Rethinking The Traditional Measure Of Inflation

Summary

• Inflation expectations are an important component of Central Bank inflation measures.
• Using household and firm survey data, the Bank of Canada show that there is a disconnect between how inflation expectations are measured and what individuals actually forecast inflation to be.
• To obtain better measures of inflation, policymakers should consider changing the way they communicate key information, or use better measures such as a market-based approach.

Introduction

Expectations of future inflation are a key driver of short-term price movements. When estimating inflation, policymakers assume these expectations are rational and made based on full information. Yet, recent survey data covered in a Bank of Canada working paper, shows that this is far from the case.

Instead of using Central Bank communicated information, households and firms are more likely to base expectations on personal consumer experiences. Not only does this make individuals forecasts more varied, it also makes them susceptible to structural changes, such as greater price discrimination by large retailers using big data.

Given the large disconnect that exists between inflation expectations and those assumed by policymakers, new statistical methods are needed. In addition, Central Banks should re-think the methods they use for disseminating key policy information.

The Phillips Curve

Developed in the mid twentieth century, the original Phillips Curve focused on an inverse relationship between the rate of inflation and unemployment (lower unemployment results in higher inflation). Although the essence of this relationship still exists, the most recent iterations (New Keynesian) try to explain short-term inflation fluctuations through changes in inflation expectations and the output gap.

The output gap is a major component in the NKPC. In theory, it is linked to the labour market and the amount of domestic demand. In an open economy setting, it also contains information regarding global prices, exchange rates and the degree to which an economy engages in global trade. For example, according to the paper, the Bank of Canada uses a lag of relative import prices, and nominal oil prices as additional variables.

The other component^[1], inflation expectations, contains the forward-looking (expectations-augmented) element of the Phillips Curve. It has traditionally been relied on a full-information rational expectations assumption, i.e. that inflation prospects of firms and workers are very close to what actually happened. This approach to expectations has come under scrutiny, especially because of the evidence reviewed by the Bank of Canada.

Overall, the Phillips curve accounts for roughly 70 percent of the consumer price index (CPI) deviations in Canada and advanced economies, according to the authors. The open-economy version does even better, improving domestic inflation forecasts for a least half of OECD countries. The importance of the global component has also more-than-doubled since the 1990s.

The Problem With The Traditional Approach

For most advanced economies, inflation over last decade has been below target. This lack of inflationary pressure has been particularly surprising given expansionary monetary and fiscal policy. The authors argue that the rise in globalisation can only justify part of this divergence, and that understanding how inflation expectations are formed may be the key to obtaining better inflation estimates.

The Global Component in the Phillips Curve Has Limitations

According to the authors, global factors can only explain a portion of domestic inflation fluctuations. Otherwise stated, globalisation cannot account for the persistently low levels of inflation over the last 10 years. There are also two limitations to using global factors within the Phillips curve.

One is that the global component of CPI inflation is driven by food and commodity prices, and therefore it is less useful for explaining fluctuations in core inflation (which excludes food and energy sectors). Another is that an increased role of global factors should be linked to higher inflation volatility through greater interconnectedness, which has not been the case (Chart 1).

Rational Expectation Not An Appropriate Assumption

A recent academic study found that the primary driver of inflation since the Great Recession was expectations of future inflation rather than economic slack. Yet, recent survey data suggests that inflation expectations of firms and households behave very differently from full-information rational expectations assumed in most policy models (in the sense that they are not able to accurately predict inflation).

As an example, the authors use the results from the Bank of Canada’s Survey of Consumer Expectations (CSCE). It shows that, while inflation expected by professional forecasters aligns with CPI inflation between 2014 and 2019, households consistently overestimate actual inflation. Furthermore, consumer expectations were not aligned with the BoC’s 2% target (Chart 2), and vary widely among households.

Surveys of inflation expectations of firms exhibit similar properties. Although generally more accurate than households, at least one-third of firms in Canada perceive inflation to be outside the 1-3% target range, and typically above it.

