Table 10: explanation of acronyms used in table 11
|
Variables_________________ |
Explanation_________________________________________________________________ |
|
bedrooms (bedr) large municip. apartment owned house partner dep. time constr. gasoline car tta76m income class 3___________ |
number of bedrooms monthly cost housing effect code 1 location (big city) effect code 2 location (small town/rural area) travel cost (fuel and rp) single trip (euro) travel time total single trip (min) dummy college/university (yes=1) dummy working partner (yes=1) dummy children (yes=1) dummy size municipality (> 50.000 inhab. =1) dummy apartment (yes=1) dummy owned house (yes=1) dummy (semi) detached house (yes=1) dummy fuel cost compensation (yes=1) dummy possibility work at home (always, sometimes=1) dummy partner (yes=1) dummy departure time constraint (yes=1) dummy car on benzene (yes=1) dummy congestion sensitive regions in Holland (yes=1) dummy actual travel time (including congestion) between 51 and 75 min (yes=1) dummy actual travel time (including congestion) > 75 min (yes=1) dummy travel time shown in experiment between 0 and 30 minutes (yes=1) dummy household income 0-28000 euro/year =1 dummy household income 28500-56000 euro/year =1 dummy household income >56000 euro/year =1_____________________________ |
Looking at the ML results in table 11 four significant random parameters can be observed.
The fit of the ML model is higher than of the MNL-model. Also, the parameter values in
general are more extreme in the ML-case, which might partly be explained by the higher
model fit. Furthermore, some significant parameters in the MNL-case are not significant on a
90 percent level in the ML-estimation, namely: the relatively lower dislike of living in a big
city for people receiving a fuel cost compensation, the relation between the province and
travel costs and the fact that people with departure time constraints value travel time less
negatively. Besides these effects, some relations with income are not significant in the ML-
estimation (e.g. big city*i2 and tc*i2). However, in general the picture between the two model
estimations is comparable.
20