135)ation and Monetary Policy. The paper presents the authors(cid:146)personal opinions and does not necessarily re(cid:135)ect the views of the Bank of Canada, the Board of Governors of the Federal Reserve System or of any other person associated with the Federal Reserve System. Address: Bank of Canada, 234 Laurier Avenue West, Ottawa ON K1A0G9, Canada. E-mail: y mehrmann@bankofcanada.ca. Address: Board of Governors of the Federal Reserve System, 20th and Constituz tion Ave, NW Washington, DC 20551, U.S.A. E-mail: damjan.pfajfar@frb.gov. Web: https://sites.google.com/site/dpfajfar/. Address: Department of Economics, University of Copenhagen, (cid:216)ster Farimagsgade 5, x Building 26, 1353 Copenhagen, Denmark. E-mail: emiliano.santoro@econ.ku.dk. Web: http://www.econ.ku.dk/esantoro/. 1

1 Introduction How do consumers form in(cid:135)ation expectations and what determines their forecast accuracy? These questions are of critical importance for central banks and macroeconomists, since in(cid:135)ation expectations are known to a⁄ect the actual evolution of in(cid:135)ation and macroeconomy more generally. Recognizing this importance, central banks have in recent decades devoted considerable e⁄ort to anchoring in(cid:135)ation expectations, for instance, by announcing in(cid:135)ation targets. Consumer in(cid:135)ation expectations have also been central in explaining the evolution of in(cid:135)ation in the aftermath of the (cid:133)nancial crisis, (cid:133)rst during the period of the (cid:147)missing disin(cid:135)ation(cid:148)(during which in(cid:135)ation was higher than would have been expected based on models with standard determinants like the magnitude of the output gap and in(cid:135)ation expectations of professional forecasters) and subsequently, when in(cid:135)ation was weaker than expected (Coibion and Gorodnichenko, 2015; Friedrich, 2014). However, while a substantial body of empirical research has extensively studied professional forecasters(cid:146)in(cid:135)ation expectations (among many others, see Capistran and Timmermann, 2009; Coibion and Gorodnichenko, 2010), much less is known about expectations by the households. Consumer expectations are known to be biased and ine¢ cient, with forecast errors being systematically correlated with demographic characteristics (Souleles, 2004). They are also a⁄ected by frequently purchased items, such as gasoline, as pointed out recently by Coibion and Gorodnichenko (2015), and they are responsive to media reporting (Carroll, 2003). In addition to these factors, the current paper tests whether consumer attitudes also shape in(cid:135)ation expectations. We (cid:133)nd that consumers who are pessimistic abouttheireconomicor(cid:133)nancialsituationarelikelytohavehigherin(cid:135)ationexpectations. When consumers struggle to make ends meet with their available budget, it may be due to a reduction in their income or to an increase in their expenditures (cid:150)which in turn couldbe due toseveral factors, one of thembeingrisingprices fortheirconsumptionbundle. Under uncertain information and information-processing constraints, it might well be that such consumers estimate in(cid:135)ation to be higher than others. In addition, it has been shown that (cid:133)nancially constrained consumers are more attentive to price changes of the goods they purchase than more a› uent consumers (Snir and Levy, 2011). Combining this with the well-known notion that agents are more receptive to bad than to good news (see, e.g., Baumeister, Bratslavsky, Finkenauer, and Vohs, 2001) might well imply that (cid:133)nancially constrained consumers arrive at a higher estimate of in(cid:135)ation. The paper uses more than 175,000 observations from the Surveys of Consumers conductedbyUniversityofMichiganovertheyears1980to2011totestthesehypotheses. We (cid:133)nd that consumers with pessimistic attitudes about major purchases (such as purchases of durables, houses or vehicles), who (cid:133)nd themselves in di¢ cult (cid:133)nancial situations, or who expect income to go down in the future do indeed have a stronger upward bias in 2

their in(cid:135)ation expectations. Also NBER recessions (another proxy for consumer pessimism and their (cid:133)nancial di¢ culties) are associated with an incremental bias in in(cid:135)ation expectations. Wealsocon(cid:133)rmtheearlier(cid:133)ndingsthatconsumersareresponsivetonews. Weemploy two news measures, the (cid:133)rst based on the survey itself (where respondents can report whether they have recently heard news about prices), and the second, following Carroll (2003), based on intensity of news coverage related to in(cid:135)ation in the New York Times and the Washington Post. While both of these measures have been used previously, e.g., in Pfajfar and Santoro (2013), how they di⁄er, and how each of them would have to be interpreted, have not been discussed. In this paper, we clarify that reporting having heard news about prices is very tightly linked to gasoline price in(cid:135)ation in the United States. This relationship is in line with earlier evidence that frequently purchased items (such as gasoline) shape the in(cid:135)ation perceptions of consumers, and also likely re(cid:135)ects the fact that gasoline prices are extremely salient due to their prominent postings at gas stations. Interestingly, our two news measures have very di⁄erent implications for consumer in(cid:135)ation expectations. Having heard news about prices (re(cid:135)ecting predominantly large increases in gasoline prices) increases the bias. In contrast, more intense media coverage tends to reduce the bias and improve forecast accuracy. In that regard, consumers with more strongly upward-biased expectations are more responsive to media coverage, and see their bias shrinking by more than the other consumer groups. These (cid:133)ndings have interesting implications for policy-makers and the media, suggesting that more reporting about in(cid:135)ation improves consumers(cid:146)in(cid:135)ation expectations, and particularly so for consumers who are in the right tail of the distribution, i.e., have a particularly strong upward bias. The paper connects to the previous literature on the determinants of consumer in(cid:135)ation expectations. In that regard, a number of factors have been identi(cid:133)ed that shape the level of in(cid:135)ation expectations. Several socio-economic characteristics are known to a⁄ect in(cid:135)ation expectations (cid:150)females tend to have higher in(cid:135)ation expectations than men, and in(cid:135)ation expectations tend to decrease with income, whereas they are often found to be lowerforolderconsumers(Jonung,1981; BryanandVenkatu,2001; LombardelliandSaleheen, 2003; Christensen, Els, and Rooij, 2006; Anderson, 2008). These socio-economic determinants of in(cid:135)ation expectations are rather stable over time, which makes it hard to explain why household in(cid:135)ation expectations, their accuracy and the magnitude of their bias are subject to substantial time variation (Coibion and Gorodnichenko, 2015). The current paper suggests a time-varying characteristic, consumer attitudes, that can help addressing this. A small number of related studies have provided some evidence in that direction. Webley and Spears (1986) show that U.K. consumers who think they have done less well (cid:133)nancially than during the previous year, as well as consumers who expect 3

to be worse o⁄in the subsequent year, have higher in(cid:135)ation expectations. Similarly, del Giovane, Fabiani, and Sabbatini (2009) and Malgarini (2009) (cid:133)nd that in(cid:135)ation expectations of Italian consumers are higher for respondents with pessimistic attitudes, and for consumers in (cid:133)nancial di¢ culties. In(cid:135)ation expectations are also determined by the in(cid:135)ation that consumers actually experience (cid:150)(cid:133)rst, in(cid:135)ation expectations are shaped much more by the in(cid:135)ation rate of consumption baskets that relate to the respective socio-economic group to which the individual belongs than by the overall in(cid:135)ation indices, at least for low-education and low-income consumers (Pfajfar and Santoro, 2009; Menz and Poppitz, 2013); second, in(cid:135)ation expectations vary positively with the in(cid:135)ation experience that individuals have undergone over their lifetime (Lombardelli and Saleheen, 2003; Malmendier and Nagel, 2013); third, more frequently purchased items have been found to have a higher impact on in(cid:135)ation perceptions and in(cid:135)ation expectations (Ranyard, Missier, Bonini, Duxbury, and Summers, 2008; Georganas, Healy, and Li, 2014). The evolution of consumers(cid:146)in(cid:135)ation expectations has also been studied. In his seminal paper, Carroll (2003) has demonstrated that consumers update their expectations only infrequently (roughly once every year), that they respond to media reporting and update toward the expectations of professional forecasters, and that inattention to news generatesstickinessinaggregatein(cid:135)ationexpectations. Subsequently, anumberofcontributions have studied the role of media reporting for in(cid:135)ation expectations in more detail. Lamla and Maag (2012) analyze the e⁄ect of media reporting on disagreement among forecasters, and (cid:133)nd professional forecaster disagreement to be una⁄ected by media coverage, whereas disagreement among households increases with higher and more diverse media coverage. Pfajfar and Santoro (2009) provide evidence that the e⁄ect of news on in(cid:135)ation expectations di⁄ers across socio-economic groups, and Easaw, Golinelli, and Malgarini (2013) demonstrate that the rate at which professional forecasts are embodied in households(cid:146)expectations depends on socio-economic characteristics. Finally, Pfajfar and Santoro (2013) highlight the importance of di⁄erentiating between media reporting on in(cid:135)ation and whether a consumer has actually heard news about prices. Their study replicatesCarroll(cid:146)s(cid:133)ndingthatin(cid:135)ationexpectationsgetupdatedtoward theprofessional forecasts using aggregate data. However, this is not the case at the individual consumer level, where most consumers who update actually revise their expectations away from the professional benchmark, but by su¢ ciently small amounts that they are dominated in the aggregate data by relatively few consumers who update toward professional forecasts by large amounts. Di⁄erences in the magnitude of revisions that take place in response to newshavebeenidenti(cid:133)edbyArmantier, Nelson, Topa, vanderKlaauw, andZafar(2012), who (cid:133)nd larger revisions for agents that start o⁄with relatively less precise expectations. These (cid:133)ndings are in line with the current paper, which suggests that media reporting about in(cid:135)ation improves in(cid:135)ation expectations particularly for consumers who are in the 4

right tail of the distribution, i.e., have a particularly strong upward bias. The remainder of the paper is structured as follows. In Section 2, we describe the data used in our empirical analysis and provide some stylized facts. Section 3 provides an overview of the econometric approach that we employ, while Section 4 reports the relevant results. Section 5 concludes. 2 The Data and Some Descriptive Analysis Our microdata contain information on a wide range of factors that in(cid:135)uence consumers(cid:146) in(cid:135)ation expectations. As such, they allow us to explore households(cid:146)forecast accuracy in great detail. In this section we describe the key features of the data set and report some preliminary evidence on consumers(cid:146)in(cid:135)ation expectations, as well as on the newspaper index proposed by Carroll and a direct measure of consumers(cid:146)receptiveness toward news on prices. Moreover, we report some descriptive statistics about consumer-level characteristics that are accounted for as determinants of the process of expectations formation. Figure 1: CPI In(cid:135)ation, MS and SPF mean forecasts. 2 1 9 6 % 3 0 3 1980 1985 1990 1995 2000 2005 2010 Year NBER Recessions CPI Inflation (+1 year) SPF Mean Forecast MS Mean Forecast Notes: The chart reports the University of Michigan, Surveys of Consumers (MS) and the Survey of Professional Forecasters conducted by the Federal Reserve Bank of Philadelphia (SPF) mean forecasts for in(cid:135)ation at t + 12, as well as in(cid:135)ation as realized at t + 12. Based on monthly data. Source: University of Michigan, Surveys of Consumers and Survey of Professional Forecasters, Federal Reserve Bank of Philadelphia. 5

2.1 In(cid:135)ation Expectations The Survey of Consumer Attitudes and Behavior is a representative survey conducted monthly by the Survey Research Center at the University of Michigan (Curtin, 2013). ParticipantsintheSurveysofConsumers(henceforth, MS)areaskedtwoquestionsabout expected changes in prices: (cid:133)rst, whether they expect prices to go up, down or stay the same in the next 12 months; second, to provide a quantitative statement about the expected change.1 The analysis will focus on the 1980M1-2011M12 period. Figure 1 reports the mean forecasts obtained in the MS against CPI in(cid:135)ation.2 To provide another benchmark, the (cid:133)gure also includes forecasts from the Survey of Professional Forecasters (SPF), a survey amongleadingprivateforecasting(cid:133)rmsthatiscurrentlyconductedbytheFederalReserve Bank of Philadelphia.3 Both the MS and the SPF appear to predict in(cid:135)ation reasonably well, although they often fail to match periods of low in(cid:135)ation. For instance, at the very end of the sample, from 2009-11, they are considerably higher than actual in(cid:135)ation turned out to be. This episode has been studied by Coibion and Gorodnichenko (2015), who suggest that, due to high oil price in(cid:135)ation, household in(cid:135)ation expectations were elevated, which in turn helps explain the "missing disin(cid:135)ation" in the United States (i.e., the fact that standardPhillips curves wouldhave predictedadisin(cid:135)ationoverthat period that did not materialize). 2.2 News on In(cid:135)ation A direct implication of Carroll(cid:146)s view is that more media reporting should imply that people are better informed and produce better forecasts. To account for this possibility, we require reliable indicators of the (cid:135)ow of news on in(cid:135)ation that the public is confronted with. Carroll computes a yearly index of the intensity of news coverage in the New York Times and the Washington Post. In this paper, we use the monthly version of this index thathasbeenconstructedinPfajfarandSantoro(2013). Itisbasedonasearchof eachof the two newspapers for in(cid:135)ation-related articles, converted into an index by dividing the 1If a respondent expects prices to stay the same, the interviewer must make sure that the respondent does not actually expect that prices will change at the same rate at which they have changed over the past12months. Inlinewithcommonpractice,wediscardobservationsiftherespondentexpectsin(cid:135)ation to be less than -5% or more than +30%. This rule only a⁄ects 0.7% of the observations in the sample under scrutiny. Curtin (1996) also adopts alternative truncation intervals, such as [-10%,50%], showing that the key statistical properties of the resulting sample are close to invariant across di⁄erent cut-o⁄ rules. 2In(cid:135)ation expectations sampled at time t are graphed with in(cid:135)ation 12 months later, so as to be in line with the forecast target. 3The SPF is a quarterly survey. In order to obtain a monthly estimate of the SPF we may consider twooptions: eitherforecasterskeeptheirforecastuntilthenextsurveyround,ortheir"monthly"forecast includesapartialadjustmenttothenextquarterforecast. Wetookbothapproachesandobtainednearly identical results. This paper is based on a linear interpolation of the data. 6

number of in(cid:135)ation-related articles by the total number of articles.4 To be more precise, we de(cid:133)ne this news measure as NEWSN = 100nt NEWS N , where n denotes the t Nt (cid:0) t number of in(cid:135)ation-related articles in a given month t, N the total number of articles, t N and NEWS the sample average of the news measure. We demean the news measure to allow for an easier interpretation of interaction terms in the regression analysis. In addition, our analysis will rely on a measure of consumers(cid:146)perceptions of new information about prices. This is intended to complement the newspaper index proposed by Carroll. In fact, the accuracy of a proxy based on the intensity of news coverage in national newspapers can be questioned on di⁄erent grounds. For instance, Blinder and Krueger(2004)suggestthatconsumersprimarilyrelyoninformationaboutin(cid:135)ationfrom television, followed by local and national newspapers. It is also plausible to expect that the volume of news about in(cid:135)ation does not necessarily match the (cid:135)ow of information that is assimilated by the public. In this respect, a non-trivial discrepancy could result from the interplay of two mutually reinforcing e⁄ects: (i) news from the media does not necessarily reach the public uniformly and (ii) the connection between news and in(cid:135)ation expectations is likely to be a⁄ected by consumers(cid:146)receptiveness to the news and the capacity to process new information. Indeed, Sims (2003) emphasizes the presence of information-processing constraints that could be compatible with such ine¢ ciencies. Finally, it is well known that consumer in(cid:135)ation perceptions are shaped (cid:150)in line with the availability heuristic (Tversky and Kahneman, 1974) (cid:150)by frequently purchased items (Ranyard, Missier, Bonini, Duxbury, and Summers, 2008), such that in periods where in(cid:135)ation of such items is high, consumers might be more aware and concerned about in(cid:135)ation, whereas media reporting (which most likely is generally concerned with overall in(cid:135)ation) need not be more intense. In light of these considerations, it is advisable to complement the analysis with a variable that accounts for consumers(cid:146)actual perceptions of in(cid:135)ation. Such a variable is directly available from the MS, where respondents are asked whether they have heard of any changes in business conditions during the previous few months. In case of an a¢ rmative response, the respondents have the possibility to give two types of news that they have heard about, among them being either higher or lower prices. Our second news variable, NEWSP, is therefore de(cid:133)ned as a dummy variable that takes the value of one i 4A potential problem connected with this type of search is that the resulting index may include articles that do not primarily cover U.S. in(cid:135)ation. Accordingly, Pfajfar and Santoro (2013) tested the robustness of this methodology by restricting the search to articles that just cover U.S. in(cid:135)ation, and found the results to be robust. 7

if the respondent cites prices as a factor that has come to their attention.5 Figure 2: Perceived news and media reporting. 0 4 3 0 3 2 % 0 2 1 0 1 0 0 1 1980 1990 2000 2010 Year NBER Recessions CPI Inflation (left axis) Heard chang. prices (left axis) News Stories (right axis) Notes: The chart reports CPI in(cid:135)ation as recorded for a given time period t, as well as the share of respondentsintheMSinperiodtansweringthattheyhaveheardnewsaboutprices("perceivednews") and the index about media reporting related to in(cid:135)ation in period t ("news stories"). Based on monthly data. Source: University of Michigan, Surveys of Consumers. Figure 2 reports the fraction of MS respondents who have heard news about prices, together with the newspaper index and CPI in(cid:135)ation. The two series display poor correlation, suggesting that they contain two distinct measures of news. The fraction of MS respondents who have heard news about prices exhibit more volatility than the newspaper index. Especially in the latter part of the sample it displays sizable (cid:135)uctuations that neither actual in(cid:135)ation nor the newspaper index presents. Splitting the series into the share of respondents who have heard news about decreasing and increasing prices, respectively, it is evident that most of the volatility in the overall series arises due to 5The MS respondents primarily report about news on unemployment, followed by news on the government (elections) and then prices. It is important to stress that 41% of the respondents report having heard no news at all and that in 28% of the cases only one type of news is reported. This is to say that, on average, only 31% of the respondents are confronted with a potentially binding limit of two options. Therefore, though some underreporting may a⁄ect our measure of perceived news about prices, this is not likely to be primarily induced by the speci(cid:133)c design of the questionnaire. 8

movements in the share of consumers who have heard about rising prices (see Figure 3). Figure 3: Perceived news about increasing / decreasing prices. 5 3 0 3 5 2 0 2 % 5 1 0 1 5 0 1980 1985 1990 1995 2000 2005 2010 Year NBER Recessions CPI Inflation Heard: decreasing prices Heard: increasing prices Notes: The chart reports CPI in(cid:135)ation as recorded for a given time period t, as well as the share of respondents in the MS in period t answering that they have heard about prices increasing / decreasing. Based on monthly data. Source: University of Michigan, Surveys of Consumers. So what is behind this measure of news? As shown in Figure 4, the correlation between the share of respondents reporting that they have heard about price increases and in(cid:135)ation of retail gasoline prices is very high (0.63).6 Based on this evidence, we interpret the survey-based news measure as capturing in(cid:135)ation perceptions originating from frequently-purchased items such as gasoline prices. In contrast, the correlation between negativein(cid:135)ationratesingasolinepricesandtheshareofrespondentsreportingthatthey have heard about decreases is much smaller (0.23), which is in line with the prospect theory pioneered by Kahneman and Tversky (1979), since agents tend to manifest higher receptiveness toward "bad" news on prices, as compared with "good" news. 6For Figure 4, we set any negative gasoline in(cid:135)ation numbers to zero, to re(cid:135)ect the fact that the survey news measure only re(cid:135)ects having heard about price increases. 9