Source: Page 7, Bank of Canada

Why Are Firms And Households Unable To Accurately Forecast Inflation?

Households or firms often lack information regarding economic developments and policy changes, for reasons related to cost or understanding. Instead, the paper argues, individuals rely on their personal experiences as consumers to form inflation expectations, such as changes in the prices of food or oil at-the-pump (which are often excluded from core inflation!). In fact, even when survey respondents are informed about the most recent inflation rate or the central bank’s inflation target, they still mainly recall prices they have seen during their own shopping.

The reason for the lack of attention to inflation may be related to the long history of low and stable inflation rates, i.e. successful inflation targeting has made inflation expectations less sensitive to disturbances. Individuals in low-inflation countries assign a lower weight to current prices when forming expectations than those in high-inflation countries.

With households placing significant weight on consumer experiences, differences in inflation expectations among households may be explained by income inequality. A recent paper finds that low-income households experience more dispersion in changes of their cost of living and thus display more heterogeneity in their inflation forecasts.

How Can Policymakers Overcome These Problems?

Firstly, communicating simple inflation information is key. In randomised control trials, individuals presented with the most recent inflation rate or the Fed’s inflation target, had on average 1-1.2pp lower inflation expectations. Firms in particular place more emphasis on these statistics.

Across all individuals, information relatable to their experiences has the most influence on expectations. For example, firms place significant emphasis on the beliefs about other firms’ expectations. According to the authors, such findings suggest that central bank’s communication should be tailored to narrower population subgroups.

Beyond communication, policymakers could use different measures. The paper suggests using market-based inflation expectations, derived from prices of financial instruments (bonds). These offer two main advantages; (i) they provide almost immediate information about the response of expectations to macroeconomic developments and policy announcements; and (ii) they can also be compared with other drivers of bond yields. This measure of inflation expectations has been well anchored around 2 percent targets (Chart 3).

Another alternative is to use financial professionals. Above we discussed the difficulties of using households and firms, but financial professionals exhibit inflation expectations that are largely in line with expectations derived from financial instruments. Indeed, the median one-year-out inflation forecast in surveys of professional forecasters is fairly accurate.

Source: Page 5, Bank of Canada

The Pricing Process Is Changing Too

The transformation of consumer-good markets, along with the expansion of new product varieties and a dramatic shift to big data are all reasons to rethink the methods for measuring inflation. Large retailers are able to price-discriminate now more than ever. They also offer a wide variety of differentiated products that are available at different prices depending on quality. Consumer baskets are, therefore, changing more rapidly. Around 40% of goods in baskets have been created in the last four years, and 20% will disappear in the next four years. New varieties tend also to be oriented to high-income households, which exacerbates inflation inequality between high- and low-income households.

Bottom Line

Policymakers face two key challenges. First, firms and households, who in theory provide the most accurate view on inflation expectations, are consistently overestimating actual inflation. This is mainly because they are unaware of recent inflation and monetary policy announcements. Second, the pricing process from which inflation is calculated is changing.

For these reasons, there is an increasing risk that inflation is becoming decoupled from its traditional fundamentals and as a result, monetary and fiscal policy may be becoming less stimulative (an issue discussed in our podcast with Pakistan’s Central Bank Governor Dr Reza Baqir). As countries attempt to eek out the full stimulatory effect of policies, the need to improve the measure of inflation expectations, the communication of key statistics and the changing nature of inflation is becoming important.

Expect this to be at the top of the 2021 agenda.

Citation

Kryvtsov, O. and MacGee, J. (2020), Has the Inflation Process Changed? Selective Review of Recent Research on Inflation Dynamics, Bank of Canada Staff Discussion Paper, https://www.bankofcanada.ca/wp-content/uploads/2020/11/sdp2020-11.pdf

1. In addition, some variants also include a backward-looking inflation component that is simply a lag of inflation. ↑