June 7, 2012

Qualitative vs quantitative

The Herald claims “NZ says no to larger schoolrooms”  based on a street survey of “more than 70” people, of whom 81% were opposed to the changes.   The current clicky poll has 74% of about 8000 responses supporting the much weaker claim ‘Less one-on-one time can’t be good for kids’,  a statement that even John Key would probably not contest. We aren’t told what the actual questions were in the street survey, or how much the respondents knew about what the actual proposed changes were.

There are two ways you can get useful data by interviewing people.  In quantitative research, where you take a proper probability sample and ask questions with simple answers unambiguously related to what you are trying to find out.  The law of large numbers then ensures that your sample results are not too different from the population results.   The Colmar Brunton polls are an example of this. In qualitative research, you are trying to find out the full range of responses and understand people’s thinking: you get smaller numbers of people and ask much more open-ended questions.  Marketing focus groups are an example of this approach.

Just as the clicky website polls are a degraded version of quantitative survey research, the street poll is a degraded form of qualitative survey research.  A good qualitative survey would try to find out how people feel about the tradeoffs in education funding, about where they would rather make cuts, and how views differed between groups of people — do people without children give similar explanations to  those with children, for example.

In this case, however, public opinion is probably clear enough that any measurement will give a similar result, at least as long as you’re not interested in working out what policies would be better.

avatar

Thomas Lumley (@tslumley) is Professor of Biostatistics at the University of Auckland. His research interests include semiparametric models, survey sampling, statistical computing, foundations of statistics, and whatever methodological problems his medical collaborators come up with. He also blogs at Biased and Inefficient See all posts by Thomas Lumley »