Figure 4: Gasoline in(cid:135)ation and perceived news about increasing prices. 0 0 4 8 0 0 3 6 %0 0% 2 4 0 0 1 2 0 0 1980 1990 2000 2010 Year NBER Recessions Heard: incr. prices (left axis) Pos. gas. infl. (right axis) Notes: The chart reports the share of respondents in the MS in period t answering that they have heard about prices increasing, as well as retail gasoline price in(cid:135)ation truncated at zero for negative values (labelled in the (cid:133)gure Pos. gas. in(cid:135).). Source: University of Michigan, Surveys of Consumers. 2.3 Consumer-level Attributes The core of our econometric analysis focuses on the connection between consumers(cid:146)in- (cid:135)ation expectations and a number of consumer-level attributes. These can be grouped in the following categories: the current and expected (cid:133)nancial situation, consumer attitudes toward major purchases, and the classi(cid:133)cations used in the previous literature, namely gender, income and age of the respondent. The attributes are constructed using the survey responses as follows: Financial situation Financial situation worse: Individuals responding "worse" to the following ques- (cid:15) tion: Would you say that you are better o⁄or worse o⁄(cid:133)nancially than you were a year ago? From this category, we exclude all individuals who name high(er) prices as one reason for being worse o⁄, in order to avoid a possible endogeneity bias. Financial expectations worse: Individuals responding "will be worse o⁄" to the (cid:15) following question: Now looking ahead - do you think that a year from now you will be better o⁄(cid:133)nancially, or worse o⁄, or just about the same as now? 10

Nominal income expectations worse: Individuals responding "lower" to the follow- (cid:15) ing question: During the next 12 months, do you expect your income to be higher or lower than during the past year? Purchasing attitudes Time for durable purchases bad: Individuals responding "bad" to the following (cid:15) question: Generally speaking, do you think now is a good or a bad time for people to buy major household items? Again, to avoid possible endogeneity, we exclude all respondents who respond "Prices are too high, prices going up" to the following question: Why do you say so? (Are there any other reasons?) Time for house purchases bad: Individuals responding "bad" to the following ques- (cid:15) tion: Generally speaking, do you think now is a good time or a bad time to buy a house? Once more, we exclude those who are pessimistic due to high(er) prices. Time for vehicle purchases bad: Individuals responding "bad" to the following (cid:15) question: Speaking now of the automobile market (cid:150)do you think the next 12 months or so will be a good time or a bad time to buy a vehicle, such as a car, pickup, van, or sport utility vehicle? Also here, we exclude individuals who give high or rising prices as a reason for their answer. Other characteristics, following the previous literature Income bottom 20%: Individuals in the bottom 20% of the income distribution (as (cid:15) identi(cid:133)ed by the MS). Elderly: Respondents who are at least 65 years old. (cid:15) Female: Female respondents. (cid:15) For each of these categories, we construct a dummy variable that is equal to one in case the attribute applies, and equals zero otherwise. For the (cid:133)nancial situation and the purchasing attitudes categories, the dummy variable is equal to one whenever the respondent is "pessimistic", i.e., the consumer describes the current situation as worse, expects a worsening, or perceives the environment as unfavorable for major purchases. For the other characteristics that had been used in the earlier literature, we expect a larger bias for low-income consumers and females, but possibly a smaller one for the elderly. 11

Figure 5: Share of pessimistic consumers (cid:150)Purchasing attitudes. 0 7 0 6 0 5 %0 4 0 3 0 2 0 1 0 1980 1985 1990 1995 2000 2005 2010 Year NBER Recessions Durables Vehicles Houses Notes: The chart reports the share of respondents in the MS in period t answering that the time for purchasing durables / vehicles / houses is bad. Source: University of Michigan, Surveys of Consumers. Figure 5 gives an impression of the time variation in consumer characteristics, for the example of purchasing attitudes. It reports the share of pessimistic consumers, and demonstrates that this share varies substantially over time. It is apparent that at the end of the sample, with the U.S. economy going through the (cid:133)nancial crisis and a major recession, many more consumers felt that times were not good for major purchases. Table 1 provides a number of summary statistics for each consumer group. The (cid:133)rst column reports the number of observations (OBS) for the full sample (which contains 175,147 observations) and separately for each consumer category. The table also provides tests for whether the news reception and the in(cid:135)ation expectations of the various respondentgroupsaresigni(cid:133)cantlydi⁄erentfromthoseoftheirpeers. Thesestatisticsare reported for the percentage of consumers who have heard news about prices (NEWSP), i the average di⁄erence between the MS consumer-speci(cid:133)c forecast and the SPF mean in- (cid:135)ation forecast (at the time of the survey, BIASF) and the average di⁄erence between the MS consumer-speci(cid:133)c forecast and CPI in(cid:135)ation (at the forecast horizon, BIAS(cid:25)). The bias statistics con(cid:133)rm that consumer in(cid:135)ation expectations are on average upward biased. Relative to actual in(cid:135)ation, the bias for the overall sample amounts to 0.8 percentage points; relative to professional forecasters, consumers overestimate in(cid:135)ation by around half a percentage point. In addition, the magnitude of this bias di⁄ers across consumer groups. With the exception of the elderly, di⁄erences in the bias are statistically signi(cid:133)cantly di⁄erent, and often by large amounts. The biggest di⁄erence is found for consumers who expect their (cid:133)nancial situation to worsen, with an upward bias that 12

.scitsitats evitpircseD :1 elbaT FSAIB p SAIB PSWEN )%(SBO SBO 35.0 87.0 24.4 %001 741,571 elpmas llarevO esrow noitautis laicnaniF ***58.0 ***50.1 ***18.3 %81 673,23 esrow noitautis laicnaniF ***64.1 ***87.1 ***92.7 %11 479,91 esrow snoitatcepxe laicnaniF ***01.1 ***43.1 ***20.5 %31 459,22 esrow snoitatcepxe emocni lanimoN rof emit dab :sedutitta gnisahcruP ***59.0 ***03.1 ***62.5 %61 192,72 sesahcrup elbaruD ***57.0 ***12.1 ***15.5 %02 374,43 sesahcrup esuoH ***01.1 ***14.1 ***25.6 %61 860,82 sesahcrup elciheV srehtO ***22.1 ***84.1 ***54.3 %41 720,42 %02 mottob emocnI *75.0 47.0 ***21.4 %61 633,82 )+56 egA( ylredlE ***88.0 ***31.1 ***81.4 %35 653,39 elameF :)%(SBO ;snoitavresbo derosnecnu fo rebmun :SBO .)swor( setubirtta suoirav no lanoitidnoc )snmuloc( scitsitats evitpircsed sniatnoc elbat ehT :setoN neewteb ecnere⁄id egareva :(cid:25)SAIB ;swen gnivresbo sremusnoc fo tnecrep egareva :PSWEN ;elpmas llarevo eht ni snoitavresbo derosnecnu fo tnecrep .stsacerof noita(cid:135)ni naem FPS eht dna stsacerof noita(cid:135)ni(cid:146)sremusnoc neewteb ecnere⁄id egareva :FSAIB ;noita(cid:135)ni IPC dna stsacerof noita(cid:135)ni(cid:146)sremusnoc eht fo tser eht morf detupmoc trapretnuoc sti naht rewol yltcirts si yrtne hcae taht tset eht fo level %01/5/1 eht ta ecnac(cid:133)ingis lacitsitats setoned = = (cid:3) (cid:3)(cid:3) (cid:3)(cid:3)(cid:3) .1102(cid:150)0891 :doirep emiT .)secnairav lauqe htiw( stset-t elpmas-owt htiw elpmas llarevo 13

.snoitalerroc esiwriaP :2 elbaT srehtO rof emit dab :.tta gnisahcruP esrow noitautis laicnaniF elameF ylredlE emocnI elciheV esuoH elbaruD lanimoN laicnaniF laicnaniF )+56 egA( mottob sesahcrup sesahcrup sesahcrup detcepxe itatcepxe noitautis %02 emocni sno esrow noitautis laicnaniF 1 noitautis laicnaniF 1 ***320.0 snoitatcepxe laicnaniF 1 ***122.0 ***271.0 emocni detcepxe lanimoN rof emit dab :sedutitta gnisahcruP 1 ***170.0 ***260.0 ***870.0 sesahcrup elbaruD 1 ***641.0 ***350.0 ***660.0 ***030.0 sesahcrup esuoH 1 ***281.0 ***752.0 ***260.0 ***560.0 ***640.0 sesahcrup elciheV srehtO 1 ***730.0 ***570.0 ***420.0 ***220.0 ***830.0 ***270.0 %02 mottob emocnI 1 ***712.0 400.0 200.0 ***110.0 ***160.0 ***490.0 ***820.0 )+56 egA( ylredlE 1 ***140.0 ***711.0 ***720.0 ***530.0 ***430.0 ***410.0 400.0 ***530.0 elameF emiT .level %1 eht ta ecnac(cid:133)ingis lacitsitats setoned .sisylana noisserger eht ni deyolpme selbairav eht gnoma snoitalerroc esiwriap stroper elbat ehT :setoN (cid:3)(cid:3)(cid:3) .1102(cid:150)0891 :doirep 14

is around 1 percentage point larger than the one of the other consumers. While these descriptive statistics are unconditional, we will see later on that the di⁄erences remain relevant also when we control for other consumer characteristics. Aquestionthatarisesistowhatextentthevariousconsumercategoriesthatwedistinguish are correlated, or in other words whether one can assume that they are reasonably independent to warrant a separate interpretation. Table 2 reports pairwise Pearson correlations among the attributes we include in the analysis. All the correlations are highly statistically signi(cid:133)cant, but surprisingly small from an economic point of view, with most of them being substantially smaller than 0.1. Based on these results, we will conduct separate regression analyses, using one characteristic at a time, and interpret the results as independent, but it is important to keep in mind that the characteristics are not entirely unrelated. 3 Econometric Framework This section explains the econometric framework employed in the analysis. We are interested whether the in(cid:135)ation expectations of our consumer groups are more upward biased than those of their peers. For that purpose, we specify the following linear regression model: BIAS = (cid:11) +c (cid:11) +NEWSP(cid:11) +NEWSN(cid:11) +x (cid:11) (1) i 1 i 2 i 3 4 i 5 +c NEWSP(cid:11) +c NEWSN(cid:11) +u ; i (cid:1) i 6 i (cid:1) 7 i BIAS = BIAS(cid:25);BIASF ; (2) i i i (cid:8) (cid:9) where BIAS(cid:25) is the di⁄erence between the MS consumer-speci(cid:133)c forecast and CPI in(cid:135)ai tion (at the forecast horizon), and BIASF is the di⁄erence between the MS consumeri speci(cid:133)c forecast and the SPF mean in(cid:135)ation forecast. A comparison with actual, realized in(cid:135)ation will tell us about the overall bias of in(cid:135)ation expectations, whereas the comparison with the SPF is meant to compare consumer expectations against a forecast that is in principle conditional on the same information set, namely the information available at the time of the forecast. (cid:11) is a constant, c denotes the consumer classi(cid:133)cation of interest, NEWSP is an 1 i i individual-speci(cid:133)cindicatorofnewsperception(whichequalsoneiftheintervieweehas,in the previous months, heard of recent changes in prices and zero otherwise), and NEWSN indexes the intensity of news coverage at the time of the survey.7 x is a vector of socioi 7In a robustness test, we will also include the last observed CPI in(cid:135)ation rate. We have furthermore consideredthepossibilitythatconsumerslookatalternativein(cid:135)ationmeasures, suchastheaveragerate of in(cid:135)ation over the six months reinterview period, but did not obtain di⁄erent results. 15

