[Home Page] [Previous Page]
[Old Roundtable Archives] [New Roundtable]
Posted by Daniel Schulman on January 16, 2000 at 21:06:23:
In Reply to: Online Research posted by Jim Villella on January 14, 2000 at 13:37:14:
The online population does not mirror the offline population. As a result, if you are looking for population frequencies--gender breakdown, presidential polling, favorite basketball team--you will get biased answers from online research.
There are two basic work-arounds. First, if you know something about the general population, you can weight the data and come up with some reasonable estimates. Second, you can make the assumption that the factors that lead someone to be online are not correlated with your variables of interest. I wouldn't have too much faith in the three examples above, but if your questions have less to do with SES and age, it might not be so bad. The first work around is expensive and takes a fair amount of sophistication; the second has limited viability (most survey questions have a fair amount to do with age and SES).
If you are not looking for profiling measures, you might be able to reduce the effects of bias through a stronger research design. For example, say you are interested in measuring your firm's brand awareness. You measure it at time t and time t+1 (say 6 months later) and see if it's different. With this quasi-experimental design, you might get a biased measure of incidence at each of your two time periods, but as long as there are no interactions between what's causing someone to be online and what raises your brand awareness, you'll get an unbiased measure of the difference between the two time periods.
Subject: