PSU are groups consisting of states counties, or groups on contiguous counties.
all surveying take place in the PSU
It's the first stage in sampling Employment
Stratas
The strata does not cross state line
the PSU are put into stratas
Ultimate Sampling Units
USU is the second stage in sampling Unemployment.
each USU consist of a cluster of about 4 adjacent housing units
each person 16 and older who is in USU is selected to represent the entire USU
Sampling
Each person 16 and older who is in PSU and USU is selected to represent the Current Population Survey.
The whole US sampling rat is 1-2000. It is bigger in state that have more people like california 1-3000
Bias
When bias operates more of less evenly across the same it cannot be detected just by looking at the sample data. Standard errors ignore that kind of bias.
Panel Bias
Peoples responses change after a result of being interviewed.
Employment Classifications
Employed (have a job)
Unemployed (Don't have a job and currently) looking for one)
Outside the Labor Force (The state of being neither) employed or unemployed
Employed asked hours work and what type
Unemployed asked last job why the left and how they are looking for work
Outside the labor forced asked if they are keeping house going to school or unable to work for some other reason
Civilian Labor Force
Labor Force = consist of those who are either working or unemployed.
Labor Force = Working + Unemployed
Employment rate: = Unemployed (Divided) by Labor Force
Half -Sampling
Standard errors for cluster samples can be obtained by the half-sample method splitting the method into two halfs and seeing how well they agree.
It makes another survey following the same rule and compare the two. The difference is then added
CPS = 150.2
Independant Survey = 150.7
(150.5 + 150.7)/2 = 150.6
the two difference was due to chance error.
Weighted
The Survey re weights the the sample so it agrees with census data on age sex race state of residence and certain other characteristics influencing employment status
If one group (white males) overrepresented is the survey more than the general population they are weighted to compensate for the Chance Error
Formula
The formulas for the standard error have to take into account the details of the probability method used to draw the sample. the formulas which apply to simple random samples will usually underestimate the standard errors in cluster samples