economic characteristics (namely gender, age, income, education, race, marital status, location in the United States)8 and u is assumed to be normally distributed. We also i interact the consumer classi(cid:133)cation variable with each of the news intensity measures. While (cid:11) will reveal whether the various consumer groups di⁄er in their bias, the para- 2 meters (cid:11) and (cid:11) will reveal whether they di⁄er in their response to news. Note that 6 7 we omitted time subscripts for simplicity. To assess the statistical signi(cid:133)cance of our estimates, we calculate robust standard errors using the sandwich estimator. 4 The Determinants of Consumer In(cid:135)ation Expectations 4.1 Benchmark Results Having speci(cid:133)ed the data and the econometric model, we next discuss the econometric results. Tables3and4con(cid:133)rmtheprevious(cid:133)ndingsthatconsumerin(cid:135)ationexpectations are biased upwards. The constant ((cid:11) ) re(cid:135)ects the bias of the benchmark consumer, i.e., 1 anagentwiththefollowingcharacteristics: white(non-Hispanic), married, male, 40years old, highschooldiploma, anincomeinthemiddlequintileofthedistributionandlivingin the North-Center of the country. The bias of the benchmark consumer is estimated to be statistically signi(cid:133)cant and positive both when we compare in(cid:135)ation expectations against realized in(cid:135)ation in Table 3 (where we (cid:133)nd a bias in the order of 0.5 to 0.6 percentage points) and when we compare against those of professional forecasters in Table 4 (with a bias of around 0.3 percentage points). While the in(cid:135)ation expectations of the representative consumer are biased upwards, the bias is substantially larger for the consumer groups that we study (with the exception of age, where a negative coe¢ cient is in line with the previous literature). The additional bias ((cid:11) ) is particularly large for consumers with pessimistic expectations about their (cid:133)- 2 nancialsituation, amountingto1additionalpercentagepoint. However, alsofortheother groups, we detect an additional upward bias, which is similar in magnitude to what we (cid:133)nd for the consumers in the bottom 20% of the income distribution and slightly smaller than for females. These results hold when comparing consumer in(cid:135)ation expectations to actual in(cid:135)ation and to professional forecasters. Having heard news about prices (which itself is heavily in(cid:135)uenced by positive gasoline 8Household income is grouped into quintiles and age is measured in integers, while education is split into six groups: (cid:147)Grade 0-8, no high school diploma,(cid:148)(cid:147)Grade 9-12, no high school diploma,(cid:148)(cid:147)Grade 0-12, with high school diploma,(cid:148)(cid:147)4 yrs. of college, no degree,(cid:148)(cid:147)3 yrs. of college, with degree(cid:148)and (cid:147)4 yrs. of college, with degree.(cid:148)Race is grouped into (cid:147)White except Hispanic,(cid:148)(cid:147)African-American except Hispanic,(cid:148)(cid:147)Hispanic,(cid:148)(cid:147)American Indian or Alaskan Native(cid:148)and (cid:147)Asian or Paci(cid:133)c Islander,(cid:148)while marital status is given as (cid:147)Married/with a partner,(cid:148)(cid:147)Divorced,(cid:148)(cid:147)Widowed,(cid:148)(cid:147)Never married.(cid:148)Finally, the region of residence is grouped into (cid:147)West,(cid:148)(cid:147)North Central,(cid:148)(cid:147)Northeast,(cid:148)(cid:147)South.(cid:148) 16

.noita(cid:135)ni lautca ot evitaler saib fo stnanimreteD :3 elbaT srehtO rof emit dab :sedutitta gnisahcruP esrow noitautis laicnaniF elameF ylredlE emocnI elciheV esuoH elbaruD lanimoN laicnaniF laicnaniF )+56 egA( mottob sesahcrup sesahcrup sesahcrup detcepxe itatcepxe noitautis %02 emocni sno ***8796.0 ***5303.0 ***5783.0 ***2965.0 ***9616.0 ***4834.0 ***0105.0 ***5090.1 ***6841.0 ) (a citsiretcarahc HH 2 )1420.0( )3140.0( )2640.0( )1230.0( )6030.0( )8130.0( )7430.0( )5730.0( )3920.0( ***3513.1 ***3582.1 ***9282.1 ***0321.1 ***3813.1 ***6891.1 ***0652.1 ***1681.1 ***0423.1 ) a( PSWEN 3 )1470.0( )0360.0( )4060.0( )8360.0( )6360.0( )0260.0( )6260.0( )3260.0( )5360.0( 5400.0 5741.0 8562.0 ***0606.0 4031.0 ***8055.0 *6403.0 *7492.0 6950.0 ) a( .hC * PSWEN 6 )0611.0( )5361.0( )7702.0( )0441.0( )4641.0( )0461.0( )5761.0( )0061.0( )5851.0( ***2244.0 ***0104.0 ***5454.0 ***4044.0 ***5624.0 ***8064.0 ***5864.0 ***0515.0 ***4694.0 ) a( NSWEN 4 )1320.0( )0910.0( )5810.0( )9810.0( )9910.0( )9810.0( )5810.0( )5810.0( )1910.0( ***7890.0 ***8936.0 ***2303.0 ***0492.0 ***6993.0 ***4681.0 ***2051.0 8660.0 0430.0 ) a( .hC * NSWEN 7 )9330.0( )1940.0( )8250.0( )3140.0( )3630.0( )8140.0( )5840.0( )2840.0( )8140.0( ***1006.0 ***8785.0 ***2595.0 ***2115.0 ***2105.0 ***3625.0 ***8535.0 ***7574.0 ***5965.0 ) a( tnatsnoC 1 )9180.0( )9180.0( )9180.0( )9180.0( )7180.0( )9180.0( )9180.0( )7180.0( )1280.0( 000.0 000.0 000.0 000.0 000.0 000.0 000.0 000.0 000.0 0= α+ α:1 tseT 6 3 000.0 000.0 000.0 000.0 000.0 000.0 000.0 000.0 000.0 0= α+ α :2 tseT 7 4 741571 741571 741571 741571 741571 741571 741571 741571 741571 N 0473 2183 3273 5414 7414 0493 3093 2264 8763 2ihC lortnocsledomllA .21+tninoita(cid:135)nilautcadnasnoitatcepxeremusnocneewtebecnere⁄idehtgninialpxe,)1(noitauqenodesabstluserstroperelbatehT :setoN .redaeh nmuloc eht ni detroper si citsiretcarahc remusnoc tnaveler ehT .setatS detinU eht ni noitacol ,sutats latiram ,ecar ,noitacude ,emocni ,ega ,redneg rof eht fi eno slauqe hcihw( noitpecrep swen fo rotacidni c(cid:133)iceps-laudividni na si PSWEN .3.2 noitceS ni debircsed era scitsiretcarahc eseht fo snoitin(cid:133)ed ehT 1 tseT .aidem eht ni egarevoc swen detaler-noita(cid:135)ni fo ytisnetni eht sexedni NSWEN ,)esiwrehto orez dna secirp ni segnahc tnecer fo draeh sah eeweivretni .snoitavresbo fo rebmun eht setoned N .0 = (cid:11) + (cid:11) fo tset )1(2ihC a fo seulav-p setoned 2 tseT .0 = (cid:11) + (cid:11) fo tset )1(2ihC a fo seulav-p setoned 7 4 6 3 .1102(cid:150)0891 :doirep emiT .level %01/%5/%1 eht ta ecnac(cid:133)ingis lacitsitats setoned */**/*** .srorre dradnats era sesehtnerap ni srebmuN 17

