Zhanga: April 12, 2009
The comments are at the bottom of the page. You can view old posts by clicking the months on the left sidebar. RSS feed.
« Apr 9, 2009 | All April 2009 posts | Apr 16, 2009 »Sunday, April 12, 2009 (1 comment)
I think it's really stupid when surveys offer stupid crap as incentives. Duke (or parts of it) often sends me emails asking me to participate in surveys in return for some sort of compensation. The econ department is apparently loaded and paid me $35 for less than an hour's worth of work (which I promptly spent on the Tux front license plate). But the following types of things just don't motivate me to participate, and also are bad for your survey:
Filling out the form enters me into a drawing where the top n winners get a new iPhone/other gadget or a $500 gift certificate. Even if the prize is something I want (e.g. a phone that is not an iPhone, a computer, a gift certificate to somewhere where I actually shop, etc), I know there is 0 chance I will actually win anything. Why bother? I'd be more likely to participate if you just handed me $1 cash.
More importantly, having more participants than prizes/compensation means that each individual taking the survey takes on the risk of losing the time investment for no return. Thus, individuals who choose to take the survey will tend to be less risk-averse (i.e. prefer to take on relatively larger amounts of risk) than those who decide to skip it. The survey now has a selection bias that was probably not intended.
Filling out the form causes you to get a $10 gift certificate to some place. This is a little better, because if the gift certificate is to a place where I can actually use it (e.g. textbook store), then it's riskless and equivalent to cash and that's great. But I keep getting emails about $10 gift certificates to the coffeeshop, and I really, really dislike coffee. Even worse, the last survey email I got was offering a drawing for $25 iTunes gift certificates. Not a good way to get a representative sample. Surely those gift certificates cost something. Can't you instead just hand out cash equivalent to whatever they cost?
When a survey hands out compensation, it will tend to get participants who most highly value the type of compensation offered. For example, I'm much more likely to fill out a survey that gives me watermelons than one that only gives out iTunes gift certificates. Dining services has surveys all the time and they always give out Apple products. Is that the respondant base that they're really looking for?
Survey makers have to hand out something to encourage people to respond, and they want to do so at minimal cost. The optimal way to do this would be to say "every respondant gets whatever we can buy them for $10," whether that's $10 cash, a (perhaps) $15 gift certificate, a flash drive, or whatever. Of course you can't actually say that, so you could offer a limited variety of things, such as the ones listed above. And when you're too lazy to do that, you should just offer plain cash — who doesn't like cash? Maybe really rich people don't, but they probably won't care about your iTunes gift card either. The point is, if you have to give out only one item as incentive, flat-rate cash causes the least distortion in your results.
Summary: minimize selection bias. Do not use stupid incentives.
/end rant
10:08PM
Older posts can be viewed through the links to the left, under the main menu.
125 hits since April 13, 2009.
Comments
when i get gift certificates for coffee shops i just buy their cakes and shit usually.
I like getting cash/gift certificates for doing shit.
pat on Monday, April 13, 2009 at 9:31 PM
Add a Comment