The history of the transport future : evaluating Dutch transport scenarios of the past

Introduction

This paper examines 42 business‐as‐usual (BAU) scenarios for future transport and traffic development in the Netherlands from the period 1970 to 2000. An important aspect of these scenarios is that they indicate the potential future state of the transport system if the policies in place at the time continue unchanged, thus raising issues which may require a policy response. This paper shows that in most of the important BAU transport scenario results of the past in the Netherlands, such as future car use and transport emissions, the policy makers were not misled. The prognoses for traffic congestion are an important exception as they underestimated the congestion problems that would arise. This paper shows that, based on the research it examines, BAU transport forecasting is an inaccurate practice. It is recommended that the BAU scenario designer communicates this high inaccuracy, enabling the policy maker to include the inherent future uncertainty in their decision‐making.

Since the 1980s, future scenarios have become an often‐used approach in strategic transport policy making worldwide (e.g. Schafer and Victor, [34]; WBCSD, [49]; Bartholomew, [1]; Petersen et al., [26]). This paper reviews 42 Dutch business‐as‐usual (BAU) transport scenarios (BAU scenarios) from the period 1970 to 2000 (Table 1). BAU transport scenarios have been utilized in the design of new strategic national transport policies since the 1970s (V&W, [43], [44], [46], [47]) and so have both helped to shape transport policy in the past and continue to do so now [see e.g. Netherlands Environmental Assessment Agency (MNP) et al. ([23]), Van Steertegem ([42]), Petersen et al. ([26])]. It is therefore of interest to review the quality of the BAU transport scenarios of the past.

Table 1. The 42 business‐as‐usual scenarios analysed

Institute/author	Short description
Ministry of Transport (1970)	A broad outlook for transport, up to 2000
Scientific Council for Government Policy (1977)—low and high	National outlook, transport included, low and high economic growth assumed, up to 2000
Hupkes (1977)—low and high	Influential Ph.D. thesis, described future transport developments, up to 2000
Ministry of Transport (1986)—high, middle and low	Future Dutch traffic and transport, up to 2010; low, middle and high economic growth
McKinsey (1986)—high, middle and low	Dutch traffic congestion, up to 2010; low, middle and high economic growth
Van den Broecke (1987)—low, middle and high	Influential research: car ownership in 2010; low, middle and high economic growth
RIVMa (1988)—high, middle and low	First Dutch environmental outlook, up to 2010; high, middle and low economic growth
Peeters (1988)—base case	Trend scenario to assess environmental measures, passenger transport, 2010
RIVM (1991)	Second Dutch environmental outlook, up to 2010
NEAb (1992), balanced growthc, European Renaissancec, global shiftc	Specific railways usage scenarios; low, middle and high economic growth
Dutch Railways (1992), base casec	Scenario to estimate rail passenger transport, 2010
RIVM (1993)—European Renaissance andglobal shift	Third Dutch environmental outlook, up to 2015; middle and high economic growth
Ministry of Transport (1993)—ER, ER‐2, GSand GS‐2	Future Dutch transport, middle and high economic growth assumed, up to 2010. Different scenario variations
Peeters (1993)—base case	Trend scenario to assess environmental measures, freight transport, 2010
NEIb (1995)—base case	A general transport outlook, up to 2010
Ministry of Transport (1997)—DE, EC and GC	Future Dutch transport, Divided Europe, European Coordination, Global Competition 2010 and 2020
RIVM (1997)—GC and DE	Fourth Dutch environmental outlook, up to 2020, Global Competition and Divided Europe
CPBd (1997)—DE, EC and GC	Broad outlook, transport included, 1995–2020; Divided Europe, European Coordination and Global Competition
RIVM (2000)—EC and GC	Fifth Dutch environmental outlook, up to 2030, European Coordination and Global Competition
aNational Institute for Public Health and the Environment.bDutch consultancy firms.cTo be found in CE et al. (2000).dThe Netherlands Bureau for Economic Policy Analysis.

There are many different types of scenarios (see e.g. European Environment Agency, [7]), of which BAU scenarios are a specific type which aims to inform the policy maker of what might happen in the future if the already‐existing policies continue to be pursued (Figure 1). By informing the policy maker about a possible future without changing policies, issues are raised which may require a policy response. As well as including only already‐existing policies, BAU scenarios are based on continuity (van der Duin et al., [40]). First, they are based on continuity in circumstances. For example, in BAU scenarios no wars or revolutions are included. Second, the scenarios are based on continuity in the nature of things and in the speed and direction of changes. For example, in BAU scenarios all kinds of future changes are possible, but in most cases neither complete new developments nor abrupt or relatively large changes are assumed. Finally, in BAU scenarios, continuity in the relationship between phenomena is assumed. For example, if it is currently empirically shown that higher income results in more car mobility, BAU scenarios assume that this relationship will also be valid in the future.

In this paper, the main question is whether the BAU scenarios of the past fulfilled their purpose, namely, informing the policy makers correctly about a potential future state of the transport system if the policies in place at the time were pursued? By evaluating the accuracy of BAU transport scenarios in the period 1970–2000, what are the main lessons for future BAU transport scenario‐making?

