Mobility innovation SCIENCE & RESEARCH International Transportation (70) 1 | 2018 33 Vehicle Stock Modelling: A new approach Forecasting based strategy development for new-mobility solutions Vehicle stock model, Mobility innovations, Market acceleration forecasts, Time-series-analysis, Planning measures Forecasting mobility and travel demand with the aid of dedicated model-based approaches is a recognized method to deal with challenges related to urban transport planning as well as the fulfilment of political goalsetting. Nevertheless, even if for the transport segment a large variety of forecasting models exists, the specification on limited purpose forecasts doesn’t meet the requirements of integrated, realistic, longterm planning measures. The presented vehicle stock model as a generic, multi-purposeoriented forecast tool closes this gap with a new, time series analysis based approach. Lea Heinrich, Felix D. Segel, Wolfgang H. Schulz H ow can new business models be developed and integrated into a saturated market? How can we manage mobility challenges such as environmental and resource constraints, infrastructure or funding - and how can change be addressed and managed? These are just three questions out of a broad set, the actors in the mobility and transport segment must deal with today. Climate change is a huge challenge for the society. Many countries of the European Union and the European Commission itself have set ambitious emission reduction targets to counteract the trend. These upheavals and paradigm shifts affect the transport sector, as a significant part of the emissions can be allocated to road transport with accounting more than 70 % of all greenhouse gas emissions in the EU28 in 2014 [1]. Nevertheless, the automotive industry - especially in Germany - on the other hand is a major driver for economic growth and bearer of innovation capacities. The automobile market therefore in recent times goes through radical structural changes that are mainly driven by increasing competition, pressure to innovate, system dynamics and a demanding business environment due to societal developments and political goal-settings. One major challenge for the automotive industry therefore is demand-focused innovation: To unlock market potentials by offering products and services that are promising to meet consumer needs with respect to regulatory guidelines and policy. The achievement of this goal is linked to various challenges regarding cross-industrial and cross-sectoral cooperation in the field of technology development as well as market deployment. The following statement with reference to autonomous vehicles (AVs) clearly demonstrates, the actors are aware of those challenges, they cannot properly predict and include appropriate measures in their strategies. “In the meantime, the industry will have to navigate through a number of difficult challenges and figure out how to take advantage of a few surprising opportunities. But few industry players are adequately prepared for — or even willing to fully acknowledge — the hurdles they must clear before AVs are able to produce real revenue.“ [2] Even if innovation activities require complete realignments of business activities and core competencies - the research and development activities must be separated from the actual market introduction decision. Political objectives as driving forces on the one hand enable faster product developments with less investments due to federal framework programmes including subsidies for cooperative, cross-sectoral technology developments. On the other hand, the actual market deployment decision of those R&D activities underlies high uncertainty, as this decision is accompanied with investments that are out of the subsidy scope and bear AUF EINEN BLICK Die Abschätzung der zukünftigen Mobilitäts- und Verkehrsnachfrage ist eine zentrale Herausforderung, sowohl bei der Entwicklung von Stadtverkehrskonzepten als auch bei der Erfüllung politischer Zielsetzungen. Um diesen planerischen Herausforderungen begegnen zu können, werden meist individuelle Berechnungsmodelle entwickelt, was zwar zu einer Vielzahl an verfügbaren Methoden führt, die jedoch auf spezifische Problemstellungen begrenzt sind. Das in diesem Artikel präsentierte Prognosetool schließt diese Lücke mithilfe eines neuen, zeitreihenbasierten Analyseansatzes der es erlaubt, Anwendungsbereich übergreifende Vorhersagen zu den technologiebasierten Entwicklungen im Transport- und Mobilitätssegment zu treffen. SCIENCE & RESEARCH Mobility innovation International Transportation (70) 1 | 2018 34 high risks in terms of user acceptance and business potentials accordingly. As the market deployment of new technologies in the mobility and