Need some stat-related advice

[Bosh]

Because I'm all geeky like that and I'll have a lot of spare time on my hands in two weeks, I'm thinking of doing an informal study of Korean english-teacher hiring practices in order to check to what extent the standard conventional wisdom that blue eyes are vastly more important than qualifications, in a statistically valid manner.

To do this what I thought I'd do would be to set up some gmail accounts and spam Korean english teacher recruiters with fake resumes (they deserve that being all evil and whatnot) with photos attached and try to see wether the contents of the resume or the attached photo matters more. To do this I'd have to grab some head shots off the internet (of, say, Cute Blonde Girl, Scary Black Dude and Fat Pimply White Guy) write up a few fake resumes (of, say, Smart Guy With Education Degree and ESL-Teaching Certification, Naive Idiot Who Didn't Spell-Check Their Resume and Freaky Wierdo Who Should Never Be Left Alone With Children). If I do three of each and mix and match that means that I have nine "job seekers" and then I'd see which one gets the most responces and go from there.

Some problems:
-I haven't studied stat since HS but I don't want to make any glaring methodological errors.
-There aren't THAT many recruiters. How many would I have to spam to get statistically-meaningful results?
-People here tend to want resumes with phone numbers and passport scans. I really don't want to fake those.
-If the recruiters check to see wether my fake people actually got real degrees they'll know my people are fake (although it would be interesting to see wether claiming degrees from non-existant colleges would have any impact on my results whatsoever). Although it is very doubtful that the recruiters would ever bother...
-Should I check one variable at a time instead of looking at photo vs. resume?
-A lot of recruiters share resumes with each other and a lot of bosses work with more than one recruiter. It might screw with the results if people get suspicious of getting the same photo with different resumes. etc. What would be the best way of dealing with that?

thanks!

[Lul Thyme]

How many you will need to show a difference depends on how big the difference that you are trying to find is (if it exists).
I would think 30 of each type (90-100 total) would be a good start.
About a few of the other points : maybe you can make the degrees from some hard to check place while still being believable...
You are planning on sending 3 cv per person?
If not, then you should juste use one cv that would make things easier.
In both cases you should just study one variable, the photo.


tell us how it goes...

[Urban Ranger]

Originally posted by Lul Thyme
In both cases you should just study one variable, the photo.

Have an idential resume, but use different photos. Maybe a borderline resume...

[Kropotkin]

As stated above, keep it simple. If you want clear results, the photo is really the only thing of importance you should change. That they check or don't check the validity of the degree could be controlled by the same behavior that makes them go for the aryan poster boys - so that's not really a huge problem.

A way to go is to get as a complete list of recuriters as possible (all would be nice - but I suppose impossible). The validiy could be questioned if someone finds a bias in your selection of agencies.