Skip to content

Tour formation

Binary logits that shape the tours of MAST: outings from the workplace, their mode, and returns home between episodes.

Reading the tables

Coefficient: maximum-likelihood estimate, with significance stars *** p < 0.01, ** p < 0.05, * p < 0.10; robust (sandwich) standard error; two-sided p-value. N: observations; K: estimated parameters; LL(β̂): log-likelihood at the estimates; ρ̄²: adjusted rho-square. Hold-out: fit on the held-out half of the data, which was also used in the specification search. "Reference": the omitted category; "—": removed from the model.

Outing from the workplace

Whether a work episode at the usual workplace contains at least one outing.

N = 8,613K = 8LL(β̂) = −3,061.5ρ̄² = 0.486Hold-out: LL −3,093.2
Variable Coefficient Robust s.e. p
Constant
Constant −4.6092*** 0.5251 <0.001
Work episode
Work duration, 0–4 h (per h) 0.4033*** 0.1439 0.005
Work duration, 4–7 h (per h) 0.5862*** 0.0668 <0.001
Secondary work episode (arrival from 15:00) −2.0156*** 0.7207 0.005
Time of day
Arrival at work 7:00–10:00 reference
Arrival at work before 7:00 −0.7452*** 0.1155 <0.001
Arrival at work from 10:00 0.0870 0.1027 0.397
Workplace and commute
Workplace zone with at least 10,000 jobs per km² 0.5497*** 0.0705 <0.001
Ln(1 + home–workplace distance, km) −0.1786*** 0.0373 <0.001

Observations: 8,613 work episodes at the usual workplace; alternatives: 2; parameters: 8. LL(β̂) = −3,061.5; BIC = 6,195.5. Hold-out half (8,561 episodes): LL = −3,093.2. Estimation: Biogeme, robust standard errors. An outing is a work episode at the usual workplace followed by one or more discretionary episodes and a return to the workplace. Work duration: piecewise linear in hours with bends at 4 h and 7 h; secondary work episode: arrival at work from 15:00.

Mode of the outing

Whether the outing is walked or keeps the mode of the work tour; 7 parameters. The term for fewer cars than drivers is not in the model.

N = 1,177K = 7LL(β̂) = −360.6ρ̄² = 0.549Hold-out: LL −299.8
Variable Coefficient Robust s.e. p
Constant
Constant (walk) 4.3240*** 0.3610 <0.001
Outing
Ln(1 + distance workplace–first stop, km) −2.4743*** 0.1870 <0.001
Arrived at work by other modes reference
Arrived at work by car, as driver −2.2785*** 0.2626 <0.001
Arrived at work by car, as passenger −1.1701** 0.5325 0.028
More than one stop on the outing −1.0898*** 0.3653 0.003
Workplace zone
Workplace zone with at least 10,000 jobs per km² 0.9061*** 0.2488 <0.001
Household
Household with at least one car reference
No car in the household 0.5059 0.4887 0.301

Observations: 1,177 outings; alternatives: walk or the mode of the work tour; parameters: 7. LL(β̂) = −360.6; BIC = 770.7. Hold-out half (1,153 outings): LL = −299.8. Estimation: Biogeme, robust standard errors. If the outing does not walk, its legs keep the mode of the work tour (the car stays at the workplace for a walking outing). Mode of the work tour: reference other modes; distance: car distance between the workplace and the first stop. A term for fewer cars than drivers is not included: it is not significant (likelihood-ratio test p = 0.34) and BIC is lower without it.

Return home between episodes

Whether a person goes home between two consecutive out-of-home episodes; 33 parameters. The term for fewer cars than drivers is kept: the household car allocation does not cover this choice.

N = 28,349K = 33LL(β̂) = −13,697.0ρ̄² = 0.301Hold-out: LL −13,771.8
Variable Coefficient Robust s.e. p
Constant
Constant 0.2860** 0.1207 0.018
Detour and distance
Detour via home (km) −0.0080** 0.0034 0.018
Ln(1 + detour via home, km) −0.4645*** 0.0410 <0.001
Distance from home (km) 0.0067*** 0.0013 <0.001
Episodes
Preceding episode: other individual discretionary reference
Preceding episode: work −0.2370*** 0.0526 <0.001
Preceding episode: school −0.0491 0.0888 0.580
Preceding episode: accompanying −0.0478 0.0554 0.388
Preceding episode: shopping 0.1653*** 0.0415 <0.001
Preceding episode: joint 0.6576*** 0.0763 <0.001
Following episode: other individual discretionary reference
Following episode: work −0.8090*** 0.0698 <0.001
Following episode: school −0.2001* 0.1064 0.060
Following episode: accompanying 0.4622*** 0.0504 <0.001
Following episode: shopping −0.7064*** 0.0392 <0.001
Following episode: joint 1.2098*** 0.0599 <0.001
Joint episodes
Both episodes joint −2.2242*** 0.1081 <0.001
Joint episode next to a work or school episode 0.5146*** 0.0822 <0.001
Person
Person type: non-worker reference
Person type: worker −0.2405*** 0.0561 <0.001
Person type: K–12 student −0.1929* 0.1099 0.079
Person type: post-secondary student −0.1478 0.1093 0.176
Person type: retiree −0.1698*** 0.0594 0.004
Person type: child under 5 −0.2725** 0.1249 0.029
Female −0.1202*** 0.0305 <0.001
Driving licence 0.4137*** 0.0848 <0.001
Household
Household size (persons) 0.1267*** 0.0167 <0.001
Household income $50–100k reference
Household income below $50k −0.1886*** 0.0491 <0.001
Household income $100k or more 0.1125*** 0.0348 0.001
Cars in the household: at least as many as drivers reference
No car in the household 0.0650 0.0745 0.383
Fewer cars than drivers in the household −0.1200*** 0.0448 0.007
Child under 16 in the household −0.2205*** 0.0487 <0.001
Day pattern
Number of out-of-home episodes in the day −0.1497*** 0.0087 <0.001
Time of day
End of the preceding episode 9:00–15:00 reference
End of the preceding episode before 9:00 −0.0793 0.0583 0.174
End of the preceding episode 15–19 −0.3433*** 0.0363 <0.001
End of the preceding episode after 19:00 −1.3199*** 0.0823 <0.001

Observations: 28,349 pairs of consecutive out-of-home episodes; alternatives: 2; parameters: 33. LL(β̂) = −13,697.0; BIC = 27,732.2. Hold-out half (28,327 pairs): LL = −13,771.8. Estimation: maximum likelihood (robust sandwich covariance). y = 1 if the person returned home between two consecutive out-of-home episodes; pairs inside an outing from the workplace and pairs of two work episodes are excluded. Income bands: household income.