|
|
 |
|
|
Sample Weighting
|
|
|
No matter how hard we try,
we are not in control of who decides to fill out our surveys and who decides not
to. There are things we can do to promote higher response rates, like
using engaging messages or various incentives, but the ultimate decision to
complete or to ignore an invitation to participate in a survey belongs to the
potential respondents. As a result, the samples we obtain are not always
perfect fits to the samples we seek.
Breakdowns on key
demographics in our returned surveys may not proportionally match the group that
we sampled from. For example:
- 10% of the people you invite to participate
in your survey may be senior executives
- 20% may be managers
- 70% may
represent employees at other levels within your company or your customers'
companies.
But, the proportions of completed surveys you get back in each
of these categories may be different - you may only get 5% senior executives and
8% managers, but wind up with 83% employees at other levels. Or,
other imbalances may occur. We typically find, for example, that customers
in Latin America or in certain countries within Europe are more likely to
respond to surveys than are customers in other world regions or in other
countries in Europe. While it is nice to know that the quality of
relationships can be so warm as to promote higher response rates in some
countries, it is important to us as researchers to know that we can often adjust
our data to compensate for imbalances. Otherwise, we could not give you
accurate estimates of what how people are feeling in various world regions or in
Europe as a whole.
By weighting data to
compensate for imbalances between the proportions we invite to participate among
subgroups in your population and the proportions in those subgroups who choose
to respond, we can help to ensure that the estimates of the population as a
whole or of large groupings within the population that we compute (e.g., the
percentage of your employees overall who love their jobs) are adjusted to
provide a better fit to what we believe to be the true characteristics of your
target group. Using the example above, the responses of senior executives
and managers would be weighted up and those of other employees would be
weighted down to approximate the known proportions in the original
mailing list for invitees. This provides us with more accurate estimates
of the group as a whole.
|
|
|
|
