Disagreement about inflation expectations: The case of Japanese households

Shusaku Nishiguchi, Jouchi Nakajima, Kei Imakubo, 2 May 2014



It is well known that inflation expectations vary across agents. Nevertheless, this fact has attracted little attention. Analysis of inflation expectations typically tend to focus on measures such as the mean and the median of such expectations. One of the distinguished exceptions is Mankiw et al. (2003), who discuss in a context of sticky information how the disagreement in US households' inflation expectations was evolving through the Volcker disinflation. Our recent research (Nishiguchi et al. 2014) has brought the disagreement issue once again to the fore. We have examined the disagreement in Japanese households' inflation expectations and its evolution over time using the Opinion Survey on the General Public's Views and Behavior (the “Opinion Survey”).1

Features of households' expectations

Households do not necessarily all share the same expectations. Figure 1 plots the distribution of five-year inflation expectations from the Opinion Survey.2  The figure shows that expectations are dispersed and do not converge on a certain value. Such dispersion can be observed also in each round of the survey and varies over time.

Figure 1 Dispersion in households’ expectations

Note: The panel plots the five-year expectations from the Opinion Survey.

Figure 2 Diversity of information referred to

Note: Herfindahl-Hirschman Index based on the information referred to, denoted as the ratio to the reference value.

A possible reason for the disagreement in inflation expectations is differences in the information households refer to when forming their expectations. This is confirmed by the results of a special survey in the Opinion Survey conducted in September 2013. The survey directly asked what information respondents referred to when forming their five-year expectations. Figure 2 depicts the extent of the dispersion of information respondents in each age group referred to represented in terms of the Herfindahl-Hirschman Index calculated from the survey results. The dispersion of the information referred to is relatively low among the young and the elderly, suggesting that they rely on limited sets of information. In contrast, the dispersion of the information referred to by the middle-aged is relatively large, suggesting that the group uses a diversified set of information. Such differences in the information households refer to likely contribute to the substantial dispersion of inflation expectations.

Forming a sharper spike of the expectations distribution

From early 2013, two changes in the features of the distribution of households' five-year expectations could be observed (Figure 3). First, the skew to the deflationary side has considerably diminished. A similar change can be observed in the distribution of market expectations (see Nishiguchi et al. 2014). Second, the right tail to the inflationary side as well as the left tail to the deflationary side both have contracted and the spike around 2% expectations has become sharper. This means that there has been a decrease in the share of respondents expecting deflation or high inflation and a substantial increase in the share of respondents expecting inflation of around 2%. What are the reasons for these two changes?

Figure 3 Distribution of households’ expectations

Note: Each panel summarises the survey results for each year. The horizontal axes represent expectations (%). Circles indicate the mean of each distribution.

The first change, the reduced skew to the deflationary side, seems to be largely due to price information relating to the past. That is, a rise in actual inflation has shifted the distribution in the direction of inflation. Not only have energy-related goods pushed up prices, but a rise has also been seen in prices of a wide range of items given that underlying upward pressure on prices has increased as the aggregate supply and demand balance has improved. However, the price information relating to the past alone seems insufficient to fully explain the second change, the sharp spike of the distribution in 2013.

A sharp spike in the expectations distribution is unique to the recent phase of rising prices. It was not observed toward 2008, another phase of a price rise (Figure 3). At that time, the skew to the deflationary side decreased from mid-2007 and the distribution as a whole shifted in the direction of inflation. And in 2008, the distribution did not show a sharp spike but became less steep to form two local peaks: one of moderate inflation expectations and another of higher inflation expectations.

Not only did the distribution during the previous episode of rising prices in 2008 not show a sharp spike like the one observed in the distribution of 5-year expectations in 2013, but the distribution of one-year expectations in 2013 also does not display such a spike (Figure 3). Rather, the distribution of one-year expectations as a whole has shifted in the direction of inflation and expectations are not highly concentrated on a specific level. Taken together, these facts suggest that certain new medium-term factors affecting only five-year expectations may be responsible for the sharper spike in the distribution from early 2013.

