The objective of this thesis is to study the sources of discrepancy between the actual traffic in motorways under concession schemes and the traffic forecast ex-ante. The demand forecast for a specific project is the main variable influencing its realization. From a public sector perspective, socio-economic evaluations are driven by demand forecasts, which gives the basis for choose and hierarchy public projects in order to maximise social welfare. From a private sector perspective, traffic forecasts are the base of financial evaluation and toll setting.Despite its importance and the numerous and important developments in the field, the differences of forecast and ex-post traffic are usually very high. Some recent studies show that differences as big as 20% are much more the rule than the exception.A huge amount of uncertainty is associated with the forecasting exercise. First because transport is a derived demand and depends on many exogenous variables, also uncertain; because modelling is and simplification exercise, implies many assumptions and rely on field data, many times incomplete or of low quality; moreover, modelling human (in this case users) behaviour is always a dangerous enterprise.Although these arguments could explain at least the larger part of errors associated with forecasts, one can wonder whether the agents implicated in the forecast would or could use this uncertainty strategically in their favour. In a competition for the field scheme (bids), the bidder may overestimate the demand in order to reduce the toll included in the bid. This strategic behaviour can introduce a high bias in forecasts. Also, overoptimistic (or overpessimistic) forecasters may introduce a bias in the forecast.We propose to focus in turn on the three main groups of agents involved in the demand forecast process. The forecasters, the project promoters and the users. Study all the issues related to them would be a too ambitious (or more concretely impossible) task. We then focus on some particular issues related to the modelling of the actors' behaviour in the context of the demand forecast for toll roads.Regarding the forecaster behaviour, we present the results of the first large sample survey on forecasters' perceptions and opinions about forecasting demand for transport projects, based on an on-line survey. We first describe the main characteristics of forecasters. We then describe the last forecast forecasters prepared. We turn to the models forecasters apply, the errors they declare on past forecasts and the main sources of errors according to them. We then describe the forecast environment in terms of pressure forecasters receive. These unique results provide a picture of the world of forecasters and forecasts, allowing for a better understanding of them. We turn then to the study of the optimism and overconfidence in transport forecasts. Optimism and overconfidence in general are recognized human traits. We analyze the overoptimistic bias by comparing the distribution of stated errors with actual errors found in literature; we also compare the own skilful of subjects in doing forecasts with studies showing self-evaluations of a common skill - driving. We finally propose a regression of the competence, quality and errors on the main forecasters' and projects' specific variables.Results show that the distribution of errors transport forecasters state has a smaller average magnitude and a smaller variance than those found in literature. Comparing forecasters perception of their own competence with the results found in literature about drivers skill self-evaluation, however, we could not find a significant difference, meaning that the forecasters' overconfidence is in line with what could be viewed as a normal human overconfidence level.The pressure for results forecasters receive and the strategic manipulation they affirm exist merit a special attention. They imply that while forecasters' behavioural biases may exist and should be take in account when evaluation forecasts, the project promoter may influence forecasts by pressuring the forecasters to produce results which better fit his expectancies.We then study the bidders' strategic behaviour in auctions for road concessions. We address three questions in turn. First, we investigate the overall effects of the winner's curse on bidding behaviour in such auctions. Second, we examine the effects of the winner's curse on contract auctions with differing levels of common-value components. Third, we investigate how the winner's curse affects bidding behaviour in such auctions when we account for the possibility for bidders to renegotiate. Using a unique, self-constructed, dataset of 49 worldwide road concessions, we show that the winner's curse effect is particularly strong in toll road concession contract auctions. Thus, we show that bidders bid less aggressively in toll road concession auctions when they expect more competition. We observe that this winner's curse effect is even larger for projects where the common uncertainty is greater. Moreover, we show that the winner's curse effect is weaker when the likelihood of renegotiation is higher. While the traditional implication would be that more competition is not always desirable when the winner's curse is particularly strong, we show