Skip to content

Validation benchmarks

The data on this page are used only to test MILES; no model is estimated on them. They describe travel and commuting in the forecast year 2024 (the time-use survey, the 2024 American Community Survey) or give an independent check of the base period (the National Household Travel Survey). The ACS microdata used for estimation and calibration are described on ACS PUMS. Bracketed numbers refer to the sources of the data page.

Shares and means are weighted with the survey weights. Cells based on fewer than 10 unweighted records are shown as "—". The study area or Baltimore region is Baltimore City and the counties of Anne Arundel, Baltimore, Carroll, Harford and Howard.

American Time Use Survey

The American Time Use Survey (ATUS) 17 records one diary day (4:00 to 4:00) of one person aged 15 or over per household, sampled from households leaving the Current Population Survey. MILES uses the multi-year files 2003–2024 (respondent, activity, summary and CPS files), public, downloaded from the Bureau of Labor Statistics. The files hold 252,808 diaries of 2003–2024. They test the number and type of activities per person and the commute in 2024 (activity/lib/reporting_2024/bench_atus.py).

Codebook

Variable Description Type Values / categories / units Used in
TUCASEID Respondent identifier Integer (14 digits) Year and month of the CPS interview, household Join key
TUYEAR Year of the diary Integer 2003–2024; used: 2017–2019 and 2021–2024 Period selection
TUDIARYDAY Day of the week of the diary Categorical 1 Sunday … 7 Saturday; used: 2–6 Weekday filter
TRHOLIDAY Diary day was a holiday Binary 0 no, 1 yes; used: 0 Weekday filter
TUFNWGTP Final person weight Continuous Person-days All weighted statistics
TELFS Labour force status Categorical 1, 2 employed; 3, 4 unemployed; 5 not in labour force Employed persons
TEAGE Age Integer Years, 15–85 (top-coded) Age filter (15+)
TESCHENR Enrolled in school Categorical 1 yes, 2 no, −1 not asked Student flag
GESTFIPS State of residence (CPS file) Categorical FIPS code; used: 11, 24, 51 Pool selection
TUACTIVITY_N Episode number within the diary Integer 1, 2, … Episode order
TUSTARTTIM Start time of the episode Time hh🇲🇲ss Activity start, departure
TUACTDUR24 Duration of the episode, truncated at 4:00 Integer Minutes Activity duration, travel time
TRCODEP Six-digit activity code Categorical 05 work, 06 education, 07 consumer purchases, 18 travel, … Activity type, travel episodes
TEWHERE Where the episode took place, or the mode of travel Categorical 1 home or yard; 12 car driver, 13 car passenger, 14 walk, 15 bus, 16 subway or train, 17 bicycle; other places Home or away, commute mode

Statistics

Sample of weekday diaries (Monday to Friday, not a holiday) of persons aged 15 or over. Single years hold about 70 Maryland weekday diaries, so four years (2021–2024) are pooled for the 2024 benchmark and three years (2017–2019) describe the base period; the year 2020 is not used.

Pool Years Weekday diaries, age 15+ Employed With a commute
Maryland 2017–2019 276 182 139
Maryland 2021–2024 279 183 87
DC, Maryland and Virginia 2017–2019 706 476 350
DC, Maryland and Virginia 2021–2024 804 521 255

Activities per person and share of persons with at least one activity of each type, weekday. An activity is a stay away from home after a trip (or a return home), as defined under Processing.

Activity MD 2017–2019: per person MD 2017–2019: with activity (%) MD 2021–2024: per person MD 2021–2024: with activity (%) DC+MD+VA 2017–2019: per person DC+MD+VA 2017–2019: with activity (%) DC+MD+VA 2021–2024: per person DC+MD+VA 2021–2024: with activity (%)
Work 0.60 49.9 0.40 34.2 0.63 51.3 0.40 34.6
School 0.12 8.2 0.10 9.9 0.09 6.6 0.07 7.2
Shopping 0.48 35.4 0.48 30.7 0.46 34.9 0.45 30.7
Other 1.44 64.2 1.10 56.2 1.36 62.4 1.15 56.8
Return home 1.10 73.4 0.91 66.7 1.03 74.5 0.91 65.3
All activities 3.74 89.8 2.98 83.0 3.57 88.9 2.99 79.1

Share of persons with each activity type, Maryland weekday diaries, 2017–2019 and 2021–2024

Share of Maryland weekday diarists aged 15 and over with at least one activity of each type, ATUS.

