MODAL CHOICE DECISION MAKING BY TRAVELLERS Embracing changing behaviour in a world of exploding information on options and ever changing prices : a forecasting perspective Luis Willumsen
KEY REQUIREMENTS The four pillars of good  forecasting  models Good future population synthesis System equilibrium Consistency of future behaviour Behavioural choice modelling Utility functions and choice models applied at different levels of aggregation The parameters in the utility functions remain the same Accurate allocation of  populations and activities in the future Appropriate feed-back through all relevant submodels to ensure consistent results Forecasting
BEHAVIOURAL MODELS Utility functions Modelling choices Homo Economicus Rational individuals  j  seeking to maximise their utility Choosing among different alternatives  q Modeller has incomplete knowledge  but  can estimate the most significant part of the “utility function” The systematic component  V jq  contains attributes like:  time, fare, income, sex  and an  ASC (comfort, convenience.....) The random component  e jq   absorbs all the unknown variable influences  and  the unexplained behaviour.
The number of options available keeps increasing Prices are becoming less crisp, fixed and predictable But some things still appear to be free Moreover, we change our mind...  Behavioural Economics and the 2008/9 crisis have established the inexistence of Homo Economicus What does this means for modelling? Have we reached the limits of forecastability? What should we do then? In a changing world... A changing context
Homo Economicus A  "rational" being that considers opportunities and seeks to optimise his/her  utility  by careful choices. W e cannot have access to these “utilities” but can infer the most important components by means of  utility functions  and  choice models Homer Sapiens A  partly rational but also emotional and collaborative being that tries to find happiness, respect from peers and a sense of purpose in what he/she does. B ehavioural Economics sheds some light on the predictatibility of his/her behaviour,  often inconsistent with our models
THE DIFFICULTIES OF HOMO ECONOMICUS Plethora of choices There is clear evidence that we cannot really consider more than a few alternatives at a time, perhaps 3-4 If more than 3 we use heuristics: habit, elimination by aspects, affect (emotion), resemblance, confirmation How do we incorporate these into our demand models? Known biases in choice Anchoring, Attentional Bias, Groupthink, Bias blind spot, Choice-supportive bias. Confirmation bias, Congruence bias, Contrast effect, Denomination effect, Distinction bias, Endowment effect, Expectation bias, Focusing effect, Framing effect, Hostile media effect, Hyperbolic discounting, Illusion of control, Impact bias, Information bias, Irrational escalation, Loss aversion, Mere exposure effect, Money illusion, Moral credential effect, Negativity bias, Neglect of probability, Normalcy bias, Omission bias, Outcome bias, Planning fallacy, Post-purchase rationalization, Pseudocertainty effect, Reactance, Restraint bias, Selective perception, Semmelweis reflex, Social comparison bias, Status quo bias, Unit bias, Wishful thinking, Zero-risk bias
THERE IS ALSO A PROBLEM WITH MONEY...
IS MONEY ALWAYS MONEY? Different kinds of money There is evidence that As we move away from cash price (money) becomes less well defined ...and willingness to pay is easier.. As we increase the time elapsed between use and payment a similar effect is found If prices change significantly over time and we pay electronically the effect is amplified
UNDERSTANDING MONEY The knowledge of prices  Is becoming more tenuous Extreme examples: information is available but it is impossible to use. Almost inevitably this happens in large and complex toll roads;  But may also happen in extensive Public Transport networks with complex fare structures. Singapore Before and After
SANTIAGO TOLL ROADS 5 + 1 concessions All with three level pricing: 6/12/18 US cents/km Interoperable tags ~ 1.2 million tags in 2006 Full toll collection started January 2005 Santiago ETC urban system
CONGESTION CHARGING IN SANTIAGO Objective: free-flow road SANTIAGO TOLL ROADS
EXAMPLE OF EXPECTED CHARGING SCHEDULE Simplified Price Structure  (changes in 30 mins intervals)
PLUS, A COUPLE OF COMPLEXITIES FROM BEHAVIOURAL ECONOMICS...
