This service is for faculty and staff.
List of FAQs that explain the Census PSEO dashboard data and its security.
#1: What is the Census PSEO tool?
The Office of Higher Education (OHE) is collaborating with the University of St. Thomas to participate in the U.S. Census Post-Secondary Employment Outcomes project (Census PSEO). The Census PSEO project is a data sharing project that joins University enrollment and awards to national employment data within the Longitudinal Employer-Household Dynamics (LEHD) program at the U.S. Census Bureau. The U.S. Census PSEO Explorer tool -- an online tool that allows users to easily interact data showing the annual national earnings for their graduates by major field, degree level, institution, and graduation cohort. Specifically, for each of those cells, the U.S. Census will provide the 25th, 50th and 75th percentile of earnings for 1, 5, and 10 years after graduation. Census PSEO will also provide the number of graduates employed in each of these cells. These tabulations contain counts of the flow of graduates from institutions and majors to employer industrial sector and geography nationally.
#2: What do the ipeds_count variables in the earnings file represent? And do the ipeds_count numbers include one record per each major (so, 2 for a double major)?
The IPEDS counts variable is just pulled directly from IPEDS, and are the number of degree records for the year span. Because it is degree conferral, I believe that it is one record per major, but it may not count secondary majors. We provide them as a useful baseline, but the timing of them vs our data is off, because the IPEDS is reported by academic year, and our data are reported at the calendar year.
#3: We see instances where the grads_earn number exceeds the corresponding ipeds_count – but if the ipeds_count is the cohort, we are thinking the corresponding grads_earn counts should never exceed the ipeds_count, is that correct?
That can certainly happen, given that we inject noise into the counts. It is uncommon, but not unheard of. Additionally, we know there are reporting issues with the IPEDS data (we used to do something about that, but the issues are too widespread), which cause there to be an undercount.
4: Icases where all but one cohort year has earnings suppressed, we are seeing that the all-years row sometimes has different earnings – why is that?
We suppress cells not for privacy reasons, but for accuracy reasons, and so when we include the aggregations, we include those suppressed cells. The details for how we protect this data are here: https://lehd.ces.census.gov/doc/pseo_appendix_tech_doc.pdf
#5: What year’s dollars are the earnings figures in? And through what year/quarter?
The earnings are expressed in 2020 dollars. We use earnings information through the end of 2020 (we always update earnings to be expressed in last year of earnings used).
#6: Do you have something that explains the perturbation applied/noise added to the data? Also, the cell sizes go down to 1. Is that intentional because perturbation is applied?
This is a great question. The answer has two parts:
Additionally, we suppress cells under 30, but not for confidentiality (since all cells have same protection), but for accuracy (results are very noisy on average in very small cells).
- for flows, we consider the entire matrix of possible destinations for workers (state x industry sector) and then add noise to each cell individually, even if it is a true zero. So a 1 is already noisy (could be zero, 2, 3 or 4, as well as one). We are actually working on improving this methodology since it could be more accurate - that will come out when we do a bunch of really big updates to PSEO in a few years (trying to roll it all out at once).
The entire protection system is described in a technical appendix which is on our website. The link is here: https://lehd.ces.census.gov/doc/pseo_appendix_tech_doc.pdf
#7: In the LEHD PSEO tool, if a student has earned a graduate degree, do they still show up in the earnings data for their bachelor degree institution by CIP?
Our unit of observation for PSEO is the person-degree-field, so if you earn multiple degrees, you show up multiple times - same thing with double majors. For example, if my schools were in the data, I would be there 4 times: BA in Math, BA in Economics, MA in Economics, PhD in Economics.
MN Questions on Data Security & Upload Process
#1: Whose sFTP is it? OHE or U.S. Census Bureau?
Census maintains the SFTP.
#2: Are SSNs encrypted in the transferred spreadsheet or file itself (encrypted at rest), not just during the sFTP upload (encrypted in transit)?
Encrypted during upload
#3: How often are incoming files cleaned off the sFTP server?
They are immediately cleaned off the server.
#4: Are sFTP servers checked periodically to ensure they are correctly configured, updated, patched, and accessible only via secure, complex credentials?
Yes.
#5: Who has access to raw data containing SSNs within the U.S. Census Bureau?
Only a team of about 5 people. Even I never see SSNs. Only the data intake team and the PVS team have access to them.
#6: How is data containing SSNs transferred securely within the U.S. Census Bureau network?
It’s all behind the Census firewall - I don't know the specifics, but it is incredibly secure. No one has access to files they are not explicitly approved for. Even I can only see less than 1% of the data Census holds.
#7: After SSNs are “stripped out,” is the original file destroyed?
It is maintained on the server, but anyone who wants to have it has to request it, and those requests are denied unless the Associate Directors approve the use. This would never happen (but we keep them if we have to re-analyze the PVS process for instance, which did happen in a recent submissions).