Essays on transport economics

Abstract

The essays that compose this dissertation address different topics on the economics of transportation. The first two essays analise the demand side of transportation services. In the first essay, we study the phenomena related to the first two steps in the traditional four-step model for traffic assignment: trip generation and trip distribution. The trip generation step determines the frequency or volume of trips in each zone, and the trip distribution step matches the origin with the destination of corresponding flows. The second essay, which focuses on transport mode choice, tackles the third step of the process in an attempt to gain knowledge about transportation customers. The final essay aims to evaluate a fundamental aspect of the supply side of the transportation sector by computing the services that the transportation infrastructure contributes to the regional economy. In the first essay, we analyse the existence of a dynamic component of road freight flows. The main contribution of this portion of the thesis is to specify a model that applies a dynamic version of the gravity equation. Two lags of the dependent variable, the log of freight goods transported from one region to another, were included as explanatory variables. It was then statistically tested whether systematic changes in the past values of road freight flows positively affect the current values. In the beginning section of the essay, we reviewed the literature related to gravity models and focused on their dynamic version. Although this type of inertia has not been extensively addressed in the transportation literature, scholars have investigated its effects in international trade. According to these authors, businesses may make investments in developing their commercial networks to increase their presence in certain markets. These investments (which can be thought of as sunk costs) signal long-term planning that might partly influence the patterns of freight movements to maintain their current trends. In the empirical portion of the essay, we explore a dynamic specification of the gravity equation with some peculiarities. The model includes common regressors, such as GDP measures and distance, as drivers and deterrents of transportation flows, respectively. In addition, a measure of the quality of transportation infrastructure was used as an explanatory variable. Its inclusion attempted to test whether better infrastructure favored the transportation of goods in an aggregate demand framework. The data used to calibrate the model consist of a balanced panel set of road freight flows among Spanish regions over the 1999-2009 period. The results obtained confirm the theory of the gravity model and find a significant and positive effect for the scale measures of the regions, measured as GDP, and a negative effect for distance. The constructed proxy of the quality of infrastructure also indicates that providing more road infrastructure benefits road transportation flows. The key result shows that there is a positive effect of lagged freight variables in the explanation of current freight variables. Although its magnitude is not large, this result indicates that the inertia of flows must be taken into account to avoid model misspecification. As a complementary exercise, CO2 emissions emanating from road freight transportation are calculated in a straightforward fashion. An equation relates the total CO2 emissions per year through the interaction of data with respect to transport operations between each pair of regions and some average values on diesel consumption and distance travelled. Using the aggregate transport operations collected in the database and, alternatively, those predicted by our model, we provide yearly data on CO2 emissions caused by the most used mode of transportation in Spain. Based on the above discussion, it appears that the volume of CO2 emissions in recent years is quite stable. This outcome is a bit surprising, as we expected a larger decrease in CO2 emissions resulting from the reduction of traffic movements. Nevertheless, it seems that the economic slowdown may have had additional effects apart from reducing the pace of growth of road freight flows. The analysis of the data indicates that the vehicles¿ renewal rate is much lower than in previous years, which is a product of the economic environment. Considering that older vehicle models are less energy efficient and pollute more than newer versions, policy-makers must take these results into account if a reduction in greenhouse gases is among their primary goals. This work is already being extended in my current study that involves performing an analysis of microdata on vehicles shipping freight goods. The main objective is to measure the impact of distance on decisions about vehicle shipment size. Many transportation models assume that the vehicles used in transportation activities are of optimal size. Therefore, it is also assumed that this optimal outcome is the result of minimizing total logistics cost in the Economic Order Quantity (EOQ) model. This function typically includes a trade-off between transport costs and inventory costs; as the shipment size increases, transport costs decrease (although inventory costs increase). In such an optimization process, economies of distance and economies of scale are likely to be important. Economies of distance occur when transport rates taper as haulage distance increases, whereas decreasing transport rates with respect to increasing haulage quantity for a given distance is typically known as economies of scale. For many years, both effects have been assumed in many transportation models without being properly tested. Currently, I am working on an empirical specification that allows us to test the latest theoretical developments related to this issue. Another potential extension of this work is the application of spatial econometric techniques to the gravity model primarily by estimating a specification of the model that accounts for spatial dependence among the regions. Conventionally, it has been accepted that distance functions in origin-destination models were enough to capture the possible spatial dependence of flows. However, recent discussions have shown that residuals of aspatial specifications of gravity models exhibit spatial dependence. The latest developments of spatial econometric models using interaction data point to the need to test the inclusion of the following three different spatial effects: ``origin-based", ``destination-based" and ``origin-destination-based" spatial dependences. The first, ``origin-based" spatial dependence, would include a spatial lag of origin flows, i.e., an average of the flows from neighbors to the origin to the destination region. The second type of spatial dependence, ``destination-based" spatial dependence, is intuitively explained as the forces leading to flows from an origin state to a destination state that create similar flows to nearby or neighboring destinations. Finally, ``origin-destination-based" spatial dependence is a combination of these two types of spatial dependence and reflects an average of flows from neighbors of the origin state to neighbors of the destination state. The model allows test whether these effects are significant and finds the appropriate empirical specification. The second essay aims to study the determinants of the choice of long-distance travel modes. In practice, this chapter focuses on how socio-economic characteristics of travelers, land use features and trip attributes shape choices among a set of ground transportation modes consisting of private car, public bus and train. Mode-choice models have been widely studied in the transport economics literature in the context of daily commuting or short-distance trips. However, data on long-distance or inter-city movements are not as common, which makes the analysis of this topic less usual. Although commuting is a daily activity for almost everyone, long-distance travel is less frequent for all except a few job positions, such as salespersons or diplomats. The set