.stsacerof lanoisseforp ot evitaler saib fo stnanimreteD :4 elbaT srehtO rof emit dab :sedutitta gnisahcruP esrow noitautis laicnaniF elameF ylredlE emocnI elciheV esuoH elbaruD lanimoN laicnaniF laicnaniF )+56 egA( mottob sesahcrup sesahcrup sesahcrup detcepxe itatcepxe noitautis %02 emocni sno ***9986.0 ***5213.0 ***4863.0 ***4605.0 ***2184.0 ***8443.0 ***9764.0 ***7540.1 ***1441.0 ) (a citsiretcarahc HH 2 )2320.0( )6930.0( )8440.0( )8030.0( )4920.0( )6030.0( )6330.0( )0630.0( )4820.0( ***0991.1 ***8002.1 ***3261.1 ***6180.1 ***8931.1 ***7051.1 ***7151.1 ***5980.1 ***7202.1 ) a( PSWEN 3 )2760.0( )1850.0( )5550.0( )5950.0( )2850.0( )8750.0( )8750.0( )3750.0( )4850.0( 5410.0 7990.0 3952.0 **9292.0 2621.0 1661.0 9481.0 0771.0 1350.0 ) a( .hC * PSWEN 6 )0701.0( )7051.0( )2691.0( )4131.0( )1731.0( )5841.0( )3451.0( )5941.0( )2841.0( ***3158.0 ***5597.0 ***1658.0 ***1278.0 ***4658.0 ***5488.0 ***5868.0 ***7949.0 ***2888.0 ) a( NSWEN 4 )6220.0( )4810.0( )0810.0( )4810.0( )3910.0( )4810.0( )0810.0( )8710.0( )5810.0( ***5211.0 ***4097.0 ***5414.0 ***9412.0 ***2023.0 ***4541.0 ***9682.0 5960.0 ***2801.0 ) a( .hC * NSWEN 7 )9230.0( )0740.0( )8050.0( )1040.0( )4530.0( )6040.0( )4740.0( )3740.0( )4040.0( ***5153.0 ***9433.0 ***0543.0 ***5172.0 ***7472.0 ***9292.0 ***9092.0 ***3822.0 ***3023.0 ) a( tnatsnoC 1 )1970.0( )0970.0( )0970.0( )0970.0( )0970.0( )1970.0( )0970.0( )8870.0( )2970.0( 000.0 000.0 000.0 000.0 000.0 000.0 000.0 000.0 000.0 0= α+ α:1 tseT 6 3 000.0 000.0 000.0 000.0 000.0 000.0 000.0 000.0 000.0 0= α+ α :2 tseT 7 4 741571 741571 741571 741571 741571 741571 741571 741571 741571 N 6116 7736 7516 9546 7736 7326 0236 0807 2016 2ihC llA .sretsaceroF lanoisseforP fo yevruS eht dna snoitatcepxe remusnoc neewteb ecnere⁄id eht gninialpxe ,)1( noitauqe no desab stluser stroper elbat ehT :setoN eht ni detroper si citsiretcarahc remusnoc tnaveler ehT .setatS detinU eht ni noitacol ,sutats latiram ,ecar ,noitacude ,emocni ,ega ,redneg rof lortnoc sledom hcihw( noitpecrep swen fo rotacidni c(cid:133)iceps-laudividni na si PSWEN .3.2 noitceS ni debircsed era scitsiretcarahc eseht fo snoitin(cid:133)ed ehT .redaeh nmuloc eht ni egarevoc swen detaler-noita(cid:135)ni fo ytisnetni eht sexedni NSWEN ,)esiwrehto orez dna secirp ni segnahc tnecer fo draeh sah eeweivretni eht fi eno slauqe fo rebmun eht setoned N .0 = (cid:11)+ (cid:11) fo tset )1(2ihC a fo seulav-p setoned 2 tseT .0 = (cid:11)+ (cid:11) fo tset )1(2ihC a fo seulav-p setoned 1 tseT .aidem 7 4 6 3 .1102(cid:150)0891 :doirep emiT .level %01/%5/%1 eht ta ecnac(cid:133)ingis lacitsitats setoned */**/*** .srorre dradnats era sesehtnerap ni srebmuN .snoitavresbo 18

in(cid:135)ation) increases the bias by around 1.2 to 1.3 percentage points when compared to actualin(cid:135)ation,andbyaround1.1to1.2percentagepointswhencomparedtoprofessional forecasts ((cid:11) ). Interestingly, this e⁄ect does not systematically di⁄er across consumer 3 groups ((cid:11) ), suggesting that the e⁄ect of gasoline price in(cid:135)ation on in(cid:135)ation expectations 6 is universal, and relatively homogeneous across di⁄erent consumer types. Contrary to having heard news about prices, more media reporting about in(cid:135)ation tends to reduce the bias in in(cid:135)ation expectations ((cid:11) ). A one-standard-deviation increase 4 in media reporting (i.e., a change in the index by around 0.8 percentage points), ceteris paribus, leads to a reduction in the bias of around 0.3 to 0.4 percentage points when measured against actual in(cid:135)ation, and of around 0.7 to 0.8 percentage points when measured against the SPF. The e⁄ect is estimated to be di⁄erent across consumer groups ((cid:11) ), 7 with a larger reduction in the bias of pessimistic consumers and those in dire (cid:133)nancial situations; to give one example, consumers who are pessimistic about house purchases see their bias relative to actual in(cid:135)ation reduced by nearly twice as much as does the average consumer. This result suggests that more news coverage is bene(cid:133)cial in that (i) it reduces the bias in in(cid:135)ation expectations of consumers more generally, and (ii) it does so particularly for those consumer groups that had a larger bias to start with. 4.2 In(cid:135)ation Expectations During Recessions Intheprevioussection, weproxiedconsumers(cid:146)pessimismbymeansoftheirownresponses totheMS.Anotherwaytogetatconsumerpessimismistotesttowhatextentconsumers(cid:146) forecast accuracy di⁄ers during recessions, i.e in times when there is generally less reason for optimism about economic prospects. Accordingly, we have enhanced our econometric model as follows: BIAS = (cid:11) +c (cid:11) +NEWSP(cid:11) +NEWSN(cid:11) +x (cid:11) +c NEWSP(cid:11) (3) i 1 i 2 i 3 4 i 5 i i 6 +c NEWSN(cid:11) +NBER(cid:11) +c NBER(cid:11) i 7 8 i 9 (cid:1) (cid:1) +NBER NEWSP(cid:11) +NBER NEWSN(cid:11) +u ; (cid:1) i 10 (cid:1) 11 i where NBER is a dummy variable that is equal to one during NBER recessions. This model tests whether the bias in in(cid:135)ation expectations di⁄ers during recessions (by means of (cid:11) ), whether there is an additional di⁄erentiation across consumer groups ((cid:11) ), and 8 9 whether the responsiveness to news changes ((cid:11) and (cid:11) ). The results are reported in 10 11 Table 5. A number of (cid:133)ndings are noteworthy. First, during recessions, there is a substantial additional upward bias in in(cid:135)ation expectations in the order of 2 percentage points presumably because consumers underestimate how much in(cid:135)ation tends to drop during 19

.snoissecer REBN gnidulcni ,noita(cid:135)ni lautca ot evitaler saib fo stnanimreteD :5 elbaT srehtO rof emit dab :sedutitta gnisahcruP esrow noitautis laicnaniF elameF ylredlE emocnI elciheV esuoH elbaruD lanimoN laicnaniF laicnaniF )+56 egA( mottob sesahcrup sesahcrup sesahcrup detcepxe itatcepxe noitautis %02 emocni sno ***2896.0 ***8672.0 ***1713.0 ***2933.0 ***6854.0 ***1852.0 ***6894.0 ***6759.0 ***6522.0 ) (a citsiretcarahc HH 2 )8420.0( )9140.0( )0840.0( )4430.0( )9230.0( )4430.0( )4630.0( )6930.0( )7030.0( ***4648.0 ***3408.0 ***9408.0 ***4737.0 ***1658.0 ***8097.0 ***1267.0 ***5647.0 ***9028.0 ) (a PSWEN 3 )4470.0( )0560.0( )2260.0( )4560.0( )7460.0( )0460.0( )0560.0( )2460.0( )5560.0( 2410.0 6302.0 3503.0 ***9963.0 8871.0 *5792.0 **4993.0 1591.0 6011.0 ) (a .hC * PSWEN 6 )7211.0( )4951.0( )8302.0( )3041.0( )6441.0( )1061.0( )1361.0( )2651.0( )0651.0( ***6844.0 ***2314.0 ***4754.0 ***8534.0 ***9493.0 ***8274.0 ***9184.0 ***5305.0 ***6315.0 ) (a NSWEN 4 )6420.0( )8020.0( )3020.0( )5020.0( )2120.0( )5020.0( )3020.0( )1020.0( )7020.0( ***5490.0 ***8285.0 ***6292.0 ***4833.0 ***0094.0 ***5641.0 *4380.0 **9401.0 ***9421.0 ) (a .hC * NSWEN 7 )9330.0( )3940.0( )9250.0( )1240.0( )8730.0( )5240.0( )9840.0( )5940.0( )0240.0( ***9749.1 ***5619.1 ***8459.1 ***3387.1 ***4588.1 ***3598.1 ***6099.1 ***6358.1 ***0090.2 ) (a REBN 8 )7172.0( )7172.0( )6172.0( )5172.0( )0172.0( )4272.0( )1272.0( )2172.0( )1272.0( 5010.0 *5991.0 ***7015.0 ***2654.0 **4461.0 8011.0 **1512.0 ***8203.0 ***8156.0 ) a( .hC * REBN 9 )7170.0( )1711.0( )1931.0( )7580.0( )6380.0( )6480.0( )7001.0( )6401.0( )0780.0( ***5098.0 ***8298.0 ***8398.0 ***6028.0 ***6329.0 ***8358.0 ***2519.0 ***9768.0 ***5568.0 ) a(PSWEN * REBN 01 )0631.0( )9531.0( )9531.0( )3631.0( )3631.0( )7631.0( )9531.0( )7531.0( )8531.0( ***6825.0 ***1225.0 ***6435.0 ***4115.0 ***9625.0 ***1215.0 ***2625.0 ***5545.0 ***6555.0 ) a( NSWEN * REBN 11 )9830.0( )8830.0( )9830.0( )9830.0( )2040.0( )0930.0( )9830.0( )9830.0( )1930.0( ***0843.0 ***2243.0 ***4343.0 ***4992.0 ***6382.0 ***9503.0 ***5182.0 ***3742.0 ***5103.0 ) a( tnatsnoC 1 )3180.0( )2180.0( )2180.0( )3180.0( )1180.0( )3180.0( )1180.0( )9080.0( )4180.0( 000.0 000.0 000.0 000.0 000.0 000.0 000.0 000.0 000.0 0=α+α:1 tseT 6 3 000.0 000.0 000.0 000.0 000.0 000.0 000.0 000.0 000.0 0=α+α :2 tseT 7 4 741571 741571 741571 741571 741571 741571 741571 741571 741571 N 5596 1307 3596 2527 7827 1407 2517 7677 9996 2ihC lortnocsledomllA .21+tninoita(cid:135)nilautcadnasnoitatcepxeremusnocneewtebecnere⁄idehtgninialpxe,)3(noitauqenodesabstluserstroperelbatehT :setoN .redaeh nmuloc eht ni detroper si citsiretcarahc remusnoc tnaveler ehT .setatS detinU eht ni noitacol ,sutats latiram ,ecar ,noitacude ,emocni ,ega ,redneg rof eht fi eno slauqe hcihw( noitpecrep swen fo rotacidni c(cid:133)iceps-laudividni na si PSWEN .3.2 noitceS ni debircsed era scitsiretcarahc eseht fo snoitin(cid:133)ed ehT REBN .aidem eht ni egarevoc swen detaler-noita(cid:135)ni fo ytisnetni eht sexedni NSWEN ,)esiwrehto orez dna secirp ni segnahc tnecer fo draeh sah eeweivretni .0 = (cid:11)+ (cid:11) fo tset )1(2ihC a fo seulav-p setoned 2 tseT .0 = (cid:11)+ (cid:11) fo tset )1(2ihC a fo seulav-p setoned 1 tseT .snoissecer rof elbairav ymmud a si 7 4 6 3 emiT .level %01/%5/%1 eht ta ecnac(cid:133)ingis lacitsitats setoned */**/*** .srorre dradnats era sesehtnerap ni srebmuN .snoitavresbo fo rebmun eht setoned N .1102(cid:150)0891 :doirep 20