OVERVALUING WHAT WE HAVE Other considerations from Behavioural Economics Price elasticities are not symmetric: a 10% loss in utility is not compensated by a 10% gain The role of big changes The final state of a system depends on the  sequence of interventions System Equilibrium may be less useful than we believe The high price of ownership: cars vs public transport
OVERUSING WHAT SEEMS TO BE FREE Other considerations from Behavioural Economics The high cost of Zero Price Gratis  blind us to rational decision making ..at a great cost in congestion, pollution and quality of life Pricing for externalities is inevitable We can start with HOT lanes Therefore  understanding how we parse variable and fuzzy prices will become paramount
THE LIMITS OF MODELLING So.... Human nature limits the accuracy of our models There is scope for improving our models, recognising Homer Sapiens decision making and the fuzziness of money We will learn more about human behaviour and improve short-term forecasting ..but the contribution to accurate long-term forecasting will be limited Interpretation and judgement  will have to become more open and transparent ..and, we need to explore new sources of data and new tools of analysis to improve forecasting
ABUNDANCE OF DATA For example..... There are a lot of sensors and data out there GPS units Bluetooth units Mobile phones CCTV cameras ATC, ITS These create new opportunities for  data & representative experiments We should seek to obtain more from this abundance of data
POINTERS FOR FUTURE RESEARCH What about the future then? We need transport demand forecasting but our existing tools are less reliable than we pretend Future models should be based more on Homer Sapiens than on Homo Economicus Interpretation  and  judgement , professional responsibility, should be more open and transparent The use of  complementary models,  that look at the future from different perspectives ,  should help long-range forecasting Exploiting the over abundance of data  out there will lead to a different approach to policy advice, experimentation and decision making support
THANK YOU

Modelling World 2011

  • 1.
    MODAL CHOICE DECISIONMAKING BY TRAVELLERS Embracing changing behaviour in a world of exploding information on options and ever changing prices : a forecasting perspective Luis Willumsen
  • 2.
    KEY REQUIREMENTS Thefour pillars of good forecasting models Good future population synthesis System equilibrium Consistency of future behaviour Behavioural choice modelling Utility functions and choice models applied at different levels of aggregation The parameters in the utility functions remain the same Accurate allocation of populations and activities in the future Appropriate feed-back through all relevant submodels to ensure consistent results Forecasting
  • 3.
    BEHAVIOURAL MODELS Utilityfunctions Modelling choices Homo Economicus Rational individuals j seeking to maximise their utility Choosing among different alternatives q Modeller has incomplete knowledge but can estimate the most significant part of the “utility function” The systematic component V jq contains attributes like: time, fare, income, sex and an ASC (comfort, convenience.....) The random component e jq absorbs all the unknown variable influences and the unexplained behaviour.
  • 4.
    The number ofoptions available keeps increasing Prices are becoming less crisp, fixed and predictable But some things still appear to be free Moreover, we change our mind... Behavioural Economics and the 2008/9 crisis have established the inexistence of Homo Economicus What does this means for modelling? Have we reached the limits of forecastability? What should we do then? In a changing world... A changing context
  • 5.
    Homo Economicus A "rational" being that considers opportunities and seeks to optimise his/her utility by careful choices. W e cannot have access to these “utilities” but can infer the most important components by means of utility functions and choice models Homer Sapiens A partly rational but also emotional and collaborative being that tries to find happiness, respect from peers and a sense of purpose in what he/she does. B ehavioural Economics sheds some light on the predictatibility of his/her behaviour, often inconsistent with our models
  • 6.
