Key findings
- We examined the frequency of non-responses to the inflation expectations question, referred to as item non-response, in our Household inflation expectations survey (H1), during a study period of 1998 to 2022. These non-responses (which include respondents who responded with “I’m not confident making a guess”) accounted for approximately 44% of the sample on average during the study period.
- During the study period, the research found evidence to suggest item non-responses lead to underrepresentation of some demographic groups in the survey. Young, female, low-income, and minority ethnic groups have lower response rates.
- Survey mode affects response behaviour. The survey response rates were found to be higher between 2018 to 2022, under the online survey mode introduced in 2018, with item non-response falling to around 24 percent and participation increasing among previously underrepresented demographic groups. Response rates also increase when inflation rates deviate from the central bank’s target range (most recently between 2021 to 2024).
- Using a sample selection model in this Discussion Paper, we quantify and demonstrate how to adjust for bias in aggregate (mean) measures of inflation expectations caused by item non-response. We show that there is a positive bias, and the aggregate inflation expectation series shifts down after the adjustment.
- Inflation expectations disagreement, both across and within subgroups, tends to decrease with the correction for non-response bias. These findings have important implications for survey design and monetary policy communication.
- The socio-demographic differences we identify in this paper suggest that some groups of the population may be less confident in responding to the inflation expectations question. A potential way to address these gaps would be to improve the outreach of monetary policy with more targeted communications.
Disclaimer
The data used in this Discussion Paper cover the period 1998 to 2022. However, the household expectations survey was redesigned in March 2022, to provide more precise estimates, increase the value of the survey and better align with international best practice. Without further research it is unclear if the non-response issues identified in the earlier period remain.
What is the household inflation expectations survey?
The primary purpose of collecting the data is to capture inflation expectations of the general public to support us with inflation forecasting and policy development. Our household inflation expectations survey is conducted quarterly via online survey and achieves approximately 1,000 household responses per quarter. The survey goes into the field after the previous quarter’s consumer price index inflation data have been released by Stats NZ.
The survey uses quota sampling across gender, age, ethnicity and region, with a blended sample used to ensure adequate representation from Māori and Pacific peoples. Final results are weighted using Random Iterative Method (RIM) benchmarking to Stats NZ Census distributions to ensure the achieved sample reflects the New Zealand population.
What data do we use?
We use data from the household inflation expectations survey. Our sample covers the period from 1998 to 2022 and contains 89,834 individual responses. The individual responses are anonymised every quarter. Independent samples are selected each quarter. Accordingly, the data follow a repeated cross-sectional design.
The survey asks for households’ perceptions of current inflation and expected inflation at varying horizons starting 1 year ahead. The inflation expectation question we focus on in this paper is formulated as follows: "As a percentage, what do you think will be the annual rate of inflation/deflation in the next 12 months?"
Why did we do this research?
We are interested in households’ inflation expectations, as they can provide useful insight into how inflation pressures have evolved in the New Zealand economy. This paper uses micro-data from the household inflation expectations survey to gain an understanding of how different demographic groups respond (or do not respond) to the inflation expectations question in the survey, allowing us to obtain a more accurate read of households’ inflation expectations. This Paper discusses ways in which we can improve item non-response bias and understanding of inflation developments in the economy.