Forecasting of road transport demand is one of the key elements in the transport planning process. Not only the final proposal but the efficiency of the transport system depends upon the quality of the transport demand forecasts. The complexity of the forecasting and/or predicting of road transport demand is acknowledged by many authors through their research on the accuracy of the forecasts as well as the constant creation of the new methods and models along with the models for higher-quality forecasting. However, despite the existence of numerous research by many authors, it is clear that notable changes in the process of transport demand forecasting are needed, shown by the fact that in the last 30, and it is safe to say even 70 years, the accuracy of the forecasts has not significantly been improved. On average, the errors of the transport demand forecasts vary between 20 and 35% depending on the research. The analysis of minimum and maximum values shows forecasted transport demand realization of only 15% up to slightly above 150%. A wide range between minimum and maximum values signifies high standard deviation which results in undercapacitating or overcapacitating of the transport system. Ultimately, both cases result in unplanned economical and financial losses.
When talking about accuracy of the road transport demand forecasts, it is firstly necessary to analyse the potential causes of the inaccuracy. If the reasons and presumptions behind the erroneous results of the road transport demand forecasting are sublimed from the research done so far, three main causes arise. For a long time, low-quality input data used in forecasting process, which does not ensure that a planner can make a quality forecast (e.g. land use, trip generation, trip distribution etc.), is considered as a first cause. It is also important to note that where reference is made to input data as a cause of inaccurate forecasts, forecasting was most often based on the results of the four-step model. In time, several authors confirm that the accuracy of the forecasts does not change significantly regardless of substantial technological progress and a scientific and professional progress of planners themselves. Encouraged by these results, this yields reconsideration of the causes of inaccuracy and assumption is made that this can be attributed to, not only causes of the technical nature such as poor input data, but also to causes relating to human, namely psychological nature. As cause relating to psychological nature, the optimistic bias of the planner is stated. As the analysis of the accuracy has determined among others, the cases of the forecasting significantly below the real values, it is apparent that the cause is not always the optimistic bias hence it is concluded that the third most common cause is that of political and economic nature.
No matter the unassailable fact of the necessity and importance of the road transport forecasting, as well as a low accuracy of the existing forecasting models, exact models for validation of the quality of the transport demand forecasts in road transport have not been ascertained. One of the main reasons for facilitated manipulation of the forecasts (no matter whether it is caused by optimism of the planner or by politico-economical nature) is difficulty of the validation of forecasts results. Accordingly, research of this thesis is directed at the establishing the link between movement of annual rate of road transport demand change and economic, demographic and transport factors. That way it is possible to analyse economic, demographic, and transport trends by the means of the result of transport forecast, and consequently, easily conclude about the probability of occurrence of the specific result, i.e. about the quality of the forecast.
In this doctoral thesis, as a first step, defining of the relevant economic, demographic and transport factors for modelling of the road transport demand is conducted. On the basis of the results of the descriptive statistics, graphical and correlation analysis, it is determined that the number of registered motorised road vehicles (average annual change rate) followed by the real net wage growth (average annual change rate), real GDP growth rate and the number of tourist arrivals (average annual change rate) bears the biggest impact on the trend of the transport demand. It was also found that the change rate of road transport demand intensity statistically significantly varies depending whether the motorways or state roads are concerned, as it was also found that, if the tourism activity is not taken into consideration, there is no statistically significant difference in relation to spatial distribution in the Republic of Croatia. This is corroborated by the fact that the statistically significant difference of GDP growth rate by county was not established. The correlation with the population has not also been proven as statistically significant and has a negative trend, which is expected when taken into consideration the fact that the population in the Republic of Croatia is in demographical decline for about 15 years now. As the majority of the previous forecasting models of road transport demand is based on population, it can be concluded that the relation between the population and the trend of transport demand has changed. The fact is that these are older models according to which owning an automobile was not the matter of need but rather prestige and with the development of the motorisation, owning at least one automobile per household is recently implied and is not even uncommon to own two or more automobiles per household. Likewise, if the population of the European Union from 1992 onwards is analysed, it is observed that it is characterised by the migrations that affect the overall increase in population, while the mortality rate exceeds birth rate. When it comes to fuel cost, the negative tendency in relation to the transport demand trend is noted, but not the significant impact, which is also not in accordance with the existing models. Based on determined connections of economic, demographic and transport trends to the changes in intensity of the road transport demand, first supporting hypothesis on the impact of these trends on the trend of road transport demand is confirmed.
