インフォテイメント研究所

Japanese Households' Inflation Attention Thresholds
(Summary)
This study analyzed the rational inattention hypothesis where households’ inflation attention become responsive to inflation rates once they exceed a threshold level. Using a threshold model, the following three points were found.First, data derived from the Google Trends analysis of the keyword “inflation” were a successful alternative indicator for households’ inflation attention level. A threshold level was found to be +3.0 percent for the U.S. and +1.5percent for Japan. In Japan, the sampling period needed to include the current inflationary phase, since the economy suffered from a prolonged period of low inflation. In addition, the word “price” in Japanese yielded much morestable threshold level than using “inflation” or “infure.” (a Japanese word equivalent to “inflation” in the U.S.). Second, threshold levels were calculated for 21 countries. Except for Switzerland and Japan, most developedcountries’ threshold levels were within the range of +2.5% to +3.5%, which were slightly above the central banks’ inflation target (2 percent or 1-3 percent). High correlations were found between the threshold levels and averageinflation rates during the sample period, not only among developed countries but also among developing countries with high inflation rates. Third, to check the robustness of the above estimation, threshold levels were estimatedby alternative data derived from the share of “don’t know” answers contained in household inflation expectations surveys. The resulting threshold levels were consistent with the Google Trends analysis. Lastly, the existence ofthe threshold levels has implications for monetary policy such as flattening Phillips curve during low inflation periods.

1. Introduction
The Japanese economy has experienced prolonged periods of low inflation. For example, the consumer price index (CPI), excluding fresh food and the effect of increased consumption tax rates, moved between -2 and +2percent. This trend suddenly disappeared at the beginning of 2021. The CPI’s rate of increase soared to +4.1 percent in January 2023 and remained at a relatively high level of +2.8 percent in August 2024. Although the recentsurge in inflation has caused problems in the Japanese economy, it offers new frontiers for analyzing the economy, such as inflation expectations (IEs). Although IEs play a vital role in the macro economy, their formationprocesses have not been fully elucidated. The Infotainment Research Center published two studies on the inflation expectations (IEs) of Japanese households: “Instability of Japanese households inflation expectations during thecurrent inflationary phase” (published in November 2023) and “Japanese households’ inflation perceptions: the formation process and their relationship with the inflation expectation” (published in February 2024). In June 2024,the center published a study on IEs in the corporate sector during the current inflationary phase. The objective of this report is to study household IEs, focusing on quantitatively analyzing the validity of the rational inattentionhypothesis (RIH). RIH is an economic model of inflation expectation formation and argues that “since people’s cognition ability is limited, it is rational to allocate cognitive resources to highly prioritized events and not to low-priority events.” Specifically, RIH implies that households are relatively insensitive to inflation rates below a certain threshold level and become increasingly attentive once they exceed the threshold level. Few studies haveconducted quantitative evaluations of RIH because there are few indicators representing households’ degree of attention to inflation. However, the aforementioned studies used the Google Trends index as a proxy and estimatedthresholds using a threshold model. This study follows their approach and applies it to Japanese households using updated data, including data from the current inflationary phase. Following three points were identified. First, dataderived from the Google Trends analysis of the keyword “inflation” were found to be a successful alternative indicator of households’ inflation attention level. The 2 threshold model indicates that the threshold levels is +3.0percent for the U.S. and +1.5 percent for Japan. In Japan, extending the sampling period to cover the current inflationary phase was necessary to estimate a robust threshold level because the Japanese economy suffered from along period of low inflation. In addition, keyword “price” in Japanese yielded much more stable threshold level estimation than using “inflation” or “infure.” (a Japanese word equivalent to “inflation” in the U.S.) Second,threshold levels were calculated for 21 countries. Switzerland (+1.0 percent) has the lowest threshold, followed by Japan (+1.5 percent). However, most developed countries’ threshold levels are within the range of +2.5% to +3.5%, which is slightly above the inflation target of central banks (2 percent or 1-3 percent). In addition, high correlations are found between the threshold levels and average inflation rates during the sample period, both amongdeveloped countries and among developing countries with high inflation rates, including Uruguay and Turkey. Third, to check the robustness of the above estimation, alternative threshold levels were calculated from data using theshare of “don’t know” answers contained in the household inflation expectations surveys. The resulting threshold levels were consistent with the Google Trends analysis. The remainder of the paper is organized as follows.Section 2 estimates the threshold levels for Japan and the U.S. using the Google Trends index as a proxy for household inflation attention. Section 3 estimates the threshold levels for the 21 countries and their relationship with theaverage inflation rates. Section 4 checks the robustness of the above estimation, using data from the household IE survey. Section 5 discusses the implications of the inflation attention threshold level for monetary policyimplementation. Finally, Section 6 presents the conclusions of this study.

2. Estimating threshold level of households’ inflation attention
2.1. Full information rational expectation hypothesis and the realityIEs play a significant role in household economic activities. For example, they affect the purchasing behavior of consumer goods and services, auto andhousing loan rates, and wage negotiations. IEs are also considered key factors in macroeconomic theory and are often assumed to satisfy full information perfect rationality (FIRE). Under the FIRE, people have full access to allthe information necessary for economic decisions and can make optimal choices based on this information. However, in reality, households’ IEs diverge significantly from professionals’ IEs and actual price development. Forexample, Japanese households’ one-year-ahead IEs are +10 percent (median figure), and +5 percent for five-year-ahead IEs2. These figures are much higher than the year-on-year increase in the CPI of +2.8 percent in August2024 or the peak value of +4.1 percent recorded during the current inflationary phase. These relationships indicate that households’ IEs do not follow the FIRE, as economic theory assumes. In addition, if households follow theFIRE, their IEs should converge to the same level, but they diverge significantly, ranging from 0 percent to more than 10 percent. The following three factors are pointed out as reasons why households do not follow FIRE. First,households lack the basic knowledge of and interest in inflation and macroeconomics, which are necessary to formulate reliable IEs. For example, in Japan, only 26 percent of surveyed households knew that the Bank of Japan hadset inflation targets at +2 percent. Similarly, only 32 percent answered that they had “interest in” the activity of the Bank of Japan3. Second, households’ price perceptions are significantly upward biased against the actual CPI.Their price perceptions are strongly influenced by the prices of frequently purchased goods such as food and gas. Such upwardly biased price perceptions naturally distort household IEs. Third, IEs differ significantly dependingon socioeconomic attributes such as gender, income level, academic background, and financial literacy. If households follow the FIRE, then these attributes should not affect their IEs. Overall, these points indicate that householdsare relatively inattentive to IEs and do not act according to the FIRE. Taking this point seriously, economists proposed new economic theories, such as the RIH and the sticky information hypothesis. As explained in Section 1,under RIH, household attention to inflation changes in a nonlinear or discontinuous manner as inflation rates increase. The following sections Data is based on “Opinion Survey” conducted by the Bank of Japan in June 2024.  

2.2. Empirical estimation of the RIH
Empirical studies on RIH are limited, particularly those that use quantitative analyses. However, the development of recent inflation rates offers favorable conditions for identifying the validity of RIH. In developed countries, aprolonged low-inflation period emerged after the global financial crisis from 2008 to 2009. Japan and other countries suffered from mild deflation at that time. According to RIH, households pay little attention to price developmentand fall into rational inattention in such circumstances. From 2021 to 2022, however, inflation rates began to significantly increase on a global scale owing to both increased demand and shortage in supply caused by Covid 19
and the invasion of Russia into Ukraine. If the RIH holds in such circumstances, households should change their attitude from the inattention phase to the increasing attention phase. The major problem in empirically analyzing theRIH is quantifying households’ attention to inflation. Recent studies suggested using the Google Trends’ keyword search of “inflation” as a proxy for inflation attention level6. Web searching can be considered an activeinformation-seeking activity compared with passively receiving information by watching TVs or reading newspapers. In addition to inflation attitudes, Google Trends keyword searches have contributed significantly to manyresearch fields such as automobile sales, unemployment, seriousness of Covid 19 infections. Figure 1 shows the results of Japanese households’ inflation attention as indicated by Google Trends. The Google Trends index rangesfrom 0 to 100, depending on the level of search activity. The sample period is from January 2004 to June 2024, and the peak value is marked in December 2022, coinciding with the highest CPI during the current inflationaryphase.  As for the keywords, instead of using “inflation” in Japanese “infureshon,” or “infure” (a shortened version), “price” in Japanese was used due to its superior performance as described in the latter section. Figure 2 is thesearch result of the word “inflation” in the U.S. The contrast between the low-inflation era and the current high-inflation phase is more striking than in Japan, as indicated in Figure 1, although regular fluctuations during the low-inflation period imply seasonal patterns.  Figure 3 shows the U.S. relation
ship between the Google Trends index on the vertical axis and the year-on-year increase in the CPI on the horizontal axis.  In Figure 3, a considerable number of dots are concentrated on the bottom-left side because the increasingrate of the CPI remained at 0 to 3 percent in the low inflation era. However, during the current inflationary phase, the dots moved toward the right upperside area. The relationship between these two variables was nonlinear, asindicated by a well-fitted exponential function7. This pattern indicates that households’ attention to inflation increases nonlinearly as the inflation rate increases. This observation is consistent with the RIH, which assumes thathouseholds move out of inflation inattention as inflation rates increase.

2.3. Estimating Thresholds level
This section empirically analyses the validity of RIH according to the estimation method described by Korenok et al. (2022) (Korenok). It uses a threshold model to statically verify that inflation attention ( Google Trends searchindex) increases nonlinearly as the inflation rate increases. Specifically, (1) when the inflation rate is low, households are inattentive to inflation and remain low responsive to an increase in the inflation rate; (2) once inflation ratesexceed a certain inflation level, which is called the threshold level, households’ inflation attention starts to increase as inflation rises. The threshold model presented below can quantitatively estimate the threshold level:
yt=α+β1xt (xt<γ)+β2xt (xt>γ)+et
where yt denotes search index at t period ranging from 0 to 100, xt represents year-on-year increase rate of CPI, (xt<γ) is a dummy variable which takes 1 when CPI increasing rate is less than the threshold level γ and 0otherwise, (xt>γ) is another dummy variable which takes 1 when CPI increasing rate is larger than the threshold level γ and 0 otherwise. β1and β2 are the coefficient of each dummy variable. Korenok requires (1) a nullhypothesis of β1＝0 should not be rejected, (2) a null hypothesis of β2＝0 should be rejected and take a positive value. As regards γ, the Korenok estimate optimal value by minimizing RMSE. This study instead searchesoptimal γ by inserting CPI by 0.5 percent point interval to maximize R2. Table 1 shows the estimation results for the threshold level for U.S. households.Table 1 shows that the threshold level for U.S. households is +3.0%, usingthe core CPI as an inflation indicator. Coefficients for +3.0 percent threshold level are β1=5.952 (t value = 28.104), β2＝0.454 (t value = 1.385）. Figure 5 plots both variables with the estimation results of the threshold model at+3.0 percent threshold. When the core CPI is under 3 percent, the trend line is nearly flat, whereas when the CPI exceeds 3 percent, the trend line shows a clear upward slope. As the RIH suggests, the threshold level divides
the low-attention phase from the increasing-attention phase. Korenok estimated the U.S. households’ threshold value as + 3.55 percent, which is 0.55 percent point higher than our estimates. This difference can be attributed totwo factors. First, Korenok used the headline CPI as inflation rates, whereas this study employed the core CPI. Second, the sample period for the Korenok ended in May 2022, whereas this study extends it to June 2024.
The same estimation method was applied to Japanese household data. At first, we estimated the threshold value using keyword of “inflation” in Japanese and “infure,” an abbreviated word for the “inflation.” The sample period isthe same as that for the U.S., from January 2004 to June 2024. The CPI excludes fresh food and is adjusted for consumption tax rate increases. Threshold level for “infure” was +1.5％. In case of “inflation,” in Japanese, the
threshold level was indetermined, since multiple peak R2 emerged at +0.5 percent, +2.0 percent and +3.0 percent. This could be related to the fact that the word “infure” is more often used than the “inflation.” As a next step,threshold level was re-estimated using more neutral keyword “price.” Although it differs in meaning compared to “inflation,” the motivation behind the reestimation was that the Japanese households experienced a prolonged lowinflation or even mild deflation during the significant portion of the sample period so that they might have searched for more neutral words such as “price” rather than “inflation.”
Table 3 above showed that the threshold level for “price” was +1.5 percent, same as the result for “infure.” Compared to “infure,” R2 jumped up from 0.395 to 0.515. Judging from the estimation results, Japanese households’inflation attention threshold level is estimated at approximately +1.5 percent. The +1.5 percent threshold level is lower than the +2 percent inflation target set by the Bank of Japan. Such a low threshold level could have beenaffected by a prolonged low inflation period. The next section discusses the relationship between the threshold level and the average inflation rate in more detail. Figure 6 indicated the dot diagram showing estimation result of thethreshold model based on the keyword “price” in Japanese. Compared to the U.S. estimate, R2 is lower, the dots are more dispersed, and the trend line has a slightly negative slope in the lowattention area. Therefore, although theJapanese estimation result satisfied Korenok’s condition, it was less textbook-like. The estimated equation for Figure 6 is as follows:
yt=33.01-1.81xt (xt<1.5)+10.79xt (xt>1.5)+et
           （-1.80）            （15.93）  
（）indicates t value.
Korenok estimated the threshold levels for 37 developed and developing countries. The average value of the threshold level for the 37 countries was + 2.09, much lower than that of the U.S. In addition, the study divides theestimation results of 37 countries into the following three categories: (1) consistent with the U.S., (2) intermediate, and (3) inconsistent with the U.S. Japan was categorized in “inconsistent  with the U.S.” and their estimatedthreshold level for Japan was 0.27, very low value compared to other countries. Switzerland also has a low threshold value of 0.36. Nonlinearities were not observed in either country. The study inferred from the above results thatboth countries’ relatively low peak inflation rates indicate that they never entered the high-attention mode during the sample period. To show that this inference was correct, the Japanese threshold model was reestimated using ashortened sample period of January 2004 to December 2021(Table 4). The results showed that although the peak R2 was marked at +1.5 percent, the R2 was quite low and the estimation results were unreliable.  

