Non-response bias in expectation surveys: Different perceptions and expectations of financial matters from “early quitters”

• Surveys requiring consecutive responses from each respondent face non-response bias. • Underrating early quitters’ views, SCE data cannot reflect the whole population. • Early quitters differ from others in demographics that can affect financial literacy. • Early quitters differ in perceptions and expectations of financial issues from others. • Response weights assigned by SCE cannot solve the non-response bias. This study demonstrates that non-response bias drives selective sample attrition in expectation surveys that require consecutive household responses. Examining 160,842 responses from 20,963 respondents in the Survey of Consumer Expectations (SCE) from June 2013 to April 2024, I show that 66.1% of respondents who quit the consecutive survey before the completion of 12 tenures—the early quitters—only contributed 44.6% of total responses. On average, they are 3.2 years younger, have a 5.14% higher proportion of females, and exhibit lower numerical ability and education level than other respondents. I construct a regression model to demonstrate that early quitters possess significantly different perceptions and expectations regarding personal, household, and national financial and economic matters when responding to SCE. Notably, the early quitters have more positive expectations regarding changes in their financial situation for the next year. They exhibit more negative perceptions of changes in loan difficulty in the U.S. over the past year and hold more pessimistic expectations for changes in loan difficulty in the U.S. for the next year. They anticipate a higher probability of an increased unemployment rate in the U.S. for the next year. All the above results are irrespective of whether the regressions are unweighted or weighted. The non-response bias renders the survey data ungeneralizable, and the weighting applied by SCE does not solve this bias.