This paper only reviews BAU transport scenarios which were intended to be used in strategic policy making. By strategic we mean that the scenarios reviewed aimed to show future transport developments on a regional or national scale in order to help shape regional or national transport policies. The ten‐year range between the last scenario study reviewed (in the Year 2000) and the time of writing (the Year 2010) allows us to evaluate long‐term future scenarios, rather than short‐term prognoses.

'Methodology to Evaluate the Accuracy of BAU Transport Scenarios' section contains the methodology used to evaluate the accuracy of the scenarios of the past. 'Driving Forces' section analyses the driving forces in the transport scenarios of the past. The quality of future transport volume indicators and the quality of transport impact outcomes are discussed in sections 'Transport Volume Outcomes' and 'Transport Impact Outcomes', respectively. A discussion and some conclusions follow in section 'Discussion and Conclusions'.

Methodology to Evaluate the Accuracy of BAU Transport Scenarios

Figure 1 depicts how BAU transport scenarios are designed and used.

Graph: Figure 1 Scientists or consultants design a business‐as‐usual scenario resulting in different transport indicator prognoses (e.g., car use, congestion, carbon dioxide emissions). Only alreadyexisting policies are taken into account. Policy makers may use these prognoses to evaluate whether the already‐existing policies are sufficient to meet their policy goals.

In this paper, we assess the accuracy of the BAU transport scenarios by comparing the outcomes of the scenarios made in the past (such as car usage, transport CO2 emissions and so forth) with actual data. In our view, a BAU scenario outcome is accurate if it provides a picture of the future which is consistent with reality, as the BAU scenario designer has then correctly informed the policy maker. However, we are not looking for a perfect fit between the past future assessment and reality, and we do not regard BAU scenario outcomes as predictions. The outcomes of a BAU scenario of the past (e.g. the outcome was the high growth of traffic air pollutants) may not give a picture of the future which is consistent with reality (e.g. an actual decline in traffic air pollutants has taken place). However, if the deviation between the BAU scenario transport outcomes and reality can, to a large extent, be explained by new policies implemented after the publication of the BAU scenario, we denote this scenario outcome as being accurate. After all, stimulating the adoption of new policy is often one of the aims of BAU scenarios.

In contrast, the outcomes of a BAU scenario may give a picture of the future which is not consistent with reality and which cannot be explained by the implementation of new policies after publication of the BAU scenario. We will consider these BAU scenario outcomes as inaccurate, as they may have misinformed the policy makers. It is possible that an unexpected event occurred after the scenario was published. Humans are unable to predict accurately when and where these events might occur. At the same time, humans will continue to face rare and unique events that are completely unexpected and beyond the realm of our imaginations—what Taleb ([38]) has labelled 'Black Swans'. Another reason for inaccuracy could be that the inputs in the BAU transport scenario were wrong (assumptions about future developments in demography, economy and technology, see Step 1 in the 'Design' of a BAU scenario, Figure 1). Or it could be that the model used to translate the inputs into the desired transport indicators (e.g. car kilometres, CO2 emission, amount of future traffic congestion) was, in retrospect, not adequate; see Step 2 in the 'Design' of a BAU scenario, Figure 1. It is also possible that both of these last two reasons—as well as the occurrence of unexpected events—explain why the future, as was assessed in the past, is not consistent with reality.

The 42 Dutch BAU transport scenarios, made in the period 1970–2000, are shown in Table 1. For an in‐depth individual examination and analysis of the scenarios, we refer the reader to de Jong ([5]). For this paper we have selected some of the main transport scenario driving forces (section 'Driving Forces'), transport scenario volume outcomes (section 'Transport Volume Outcomes') and transport scenario impact outcomes (section 'Transport Impact Outcomes'). For each driving force and outcome indicator, we have plotted the scenario estimates as well as the actual realization. This approach gives a more general picture of the accuracy of the Dutch transport scenario results. In the explanatory text alongside each picture we will identify striking details.

The 42 scenarios reviewed were designed by different parties: consultants, ministries and research institutes. In most cases, more than one BAU scenario was designed in each scenario study, in the early days with prosaic names such as 'low' and 'high'. Later, more elaborate scenario story lines were used with names like 'balanced growth', 'European Renaissance' and 'global shift'. Sometimes, scenario studies were inspired by older scenario studies, and sometimes completely new scenario studies were carried out.

Driving Forces

Every transport scenario is required to make assumptions about future demographic, cultural and economic developments (Step 1, Figure 1). Some of these determinants, as used in the 42 Dutch transport scenarios studied, are analysed in this section (Figure 2).

Graph: Figure 2 The assumed future development of Dutch population, economic growth (ten‐year average) and jobs in Dutch transport scenarios of the past. Every little sign (the squares, rounds, triangles, etc.) represents an assumption about the future in one of the 42 Dutch transport scenarios studied. The black lines represent the realized developments as observed by Statistics Netherlands (www.cbs.nl/statline).