3. The relationship between the threshold level and the average inflation rates
The last section reveals that the inflation attention threshold levels differ among countries, such as +3.5 percent for the U.S. and +1.5 percent for Japan. This section estimates the threshold levels of the 21 countries using themethod described in the previous section (Table 5). Countries with undermined threshold levels were omitted from the table. The sample period was from January 2004 to May or June 2024, depending on the data availability. Thekeywords used for estimation were “inflation” translated to each country’s official language. In some countries, the Google Trends index shows unnatural responses such as a continuing zero value. In such case, “inflation” inEnglish was alternatively used. The estimation results indicate that Switzerland has the lowest threshold level (+1.0 %), followed by Japan (+1.5%). Most of the other developed countries’ threshold
levels lie within the range of +2.5% to +3.5%, which is slightly above the inflation target range of 2% or 1-3%, set by their central banks. The threshold level for developing countries such as Uruguay, Argentina, and Turkeyexceeded 5%, while the latter two countries showed double-digit figures. These three countries also have high inflation rates. The Korenok paper categorized Japan and Switzerland as “inconsistent with  the U.S.” However,extending the sample period to June 2024 enabled us to estimate stable threshold levels. These results imply that to derive stable results, the estimation periods should contain both low- and high-attention periods.
The country order indicated in Table 5 and their inflation conditions imply that the threshold level and the inflation rate are correlated, as Korenok points out. Figure 7 plots the variables for the developed countries in Table 5.
Figure 7 shows a relatively high R2 value of 0.795. Japan is located at the left end near the trend line. Switzerland’ threshold level was lower than Japan and less than their central bank’s inflation target “less than 2 percent.”
During the sample period, the Japanese CPI increase rate, adjusted for the consumption tax increase, exceeded the threshold level only twice: (1) just before the global financial crisis and (2) the current inflation phase, or just 12.5percent of the sample period (Figure 8).  Figure 9 shows the data from 21 countries, including high-inflation countries. A strong correlation between the threshold level and average inflation rates was observed,
even in these countries. This figure shows that the threshold level at which households turn to the high-attention mode is significantly affected by actual inflation rates.

4. Robustness checks
4.1. Alternative threshold level estimation using DK answers
The last section uses the Google Trends index as a proxy for household inflation attention. To check the robustness of the estimation in the previous section, we use other variables representing the degree of households’ attentionto inflation. As a robustness check, the Korenok paper relied on the X (former twitter), especially volume of using “inflation.” These results confirm the robustness of the Google Trends estimation results. This study relied on themethod used by Bracha and Tang (2022) (BT Paper). BT Paper used the share of respondents choosing “don’t know” answer (DT ratio) in household IE survey as a proxy for households’ degree of attention to inflation. TheDT ratio remains high when inflation rates are below the threshold level nd households are inattentive to inflation. However, once inflation rates exceed the threshold level and household attitudes change to the high attention mode,the DT ratio begins to fall. For the U.S. households, DT ratio is derived from the “Survey of Consumers” published by the Michigan University. The sample period and type of CPI were matched to the Google Trends estimation(Table 6). Although R2 was lower than that of the Google Trends index, the threshold level derived from the DK ratio was +3.0%, similar to the Google Trends estimation. This result confirms that the U.S. threshold level wasapproximately +3.0%. In case of Japan, although the “Opinion Survey” conducted by the Bank of Japan does not allow “don’t know” choice, Cabinet Office’s “Consumer Confidence Survey” has such option. Thresholdmodel was run by sample period from April 2004 to June 2024 and CPI was consumption tax rate increase adjusted base. Table 7 shows the estimation results. The estimated threshold level was +1.5%, matching the resultsobtained using the Google Trends index shown in Table 3. Therefore, the robustness of the estimation was confirmed and the threshold level for Japan was estimated to be approximately +1.5%. The central banks in bothcountries set their inflation targets at 2 percent. Interestingly, in Japan, households become highly attentive before the CPI reaches the policy target, while in the U.S., changes in attitude begin 1 percent above the inflation
target. If households trust the policy target, then the U.S. type of threshold, which is located above the policy target, seems natural.

4.2. Threshold level by households’ attribution
The Consumer Confidence Survey” publishes DK ratios by socioeconomic attributes, which enables us to estimate the threshold levels by different attributes (Table 8). For every class sorted by sex and income level, thethreshold levels were identical at +1.5 percent. According to the employment type, the threshold level for the unemployed was 0.5 percentage point higher than the average of +2.0 percent. According to household classification,the threshold level of single-person households was 0.5 percent point lower than the average of +1.0 percent. Based on the household head’s age classification, the estimation results were unstable. For example, the class age 20-29 was unable to estimate the threshold level owing to the small sample size, and the class age 40-49’s threshold level was relatively low at +0.5 percent. The threshold level for the classes of age 60-69 and 70 and over was +2.5percent, which was much higher than for other age classes. This could have been influenced by cohort effects, as people in these age categories experienced periods of high inflation in the early 1970s.  The estimation results inTable 8 are consistent with the initial estimate of the Japanese threshold level shown in Table 3, since more than half of the classifications showed a +1.5 percent threshold level.

5. Implication for the monetary policy
The above estimate shows that households’ attention to inflation has a threshold level and a nonlinear relationship with inflation. These findings have implications for monetary policy. First, when the inflation rate is low andhouseholds are in the low-attention mode, their response to inflation is low. This indicates that, in such circumstances, the slope of the Phillips curve becomes flatter. In fact, when developed countries faced continuously/ Third,central banks’ communication with the public should consider the existence of a threshold level. When the inflation rate is lower than the threshold level and central banks must introduce a non-traditional monetary policy,households’ responses to communication decline. For example, forward guidance and quantitative easing rely significantly on the IEs. When inflation rates move above the threshold level, effective central bank communication, inaddition to a policy interest rate hike, is necessary to reduce household IEs. Therefore, communicative policies should consider household attention to inflation rates. Second, when households are in the high-attention mode, suchas during the current inflationary phase, they become sensitive to inflation development. Even after inflation rates peak, IEs may remain high or take longer than usual to return to normal levels. In fact, Japanese households’ five-year-ahead IEs have remained at 5 percent for the past two years, even though CPI increase rates have been declining since the fourth quarter of 2022 (see Figure 10). In addition, in the high-attention mode, pass-through fromincreased inflation rate to wages and price-setting behavior also increases. low inflation after the global financial crisis, the flattened Phillips curve became widely debated.

6.Conclusion
This study analyzed the RIH, according to which households’ inflation attention becomes responsive to inflation rates once they exceed a threshold level. Recent studies started using the Google Trends index as a proxy forhouseholds’ degree of attention to inflation and succeeded in estimating the threshold level. Our study applied this method to Japanese households using updated data, including the current inflationary phase. The following threepoints were identified. First, data derived from the Google Trends analysis of the keyword “inflation” were a successful alternative indicator of households’ inflation attention level. The threshold model yielded the U.S. andJapanese household threshold levels of +3.0 percent and +1.5%, respectively. In Japan, extending the sampling period to cover the current inflationary phase was necessary to estimate a robust threshold level because the Japaneseeconomy suffered from a prolonged period of low inflation. In addition, keyword “price” in Japanese yielded much more stable threshold level estimation than using “inflation” or “infure.” (a Japanese word equivalent to“inflation” in the U.S.). Second, the threshold levels were calculated for 21 countries. Switzerland (+1.0 percent) had the lowest threshold, followed by Japan (+1.5 percent). However, most developed countries’ threshold levelswere within the range of +2.5% to +3.5%, which is slightly above the inflation target of central banks (2 percent or 1-3 percent). In addition, high correlations are found between the threshold levels and average inflation ratesduring the sample period, among developed countries and among developing countries with high inflation rates, including Uruguay and Turkey. Third, to check the robustness of the above estimation, threshold levels were
estimated by alternative data derived from the share of “don’t know” answers contained in household inflation expectations surveys. The resulting threshold levels were consistent with the Google Trends analysis. Households’threshold level for inflation attention also has the following implications for monetary policy: central banks should (1) be aware of the flattening Phillips curve during low inflation periods, (2) properly implement monetary policywhen households are in a high-attentive mode, and (3) consider households’ inflation attitudes when communicating with the general public.

(References)
Tomiyuki Kitamura and Masaki Tanaka (2019) “Firms inflation expectations under rational inattention and sticky information: An analysis with a small-scale macroeconomic model” Bank of Japan Working Paper Series No.19-E-  16, November 2019
Toshiyasu Fukuhara (2024) “Japanese households’ inflation perceptions: the formation process and their relationship with the inflation expectations” Infotainment Research Center, February 2024  
Bracha and Tang (2022) “Inflation levels and (in)attention,” Anat Bracha and Jenny Tang, Federal Reserve Bank of Boston Working Paper No,22-4, January 2022
Buelens (2023) “Googling “inflation”: What does internet search behavior revel about household (in) attention to inflation & monetary policy,” European Commission Discussion Paper 183, March 2023
Korenok et al. (2022) “Inflation and attention thresholds,” O. Krenok, D. Munro and J. Chen, GLO Discussion Paper No. 1175, Global Labor Organization, Essen
Sims (2003) “Implication of rational inattention,” Journal of Monetary Economics, Christopher A. Sims, Vol.3, Issue 50, April 2003




Japanese Firms' Inflation Expectations during the Current Inflationary Phase"
1. Introduction
The Japanese economy suffered from a prolonged period of low inflation. For example, the consumer price index (CPI), excluding fresh food and the effect of increased consumption tax rates, moved between -2 and +2 percent.This trend suddenly diverged at the beginning of 2021. For instance, the rate of increase of CPI soared to +4.1 percent in January 2023 and remained at a relatively high level of +2.6 percent in March 2024. Although the recentsurge in inflation has caused various problems in the Japanese economy, it has also offered new frontiers for analyzing Japan’s economy, such as breaking the persistent deflationary trend in the pre-pandemic period.
The Infotainment Research Center published two studies on the inflation expectations (IEs) of Japanese households: “Instability of Japanese households inflation expectations during the current inflationary phase” (published inNovember 2023) and “Japanese households’ inflation perceptions: the formation process and their relationship with the inflation expectation” (published in February 2024). These studies pointed out that: (1) Japanesehouseholds’ short-term IEs significantly exceeded the rate of increase of the CPI during the current inflationary phase; and (2) longer-term IEs also soared compared to those of U.S. and European households.
This study also focuses on IEs, albeit on those of the corporate sector. The main results are summarized as follows:
First, based on Tankan data, Japanese firms' IEs moved around 1 percent before the pandemic. This rose to 3 percent for one-year-ahead and about 2 percent for five-year-ahead expectations during the current inflationary period.Although larger firms showed lower IEs than smaller firms, no distinct differences were observed between the types of industries. Data on different economic sectors provided supporting evidence for the relationship “IEs ofeconomist and financial market participants < IEs of firms < IEs of households.” About the half of the firms answered “don’t know” about five-year-ahead IEs due to economic uncertainty; this share increased by additional 5 to10 percentage points after the pandemic.
Second, firms’ IE mechanisms do not completely follow either rational or adaptive expectations. Rational expectations do not hold because of: (1) a wide dispersion of IEs by firm size, (2) a clear upward bias of IEs compared tothe CPI, and (3) observed correlations between firms’ expected output prices and IEs. In addition, pure adaptive expectations are not supported because: (1) firms assign higher priority to IEs when actual inflation soars, and (2)IEs are affected by the movement in input costs.
“Rational inattention” (RI) theory became a useful framework for explaining such seemingly inconsistent firms’ behaviors. RI assumes that firms intentionally and rationally ignore information that has relatively small value for theiractivities. According to the RI theory, firms assign a higher priority to IEs in business operations when inflation rates or input costs rise significantly, whereas firms assign a lower priority to IEs if deflationary conditions last for asignificant period.
Third, IEs are not anchored at the 2 percent inflation target (a state of IE moving stably around the central bank’s inflation target). Even longer-term IEs, such as three- and five-year-ahead IEs, have clearly risen during the currentinflationary phase. Similar patterns are observed in the U.K., Italy, and other major developed countries.
Data and previous studies on this subject in and outside Japan are limited compared to those on household sectors, economists, and financial markets. Therefore, further research challenges include accumulating related data andanalyzing the formation mechanisms of IEs.
The remainder of this article is organized as follows. Section 2 summarizes the increased attention towards firms’ IE. Section 3 uses Tankan data to analyze firms’ IEs. Section 4 discusses the IE formation mechanism andwhether firms’ IEs are anchored around the inflation target. Finally, section 5 presents the conclusions of the study.