recessions (or because the pessimism rises). Second, having heard news about prices during recessions substantially increases the bias, by nearly one additional percentage point. Third, additional media reporting is bene(cid:133)cial in the sense that it reduces the bias signi(cid:133)cantly. Fourth, while some of the interaction terms with our consumer characteristics are statistically signi(cid:133)cant, they are not consistently signi(cid:133)cant, and have di⁄erent signs, such that no clear pattern is emerging. Finally, it is important to note that the results of the previous section all remain valid (cid:150)the consumer characteristics themselves matter as before, and the way they interact with news about in(cid:135)ation. This suggests that both proxies for pessimism, via the responses in the MS and via the recession dummy, provide us with independent evidence pointing in the same direction. 4.3 Determinants of the Bias According to Jonung (1981) and Bryan and Venkatu (2001), taking into account demographic characteristics reduces the unexplained bias in the level of consumer forecasts. In this section, we look at the connection between the bias and the set of explanatory variables in the regression models we have considered so far. To this end, we plot the estimated constant terms. According to Figure 6, when regressing BIAS on a constant only, the resulting uni conditional bias is around 0:8. When we account for demographics, the bias for the benchmark consumer reduces to about 0:6. Including the NBER recessions reduces this bias further (cid:150)by about 0:2 (cid:150)while adding consumer attitudes reduces the unexplained part of the bias to about 0:23, on average. Notably, when accounting for consumers that declare to have negative nominal income expectations, the resulting constant is not statistically di⁄erent from zero. Overall, the picture emerging from this exercise is that our set of explanatory variables allows compressing the unexplained bias that previous 21

contributions have typically reported. Figure 6: The unexplained bias in the level of consumer forecasts. 8 . 6 . 4 . 2 . 0 2 . Unc. Bias Dem. Rec. Dem. + Rec. HH Full Model Mean Prediction 95% Conf. Int. Notes: Thechartreportstheunexplainedbiasinthelevelofconsumerforecastsfromamodelcontaining: (1)aconstant(Unc. Bias);aconstantandthedemographiccharacteristicsoftherepresentativeconsumer (Dem.); (2) a constant and the NBER recession dummy (Rec.); (3) a constant, consumers(cid:146)demographic characteristicsandtheNBERrecessiondummy(Dem. +Rec.); (4)aconstant,consumers(cid:146)demographic characteristics, the NBER recession dummy and consumer attitudes (HH-Full). In the last column the height of the shaded area indicates the average of the constants in the models obtained by alternatively including six di⁄erent types of consumer attitudes. 4.4 Robustness We have conducted several robustness checks to investigate the sensitivity of our results to our modelling choices. For brevity, we will only show those that relate to the bias of consumersrelativetoactualin(cid:135)ation(i.e.,thosereportedinTable3),butresultsgenerally hold also for the other analyses. For the (cid:133)rst robustness check, we added lagged actual in(cid:135)ation as an explanatory variable to the regression (see Table 6). It turns out that the magnitude of the bias is not responsive to past developments of in(cid:135)ation. Accordingly, all our results go through.9 Another robustness test checks for those consumers who are pessimistic about major 9In an alternative regression, we have also included gasoline price in(cid:135)ation in the set of regressors. However,despitethecloseconnectionbetweenhearingnewsaboutpricesandincreasesingasolineprices, the coe¢ cient attached to NEWSP remains statistically signi(cid:133)cant and preserves its sign. The same is true when we add consumers(cid:146)expectations about gasoline price developments based on a question in the MS. As there is substantial item non-response to that particular question, this estimation is based on far fewer observations and therefore not considered as the benchmark regression. 22

.noita(cid:135)ni lautca gnidulcni ssentsubor ,noita(cid:135)ni lautca ot evitaler saib fo stnanimreteD :6 elbaT srehtO rof emit dab :sedutitta gnisahcruP esrow noitautis laicnaniF elameF ylredlE emocnI elciheV esuoH elbaruD lanimoN laicnaniF laicnaniF )+56 egA( mottob sesahcrup sesahcrup sesahcrup detcepxe itatcepxe noitautis %02 emocni sno ***8796.0 ***7303.0 ***6683.0 ***3375.0 ***7026.0 ***1044.0 ***8005.0 ***0490.1 ***3841.0 ) (a citsiretcarahc HH 2 )1420.0( )3140.0( )2640.0( )3230.0( )9030.0( )9130.0( )7430.0( )5730.0( )3920.0( ***3913.1 ***1392.1 ***2982.1 ***1431.1 ***8623.1 ***9502.1 ***6852.1 ***5002.1 ***1723.1 ) a( PSWEN 3 )4470.0( )4360.0( )9060.0( )2460.0( )9360.0( )4260.0( )0360.0( )7260.0( )9360.0( 6400.0 4541.0 0462.0 ***1506.0 3921.0 ***1055.0 *8303.0 *2392.0 4060.0 ) a( .hC * PSWEN 6 )0611.0( )5361.0( )7702.0( )0441.0( )4641.0( )0461.0( )5761.0( )0061.0( )5851.0( 9300.0 2700.0 8500.0 8010.0 7800.0 9600.0 3200.0 *9310.0 9200.0 noitalfnI )6700.0( )5700.0( )6700.0( )6700.0( )7700.0( )6700.0( )6700.0( )6700.0( )6700.0( ***9134.0 ***4183.0 ***7834.0 ***5214.0 ***4504.0 ***6244.0 ***2264.0 ***9874.0 ***6884.0 ) a( NSWEN 4 )6720.0( )3420.0( )7320.0( )6320.0( )8320.0( )6320.0( )7320.0( )4320.0( )3420.0( ***8890.0 ***5146.0 ***9403.0 ***0092.0 ***7393.0 ***3481.0 ***5051.0 6060.0 4330.0 ) a( .hC * NSWEN 7 )9330.0( )1940.0( )8250.0( )5140.0( )9630.0( )9140.0( )5840.0( )4840.0( )8140.0( ***6416.0 ***7416.0 ***9616.0 ***7055.0 ***6235.0 ***9155.0 ***6445.0 ***0725.0 ***3085.0 ) a( tnatsnoC 1 )8580.0( )7580.0( )8580.0( )7580.0( )6580.0( )8580.0( )8580.0( )5580.0( )0680.0( 000.0 000.0 000.0 000.0 000.0 000.0 000.0 000.0 000.0 0= α+ α:1 tseT 6 3 000.0 000.0 000.0 000.0 000.0 000.0 000.0 000.0 000.0 0= α+ α :2 tseT 7 4 741571 741571 741571 741571 741571 741571 741571 741571 741571 N 5283 4983 0083 5124 6124 8104 6993 8864 1773 2ihC lortnocsledomllA .21+tninoita(cid:135)nilautcadnasnoitatcepxeremusnocneewtebecnere⁄idehtgninialpxe,)1(noitauqenodesabstluserstroperelbatehT :setoN .redaeh nmuloc eht ni detroper si citsiretcarahc remusnoc tnaveler ehT .setatS detinU eht ni noitacol ,sutats latiram ,ecar ,noitacude ,emocni ,ega ,redneg rof eht fi eno slauqe hcihw( noitpecrep swen fo rotacidni c(cid:133)iceps-laudividni na si PSWEN .3.2 noitceS ni debircsed era scitsiretcarahc eseht fo snoitin(cid:133)ed ehT 1 tseT .aidem eht ni egarevoc swen detaler-noita(cid:135)ni fo ytisnetni eht sexedni NSWEN ,)esiwrehto orez dna secirp ni segnahc tnecer fo draeh sah eeweivretni .snoitavresbo fo rebmun eht setoned N .0 = (cid:11) + (cid:11) fo tset )1(2ihC a fo seulav-p setoned 2 tseT .0 = (cid:11) + (cid:11) fo tset )1(2ihC a fo seulav-p setoned 7 4 6 3 .1102(cid:150)0891 :doirep emiT .level %01/%5/%1 eht ta ecnac(cid:133)ingis lacitsitats setoned */**/*** .srorre dradnats era sesehtnerap ni srebmuN 23

