Using ordered attitudinal indicators in a latent variable choice model: a study of the impact of security on rail travel behavior

Daly, Hess, Patruni, Potoglou, Rohr, 2012, in Transportation

Location	UK
Population	General
Sample size	1961
Factor analysis type	confirmatory factor analysis, nan rotation
Stepwise regression	No
Removal of insignificant variables	No
Reviewed by	MWC

Abstract

There is growing interest in the use of models that recognise the role of individuals’ attitudes and perceptions in choice behaviour. Rather than relying on simple linear approaches or a potentially bias-inducing deterministic approach based on incorporating stated attitudinal indicators directly in the choice model, researchers have recently recognised the latent nature of attitudes. The uptake of such latent attitude models in applied work has however been slow, while a number of overly simplistic assumptions are also commonly made. In this article, we present an application of jointly estimated attitudinal and choice models to a real-world transport study, looking at the role of latent attitudes in a rail travel context. Our results show the impact that concern with privacy, liberty and security, and distrust of business, technology and authority have on the desire for rail travel in the face of increased security measures, as well as for universal security checks. Alongside demonstrating the applicability of the model in applied work, we also address a number of theoretical issues. We ﬁrst show the equivalence of two different normalisations discussed in the literature. Unlike many other latent attitude studies, we explicitly recognise the repeated choice nature of the data. Finally, the main methodological contribution comes in replacing the typically used continuous model for attitudinal response by an ordered logit structure which more correctly accounts for the ordinal nature of the indicators.

Factors

Variable	Loading
Protecting the privacy of my information (Lifestyle)	1
Taking action against important security risks (e.g. international terrorism, organized crime) (Other)	0.04
Defending current liberties and human rights (Lifestyle)	1.411

Variable	Loading
Protecting the privacy of my information (Lifestyle)	1
Taking action against important security risks (e.g. international terrorism, organized crime) (Other)	-1.191
Defending current liberties and human rights (Lifestyle)	1.399

Variable	Loading
Protecting the privacy of my information (Lifestyle)	0.101
Taking action against important security risks (e.g. international terrorism, organized crime) (Other)	-0.12
Defending current liberties and human rights (Lifestyle)	0.141

Variable	Loading
Protecting the privacy of my information (Lifestyle)	1
Taking action against important security risks (e.g. international terrorism, organized crime) (Other)	0.068
Defending current liberties and human rights (Lifestyle)	1.078

Variable	Loading
Technology has almost gotten out of control (Other)	1
Government can generally be trusted to look after our interests (Lifestyle)	2.084
The way one votes has no effect on what the government does (Other)	0.755
In general business helps us more than it harms us (Other)	1.102

Variable	Loading
Technology has almost gotten out of control (Other)	1
Government can generally be trusted to look after our interests (Lifestyle)	1.941
The way one votes has no effect on what the government does (Other)	0.76
In general business helps us more than it harms us (Other)	0.894

Variable	Loading
Technology has almost gotten out of control (Other)	0.348
Government can generally be trusted to look after our interests (Lifestyle)	0.676
The way one votes has no effect on what the government does (Other)	0.265
In general business helps us more than it harms us (Other)	0.312

Variable	Loading
Technology has almost gotten out of control (Other)	1
Government can generally be trusted to look after our interests (Lifestyle)	2.089
The way one votes has no effect on what the government does (Other)	0.759
In general business helps us more than it harms us (Other)	1.104

Models

Dependent variable	Use of rail system under different security regimes
Model type	Multinomial logit
Sample size	1961.0
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	-19535.1

Variable	Coefficient	p-value
Price (pounds)	-0.295	0.0
Travel time (minutes)	-0.073	0.0
Camera type 0 (no security cameras)	0.0	nan
Camera type 1 (standard CCTV)	0.562	0.0
Camera type 2 (CCTV and new cameras that automatically identify individuals)	0.845	0.0
Security check type 0 (no checks)	0.0	nan
Security check type 1 (Pat down and bag search for 1 in 1,000 travelers)	0.214	0.0
Security check type 2 (Pat down and bag search for 2 in 1,000 travelers)	0.213	0.0
Security check type 3 (Pat down and bag search for 10 in 1,000 travelers)	0.316	0.0
Security check type 4 (Metal detector/X-ray for all)	0.671	0.0
Presence of security personnel: rail staff	0.0	nan
Presence of security personnel: rail staff and British transport police	0.233	0.0
Presence of security personnel: rail staff, British transport police, and armed police	0.126	0.0
Presence of security personnel: rail staff, British transport police, armed police, and uniformed military	0.102	0.002
Number of known terrorist plots disrupted	0.31	0.0
Number of known terrorist plots disrupted: > 2.5	-0.24	0.0
Number of known terrorist plots disrupted: > 10	-0.05	0.0
Visibility of response: if an incident occurs you are not aware of it	0.0	nan
Visibility of response: if an incident occurs then you are aware of that when you get back home	-0.052	0.134
Visibility of response: if an incident occurs things are handled with minimal disruption	0.007	0.842
Visibility of response: if an incident occurs there is some disruption and chaos	-0.369	0.0
Visibility of response: if an incident occurs there is lots of disruption and chaos	-0.586	0.0
No travel by rail (ASC)	0.221	0.003