2. The purpose of analyzing firms’ IEs
2.1. Increased attention towards firms’ IEs
As mentioned in Section 1, studies on firms’ IEs in and outside Japan are limited. This is mainly due to the following two reasons.
First, data on firms’ IEs are limited. Existing problems in gathering data on firms’ IEs are difficulties in (1) increasing the sample size due to firms’ reluctance to participate, (2) making samples representative of the population interms of type of industry and size of firms due to sample size limitations, and (3) unifying respondents’ job positions.
Second, academics have focused their attention on the IEs of financial market participants and professional economists. This is because they seem to be better informed about the development of future inflation than ordinarypeople. Some studies assumed that every participant had perfect rational expectations and used the IEs of professional economists, even when analyzing the IEs of households and firms. However, in reality, as businesspersonshave valuable information on their firms’ input and output prices, they have their own perspectives on future inflation which are different from those of professional economists.
The tendency to ignore firms’ IEs began to change in the 2010s supported by the following reasons. First, major economic theories began to explicitly treat firms’ price-setting behaviors in their models. Second, firms’ uniqueprice-setting behaviors attracted economists’ attention when analyzing deflationary economies during the pre-pandemic period. Third, during the current inflationary period, when most developed countries had marked dual-digitinflation rates, economists were curious about how the surge in CPI would affect various types of IEs, including households, firms, and financial market participants. Moreover, central bank economists, in addition to academicsand professional economists, began paying attention to firms’ IEs, since the movements of IEs are one of the important factors in determining the effects of monetary tightening policy under high inflation.  
Reflecting such movements, since 2010s, central banks in major developed countries began surveying firms’ IEs (Table 1). The Bank of Japan added questionnaires on firms’ IEs as part of the Tankan survey that began in March2014. This effort was ahead of most major developed　countries central banks, which started their surveys in the late 2010s or during 2020s.

2.2. The relationship between firms’ IEs and the macroeconomy
Before proceeding to the data analysis, this section discusses how firms’ IEs affect the macro economy.
First, firms’ IEs affect the price-setting behaviors of products and services. Final output prices are also affected by the demand and supply conditions in the market. However, as the current inflationary phase shows, firms’ price-setting behaviors played a larger role than usual due to the cost-push inflation caused by price surges in raw materials, depreciating yen, and weakening global supply chains under the influence of the U.S.?China trade conflict.
Second, IEs significantly affect wage-setting behaviors. A typical case was the U.S. economy in the 1970s, when surging inflation increased IEs and caused an upward wage-price spiral. Once the economy falls into such acondition, containing inflation requires a prolonged tight monetary policy and enduring high unemployment rates, thereby significantly damaging the economy.
Third, firms’ IEs influence real interest rates. Even if the nominal interest rates are constant, the real interest rates for individual firms are affected by their IEs. Because real interest rates are an important factor in deciding fixedinvestments, they in turn affect the business cycle.
Fourth, IEs affect monetary policy efficiency. Firms’ IEs are anchored when their long-term IEs move close to the inflation target and are not significantly affected by changes in the actual inflation rate. If this holds, an upwardwage-price spiral is unlikely to occur and high inflation rates will swiftly converge to the inflation target (whether Japanese firms’ IEs are anchored is discussed in Section 4.4).  


3. Detailed analysis of firms’ IEs
3.1. Japanese firms’ IEs
This section analyzes firms’ IEs using Tankan data. Tankan is the only quarterly survey conducted in Japan on firms’ IEs. The Bank of Japan conducts the survey, covering more than 10,000 firms located in Japan. The surveyaskes firms about Diffusion Index (D.I.), such as business conditions, and supply and demand conditions, as well as figures for sales and current profits. It is considered important survey data in Japan to judge businessconditions. In 2014, questions about expected inflation (one-, three-, and five-years ahead) and output prices (one-, three-, and five-years ahead) (Figures 1 and 2) were added.
Japanese firms’ IEs moved stably at approximately 1 percent for every term during the pre-pandemic period (see Figure 2). They remained positive even when the CPI fell into negative territory in 2020. Since 2021, IEs begansurging and peaked at 3 percent for one-year-ahead IEs and approximately 2 percent for five-year-ahead IEs. One-year-ahead IEs continued to decline moderately after March 2023, whereas five-year-ahead IEs remained aroundthe peak level. Meanwhile, the three-year-ahead IEs moved closer to the five-year-ahead IEs.

3.2. Relationship between IEs and output prices
In addition to general prices, Tankan asks about firms’ outprice expectations (Figure 2). The one-year-ahead expectations in both figures exhibit similar trends. Expected output prices fell into negative territory in 2020, similar tothe CPI, whereas IEs did not.
Tankan’s figures for the three- and five-year-ahead expected output prices are compiled as the cumulative change rates relative to the current level. Figure 3 modifies the released figures and converts them into annualized changerates from one to five years ahead.  
Only the one-year-ahead expected output prices are affected by actual prices. Annualized changes in expected prices for one-to-three years and three-to five years moved stably at approximately 0.5 percent, despite significantchanges in the CPI. Looking closely, in the pre-pandemic period, three-to-five years ahead expectations fell to 0.2 to 0.3 percent. In the recent period, one-to three-year expectations showed slight increases. Overall, whileJapanese firms expected a 1 to 2 percent increase in general prices, they assumed a modest 0.5 percent annual increase in their output prices.
Outside Japan, expectations of output prices are also available in the survey conducted by the central bank of Italy (Figure 4). The survey began in 2013, a year earlier than in Tankan. Figure 4 shows that expectations of outputprices and general prices moved very closely. One difference from the Tankan survey results is that Italian IEs fell significantly faster than their output price expectations during the current inflationary phase.

3.3. The relationship between the firms’ inflation perceptions and the IEs
Households' IEs are significantly influenced by how they perceive current inflation rates, or inflation perceptions. Although Tankan does not cover inflation perceptions, surveys conducted in the U.S., U.K., and France do.Although these surveys commonly suffer from relatively short sample periods, the inflation perception of firms moved closely with the CPIs, except in the U.K., where perceptions exceeded the CPI. Since household inflationperceptions are significantly above the actual CPI in most developed countries (Japanese households shown in Figure 8), this suggests that firms and households have different IEs formation mechanisms.

3.4. IEs by the size of firms
Tankan’s IE survey has a unique advantage over other surveys because of its large sample size, which exceeds 10,000 firms. This enables a detailed analysis of IEs by firm size and industry type. This section analyzes the firms’IEs according to their size.
Figure 9 clearly shows that levels of IEs have been stable at “large firms <medium sized firms <small firms” for the whole sample period. Since larger firms have relatively more information on general prices and macroeconomicconditions than smaller firms, larger firms’ upward bias of IEs relative to the CPI is clearly less than that of smaller firms.  
Figure 10 shows the expected output prices according to firm size. The output prices of large and medium-sized companies moved similarly, while those of small-sized firms have been higher than those of the other two, especiallyduring the current inflationary phase. Because the IEs of small firms are higher than those of large firms, as shown in Figure 9, small firms seem to expect a larger increase in their output prices in the future.  
The Bank of Italy conducts a survey of IEs depending on the size of the firms (Figure 11). During the current inflationary phase, although the IEs of small firms exceeded those of larger firms, the difference seems smaller thanthose from Tankan. One of the main reasons behind this phenomenon is its unique survey method, in which information on the current inflation rate is given to respondents just before asking their IEs.
Economists criticize this unique survey method because it significantly reduces the upward bias of IEs relative to the CPI. In fact, if we compare Italian IEs (Figure 12) with those in Tankan (Figure 1), the Italian survey’s gapsbetween long- and short-term IEs are unnaturally smaller than those in Tankan, especially during the pre-pandemic period.

3.5. IEs by type of industries
The Tankan data can be broken down by industry type. Figure 13 shows the IEs of the manufacturing and nonmanufacturing sectors. The difference between the two sectors is only 0.1 percent point, which is relatively smallcompared to the 0.6 percentage point difference in IEs between large and small firms.
Figure 14 shows the standard deviations (SDs) of 36 industries (19 manufacturing and 17 non-manufacturing industries). In this graph, an increase in the SDs implies that the IEs of different types of industries are more widelydispersed.
Although the SD level was relatively stable during the pre-pandemic period, it began to surge in tandem with IEs during the current inflationary phase. The SDs peaked in the middle of 2022 and fell significantly afterwards, suchthat the level of SD in March 2024 was almost equal to that at the beginning of 2021. On the other hand, IEs peaked at the beginning of 2023 but remained at 2.5 percent even in March 2024. Consequently, the gap between IEsand SDs has widened. It seems that at the beginning of 2021, the SDs were pushed up significantly by huge uncertainties created by the pandemic and then peaked and fell sharply as the pandemic was gradually contained anduncertainty diminished.
Figure 15 plots the SD for one, three, and five years ahead for the Tankan firms. This clearly indicates that the sharp increase in the SDs was limited to expectations one year ahead, which showed synchronized movement with asurge in actual inflation and IEs.
No survey outside Japan provides detailed IEs by industry type. However, a survey conducted by the Federal Reserve Bank of Cleveland revealed SDs and IEs (Figure 16). Similar to Tankan, the SDs increased significantlyduring the current inflationary phase.

3.6. Comparing IEs with other economic entities
This section compares firms’ IEs with those of other economic entities, such as households and economists. In case of Japan, households’ IEs are available using the “Opinion Survey” conducted by the Bank of Japan. Figure17 shows the IEs of firms and households, along with the CPI. This clearly indicates that the IEs of firms are consistently lower than those of households.
The figure also shows that (1) the IEs of firms comoved with those of households and CPI, (2) during the current inflationary phase, the increasing pace of households’ IEs exceeded those of firms; therefore, the peak rate forhouseholds reached 10%, while those of firms remained at 3 %, and (3) during the current inflationary phase, the peak period of IEs of firms was several quarters earlier than that of households.
The Bank of Japan also published a similar graph showing various IEs in the “Outlook for economic activities and prices” (Figure 18). According to the figure, firms’ IEs moved consistently above those of professionals,including economists, financial market participants, and households, especially during the current inflationary phase (shapes of IEs in Figure 18 differ from Figure 17 because the former plots five-year-ahead IEs, while the lattershows those of one-year-ahead).
Although both Figures 17 and 18 depend on the same “Opinion Survey” data for households IEs, the significant differences in the level of IEs were caused by the data themselves and the calculation methods. The survey askedthe respondents to respond to their IEs in two ways. The first directly asked about IEs qualitatively, and the second asked respondents to select from five choices depending on the magnitude and direction of IEs. Figure 17 usesthe median figure obtained from the quantitative answers, whereas figure 18 depends on a five-choice answer and uses an econometric method to quantify the responses. The Bank of Japan explained the reason for choosingqualitative answer rather than quantitative ones by suggesting the facts that quantitative answers contain many “bias” such as answers concentrated on integers, multiplier of 5, and 0 percent. The Bank concluded that using themedian figure of quantitative answers could not precisely represent the complex distribution of household IEs.
However, the unique response patterns of qualitative answers are common for household IEs, not only in Japan but also in many major developed countries. Such a pattern is considered to be closely related to psychologicalfactors and frequently occurs when respondents are not confident in their answers. Therefore, it is unlikely that a unique pattern emerged because of technical problems, such as survey methods or respondents’ irresponsibility.Because respondents are not confident about their responses, it is hopeless to improve the quality of the outcome of the five-choice responses by mathematically quantifying the data. This mathematical modification seems to havecreated an unrealistic situation, as shown in figure 18, where households’ IE was as low as that of professionals.
Other countries offer data to compare firms’ IEs to those of other economic entities. First, the Bank of Canada publishes the IEs of households, firms, and economists (Figure 19). This indicates that the level of firms’ IEs islocated between those of households and professionals and that the IEs of households surged in the current inflation phase, similar to Japan in Figure 17.
Second, in the U.S., firms’ IEs are also located between households and economists, similar to Japan and Canada. However, in the current inflationary phase, firms’ IEs are closer to those of households than those ofeconomists (Figure 20).
Third, in Norway, firms’ IEs also lie between economists and households, except in the current inflationary phase(Figure 21). The peak rate of IEs of firms is 6 percent and that of households is approximately 4.5 percent.
Fourth, the Bank of France published a research paper in 2021 that presented a table indicating that firms’ IEs lie between those of households and economists (Figure 22).
Other studies also state that firms’ IEs lie between those of households and firms. In addition, a study published in 2022 in the U.S. pointed out that firms’ IE formation mechanism differs from those of households andeconomists. Another study, published in the U.S. by the Federal Reserve Bank of Cleveland, calculated the SDs of firms’ IEs and found that they were much smaller than those of households.
This evidence suggests that the in terms of IEs, the relationship “economists<firms<households” holds. Thus, the Bank of Japan’s chart (Figure 18) seems to be an outlier.