that, in toll road concession contract auctions, more competition may be always desirable. Modelling aggregated users' behaviour, we study the long term traffic maturity. We argue that traffic maturity results from decreasing marginal utility of transport. The elasticity of individual mobility with respect to the revenue decreases after a certain level of mobility is reached. In order to find evidences of decreasing elasticity we analyse a cross-section time-series sample including 40 French motorways' sections. This analysis shows that decreasing elasticity can be observed in the long term. We then propose a decreasing function for the traffic elasticity with respect to the economic growth, which depends on the traffic level on the road. Although “unconditional” decreasing elasticities were already proposed in the literature, this is the first work, as far as we know, putting this idea in evidence and giving it a functional form. This model provides better interpretation of the coupling between traffic and economic growth, and a better long-term forecast. From the disaggregate perspective, we study the main individual modal choice variable, the value of time. The value of travel time savings is a fundamental concept in transport economics and its size strongly affects the socio-economic evaluation of transport schemes. Financial assessment of tolled roads rely upon the value of time as the main (or even the unique) willingness to pay measure. Values of time estimates, which primarily represent behavioural values, as then increasingly been used as measures of out-of-pocket money. In this setting, one of the main issues regarding the value of time is its distribution over the population. We apply the Logit, the Mixed Logit and the Bayesian Mixed Logit models to estimate the value of time in freight transport in France. Estimations with mixed logit faced many difficulties, as expected. These difficulties could be avoided using the Bayesian procedures, providing also the opportunity of properly integrating a priori beliefs. Results show that 1) using a single constant value of time, representative of an average, can lead to demand overestimation, 2) the estimated average value of time of freight transport in France is about 45 Euro, depending on the load/empty and hire/own account variables, which implies that 3) the standard value recommended in France should be reviewed upwards.

Le problème de tarification qui nous intéresse ici consiste à maximiser le revenu généré par les usagers d'un réseau de transport. Pour se rendre à leurs destinations, les usagers font un choix de route et utilisent des arcs sur lesquels nous imposons des tarifs. Chaque route est caractérisée (aux yeux de l'usager) par sa "désutilité", une mesure de longueur généralisée tenant compte à la fois des tarifs et des autres coûts associés à son utilisation. Ce problème a surtout été abordé sous une modélisation déterministe de la demande selon laquelle seules des routes de désutilité minimale se voient attribuer une mesure positive de flot. Le modèle déterministe se prête bien à une résolution globale, mais pèche par manque de réalisme. Nous considérons ici une extension probabiliste de ce modèle, selon laquelle les usagers d'un réseau sont alloués aux routes d'après un modèle de choix discret logit. Bien que le problème de tarification qui en résulte est non linéaire et non convexe, il conserve néanmoins une forte composante combinatoire que nous exploitons à des fins algorithmiques. Notre contribution se répartit en trois articles. Dans le premier, nous abordons le problème d'un point de vue théorique pour le cas avec une paire origine-destination. Nous développons une analyse de premier ordre qui exploite les propriétés analytiques de l'affectation logit et démontrons la validité de règles de simplification de la topologie du réseau qui permettent de réduire la dimension du problème sans en modifier la solution. Nous établissons ensuite l'unimodalité du problème pour une vaste gamme de topologies et nous généralisons certains de nos résultats au problème de la tarification d'une ligne de produits. Dans le deuxième article, nous abordons le problème d'un point de vue numérique pour le cas avec plusieurs paires origine-destination. Nous développons des algorithmes qui exploitent l'information locale et la parenté des formulations probabilistes et déterministes. Un des résultats de notre analyse est l'obtention de bornes sur l'erreur commise par les modèles combinatoires dans l'approximation du revenu logit. Nos essais numériques montrent qu'une approximation combinatoire rudimentaire permet souvent d'identifier des solutions quasi-optimales. Dans le troisième article, nous considérons l'extension du problème à une demande hétérogène. L'affectation de la demande y est donnée par un modèle de choix discret logit mixte où la sensibilité au prix d'un usager est aléatoire. Sous cette modélisation, l'expression du revenu n'est pas analytique et ne peut être évaluée de façon exacte. Cependant, nous démontrons que l'utilisation d'approximations non linéaires et combinatoires permet d'identifier des solutions quasi-optimales. Finalement, nous en profitons pour illustrer la richesse du modèle, par le biais d'une interprétation économique, et examinons plus particulièrement la