ssentsubor ,noita(cid:135)ni lautca ot evitaler saib fo stnanimreteD :7 elbaT .secirp gnisir yb denimreted gnieb sedutitta remusnoc rof tset rof emit dab :sedutitta gnisahcruP ,sesahcrup elciheV ,sesahcrup esuoH ,sesahcrup elbaruD ,noitautis laicnaniF secirp ot eud secirp ot eud secirp ot eud secirp ot eud ***2425.0 ***6065.0 ***3267.0 ***7776.0 ) (a citsiretcarahc HH 2 )4430.0( )5740.0( )6750.0( )5520.0( ***4773.1 ***9104.1 ***9033.1 ***5702.1 ) a( PSWEN 3 )4160.0( )7060.0( )9950.0( )9860.0( ***9584.0 ***4949.0 *5504.0 5991.0 ) a( .hC * PSWEN 6 )0381.0( )8502.0( )0832.0( )1521.0( ***3405.0 ***0284.0 ***3064.0 ***0354.0 ) a( NSWEN 4 )7810.0( )5810.0( )1810.0( )5020.0( ***6421.0 ***0782.0 ***9066.0 ***8151.0 ) (a .hC * NSWEN 7 )0740.0( )7350.0( )7560.0( )3430.0( ***5325.0 ***4065.0 ***7975.0 ***8883.0 ) a( tnatsnoC 1 )1280.0( )0280.0( )9180.0( )1280.0( 000.0 120.0 000.0 000.0 0= α+ α:1 tseT 6 3 000.0 000.0 000.0 000.0 0= α+ α :2 tseT 7 4 741571 741571 741571 741571 N 1783 2873 6483 5754 2ihC lortnocsledomllA .21+tninoita(cid:135)nilautcadnasnoitatcepxeremusnocneewtebecnere⁄idehtgninialpxe,)1(noitauqenodesabstluserstroperelbatehT :setoN ,redaeh nmuloc eht ni detroper si citsiretcarahc remusnoc tnaveler ehT .setatS detinU eht ni noitacol ,sutats latiram ,ecar ,noitacude ,emocni ,ega ,redneg rof .3.2 noitceS ni debircsed era scitsiretcarahc eseht fo snoitin(cid:133)ed ehT .tnemssessa rieht rof nosaer gniylrednu eht sa secirp gnisir evig taht sremusnoc gniredisnoc ,)esiwrehto orez dna secirp ni segnahc tnecer fo draeh sah eeweivretni eht fi eno slauqe hcihw( noitpecrep swen fo rotacidni c(cid:133)iceps-laudividni na si PSWEN setoned 2 tseT .0 = (cid:11)+ (cid:11) fo tset )1(2ihC a fo seulav-p setoned 1 tseT .aidem eht ni egarevoc swen detaler-noita(cid:135)ni fo ytisnetni eht sexedni NSWEN 6 3 lacitsitats setoned */**/*** .srorre dradnats era sesehtnerap ni srebmuN .snoitavresbo fo rebmun eht setoned N .0 = (cid:11)+ (cid:11) fo tset )1(2ihC a fo seulav-p 7 4 .1102(cid:150)0891 :doirep emiT .level %01/%5/%1 eht ta ecnac(cid:133)ingis 24

.sremusnoc deweivretni-er rof tset ssentsubor ,noita(cid:135)ni lautca ot evitaler saib fo stnanimreteD :8 elbaT srehtO rof emit dab :sedutitta gnisahcruP esrow noitautis laicnaniF elameF ylredlE emocnI elciheV esuoH elbaruD lanimoN laicnaniF laicnaniF )+56 egA( mottob sesahcrup sesahcrup sesahcrup detcepxe itatcepxe noitautis %02 emocni sno ***0323.0 ***2343.0 ***1862.0 ***7774.0 ***1974.0 ***8823.0 ***1234.0 ***8439.0 9770.0 ) (a citsiretcarahc HH 2 )3430.0( )1160.0( )2470.0( )9250.0( )7050.0( )6250.0( )4650.0( )2560.0( )4740.0( ***7842.1 ***3751.1 ***3742.1 ***3479.0 ***4432.1 ***9941.1 ***5141.1 ***3501.1 ***0371.1 ) a( PSWEN 3 )4611.0( )7001.0( )4790.0( )4101.0( )5101.0( )8890.0( )7001.0( )7990.0( )4201.0( 8970.0 0492.0 3453.0 ***8058.0 5602.0 5792.0 1014.0 2442.0 6712.0 ) a( .hC * PSWEN 6 )1981.0( )3382.0( )9263.0( )6942.0( )7152.0( )5982.0( )9382.0( )0482.0( )1462.0( ***0734.0 ***6973.0 ***9354.0 ***0624.0 ***5014.0 ***0544.0 ***4264.0 ***7764.0 ***2605.0 ) a( NSWEN 4 )2430.0( )3820.0( )7720.0( )7820.0( )4030.0( )6820.0( )8720.0( )7720.0( )7820.0( 1380.0 ***3596.0 **7812.0 ***7703.0 ***4683.0 ***1802.0 8590.0 ***8162.0 **1371.0 ) a( .hC * NSWEN 7 )6050.0( )2570.0( )4580.0( )6860.0( )0060.0( )7960.0( )4280.0( )5580.0( )0070.0( *0902.0 8002.0 *5502.0 1931.0 9031.0 2061.0 6351.0 4331.0 9391.0 ) a( tnatsnoC 1 )4421.0( )3421.0( )3421.0( )6421.0( )5421.0( )3421.0( )3421.0( )2421.0( )7421.0( 000.0 000.0 110.0 000.0 000.0 000.0 000.0 000.0 000.0 0= α+ α:1 tseT 6 3 000.0 000.0 000.0 000.0 000.0 000.0 000.0 000.0 000.0 0= α+ α :2 tseT 7 4 ***859.0 ***859.0 ***859.0 ***859.0 ***859.0 ***859.0 ***859.0 ***759.0 ***959.0 ohR 23717 23717 23717 23717 23717 23717 23717 23717 23717 N 698 649 098 3101 999 939 849 2211 588 2ihC lortnocsledomllA .21+tninoita(cid:135)nilautcadnasnoitatcepxeremusnocneewtebecnere⁄idehtgninialpxe,)1(noitauqenodesabstluserstroperelbatehT :setoN .redaeh nmuloc eht ni detroper si citsiretcarahc remusnoc tnaveler ehT .setatS detinU eht ni noitacol ,sutats latiram ,ecar ,noitacude ,emocni ,ega ,redneg rof eht fi eno slauqe hcihw( noitpecrep swen fo rotacidni c(cid:133)iceps-laudividni na si PSWEN .3.2 noitceS ni debircsed era scitsiretcarahc eseht fo snoitin(cid:133)ed ehT 1 tseT .aidem eht ni egarevoc swen detaler-noita(cid:135)ni fo ytisnetni eht sexedni NSWEN ,)esiwrehto orez dna secirp ni segnahc tnecer fo draeh sah eeweivretni .snoitavresbo fo rebmun eht setoned N .0 = (cid:11) + (cid:11) fo tset )1(2ihC a fo seulav-p setoned 2 tseT .0 = (cid:11) + (cid:11) fo tset )1(2ihC a fo seulav-p setoned 7 4 6 3 osla hcihw( noisserger niam eht ni noitauqe noitceles eht fo slaudiser eht no tneic ¢eoc eht setoned ohR .detnemelpmi si )9791 namkceH( noitcerroc namkceH dradnats era sesehtnerap ni srebmuN .)scitsiretcarahc cihpargomed-oicos(cid:146)sremusnoc tnere⁄id neewteb smret noitcaretni emos srosserger fo tes eht ni sedulcni .level %01/%5/%1 eht ta ecnac(cid:133)ingis lacitsitats setoned */**/*** .srorre 25

.PSWEN gnidulcxe tset ssentsubor ,noita(cid:135)ni lautca ot evitaler saib fo stnanimreteD :9 elbaT srehtO rof emit dab :sedutitta gnisahcruP esrow noitautis laicnaniF elameF ylredlE emocnI elciheV esuoH elbaruD lanimoN laicnaniF laicnaniF )+56 egA( mottob sesahcrup sesahcrup sesahcrup detcepxe itatcepxe noitautis %02 emocni sno ***8296.0 ***2503.0 ***2193.0 ***8436.0 ***6726.0 ***3874.0 ***4425.0 ***8541.1 ***2831.0 ) (a citsiretcarahc HH 2 )9320.0( )1140.0( )0640.0( )5130.0( )1030.0( )5130.0( )2430.0( )7630.0( )0920.0( ***2634.0 ***2193.0 ***5544.0 ***2034.0 ***6814.0 ***2254.0 ***4854.0 ***4705.0 ***3884.0 ) a( NSWEN 4 )1320.0( )0910.0( )6810.0( )0910.0( )9910.0( )0910.0( )5810.0( )5810.0( )1910.0( ***8290.0 ***0546.0 ***8103.0 ***6403.0 ***7004.0 ***6281.0 ***3951.0 9660.0 0930.0 ) a( .hC * NSWEN 7 )0430.0( )1940.0( )8250.0( )3140.0( )3630.0( )8140.0( )6840.0( )2840.0( )8140.0( ***2546.0 ***8136.0 ***3936.0 ***5445.0 ***8345.0 ***3465.0 ***7775.0 ***7115.0 ***8616.0 ) a( tnatsnoC 1 )2280.0( )1280.0( )2280.0( )1280.0( )0280.0( )2280.0( )2280.0( )9180.0( )3280.0( 000.0 000.0 000.0 000.0 000.0 000.0 000.0 000.0 000.0 0= α+ α :1 tseT 7 4 741571 741571 741571 741571 741571 741571 741571 741571 741571 N 8513 3723 8713 1563 6953 1243 9633 7414 1313 2ihC sledom llA .21 + t ni noita(cid:135)ni lautca dna snoitatcepxe remusnoc neewteb ecnere⁄id eht gninialpxe ,)1( noitauqe no desab stluser stroper elbat ehT :setoN nmuloc eht ni detroper si citsiretcarahc remusnoc tnaveler ehT .setatS detinU eht ni noitacol ,sutats latiram ,ecar ,noitacude ,emocni ,ega ,redneg rof lortnoc .aidem eht ni egarevoc swen detaler-noita(cid:135)ni fo ytisnetni eht sexedni NSWEN .3.2 noitceS ni debircsed era scitsiretcarahc eseht fo snoitin(cid:133)ed ehT .redaeh */**/*** .srorre dradnats era sesehtnerap ni srebmuN .snoitavresbo fo rebmun eht setoned N .0 = (cid:11)+ (cid:11) fo tset )1(2ihC a fo seulav-p setoned 1 tseT 7 4 .1102(cid:150)0891 :doirep emiT .level %01/%5/%1 eht ta ecnac(cid:133)ingis lacitsitats setoned 26

