Group Quarters Handling in the PopulationSim Pipeline

Overview

The pipeline uses a single unified process for both regular households and non-institutional group quarters (GQ) persons. Because person totals in TM2 include both household persons and non-institutional GQ persons, running them through a single PopulationSim pass is simpler and keeps control totals internally consistent.

A single PopulationSim run means:

• All person records come out of one synthesis step
• Control totals are internally consistent by construction
• The hhgqtype field on each record already carries the information needed to distinguish regular households from GQ in any downstream analysis

---

hhgqtype Values

hhgqtype	Meaning
0	Regular household (TYPEHUGQ == 1)
1	University / college GQ (noninstitutional, identified by college enrollment SCHG 15–16)
2	Military barracks, group homes, and other noninstitutional GQ

Institutional GQ (TYPEHUGQ == 2: nursing homes, correctional facilities, etc.) are excluded from the seed entirely and never synthesized.

---

Implementation Details

1. Seed Population (create_seed_population.py)

Household-level pass — _create_group_quarters_type():

1. All records with TYPEHUGQ == 2 (institutional GQ) are dropped.
2. All remaining noninstitutional GQ records (TYPEHUGQ == 3) are assigned hhgqtype = 2 as a placeholder.
3. Regular housing units (TYPEHUGQ == 1) are assigned hhgqtype = 0.

Person-level refinement — done during person processing:

• Persons in a GQ household who are enrolled in college (SCHG 15 or 16) are reclassified to hhgqtype = 1 (university GQ).
• All other noninstitutional GQ persons remain hhgqtype = 2.

This two-pass design is necessary because university GQ identification requires person-level enrollment data that is not available at the household record level.

2. Controls (create_baseyear_controls.py)

Three GQ-related controls are generated:

Control	Geography	Covers
numhh_gq	MAZ	All synthesized household records, including noninstitutional GQ units
numhh	MAZ	Regular households only (hhgqtype == 0)
gq_pop_region	Region	Non-institutional GQ persons (soft constraint; ~85% of ACS total to exclude institutional)

Household-level demographic controls (size, income, workers, children) are computed against numhh (regular households only). Person-level controls (age, occupation) apply to the full synthesized population, including GQ persons.

3. Post-processing (postprocess_recode.py)

Both synthetic households and synthetic persons come out of a single PopulationSim run. postprocess_recode.py recodes and renames columns and writes:

• synthetic_households_recoded.csv — all household-type records (including GQ units)
• synthetic_persons_recoded.csv — all persons; hhgqtype is preserved so downstream tools can filter to regular-household persons or GQ persons as needed

---

Last updated: February 2026