Summary: Part Molin

Read the summary and the most important questions on Part Molin

• Lecture 1

  This is a preview. There are 12 more flashcards available for chapter 29/01/2018
  Show more cards here

• What is the main difference between the stated and the revealed data collection paradigm?

  Revealed preference -> real-market paradigms
   Stated preference  -> hypothetical alternatives

• What are:AttributesAttribute levelsChoice alternativesChoice set

  • attributes: characteristics of choice alternatives: cost, time, distance, comfort
    
  • attribute levels: particular values of attributes :€10, 15 minutes, 2 km., second class 
    
  • choice alternative: choice option with scores (levels) on attributes
    
  • choice set:  the group of choice alternatives that can be chosen in a particular moment in time    
    

• What is the strongest point of the Revealed Preference collection paradigm?

  It focuses on what people actually did: this results in high validity of models.
  
  (it is suitable for forecasting choice if: choice alternatives and market conditions do not change much)

• What are the limitations of Revealed preference?

  • No new alternatives: especially if the new alternatives have new attributes that do not yet exist; have attribute values outside current value range
    

  
  

  • Insufficient variation: for example mode choice in extreme weather.
    

  
  

  • Unknown choice set: what was the choice set?; there is only information about the chosen alternative. 
    

  
  

  • Multicollinearity problems: High multicollinearity among attributes: e.g. price & distance & time of railway ticket are highly correlated. this makes it impossible to disentangle effects; 1. invalid parameters 2. unreliable parameters.
    

  
  

  • Many respondents: many respondents required because you have only one choice per person observed.
    

• In what way does the Stated Preference approach provide a solution for the shortcomings of Revealed Preference approach?

  Stated preference: More choices per respondent & Control of correlations

• Why is multicollinearity a problem in Choice Modelling?

  Not valid = biased
  

  • Parameters may even have wrong sign: i.e. positive price parameter

  Low reliability = high standard errors
  

  • Low t-values -> insignificant parameters
    

• What is the solution for multicollinearity?

  • Construct own choice alternatives with zero (or low) correlations between attributes.
    
  • This results in Orthogonal; zero correlation between attributes -> results in low standard errors (reliable parameters) 
    

• What solutions do Stated Choice provide?

  1. alternatives are hypothetical, so you can include any new alternative (or new attributes & values beyond current levels)
  
  2. researcher constructs any variation necessary and observe choices that have low level of occurrence in real life
  
  3. researcher constructs the choice sets, so these are always fully known
  
  4. experimental design controls correlations, so multicollinearity may be avoided
  
  5. respondents make multiple choices, even small samples may result in reliable models        

• What is the most important drawback of SC experiments?

  main drawback: hypothetical bias: would people really choose the option in real life?

• What are the main challenges when constructing SC experiments?

  1. Create sufficient variation in choice situations
     
  2. Estimated parameters should be reliable -> small standard errors
     
  3. Choice task does not exhaust respondents
     
  4. Choice situations should resemble real world choices -> this will increase validity.