of reported usual commutes are by private vehicle.
Analysis snapshot
One survey, three analytical levels.
The analysis keeps households, people, and trips separate so each estimate uses the correct weight and denominator.
Weighted descriptive results
The commute is car-dominant. Access is not evenly distributed.
Income band
Usual commute mode
All reported commuters
Whisker labels show Taylor-linearized 95% confidence intervals.
Vehicle access
All households
8.4%
of households have no vehicle
Mean household vehicles
1.83
95% CI: 7.5%–9.3%
18.6%
of low-income households have no vehicle, compared with 2.9% of high-income households.
Weighted multinomial logit
Income effects after adjustment.
Private vehicle is the reference outcome. Low income is the reference income band. The model also controls for age, sex, education, household vehicles, and urban residence.
Comparison
Adjusted odds ratio
Analysis integrity
The result has an audit trail.
Sample exclusions, model diagnostics, sensitivity analysis, source checksums, and numerical regression tests are generated with the results.
Model sample flow
Primary model diagnostics
Household clusters4,255
Design strata20
McFadden R²0.168
Gradient norm1.4e-7
Sensitivity check
What changes without vehicle count?
Household vehicle count may partly sit on the pathway between income and mode choice. The alternate model removes it while retaining the same 6,044-person cohort.
Reproducible workflow
One question, five tools.
Each tool has a defined role, and every published metric traces back to the official V2.1 public-use files.
- 01SAS
Survey-weighted estimates, Taylor standard errors, and sample sizing.
- 02SPSS
Portable weighted NOMREG specification; Python remains the primary complex-sample inference.
- 03Python
Data validation, clustered sandwich inference, automation, and tests.
- 04Power BI
Star-schema extracts and DAX measures for an interactive report.
- 05Excel
Executive summary, editable charts, model tables, and sample calculator.
Methods that hold up in an interview
The design is part of the answer.
The 2022 public-use release does not contain replicate weights. Estimates therefore use the survey design variables and Taylor-series linearization recommended in the official weighting documentation.
View the official NHTS source- Income
- Imputed household income grouped into <$50k, $50k–$99,999, and $100k+.
- Variance
- STRATUMID with HOUSEID clustering for person-level estimates.
- Mode groups
- Private vehicle, transit/shared, and walk/bike.
- Validation
- Source-file checksums, unique keys, positive weights, sample flow, numerical baselines, optimizer convergence, and automated output tests.
Inspect the work
Download the complete analysis package.
Includes official source data, reproducible Python, SAS and SPSS programs, Power BI extracts, the Excel dashboard, sensitivity analysis, provenance manifest, tests, survey instrument, and interview guide.