Since it is presently impossible to determine the actual number of unemployed persons in America with a 100% certainty, then it is impossible to determine how accurate a measurement is compared to another measurement.[/qutoe] First, you're changing the topic from Employed to Unemployed. Second you are again ignoring the issue of different sample sizes. Third, you're ignoring that, for the Establishment survey we do have a near exact count from the QCEW.
It's like if I think of a number and you come up with two methods to estimate what that number is. Then you say that method 'a' is more accurate then method 'b'...even though you have absolutely no idea what the number is.
It is impossible to know which method is more accurate since you do not even know the number you are trying to determine.
No, it is nothing like that. It's like there is a giant jar with about 1 million marbles in it of red, white, and blue. You pick a sample of 500 marbles, count the number of each color, and use the percent of your sample to estimate the total number of each color (for example, out of the 500 you pick, there are 100 white, which is 20%....20% of of 1 million is 200,000 so that's your estimate). I pick a sample of 320,000 marbles and get a count of 48,000 white marbles (15%) giving me an estimate of 150,000 white marbles.
Whose estimate is more likely to be closer to the true number?
Do you really want to claim sample size doesn't matter and we could go with a sample of 100 households and get the same accuracy as the current 60,000 households?