>>>>I think that in order for the Confidence level Calc to be reliable, you have to assume the sample was completely random. IOW, if the sample was skewed, i.e. just call households in Scarsdale, NY, that would definitely skew the results.
>>>
>>>Just call at random and you get a random sample of those who still keep landlines. I know a few people who never bothered to get one, cell phone is their only phone. So these are out.
>>>
>>>Next, there are people who never answer any surveys, who have their land line cut off (because they're unemployed and aren't paying their phone bill), those hiding from the law, those whose last phone line was in their repo'd home,... and you end up with a very random sample of people who aren't in any of these categories.
>>
>>Most of the questions raised on this thread regarding these statistics and the way they are compiled are answered if anyone bothered to read the original link : http://www.bls.gov/cps/cps_htgm.htm#where
>>See: "Where do the statistics come from" and "Who is counted as employed/unemployed"
>>And, yes, some in the sample are interviewed by phone - but they are *not* in the sample because they have a phone......
>
>Ah, OK, this is a different survey(*). The phone sample is what most universities do - I know I was in the sample once, and it was obvious from the content of the questions that the survey had a rather biased thesis to prove or refute. Then guess what, within the next few months I got several phone calls from other universities (or the services they hired), which I simply refused to engage, being rather busy at the time. Then they stopped calling altogether. So how random is this? It was maybe random only the first time, which got me on the list, then my refusals got me permanently off the list. It's as random as targeted advertising - you get tagged as "does answer, keep in sample" until you get tagged as "doesn't answer, do not call". That's availability bias and willingness bias, you pick those who you know want to be picked.
>
>(*) the bias in this case is "only where can you send someone with a laptop". Most of the trailer parks are out, communes in the desert are out, unsafe parts of towns are out, people who are too rich and have guards to keep the guys with the laptops away are out, etc etc.

But the sample is selected before there are any attempts to interview. From BLS Handbook of methods:

"Each month, about 72,000 housing units are assigned for data collection, of which about 60,000 are occupied and thus eligible for interview. The remainder are units found to be destroyed, vacant, converted to nonresidential use, containing persons whose usual place of residence is elsewhere, or ineligible for other reasons. Of the 60,000 housing units, about 7 to 8 percent are not interviewed in a given month due to temporary absence (vacation, for example) of the occupants, other failures to make contact after repeated attempts, inability of persons contacted to respond, unavailability for other reasons, and refusals to cooperate (about half of the noninterviews).

Whether the ~8% who were not interviewed results in a real bias to the results is moot. However it seems odd to me that of the 72K houses selected 12K were not occupied - is such a large percentage of U.S. housing stock empty ?

>There's a much simpler way to count unemployment: estimate the workforce by the percentage of people capable of work, based on the last census,

That can't take account of factors such as seasonal variations ?

> and compare that with the social security data. It's been done, actually, and nobody liked the results, because it turned out that the employment is far lower than the sample method shows. The unemployment in the US, calculated as number of able people who don't have a dime paid on their SSNs, was over 25% last time I saw this.

That's just back to the definition of what constitutes 'un-employed'......