    THE DIFFICULTIES OFHOMO ECONOMICUS Plethora of choices There is clear evidence that we cannot really consider more than a few alternatives at a time, perhaps 3-4 If more than 3 we use heuristics: habit, elimination by aspects, affect (emotion), resemblance, confirmation How do we incorporate these into our demand models? Known biases in choice Anchoring, Attentional Bias, Groupthink, Bias blind spot, Choice-supportive bias. Confirmation bias, Congruence bias, Contrast effect, Denomination effect, Distinction bias, Endowment effect, Expectation bias, Focusing effect, Framing effect, Hostile media effect, Hyperbolic discounting, Illusion of control, Impact bias, Information bias, Irrational escalation, Loss aversion, Mere exposure effect, Money illusion, Moral credential effect, Negativity bias, Neglect of probability, Normalcy bias, Omission bias, Outcome bias, Planning fallacy, Post-purchase rationalization, Pseudocertainty effect, Reactance, Restraint bias, Selective perception, Semmelweis reflex, Social comparison bias, Status quo bias, Unit bias, Wishful thinking, Zero-risk bias
  • 7.
    THERE IS ALSOA PROBLEM WITH MONEY...
  • 8.
    IS MONEY ALWAYSMONEY? Different kinds of money There is evidence that As we move away from cash price (money) becomes less well defined ...and willingness to pay is easier.. As we increase the time elapsed between use and payment a similar effect is found If prices change significantly over time and we pay electronically the effect is amplified
  • 9.
    UNDERSTANDING MONEY Theknowledge of prices Is becoming more tenuous Extreme examples: information is available but it is impossible to use. Almost inevitably this happens in large and complex toll roads; But may also happen in extensive Public Transport networks with complex fare structures. Singapore Before and After
  • 10.
    SANTIAGO TOLL ROADS5 + 1 concessions All with three level pricing: 6/12/18 US cents/km Interoperable tags ~ 1.2 million tags in 2006 Full toll collection started January 2005 Santiago ETC urban system
  • 11.
    CONGESTION CHARGING INSANTIAGO Objective: free-flow road SANTIAGO TOLL ROADS
  • 12.
    EXAMPLE OF EXPECTEDCHARGING SCHEDULE Simplified Price Structure (changes in 30 mins intervals)
  • 13.
    PLUS, A COUPLEOF COMPLEXITIES FROM BEHAVIOURAL ECONOMICS...
  • 14.
    OVERVALUING WHAT WEHAVE Other considerations from Behavioural Economics Price elasticities are not symmetric: a 10% loss in utility is not compensated by a 10% gain The role of big changes The final state of a system depends on the sequence of interventions System Equilibrium may be less useful than we believe The high price of ownership: cars vs public transport
  • 15.
    OVERUSING WHAT SEEMSTO BE FREE Other considerations from Behavioural Economics The high cost of Zero Price Gratis blind us to rational decision making ..at a great cost in congestion, pollution and quality of life Pricing for externalities is inevitable We can start with HOT lanes Therefore understanding how we parse variable and fuzzy prices will become paramount
  • 16.
    THE LIMITS OFMODELLING So.... Human nature limits the accuracy of our models There is scope for improving our models, recognising Homer Sapiens decision making and the fuzziness of money We will learn more about human behaviour and improve short-term forecasting ..but the contribution to accurate long-term forecasting will be limited Interpretation and judgement will have to become more open and transparent ..and, we need to explore new sources of data and new tools of analysis to improve forecasting
  • 17.
    ABUNDANCE OF DATAFor example..... There are a lot of sensors and data out there GPS units Bluetooth units Mobile phones CCTV cameras ATC, ITS These create new opportunities for data & representative experiments We should seek to obtain more from this abundance of data
  • 18.
    POINTERS FOR FUTURERESEARCH What about the future then? We need transport demand forecasting but our existing tools are less reliable than we pretend Future models should be based more on Homer Sapiens than on Homo Economicus Interpretation and judgement , professional responsibility, should be more open and transparent The use of complementary models, that look at the future from different perspectives , should help long-range forecasting Exploiting the over abundance of data out there will lead to a different approach to policy advice, experimentation and decision making support
  • 19.