Non-response Bias
Lifestyle surveys conducted by post are likely to suffer from a certain element
of non-response bias – where the lifestyle or other factors among responders to
the survey are different from those who do not respond. For example, we know
from available evidence across many lifestyle surveys that responders to postal
surveys are more likely to be from older age groups (than proportionate to the
actual population). We also know that women are ‘better’ responders than men.
Both of these factors can be ‘corrected’ by weighting the data. Other factors
are more difficult (or impossible) to correct using weighting.A paper exploring the potential size of non-response bias in a lifestyle survey
was reported in Hill, Roberts, Ewing and Gunnell (1997). They reported on a
survey in Somerset where non-responders to the postal element were follow-up in
a telephone survey. Some large differences were found between respondents and
non-respondents.
| Males |
24.4 |
34.1 |
| Females |
24.0 |
30.0 |
| All Adults |
24.2 |
32.3 |
For a copy of the paper,
click here
|
