> Just learned an interesting polling term: acquiescence bias.
> Responses to agree/disagree statements are considered unreliable
> because people (Americans?) tend towards agreement.
While I was originally introduced to acquiescence bias and its importance in understanding consumer satisfaction polls, there is a related bias that is also called "acquiescence bias" that I learned about when looking at health insurance surveying.
I questioned the "conventional wisdom" amongst certain health policy people that the Current Pop Survey (the most widely used source for health insurance info) was overestimated the uninsured because some people are confused about health insurance. The "confusion" argument is a fair one as people often can gain/lose coverage, have restrictions, have not been sick recently, or get free care. The question is why would the CW point to an overstatement of the uninsured; thinking purely logically wouldn't be people when asked if they had insurance just say yes because most people sort of think they do (even when their employer or government may have booted them off coverage with varying levels of clarity after the fact)? Why would ALL the confusion be people who really had coverage failing to remember they did? There was some major hackery spilled over that question.
Anyway, the good people at the Urban Institute and eventually the CPS people ADDED a confirmation question of all people who indicated they did not have insurance: that is, "Hey you said you were some sort of loser without insurance, are you sure?" That leads me to the second type of acquiescence bias--repeating a question leads significant percentages of people to "Assent" (this is also called "assention bias")--change their answer in order to "please" the apparent "need" of the surveyor. Children seems to intuit "assention" bias at a pretty young age and often continue to ask their parents ("can I have a cookie?") repeatedly.
There is also a "negation bias" in other people--they change their answer in order to go against the surveyor. Anyway, while the majority of people who are requeried on any question will repeat themselves, a lot of people change their answer independent of actually changing their mind (or recalling differently).
Note that people who say they are insured are NOT requeried. Even if the amount of "correction" in requerying the uninsured is greater than the amount of "assention/negation" noise, it is pure hackery not to ask both ways.
Luckily, I personally don't think the number of the insured is that political of a number (it is political but it isn't central to understanding real access or insurance problems). There is just so much crappy insurance out there right now and looks like crappy insurance is the wave of the future.
Jim
"5am Here I am Walking the block To TableTalk You could cry or die Or just make pies all day I'm making pies"
--Patty Griffin