With this chapter, we take our approach to a wider applicability. So far, the expectations elicited in the laboratory pertained to a set of inflation patterns that was observed in countries like the USA and Germany over the years from the 1950s to the first decade of the 21st century. When we want to model expectations for a wider set of countries and historical periods, we need to generalize our approach. More concretely, it becomes necessary to consider pattern extrapolation for a wider set of possible courses of the price level. Notably, the set of 22 patterns used in the laboratory as described in Chapter 3 barely includes any cases of deflation. For the modeling of historical expectations, this can be quite limiting. Consider only the course of aggregate prices with periods of deflation in countries like the USA and the United Kingdom in the 19th century.

The approach explored in this book poses a challenge to various notions of conducting economic research. First, the concept of rationality that is commonly used in economics is examined. Researchers who take it for granted that a theory of rational expectations should start from a normative standpoint find their assumption questioned. The work presented here documents a theory of bottom-up rationality. A second group of scholars find their beliefs questioned. The testing of hypotheses is not the only task to be addressed in the controlled environment of a laboratory. Instead, the economics laboratory can be fruitfully used for quantifying behavioral models. Such bottom-up models can serve as a basis for many forms of applied empirical work.

This chapter details the procedures used for eliciting expectations in the laboratory as well as the data gathered. The expectations hypothesis set out here builds on the idea that people extrapolate univariate patterns in time series. The task, therefore, is to capture the general tendencies of extrapolation across different patterns or shapes of the time series. In the following, a pattern will be defined as an ordered sequence of observations over the recent past of an economic time series. The reliance on simple patterns by ordinary people should not be confused with the so-called technical analysis used by professional chartists. If anything, we could refer to humans under everyday conditions as “natural born” chartists.

Expected inflation is the major topic of this book. Yet, as we have noted before, the approach of pattern-based expectations can be applied to other macroeconomic variables as well. The second key variable besides inflation expectations that we have introduced and studied is the expected income. More concretely, Chapter 3 documents the laboratory data concerning subjective expectations of real GDP for patterns of length four.1 In this chapter, the generalized approach set out in Chapter 10 is extended to the case of income expectations. Accordingly, a new group of subjects sees the 25 possible patterns of length three, together with the information that the variable in question depicts the real GDP.2 Otherwise, the same instructions known from treatment (iii) are used again.

An obvious question to ask is how pattern-based measures of inflation expectations compare with survey data of expectations. We start this chapter by comparing the forecast performance of pattern-based inflation expectations with the expectations available from the survey center of the University of Michigan. The second part of the chapter then presents a different way of comparing the two sets of expectations. Following the lead of earlier contributions, the Michigan survey of expectations is taken as data to be explained. We will document that survey expectations can be tracked by our model of pattern-based expectations. In fact, pattern-based expectations appear as the best among competing explanations.

This chapter takes the pattern-based approach of modeling inflation expectations to higher levels of inflation. Remember that subjects in the generalized approach as developed in Chapter 10 were shown price-level paths with percentage changes in the range of ±2 percent. We then relied on the estimated scaling relationship to apply the elicited laboratory expectations to real-world inflation rates beyond this range. When turning to the modeling of expectations in times of high inflation, this approach is put to the test. Is it really warranted to rely on an estimated scaling relationship that was estimated with low rates of inflation when looking at annual inflation rates of 40 percent or more?

In the following, estimates of the real interest rate will be presented for the 10 largest economies according to the World Bank comparison for 2017. The countries included are the USA, China, Japan, Germany, United Kingdom, India, France, Brazil, Italy, and Canada. The interest rate used for this purpose is the nominal lending rate reported by the World Bank. Besides data on the nominal rates, the World Bank also reports real rates of interest. For these official calculations, inflation is measured as the percentage change of the GDP Deflator. For comparison and for the use by analysts and historians, we offer estimates of ex ante real rates based on the pattern approach. For this purpose, pattern-based expected inflation rates are computed (based on GDP-deflator data published by the World Bank) and used to correct the nominal lending rate for the expected inflation.1 Evidently, the ex ante real rate of interest (i.e., neither the nominal rate nor the ex post real rate) appropriately characterizes the circumstances under which economic decisions, for example, by investing firms, are taken.

The research literature has investigated the validity of the Fisher effect for a number of Asian economies. Among the countries that have been studied, Indonesia, South Korea, Malaysia, the Philippines, Singapore, and Thailand figure prominently. Overall, the evidence regarding the Fisher effect is mixed. Berument and Jelassi (2002) find no statistically significant effects of expected inflation on nominal interest rates for Korea and the Philippines. Kim et al. (2018) report statistically nonsignificant effects of expected inflation for the Philippines and Singapore and a significant effect, albeit not the strong form, for Thailand. Said and Janor (2001) find signs supporting the hypothesis for Indonesia but no evidence favoring Fisher’s proposition for South Korea, Malaysia, the Philippines, Singapore, and Thailand. Finally, Nusair (2008) documents a strong-form Fisher effect for South Korea and weak-form effects for Malaysia and Thailand. Summing up, the hypothesis under investigation has so far gained limited support for Asian countries.

In previous chapters, we have focused on the mean of inflation expectations. Thus, we have answered the question, for example, where the average person sees the price level a year hence. Here we want to turn to further dimensions of expectations. In particular, it is obvious that people differ with respect to their expectations. Clearly, the standard rational expectations hypothesis is silent regarding expectations heterogeneity. It assumes that all agents use the same relevant information. The expectations model of Mankiw et al. (2003) is an interesting departure from the standard model. These authors propose that only a fraction of the public makes an informed forecast at any point in time. The rest stick to the expectations they have formed earlier. This approach is similar to the model already discussed by Carroll (2003) where experts play the role of providing informed forecasts. Again, the model of pattern-based expectations offers a clear alternative.

The Fisher effect, that is, the theoretical prediction that nominal interest rates should adjust to changes in inflation expectations has been the subject of a variety of empirical studies. Chapter 9 has already presented evidence on this hypothesis with newer data. Interestingly, this theoretical proposition has not fared well for the historical data that Fisher (1930) himself studied. Irving Fisher found for the USA that the adjustment of interest rates takes about 20 years and even then is only partial. Furthermore, nominal interest rates for various countries, according to Fisher, do not adjust one-to-one to changes in inflation as witnessed by the fact that real rates of interest are lower in times of high inflation than in times of low inflation. Researchers after Fisher have largely confirmed this finding.1 A particularly interesting historical episode – from 1869 to 1913, under the gold standard – even suggests an outright contradiction of the Fisher hypothesis as we will explain. It is to these intriguing findings that this chapter turns.