Measurement error in estimating inflation expectations from survey data: An evaluation by Monte Carlo simulations

This paper discusses the measurement error of conversion methods used to convert survey data to a quantitative index, especially the Carlson and Parkin (1975) method. When we want to summarise economic conditions using a numerical value, we often have to depend on survey data and convert them to a quantitative index. However, because survey research restricts responses into specific classifications and respondent’s response density may not be uniform, survey data surely include a specific error. In addition, because the distribution assumed in the Carlson–Parkin method may not fit the respondent’s distribution, this may also produce measurement error. This paper computes the measurement error of the Carlson and Parkin method in order to clarify its properties by Monte Carlo simulation. First, this paper finds that the "balance approach" contains significant error. Second, the error is large when true inflation expectations are large. Third, the error can be decreased by increasing the number of respondents. Fourth, changes in the response classification do not bring about dramatic changes compared with an increase in the number of respondents.