Although the above consideration cannot identify what these medium-term factors are, one candidate is perceptions regarding monetary policy. The September and December 2013 rounds of the Opinion Survey contained a question asking respondents whether they were aware of the "price stability target" of 2% inflation in terms of the year-on-year change in the consumer price index. Figure 4 shows the distribution of five-year expectations depending on whether respondents knew about the target. The expectations distribution of households that knew about the target has a sharp spike centred around 2%, while the distribution for households that had not heard of the target has a lower peak and fatter tails on both sides. This contrast suggests that communication of monetary policy can help to collect various inflation expectations of households into 2%.

Figure 4 Expectations distribution depending on awareness of monetary policy

Note: The panel summarises the survey results for September and December 2013.

Other factors that may sharpen the spike of the expectations distribution are employment situation. According to the Opinion Survey, while the expectations distribution for those who are quite worried about their employment situation has a fatter right tail, the distribution of those who are not worried has a sharp spike around 2% expectations. Thus, easing concerns about the employment situation due to improvements in business conditions in 2013 may be another reason for restraining from increasing the skew of the distribution to the inflationary side and generating the spike.

Have deflation expectations disappeared?

Although the skew to the deflationary side has contracted, the left tail of the distribution extending into negative territory has not completely disappeared since early 2013. This suggests that deflation expectations still linger despite the fact that recent circumstances mean that moderate inflation over the medium horizon has become more likely.

These sticky deflation expectations may possibly reflect households' experience of inflation. Figure 5 shows the distribution of five-year expectations for different age groups. In the distribution for those aged 20-29, the peak is lower and the left tail is fatter, which means that this age group is more likely to forecast deflation than older age groups. Those aged 20-29 in 2013 were too young to have consciously experienced inflation before the onset of deflation. This suggests that the hypothesis that households form inflation expectations taking their own inflation experience into account also holds in Japan (Malmendier and Nagel 2013). This, in turn, means that households' own experience of inflation is another factor contributing to the skew in the distribution of five-year expectations.

Figure 5 Expectations distribution by age group

Note: The panel summarises the survey results for 2013.

But inflation experience is not written in stone. In the case of the US, until the mid-1980s, the young and the middle-aged at the time only possessed high-inflation experience and hence tended to forecast higher inflation than the elderly. However, with all households in the US subsequently experiencing low and stable inflation (the Great Moderation), the dispersion of inflation expectations across age groups has diminished since the 2000s. Thus, given that inflation experience changes as circumstances improve, the sticky deflation expectations as a result of Japan's past experience may gradually disappear depending on future developments in prices.


Kamada, K. (2013) "Downward rigidity in households' price expectations: An analysis based on the Bank of Japan's 'Opinion Survey on the General Public's Views and Behavior'," Bank of Japan Working Paper Series, No.13-E-15.

Malmendier, U. and S. Nagel (2013) "Learning from inflation experiences," presented at the NBER Summer Institute 2013.

Mankiw, N. G., R. Reis, and J. Wolfers (2003), "Disagreement about inflation expectations," NBER Macroeconomics Annual, Vol. 18, pp.209-268.

Nishiguchi, S., J. Nakajima, and K. Imakubo (2014), "Disagreement in households' inflation expectations and its evolution," Bank of Japan Review, 2014-E-1.

1 The survey is conducted quarterly by the Bank of Japan. For each survey round, 4,000 individuals aged 20 and over are randomly sampled in an unbiased manner. The survey asks respondents about their inflation expectations over a medium horizon (the next five years) and a short horizon (the next one year) as well as inflation perceptions (perceived price changes compared to the last year). Responses are collected both in quantitative and in qualitative terms.
2 The Opinion Survey suffers from downward rigidity and reporting biases. Inflation expectations plotted in the charts in this column are bias-adjusted following the approach suggested by Kamada (2013). Moreover, the plots in the charts are smoothed using the Kernel method after the bias adjustment.

Topics: Monetary policy
Tags: inflation, inflation expectations, Japan

Senior Economist, Bank of Japan
Economist, Bank of Japan
Economist, Bank of Japan