As response rates decline, the risk of polling errors rises
KNOWLEDGE of many facets of American life comes from surveys. Every ten years the Census Bureau asks adults to tally themselves and their demographic information in an online or mail-in form: a survey. The Bureau of Labour Statistics (BLS) produces monthly estimates of the unemployment rate and other economic numbers that are derived from interviews with tens of thousands of households: another survey. And knowledge of political issues from opinion polls, of course, comes from surveys. That would not be a problem if everyone answered the pollsters. But not everyone does, and the people who don’t can be very different from those who do.
Take a recent survey conducted by the Wall Street Journal and the National Opinion Research Centre (NORC) at the University of Chicago. The poll, published in March, purportedly showed that the popularity of many American values had fallen precipitously over the past few decades. In 1998 a poll by the Journal and NBC News found that 70% of adults said “patriotism” was “very important” to them; now just 38% do. Similarly, religion’s score had declined from 62% to 39%; bringing up a family from 59% to 30%; and community involvement, which rose from 47% to 62% by 2019, had crashed to 27%. Only money had become more important to the average adult since 1998: 31% then, 43% now.
Analysts at the US Census Bureau found that plunging CPS response rates during the covid-19 pandemic made the survey much less representative of poor Americans. Their best guess is that the estimate of median household income in 2020 is inflated by $2,000 purely because of a changing population of survey-takers. Another paper, from the Global Labour Organisation, an international non-profit group, found that CPS response rates are lower in cities and hubs for manufacturing jobs.
2023-06-22 08:46:55
Source from www.economist.com