contribution au revenu des différents groupes d'usagers. / The network pricing problem consists in finding tolls to set on a subset of a network's arcs, so to maximize a revenue expression. A fixed demand of commuters, going from their origins to their destinations, is assumed. Each commuter chooses a path of minimal "disutility", a measure of discomfort associated with the use of a path and which takes into account fixed costs and tolls. A deterministic modelling of commuter behaviour is mostly found in the literature, according to which positive flow is only assigned to \og shortest\fg\: paths. Even though the determinist pricing model is amenable to global optimization by the use of enumeration techniques, it has often been criticized for its lack of realism. In this thesis, we consider a probabilistic extension of this model involving a logit dicrete choice model. This more realistic model is non-linear and non-concave, but still possesses strong combinatorial features. Our analysis spans three separate articles. In the first we tackle the problem from a theoretical perspective for the case of a single origin-destination pair and develop a first order analysis that exploits the logit assignment analytical properties. We show the validity of simplification rules to the network topology which yield a reduction in the problem dimensionality. This enables us to establish the problem's unimodality for a wide class of topologies. We also establish a parallel with the product-line pricing problem, for which we generalize some of our results. In our second article, we address the problem from a numerical point of view for the case where multiple origin-destination pairs are present. We work out algorithms that exploit both local information and the pricing problem specific combinatorial features. We provide theoretical results which put in perspective the deterministic and probabilistic models, as well as numerical evidence according to which a very simple combinatorial approximation can lead to the best solutions. Also, our experiments clearly indicate that under any reasonable setting, the logit pricing problem is much smoother, and admits less optima then its deterministic counterpart. The third article is concerned with an extension to an heterogeneous demand resulting from a mixed-logit discrete choice model. Commuter price sensitivity is assumed random and the corresponding revenue expression admits no closed form expression. We devise nonlinear and combinatorial approximation schemes for its evaluation and optimization, which allow us to obtain quasi-optimal solutions. Numerical experiments here indicate that the most realistic model yields the best solution, independently of how well the model can actually be solved. We finally illustrate how the output of the model can be used for economic purposes by evaluating the contributions to the revenue of various commuter groups.

In the modern industrialized society, a long commuting time is becoming more and more common. However, commuting results in a number of different costs, for example, external costs such as congestion and pollution as well as internal costs such as individual time consumption. On the other hand, increased commuting opportunities offer welfare gains, for example via larger local labor markets. The length of the commute that is acceptable to the workers is determined by the workers' preferences and the compensation opportunities in the labor market. In this thesis the value of travel time or commuting time changes, has been empirically analyzed in four self-contained essays. First, a large set of register data on the Swedish labor market is used to analyze the commuting time changes that follow residential relocations and job relocations. The average commuting time is longer after relocation than before, regardless of the type of relocation. The commuting time change after relocation is found to differ substantially with socio-economic characteristics and these effects also depend on where the distribution of commuting time changes is evaluated. The same data set is used in the second essay to estimate the value of commuting time (VOCT). Here, VOCT is estimated as the trade-off between wage and commuting time, based on the effects wage and commuting time have on the probability of changing jobs. The estimated VOCT is found to be relatively large, in fact about 1.8 times the net wage rate. In the third essay, the VOCT is estimated on a different type of data, namely data from a stated preference survey. Spouses of two-earner households are asked to individually make trade-offs between commuting time and wage. The subjects are making choices both with regard to their own commuting time and wage only, as well as when both their own commuting time and wage and their spouse's commuting time and wage are simultaneously changed. The results show relatively high VOCT compared to other studies. Also, there is a tendency for both spouses to value the commuting time of the wife highest. Finally, the presence of hypothetical bias in a value of time experiment without scheduling constraints is tested. The results show a positive but not significant hypothetical bias. By taking preference certainty into account, positive hypothetical bias is found for the non-certain subjects.

