State and Local Government Employee-Retirement Systems Fiscal Year 2004 |
||||
![]()
![]() |
Sample Design Description of design: |
|
Prior to Fiscal Year (FY) 2004, the Annual Retirement Survey (ARS) was based on a non-probability sample
design. In 2004 it became a probability sample and was designed to produce estimates for Local as well as State and Local Retirement
systems with a relative standard error of 3 percent or less on the following categories: receipts by source, payments, financial holdings,
membership, and benefit payments. For this sample, three data items of interest helped determine the initial sample sizes: total earnings
on investments, total benefits paid and withdrawals, and total holdings and investments. The sampling frame for the 2004 Annual Retirement
Survey is the 2002 Government Finance Census file, updated with births, deaths, and mergers since fiscal year 2002. Units satisfying the
following criteria were automatically included in the sample with a probability of one. These units represent themselves only and are
called certainty units. |
|
![]() ![]() ![]() ![]() ![]() |
|
After determining the initial sample size, non-certainty governments were selected for the sample based on a probability proportional
to its size, which was based on the variable total holdings and investments. Prior to selecting the sample, the file was ordered by
state and probability of selection. |
|
Weighting: |
|
The weight for each unit in the sample is the inverse of that unit’s probability of being selected into the sample. For example, for
units that were included in the sample with a probability of .02, the weight is (1/.02)=50. For units that were included in the sample
with a probability of 1, the weight is 1. |
|
Sample size: |
|
The sample size is 1076 units of which 830 had a probability of selection equal to one. |