The Current Population Survey is a monthly survey of a sample of households from across the country, conducted by the United States Census Bureau and used extensively by the Bureau of Labor Statistics as a tool to characterize the labor force. Each household selected for participation is interviewed eight times:
If the people in a house move, the survey does not follow them, and the collection of their information ceases. Generally one member of the household reports for the all members included, but each member is a single record in the public data releases. On average, the CPS samples one person for every 2500 people in the population.
The CPS provides detailed information on employment based on a number of metrics, including work experience, occupational mobility, employer benefits, labor union participation and more. The data also includes information on potential workers who are unemployed, defined as those who are available to work and are searching for work, but are unable to find it. Each person or household surveyed is assigned several weights to represent how much of the population they are estimated to represent, with the choice of weight dependent on the aggregation desired. It's tempting to pull highly specific information about employment from the CPS surveys, but we find that multiple years should be combined to create meaningful detailed statistics.
As with other Census data, these space-delimited files are not easily accessible to novice users, but the data can be viewed in a relatively friendly format through IPUMS. We download the microdata files from The Data Web at the Census Bureau. All monthly survey results along with supplemental surveys are available on this page.
Since CPS interviews its participants eight times over two calendar years, the data affords us a terrific opportunity to explore career changes.
Anecdotally, we observed that some people seemed to swap careers back-and-forth over two years, or float in and out of retirement. We are not sure as end-users of the household data the extent to which these people are truly changing jobs and status, or whether their opinion of their exact classification wavers instead. Perhaps this might explain why many of job transitions we observe show a similar percentage of certain jobs as both prior and next careers; perhaps some jobs are really quite similar rather than part of career advancement.
Because we made a number of processing decisions to create the conclusions in our career transitions section, we provide notes that would allow replication of our results following.
In determining the percentages of prior and next careers, we first grouped all eight (or fewer if the participant moved during this process) interviews for each person that were conducted between January 2013 and January 2018. We experimented with several approaches, and in the end we felt annual rates gave the most reasonable information. Therefore, we recorded each person's career in their first interview, and their career in their fifth interview which occurred one calendar year later. Any people lacking a fifth interview were omitted from the data. There are multiple choices of weights over the course of the eight interviews, and we chose to weight the career pairs by the final weights for the fifth interview. These weights reflect any re-balancing needed as households drop out of the longitudinal survey. We considered using the longitudinal weights, but with our choices of pairs these did not seem valid in our application.
By aggregating our pairs by their respective weights, and treating the first entry as a prior career and the second as a current career, and separately using the same pairs to suggest a journey from a current career to a next career, we were able to collect counts for each job of the weighted sums of prior and next careers. From these counts, we create the percentages reported in our careers pages. We set 1% of the total weight as an arbitrary cut-off in the data we present. Otherwise, it's easy to suggest that someone could become an astrophysicist by working as a veterinarian, becoming a dog-walker, and then transitioning to astrophysics. Individuals do amazing things, and we do not want to share unique personal data in our career transitions pages.
Every March, as a supplement to the CPS, the US Census Bureau issues a survey covering social and economic characteristics of approximately 75,000 households. This survey provides great detail on income, benefits, taxation, and related information. This March supplement is called the Annual Social and Economic Supplement (ASEC).
In an effort to obtain as much data as possible on which to report statistics, and with an eye to the five year time-frame of our ACS microdata data which we use to report information by various demographics, we combined the most recent five years of ASEC data to provide occupation-specific information on union participation, certification requirements, and employer/union healthcare and pension benefits.
The percentages we report are based on the ASEC-weighted sum of all responses to the relevant questions in the ASEC survey. While ASEC, as well as other parts of CPS, reports demographic and salary information, Ididio uses the larger ACS survey for those estimates as well as BLS-reported totals, some of which include CPS data as part of the BLS analysis.