Can Someone Put This In Perspective?

Well, first of all they interviewed people who’ve been working for four years or longer, so it’s at least misleading to generalize on the general working population (”Befragt wurden 1.063 Personen, die im Berufsleben stehen und über mindestens vier Jahre Berufserfahrung verfügen”). It says ‘representative study’, but doesn’t tell for which general population: Representative for the whole german population? Only the working population? Only the working population with 4 years or more experience? So there might actually be a bias, but you can’t tell from those information.

Regarding your question: Assuming this is a representative study (every member of the general population you want to generalize on has had the same chance to get into the sample), then the number of interviews would not matter for the percentages they give – it would only influence the margin of error (since the more people you ask in a representative study, the smaller is the probability you miss the real distribution of openBC-users etc. by chance).

So a higher number of respondents does not necessarily guarantee results that are “more true”, since you might have a systematic bias in the way you collect those answers. By announcing a survey on a website, for example, not every Internet User has the same chance to get into the sample – you will miss those who don’t use the web or that particular website. Asking 100 or 100.000 people will not change that.

[probably not the best answer, and there might be people able to express this better in english.. ]

Jan, this is a pretty interesting feedback and thank you for putting this in perspective. I am reading your blog from time to time, so I know that you are a kind of expert for this topic

PS: I changed your status on openBC/XING ;-D

Daniela,

the number of observations is sufficient to be representative, if the participants of the study are selected carefully, namely stratified along different dimensions, as geography, education, age, and the like. The G-SOEP for example only consists of a couple of thousand observations, and is nevertheless representative for Germany, and I would bet that the restrictions imposed (like 4 years of experience) would lead to a comparable number of observations in this “scientific” data base.

You are right Stefan, got it!

Although I have to admit that for me it always leaves a bit of a strange taste, if only “a few” people have been asked and the message sounds like “everyone”