Commute on the diary day (persons with a work stay away from home):

Pool Years Commuters Employed with work away from home (%) Mean travel time (min) Median (min) Mean departure
Maryland 2017–2019 139 74.5 41.4 35 7:45
Maryland 2021–2024 87 47.1 33.3 30 7:49
DC, Maryland and Virginia 2017–2019 350 73.7 34.7 30 7:45
DC, Maryland and Virginia 2021–2024 255 50.5 36.5 30 7:39

Mode of the trip arriving at work, DC, Maryland and Virginia (share of commuters):

Mode of the trip arriving at work 2017–2019 (%) 2021–2024 (%)
Car, driver 80.9 81.9
Car, passenger 5.3 —
Bus — —
Rail — —
Walk 5.9 3.2
Bicycle — —
Other 4.6 8.1

Processing

  • Diary days on Monday to Friday that are not holidays, respondents aged 15 or over living in Maryland (pool Maryland) or in DC, Maryland or Virginia (larger pool, used where the Maryland sample is small).
  • The diary is cut into stays at every travel episode (activity code 18xxxx). A stay after a travel episode is an activity at the trip's destination; the first stay of the day is not an activity. A stay is at home when its episodes are at home or in the yard (TEWHERE 1); episodes without a place take the place of the previous episode.
  • The type of a stay away from home is the category with the most minutes: work (05), education (06), shopping (07) or other.
  • The commute is the last trip leaving home before the first work stay. Its travel time is the sum of the travel episodes up to the arrival at work, without stops; its mode is that of the trip arriving at work.

National Household Travel Survey

The National Household Travel Survey (NHTS) 18 is the national one-day travel diary of the Federal Highway Administration. Two public files are used:

  • NHTS 2017 (household, person and trip files): an independent check of the base period of the regional survey, for residents of the Baltimore–Columbia–Towson metropolitan area (activity/lib/reporting_2024/obs_surveys.py). The national file holds 129,696 households, 264,234 persons and 923,572 trips.
  • NHTS 2022 (trip file): the trip length of car drivers and the share of business driving, used to check the traffic assignment (assignment/validation/regional_validation.py, coverage_estimate.py). The public 2022 file does not identify the metropolitan area; the closest class is the South Atlantic division, metropolitan areas of 1–3 million with heavy rail. The national file holds 31,074 trips.

Codebook

Variable Description Type Values / categories / units Used in
HOUSEID, PERSONID Household and person identifiers String, integer — Join keys (2017)
HH_CBSA Core-based statistical area of the household Categorical 12580 Baltimore–Columbia–Towson Area selection (2017)
TRAVDAY Day of the week of the travel day Categorical 1 Sunday … 7 Saturday; used: 2–6 Weekday filter (2017)
R_AGE Age of the person Integer Years; used: 5 and over Person filter (2017)
WORKER Person is a worker Categorical 1 yes, 2 no Description (2017)
WTPERFIN, WTTRDFIN Person and trip weights Continuous Persons; trips per year Weighted statistics (2017)
TDTRPNUM Trip number within the travel day Integer 1, 2, … Trip order (2017)
STRTTIME, ENDTIME Trip start and end time Integer hhmm Departure and arrival (2017)
WHYFROM, WHYTO Purpose at the origin and destination Categorical 1, 2 home; 3 work; 4 work-related; 7 change of mode; 8 school; 11 buy goods; others Activity type; business trips (2022)
TRPTRANS Mode of the trip Categorical 2017: 1 walk, 2 bicycle, 3–6 car, van, SUV, pickup, 11 public bus, 15, 16 rail, …; 2022: 1–4 car, van, SUV, pickup Mode; car filter (2022)
DRVR_FLG Person drove on the trip Binary 1 yes, 2 no Car driver
TRPMILES Trip distance (network) Continuous Miles Trip length
TRVLCMIN Trip duration Integer Minutes Travel time
CENSUS_D, MSASIZE, RAIL Census division; size class of the metropolitan area; area has heavy rail Categorical 5 South Atlantic; 4 = 1–3 million; 1 yes Area selection (2022)
TDWKND Travel day on a weekend Binary 1 weekend, 2 weekday Weekday filter (2022)
WTTRDFIN5D Trip weight, weekday Continuous Weekday trips Weighted statistics (2022)