4. Formation mechanism of firms’ IEs
4.1. How confident are firm in responding to IEs?
This section describes the formation mechanisms of firms’ IEs.
First, this subsection uses Tankan data to examine how confident are firms in answering about their IEs. Tankan’s questionaries allow a “don’t know” answer if firms are unable to come up with numerical IEs. Figure 22 plots theshare of “don’t’ know” responses. This share increases as the duration of the expectation periods becomes longer. For example, before the pandemic, the share was approximately 15 percent for one-year-ahead, jumped to 30percent for three-year-ahead, and reached approximately half of respondents for five-year-ahead. Understandably, the longer the expectation period, the more difficult it is to provide a quantified answer about IEs owing toincreased uncertainties. Still, it is remarkable that almost half of the firms have no view about five-year-ahead. In addition, the share of “don’t know” has increased an additional 5 to 10 percentage point since the first half of 2020for every expectation period. At the time of writing this article in March 2024, the share of three- and five-year-ahead remained high. Tankan further asks the reasons for choosing the “don’t know” answer. Figure 23 showed thatmost respondents picked “uncertainty.” Judging by the timing, the outbreak of the pandemic may have caused the increased uncertainty among firms. Next, Figure 24 depicts the share of “don’t know” by manufacturing and non-manufacturing sectors. Before the pandemic, the manufacturing sector’s share was higher than that of the non-manufacturing sector. This might reflect the fact that (1) the demand and supply conditions of manufacturing goodsfluctuate more widely than those of service goods and (2) the manufacturing sector faces higher uncertainty due to fluctuations in overseas demand and volatile foreign exchange rates.
Figure 25 shows the share of “don’t know” answer by the firm size. The share of large firms is approximately 10 percent point higher than that of SMEs. This is odd because large firms generally seem to have a relative advantagein gathering macroeconomic information and developing detailed long-term business plans.

4.2. Business operations and IEs
Using prior studies, this subsection examines how IEs affect firms’ business operations. A U.S. study asked firms about how much general prices affected their business operations (Figure 26). The figure plots the responses fortwo different periods: 2015, when the inflation rate was modest, and 2022, when the inflation rate increased rapidly during the pandemic. The 2015 responses showed that almost 90 percent of firms chose low effect responseanswers such as “no effects” or “medium effects.” In contrast, the responses of 2022 shifted rightward. The share of “no effects” declined while that of “strong effects” increased. This indicates that the information value ofgeneral prices for business operations varies according to the level of inflation rate.
The same study asked firms about the effects of general prices on product prices (Figure 27). Surprisingly, approximately half of the surveyed firms answered “no effect.” Thus, input prices, rather than general prices, are themajor factors determining product prices.

4.3. Formation process of firms’ IEs
This subsection uses Tankan and other overseas surveys to analyze the formation process of firms’ IEs.
The following three pieces of evidence suggest that firms are not perfectly rational when forming IEs. First, although rational firms should form identical IEs based on the same information, Figure 9 using Tankan data indicates thatthe level of IEs diverges with firm size. Second, the level of firms’ IEs is consistently higher than the actual CPI or professional forecasts both inside and outside Japan (Figures 17, 19, 20, and 21). Such upward bias frequentlyoccurs when entities’ inflation perceptions, which provide the basis for IEs, are inaccurate. Third, IEs by industry type in the Tankan survey are significantly affected by their own output prices (R2=0.65). If the firms are perfectlyrational, the two variables should show no correlation (see Figure 28).
These observations contradict rational expectation theory, where economic entities grasp all economic conditions perfectly and decide on IEs with full information.
Since firms are not perfectly rational, are they following adaptive expectations theory, where IEs are calculated mechanically depending only on past and present data on inflation rates? This is not the case, since Figure 26indicates that firms do care about IEs when they face large shocks, such as pandemics or inflation surges. In addition, the Tankan data suggest that fluctuations in firms’ input costs affect their IEs (R2=0.72), which contradictsthe adaptive expectation theory where only past and present inflation rates matter (Figure 29). Another survey conducted in New Zealand in 2013 found that firms set priorities in monitoring macroeconomic data. For example, theshare of firms closely monitoring GDP reached approximately 80 percent, whereas that of inflation rates was about 50 percent. The survey also pointed out that: (1) firms’ forecast errors for highly prioritized GDP were smallerthan those for inflation rates, (2) firms belonging to highly competitive industries formed accurate IEs, and (3) firms generally collected abundant price information about their own industries. Such observations also imply thatfirms do not form IEs through simple calculations, as adaptive expectations theory suggests.
The RI theory is a useful framework to explain such seemingly inconsistent firms’ behaviors where firms are not perfectly rational nor simply adaptive. RI assumes that firms are generally rational and motivated to efficientlyallocate limited resources. Hence, they intentionally neglect information that has a relatively small impact on their activities and requires non-negligible costs to make accurate estimations. Thus, RI’s basic idea is that firms arerationally inattentive to low-value information.
According to the RI, firms assign priorities among macroeconomic indicators according to their impact on business activities at the time; hence, priorities can vary. For example, the priority of IEs increases when an inflation surgeor a pandemic occurs. However, the long-term deflationary period seen in the pre-pandemic period lowered the priority of IEs. This priority also increases when competitiveness within the industry becomes fierce and fluctuationsin input costs increase.
Another study showed that firms in high-inflation countries, such as Uruguay, Ukraine, and Argentina, place higher priority on inflation rates and analyze IEs more carefully than firms in low-inflation countries, such as the U.S. andNew Zealand. Another study conducted in the U.S. reported that firms expected a prolonged inflation surge after the pandemic earlier than economists. This is because economists regarded the inflation surge as a temporalphenomenon due to supply shortages; meanwhile, firms were aware of the ongoing increasing vulnerability in global supply chains, and therefore, expected prolonged inflation surges.

4.4　Unanchored firms’ IEs
Anchored IEs imply that long-term IEs move stably around the central bank’s inflation target (often set at 2 percent) without being significantly affected by the fluctuations in short-term IEs. During the pre-pandemic period, sinceinflation rates moved rather stably at low levels in many developed countries, IEs were also relatively stable. Studies focusing on this period concluded that IEs were anchored. The situation changed dramatically after the currentinflationary phase, where short-term IEs notably responded to the inflation surge. Furthermore, the long-term IEs, which should be stable regardless of the movements in short-term IEs, moved upward.
This “de-anchoring” of long-term IEs is evident in Japan (Figure 1), Italy (Figure 12), the U.K. (Figure 7), and Canada (figure 19). Most studies also reported that firms’ IEs are not anchored.
First, only 25 percent of firms were aware of a 2 percent inflation target in the U.S. Meanwhile, only one-third knew the target in New Zealand, where the central bank has continued its inflation targeting policy for 25 years.
Lastly, since Candia et al. (2021) pointed out five conditions for anchored IEs to meet, the question of whether Japanese firms’ IEs are anchored is examined below based on these conditions:
(1) The average IEs should be close to the inflation target: In Japan, the target is 2 percent. Long-term IEs hovered around 1 percent before the pandemic and rose to 2 percent only recently.
(2) Fluctuations in IEs should be relatively small: The SD of Tankan firms is shown in Figure 15. The SD of one-year-ahead IEs significantly increased during the current inflation phase. An overseas study also showed that the SDof IEs was much larger than that of professionals.)
(3) Firms should have confidence in their IEs: As Figure 22 shows, approximately half of Tankan firms chose “don’t know.” This suggests that firms are not confident in their forecasts.
(4) IEs, especially those in the long term, should be stable: Figure 1 indicates that the five-year-ahead IEs shifted substantially from the 1 percent level prevalent in the pre-pandemic period to a 2 percent level in the currentinflationary phase.
(5)  Long- and short-term IEs should not exhibit co-movement: Plotting the first difference of long/short IEs showed significant correlations (R2 = 0.77, Figure 30), indicating that both variables exhibit co-movement.
Thus, Japanese firms’IEs are not anchored.

5. Conclusion
This study analyzed firms’ IE and their formation process during the current inflation phase. The main results are summarized below.
First, analyzing Tankan data revealed that Japanese firms' IEs moved around 1 percent before the pandemic. This rose to 3 percent for one-year-ahead and about 2 percent for five-year-ahead expectations during the currentinflationary period. Although larger firms showed lower IEs than smaller firms, no distinct difference was observed between the types of industries. Data from different economic sectors showed that firms' IEs lie between those ofprofessionals and households. About the half of firms answered “don’t know” for five-year-ahead IEs due to economic uncertainty; this share increased an additional 5 to 10 percentage points after the pandemic.
Second, firms’ IEs mechanisms do not follow a completely rational or adaptive expectations model. Rational expectations do not hold because of: (1) a wide dispersion of IEs by firm size, (2) a clear upward bias of IEscompared to the CPI, and (3) observed correlations between firms’ expected output prices and IEs. Firms do not follow pure adaptive expectations either because (1) they assign higher priority to IEs when actual inflation soarsand (2) IEs are affected by the movement in input costs.
RI theory became a useful framework for explaining such seemingly inconsistent firms’ behaviors. RI assumes that firms intentionally and rationally ignore information that has relatively small value for their activities. According tothe RI, firms assign a higher priority to IEs when inflation rates or input costs rise significantly, whereas they assign a lower priority to IEs if deflationary conditions last for a significant period.
Third, IEs are not anchored at the 2 percent inflation target (a state of IE moving stably around the central bank’s inflation target). Even longer-term IEs, such as three- and five-year-ahead IEs, clearly rose during the currentinflationary phase. Similar patterns are observed in the U.K., Italy, and other major developed countries.
Data and previous studies on this subject in and outside Japan are limited compared to those of households, professionals, and financial markets. This trend gradually began to change in the 2010s, as firms’ pricing behaviorduring the low-inflation period attracted economists’ interest. Accordingly, much data on firms’IEs, including those in Japan, was now being collected, albeit mainly by central banks.
Finally, studies on firms’ IEs have several implications for central bank policy communication. First, central banks need to consider how they can reach medium and small firms whose IEs contain a larger upward bias than largefirms. Second, methods or tools to promote firms’ interest in IEs during the low-inflation period, when firms follow RI and the priority of inflation is lowered, should be examined. Third, central banks should explore how theycan establish repetitive communication tools for firms that can help sustain the effects of policy communication.  

References
Albrizio et al. (2023) “Mining the gap: Extracting firms’ inflation expectations from earnings calls,” S. Albrizio, A. Diziori, P. Simon, IMF-WPs 202/23, October 2023
Bank of Japan (2016) “Firms' inflation expectations and wage-setting behaviors” BOJ-WP series, No. 16-E-10, July 2016
--- (2017) “New facts about firms’ inflation expectation formation: Part 1 revisiting sticky information model,” Y. Uno, H. Naganuma, H. Hara, Bank of Japan WP series No. 17-J-3 March 2017 (in Japanese)
--- (2018) “New facts about firms’ inflation expectations: Simple tests for sticky information model,” Y. Uno, S. Naganuma, N. Hara, Bank of Japan WP series, No. 18-E-14, October 2018
--- (2019) “The formation of Japanese firms’ inflation expectations: A survey data analysis,” H. Inatsugu, T. Kitamura, T. Matsuda, BOJ WP series, No. 19-E-15, November 2019
Bouche et al. (2021) “Measuring firms’ inflation expectations” P. Bouche, M. Gerardin, E. Gautier, E. Savignac, Bank of France Bulletin, August 2021
Bryan et al. (2015) “The inflation expectations of firms: What do they look like, Are they accurate, and Do they matter?” M. Bryan, B. Meyer, N. Parker, Federal Reserve Bank of Atlanta Working Paper Series 2014-27a
Candia et al. (2021) “The inflation expectations of U.S. firms: Evidence from a new survey,” B. Candia, O. Coiboin, Y. Gorodnichenko, NBER WP 28836, May 2021
--- (2022) “Macroeconomic expectations of firms,” B. Candia, O. Coiboin,
Y. Gorodnichenko, NBER WP 30042, May 2022
Coibion et al. (2015) “How do firms form their expectations? New survey evidence,” O. Coibion, Y. Gorodnichenko, S. Kamer, NBER WP No. 21092, April 2015
Fukuhara (2024) “Japanese households’ inflation perceptions: the formation process and their relationship with inflation expectations,” T. Fukuhara, Infotainment Research Center, February 2024
Garciga et al. (2023) “The survey of firm’s inflation expectations,” C. Garciga, E. Knotek II, M. Pedemonte, T. Schiroff, Economic Commentary, Federal Reserve Bank of Cleveland, May 2023
Meyer and Sheng (2021) “Unit Cost Expectations and Uncertainty: Firms’ perspective on inflation,” B. Meyer, X. Sheng, Federal Reserve Bank of Atlanta WP Series 2021-12b, Revised March 2024
Richards and Verstraete (2016), “Understanding firms’ inflation expectations using the Bank of Canada’s Business Outlook Survey,” S. Richards, M. Verstraete, Bank of Canada WP No. 2016-7, February 2016
Savagnac et al. (2021) “Firms’ inflation expectations; New evidence from France,”
F. Savagnac, E. Gautier, Y. Gorodnichenko, O. Coibion, NBER WP No. 29376, October 2021
Weber et al. (2022) “The subjective inflation expectations of households and firms: Measurement, determinants and implications,” M. Weber, F. D’Acunto, Y. Gorodnichenko, O. Coibion, NBER WP No. 30046, May 2022



