Weights in the CPS
The CPS is designed to be representative of the U.S. population without actually surveying the U.S. population every month. Each individual in the survey is assigned a weight that roughly estimates how many people that individual would “represent” in the entire U.S. population. This is similar to an inverse probability of that individual’s selection out of a group of people demographically similar to them.
Why use weights?
A CPS sample without weights can be useful in certain types of analysis, but it cannot be used to generate many of the macro labor statistics reported monthly by the Bureau of Labor Statistics (BLS). Using weights allows researchers to aggregate the sample up to match macro numbers representative of the U.S. population. For example, an estimate of the population of the United States can be obtained by adding up the weights for all individuals in a given month.
How many weights are in the CPS?
Although the number of weights vary based on the year of the survey (earlier years will have fewer weight options) the most recent iteration of the CPS contains the following weights:
PWFMWGT – Family weight, used for tallying family characteristics in variables that pertain to a family rather than a household or an individual
PWLGWGT – Longitudinal weight, used for adult records matched from month to month in gross flows analysis
PWORWGT – Outgoing rotation weight, used for tallying information collected only during outgoing rotation months such as earnings, detailed information on not in the labor force, and second job industry and occupation
PWSSWGT – Final weight, used for most tabulations, controlled to independent estimates for states, origin, sex, age, and race
PWVETWGT – Veterans weight, used for tallying veterans’ data only, controlled to estimates of veterans supplied by the Veterans’ Administration. This weight should be used to tabulate variables pertaining to veterans.
PWCMPWGT – Composited final weight, used to create the BLS’s published labor force statistics
HWHHWGT – Household weight, used for tallying household characteristics in variables that refer directly to a household
Which weight is the “correct” weight?
There is no single correct weight to use in all circumstances. Weight selection depends on the research question. For instance, research that involves the family or household unit of the survey might select those respective weights. Similarly, research that involves veterans exclusively might consider the veterans weight.
Researchers using outgoing rotation groups or a longitudinal construction should consider those weights. In general, calculations performed on CPS data that involve individuals with no outgoing rotation or longitudinal construction should consider using either the composited final weight or the final weight.