Statistics

NHTS 2017, Baltimore metropolitan area, weekday. 290 households, 596 persons aged 5 or over and 1,982 unlinked trips (1,959 trips after joining changes of mode). 55.1 % of the weighted persons are workers. Trip distance: mean 13.0 km, median 5.3 km, standard deviation 25.5 km; trip duration: mean 25.5 min, median 15 min.

Activities per person (one activity per linked trip, at its destination):

Activity Per person Persons with the activity (%)
Work 0.37 33.1
School 0.14 14.2
Shopping 0.36 26.2
Other 1.27 57.3
Return home 1.25 80.1
All activities 3.40 83.0

Mode of the unlinked trips (share of weighted trips):

Mode (unlinked trips) Trips Share (%)
Car, driver 1,343 61.6
Car, passenger 285 15.0
Public bus 42 5.1
Rail 7 —
School bus 70 3.9
Walk 187 10.6
Bicycle 14 0.4
Other 34 3.3

NHTS 2022, car-driver trips on weekdays. Of 15,300 car-driver trips with a distance in the national file, 131 are in the South Atlantic class used for the comparison and have a positive duration. Their distance has a mean of 16.4 km (median 8.2 km, standard deviation 19.7 km); their duration has a mean of 23.7 min (median 20 min). Work-related trips (WHYTO or WHYFROM 4) make up 3.9 % of the car-driver distance nationally and 0.6 % in the South Atlantic class.

Processing

  • NHTS 2017: households of the Baltimore–Columbia–Towson area (which adds Queen Anne's County to the study area) with a weekday travel day; persons aged 5 or over. Legs that end in a change of mode (WHYTO 7) are joined to the next trip. Destination purpose: home (1, 2), work (3), school (8), shopping (11), other. Distances are capped at 300 km.
  • NHTS 2022: weekday car-driver trips (DRVR_FLG 1, TRPTRANS 1–4) with a positive distance and duration.

American Community Survey 2024, journey to work

The journey-to-work questions of the ACS person file 1 describe commuting in the forecast year: whether a worker worked from home, the mode, the departure time and the travel time to work on the reference week. MILES uses the Maryland person files of the 2024 1-year sample (62,777 records) and of the 2020–2024 5-year sample (299,422 records), both public (activity/lib/reporting_2024/bench_acs.py). The same file series is described for estimation on ACS PUMS.

Codebook

Variable Description Type Values / categories / units Used in
SERIALNO Housing unit identifier String Year of interview in the first four characters Join key; years of the 5-year file
PUMA Public use microdata area of residence (2020 definition) Categorical 5-digit code; nested in counties of the study area County, Baltimore region
PWGTP Person weight Integer Persons All weighted statistics
PWGTP1–PWGTP80 Replicate weights Integer Persons Standard errors (successive differences)
AGEP Age Integer Years Workers aged 16+
ESR Employment status Categorical 1, 2 civilian employed (at work, with a job not at work); 4, 5 armed forces; 3 unemployed; 6 not in labour force Workers; at work last week (1, 4)
RELSHIPP Relationship to the reference person Categorical 20–36 household members; 37, 38 group quarters Household population
JWTRNS Means of transportation to work Categorical 1 car, truck or van; 2 bus; 3 subway; 4 commuter rail; 5 light rail; 6 ferry; 7 taxi; 8 motorcycle; 9 bicycle; 10 walked; 11 worked from home; 12 other Mode; worked from home
JWRIP Vehicle occupancy Integer 1 drove alone, 2–10 persons in the carpool Driver and passenger shares
JWMNP Travel time to work Integer Minutes, 1–200 Travel time
JWDP Time of departure for work Categorical 150 intervals of 5–30 min (decoded with the data dictionary) Departure time
POWSP, POWPUMA Place of work, state and PUMA Categorical FIPS state; 5-digit place-of-work PUMA Commute flows

Statistics

Household population (group quarters excluded). Worked from home: share of workers at work in the reference week. Travel time and departure: commuters (at work, not working from home).

