Residential self-selection and the reverse causation hypothesis: Assessing the endogeneity of stated reasons for residential choice
Kroesen, 2019, in Travel Behaviour and Society
doi:10.1016/j.tbs.2019.05.002
Location |
Netherlands |
Population |
General |
Sample size |
1824 |
Factor analysis type |
none, none rotation |
Stepwise regression |
no |
Removal of insignificant variables |
no |
Reviewed by |
LCM |
Abstract
Residential self-selection is a well-recognized potential bias in estimating the true effects of the built environment on travel behavior. A popular method to account for residential self-selection is by including people's attitudes towards various modes as additional control variables in the regression. Yet, while attitudes may indeed influence both residential location choice and travel behavior, they may, in turn, also be affected by these factors. This paper aims to assess to what extent the built environment and travel behavior influence people's stated reasons for living in a certain location over time, which would mean that these reasons are actually endogenous to the built environment and travel behavior. To achieve this aim panel data are used from the same respondents (who did not move house)asking them at two points in time (two years apart)to state their reasons for their current residential choice. The data are modeled using a latent transition model. The results indicate that approximately 39% of the Dutch population belongs to a class which attaches importance to short distances to public transport and shops. Moreover, the distance to the train station, the amount of travel by train and car ownership at the first point in time are found to influence the probability that a person (still)belongs to this class at the second point in time, providing evidence that the built environment and travel behavior temporally precede travel related residential preferences. The results suggest that the use of stated reasons for residential choice as control variables is problematic. © 2019 Hong Kong Society for Transportation Studies
Factors
Models
Dependent variable |
Class membership |
Model type |
LCM |
Sample size |
1824.0 |
R2 |
nan |
Adjusted R2 |
|
Pseudo R2
(nan)
|
nan |
AIC |
nan |
BIC |
nan |
Log-likelihood at zero |
nan |
Log-likelihood at constants |
nan |
Log-likelihood at convergence |
nan |
High access class |
Variable |
Coefficient |
p-value |
Intercept |
-0.25
|
0.81 |
Level of urbanization |
-0.239
|
0 |
Distance between train station and residential location (km) |
-0.005
|
0.02 |
Frequency of car use |
-0.114
|
0.06 |
Frequency of train use |
0.206
|
0 |
Frequency of bicycle use |
0.1
|
0 |
Number of cars in household |
-0.048
|
0.12 |
Gender |
0.101
|
0.45 |
Age 18-29 |
0.093
|
0.26 |
Age 30-59 |
-0.125
|
nan |
Age 60+ |
0.033
|
nan |
Level of education |
0.105
|
0.01 |
Presence of young children (< 12 years of age) |
-0.212
|
0.01 |
Household income |
0.028
|
0.29 |
Moderate access class |
Variable |
Coefficient |
p-value |
Intercept |
-0.058
|
0.81 |
Level of urbanization |
0.118
|
0 |
Distance between train station and residential location (km) |
-0.008
|
0.02 |
Frequency of car use |
0.074
|
0.06 |
Frequency of train use |
0.114
|
0 |
Frequency of bicycle use |
-0.015
|
0 |
Number of cars in household |
-0.092
|
0.12 |
Gender |
-0.072
|
0.45 |
Age 18-29 |
-0.012
|
0.26 |
Age 30-59 |
0.095
|
nan |
Age 60+ |
-0.084
|
nan |
Level of education |
0.036
|
0.01 |
Presence of young children (< 12 years of age) |
0.278
|
0.01 |
Household income |
0.039
|
0.29 |
No access class |
Variable |
Coefficient |
p-value |
Intercept |