transport segment in most cases goes along with the adoption of existing infrastructure settlements and regulations. A comprehensive market deployment of promising technologies therefore mainly depends on adequate framework conditions that meet the requirements of working systems that - in additionare accepted by the society to exploit full potentials. In the past two decades, several researchers have sought to determine which approaches, methods and models are suitable for emission value control and for successfully reducing emissions in the transport sector, refocusing on vehicle technologies such as electric vehicles (EVs) and ensuring market deployment in terms of creating user benefits and realizing economic feasibility. The different methodologies and research approaches concerning market potentials are reflected in different forecasting models, whereby the focus areas and goal-settings regarding the targeted outcome of findings diverge in the level of specification on certain research topics. Policy assessment models, just as TREMOVE [3] are representative approaches that support decision-making on governmental level. As the focus herewith is clearly set on emission value and policy objective scenarios, the practicability of this approach must be questioned in terms of the applicability and transferability on other investigation scopes such as potential business models or customer acceptance. Further approaches (e.g. SEiSS, COPERT, BIG) [3, 4, 5, 6] that have been examined are using potential evaluation tools and forecasting methods based on comprehensive problem analyses. Nevertheless, a specific demand-oriented potential assessment that addresses overall economic interests including macroeconomic as well as business economic forecast scopes and that allows to derive policy-, industryand market-oriented guidance on specific geographic levels (city, state, federal government) considering existing and desired framework conditions is not available yet. The aim of the presented model approach therefore is to close this gap with the conception of a vehicle stock model (VSM) that delivers forecasts on new technology market success with the inclusion of external impact factor’s effects. This overall approach is not limited to a single investigation objective and offers high transferability potentials. The Model The proposed vehicle stock model (VSM) builds upon four driving forces which are assumed to determine the number of vehicles equipped with new technology capabilities, namely (1) OEM strategy, (2) customer demand, (3) political influence and (4) infrastructural factors. Since the first force raises needs for explanation, it is worth to add, that the VSM is based on the assumption, that development of new vehicle technologies must be completed before customers can decide whether to buy or use them. Therefore, market launches of new models are included within the OEM’s roadmaps and hence can be observed for a targeted forecasting period. Difficulties regarding this task are likely to occur due to data availability issues, because OEMs traditionally keep their specific roll-out dates secret. Further, uncertainties exist considering the type of technology introduction as an upgrade, basic or premium feature, leading to significantly different rates of equipped vehicles. By analysing historical data of vehicle registrations or stocks, the authors aim to build a base case forecast Figure 1: Basic model, refinement and evaluation scheme Input sources Form of output Qualitative survey Expert judgement Time series analysis • New vehicle sales split along three dimensions: 1. Private/ business owners. 2. Type of engine. 3. Geographical distribution. Identification of blind spots • Household survey to identify • demand (private and commercial use), • replacement rates, • price sensitiveness, purchasing intention • geographic “hotspots”, based on revealed and stated preferences. • Business model evaluation. • Political and legal developments. • Demand development. • Technology evaluation and technological developments. • Integration of interaction effects (e.g. more business owners in urban areas). Base case scenario Refined scenarios Use cases • Expected development of upper class models until December 2020. • Static/ dynamic development of split dimensions. • Provides insights on an averaged basis. • Identification of relevant use cases with regards to estimated ramp-up development. • Allow final recommendations. • Enables market size estimation of different cases. Mobility innovation SCIENCE & RESEARCH International Transportation (70) 1 | 2018 35 which is enriched by qualitative methods (e.g. survey, expert interview). Afterwards, the forecast can be modified with