Japanese Households' Inflation Perceptions: The Formation Process and Their Relationship with Inflation Expectations
This study examined the formation process of household inflation perceptions as well as their relationship with inflation expectations and implications for monetary policy implementation. Inflation perceptions attracted relativelylittle attention until recently because they had constantly exceeded the actual CPI and were considered unreliable. This attitude has changed as inflation perceptions became considered the basis for forming inflation expectations.At the same time, the following reasons are offered to account for the inflation perceptions constantly exceeding the CPI. The first is a lack of households’ basic knowledge regarding the CPI. The second is people’s tendencytoward loss aversion. The third is households’ lack of awareness of the CPI’s quality adjustments. Regarding the relationship between inflation perceptions and expectations, the following evidence was found. First,socioeconomic attributes of consumers affect both inflation perceptions and expectations in the same way. Second, according to household surveys, households heavily rely on inflation perceptions when responding to inflationexpectations. Third, crosssectional analysis using individual sample data shows strong correlations between the inflation expectations and perceptions. Such features of inflation perceptions have monetary policy implications.First, although food and gas prices are excluded from the “core-core CPI,” they should be monitored closely. Second, differences in inflation expectations caused by socio-economic attributes distort efficient resource allocation.Third, households’ financial literacy should be improved. Future challenges in this field include further data accumulation, which enables detailed studies on the formation process of inflation perceptions

1. Introduction
The Japanese economy has experienced a long period of low inflation. During this period, the Consumer Price Index (CPI), excluding perishables and the effects of increases in consumption tax, has hovered around the minus-2-to-plus-2-percent range. This low-inflation phase abruptly ended at the beginning of 2021, as the corporate sector began to pass on the jump in import goods prices to the prices of its own products. In January 2023, the CPI peakedat + 4.1 percent on a year-on-year basis and remained high at +2.8 percent as of October 2023 (Figure 1). Such increases in the inflation rate have caused various economic problems in Japanwhile providing a unique opportunityto analyze households’ inflation perceptions and expectations. The Infotainment Research Center published a study in December 2023 titled “Instability of Japanese Households’ Inflation Expectations During the CurrentInflationary Phase” (December Study). The study highlighted two findings. First, Japanese households’ inflation expectations significantly overshot the CPI increase rates (CPI: +2.9 percent; one-year-ahead inflation expectation: +10 percent; five-years-ahead inflation expectations: +5 percent).  Second, even though the U.S. and European CPI increased to around +10 percent, their one-year-ahead expectations remained at a maximum of +5 to 6 percent,and the five-years-ahead inflation stayed at +3 to 4 percent. The study concludes from the above evidence that Japanese households’ inflation expectations are more unstable than those in the U.S. and Europe.The DecemberStudy also showed that the movement of inflation expectations has been closely related to inflation perception, which is households’ view of current price development. Based on the December Study, this study examines theformation process of inflation perceptions. Simultaneously, the relationship between inflation perceptions and expectations and the implications for monetary policy implementation are discussed. The main results of this studyshowed that inflation perceptions have attracted relatively little attention because they constantly exceed the actual CPI and are considered unreliable. However, their information value has recently become highly appreciated, asinflation perceptions have been considered the basis for forming inflation expectations. In this process, the following explanations are offered for inflation perceptions constantly exceeding the CPI. the first is the lack of basichousehold knowledge of the CPI. Households rely on the prices of frequently purchased goods such as food and gas as the most reliable sources of inflation perceptions. The synthesized CPI, consisting of food and gas prices,fits more closely with inflation perceptions than the overall CPI.Second, people have a tendency toward loss aversion. People are said to be 2.5 times more sensitive to price increases, which lead to a loss in purchasing power,than price decreases. Therefore, price increases cause significant damage to consumer psychology and boost the perception of inflation above the actual CPI.Third, there is a lack of awareness of households’ CPI qualityadjustments. Although such adjustments reduce the price index, consumers simply ignore them, leading to a wider divergence between the CPI and inflation perceptions.Regarding the relationship between inflation perceptions andexpectations, the following evidence was found. First, both inflation expectations and perceptions are affected in the same way by differences in consumers’ social attributes, such as gender, income level, and academicbackground. Second, according to household surveys, households rely heavily on inflation perceptions when responding to inflation xpectations. Third, a cross-sectional analysis using individual data revealed strong correlationsbetween inflation expectations and perceptions.These findings have implications for monetary policy implementation. First, although food and gas prices are excluded from the “core-core CPI,” their developments should bemonitored closely as they are tightly related to the inflation perceptions. Second, differences in inflation expectations due to socioeconomic attributes also affect real interest rates and resource allocation. Third, households’financial literacy should be improved as financially literate persons actively gather macroeconomic information,which, in turn, narrows the divergence between the CPI and inflation perceptions.Because research on inflationperceptions remains scarce, future challenges in this field include further data accumulation, which enables studies on the formation process of inflation perceptions and their relationship with inflation expectations.The remainderof this paper is organized as follows. Section 2 presents the relationship between perceptions of inflation and expectations. Section 3 highlights the factors contributing to higher inflation perceptions than the CPI, includinghouseholds’lack of knowledge regarding the CPI, reliance on prices of frequently purchased goods, and the effect of CPI quality adjustments. Section 4 provides evidence of the close relationship between inflation expectationsand perceptions. Section 5 discusses the implications of monetary policy implementation. Finally, section 6 concludes the study.

2. Inflation Perceptions Exceed Changes in CPI
2.1. Inflation Perceptions
Figure 2 shows the relationship between inflation perceptions and the CPI in Japan. The features of the inflation perceptions are twofold. First, they consistently exceeded the CPI. The divergence amounts to approximately 2?4percent, even before the current inflationary period. During the current inflationary phase, even though the peak CPI rate remained at +4.1 percent, inflation perceptions reached 10 percent. Second, they closely followed CPImovement. Third, even when the CPI decreased to negative values, inflation perceptions remained at zero percent. Inflation perceptions exceed the CPI not only in Japan but also in other major developed countries, a phenomenonoften called the “inflation perception conundrum”. Figure 3 shows the Eurozone case. Although inflation perceptions closely follow CPImovement, their level is 5 to 10 percent higher than that of the CPI. Particularly during thecurrent inflationary phase, inflation perceptions reached 20%, whereas the CPI remained at approximately 10 percent.

2.2. The Relationship Between Inflation Perceptions and Expectations
Figure 4 illustrates the relationship between inflation perceptions and expectations in Japan. It illustrates several features. First, both variables move closely. Second, inflation expectations are more stable than inflation perceptions.Third, during the current inflation phase, inflation expectations increased almost as much as inflation perceptions. Fourth, inflation expectations seem to lead perception by three months.Figure 5 illustrates the situation in U.K.households. It also exhibited several features. First, inflation perceptions and expectations behave similarly. Second, except for the current inflationary phase, the levels of inflation perceptions and expectations are very close toeach other. Third, unlike Japan, during the current inflation phase, inflation expectations remained below 5 percent, while perceptions increased to approximately 10 percent. Fourth, inflation expectations lead perception to acertain degree.Next, Figure 6 shows the case of Eurozone households and shows four features. First, similar to households in Japan and the U.K., inflation perceptions and expectations behave similarly. Second, before the currentinflation phase, the gap between expectations and perceptions was wider than in Japan and the U.K. Third, during the current inflation hase, the gap widened to more than 10 percent point. Fourth, inflation expectations seem tolead perceptions, as in the cases of Japan and the U.K.Based on the above three examples, three conclusions can be drawn. First, the two variables moved closely together and followed the movement of the CPI. Second, inflationexpectations show more moderate movements than do perceptions. Third, inflation expectations seem to lead perceptions to a certain degree. As the December Study showed, Japanese households’ inflation expectations aremore unstable than those of the U.K. and the Eurozone.Until recently, most economists in developed countries rarely paid attention to households’ inflation perceptions as they are too high to “justify” or “impractical to use ineconomic analysis.” Therefore, studies on the gap between the CPI and inflation perceptions are scarce. However, calculating the statistical relationship between inflation expectations and CPI/inflation perceptions revealed thatinflation perceptions are much more closely correlated with expectations than the CPI. In Japan, the correlation between inflation perceptions and one-year-ahead inflation expectations is higher (0.84) than that between the CPI andexpectations (0.77). Inflation expectations, on the other hand, are regarded as one of the important macroeconomic variables. Professional forecasts of inflation expectations, such as surveys of economists and people engaged infinancial markets, or inflation expectations derived from yields on inflation-linked bonds are often used as typical inflation expectations rather than those of households. When the inflation expectations are adapted inmacroeconomic models, they often assume rational expectations (people act according to the assumption that they know all of the information about the macroeconomic variables) or adaptive expectations (people rely on pastinflation rates as a proxy for future inflation rates). Nonetheless, as shown in Figures 4-6, inflation perceptions follow CPI movement, although the level and degree of fluctuation are somewhat different. Thus, economists havepaid increasing attention to inflation perceptions since the early 2000s. In doing so, the first task was to uncover the reasons why inflation perceptions constantly exceed the CPI.

3. The Formation Process of the Households Inflation Perceptions
This section discusses the formation process of inflation perceptions, focusing on the fact that they constantly exceed the CPI. Inflation perceptions are affected by various factors such as consumers’ purchasing patterns,information-gathering behavior, and knowledge of the economy. This section identifies five factors that affect inflation perceptions. The first is the price of frequently purchased goods. Second, we consider food and gas prices.The third is consumers’ tendencies toward loss aversion, the fourthis the CPI’s quality adjustment, and the fifth is individual experiences with inflation and deflation.

3.1. Correlation between inflation perceptions and the CPI
As Figure 2 shows, inflation perceptions move closely in relation to the CPI. Their correlation is high at 0.76 (observation period: June 2007 to September 2023 on a quarterly basis). In the U.K., the correlation is 0.90 (observationperiod: November 1999 to September 2003 on a quarterly basis). The Eurozone’s correlation is 0.82 (observation period: March 2004 to September 2003 on a quarterly basis).  In case of the U.S., for which the observationperiod is the longest among developed countries, the correlation between the CPI and inflation expectations using the University of Michigan Survey was 0.91(observation period: January 1978 to September 2003 on a monthlybasis).

3.2. Households’ Source of Information in Forming Inflation Perceptions The above subsection describes the strong correlations between inflation perceptions
and the CPIs in major developed countries. The next question is what type of information households rely on when responding to household surveys. First, we examine how knowledgeable households are regarding inflation. The“Financial Literacy Survey,” conducted by the Central Council for Financial Service Information in Japan, contains basic questions on inflation. The first item is as follows:
“When inflation is high, overall price level, including that of necessities and service, will go up swiftly.”
Respondents are given three choices: “correct,” “incorrect,” and “don’t know.” Survey results shows that 64 percent answered “correct,” while 3 percent chose “incorrect” and 29 percent responded “don’t know.”Surprisingly, approximately 40 percent of respondents who chose “don’t know” and “incorrect” do not understand the basic concept of inflation. Some argue that continued deflation in Japan may have affected theseresponses.The next question is more neutral.
“If the inflation rate is 2 percent and you had a deposit interest rate of 1 percent, how much goods and service you can buy one year from now?”
Only 55% of the respondents answered this question correctly, while 11 percent chose the wrong answer, and 33 percent responded that they did not know the answer. As this question shows, about 40 percent of consumers(“incorrect” plus “don’t know”) lack a basic understanding about the inflation.The responses to the two questions above suggest that at least 40 percent of consumers lack a basic understanding of inflation and its economicconsequences. This implies that these households are incapable of understanding the role of the CPI or monitoring its actual value. In other words, consumers who can respond regarding inflation perception and expectationsbased on accurate information related to the CPI account for only 60 percent at most, a disappointment for economists. Next, the Bank of Japan’s “Opinion Survey,” conducted in September 2013, directly inquired into sourcesof information for inflation perceptions. Figure 7 presents the survey results. The most popular answer was “price development of frequently purchased goods,” which included food and gas prices, followed by “informationfrom TV news and newspapers” and “regular income and payment.” However, each share of the total response is less than half that of frequently purchased goods. Although this question allows multiple responses (up to threeresponses), approximately half of the respondents relied on the price of frequently purchased goods when asked about their inflation perceptions. Similar surveys have been conducted in other developed countries; Figure 8 showsan example of a survey conducted in the U.S. The most popular responses when asked about the inflation expectations is “own shopping,” followed by “family and friends,” which is also related to daily shopping. Masscommunications, such as TV, radio, and newspapers, follow, but their share is much lower than the top two.Figure 9 shows the results of a similar survey conducted in Germany. As the survey was conducted in 2023, the topchoice was “war in Ukraine,” but the second choice was “fuel,” and the third was “food prices.” “Media about inflation” was fifth, and its share was only about the half that of food prices and gas prices. An European CentralBank (ECB) study segregated consumers into “better performers” and “worse performers” depending on differences in the formation of inflation perceptions. The former are clever consumers who refer to objective informationsuch as mass media reports when forming inflation perceptions. The latter are unwise consumers who depend on their own experiences, such as shopping and the price development of frequently purchased goods. In Japan, theshare of “worse performers” seems to reach 40 to 50 percent, which may be a reason for the significant divergence between the CPI and inflation perceptions. A survey by the central bank of Germany showed that even thoughconsumers knew about mass media reports regarding inflation, 89-92 percent of them still relied mainly on information acquired through their own shopping experience. Figure 10 shows the results of the U.K. survey conductedby the Bank of England (BOE). As in other countries, the first to sixth choices were related to the prices of frequently purchased goods. Media reports on inflation development and the actual CPI ranked seventh and last,respectively. In another survey, 45 percent of households that were asked about the CPI inflation rate answered “don’t know,” and only 10-20 percent answered correctly. Figure 11 shows the distribution of inflation perceptionsin the Eurozone  and reflects several features. First, the responses are concentrated at multiples of five such as 0, 5, 10, 15, and 20. Second, even for the range between 0and 5 percent, which is close to the actual CPI, responsesare concentrated on integers such as 1, 2, and 3. Third, although extremely high responses (such as 30 percent) were observed to a certain degree, responses below zero were rare. These finding align with “rounding responses,” aconcept used in psychology, which often emerges when respondents are not confident with their own responses. Most of the“worse performers” may have given such responses.