of differences between short- and long-distance trips includes differences in travel times and costs incurred. Generally, more time and out-of-pocket costs are required for inter-city trips, which make the traveler more likely to study the supply of services more closely. In addition, travel purposes and mode availability might differ. In the literature review section, we provide up-to-date empirical evidence about the effect of socio-demographic, land use and trip attribute variables on the selection of the mode of transportation regarding inter-city trips. The data description of mobility survey (Movilia 2007) plays a fundamental role in this essay. The objective of this survey is to study the main features of long-distance trips and travelers using telephone interviews as the main data collection method. More than 19,000 microdata observations were analyzed and described, and these provided the main substance of the database. This preliminary analysis combined with a literature review help properly address the research question in two ways. First, it helped check the consistency of the results to be obtained by the multivariate analysis. Second, it allowed us to identify the drawbacks of the database. In general, the data are rather powerful and rich, although some caveats from the spatial point of view must be flagged. An ideal database would include data on time of the trip, in addition to accurate geographical information on the points of origin and destination. With this information in hand, it would be possible to calculate interesting variables, such as accessibility measures to public transportation services, origin and destination land-use indices and ingress and egress times to transport stations. The empirical approach adopted attempts to overcome these drawbacks by exploiting the hierarchical structure of the data. The multinomial model allows the probabilities of choosing one specific mode of transportation over another that was previously set as the base category to be computed. The multilevel version of the multinomial model permits estimating random intercepts associated with higher levels that nest individual observations. In this essay, these intercepts are related to the departure provinces collecting information about unmeasured characteristics that are shared by all travelers with origins in a certain province. In this way, we are able to calibrate the model and take the spatial heterogeneity of trip-makers into account. Among the avenues for future research, I would like to explain the variation in these random intercepts to find the drivers that cause spatial heterogeneity in the choice of long-distance transport modes. In particular, future arrangements might include studying the effects of accessibility to transport stations and the impact of the degree of land-use characteristics. Controlling for these variables, almost all other spatial information except preferences would be erased from the provincial intercepts. My interests now proceed into the details of long-distance mobility of low-income populations and particularly the elderly and the unemployed. A large portion of these population groups make no trips at all because of economical reason (in all likelihood), but I am also interested in analyzing the importance of other explanatory variables. Using a double-hurdle type of model, I aim to investigate the socio-demographic and spatial factors that affect the decision whether to travel or not and to model the choice of transportation mode in a second stage. The third issue attempts to answer a question that has been at the center of the economic debate for a long time. In particular, we are concerned with measuring the economic effects of public infrastructure on the productivity of regions. Much emphasis is put on capital related to transportation infrastructure, particularly in the Spanish road transportation network. The contribution to the vast literature is twofold. First, we apply the latest advances in spatial econometrics modeling that allows us to test whether spatial dependence of the dependent and the independent variables must be included in the model specification, which has been labeled the Spatial Durbin Model in the literature. The second contribution attempts to overcome another drawback in the literature, i.e., the use of stock variables instead of flows. Measuring private and public capital as a stock adds information about the quantitative properties of the infrastructure but not the services it provides. In this sense, stock measures are combined with information about the vehicles potentially using the infrastructure to obtain a proxy of the transportation services provided by this type of public capital. Although this is not the first time this Spatial Durbin Model has been applied to regional production functions, previous authors have typically focused on its capabilities to account for the spillover effects of independent variables by using spatial lags of these variables. For studying transportation infrastructure spillovers, this feature might be particularly important because it allows the researcher to capture those effects of the infrastructure that go through the network and affect neighboring regions. The specification with the lag of the dependent variable has been supported by econometric reasons but has remained unexplained from the economic point of view. In this chapter, we provide theoretical justification for its inclusion by showing how the use of private capital in a region might be influenced by the business cycles of its neighboring regions. The resulting model specification is applied to Spanish provinces (NUTS-3) during the 1986-2006 period, which was a period that witnessed heavy investments to increase the quality of Spain's road transportation network. Our main empirical result appears to support the idea that better transportation services increase regional productivity. However, this result might be a straightforward justification to maintain investments in upgrading the infrastructure, and we suggest it be taken with caution. From our perspective, this outcome of the model is important in explaining the productivity increases in that period; however, we are aware that the marginal contribution of the early rounds of upgrading certain corridors might be much more productive than subsequent contributions. Additional results are related to the significant spillover effects caused by transportation infrastructure. According to the estimations, an increase in transportation services provided by road infrastructure in one province raises productivity in neighboring units by approximately one-half the amount of the improvement in the spatial unit in which the infrastructure is located. This result supports the existence of spillover effects and shows that the potential impacts of some public infrastructure projects are not confined to the geographical area in which they are located. Thus, the network characteristics of this public capital are confirmed and might suggest the need for coordination among administrators in charge of transportation infrastructure planning. In the case of Spain, national, regional and provincial governments share this task. As a policy matter, we suggest testing the inclusion of spillover effects even in the analysis of those projects that are not particularly inter-regionally oriented. An alternative version of this essay, which has recently been recently chosen to be part of a book edited by Edward Elgar Publishing, uses recent findings on the role of accessibility to study its consequences on regional economy. In addition, we are undertaking the development of a formal background of the Spatial Durbin Model based on New Economic Geography foundations. The objective is to provide other scholars with a robust theoretical framework when using this technique. Other potential work derived from this last chapter might involve the estimation of a spatial stochastic frontier to study the efficiency of provinces. Finally, we would also like to calculate the usage level of public and private capital in the Spanish provinces.