Dependent variable	Use of rail system under different security regimes
Model type	Integrated choice and latent variable
Sample size	1961.0
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	-36128.2

Increased concern (Ben-Akiva normalization)
Variable	Coefficient	p-value
Age	-0.0081	0.016
Gender (male)	0.0237	0.007
Std. dev	0.1008	0.0

No travel by rail
Variable	Coefficient	p-value
No travel by rail (ASC)	-2.985	0.0
Increased concern (Ben-Akiva normalization)	37.038	0.0
Reduced distrust (Ben-Akiva normalization)	-6.25	0.0

Reduced distrust (Ben-Akiva normalization)
Variable	Coefficient	p-value
Age	-0.0427	0.0
Gender (male)	-0.0001	1.0
Std. dev	0.3484	0.0

Variable	Coefficient	p-value
Price (pounds)	-0.323	0.0
Travel time (minutes)	-0.081	0.0
Camera type 0 (no security cameras)	0.0	nan
Camera type 1 (standard CCTV)	0.578	0.0
Camera type 2 (CCTV and new cameras that automatically identify individuals)	0.885	0.0
Security check type 0 (no checks)	0.0	nan
Security check type 1 (Pat down and bag search for 1 in 1,000 travelers)	0.255	0.0
Security check type 2 (Pat down and bag search for 2 in 1,000 travelers)	0.255	0.0
Security check type 3 (Pat down and bag search for 10 in 1,000 travelers)	0.356	0.0
Security check type 4 (Metal detector/X-ray for all)	0.763	0.0
Presence of security personnel: rail staff	0.0	nan
Presence of security personnel: rail staff and British transport police	0.235	0.0
Presence of security personnel: rail staff, British transport police, and armed police	0.153	0.0
Presence of security personnel: rail staff, British transport police, armed police, and uniformed military	0.104	0.003
Number of known terrorist plots disrupted	0.318	0.0
Number of known terrorist plots disrupted: > 2.5	-0.241	0.0
Number of known terrorist plots disrupted: > 10	-0.057	0.0
Visibility of response: if an incident occurs you are not aware of it	0.0	nan
Visibility of response: if an incident occurs then you are aware of that when you get back home	-0.067	0.072
Visibility of response: if an incident occurs things are handled with minimal disruption	0.016	0.617
Visibility of response: if an incident occurs there is some disruption and chaos	-0.405	0.0
Visibility of response: if an incident occurs there is lots of disruption and chaos	-0.604	0.0

Dependent variable	Use of rail system under different security regimes
Model type	Integrated choice and latent variable
Sample size	1961.0
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	-16593.3

Increased concern (Bolduc normalization)
Variable	Coefficient	p-value
Age	-0.0805	0.004
Gender (male)	0.2351	0.001
Std. dev	1.0	nan

No travel by rail
Variable	Coefficient	p-value
No travel by rail (ASC)	-2.985	0.0
Increased concern (Bolduc normalization)	3.733	0.0
Reduced distrust (Bolduc normalization)	-2.178	0.0

Reduced distrust (Bolduc normalization)
Variable	Coefficient	p-value
Age	-0.1226	0.0
Gender (male)	-0.0004	1.0
Std. dev	1.0	nan

Variable	Coefficient	p-value
Price (pounds)	-0.323	0.0
Travel time (minutes)	-0.081	0.0
Camera type 0 (no security cameras)	0.0	nan
Camera type 1 (standard CCTV)	0.578	0.0
Camera type 2 (CCTV and new cameras that automatically identify individuals)	0.885	0.0
Security check type 0 (no checks)	0.0	nan
Security check type 1 (Pat down and bag search for 1 in 1,000 travelers)	0.255	0.0
Security check type 2 (Pat down and bag search for 2 in 1,000 travelers)	0.254	0.0
Security check type 3 (Pat down and bag search for 10 in 1,000 travelers)	0.355	0.0
Security check type 4 (Metal detector/X-ray for all)	0.763	0.0
Presence of security personnel: rail staff	0.0	nan
Presence of security personnel: rail staff and British transport police	0.235	0.0
Presence of security personnel: rail staff, British transport police, and armed police	0.153	0.0
Presence of security personnel: rail staff, British transport police, armed police, and uniformed military	0.104	0.003
Number of known terrorist plots disrupted	0.318	0.0
Number of known terrorist plots disrupted: > 2.5	-0.241	0.0
Number of known terrorist plots disrupted: > 10	-0.057	0.0
Visibility of response: if an incident occurs you are not aware of it	0.0	nan
Visibility of response: if an incident occurs then you are aware of that when you get back home	-0.067	0.072
Visibility of response: if an incident occurs things are handled with minimal disruption	0.016	0.617
Visibility of response: if an incident occurs there is some disruption and chaos	-0.405	0.0
Visibility of response: if an incident occurs there is lots of disruption and chaos	-0.604	0.0