a bottom-up (e.g. fixed rate of newly registered electric vehicles) as well as a top-down approach (e.g. reaching a specified market share at a certain point in time), resulting in different scenarios and use cases. This capability ensures that the VSM is resilient in in terms of changing framework conditions. At this stage, it is important to note that the authors assume that new car model registrations take place in an efficient market. Analogues to financial stock data, these numbers reflect current demand, prices and every other possible factor. As an example, if an OEM promotes a certain car model with substantial discounts, model registrations of this vehicle are assumed to increase given that fact that this is a relevant variable for consumers. In contrast to time series analyses, a factorial regression model which e.g. includes list prices that are not affected by these promotional activities would not be able to capture such developments. Therefore, trying to analyse this data with a finite set of explanatory variables could let to significant blind spots and hence biased forecasts. Applying time series analyses as the primary tool circumvents such inherent model risks. Additionally, being able to expand the analyses with explanatory variables provides sufficient flexibility as well as granularity to account for certain events like the German scrapping incentives, OEM pricing strategies or other qualitative features of individual car models. In this case, the model would combine the best of both i.e. time series and conventional regression. Nevertheless, the evaluations might leave certain questions unsolved when it comes to the qualitative analysis of the forecasting data. Even if the intention of the presented model is to be independent from concrete information on model prices and related market acceleration effects, a validation of those assumptions seems necessary. The research gap by now is the lack of proof concerning a concrete use-case that currently is of high relevance and that clearly demonstrates the benefits of the presented model approach including the altering influencing parameters (refinement factors) that must be investigated. Figure 1 presents the general architecture of the model, including influencing variables and expected outputs. For the time series analysis, we used data from the German Kraftfahrtbundesamt (KBA) [7, 8, 9] which is the Federal Motor Transport Authority. The focus within this exemplified calculation lies on so called upper class vehicles. To determine which car models fit into this segment, we followed the categorization of the KBA. Based on their recent publications, we determined a selection of nine vehicles representing over 90 % of the “upper class” segment since 2014. Beside one model (Volkswagen Phaeton) the subset is likely to be in production until 2020 and hence provides a sufficient basis for our forecasts. The data set covers the time span from January 2004 until December 2017 and includes the monthly number of new registrations into the German car register for each model. Looking at these numbers, one can clearly see that they are quite noisy (see figure 2). To identify a long-term trend development, time series decomposition is applied to extract seasonality as well as random noise. Forecasting a Scenario This part will illustrate the basic capabilities for scenario forecasting of our VSM. For this purpose, we will examine the hypothesis that the German Government will introduce regulations which predetermine hybrid and electro vehicle registration rates at a fixed level from January 2018 on. We will have a specific look at BMW 7 Series registrations and how these would be affected by the regulation. First, we forecast car model registration numbers. An ARIMAX (autoregressive integrated moving average with exogenous inputs) model is applied to the time series of BMW 7 Series registrations. Since we already observed that the registration development is not just influenced by seasonality but also driven by product life cycles, we need to account for this dependency via an exogenous input. Specifically, we integrate the age of the production series measured in years. Using a regression analysis, it can be 0 200 400 600 800 1000 2004 2006 2008 2010 2012 2014 2016 2018 Monthly car registrations: BMW 7 Series Actual Registrations Trendline (MA with n = 6) Figure 2: Historic development of BMW 7 Series registrations. Considering that this model has seen major upgrades in November 2008 as well as in October 2015, the trendline is driven by the model’s product life cycle. This relationship is typical for most of the analyzed models. 