1 of 25

ActivitySim Explicit Telecommute Implementation

Sijia Wang

June 30, 2026

WSP | Explicit Telecommute | June 30, 2026

1

This Photo by Unknown Author is licensed under CC BY-SA-NC

2 of 25

Agenda

  1. Recap
  2. Model Design, Software Implementation, and Model Estimation
  3. Next Steps

2

WSP | Explicit Telecommute | June 30, 2026

3 of 25

1. Recap

3

WSP | Explicit Telecommute | June 30, 2026

4 of 25

Two New Components

  • Telecommute Arrangement
    • Applied to all workers
    • Do they have in-home work activity: yes or no?
    • The model classifies workers on simulation days as:
      • Commuting: CDAP == “M” and no in-home work
      • Hybrid: CDAP == “M” and yes in-home work
      • Teleworking: CDAP == “N” or “H” and yes in-home work
      • Not working: CDAP == “N” or “H” and no in-home work
  • In-Home Work Activity Duration
    • Total duration of time spent working while at home

4

Work From Home

Coordinated Daily Activity Pattern (CDAP)

Workplace Location

Telecommute Frequency

Telecommute Arrangement

Conceptual Flow of Selected ActivitySim Components

Workers

Yes

No

Workplace Zone

0, 1, 2, 3, or 4+

Mandatory

Non-mandatory

Home

Commuting

Hybrid

Teleworking

Not Working

In-Home Work Activity Duration

Total Teleworking Time

Out-of-Home Tour Models

WSP | Explicit Telecommute | June 30, 2026

5 of 25

Updated Names for Clarity

  • Telecommute Telework Arrangement
  • In-Home Work Activity Telework Duration

5

Work From Home

Coordinated Daily Activity Pattern (CDAP)

Workplace Location

Telecommute Frequency

Telework Arrangement

Conceptual Flow of Selected ActivitySim Components

Workers

Yes

No

Workplace Zone

0, 1, 2, 3, or 4+

Mandatory

Non-mandatory

Home

Commuting

Hybrid

Teleworking

Not Working

Telework Duration

Total Teleworking Time

Out-of-Home Tour Models

WSP | Explicit Telecommute | June 30, 2026

6 of 25

Requirements

  • Telework Arrangement will be an MNL model
  • Telework Duration will allow a switch between:
    • Simple probability distribution based on empirical analysis, or
    • Choice model
  • Use SANDAG’s 2022 post-COVID household travel survey for estimation
  • Minimal increases in runtime and memory

6

WSP | Explicit Telecommute | June 30, 2026

7 of 25

2. Telework Arrangement�Model Design, Software Implementation, and Model Estimation�

7

WSP | Explicit Telecommute | June 30, 2026

8 of 25

Telework Arrangement – MNL

  • Choosers
    • All workers
  • Alternatives
    • Has in-home work activity
    • No in-home work activity
  • General Sequence
    • After CDAP, before tour and trip models
  • Output
    • persons["has_in_home_work_activity"] = True or False

8

WSP | Explicit Telecommute | June 30, 2026

9 of 25

Annotation

  • Annotate persons[“telework_arrangement”] based on persons[“cdap_activity"] and persons["has_in_home_work_activity"]

9

WSP | Explicit Telecommute | June 30, 2026

10 of 25

Source Code & Unit Test

10

WSP | Explicit Telecommute | June 30, 2026

abm/models/telework_arrangement.py

abm/test/test_misc/test_telework_arrangement.py

11 of 25

What is observed “has in-home work activity”?

  • Reported `telecommute_time` >= 180 minutes
  • Consistent with Phase 9 definition
  • Ignores short spillover tasks

11

WSP | Explicit Telecommute | June 30, 2026

43%

27%

24%

0%

12 of 25

Observed “Telework Arrangement”

  • 6% of workers are hybrid working on the day
  • 35% of workers are teleworking on the day

​

12

WSP | Explicit Telecommute | June 30, 2026

13 of 25

Explanatory Variables Considered

13

WSP | Explicit Telecommute | June 30, 2026

  • Person
    • age, sex, person type, telecommute frequency, work from home, workplace mode choice logsum, industry, free parking at work, disaggregate accessibility
  • Household
    • hh size, income, num children, non-family, home is urban, aggregate accessibility
  • Day
    • cdap, has preschool children at home, has school children at home

​

14 of 25

Survey Issues

  • 727 non-home-based workers have no recorded workplace location
    • Added a “missing workplace location” indicator variable to the estimation
  • 628 workers who selected “Never Telecommute” nonetheless logged at least 3 hours of telecommute time
    • 203 of them also recorded work tours that day, which may reflect a difference in interpretation
    • The `telecommute_frequency` value could be revised for these workers, but it was left unchanged to stay consistent with other efforts using the same data
  • 249 workers reported over 7 hours of telecommute time while also having out-of-home work tour(s)
    • Still investigating, but it appears that they might be teleworking from a non-home place and it’s coded as a work tour

​

14

WSP | Explicit Telecommute | June 30, 2026

15 of 25

Estimated Model - Draft

15

WSP | Explicit Telecommute | June 30, 2026

16 of 25

3. Telework Duration�Model Design, Software Implementation, and Model Estimation�

16

WSP | Explicit Telecommute | June 30, 2026

17 of 25

Telework Duration – MNL

  • Choosers
    • Workers who has in-home work activity (hybrid working or teleworking)
  • Alternatives
    • 3-5 hours
    • 5-7 hours
    • 7-9 hours
    • 9+ hours
  • General Sequence
    • After telework arrangement, before tour and trip models
  • Output
    • persons["telework_duration_category"]

17

18 of 25

Telework Duration – Probabilistic

  • Difference from MNL
    • Make direct draws from observed distribution

18

Default to one aggregate distribution:

Disaggregate distribution by segments. E.g., PROB_JOIN_COLS: [‘ptype’]

19 of 25

Source Code & Unit Test

19

WSP | Explicit Telecommute | June 30, 2026

abm/models/telework_duration.py

abm/test/test_misc/test_telework_duration.py

20 of 25

Observed Telework Duration

  • The observed 7+ hours teleworking is unexpectedly high for hybrid working – need further investigation
    • As noted on slide 14

20

21 of 25

Explanatory Variables Considered

  • Person
    • age, sex, person type, telecommute frequency, work from home, workplace mode choice logsum, industry, free parking at work, disaggregate accessibility, telework arrangement
  • Household
    • hh size, income, num children, non-family, home is urban, aggregate accessibility
  • Day
    • cdap, has preschool children at home, has school children at home

​

21

WSP | Explicit Telecommute | June 30, 2026

22 of 25

Estimated Model - Draft

22

WSP | Explicit Telecommute | June 30, 2026

23 of 25

4. Next Steps�

23

WSP | Explicit Telecommute | June 30, 2026

24 of 25

Next Steps

  • Investigate the large telecommute time in hybrid working workers
  • Finalize model estimation
  • Documentation

24

WSP | Explicit Telecommute | June 30, 2026

25 of 25

QUESTIONS?

wsp.com