3.3 Inflation Perceptions and Price Index Based on Frequently Purchased GoodsI
Figure 13 plots the price index of frequently purchased goods and inflation perceptions. Compared to the overall CPI, shown in Figure 2, the gap between price index and inflation perceptions has narrowed. However, before thecurrent inflationary period, inflation perceptions are much more stable than the frequently purchased goods index.The “consumer price index” statistics refer the “family income and expenditure survey” for consumers’ frequencyof purchases. According to these statistics, six items were purchased once or more per week, all of which were food. Goods purchased once or more per month include 80 items, such as food, gas, medical expenses, and dailynecessities, and their share of total expenditure amounts to slightly more than 10 percent.The fact that the most frequently purchased goods consist of food and gas is consistent with the statistical analysis and anecdotal evidence.In fact, a paper published by the Bank of Japan showed the effect of food and gas prices on inflation perceptionsusing statistical models. Another study by the Central Bank of Germany mentioned food and gas as frequentlypurchased goods. A study by the Ohio State University conducted a simulated consumer purchasing behavior test and concluded that food and gas prices are affecting inflation perceptions (they termed this “frequency bias,” andit exists in perception). Furthermore, a stus described in the previous subsection, a “worse performer” relies heavily on the price development of frequently purchased goods rather than macroeconomic information when forminginflation perceptions. This tendency has been reported not only in Japan but also in the U.S., the Eurozone, Germany, and Denmark. The Japanese Consumer Price Index publishes a price index according to purchase frequency(Figure 12). The price development of frequently purchased goods has three features. First, it is more unstable than the overall CPI, or goods purchased less than once per month. Second, during the entire period shown in thefigure; the average rate is 1.4 percent, which is higher than the overall CPI average of 0.6 percent. Third, during the current inflationary period, although the overall CPI has already peaked, the frequently purchased price indexcontinues to rise.  Study by the Central Bank of Denmark stresses the significant effect of food prices on inflation perceptions. Another study by the Bank of Italy showed that consumers cannot precisely recallinformation onrarely purchased goods such as durables, and this phenomenon distorts inflation perceptions.

3.4. Inflation Perceptions and Food and Gas Prices
The last subsection reveals that consumers refer to the prices of frequentlypurchased goods, such as food and gas, in forming inflation perceptions. In this subsection, we calculate the weighted average index based on the foodand gas prices in Japan. The weight of food is 26.2%, and that of gas is 1.8 percent, according to the CPI statistics (Figure 14). Comparing the food and gas price index (FGPI) shown in Figure 14 with the overall  CPI (Figure 2)and the prices of frequently purchased goods (Figure 13) yielded three features. First, the gap between inflation perceptions and FGPI is smaller than thoseamong the other indices. Second, the FGPI closely follows the surge ininflation perceptions, especially during the current inflationary phase. Third, the correlation between inflation perceptions and the FGPI increases to 0.84, whereas that with the overall CPI is 0.76. These findings indicate that mosthouseholds refer to the FGPI when forming inflation perceptions.The CPI is often criticized for its significant divergence from inflation perceptions. The CPI is a weighted average of the prices of individual items using the shareof their expenditure as their weights. Although such an approach is reasonable as an official statistic, there is room to develop a new price index that can follow inflation perceptions more closely than the official CPI by changingthe weights from expenditure share to frequency of purchases and focusing only on frequently purchased goods and services. Some Japanese economists developed a new price index based on this notion. They calculated theirown price index based on (1) limiting items to goods and services purchased more than once a week and (2) changing weights from expenditure to frequency of purchases. This new index amounted to 14.0 percent in June 2023,which is much closer to inflation perceptions than the overall CPI.Furthermore, some argue that the gap between the inflation perceptions and overall CPI is partly due to the fact that the CPI does not cover real estate prices.16Many consumers have misunderstood that the CPI included real estate prices because they did not understand its definition and coverage. An Italian survey revealed that more than half of the respondents (51.4%) incorrectlyreported inclusion of house prices in the CPI definition.Figure 15 shows households’ outlook for real estate prices surveyed by the Bank of Japan’s “Opinion Survey.” “Land Prices D.I.” is calculated as deducting theproportion of respondents who answered “land prices will go down” from the proportion who responded “land prices will go up.” The D.I. increased steadily from the beginning of 2021 until the most recent survey, and thisupward trend may have contributed to heightened inflation perceptions during the current inflationary phase. This finding implies that real estate price movements should also be considered when developing a new price index.

3.5. Inflation Perceptions and Loss Aversion of Consumers
The last subsection discusses the effects of the FGPI as a cause of the divergence between the CPI and inflation perceptions. However, this factor alone cannot explain the overall divergence between the two indicators. This isbecause, as Figure 2 indicates, although inflation perceptions constantly exceed the overall CPI, the FGPI is not guaranteed to move constantly above the overall CPI. Therefore, additional factors are required to explain thisphenomenon. From this point of view, this subsection discusses consumers’ loss aversion, and the next subsection considers the effect of the quality adjustment of CPI statistics as a factor that provides upward bias to inflationperceptions. Loss aversion is a concept often used in behavioral economics and psychology. It is a psychological phenomenon in which people are 2.5 times more sensitive to suffering the same amount of loss than gainingpleasure. It also applies to daily purchasing behaviors, where increases in the prices of frequently purchased goods, which is a loss to the household budget, cause much more damage to consumers than the same amount of pricedecline, which is a gain to the household budget. A study by the Bank of England referred to this phenomenon as “salient events,” and it may have increased inflation perceptions significantly. As Figure 14 shows, the FGPI ismore unstable than the overall CPI, and this, combined with loss aversion, may have increased inflation perceptions. Such a possibility has been pointed out in many related studies, including those by the Economic and SocialResearch Institute in Japan, the Central Banks of Canada, Denmark, and England, the Federal Reserve Board in the U.S., and the European Central Bank. Figure 16-1 shows the share of price-increasing items in Japan. Althoughraw data were not available, and statistical analysis could thus not be conducted, the timing of the widening gap between inflation perceptions and the overall CPI in 16-2 and the timing of the growing proportion of price-increasingitems in 16-1 nearly match. During the current inflationary phase, the gap between inflation perceptions and the CPI widened significantly, which was not seen in the low-inflation period. This may partly reflect consumers’ lossaversion, stimulated by a surge in consumer goods prices and an increasing proportion of price-increasing items. Figure 17 showed the relationship between news heard of price decreases/increases and inflation expectations in theU.S.19 Consumers’ response to price development is asymmetrical and much more sensitive to “price increases” than “price decline,” which is consistent with the loss aversion theory. Additionally, one-year-ahead inflationexpectations, as a proxy for inflation perceptions, are significantly affected by news about price increases.

3.6. Inflation Perceptions and the Quality Adjustment of the CPI
The CPI statistics adjust each item’s price index according to changes in quality. For example, even if the retail prices of PCs remain unchanged, their price index will decrease as their performance improves. On the other hand,consumers focus only on changes in retail prices. Therefore, these differences in price perceptions cause an upward bias in inflation perceptions. Shimizu, 20 the Federal Reserve Board, 21 and the central banks of Denmark,Ireland, and Canada refer to the effect of quality adjustment. The Bank of Canada estimates that quality changes lower the CPI by 0.2 percent.

3.7. Consumers’ Past and Present Experiences of Inflation
Consumers’ past and present experiences of inflation may affect inflation perceptions and expectations. For example, the “Great Inflation” in the U.S. in the 1970s and the “Oil Price Shock” in Japan in the same decade havecontributed to increased inflation expectations. The December Study showed that inflation expectations as surveyed by the University of Michigan rose significantly in the 1970s and remained high until the first half of the 1980s.By contrast, economies have suffered from a long period of low inflation or mild deflation since the late 1990s in Japan and the late 2000s in other developed countries. Because price fluctuations remained modest during thisperiod, consumers paid less attention to price movements. Instead, their attention shifted to other, more urgent economic issues such as the employment situation and wage increases. Such changes in attention occur due topsychological mechanisms that efficiently allocate limited recognition resources according to the priority of economic issues. The consumers’ tendency to pay less attention to price development during low-inflation periods iscalled “rational inattention.” This is rational behavior in the sense that when prices are stable, it is rational to allocate limited cognitive resources to other economic issues and intentionally pay less attention to price development.For example, a study by the Bank of Canada pointed out that the 25 years of low inflation period has caused Canadian consumers “rational inattention” to price developments (Figure 18).  The Bank of England’s survey showedthat the percentage of respondents who chose “inflation is significant problem” was more than 50 percent in the high-inflation period of the 1970s but had declined to less than 10 percent just before the pandemic began in 2019.Another study by the Central Bank of Germany argued that because the young generation only experienced a low-inflation period, rational inattention became dominant. 24 However, this rational inattention suddenly disappeared atthe beginning of the current inflationary period, and people began to pay attention to inflation (Figure 19). Some economists argue that significant inflation and deflation have long-term effects on inflation perceptions andexpectations. For example, the inflation perceptions of those who experienced high inflation remain relatively high for a long period compared to those who did not experience such a situation. However, statistically speaking, theresults of testing the relationship between the cohort and inflation perceptions are mixed depending on the surveyed countries and periods; thus, it is too early to determine whether inflation perceptions are affected by cohorts.

4. The Relationship between Inflation Expectations and Inflation Perceptions
As pointed out in Section 2, studies on inflation perceptions have recently become popular because inflation perceptions have been increasingly considered the basis of inflation expectations. This approach has the advantage ofexplaining households’ inflation expectations better than the traditional approach.25 The evidence of a close relationship between inflation perceptions and expectations is outlined below.First, consumers’ socioeconomicattributes affect both inflation expectations and perceptions similarly. For example, several socioeconomic attributes affect inflation perceptions. First, women’s perceptions of inflation were generally higher than those of men.Second, the higher the educational background, the lower is the perception of inflation. Third, people with higher incomes have lower inflation perceptions. Fourth, inflation perception decreases as financial literacy improves.Asurvey by the Federal Reserve Bank of Cleveland in the U.S. yielded several results.  First, men’s inflation perception was 4.6 percent from August 1998 to November 2001, whereas that of women was 6.9 percent. The inflationexpectations of men and women were 4.0 percent and 6.4 percent, respectively. Second, as noted in the December Study, households refer to not only inflation perceptions but also other macroeconomic information available tothem. This is supported by the time difference correlation analysis showing inflation expectations, which contains information about the future outlook in addition to inflation perceptions. When households are divided into fiveincome levels, and the lowest-income group responded with inflation perceptions of 9.2 percent, whereas for those in the highest-income group, the percentage was 4.7. In terms of inflation expectations, the lowest-income groupresponded with 8.4 percent, whereas the highest-income group responded with 4.4 percent. Third, the inflation perception of university graduates was 4.8 percent, and that of those whose education level was high school graduateor below was 7.5 percent. Similarly, the inflation expectation of those who had graduated from university was 4.4 percent, whereas that of people who had graduated from high school or less was 6.6 percent.The mechanismbehind the relationship between inflation expectations/perceptions and socioeconomic attrition, especially gender differences, has been widely studied, but no convincing explanations have yet been offered. Influential hypothesesinclude that, first, because women are still more often engaged in daily shopping than men, women are more aware of changes in the price movements of frequently purchased goods, as described in subsections 3.3. and 3.4.Second, because the portion of women still not engaged in jobs is greater than that of men, they are not interested in inflation development and thus become “worse performers” more often than men. Moreover, household surveyresults have shown that consumers rely heavily on inflation perceptions when responding to questions on inflation expectations. Figure 20 is a result of the Bank of Japan’s “Opinion Survey” asking information sources ofinflation perceptions and expectations. Compared to the responses for inflation perceptions, answers for five-years-ahead inflation expectations are tilted toward macroeconomic variables affecting future inflation development,such as exchange rates, stock/land prices, and monetary policy. This tendency is consistent with the results of a December Study, which applied time-difference correlation analysis and found that inflation expectations containunique information not included in inflation perceptions. However, more than half of responses were assigned to “backward-looking factors,” implying that households rely heavily on inflation perceptions (backwardlookingfactors) when forming inflation expectations (Figure 21).  Similarly, a survey conducted by the Bank of England shows that households rely heavily on backward-looking information when forming inflation expectations (Figure22). The survey compares the important factors in determining inflation expectations for one and five years ahead. In the figure, long-term inflation expectations were determined by relatively more forward-looking factors, such as“expectations of future economic strength,” “level of interest rates,” and “discussion in the mass media,” than short-term Third, there is a strong correlation between inflation expectations and perceptions when using individualhousehold data. Figure 24 shows an example of an analysis conducted by the Federal Reserve Board. Regrettably, as yet, no individual data have been disclosed in Japan. These figures plot the correlation between inflationperceptions on the horizontal axis and inflation expectations on the vertical axis. The figure on the left side indicates the long-term relationship (5?10 years), and the right side shows the short-term (1 year) relationship. The greenline is the result of a linear regression, limiting the X variables from -10 to +20 percent. Another study by the Central Bank of Germany calculated the pass-through from inflation perceptions to expectations.29 A 1 percent increasein inflation perceptions increased inflation expectations by 0.74 percent in the long term and 0.53 percent in the short term. In addition, the pass-through was higher when the actual inflation rate was lower. A similar statisticalanalysis was conducted by the Bank of England, which reported a 0.4 percentage point pass-through in five-years-ahead expectations.