The population prognoses in most cases were accurate. There was a tendency towards underestimation, but the direction of the development was estimated rather well. An interesting exception is the prognosis of the Ministry of Transport ([18]): the grey squares (x‐value = 1990, y‐value = 16; x‐value = 2000, y‐value = 18) (in Figure 2) show the population growth forecasts for 1990 and 2000. The 1970 prognosis could be denoted as less accurate as it greatly overestimated population growth. The reason is a 'surprise' (as explained in the section 'Methodology to Evaluate the Accuracy of BAU Transport Scenarios'). The 1970 study was carried out right at the end of the period 1945–68 which is now considered to have been a baby boom period in the Netherlands, as in most Western countries. The 1970 scenario designers thought this population growth would more or less continue. However, in the late 1960s and in the 1970s, great cultural changes occurred, such as the sexual revolution and the related breakthrough of the use of the contraceptive pill. As a consequence, the fertility rate in the Netherlands decreased from around three at the end of the 1960s to around 1.7 in 1980. Later transport scenario studies were able to take these cultural changes into account.

The assumptions about economic growth in the scenarios of the past (Figure 2, bottom left) are diverse, resulting in a spaghetti‐like picture. The gross national product (GNP) growth assumptions in the scenarios of the past vary between a ten‐year average growth of 1.5% and 4.5%. The scenario designers, in particular those, who chose the more extreme values (1.5 and 4.5) could be accused of having assumed rather inaccurate long‐range estimates. As shown, the ten‐year average economic growth values realized between 1970 and 2009 show a cycle of higher and lower growth periods of around 2–3% per year. As economic growth is an important determinant for transport volumes—especially for freight transport—the scenarios which used the more extreme economic growth values could have led to biased BAU transport forecasts. As well as making assumptions about the aggregate economic growth number (Figure 2), it is also important for a transport forecast to make assumptions about the economic growth per economic sector. The total economic growth in a country may be high whilst the freight transport volumes may grow relatively slowly. This could occur in particular if the growth of the not very transport‐intensive 'provision of services' sector explained the aggregate economic growth. This phenomenon actually occurred in the Netherlands from the 1970s onward (KiM, [13]). The transport scenario studies carried out in the 1970s often implicitly assumed that hardly any shift would take place in the future economic growth per sector. They missed the increasing importance of the 'provision of services' sector.

In the transport scenarios studied, future employment growth for the Year 2000 is significantly underestimated (Figure 2, bottom right). This underestimation took place despite the fact that these scenarios assumed future economic growth. So, it seems logical that some future employment growth would also have been assumed. Unfortunately, it is not very clear from these studies why they assumed hardly any employment growth. We can only speculate. Perhaps the main reason is that when these—in retrospective rather inaccurate—prognoses for 2000 were made (in the 1970s and 1980s), the amount of jobs in the Netherlands was rather stable at around 5 million. This long‐standing stability could have enticed the scenario designers to assume a future of continuing stability or even a small decline. However, in reality, employment has risen (Figure 2, bottom right), because of the economic growth in combination with labour market policies implemented in the 1980s (the Dutch labour market became more flexible, Remery et al., [28]). The number of jobs rose quickly after 1985 to 6.5 million in 2000 (De Beer, [4]). Unforeseen cultural changes may also have played a role. In this case, it was the changing cultural ideas about the desirability of women working. In 1970, only 29.4% of the women in the Netherlands had a paid job (SCP, [36]). In 2007, this percentage had grown to 70% (Keuzenkamp et al., [12]), for mostly part‐time jobs.

Transport Volume Outcomes

In transport scenarios, the assumptions about driving forces (section 'Driving Forces') are translated into transport volume indicators. Some of these volume indicators, estimated in the 42 Dutch transport scenarios studied, are analysed in this section (Figure 3).

Graph: Figure 3 The assumed future development of Dutch car use, lorry use and inland shipping in past Dutch transport scenarios. Every little sign (the squares, circles, triangles, etc.) represents an assumption about the future in one of those 42 Dutch transport scenarios. The black lines represent the realized developments as observed by Statistics Netherlands.

The accuracy of the car use prognoses varies (Figure 3, top). Many of the future estimates for 2000 and 2010 do not differ greatly from reality, although there are some estimates which can be considered highly inaccurate. Generally speaking, the later studies from the 1980s and 1990s, which forecasted the situation in 2010, seem to be more accurate than the older studies from the 1970s which made forecasts for 2000. Overall, there seems to be a general tendency to underestimate the growth of car use. This is not too surprising if section 'Driving Forces' is taken into account, which showed that the number of jobs, a driving force for car use, was underestimated in the scenarios. However, highly inaccurate scenarios are often especially interesting from a learning perspective. In this case, the very low car use prognosis for the Year 2000 made in 1977 by the Scientific Council for Government Policy ([35]) (scenario 'Low', see black dot [x‐value = 2000; y‐value = 48] in figure top left) is an interesting scenario because an attempt was made to include political and societal cultural changes in the future story. One can wonder if this is a genuine BAU scenario because, instead of continuity (see section 'Methodology to Evaluate the Accuracy of BAU Transport Scenarios'), breaks in particular trends seemed to have been assumed. However, we think that related to the spirit of the times, this 'Low' scenario by the Scientific Council for Government Policy ([35]) was really meant to provide a kind of BAU scenario. The changes in this 'Low' scenario included a strong redistribution of welfare in the Netherlands (politically driven), resulting in a relatively low future income development, and, consequently, a low growth in car mobility. Also, it was assumed that people would change their mobility behaviour from car use to public transport use. In retrospect, these assumptions could be considered to be wrong. Why were they made? First, in the 1970s, the Netherlands had one of the most left‐wing governments that—as it stated in its government policy statement—aimed to create a 'fair distribution of knowledge, income and power' in the Netherlands (Ramaker et al., [27]). Second, in 1972, the report Limits to Growth (Meadows et al., [17]) was published, which stated that humankind was squandering natural resources by, for example, the seemingly unrestrained growth in the use of oil for cars. This message was for one reason or another highly influential in the Netherlands. The spirit of the time—the rather left‐wing political atmosphere in the mid‐1970s and the influential message of limits to growth—may explain why this, in retrospect highly inaccurate 'Low' car use scenario, was designed.