Vintage Area Persons (records) Workers (records) Workers (weighted) Worked from home (%) Commuters (records) Mean travel time (min) Median (min) SD (min) Mean departure
2024 (1-year) Maryland 60,090 31,485 3,260,281 15.7 25,566 32.3 30 24.0 8:14
2024 (1-year) Baltimore region 25,966 13,693 1,452,629 14.9 11,243 30.4 25 22.9 8:16
2020–2024 (5-year) Maryland 284,800 146,634 3,168,183 19.3 113,871 31.2 25 23.0 8:15
2020–2024 (5-year) Baltimore region 122,483 63,327 1,421,624 18.1 50,061 29.3 25 21.8 8:19

Mode to work of commuters, 2024 1-year (a car-pool of k persons counts 1/k as driver):

Mode Maryland Baltimore region
Car, driver 84.1 85.9
Car, passenger 5.9 5.5
Bus 2.6 2.5
Rail 3.2 1.5
Walk 2.1 2.3
Bicycle 0.4 0.4
Other 1.9 1.9

By county of residence, 2024 1-year:

County Persons (records) Workers (weighted) Worked from home (%) Commuters (records) Mean travel time (min) Car driver (%) Bus and rail (%)
Baltimore City 4,436 280,624 14.1 1,935 30.5 72.0 10.8
Anne Arundel 5,787 321,571 15.2 2,622 30.9 89.6 2.2
Baltimore County 7,669 438,021 13.7 3,288 27.9 87.3 2.8
Carroll 2,007 89,114 12.9 853 36.7 91.1 0.8
Harford 2,759 139,761 11.4 1,212 32.9 92.7 0.8
Howard 3,308 183,538 22.0 1,333 29.8 90.1 3.1

Distribution of the travel time to work, Maryland and the Baltimore region, ACS 2024 1-year

Travel time to work of commuters living in Maryland and in the Baltimore region, ACS 2024 1-year PUMS (weighted; the classes have unequal widths).

Processing

  • Workers: ESR 1, 2, 4 or 5 and age 16 or over. Counties are assigned from the 2020 PUMA with the PUMA-to-county lookup of the 2024 code list; in the study area every PUMA lies in one county.
  • The departure time is the midpoint of the JWDP interval. Mode: drove alone counts as driver, a car-pool of k persons as 1/k driver and (k − 1)/k passenger, motorcycle as driver; subway, commuter and light rail as rail; ferry, taxi and other as other.

American Community Survey summary tables

Published ACS estimates 1 (detailed tables, with 90 % margins of error) are the benchmark where the microdata are too coarse: by county for commuting and by census tract for households and housing. They were downloaded from the data.census.gov table service as one JSON file per table and county (60 county files; 114 tract files).

  • County tables, 2024 1-year and 2020–2024 5-year (assignment/validation/regional_validation.py): B08134 (travel time to work by mode), B08136 (aggregate travel time), B08301 (means of transportation), B08302 (departure time) and B08303 (travel time) for the six study-area counties. Tables the Census Bureau does not publish in the 1-year release are empty: B08134 Carroll, B08134 Harford, B08134 Howard, B08136 Carroll, B08136 Harford, B08301 Harford.
  • Tract tables (reports/validation_tracts/tract_validation.py): 2012–2016 5-year on 2010 tracts (B08202, B11001, B19001, B25003, B25007, B25008, B25009, B25032, B25044) and 2020–2024 5-year on 2020 tracts (B01003, B08201, B08301, B11001, B19001, B19013, B25003, B25008, B25032, B25044), all tracts of the six counties.