0.309
|
0.81 |
Level of urbanization |
0.121
|
0 |
Distance between train station and residential location (km) |
0.012
|
0.02 |
Frequency of car use |
0.041
|
0.06 |
Frequency of train use |
-0.092
|
0 |
Frequency of bicycle use |
-0.085
|
0 |
Number of cars in household |
0.14
|
0.12 |
Gender |
-0.026
|
0.45 |
Age 18-29 |
-0.081
|
0.26 |
Age 30-59 |
0.03
|
nan |
Age 60+ |
0.051
|
nan |
Level of education |
-0.141
|
0.01 |
Presence of young children (< 12 years of age) |
-0.066
|
0.01 |
Household income |
-0.067
|
0.29 |
Dependent variable |
Class membership |
Model type |
LCM |
Sample size |
1824.0 |
R2 |
nan |
Adjusted R2 |
|
Pseudo R2
(nan)
|
nan |
AIC |
nan |
BIC |
nan |
Log-likelihood at zero |
nan |
Log-likelihood at constants |
nan |
Log-likelihood at convergence |
nan |
High access class |
Variable |
Coefficient |
p-value |
Intercept |
-0.784
|
0.39 |
Class membership in 2014: class 1 |
1.493
|
0 |
Class membership in 2014: class 2 |
-1.013
|
nan |
Class membership in 2014: class 3 |
-0.479
|
nan |
Level of urbanization |
0.013
|
0.66 |
Distance between train station and residential location (km) |
-0.029
|
0 |
Frequency of car use |
0.012
|
0.15 |
Frequency of train use |
0.129
|
0.01 |
Frequency of bicycle use |
-0.029
|
0.77 |
Number of cars in household |
-0.37
|
0 |
Gender |
0.131
|
0.47 |
Age 18-29 |
-0.112
|
0.41 |
Age 30-59 |
0.064
|
nan |
Age 60+ |
0.048
|
nan |
Level of education |
0.014
|
0.5 |
Presence of young children (< 12 years of age) |
0.2
|
0.23 |
Household income |
0.158
|
0.02 |
Job change |
-0.352
|
0 |
Started working |
-0.317
|
0.23 |
Stopped working |
0.639
|
0.22 |
Birth of a child |
0.403
|
0.01 |
Divorce |
0.521
|
0.73 |
Started living together |
-0.882
|
0.24 |
Moderate access class |
Variable |
Coefficient |
p-value |
Intercept |
0.708
|
0.39 |
Class membership in 2014: class 1 |
-0.259
|
0 |
Class membership in 2014: class 2 |
0.798
|
nan |
Class membership in 2014: class 3 |
-0.539
|
nan |
Level of urbanization |
0.024
|
0.66 |
Distance between train station and residential location (km) |
0.009
|
0 |
Frequency of car use |
-0.094
|
0.15 |
Frequency of train use |
-0.089
|
0.01 |
Frequency of bicycle use |
0.007
|
0.77 |
Number of cars in household |
0.209
|
0 |
Gender |
-0.01
|
0.47 |
Age 18-29 |
0.074
|
0.41 |
Age 30-59 |
0.05
|
nan |
Age 60+ |
-0.124
|
nan |
Level of education |
0.035
|
0.5 |
Presence of young children (< 12 years of age) |
0.019
|
0.23 |
Household income |
-0.037
|
0.02 |
Job change |
-0.151
|
0 |
Started working |
0.647
|
0.23 |
Stopped working |
-0.603
|
0.22 |
Birth of a child |
0.79
|
0.01 |
Divorce |
-0.443
|
0.73 |
Started living together |
0.281
|
0.24 |
No access class |
Variable |
Coefficient |
p-value |
Intercept |
0.076
|
0.39 |
Class membership in 2014: class 1 |
-1.234
|
0 |
Class membership in 2014: class 2 |
0.215
|
nan |
Class membership in 2014: class 3 |
1.019
|
nan |
Level of urbanization |
-0.036
|
0.66 |
Distance between train station and residential location (km) |
0.021
|
0 |
Frequency of car use |
0.082
|
0.15 |
Frequency of train use |
-0.04
|
0.01 |
Frequency of bicycle use |
0.022
|
0.77 |
Number of cars in household |
0.161
|
0 |
Gender |
-0.122
|
0.47 |
Age 18-29 |
0.037
|
0.41 |
Age 30-59 |
-0.114
|
nan |
Age 60+ |
0.077
|
nan |
Level of education |
-0.048
|
0.5 |
Presence of young children (< 12 years of age) |
-0.219
|
0.23 |
Household income |
-0.121
|
0.02 |
Job change |
0.503
|
0 |
Started working |
-0.33
|
0.23 |
Stopped working |
-0.036
|
0.22 |
Birth of a child |
-1.193
|
0.01 |
Divorce |
-0.078
|
0.73 |
Started living together |
0.601
|
0.24 |