The bottom left picture (Figure 3) depicts the old BAU prognoses for lorry use (in road vehicle kilometres), showing the tendency to overestimate future growth. Lorries are undivided freight vehicles which are allowed to carry a load of at least 3500 kilograms (including the empty weight of the vehicle). The most important reason for overestimation is that the scenario designers of the past made wrong assumptions about the developments in logistics. Their perspective was that the amount of lorry kilometres would significantly increase in the future, because of increasing distances between goods' production and consumption and the rising popularity of the logistic concept of 'just‐in‐time‐delivery'. Their idea was that the relatively small lorries would be especially suitable for carrying out this logistic concept. In fact, road tractors and vans became relatively popular during the 1980s and the 1990s, and this shift to vans and larger road freight vehicles in combination with improved load factors explains why the lorry kilometres grew more slowly than expected.

For inland shipping, the scenarios of the past also tell an interesting story (Figure 3, bottom right). The 1970 scenario studies estimated very high and inaccurate growth figures, see the 1990 and 2000 forecasts. The scenarios made in the 1980s and 1990s (forecasts for the Year 2010) are of far better quality. The reason that the 1970 forecasts are inaccurate is that they assumed higher economic growth (3–4% yearly average in the long term) compared with what actually happened. Moreover, they missed the trend that led to the Dutch economy shifting to a more 'provision of services' economy (see section 'Driving Forces'), which slowed down inland shipping growth even more.

Transport Impact Outcomes

The scenario outcomes for the impacts of transport are particularly important politically. Environmental and accessibility concerns (related directly to congestion) are major topics in transport politics. Some of these impact indicators, as estimated in the 42 Dutch transport scenarios studied, are analysed in this section (Figure 4).

Graph: Figure 4 The assumed future development of traffic congestion on Dutch trunk roads, road transport carbon dioxide emissions and road transport nitrogen oxides emissions in Dutch transport scenarios. Every little sign (the squares, circles, triangles, etc.) represents an assumption about the future in one of the 42 Dutch transport scenarios studied. The black and grey lines represent the realized developments for traffic congestion as observed by van Mourik ([41]) and for the emissions by Statistics Netherlands (www.cbs.nl/statline).

Traffic congestion forecasting appears to be a very difficult art (Figure 4, top). The strong growth that has occurred was not foreseen in any of the scenario studies. In Figure 4, two lines representing the actual situation are shown: the top black line represents the development of heavy traffic congestion on trunk roads, where the average car speed is less than 50 km/h; the grey bottom line represents the hours lost by vehicles which are unable to travel below the 'free flow' average speed of 100 km/h on. Why has the traffic congestion forecasting in the Netherlands been of such poor quality? We can identify different reasons. One reason is the underestimation of both car use growth in some scenarios (Figure 3) and employment (Figure 2). The employment figures that were forecasted were too low, resulting in the car‐commuting kilometres forecasted being too low. Another reason is that the expansion of the actual road capacity in the Netherlands has been carried out a far slower rate than was expected in the past. Not even all road extension projects announced in V&W ([44]) have been carried out yet. An important explanation for this is that the Western part of the Netherlands is a very densely‐populated area (around 900 inhabitants per square km). On the one hand, this density makes the traffic in this area highly congested. On the other hand, it also makes it politically difficult to extend roads because of the potential negative impacts of new roads, like air and noise pollution, on the many local residents. One could state that the traffic congestion scenario designers of the past were too optimistic about the political and practical possibilities of road extensions. A third reason for the poor traffic congestion scenario outcomes of the past is road pricing. Since 1990 (V&W, [44]), a huge political debate on implementing road charging has taken place in the Netherlands, even though an actual scheme for road charging has not yet been implemented. Some scenario designers in the past included some modest form of road pricing in their BAU scenarios as an existing policy although they were neither large nor very effective forms of road pricing. The last reason for the poor quality of traffic congestion forecasting is related to the transport models used. All old traffic congestion estimates are based on versions of the so‐called National Model System (NMS) (V&W, [45]). The NMS consists of a series of state‐of‐the‐art disaggregate choice models. Geurs et al. ([8]) recently found that the NMS traffic congestion predictions in particular are highly uncertain and tend to be underestimated. The explanation is that NMS overestimated the attractiveness in travel time gains of alternative routes compared with congested trunk road routes. So, the 'model world' predicts that car drivers will shift to alternative routes when confronted with traffic jams more than actually happens in the 'real world'.