5. Implication for Monetary Policy Implementations
As inflation perceptions significantly affect the formation of inflation expectations, they have important implications for monetary policy implementation.

5.1. Demerits of Focusing Only on Core CPI as a Policy Indicator
When discussing monetary policy implementation, often-used price indicators are the “core-core CPI,” which exclude food and gas prices to reduce short-term fluctuation of the overall CPI. Although it is of the utmostimportance to understand correctly the fundamental movement of prices, we should also bear in mind that the “non-core CPI”consisting of food and gas prices significantly affect inflation perceptions.Therefore, both core CPIand non-core CPI should be included when monitoring price development (Figure 25). Additionally, the development of a new type of price index is required. The weight of this index should reflect the purchasing frequency andconsumers’ loss aversion (asymmetric weights).

5.2. Macroeconomic Effect of Inflation Perceptions/Expectations
As discussed in section 4, inflation expectations differ significantly depending on socioeconomic attributes such as gender, school career, and income level. For example, women have higher inflation expectations than men.Because nominal interest rates are equal for both sexes, real interest rates are lower for women than for men. This causes distortions in household behavior, such as expenditure, savings, and investment. In addition, inflationexpectations significantly affect wage negotiations. This, in turn, affects the efficiency of macroeconomic resource allocation. The central bank of Germany’s latest study stresses the importance of policymakers paying moreattention to analyzing and monitoring inflation perceptions. In Japan too, further effort should be devoted to analyzing households’ inflation expectations and perceptions and their application to monetary policy decisions.

5.3. The Central Bank’s Actions to Guide Household Inflation Expectations
Central banks should not only monitor households’ inflation expectations and perceptions but also promote the proper formation of inflation expectations by following practices to improve the efficiency of monetary policyimplementation.First, central bank communication with the general public account for the fact that inflation expectations differ significantly depending on socioeconomic attributes, such as gender and income level. Second, toreduce the upward bias of inflation expectations compared with the CPI, central banks need to promote financial education for households so that they can actively gather and understand macroeconomic information, including theCPI. The background factor for such efforts is that the higher the financial literacy, the lower are the inflation expectations. In other words, as the previously mentioned study by the ECB stressed, improving financial literacywould increase the number of “better performers” and decrease that of “worse performers.” Other studies have highlighted similar views.

6. Conclusion
This study analyzed the formation process of inflation perceptions and discussed the relationship between inflation expectations and perceptions as well as the implications for monetary policy. Research on inflation perceptions isscarce because household inflation perceptions constantly exceed the CPI at a significant level. This attitude gradually changed as it was considered to form the basis for inflation expectations. At the same time, the followingreasons explain the divergence between CPI and inflation perceptions. First, households lack an understanding of CPI statistics. They rely mostly on the price development of frequently purchased goods when responding toinflation perceptions. In fact, a price index composed of food and gas prices fits inflation perceptions more closely than does the overall CPI.Second, consumers tend to be loss averse. They are 2.5 times more sensitive to thesame amount of loss compared to gain. This sensitivity increases inflation perceptions, particularly when large portions of goods and services are experiencing rising prices. Third, the CPI adjusts the quality improvements for eachitem by lowering the price index. Households, however, only refer to retail prices, which do not reflect quality  changes; thus, the divergence between the CPI and inflation perceptions widens.The following evidence emergedregarding the relationship between inflation expectations and perceptions. First, differences in socioeconomic attributes, such as gender, income level, and educational background, affect inflation expectations and perceptions inthe same manner. Second, according to household surveys, households rely heavily on inflation perceptions when responding to inflation expectations. Third, a crosssectional analysis using individual samples showed a strongcorrelation between inflation expectations and perceptions. The significant role played by inflation perceptions in forming expectations has implications for monetary policy implementation. First, food, and gas prices, which areexcluded from the “core-core” CPI, are important components of price movements when assessing households’ inflation perceptions. Second, socioeconomic attributes affect inflation expectations and thus distort real interestrates and efficient resource allocation. Third, households’ financial literacy should be improved through financial education to encourage the formation of inflation perceptions in response to macroeconomic situations.As researchon inflation perceptions is scarce, further studies are required to accumulate survey data on the formation mechanism of inflation perceptions and their relationship with inflation expectations. The urgent task is to disclose detailedsurvey results, such as individual sample data, data sorted by socioeconomic attributes, and distributions of inflation perceptions and expectations.  


(References)
Bank of Canada (2020) “Perceived Inflation and Reality: Understanding the Difference” Remarks by Deputy Governor L. Schembri
Bank of Italy (2008) “What’s Behind “Inflation Perceptions? A Survey-Based Analysis of Italian Consumers,” P. D. Giovane, S. Fabiani and R. Sabbatine, Working papers No.655, January 2008
Bank of Japan (2022) “Households’ Perceived Inflation and CPI Inflation: The Case of Japan,” Y. Takahashi, Y. Tamanyu, No.22 E-1, March 2022
--- (2023a) “Outlook of Economic Activity and Prices,” October 2023
--- (2013), (2023b) “Opinion Survey”
BOE (2016) “How Are Households’ Inflation Expectations Formed?” Quarterly Bulletin 2016 Q2, J. Rowe
--- (2019) “Understanding Inflation: Expectation and Reality,” Silvania Tenreyro, July 10, 2019
Bruine de Bruin W et al. (2013) “Measuring Inflation Expectations,” Annual Review of Economics, 2013
CCFSI (2022) “Financial Literacy Survey,” Central Council for Financial Services and Information
Central Bank of Ireland (2022) “What Drives Consumers’ Inflation Perceptions in the Euro Area?” Economic Letter Vol. 2022, No.6
D’Ancuto et al. (2019) “Exposure to Daily Price Changes and Inflation Expectations,” E. D’Ancuto, U. Malmendier, J. Ospia, M. Weber, NBER Working Paper 26273, September 2019
Demarks National Bank (2019) “Revisiting the Inflation Perception Conundrum,” K. Ablidgen, A. Kuchier
Deuche Bundes Bank (2023) “The Pass-through from Inflation Perceptions into Inflation Expectation,” Discussion Paper No17/2023, S. Huber, D. Minira, T. Schmidt
ECB (2017) “EU Consumers’ Quantitative Inflation Perceptions and Expectation: An Evaluation,” R. Arioli, C. Bates, H. Dieder, I. Duca, R. Friz, C. Gayer, G.
Kenny, A. Meyler, I. Pavlova, ECB Occasional Paper Series No.187, April 2017
Ehrmann et al. (2017) “Consumers’ Attitudes and Their Inflation Expectations,” M. Ehrmann, D. Pfajfar, E. Stantoro, International Journal of Central Banking February 2017
ESRI (2013) “Disagreement and Biases in Inflation Expectations of Japanese Households,” Y.Ueno and R. Namba, Discussion Paper Series No.300, Economic and Social Research Institute, Cabinet Office of Japan,
--- (2022) ”Micro-shock and Macroeconomic Forecast,” Y.Iizuka, J. Kikuchi, Y. Nakazono, “Keizai-Bunseki,” No. 204 (Japanese only)
FRB (2001) “The Demographics of Inflation Opinion Surveys,” Federal Reserve Bank of Cleveland, October 2001
--- (2016) ” Inflation Perception and Inflation Expectations,” FEDS Notes 2016/12, A. Detmeister, D. Lebow, E. Peneva
--- (2018) “Perceptions and Expectations of Inflation by U.S. Households,” Finance and Economics Discussion Series 2018-073, A. Axerlrod, D. Lebow, E. Peneva
FRB of Cleveland (2001) “The Demographic of Inflation Opinion Surveys,” M. Brayan, G. Venkatu, October 15, 2001
Georgans et al. (2014) “Frequency Bias in Consumers’ Perception of Inflation: An Experimental Study,” European Economic Review 67, S. Georgans, P. Healy, N. Li
Hayo and Neumeier (2022) “Households’ Inflation Perceptions and Expectations: Survey Evidence from New Zealand,” Bearnd Hayo and Florian Neumeier,
International Economics and Economic Policy (2022)
NBER (2020) “Gender Roles and the Gender Expectation Gap,” NBER 26837, March2020
Nihon-Keizai-Shinbun (2023) “Price Development Analysis Needs to Consider Degree of Frequent Purchasing Behavior: Inflation Perspective and the Monetary Policy,” S. Abe and N. Inekura, August 25, 2023 (Available inJapanese only)
Shimizu (2008) “Quantitative Analysis on Recent Inflation Perceptions,


Instability of Japanese Household Inflation Expectations During the Current Inflationary Phase

This study analyzed Japanese household inflation expectations during the current inflationary phase and concluded that they are relatively unstable compared with those of European and U.S. households. Japanese householdinflation expectations show certain unique features. First, they consistently move above the Consumer Price Index (CPI) and closely follow the movement of CPI, especially in the case of one-year-ahead expectations. Second,during the current inflationary phase, inflation expectations suddenly became unstable, since they rose to +10percent for one-year-ahead expectations and +5 percent for five-years-ahead expectations. Both expectations exceededthe actual CPI at the time. Third, although the peak level of CPI was lower in Japan than in the U.S. and European countries,Japanese inflation expectations were significantly higher than other countries during the currentinflationary phase. The contrast is especially remarkable in the case of five-years-ahead expectations, where those of the U.S. and European countries were barely affected by the rises in CPI. Further analysis on the stability ofJapanese inflation expectations in different economic circumstances is required, such as an examination of the mechanisms behind household inflation perceptions, since these are closely related to inflation expectations.

1. Introduction
The Japanese economy has experienced a period of long-term low inflation. During this period, the Consumer Price Index (CPI), excluding perishables and the effects of increases in consumption tax, has hovered around theminus 2 to plus 2 percent range. This low-inflation phase abruptly ended at the beginning of 2021 as the corporate sector started to pass on the jump in import goods prices to the prices of their own products. In January 2023,CPI peaked at + 4.1 percent on a year-on-year basis and remained high at +2.8 percent as of October 2023 (Figure 1). The CPI is projected to remain above +2 percent on a year-on-year basis, reflecting the rise in the corporatesector’s cost-push price rises and surges in crude oil prices. The Bank of Japan projects CPI to remain at a relatively high rate of +2.8 percent in both the fiscal years 2023 and 2024.Such increases in the inflation rate have causedvarious economic problems in Japan, while providing a unique opportunity to analyze household inflation expectations. This study analyzed Japanese household inflation expectations and compared them with those of Europeanand U.S. households during the current inflationary phase. The results of our analysis show that Japanese household expectations of inflation are unstable. This conclusion is supported by the following observations: First, beforethe current inflationary phase, they moved around the +2 to +3 percent range, which was higher but more stable than CPI. In addition, the five-years-ahead expectations were more stable than the one-year-ahead expectations.Second, during the current inflationary phase, inflation expectations suddenly became unstable as they rose to +10 percent for one-year-ahead expectations and +5 percent for five-years-ahead expectations, much higher than theactual CPIat that time. Third, compared with the U.S. and European countries, Japanese expectations became much more unstable than their expectations, especially for the five-years-ahead inflation expectations. This contrast isastonishing, considering that the peak levels of the U.S. and European CPI (+8 to 10 percent) were much higher than those of Japan. The remainder of this paper is organized as follows: Section 2 focuses on inflationexpectations, which have consistently moved above CPI and suddenly became unstable during the current inflationary phase. Section 3 demonstrates that inflation expectations in Japan are more unstable than in the U.S. andEuropean areas. Section 4 tests statistically whether household inflation expectations contain forward-looking information. Section 5 concludes the paper.

