Demand Impacts of Bus Quality Improvements

Demand Impacts of Bus Quality Improvements


J Shires, M Wardman, ITS, University of Leeds, UK


The research reported in this paper contrasts with much of the previous research in the area in that it is focussed not on the valuation of improvements to bus service quality but rather on the demand impacts of it.


The research reported in this paper contrasts with much of the previous research in the area in that it is focussed not on the valuation of improvements to bus service quality but rather on the demand impacts of it.

In recent years, many of the privately owned UK bus operating companies have invested heavily in new and improved buses with higher levels of quality of service. Financial support from local authorities, particularly in terms of infrastructure, provides 'leverage' to the private sector spend.

The research reported here was funded by the UK Department of Transport who were interested to establish the impact on demand of the many and varied bus improvement schemes that had been implemented.

Thus whilst there are many studies that have estimated values for improving many different aspects of bus services, and this paper provides a review of key studies and findings, this study is treading a far more difficult and original path in attempting to identify the impacts on demand.

We report here two exercises which we believe will contribute significantly to the existing body of evidence. The analysis is based on a sample of over 5,000 travellers obtained through computer assisted home interviews in 10 case study areas where bus quality improvements had occurred.

The first analysis we report is the estimation to SP data of models of the direct demand type that are more typically estimated to measures of demand volumes. The SP exercises have included amongst their scenarios instances where for bus commuters the bus is made worse and where for car commuters the bus is improved. Whilst we can and do develop standard choice models to this data, the aim was to estimate models which relate bus demand directly to its fare, journey time, headway, average late time, quality of bus and the time and cost of car. Such models, based on pooling of the SP responses, recover demand elasticities directly, revealing whether and to what extent new buses impact on demand.

Without the need for transformation, and without the scaling problems of discrete choice models, we have obtained a wide range of plausible own and cross elasticity estimates. For example, the removal of quality buses would reduce bus demand by around 12%. The long run bus fare elasticity is around -0.7 whilst the time elasticity is around -0.2. Amongst car users, the introduction of a quality bus reduces car demand by around 2% with the fuel and time elasticities each estimated to be around -0.1. The overall demand impacts will be disaggregated into effects due to the precise attributes that make up the quality bus improvement in each area, taking account of any package effects.

The standard choice models recovered sensible valuations for a range of attributes. The valuations are lower than are typically obtained in SP studies and this may be because in this mode choice context the purpose of the study was less transparent and hence did not offer such an incentive to strategic bias.

The second piece of analysis is based around commuters having a real choice between a high quality bus on one route near their home and an alternative low quality bus on another route. This facilitates the development of Revealed Preference (RP) models as well as providing a realistic basis for an SP exercise. The purpose of this exercise was, given the scepticism that sometimes surrounds the valuations and implied demand impacts obtained from SP studies of bus quality improvements, to determine whether bus quality had any impact on bus users? actual behaviour. These respondents, if anyone, should react to such changes. The comparison of RP and SP models is also a valuable exercise.

Through judicious choice of home interview location, we have selected respondents who are in a position to use a high quality service but at the expense of possibly lower frequency or higher cost or higher access time. It has been possible to detect an effect from walk time, fare and bus quality on bus users? actual behaviour. The SP exercise, offering more variation in attributes, discerns significant effects additionally from journey time and service headway. The SP results are broadly corroborated by the RP evidence. In terms of bus quality the models have estimated positive valuations that would reduce the generalised cost of travellers, by around 5 minutes (SP model) and 6 minutes per journey (RP model).


Association for European Transport