Controlling the driving and road safety training sector through the driving test success rate?

Abstract

The driving training sector in France is subjected to an increasing interest from the French government. The regulator notably wonders about the performance of the training given by services companies. He plans to use more widely the driving test success rate indicator by enterprise to guide the public action.

This article explores how the driving examination rate depends on a set of socio-economic variables. With a factor analysis and an analysis of variance and a CART method, we highlight that the companies' success rate depends more on the characteristics of the regional demand than to their inner characteristics in terms of provision and supply. We can thus deduce that the control through this indicator proves to be ineffective and that it is necessary to appeal to a more multidimensional definition of the performance of the companies of this sector.

Keywords:

• driving sector
• evaluation
• factor analysis
• graduated driver licensing
• public policy
• variance analysis

Notes

1. For further details on the socio-economic data of this sector, see the report of the Laboratoire OEP (2005) for the ‘Direction à la Sécurité et la Circulation Routière’; for a synthesis, see Abramovici, Bancel-Charensol, Chevrier, Jougleux, and Maman (2006).

2. The main activity of the traffic education cells concerns the organisation of the driving tests of all categories (A, B, C, EC, D) in the driving schools in their own region. Besides, the agents of the cells participate in actions aiming at the improvement of road safety in the region internally as well as externally.

3. The work on the variables of the pole ‘characteristics of the demand’ is not presented in this article, only the results are used. The detailed results can be supplied by the authors on request.

4. The list of the different licences that exist in France is presented in Table A1 in the Appendix.

5. The notion of driving training or driving school, which is often used, is rather vague for it covers at the same time the premise and the company. In our study, the company is marked off by its SIREN number (Système d‘Identification du Répertoire des ENterprises, company identification number) and a number given by the DSCR that is linked to the name of the person that runs the company or the representative of a legal entity in a special region. A premise in a region is marked off with its SIRET number (Système d‘Identification du Répertoire des ETablissements, premises registration number). A company located in n regions will be counted n times. Less than 40 companies over the 8500 counted are concerned.

6. The RAFAEL application aims to ‘control the conditions to access the profession, to make a better follow up of this profession and to obtain socio-economic data on the driving schools and their trainers’. This database allows agrements' applications and monitoring to be dealt with. The AURIGE database concerns the follow-up and the management of the public service activity of the driving tests, of the trainers and of the companies.

7. As we shall see further on, researchers do not downplay here the importance of the individual ability of the candidate to pass the driving test more or less easily. These researches are related to the science of the education field. Here, our purpose is to focus on the influence of the variables that are little studied in this sector and yet directly concerned by any public political action.

8. We notice the Guttman effect when the projections of initial variables (and of their clusters) are parallel curves between themselves and represent an arc of a circle. From a statistical point of view, it means that the variables are bound (parallel curves) and would be summarised in a sole variable.

9. A classification tree or ‘dendrogramme’ allows one to visualise the categories of companies that are created from the variables of each pole. The proximity between two groups can be noticed by the height of the branches that connect them. The higher the level of the grouping, the greater the difference (or heterogeneity) between the two groups of companies. Conversely, the earliest the grouping is made the closest the groups are – they are homogeneous.

10. However, one has to take notice of the fact that in one of the categories, the companies (4% of the companies) do not register candidates to the B and AAC licence, though in the other category (2% of the companies) this is a complete diversification.

11. The variable price of the service on the market is not available just as it is. Nevertheless, it seems it will be possible to evaluate it through the connection between the annual turnover of the company (from the individual data of the file SUSE) and the number of candidates registered or that pass the examination (file AURIGE), which is not possible for this research.

12. The fact of having worked on just one variable for each group presented above allows us to build more reliable trees. The price to pay, as in a regression, is a lower quality of the final tree (we notice an increase in the bad classification from 0.31 to 0.35).