Parmi les méthodes d’estimation de paramètres de loi de probabilité en statistique, le maximum de vraisemblance est une des techniques les plus populaires, comme, sous des conditions l´egères, les estimateurs ainsi produits sont consistants et asymptotiquement efficaces. Les problèmes de maximum de vraisemblance peuvent être traités comme des problèmes de programmation non linéaires, éventuellement non convexe, pour lesquels deux grandes classes de méthodes de résolution sont les techniques de région de confiance et les méthodes de recherche linéaire. En outre, il est possible d’exploiter la structure de ces problèmes pour tenter d’accélerer la convergence de ces méthodes, sous certaines hypothèses. Dans ce travail, nous revisitons certaines approches classiques ou récemment d´eveloppées en optimisation non linéaire, dans le contexte particulier de l’estimation de maximum de vraisemblance. Nous développons également de nouveaux algorithmes pour résoudre ce problème, reconsidérant différentes techniques d’approximation de hessiens, et proposons de nouvelles méthodes de calcul de pas, en particulier dans le cadre des algorithmes de recherche linéaire. Il s’agit notamment d’algorithmes nous permettant de changer d’approximation de hessien et d’adapter la longueur du pas dans une direction de recherche fixée. Finalement, nous évaluons l’efficacité numérique des méthodes proposées dans le cadre de l’estimation de modèles de choix discrets, en particulier les modèles logit mélangés. / Maximum likelihood is one of the most popular techniques to estimate the parameters of some given distributions. Under slight conditions, the produced estimators are consistent and asymptotically efficient. Maximum likelihood problems can be handled as non-linear programming problems, possibly non convex, that can be solved for instance using line-search methods and trust-region algorithms. Moreover, under some conditions, it is possible to exploit the structures of such problems in order to speedup convergence. In this work, we consider various non-linear programming techniques, either standard or recently developed, within the maximum likelihood estimation perspective. We also propose new algorithms to solve this estimation problem, capitalizing on Hessian approximation techniques and developing new methods to compute steps, in particular in the context of line-search approaches. More specifically, we investigate methods that allow us switching between Hessian approximations and adapting the step length along the search direction. We finally assess the numerical efficiency of the proposed methods for the estimation of discrete choice models, more precisely mixed logit models.

Parmi les méthodes d’estimation de paramètres de loi de probabilité en statistique, le maximum de vraisemblance est une des techniques les plus populaires, comme, sous des conditions l´egères, les estimateurs ainsi produits sont consistants et asymptotiquement efficaces. Les problèmes de maximum de vraisemblance peuvent être traités comme des problèmes de programmation non linéaires, éventuellement non convexe, pour lesquels deux grandes classes de méthodes de résolution sont les techniques de région de confiance et les méthodes de recherche linéaire. En outre, il est possible d’exploiter la structure de ces problèmes pour tenter d’accélerer la convergence de ces méthodes, sous certaines hypothèses. Dans ce travail, nous revisitons certaines approches classiques ou récemment d´eveloppées en optimisation non linéaire, dans le contexte particulier de l’estimation de maximum de vraisemblance. Nous développons également de nouveaux algorithmes pour résoudre ce problème, reconsidérant différentes techniques d’approximation de hessiens, et proposons de nouvelles méthodes de calcul de pas, en particulier dans le cadre des algorithmes de recherche linéaire. Il s’agit notamment d’algorithmes nous permettant de changer d’approximation de hessien et d’adapter la longueur du pas dans une direction de recherche fixée. Finalement, nous évaluons l’efficacité numérique des méthodes proposées dans le cadre de l’estimation de modèles de choix discrets, en particulier les modèles logit mélangés. / Maximum likelihood is one of the most popular techniques to estimate the parameters of some given distributions. Under slight conditions, the produced estimators are consistent and asymptotically efficient. Maximum likelihood problems can be handled as non-linear programming problems, possibly non convex, that can be solved for instance using line-search methods and trust-region algorithms. Moreover, under some conditions, it is possible to exploit the structures of such problems in order to speedup convergence. In this work, we consider various non-linear programming techniques, either standard or recently developed, within the maximum likelihood estimation perspective. We also propose new algorithms to solve this estimation problem, capitalizing on Hessian approximation techniques and developing new methods to compute steps, in particular in the context of line-search approaches. More specifically, we investigate methods that allow us switching between Hessian approximations and adapting the step length along the search direction. We finally assess the numerical efficiency of the proposed methods for the estimation of discrete choice models, more precisely mixed logit models.