0 200 400 600 800 1000 2004 2006 2008 2010 2012 2014 2016 2018 2020 Forecast of new BMW 7 Series registrations Actual Registrations Trendline (MA with n = 6) Forecast Figure 3: Forecast of BMW 7 Series registrations including empirically observed development. Based on an ARIMAX (0,1,2) (2,0,0) model chosen by using the Akaike information criterion. Coefficients are significant on a 90 % level of significance. Evaluating the model accuracy with estimated coefficients based from 2004 to 2014, we calculate approx. 3 % deviation in cumulated registrations from January 2015 to December 2017. SCIENCE & RESEARCH Mobility innovation International Transportation (70) 1 | 2018 36 shown that after eliminating the effect of the product cycle, the time series is stationary over time. After selecting an appropriate model, a forecast is conducted until the end of 2020. Figure 3 presents the empirically observed trend up to December 2017 as well as the subsequent forecast. In accordance with the analysis of historic registrations, it is not surprising that the model’s forecast primarily follows the expected product life cycle, not including a major model upgrade. Returning the focus to our example scenario, we applied a fixed minimum registration for hybrid and electro vehicles rate of 10 %, with the results illustrated in figure 4. Considering that China is assumed to introduce similar regulations [10] and historic hybrid registrations of BMW 7 Series cars amount to approx. 7 % on average, this rate seems to be within a realistic probability space. Conclusion and Outlook The overall assumption regarding the applicability of the presented VSM approach was, that with the used methods a far more precise and realistic forecast on new technology market deployment and diffusion can be given, as the likelihood of disregarding continuous and biasing occurrences is lowered. This on the one hand can be reasoned by an extensive observation period. On the other hand, the authors use time series analysis that allows to take account for the characteristics related to new car model registrations. Therefore, the presented model can be regarded as a more detailed bottom-up approach compared to those which have been developed by now. With the presented VSM, more detailed assumptions with shorter forecasting periods can be made. Furthermore, the geographical aspect is addressed by converting the federal state’s vehicle stock share (see figure 5) to the forecasted scenarios, whereby detailed geographical splits and local characteristics and related effects will be included with the model refinement upon data availability. This allows the authors to improve forecasts on both, the macroand the micro-level. Overall, the improvements given by applying the presented method are mainly the benefits that can be realized in terms of specific information distribution for specific stakeholders. For market-oriented stakeholders, e.g. automotive OEMs, the model is beneficial especially in markets characterized by high dynamics and changing framework conditions. In a next step, the authors are in a next step forecasting the market acceleration and business potentials for standardized inductive charging systems for electric vehicles. The technology development by now is conducted within a research project funded by the German government (STILLE) [11], whereby the market introduction is announced for 2020. This use case will clearly outline a representative case that unions all the impact variables influences and the importance for decision making of major stakeholder on industry, consumer, related market actors and governmental level. The basic model’s forecast therefore will be refined by a qualitative analyses of user preferences and market potentials by conducting expert-interviews and consumer-surveys that shall enable the authors to identify the actual influences of user behaviour, price sensitiveness, technology 0 250 500 750 1.000 1.250 1.500 2017 2018 2019 2020 Scenario analysis: Additional electro and hybrid BMW 7 Series registrations (cumulated) Additional electro + hybrid registrations (cumulated) Forecast Figure 4: Additional electro and hybrid BMW 7 Series registrations (cumulated starting at January 2017) based on the scenario that there will be a minimum registration rate of 10 % for electro and hybrid vehicles starting at January 2018. The forecast is based on the model presented in figure 3. Figure 5: Heatmap of historic electro and hybrid registration share in Germany from 2006 to 2016 Mobility innovation SCIENCE & RESEARCH International Transportation (70) 1 | 2018 37 acceptance as well as needs and strategies of local authorities. With applying these aspects, an overall significant demonstration of the model’s capabilities will be made. ■ LITERATURE [1] European