Consumers are less accurate than experts at predicting inflation because of biased surveys
25 August 2021
Economists are better at predicting inflation than consumers when surveyed; but only because they are given more information, according to new research which suggests improvements in survey design could help to improve the accuracy of inflation forecasts.
Together with his PhD student, Xiaoxiao Niu, Nigel Harvey, Professor of Judgement and Decision Research from UCL PALS, measured the effectiveness of inflation surveys between economists and the general public, focusing on how people make judgements and decisions on the basis of information that is given to them and information received over time.
Context is known to affect responses to both online and traditional surveys but its impact on inflation surveys has not been previously studied. Nigel Harvey and Xiaoxiao Niu looked at the effects of answering one question about inflation before answering another and of providing additional information to expert but not lay survey respondents.
When asked to make judgments of what inflation will be in the current year and in future years, lay respondents usually give much poorer predictions than those of experts: their predictions are typically much too high. But this is still of interest to central banks because lay expectations affect consumer behaviour: the more people expect inflation to increase, the more likely they will bring forward their purchase of durable goods, which will, in turn, increase the price of those goods and raise inflation.
What Professor Harvey found is a crucial fault in the design of inflation surveys: experts, but not lay respondents, are provided with information about past levels of inflation and other economic variables.
“Surveys of both lay people and experts ask respondents to estimate inflation for more than one year. For example, they are required to estimate inflation for current year and for next year (and, typically, for later years too). We show that, when people are asked to do this, their estimates of inflation typically increase into the future. However, when different respondents make inflation judgments for different years, their estimates are not significantly different: they do not increase into the future.
“Thus, it appears that people providing responses for a number of years do not make their judgments using only their memory and the information they are given in the survey but are also influenced by their expectation of how inflation changes over time. Perhaps because news media covers the possibility of inflation increasing more thoroughly than the possibility of it decreasing, they tend to expect that increases in inflation are more likely than decreases.”
Professor Harvey found further biases where inflation surveys only provide additional information in expert but not lay surveys.
“Crucially, experts and lay people have been required to respond to different surveys. The notion that there is a difference between lay and expert judgments of inflation that is in need of explanation is predicated on the assumption that these different surveys are equally good at eliciting judgments of inflation.
“When responding to their surveys, experts are provided with information about past levels of inflation and other economic variables. Lay respondents are not given this information. We show that when they are given it, the accuracy of their judgments improves a lot and the heterogeneity of their judgments declines markedly. This suggests that, if lay and expert respondents were given the same survey, the difference between their estimates of inflation would decrease considerably: “To increase question validity, designers of inflation surveys need to ensure that the information provided within surveys corresponds to the information that consumers naturally use when making their economic decisions. Improvements in survey design based on this principle could help to improve the accuracy of inflation forecasts.”