Op-Ed: Estimating Puerto Rico’s population after Hurricane María
On Sept. 20, 2017, Hurricane María made landfall on Puerto Rico, causing widespread devastation.
This disaster not only impacted residents of the island, resulting in increased net out-migration from Puerto Rico to other parts of the United States, but also the quality of data typically used to measure migration patterns.
During 2018, people began to move back to Puerto Rico as the island recovered from Hurricane María, which was not immediately reflected in typically utilized data sources.
As a result, the U.S. Census Bureau modified its existing migration estimation method to produce its Puerto Rico population estimates for both 2018 and 2019.
At the time of the hurricane, the methodology for estimating migration to and from Puerto Rico and stateside used the American Community Survey (ACS) and its counterpart, the Puerto Rico Community Survey (PRCS).
In the wake of the hurricane, the PRCS halted data collection from October to December 2017, and used data from the previous nine months to extrapolate patterns for the entire 12-month period, resulting in an overestimation of in-migration to Puerto Rico.
In addition, the 2017 ACS estimate did not reflect a substantial increase in out-migration from Puerto Rico to other parts of the United States. This was perhaps due to the late timing of the hurricane during the calendar year, the ACS two-month residency requirement for inclusion in the survey, and the tendency of migration data collected from surveys to lag measurement of actual migration events.
Similarly, while the 2018 ACS picked up increased out-migration from Puerto Rico to other parts of the United States, it did not reflect the substantial return migration to Puerto Rico that occurred during the first part of the year.
As a result, the Census Bureau’s usual methodology for estimating net migration between stateside and Puerto Rico could not reliably capture the periods between July 2017 and July 2018 and July 2018 to July 2019.
Incorporating travel data for a better estimate
To address these issues, the Census Bureau modified the ACS/PRCS method by incorporating 2017-2018 monthly Airline Passenger Traffic (APT) data from the Bureau of Transportation Statistics to create the 2018 Puerto Rico population estimates. It then used 2018-2019 monthly APT data to produce its 2019 Puerto Rico estimates.
Given that most travel to and from Puerto Rico is via airplane, we make the assumption that APT data are a reasonable proxy for measuring changes in net migration trends.
While the 2017 ACS did not appear to fully capture the scale of out-migration in the immediate aftermath of Hurricane María, monthly airline data did show a higher level of out-migration from the island to stateside during this time period.
Similarly, while the 2018 ACS did not seem to fully measure return migration to Puerto Rico, monthly airline data showed a high level of in-migration during the calendar year.
The revised method is as methodologically consistent as possible with our previous ACS/PRCS-based estimates. It uses approximately the same time period (calendar year) and migration universe (stateside United States to Puerto Rico) as the ACS/PRCS method, for both the 2018 and 2019 estimates.
Historically, APT data have consistently shown higher net out-migration from Puerto Rico to other parts of the United States than ACS/PRCS estimates.
To account for this inherent difference between data sources, the revised method “blends” ACS/PRCS and APT data. This was accomplished by calculating the ratio of ACS/PRCS-to-APT net migration for Puerto Rico for two recent years, 2015 and 2016, and applying it to net Puerto Rico-U.S. stateside migration measured from both 2017 and 2018 APT data.
Net outflow from Puerto Rico after Hurricane María
The ACS/PRCS method would give us a net outflow from Puerto Rico of 77,321 between 2017 and 2018. By comparison, 2017 APT data alone show a net outflow of 301,304 (and 195,957 for the months of September to December).
The latter is a more accurate reflection of the hurricane’s impact but somewhat high, given that APT data show sizeable return migration in January and February of 2018.
To account for the January return migration, we use the APT time period from February 2017 to January 2018, before applying the two-year (2015-2016) ratio to make it ACS-equivalent.
Shifting our time period one month helps account for return migrants yet keeps most months (11 of 12) within the ACS/PRCS-equivalent calendar year. This modification yields a net out-migration of 215,166 people.
After applying the ratio method and accounting for the return migration shown in the APT data for early 2018, the new method gave us an annual net out-migration of 123,399 people from Puerto Rico to other parts of the United States.
Net inflow to Puerto Rico during post-Hurricane María recovery
To account for the impact of return migration to Puerto Rico, we applied a similar methodology for our 2019 estimates.
The ACS/PRCS method gave a net outflow from Puerto Rico of 112,000 between 2018 and 2019. However, 2018 APT data showed a net inflow of 88,194 for the calendar year, reflecting sizable return migration which required another adjustment to our methodology.
Given that we already included January 2018 APT data (and its sizeable return migration) in our 2018 estimate, we again shifted the time period one month to reflect the period from February 2018 to January 2019 and applied the two-year (2015-2016) ratio to make it ACS-equivalent.
After applying the ratio method, our new estimate gave us an annual net in-migration of 7,773 people to Puerto Rico from other parts of the United States.
This is the first time Puerto Rico has shown positive migration from other parts of the United States since the Census Bureau began producing estimates for Puerto Rico in 2000.
The results of this method change are July 1, 2018 and 2019 estimates of the Puerto Rico population that take into account the impact of Hurricane María, as well as return migration during the island’s recovery period.
There were no methodological changes to adjust the Puerto Rico estimates of births and deaths for hurricane impacts. We used the information on deaths from the Puerto Rico Institute of Statistics as we have done in previous years.
Author Jason Schachter is chief of the Census Bureau’s Net International Migration Branch, while Antonio Bruce is a mathematical statistician in the Population Evaluation, Analysis and Projections Branch.