The road transport nitrogen oxide (NOx) emission scenario outcomes are fairly accurate (see Figure 4, bottom right). The scenario designers forecasted the large reductions rather well, with the exception of the very high 2010 estimate made by RIVM ([29], the middle scenario, the black dot [x‐value = 2010; y‐value = 320]). However, this is a genuine BAU scenario. In RIVM's ([29]) forecast it was decided to include only policy measures that had been agreed upon. In 1988, the only concrete plans were for not very strict EU air polluting emission standards for new heavy passenger cars. The RIVM middle scenario showed policy makers that if they implemented this emission policy, there would be no absolute reduction in NOx emission compared with the 1980 levels. In reality, the EU chose far stricter car emission standards for all cars, vans, lorries and road tractors (these standards were successively made stricter approximately every five years between 1990 and 2010).

The 2010 CO2 emission forecasts are also quite accurate, although there is a tendency to somewhat underestimate emission growth (Figure 4, bottom left). This tendency can be partly explained by the underestimation of car use in some scenarios (see section 'Transport Volume Outcomes'). However, a more important reason is that most scenario studies assumed energy efficiency improvements for cars, vans, lorries and road tractors. For example, in the Ministry of Transport ([21]) scenario study, it was assumed that technological improvements would make cars, vans and lorries roughly 5–10% more energy efficient by 2010 compared with 1986. In reality, there has been hardly any improvement in the average fuel use per kilometre driven since 1986 (Statistics Netherlands, [37]), despite the fact that car manufacturers have made more fuel‐efficient cars in this period. This is because Dutch consumers have bought increasingly bigger and heavier cars, so the overall average car fuel use has not decreased.

Discussion and Conclusions

The future developments of BAU transport are estimated with differing levels of accuracy, as sections 'Driving Forces', 'Transport Volume Outcomes' and 'Transport Impact Outcomes' show. In most of the cases studied, the important outcomes of the Dutch BAU transport scenarios, such as future car use and transport emissions, did not mislead policy makers. However, the traffic congestion forecasts all underestimated the problems that would be faced.

Visible progression has been made in the art of traffic and transport forecasting in the Netherlands. The scenarios from 1970 were sometimes flawed. The BAU scenarios made in the 1980s and 1990s provided, in most cases, more accurate predictions (Figures 2–4), again except for the traffic congestion predictions in particular. We think that there are two main causes for the improvement over time. First, in the 1980s and the 1990s, there was a better overview of the developments involved. During the 1970s, everything was less clear. There were many turbulent developments: cultural, economic, social and demographic as well as policy developments. Second, the quality of forecast preparation has improved. Since 1970, far more knowledge has become available, and traffic models have improved.

It can be considered a strong quality of the Dutch transport scenario practice that in most future studies (Table 1), more than one BAU scenario is given. By doing so, the scenario designers clearly showed the policy makers that a future estimate is highly uncertain. After all, Figures 2–4 show that scenario results for one transport indicator in a certain future year are indeed highly uncertain. Estimates per transport indicator for one future year may differ by several tens of percents from each other.

Many choices have to be made in order to design a BAU scenario, such as what model will be used, what assumptions will be made regarding transport determinants (e.g. population and economic growth), what existing policies will be taken into account and so forth. If researchers paint too rosy a picture of how things will be, this is called 'optimism bias' (Kahneman and Lovallo, [11]). Perhaps it is optimism bias that can be seen here in this review of forecasts of the past in the traffic congestion. All the researchers have been too optimistic about the speed with which new roads (or new lanes for an existing road) would be decided on and built. Also, due to choosing the same model (NMS), all the researchers were, in retrospect, too optimistic about car driver behaviour. In the NMS, car drivers chose alternative routes to avoid congested roads more easily than they would in reality. We think that the optimism bias in this case was unintended. The reason is that all the researchers were very cautious about including highly effective road pricing schemes in their BAU scenarios, despite the fact that Dutch policy makers had proposed implementing road pricing schemes several times in the 1990s. Reality has proven the old BAU scenario designers to be right in that respect.

The occurrence of 'Black Swans' has also played a role in explaining the inaccuracy of the Dutch transport scenarios studied. For example, in the Ministry of Transport ([18]) scenario study, the large cultural changes of the 1970s were not foreseen. However, in our view, the underestimations and overestimations which took place in the Dutch BAU transport scenarios made between 1970 and 2000 were caused mainly by inaccurate assumptions about economic, cultural and logistical changes than by the occurrence of 'Black Swans'. This observation that Black Swans have not played a major role in explaining the quality of Dutch transport scenarios in the past decades does not mean that their role will also be limited in the future. After all, Black Swans are by definition unexpected. Furthermore, even in the absence of Black Swans, Figures 2–4 show that transport forecasting, like all forecasting, is a highly uncertain business. The Dutch traffic congestion forecasting (Figure 4) is the most important example of this statement. So, this paper shows that it is of paramount importance to stress that BAU scenarios should play only a modest role in policy making. Policy making for the future always involves a gamble. Policy makers should be aware of this. But as Makridakis and Taleb ([15]) remark: "the big problem is, however, that the great majority of people, decision and policy makers alike, still believe not only that accurate forecasting is possible, but also that uncertainty can be reliably assessed" (p. 716).