Codebook

Variable Description Type Values / categories / units Used in
B08134_001E, B08134_021E–_030E Workers 16+ who did not work from home; those who drove alone, by travel time Count Workers; bins of minutes Travel time to work by county
B08136_001E, B08136_003E Aggregate travel time to work, all and drove alone Count Minutes Mean travel time (aggregate ÷ workers)
B08301_001E–_021E Means of transportation to work Count 003 drove alone, 004 car-pooled, 010 public transport, 018 bicycle, 019 walked, 021 worked from home Mode shares
B08302_002E–_015E Time leaving home to go to work Count 14 intervals, 0:00–23:59 Departure profile
B08303_002E–_013E Travel time to work Count 12 bins, < 5 to 90+ min Travel-time distribution
B11001_001E Households Count Households Households by tract
B25003_001E–_003E Tenure Count 002 owner, 003 renter occupied Tenure
B25008 Population in occupied units by tenure Count Persons Household population
B25044_001E–_010E Tenure by vehicles available Count 003, 010 no vehicle (owner, renter) Car ownership
B19001, B19013 Household income in bins; median household income Count; dollars 16 bins; inflation-adjusted dollars Income
B25032 Tenure by units in structure Count Dwelling types Dwelling type
B01003, B08201, B08202, B25007, B25009 Population; household size by vehicles; household size by workers; tenure by age of householder; tenure by household size Count Persons, households Tract composition
…M 90 % margin of error of each estimate Count Same as the estimate Tolerance of the comparison

Statistics

County commuting tables (estimates; "—" = not published in the 1-year release):

County 5-year: workers not at home 5-year: mean travel time (min) 5-year: drove alone (%) 5-year: public transport (%) 5-year: worked from home (%) 1-year: workers not at home 1-year: mean travel time (min) 1-year: drove alone (%) 1-year: public transport (%) 1-year: worked from home (%)
Baltimore City 222,441 29.1 55.5 10.3 17.8 243,705 30.2 56.5 9.7 14.0
Anne Arundel 257,880 28.9 70.0 1.6 18.5 273,560 30.6 70.6 1.8 15.6
Baltimore County 351,951 28.0 69.5 2.5 16.5 375,305 28.1 70.8 2.6 13.6
Carroll 75,334 34.4 73.9 0.3 17.3 — — 74.7 0.6 13.2
Harford 115,350 32.1 75.1 0.6 14.9 — — — — —
Howard 129,859 28.5 64.9 1.5 26.0 — — 66.2 2.5 23.0

Tract tables, households and housing:

Vintage County Tracts Tracts with households Households Mean households per tract Owner occupied (%) No vehicle (%)
2012–2016 Baltimore City 200 198 242,416 1,224 46.6 29.4
2012–2016 Anne Arundel 105 102 204,829 2,008 73.9 3.9
2012–2016 Baltimore County 214 211 312,826 1,483 65.8 8.0
2012–2016 Carroll 38 38 60,449 1,591 81.2 4.4
2012–2016 Harford 57 57 92,497 1,623 78.1 4.9
2012–2016 Howard 55 55 109,872 1,998 73.4 3.6
2012–2016 Study area 669 661 1,022,889 1,547 65.7 11.3
2020–2024 Baltimore City 199 198 255,668 1,291 47.5 26.7
2020–2024 Anne Arundel 130 127 224,748 1,770 75.1 4.1
2020–2024 Baltimore County 219 217 332,801 1,534 66.4 7.8
2020–2024 Carroll 37 37 64,285 1,737 84.3 3.3
2020–2024 Harford 59 59 101,795 1,725 79.7 4.4
2020–2024 Howard 59 59 121,065 2,052 71.5 3.6
2020–2024 Study area 703 697 1,100,362 1,579 66.6 10.4

Processing

  • Values below −100,000 are the Census Bureau's annotation codes (for example a median that cannot be computed) and are treated as missing.
  • Tract results are compared on clusters of tracts and zones, so that 2010 and 2020 tracts and the zones nest; margins of error of sums are combined as the square root of the sum of squares.