purchases, or see themselves in a di¢ cult (cid:133)nancial situation, but who mention that this is due to increasing prices (whereas, so far, these had been excluded from the consumer groups). Of course, we would expect these consumers to have a substantially larger bias, and this is indeed the case, as shown in Table 7. The exception is consumers who think that it(cid:146)s a bad time to purchase a house due to prices (cid:150)which is intuitive, since these respondents most likely have house prices in mind when answering that question, so they need not have a larger bias with regard to consumer prices. All other results go throughwiththisrobustnesstest(cid:150)perceivednewsincreasesthebias,andmediareporting decreases it, and particularly so for the pessimistic consumers. A third robustness test is concerned with the fact that around 40% of the consumers in the MS get re-interviewed. The response behavior of these re-interviewed consumers has been studied by Anderson (2008) and Madeira and Zafar (2012), and seems to be characterized by some learning over time. Accordingly, it is interesting to restrict the analysis to re-interviewed consumers only, as their in(cid:135)ation expectations might be less biased than those of the entire sample.10 While the overall bias does indeed shrink somewhat when comparing Table 8 to Table 3, we still (cid:133)nd an elevated bias for our selected consumer groups; as well, the responsiveness to news is qualitatively unchanged. Our benchmark model contains a variable that indicates whether a respondent has heard news about prices. Our fourth robustness test drops this variable, NEWSP, with i resultsreportedinTable9. Theestimatedcoe¢ cientschangeonlymarginally, whilequalitatively all results remain intact, suggesting that both news variables exert independent e⁄ects on in(cid:135)ation expectations. Finally, one might wonder whether the e⁄ect would be more prominent had we only included respondents who have heard news about rising prices. As discussed earlier, most of the observations for this variable originate from respondents who have heard about rising prices, whereas very few report to have heard about declining prices. Replacing our variable for perceived news to include only news about rising prices does not alter our results (which are not shown, for brevity). 5 Conclusions What are the determinants of consumers(cid:146)in(cid:135)ation-forecast errors? This paper has used the microdata of the Surveys of Consumers to shed further light on this important question. While it is well known that a number of socio-economic characteristics such as gender, age or income a⁄ect in(cid:135)ation expectations, we have shown that the same also 10From a total of 71,629 re-interviews, we lose 6.3% of observations due to question attrition (i.e., 4,513 individuals decided not to provide a year-ahead in(cid:135)ation expectation). This may represent a potentialsourceofbias. Inordertoaccountforquestionattrition,weimplementtheHeckmancorrection (Heckman, 1979), a procedure that o⁄ers a means of correcting for non-randomly selected samples. 27

holds true for consumer attitudes. Having pessimistic attitudes (cid:150)for example, towardthe purchase of durables or homes, experiencing or expecting (cid:133)nancial di¢ culties, as well as expectations that household income will go down in the future (cid:150)a⁄ects in(cid:135)ation expectations in substantially, increasing the upward bias that is anyway inherent in consumer in(cid:135)ation expectations. The e⁄ects are not only statistically signi(cid:133)cant, they are substantial in magnitude, and thus help explain time variation in the evolution of consumer in(cid:135)ation expectations. Generally, consumer in(cid:135)ation expectations are highly sensitive to perceived news about rising prices, which themselves are tightly connected to the evolution of gasoline prices. Rising gasoline prices are noticed much more than falling gasoline prices, and they lead consumers to revise their expectations more frequently, but worsen their bias. This is in contrast to media reporting about in(cid:135)ation. More intense media reporting lowers the bias, and especially so for pessimistic consumers and consumers in dire (cid:133)nancial situations. The (cid:133)ndings have important implications for policy-makers. They suggest that more communication about in(cid:135)ation improves consumers(cid:146)in(cid:135)ation expectations, and particularly so for consumers who are in the right tail of the distribution, i.e., those who have a particularly strong upward bias. References Anderson, R.D. J.(2008): (cid:147)USConsumerIn(cid:135)ationExpectations: EvidenceRegarding Learning, Accuracy and Demographics,(cid:148)University of Manchester Discussion Paper 2008/099, University of Manchester. Armantier, O., S. Nelson, G. Topa, W. van der Klaauw, and B. Zafar (2012): (cid:147)The price is right: updating of in(cid:135)ation expectations in a randomized price information experiment,(cid:148)Sta⁄Reports 543, Federal Reserve Bank of New York. Baumeister, R., E. Bratslavsky, C. Finkenauer, and K. Vohs (2001): (cid:147)Bad is stronger than good,(cid:148)Review of General Psychology, 5, 323(cid:150)370. Blinder, A. S., and A. B. Krueger (2004): (cid:147)What Does the Public Know about Economic Policy, andHowDoes It KnowIt?,(cid:148)Brookings Papers on Economic Activity, 35(2004-1), 327(cid:150)397. Bryan, M. F., and G. Venkatu (2001): (cid:147)The demographics of in(cid:135)ation opinion surveys,(cid:148)Economic Commentary, (15). Capistran, C., and A. Timmermann (2009): (cid:147)Disagreement and Biases in In(cid:135)ation Expectations,(cid:148)Journal of Money, Credit and Banking, 41(2-3), 365(cid:150)396. 28

Carroll, C. D. (2003): (cid:147)Macroeconomic Expectations Of Households and Professional Forecasters,(cid:148)The Quarterly Journal of Economics, 118(1), 269(cid:150)298. Christensen, C., P. Els, and M. Rooij (2006): (cid:147)Dutch Households(cid:146)Perceptions of Economic Growth and In(cid:135)ation,(cid:148)De Economist, 154(2), 277(cid:150)294. Coibion, O., and Y. Gorodnichenko (2010): (cid:147)Information Rigidity and the Expectations Formation Process: A Simple Framework and New Facts,(cid:148)NBER Working Papers 16537, National Bureau of Economic Research, Inc. (2015): (cid:147)Is the Phillips Curve Alive and Well after All? In(cid:135)ation Expectations and the Missing Disin(cid:135)ation,(cid:148)American Economic Journal: Macroeconomics, 7(1), 197(cid:150)232. Curtin, R. (1996): (cid:147)Procedure to estimate price expectations,(cid:148)Mimeo, University of Michigan. (2013): (cid:147)Survey of Consumers,(cid:148) Discussion paper, Survey Reserach Center, University of Michigan. del Giovane, P., S. Fabiani, and R. Sabbatini (2009): (cid:147)What(cid:146)s Behind (cid:146)In(cid:135)ation Perceptions(cid:146)? A Survey-Based Analysis of Italian Consumers,(cid:148)Giornale degli Economisti, 68(1), 25(cid:150)52. Easaw, J., R. Golinelli, and M. Malgarini (2013): (cid:147)What determines households in(cid:135)ation expectations? Theory and evidence from a household survey,(cid:148)European Economic Review, 61(C), 1(cid:150)13. Friedrich, C. (2014): (cid:147)Global In(cid:135)ation Dynamics in the Post-Crisis Period: What Explains the Twin Puzzle?,(cid:148)Working Papers 14-36, Bank of Canada. Georganas, S., P. J. Healy, and N. Li (2014): (cid:147)Frequency bias in consumers? perceptions of in(cid:135)ation: An experimental study,(cid:148)European Economic Review, 67(C), 144(cid:150)158. Heckman, J. J.(1979): (cid:147)SampleSelectionBiasasaSpeci(cid:133)cationError,(cid:148)Econometrica, 47(1), 153(cid:150)61. Jonung, L. (1981): (cid:147)Perceived and Expected Rates of In(cid:135)ation in Sweden,(cid:148)American Economic Review, 71(5), 961(cid:150)68. Kahneman, D., and A. Tversky (1979): (cid:147)Prospect Theory: An Analysis of Decision under Risk,(cid:148)Econometrica, 47(2), 263(cid:150)91. 29

Lamla, M. J., and T. Maag (2012): (cid:147)The Role of Media for In(cid:135)ation Forecast Disagreement of Households and Professionals,(cid:148)Journal of Money, Credit and Banking, 44(7), 1325(cid:150)1350. Lombardelli, C., and J. Saleheen (2003): (cid:147)Public expectations of UK in(cid:135)ation,(cid:148) Bank of England Quarterly Bulletin, 43, 281(cid:150)290. Madeira, C., and B. Zafar (2012): (cid:147)Heterogeneous In(cid:135)ationExpectations andLearning,(cid:148)Federal Reserve Bank of New York Sta⁄ Report 536, Federal Reserve Bank of New York. Malgarini, M. (2009): (cid:147)Quantitative In(cid:135)ation Perceptions and Expectations of Italian Consumers,(cid:148)Giornale degli Economisti, 68(1), 53(cid:150)80. Malmendier, U., and S. Nagel (2013): (cid:147)Learning from In(cid:135)ation Experiences,(cid:148) Mimeo, UC Berkeley and Stanford University. Menz, J.-O., and P. Poppitz (2013): (cid:147)Households(cid:146)disagreement on in(cid:135)ation expectations and socioeconomic media exposure in Germany,(cid:148)Discussion Papers 27/2013, Deutsche Bundesbank, Research Centre. Pfajfar, D., and E. Santoro (2009): (cid:147)Asymmetries in In(cid:135)ation Expectations Across Sociodemographic Groups,(cid:148)Mimeo, Tilburg University. (2013): (cid:147)News on In(cid:135)ation and the Epidemiology of In(cid:135)ation Expectations,(cid:148) Journal of Money, Credit and Banking, 45(6), 1045(cid:150)1067. Ranyard, R., F. D. Missier, N. Bonini, D. Duxbury, and B. Summers (2008): (cid:147)Perceptions and expectations of price changes and in(cid:135)ation: A review and conceptual framework,(cid:148)Journal of Economic Psychology, 29(4), 378(cid:150)400. Sims, C. A. (2003): (cid:147)Implications of rational inattention,(cid:148)Journal of Monetary Economics, 50(3), 665(cid:150)690. Snir, A., and D. Levy (2011): (cid:147)Shrinking Goods and Sticky Prices: Theory and Evidence,(cid:148)Working Paper Series 17/11, The Rimini Centre for Economic Analysis. Souleles, N. S. (2004): (cid:147)Expectations, Heterogeneous Forecast Errors, and Consumption: Micro Evidence from the Michigan Consumer Sentiment Surveys,(cid:148)Journal of Money, Credit and Banking, 36(1), 39(cid:150)72. Tversky, A., and D. Kahneman (1974): (cid:147)Judgment under uncertainty: Heuristics and biases,(cid:148)Science, 185, 1124(cid:150)1131. Webley, P., and R. Spears (1986): (cid:147)Economic preferences and in(cid:135)ationary expectations,(cid:148)Journal of Economic Psychology, 7(3), 359(cid:150)369. 30