**The Current Population Survey**
The *Current Population Survey* is a survey of households conducted by the
Census Bureau each month for the Bureau of Labor Statistics. The Census Bureau
carries out all phases of the CPS data collection process, while BLS analyzes
and publishes the data. The data from the CPS are used to produce the labor
force estimates directly for the United States and indirectly for the individual
states. Nationally, there are approximately 78,000 eligible households in the
sample. In Iowa, the sample size is approximately 1,288 households.
The CPS household survey
manages to reduce costs while collecting information about the characteristics
of an entire population. There are several ways to obtain information about a
group, ranging from asking one individual to asking every individual within the
group. In the LAUS program, the labor force status of individuals is classified
according to the concepts defined in the previous section. The most accurate way
to do this would be to survey all individuals within an area to determine their
labor force status. Unfortunately, this process would be time-consuming and
very expensive. Therefore, a smaller group within the larger population is
interviewed. Statistical theory asserts that if the smaller group is an accurate
representation of the larger group, then generalizations can be made about the
larger group, using the data collected from interviews with the smaller group.
This method of obtaining information saves time and money, while producing
results comparable to what would be expected had the entire population been
interviewed.
Households in the CPS are
initially selected using a probability design. In theory, each household in Iowa
has an equal chance of being chosen initially to participate in the survey. The
basic rationale behind the selection process is that, within each area, strata
are identified and targeted to ensure that the CPS is reflective of the area's
demographic and socioeconomic characteristics.
An important feature of the
CPS is that there is a "4-8-4" sample rotation methodology. Using this process,
a selected household is questioned for four consecutive months, dropped from the
sample for eight months, and then returned to the sample and questioned for
another four months. Households are rotated in and out of the sample at
different times. Seventy-five percent of the households remain the same from
month to month. Fifty percent of the households remain in the sample from the
prior year. Once a household permanently leaves the sample, it is replaced by
its neighbor to ensure that the CPS remains reflective of the area's demographic
and socioeconomic characteristics.
Two important
characteristics of the CPS are changing reliability and sample overlap. The
design of the CPS in regard to the states raises important questions about the
reliability of the results. "Reliability" can be defined as how confident we are
that our results meet a certain level of accuracy and are sensitive enough to
detect real changes of a given magnitude. Due to the sample design of the CPS,
there is a low reliability in these estimates. The low number of households
surveyed in Iowa, given the way that the sample is conducted, causes volatile
swings in the month-to-month data, thereby masking real change.
The sample overlap (4-8-4
rotation method) causes something called "autocorrelated sampling error."
Because of the way households are rotated in the CPS, periods of overestimation
and underestimation of the "true" value result. The "true" value is the
unobserved value that would result if the entire population were surveyed. In
other words, since 75 percent of the households remain the same from
month-to-month, a sample that is not representative of the entire population one
month is likely to carry over into the next month. This "error" that carries
over to the next month results in an extended period of time where the estimated
value from the CPS diverges from the true value.
The design of a survey
determines how much error the results will contain. The design of the CPS allows
for high levels of error to show up in the results. Although this causes a high
level of variation in CPS data, this does not necessarily mean that the results
completely lack value. There is some valuable information within CPS estimates.
Since this valuable information is not immediately salient, it must be extracted
using statistical methods. In 1989, a time-series model was developed to provide
labor force estimates for 39 small states and the District of Columbia. In
1996, this approach was extended to the larger states as well. The reason for
using the time-series modeling was to reduce the high variability in monthly CPS
estimates due to small sample sizes. |