After determining the relevant indicators, a model for forecasting the road transport demand was created. Considering the goal of the thesis and a relation between dependent and independent variables, multiple linear regression was used for creation of the model. Seeing as almost all established relevant indicators can be placed into economic category, with the purpose of avoiding multicollinearity and attaining high-quality results, three models were created (V1, V2 and V3). The first model as independent variable uses number of registered motorised road vehicles (average annual change rate), number of tourist arrivals (average annual change rate) and road category (motorway or state road), the second model uses real net wage growth (average annual change rate) and road category (motorway or state road), while the third model for that purposes utilises real GDP growth rate and road category (motorway or state road).
In the second step, based on the created forecasting models of the road transport demands, on established distributions of their independent variables and by applying the Monte Carlo method, models for validation of results of the road transport demand forecast were created. Performance testing of the created validation models was conducted by analysing a success of the validation models not used in the creation of this model, in determining the values of the independent variables needed for the occurrence of the actual trend of transport demand in the period from 2017 to 2019 (annual average rate of change in the intensity of road transport demand (AADT)). Based on the success rate, it has been concluded that the discrepancies are in accordance with the statistical quality indicators of individual models. The results are also indicative of validation models recording smaller discrepancies for validation of forecasts in the longer periods and are more effective in analysis of the state roads by comparison with motorways. Based upon the created models and success rates, the second supporting hypothesis was confirmed stating that by combining deterministic models of economical, demographic and transport historical trends into unique stochastic model, evaluation of the results of transport demand forecasts can be made.
In addition to performance testing, two additional analysis were carried out, in order to test the possibilities of the created validation models. First additional analysis pertains to validation of the forecasts results using other determined/forecasted/envisaged distributions of independent variables within a set period of forecast. Objective of the analysis in question is to determine the quality of the validation model in a case of change in distribution of the independent variables and/or in a case of having quality forecasting data of independent variables. Second additional analysis pertains to the probability of the occurrence (validation) of the established, most commonly used road transport demand forecast in the Republic of Croatia. The purpose of this analysis is to determine the accuracy of road transport demand forecasting in the Republic of Croatia in respect to achieved values and best practice worldwide. According to the obtained results of the first additional analysis it can be concluded that by using higher-quality distributions/forecasting data on the trends of the independent variables in the forthcoming period, a significant impact can be made on the quality of the interpretation of the validation model results and consequently on the correct reasoning, even more so if the results of all three models, as well as trend of independent variables, are analysed and interpreted. The second analysis shows that according to the validation models in 57% of the analysed forecasts, the probability of occurrence of the forecasted value is lesser than 50%, while in 86% of the analysed forecasts the probability of the forecasted value occurring is lesser than 85%. In line with the validation models, low probabilities of the forecasted values occurring, support the necessity of performing detailed analysis of developed forecasts. This is also supported by the fact that, on average, analysed forecasted values differ from actual values by 42%. Based on the success rates of the model and performance testing, it can be concluded that the suitability of the model application is confirmed and so is the hypothesis of this thesis affirming that it is possible to perform validation of the results of road transport demand forecasts using the data on economic, demographic and transport indicators.
In relation to the existing models in the Republic of Croatia and worldwide, created validation model of the results of the road transport demand forecasts allows conducting the analysis of probability of occurrence of the specific road transport demand scenario. Existing models forecast the road transport demand trend without insight into the probability of the forecasted result occurring. Besides that, alterations of thus far established connection between population and road transport demand trend were determined in the process of creating the validation model. Additionally, it is established that the fuel price does not have statistically significant impact on the road transport demand trend, which is not the case for some of the existing validation models for road transport demand trends. In addition, three variants of the validation models were created in this thesis, which reduces the likelihood of faulty reasoning and facilitates the utilization of the model.
Future research should be aimed at the possibility of introducing additional parameters, which will have the effect on increase of accuracy of created models. In relation to that, during the research, it is found that the transport demand trend differs between motorways and state roads and it would be necessary to consider the parameters which would describe the changes on the motorways with more precision. In addition and taking into consideration the fact that the created models are based on the transport demand trends outside of the urban environments, it is necessary for future research to be targeted at forecasts in urban environments with significantly developed public transport. Transport policy can have a significant impact on the road transport demand trends in urban environments regardless of economic indicators, i.e. stimulation of public transport will also have a strong impact on road transport demand trend. Taking that into consideration, it is necessary to conduct research and quantify relations between specific transport policy measures and road transport demand change rate.