Dependent variable	Use of rail system under different security regimes
Model type	Integrated choice and latent variable
Sample size	1961.0
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	-31699.2

Increased concern (ordered logit attitudinal model)
Variable	Coefficient	p-value
Age	-0.0378	0.317
Gender (male)	-0.13	0.271
Std. dev	1.7668	0.0

No travel by rail
Variable	Coefficient	p-value
No travel by rail (ASC)	-3.789	0.0
Increased concern (ordered logit attitudinal model)	0.338	0.004
Reduced distrust (ordered logit attitudinal model)	-14.066	0.0

Reduced distrust (ordered logit attitudinal model)
Variable	Coefficient	p-value
Age	-0.0138	0.016
Gender (male)	-0.0504	0.007
Std. dev	0.3139	0.0

Variable	Coefficient	p-value
Price (pounds)	-0.323	0.0
Travel time (minutes)	-0.081	0.0
Camera type 0 (no security cameras)	0.0	nan
Camera type 1 (standard CCTV)	0.577	0.0
Camera type 2 (CCTV and new cameras that automatically identify individuals)	0.885	0.0
Security check type 0 (no checks)	0.0	nan
Security check type 1 (Pat down and bag search for 1 in 1,000 travelers)	0.255	0.0
Security check type 2 (Pat down and bag search for 2 in 1,000 travelers)	0.254	0.0
Security check type 3 (Pat down and bag search for 10 in 1,000 travelers)	0.355	0.0
Security check type 4 (Metal detector/X-ray for all)	0.763	0.0
Presence of security personnel: rail staff	0.0	nan
Presence of security personnel: rail staff and British transport police	0.235	0.0
Presence of security personnel: rail staff, British transport police, and armed police	0.154	0.0
Presence of security personnel: rail staff, British transport police, armed police, and uniformed military	0.104	0.003
Number of known terrorist plots disrupted	0.318	0.0
Number of known terrorist plots disrupted: > 2.5	-0.241	0.0
Number of known terrorist plots disrupted: > 10	-0.057	0.0
Visibility of response: if an incident occurs you are not aware of it	0.0	nan
Visibility of response: if an incident occurs then you are aware of that when you get back home	-0.066	0.072
Visibility of response: if an incident occurs things are handled with minimal disruption	0.015	0.689
Visibility of response: if an incident occurs there is some disruption and chaos	-0.404	0.0
Visibility of response: if an incident occurs there is lots of disruption and chaos	-0.604	0.0

Dependent variable	Use of rail system under different security regimes
Model type	Integrated choice and latent variable
Sample size	1961.0
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	-31696.7

Increased concern (attitudinal interaction model)
Variable	Coefficient	p-value
Age	-0.0365	0.23
Gender (male)	-0.073	0.424
Std. dev	1.53	0.0

No travel by rail
Variable	Coefficient	p-value
No travel by rail (ASC)	-3.788	0.0
Increased concern (attitudinal interaction model)	0.38	0.001
Reduced distrust (attitudinal interaction model)	-14.049	0.0

Reduced distrust (attitudinal interaction model)
Variable	Coefficient	p-value
Age	-0.0136	0.23
Gender (male)	-0.0517	0.005
Std. dev	0.3138	0.0

Variable	Coefficient	p-value
Price (pounds)	-0.322	0.0
Travel time (minutes)	-0.081	0.0
Camera type 0 (no security cameras)	0.0	nan
Camera type 1 (standard CCTV)	0.576	0.0
Camera type 2 (CCTV and new cameras that automatically identify individuals)	0.885	0.0
Security check type 0 (no checks)	0.0	nan
Security check type 1 (Pat down and bag search for 1 in 1,000 travelers)	0.255	0.0
Security check type 2 (Pat down and bag search for 2 in 1,000 travelers)	0.254	0.0
Security check type 3 (Pat down and bag search for 10 in 1,000 travelers)	0.354	0.0
Security check type 4 (Metal detector/X-ray for all)	0.74	0.0
Presence of security personnel: rail staff	0.0	nan
Presence of security personnel: rail staff and British transport police	0.235	0.0
Presence of security personnel: rail staff, British transport police, and armed police	0.154	0.0
Presence of security personnel: rail staff, British transport police, armed police, and uniformed military	0.104	0.003
Number of known terrorist plots disrupted	0.319	0.0
Number of known terrorist plots disrupted: > 2.5	-0.242	0.0
Number of known terrorist plots disrupted: > 10	-0.057	0.0
Visibility of response: if an incident occurs you are not aware of it	0.0	nan
Visibility of response: if an incident occurs then you are aware of that when you get back home	-0.066	0.072
Visibility of response: if an incident occurs things are handled with minimal disruption	0.015	0.617
Visibility of response: if an incident occurs there is some disruption and chaos	-0.404	0.0
Visibility of response: if an incident occurs there is lots of disruption and chaos	-0.602	0.0
Metal detector/X-ray for all * increased concern (attitudinal interaction model)	-0.089	0.036