Commission (2018): A European Strategy for low-emission mobility. Retrieved 2018-04-04 from https: / / ec.europa.eu/ clima/ policies/ transport_en [2] Hirsh, Evan; Jullens, John; Kalpundi, Ganesh (2016): The Auto Industry’s Real Challenge. Sep 29, 2016. Forbes Media. Retrieved 2017-12-04 from https: / / www.forbes.com/ sites/ strategyand/ 2016/ 09/ 29/ the-auto-industrys-real-challenge/ #2f2c017fc69f [3] Samaras, Zissis (2008): European Database of Vehicle Stock for the Calculation and Forecast of Pollutant and Greenhouse Gases Emissions with TREMOVE and COPER. Final Report. Retrieved 2017-12-07 from http: / / emisia.com/ sites/ default/ files/ 08RE0009V2_ Fleets_Final.pdf [4] Fridstrøm, Lasse; Østli, Vegard; Johansen; Kjell Werner (2016): A stock-flow cohort model of the national car fleet. Retrieved 2017-12-12 from https: / / link.springer.com/ article/ 10.1007/ s12544-016-0210-z [5] Fridstrøm, Lasse; Østli, Vegard (2016): Vehicle fleet forecasts based on stock-flow modeling. Institute of Transport Economics. Retrieved 2017-12-10 from https: / / www.toi.no/ getfile.php/ 1 3 43 8 59/ P ublika sjoner/ T % C 3 % 98 I % 20rapporter/ 201 6/ 1 51 8-201 6/ 1518-2016-sum.pdf [6] Abele, Johannes et al. (2005): Exploratory Study on the potential socio-economic impact of the introduction of Intelligent Safety Systems in Road Vehicles (SEiSS). VDI/ VDE Innovation + Technik GmbH [7] Kraftfahrt-Bundesamt. (2018). Neuzulassungen von Personenkraftwagen nach Segmenten und Modellreihen. Retrieved 2018-03-31 from https: / / www.kba.de/ DE/ Statistik/ Produktkatalog/ produkte/ Fahrzeuge/ fz11/ fz11_gentab.html? nn=1146130 [8] Kraftfahrt-Bundesamt (2018): Neuzulassungen von Personenkraftwagen nach Marken und Modellreihen. Retrieved 2018-03-31 from https: / / www.kba.de/ DE/ Statistik/ Produktkatalog/ produkte/ Fahrzeuge/ fz10/ fz10_gentab.html? nn=1146130 Felix D. Segel Student Assistant at the Chair for Mobility, Trade and Logistics, Zeppelin University, Friedrichshafen (DE) f.segel@zeppelin-university.net Wolfgang H. Schulz, Univ.-Prof. Dr. Owner of the Chair for Mobility, Trade and Logistics and Director of the Center for Mobility Studies | CfM, Zeppelin University, CEO and founder of the Institute for Economic Research & Consulting GmbH (IERC), Friedrichshafen (DE) wolfgang.schulz@zu.de Lea Heinrich, MA Project Coordinator, Zeppelin University, Center for Mobility Studies | CfM, Friedrichshafen (DE) lea.heinrich@zu.de FACE THE CHALLENGES OF MOBILITY Founded in 1949 - bound forward to face the challenges of tomorrow‘s mobility: With an editorial board of renowned scientists and an advisory board of directors, CEOs and managers from all transport industry areas, »Internationales Verkehrswesen« and »International Transportation« - the worldwide distributed English-language edition - rank as leading cross-system transport journals in Europe for both academic research and practical application. Rail and road, air transport and waterway traffic — »International Transportation« and »Internationales Verkehrswesen« stimulate a worldwide interdisciplinary discussion of the numerous defiances in mobility, transport, and logistics. The magazines are targeted at planners and decision makers in municipalities, communities, public authorities and transportation companies, at engineers, scientists and students. With peer-reviewed scientific articles and technical contributions the magazines keep readers abreast of background conditions, current trends and future prospects - such as digitalization, automation, and the increasing challenges of urban traffic. Read more about the magazines and the subscription conditions: www.internationales-verkehrswesen.de www.international-transportation.com INTERNATIONALES VERKEHRSWESEN AND INTERNATIONAL TR ANSPORTATION »Internationales Verkehrswesen« and »International Transportation« are published by Trialog Publishers Verlagsgesellschaft, D-Baiersbronn VISIT US IN BERLIN 26 - 27 June 2018 IV_Image_halb_quer.indd 1 30.04.2018 14: 28: 35 [9] Kraftfahrt-Bundesamt. (n.d.). Neuzulassungen von Personenkraftwagen nach Bundesländern und ausgewählten Kraftstoffarten absolut. Retrieved 2018-03-31 from www. kba.de [10] Frankfurter Allgemeine Zeitung (28.09.2017): Die E-Auto-Quote in China kommt. Retrieved 2018-03-31 from http: / / www.faz.net/ aktuell/ wirtschaft/ unternehmen/ elektroautos-china-fuehrt-die-elektroquote-ab-2019-ein-15222043.html [11] Bundesministerium für Wirtschaft und Energie (BMWi) (2016): ELEKTRO POWER II: Elektromobilität - Positionierung der Wertschöpfungskette Retrieved 2018-03-20 from https: / / www.bmwi.de/ Redaktion/ DE/ Publikationen/ Industrie/ elektromobilitaet-positionierung-der-wertschoepfungskette.pdf? __blob=publicationFile&v=5, p. 15-15