This analysis indicates that for nonlinear phenomena such as traffic congestion predicting the future seems to be especially difficult. What we notice is that the Dutch practice of traffic congestion forecasting in the 1980s and 1990s (Figure 4) shows that using newer and better modelling does not necessarily lead to better results. The most worrying aspect seems to be that the poor performance of Dutch traffic congestion forecasting has gone unnoticed for so long. De Jong et al. ([6]) carried out uncertainty analyses on the NMS results, but they have limited their study to the sensitivity analysis of transport volume outcomes of the model. These kinds of studies are very useful but, in our view, should perhaps be carried out more often and expanded to also include an uncertainty analysis of transport impact outcomes, as much policy making is related to solving negative transport impacts.

Learning from Scenarios of the Past

We have identified six main lessons:

• 1. This analysis of Dutch scenario studies shows that BAU transport forecasting is an uncertain practice. We think it is highly important that scenario designers should communicate this uncertainty to the policy makers. We, therefore, recommend that BAU scenario designers give BAU forecast outcomes in different future scenarios. This practice avoids the pitfall that the policy maker thinks that the scenario designer really can predict the future. In our view, our paper indicates that the BAU scenarios chosen should provide a broad range of results for the future. By doing so, the scenario designers clearly communicate to the policy makers that they should base their decision on future uncertainty instead of on a false feeling of certainty (see Lesson 2). To be more precise, if policy makers ask for BAU scenario outcomes, all the BAU scenarios presented to the policy makers should be genuinely BAU. This means that only existing policies are taken into account in the scenarios (Lesson 4 gives an exception to this rule). The range in the scenarios should be related to different assumptions on economic growth, oil price, logistics developments, transport modelling assumptions and so forth.
• 2. Related to the previous point, policy makers should learn to include future uncertainty in decision‐making and not rely on the 'precise' outcomes of one BAU scenario. A large number of scientists have thought about policy strategies which take high future uncertainty into account. The key idea is not to develop precise plans for one precise future, but rather construct policies which are flexible, adaptable and devised not to be optimal for a best estimate future, but robust across a range of plausible futures (Walker et al., [48]; Makridakis et al., [14], Wright and Goodwin, [50]).
• 3. The Dutch prognoses from the 1970s show that a strong belief in social engineering and in people changing their behaviour because of general ideas about 'limits to growth' has little value for forecasting. On the other hand, simply extrapolating growth figures also runs this risk. The lessons here are perhaps stating the obvious: it is not easy to design a BAU scenario (or any kind of future scenario whatsoever), and there is no general recipe for designing a scenario. The Dutch practice shows that a keen eye on which determinants are important for the policy‐relevant outcomes (and varying those determinants to end up with a broad range of possible future states, see Lessons 1 and 2) as well as using good quality transport models (see Lessons 4, 5 and 6) are important aspects of BAU scenario designing.
• 4. The underestimation in the traffic congestion forecasts indicates that we should exercise caution when implementing policy in BAU scenarios, which is not yet fully developed. There are several possible lessons to learn here. First, scenario designers should be careful not to succumb to political pressure. Second, it may be an idea to construct pure BAU scenarios and scenarios including some planned policies (e.g. the highly debated ones). Groot and van Mourik ([9]) used this approach in a cost–benefit analysis (CBA) for a new Dutch national road investment programme (aimed at new investments compared with existing investment plans for the period 2014–20). They designed two pure BAU scenarios (differing in assumptions on demography and economy, low and high) as the base cases for future road transport demand and congestion levels. In 2007, there was still a great debate in the Netherlands as to whether road pricing was an existing or planned policy. To end all discussions, Groot and van Mourik ([9]) decided to design an extra 'semi' BAU scenario: in the pure BAU scenario with the lowest assumed population and economic growth, they included road pricing. Of all the possible future states distinguished in their study, this 'semi' BAU scenario would obviously result in the lowest future road transport demand and lowest congestion levels. Using this base case in their CBA, they found that for new road extensions (mainly adding extra lanes to existing infrastructure) with investment costs of 4.1 billion Euros (in 2005 prices), the benefits would outweigh the costs. They recommended that the Ministry of Transport (which asked for this analysis) included the 4.1 billion road investments in the national road investment plan 2014–20. It was argued that if these new roads provided societal benefits in this relatively low transport demand scenario, they would also be beneficial in all the other possible futures distinguished. The 4.1 billion extra road investments is a 'no regret' policy, as Groot and van Mourik ([9]) commented. At the same time, it is possible that the Dutch economy would grow faster compared with their 'low' base case and that road pricing decision‐making would again be postponed for several years. It was therefore recommended that legal procedures were started—which can take years in the Netherlands—for those road projects which provided societal benefits in the high‐growth scenario without road pricing. By doing so, a relatively fast policy response would be possible if the future turned out to be more like the 'high' scenario. Waiting for the final decision‐making on building the 'high BAU scenario' road lanes was recommended whilst monitoring the actual economic development (actually low growth since 2008 due to the worldwide financial crisis) and the road pricing decision‐making process (still undecided in July 2010). This approach is an example of using pure and semi BAU scenarios and also of flexible decision‐making which takes future uncertainty into account (see Lesson 2).
• 5. In the Dutch practice of BAU transport forecasting, many scenario studies are based on the same model, see sections 'Transport Volume Outcomes' and 'Transport Impact Outcomes'. It may be useful in such a situation to use other models alongside the dominant model. The strong dependence on one model can lead to certain developments being disregarded or to the same model flaws recurring as shown in the Dutch congestion forecasting history (Figure 4, top).
• 6. Carrying out ex‐post analyses like this on a regular basis is recommended. These evaluations may show the weaknesses in the practice of BAU transport forecasting, such as the poor performance of traffic congestion forecasting in the Netherlands. Improvements can then be implemented, such as improving traffic congestion modelling.