2. Household Inflation Perceptions
The Bank of Japan conducts the “Opinion Survey” on a quarterly basis, and it includes household inflation perceptions and expectations. The survey asks about household perceptionsof inflation in two ways. One is by choosingfrom five categories. The other asks for direct changes in the price level compared to the previous year. For example, from the September 2023 survey, 68.4 percent of households responded that “present price levels have goneup significantly” and the median value3 of “changes in price levels compared with one year ago” was +10 percent. This value is significantly higher than the actual CPI at the time, which was+2.8 percent. Figure 2 shows the time-series relationship between changes in CPI and inflation perceptions. It exhibits several distinctive characteristics. First, inflation perceptions have been consistently higher than changes in CPI. Second, inflation perceptions closelyfollow CPI movements. Third, the gap between the two variables widens when CPI increases, including during the current inflationary phase. Fourth, inflation perceptions remain at zero or above, even when CPI falls into negativeterritories.The first feature of inflation perceptions mentioned above, that is, inflation perceptions have been consistently higher than the changes in CPI, caused economists inside and out of Japan to question the reliability ofinflation perceptions. Therefore, only small number of studies have been conducted in this field. In fact, a paper published by the Bank of Japan in 2022 pointed out that although the reasons behind the gap between CPI andinflation perceptions have not become apparent, only a few studies have tried to explain the cause of the gap. Figures 3 and 4 show that inflation perceptions have been higher than the actual price index, not only in Japan, but alsoin the U.K. and Eurozone.
Both U.K. and Eurozone households share the same features asthose of Japan. First, inflation perceptions constantly exceed CPI (HICP), but closely follow CPI (HICP). Second, in the Eurozone, the gap between inflationperceptions and HICP widens during inflationary periods. Third, inflation perceptions remain at zero percent or higher even when CPI (HICP) falls into negative territory. The first point mentioned above, where inflationperceptions constantly exceed CPI, is often called the “inflation perception conundrum” and has been observed in many developed countries, including Sweden, Italy, and Canada. However, research activities to analyze inflationperceptions are scarce, as studies in this field only began in the 2000s.

3. Inflation Expectations in Japanese Households Are Unstable Compared to Those of the U.S. and European Countries
3.1. Japanese Household Inflation Expectations
While research on inflation perceptions is relatively scarce, inflation expectations have attracted the attention of academics and central bankers. This marked contrast arises because movements in inflation expectations greatlyinfluence aggregate demand, such as personal consumption and corporate fixed investments, through changes in real interest rates. Real interest rates are defined as nominal interest rates minus expected inflation. If this rises, theactivities of households and the corporate sector fall, and vice versa. Inflation expectations have attracted academic attention since the 1980s, and particularly since the global financial crisis. This is because monetary policies inmajor developed countries maintained nominal interest rates at zero or in the negative range for a substantial period, and real interest became closely linked to movements in inflation expectations.Since inflation expectations are notdirectly observable variables, professional forecasts of inflation expectations, such as surveys of economists and people engaged in financial markets or inflation expectations derived from yields on inflation-linked bonds, areoften used as alternative variables (Figure 5). As of October 2023, these alternative figures ranged between 1 and 1.5 percent. When the inflation expectations are used as a variable in macroeconomic models, they often assumerational expectations (people act according to the assumption that they know all the information about the macroeconomic variables) or adaptive expectations (people rely on past inflation rates as a proxy for future inflation rates).As for inflation expectations of Japanese households, the Bank of Japan’s “Opinion Survey”is conducted on a quarterly basis. The latest survey results (September 2023) show that the most common choice among fivecategories for one-year-ahead inflation expectations was “prices will go up significantly,” and that the median value was +10 percent, which was much higher than the Bank’s projection of +2.8% for fiscal 2023. In addition, thesurvey indicated that the median value for five-years-ahead inflation expectations was +5 percent, which was much higher than the professional economists’ forecast of +1 to 2 percent. These results have led economists toconclude that the household inflation expectations were too high to “justify” or “impractical to use in economic analysis.”Household inflation expectations were surveyed using the Cabinet Office’s “Consumer ConfidenceSurvey.” It asked about respondents’ inflation expectations by offering choices from seven categories (Figure 6-2). The distribution of inflation expectations is similar to that found in the Bank of Japan’s survey. The mostcommon choice was that “prices will go up greater or equal to five percent.” The survey does not directly ask for the numerical values of inflation expectations. However, calculating a weighted mean yielded +4.7 percent, whichwas lower than that of the Bank of Japan’s survey, but much higher than the professional survey results. The Bank of Japan’s “Opinion Survey” asks inflation expectations by choosing from five categories. Based on thedistributions of the responses, it is possible to calculate a numerical value of inflation expectation under a certain assumption. However, expectation values generated by such statical procedures are quite often much lower than thevalue from asking directly about inflation expectations. This inconsistency suggest that statistical procedures or assumptions need further review. Comparing Figure 7 with Figure 2, movements in inflation expectations are morestable than those in inflation perceptions. In addition, both variables exceed the CPI increase rates in almost all observation periods. The features of one-year-ahead inflation expectations are as follows: First, inflation expectationsdid not increase with inflation perceptions during the 2008 inflation period. Second, before the current inflationary period, inflation expectations were stable, ranging between approximately 2 to 3 percent. Third, during the currentinflationary phase, inflation expectations increased 10 percent, similar to inflation perceptions.Next, the five-years-ahead inflation expectations also constantly exceed the CPI increasing rate, except during the consumption taxincrease period. Before the current inflationary phase, five-year expectations moved moderately by approximately 2 percent and were more stable than one-year-ahead inflation expectations. However, during the current inflationaryphase, five-year expectations increased swiftly to the 5 percent level and exceeded actual CPI. This contradicts the basic feature of five-years-ahead inflation expectations, which should reflect the long-run average of inflation ratesrather than short-term fluctuations in actual inflation. In other words, Japanese household inflation expectations have not been anchored firmly.

3.2. Comparing Japanese Household Inflation Expectations with Those of U.S. and European Households
In this subsection, the inflation expectations of Japanese households are compared with those of U.S. and European households. Figure 8 shows one- and five-years-ahead inflation expectations of U.K. households. Even in the U.K., inflation expectations consistently exceed CPI. However, in contrast to Japanese households, five-years-ahead expectations exhibit only minor responses to changes in the CPI. In addition, the peak level of one-year-aheadexpectations was approximately 4 percent, which was about half of the CPI increase. Since U.K. CPI rose to about 10 percent year-on-year, which was about double that of Japan, the inflation expectations of U.K. householdswere much more stable than those of Japanese households. In Figure 7, inflation expectations in 2014 were lower than the actual CPI. This is due to the effect of a consumption tax rate increase at the time. Two different surveysare available regarding the inflation expectations of Eurozone households. A survey conducted by the European Central Bank (ECB) asked for both one- and ive-years-ahead inflation expectations, although the survey only wentback as far as 2020 (Figure 9). Another survey conducted by the European Commission offers a long observation period; however, it lacks long-term inflation expectations (Figure 10). Bearing in mind such statistical limitations,the ECB survey during the current inflationary phase showed that, although HICP moved up to about 10 percent, the increase in the one-year-ahead expectations was much more modest, and their peak level was only about 4percent. In addition, five-year expectations did not seem to be significantly influenced by the surge in HICP. Both one- and five-years expectations are much more stable than inflation perceptions, jumping to about 25 percent atthe peak. Figure 10 shows a survey conducted by the European Committee. The one-year-ahead inflation expectations go back to 2004. As with the ECB survey, inflation expectations consistently exceed CPI and follow it moreclosely than inflation perceptions. However, in the current inflationary phase, inflation expectations from the two surveys diverge significantly.  According to by D’Ancuto et al. (2019), until recently, the ECB was reluctant topublish household inflation expectations, since the high expectation results might cause damage to the reliability of the central bank.For example, the peak level of inflation expectations in the ECB survey was 5 percent, whereasthat in the European Committee survey was 12 percent. Although it is well known that differences in survey methods, such as how questionnaires and answers are provided, affect survey results significantly, the gap between thetwo surveys is too large to justify such factors alone.Regarding U.S. household inflation expectations, surveys conducted by the Federal Reserve Bank of New York (FRBNY) and Michigan University are considered reliable. TheFRBNYsurvey, shown in Figure 11, has the following characteristics. First, before the current inflationary phase, even one-year-ahead inflation expectations were quite stable at around 3 percent. Second during the same period,the levels of one- and three-year-ahead inflation expectations were remarkably close to one another. Third, during the current inflationary phase, three-year-ahead expectations swiftly returned to a 3 percent level in the latter half of2023, while one-year-ahead expectations rose to 6 percent. Overall, U.S. household inflation expectations seem to be much more stable than those of Japan. Figure 12 illustrates the results of the Michigan University survey. As thesurvey goes back to 1978 on a monthly basis, it covers the famous “great inflation” period of the latter half of 1970s. During this period, the inflation expectations of U.S. households increased to an astonishingly high level ofabout 10 percent, while actual CPI rose to about 15 percent at the time. Then, inflation subdued in the first half of the 1980s, as did inflation expectations, which hovered around 2-3 percent. The exceptions were the pre-GreatRecession period and the current inflationary phase. Comparing these two periods, the peak level of inflation expectations was approximately the same, whereas the actual CPI increased more during the current phase. This impliesthat inflation expectations were more stable during the current period. The differences in survey methods contributed to the fact that peak inflation expectation levels during the current inflation phase were lower than those of theFRBNY. A comparison of the inflation expectations in the three countries and regions strongly suggests that Japanese household inflation expectations show much lower stability than those in the other two countries and regions.This conclusion follows from the following observations during the current inflationary phase. First, it was only in Japan and the Eurozone (and only by the ECB survey) that one-year inflation expectations exceeded the actualCPI. Second, only the Japanese long-term inflation expectations overshot changes in CPI. Third, the peak level of longterm expectations in Japan jumped up to 5 percent during the current inflationary phase, which was higher thanin the other two countries/regions, where CPI rose significantly more than in Japan.

4. Forward-looking Information Contained in Household Inflation Expectations
Recent studies suggest that inflation perceptions are the key determinants of inflation expectations. The Infotainment Research Center conducted another study to examine the relationship between these two factors and the factorsdetermining inflation perceptions. This section examines whether inflation expectations contain forward-looking information by following the ECB’s analysis. The statistical method employed by the ECB calculates the timedifference correlations between the CPI and inflation perceptions/expectations and identifies the peak correlation period for each variable. The ECB’s analysis indicated that, in the Eurozone, although the inflation perceptionslagged behind the movement in CPI by three months, the lag in inflation expectations was only one month. Judging from the result, the ECB study concludes that inflation expectations contain “limited but useful forward-lookinginformation.”The results of applying the ECB method to the Japanese data are shown in Figure 13. The peak of the inflation expectations correlation had no lag against the CPI, while inflation perceptions lagged behind CPI bythree months. This result indicates that similar to the ECB analysis, Japanese household inflation expectations contain limited but useful forward-looking
information.

5. Conclusion
This study analyzed Japanese household inflation perceptions and expectations and compared them with those of European and U.S. households during the current inflationary phase. From the observations below, we concludethat Japanese household inflation expectations are unstable. First, during the current inflationary phase, inflation expectations jumped to a level well beyond the CPI’s increasing rate. In particular, both one-year-ahead and five-years-ahead inflation expectations exceeded the CPI rate, diverging significantly from the previous trend.Second, comparing such developments with those of the U.S. and the Eurozone, although the peak increasing rates of the CPIwere higher in those countries, their inflation expectations were more stable than those of Japan. The contrast is more distinctive in the case of five-yearsahead expectations, as their expectations are barely affected by the CPI hike.Further analysis of the stability of inflation expectations under different economic circumstances and comparisons among countries is required. One area to examine is the mechanism behind household inflation perceptions, asinflation expectations seem to be significantly affected by inflation perception movements. Inflation perceptions appear to have significantly affected heightened inflation expectations during the current inflation phase. Aspreviously mentioned, the Infotainment Research Center prepared a new study based on this view.




(References)
Bank of Japan (2008) “About the inflation expectations” BOJ review series 2008 J-15, T. Sekine, K. Yoshimura, C. Wada, December 2008, (available only in Japanese)
--- (2022) “Households’ perceived inflation and CPI inflation: A case of Japan” Y. Takahashi, Y. Tamanyu, Bank of Japan WP series No.22-J-2
--- (2023a) “Outlook of Economic Activity and Prices” October 2023, Bank of Japan
--- (2023b) “Opinion Survey” September 2023, Bank of Japan
Cabinet Office (2013) “Disagreement and Biases in Inflation Expectations of Japanese Households,” ESRI Discussion Paper Series No.300、Y. Ueno, R. Namba
Central Bank of Ireland (2022) “What Drives Consumers’ Inflation Perceptions in the Euro Area?”Economic Letter Vol. 2022, No.6
Consumer Affairs Agency (2019) “An Analysis of inflation Expectation Using “Price Monitoring Survey in Japan” May 2019
D’Ancuto et al. (2019) “Exposure to Daily Price Changes and Inflation Expectations,” E. D’Ancuto, U. Malmendier, J. Ospia, M. Weber, NBER Working Paper 26273, September 2019
Demarks National Bank (2019) “Revisiting the Inflation Perception Conundrum,” K. Ablidgen, A. Kuchier
ECB (2016) “EU Consumers’ quantitative inflation perceptions and expectations: An Evaluation” Discussion Paper 38, November 2016
--- (2021) “Making Sense of Consumers’ Inflation Perceptions and Expectations?The Role of (un)certainty” A. Meyler, L. Reiche, ECB Economic Bulletin February 2021
FRB (2001) “The Demographics of Inflation Opinion Surveys” Federal Reserve Bank of Cleveland, October 2001
--- (2016) “Inflation Perceptions and Inflation Expectations” A. Detmister, D. Lebow, E. Peneva, FED Notes December 5, 2016
Statistics Bureau of Japan “Consumer Price Index"