The second myth I encountered at my recent EARNet conference is that
2. You cannot generalize from 22 interviews but rather you need 100’s! or at least 30 per the central limit theorem – an argument I am not certain comes close to applying to a non-randomly selected purposive sample.
Here is the opposite side of the coin that “12 is enough” made by some. Just as I poured cold water on the “12 is enough” myth in the previous post, I also need to pour cold water on this one!!!!
Generalization as always depends on theory and its interplay with the data. If the theory is based on expertise and there are only a limited number of such experts, 12 might be a substantial portion thereof in many audit applications. If theory suggests (see for example my negotiation dyad study in 2008 in AOS based on 5 dyads) generalizability dependent on the existence of the phenomenon with hard to access participants (like real dyads in ACM negotiation) and it can be confirmed with other data then 10 might be enough to generalize.
One thing is certain, you do not need 100’s of audit partners to generalize to the audit partner population. Heck, even the US presidential pollsters had it right 48% Clinton 46% Trump (but because of the strange quirk of the US system (one person one vote it is NOT) Trump won the electoral college with 3 million fewer votes than Clinton). These polls of the US population generally had 1200 to 2400 interviews – enough that one can generalize to 130 million Americans who voted and have a reasonably accurate breakdown on gender, ethnicity, etc. So do not tell me we need 100’s of interviews of audit partners to generalize in qualitative studies unless this is purely an inductive study with no theorizing at all. Even then I would find it hard to believe that audit partners are so heterogenous that more than 30 or 40 are needed.