1 Bartholomew, K.2005. Integrating Land Use Issues into Transportation Planning: Scenario Planning, Summary Report, Salt Lake City: College of Architecture + Planning, University of Utah. 2 CE, RIVM, TNO Inro. 2000. Milieuwinst op het spoor? Synthese van onderzoeken naar milieu‐effecten van het goederenvervoer per spoor [Synthesis of Research towards Environmental Impacts of Rail Freight Transport], Delft: CE Delft. 3 CPB. 1997. Economie en fysieke leefomgeving [Economy and Environment], Den Haag: Centraal Planbureau. 4 De Beer, P.2001. Banengroei lost niet alle sociale problemen op [Not all Social Problems are Solved with Job Growth], The Hague: The Netherlands Institute for Social Research. 5 de Jong, M. and Annema, J. A.2010. De geschiedenis van de toekomst [The history of the future], The Hague: Kennisinstituut voor Mobiliteitsbeleid. 6 De Jong, G., Pieters, M., Miller, S., Daly, A., Plasmeijer, R., Graafland, I., Lierens, A., Baak, J., Walker, W. and Kroes, E.2005. Uncertainty in Traffic Forecasts: Literature Review and New Results for the Netherlands, Leiden: Rand Corporation. WR‐268‐AW 7 European Environment Agency (EEA). 2009. Looking Back on Looking Forward: A Review of Evaluative Scenario Literature, Copenhagen: EEA. EEA Technical Report No. 3/2009 8 Geurs, K. T., Annema, J. A. and van Mourik, H.2007. Analyse van de onzekerheden in de Verkeerskundige en wagenparkeffecten van de Eerste stap Anders Betalen voor Mobiliteit [Uncertainty Analysis of Road Charging Impacts on Traffic and Fleet of Cars], Bilthoven: Netherlands Environmental Assessment Agency/The Hague: The Netherlands Institute for Transport Policy Analysis. 9 Groot, W. and van Mourik, H.2007. Welvaartseconomische toets investeringspakketten hoofdwegen [Economic Welfare Analyses of Trunk Road Investment Proposals], The Hague: Kennisinstituut voor Mobiliteitsbeleid. Hupkes, G.1977. Gasgeven of afremmen: toekomstscenario's voor ons vervoerssysteem [Step on the Gas or Put on the Brakes: Future Scenarios for Our Transport System], Deventer: Kluwer. Kahneman, D. and Lovallo, D.1993. Timid choices and bold forecasts: A cognitive perspective on risk taking. Management Science, 39: 17–31. Keuzenkamp, S., Hillebrink, C., Portegijs, W. and Pouwels, B.2009. Deeltijd (g)een probleem. Mogelijkheden om de arbeidsduur van vrouwen met een deeltijdbaan te vergroten [Part‐Time no Problem or a Problem. Possibilities to Increase Working Hours of Women Who Work Part‐Time], The Hague: The Netherlands Institute for Social Research. KiM. 2009. Mobiliteitsbalans 2009 [Mobility Report 2009], The Hague: The Netherlands Institute for Transport Policy Analysis. Makridakis, S., Hogarth, R. M. and Gaba, A.2009. Forecasting and uncertainty in the economic and business world. International Journal of Forecasting, 25: 794–812. Makridakis, S. and Taleb, N.2009. Decision making and planning under low levels of predictability. International Journal of Forecasting, 25: 716–733. McKinsey. 1986. Afrekenen met files, samenvatting conclusies en aanbevelingen [Being Finished with Traffic Jams: Summary and Conclusions], Amsterdam: McKinsey & Company. Meadows, D. H., Meadows, D. L., Randers, J. and Behrens, W. W.III. 1972. The Limits to Growth, New York: Universe Books. Ministry of Transport. 1970. TP 2000: op weg naar 2000, een toekomstprojectie [TP2000: On the Road to the Year 2000, a Future Outlook], The Hague: Staatsdrukkerij. Ministry of Transport. 1986. SVV‐verkenning [Transport Policy Plan Studies], Rotterdam: Adviesdienst Verkeer en Vervoer. Ministry of Transport. 1993. SVV‐verkenning. Analyses en prognoses [Transport Policy Plan 1990 Studies. Analysis and Prognoses], Rotterdam: Adviesdienst Verkeer en Vervoer. Ministry of Transport. 1997. Personen‐ en goederenmobiliteit in 2010 en 2020 [Persons and Freight Mobility in 2010 and 2020], Rotterdam: Adviesdienst Verkeer en Vervoer. NEI. 1995. De prijs van een reis, enkele scenarioberekeningen voor 2010 [The Price of a Trip, Some Scenario Estimates for 2010], Rotterdam: Nederlands Economisch Instituut. Netherlands Environmental Assessment Agency (MNP), Netherlands Bureau for Economic Policy Analysis (CPB), National Institute for Spatial Research (RPB). 2006. Welfare, Prosperity and the Living Environment, Netherlands: MNP, CPB, RPB. Available at: http://www.welvaartenleefomgeving.nl/authors%5fUK.html (accessed January 2010) Peeters, P. M.1988. Schoon op weg, naar een trendbreuk in het personenverkeer [Clean on the Road, towards Major Changes in Personal Travel], Amsterdam: Milieudefensie. Peeters, P. M.1993. Goed op weg, naar een trendbreuk in het goederenvervoer [Good on the Road, towards Major Changes in Freight Transport], Amersfoort: Stichting Werkgroep '2duizend. Petersen, M. S., Enei, R., Hansen, C. O., Larrea, E., Obisco, O., Sessa, C., Timms, P. M. and Ulied, A.2009. Report on Transport Scenarios with a 20 and 40 year Horizon, Final ReportCopenhagen, , DenmarkFunded by DG TREN Ramaker, J., Voerman, G. and Zwart, R.1998. Illusies van Den Uyl? De spreiding van kennis, inkomen en macht [Den Uyl's Illusions? The Distribution of Knowledge, Income and Power], Amsterdam: Het Spinhuis. Remery, C., Van Doorne‐Huiskes, A. and Schippers, J.2002. Labour market flexibility in the Netherlands: Looking for winners and losers. Work, Employment and Society, 16(3): 477–492. RIVM. 1988. Zorgen voor morgen [Concerns for Tomorrow], Alphen aan den Rijn: Samsom HD Tjeenk Willink. RIVM. 1991. Nationale Milieuverkenning 2, 1990–2010 [National Environmental Outlook 2, 1990–2010], Alphen aan den Rijn: Samsom HD Tjeenk Willink. RIVM. 1993. Nationale Milieuverkenning 3, 1993–2015 [National Environmental Outlook 3, 1993–2015], Alphen aan den Rijn: Samsom HD Tjeenk Willink. RIVM. 1997. Nationale Milieuverkenning 4, 1997–2020 [National Environmental Outlook 4, 1997–2020], Alphen aan den Rijn: Samsom HD Tjeenk Willink. RIVM. 2000. Nationale Milieuverkenning 5, 2000–2030 [National Environmental Outlook 2, 2000–2030], Alphen aan den Rijn: Samsom BV. Schafer, A. and Victor, D. G.2000. The future mobility of the world population. Transportation Research Part A: Policy and Practice, 34(3): 171–205. Scientific Council for Government Policy [Wetenschappelijke Raad voor het Regeringsbeleid]. 1977. De komende vijfentwintig jaar: Een toekomstverkenning voor Nederland [The Next 25 Years. A Future Outlook for the Netherlands], The Hague: Staatsuitgeverij. SCP. 1998. Sociaal en Cultureel Rapport 1998 [Report on Social and Cultural Developments], The Hague: The Netherlands Institute for Social Research. Statistics Netherlands. 2009. Statline (electronic databank of Statistics Netherlands)Available at: http://statline.cbs.nl/statweb/ (accessed 2009) Taleb, N. N.2007. The Black Swan: The Impact of the Highly Improbable, New York: Random House. Van den Broecke. 1987. Invloed mate van verstedelijking op rijbewijsbezit en op autobezit [Impact of Urbanization on Driver License and Car Ownership], The Hague: Projectbureau Integrale Verkeers‐ en Vervoersstudies. van der Duin, P., Hazeu, C., Rademakers, P. and Schoonenboom, J.2006. The future revisited: Looking back at the next 25 years by the Netherlands Scientific Council for Government Policy (WRR). Futures, 38: 235–245. van Mourik, H.2008. Verkenning autoverkeer 2012 [Road Transport Outlook 2012], The Hague: The Netherlands Institute for Transport Policy Analysis. Van Steertegem, M., ed. 2009. Milieuverkenning 2030 [Environmental Outlook 2030, chapter 6 is devoted to mobility], Aalst: Milieurapport Vlaanderen, VMM. V&W. 1979. Structuurschema verkeer en vervoer [Strategic Transport Plan], The Hague: Ministry of Transport, Public Works and Water Management. V&W. 1990. Tweede structuurschema verkeer en vervoer [Second Strategic Transport Plan], The Hague: Ministry of Transport, Public Works and Water Management. V&W. 1992. The National Model System for Traffic and Transport: Outline, Rotterdam: Ministry of Transport, Public Works and Water Management, Transportation and Traffic Research. V&W. 2005. Nota Mobiliteit [Mobility Document], The Hague: Ministry of Transport, Public Works and Water Management. V&W. 2008. Mobiliteitsaanpak [Transport Plan], The Hague: Ministry of Transport, Public Works and Water Management. Walker, W. E., Rahman, S. A. and Cave, J.2001. Adaptive policies, policy analysis and policy‐making. European Journal of Operational Research, 128(12): 282–289. WBCSD. 2004. Mobility 2030: Meeting the challenges to sustainabilityAvailable at: http://www.wbcsd.org (accessed 25 October, 2010) Wright, G. and Goodwin, P.2009. Decision making and planning under low levels of predictability: Enhancing the scenario method. International Journal of Forecasting, 25: 813–825.

By Jan Anne Annema and Mig De Jong

Reported by Author; Author