Bạn đang xem bản rút gọn của tài liệu. Xem và tải ngay bản đầy đủ của tài liệu tại đây (786.93 KB, 30 trang )
<span class='text_page_counter'>(1)</span><div class='page_container' data-page=1>
Lancaster’s attribute based utility theory
The Law of Comparative Judgment (Thurstone 1927)
Individuals react to stimuli
When making choices among alternatives, individuals choose
the one with highest level of stimulus
Stimulus comprises an objective level and a random error
Economists interpret stimulus as utility (Marschak
1960, Manski1977)
systematic component
the random component
Individuals choose the alternative with the highest level
The specification of random and systematic
utility
makes the model probabilistic
estimate the utility function
The model was made popular by McFadden
(1974)
multinomial logit (MNL) model
nested logit model
Transportation demand: choice among
transportation modes
Environmental valuation:
choice data generated from real market (actual
choice, or Revealed Preference data)
choice data from hypothetical market (Stated
Preference data)
Experimental data that involves choice among
options
Utility function
Alternative specific constants
Error terms
The probabilistic choice
Alternatives
Alternative 1 Alternative 2
… Alternative J
1 1 1
<i>Utility from alternative j include</i>
systematic component
the random component
j
<i>V</i>
j
j j j
Utility from alternative j is assumed to be a
function of attributes of alternative j
utility of from alternative j
level of attribute k of alternative j
marginal utility of attribute k (to be estimated)
Note: is constant specific to alternative j
j 0 <i>j</i> 1 <i>j</i>1 1 <i>j</i>2 ... <i>K</i> <i>jK</i>
<i>V</i>
j
<i>V</i>
<i>jk</i>
<i>x</i>
<i>k</i>
<i>0 j</i>
The constant term of each alternative is ASC
The model is unidentifiable if all the ASCs are
estimated. We have to fix one of them at 0
1 1
<i>V</i> <i>X</i>
2 2 2
<i>V</i> <i>ASC</i> <i>X</i>
J
<i>J</i> <i>J</i>
<i>V</i> <i>ASC</i> <i>X</i>
...
<b>ASC reflects </b>
<b>the preference </b>
<b>on each </b>
Gumbel distributed random
variable
location parameter
scale parameter
<i>e</i>
<i>e</i>
Properties of the Gumbel distribution
1. Mode ; Mean where
(Euler’s constant); and variance
2. If then
3. If and then
is logistically distributed
2
2
1
*
1 2
Properties of the Gumbel distribution
4. If ; ; …;
are independent, then
1 <i>G</i> 1,
: <sub>2</sub> : <i>G</i>
<i>J</i> <i>G</i> <i>J</i>
:
1
<i>J</i>
<i>J</i>
<i>j</i>
The utility function
The probability of choosing j
<i>j</i> <i>j</i> <i>j</i>
<i>U</i> <i>V</i>
Pr j is chosen among C
<i>j</i>
<i>p</i>
Pr U , , j,l
<i>j</i> <i>j</i> <i>l</i>
<i>p</i> <i>U</i> <i>l</i> <i>j</i> <i>C</i>
*
Pr U max Pr U
<i>j</i> <i>j</i> <i>l</i> <i>j</i>
<i>l C</i> <i>j</i>
<i>p</i> <i>U</i> <i>U</i>
<sub></sub> <sub></sub>
Pr V + Pr V
<i>j</i> <i>j</i> <i>j</i> <i>j</i> <i>j</i>
<i>p</i> <i>V</i> <i>V</i>
*
1
1 <i>j</i>
<i>j</i> <i><sub>V</sub></i> <i><sub>V</sub></i>
<i>p</i>
<i>e</i>
Properties of the Gumbel distribution result in
1
<i>j</i>
<i>l</i>
<i>V</i>
<i>j</i> <i><sub>J</sub></i>
<i>V</i>
<i>l</i>
Log-likelihood
the choice of individual i on alternative j (1 =
chosen)
Coefficients of the utility functions are estimated by
maximizing the log-likelihood function
1
<i>j</i>
<i>l</i>
<i>V</i>
<i>j</i> <i>J</i>
<i>V</i>
<i>l</i>
<i>e</i>
<i>p</i>
<i>e</i>
<i>L</i> <i>Y</i> <i>p</i>
<i>ij</i>
Method of estimation
Data: Revealed preference and Stated preference
Data organization
Assume location and scale
Likelihood function
Log-likelihood function
by maximizing the log-likelihood function
0
1
1
<i>j</i>
<i>l</i>
<i>V</i>
<i>j</i> <i>J</i>
<i>V</i>
<i>l</i>
<i>e</i>
<i>p</i>
<i>e</i>
<i>N</i> <i>J</i> <i><sub>Y</sub></i>
<i>ij</i>
<i>i</i> <i>j</i>
<i>L</i> <i>p</i>
<i>L</i> <i>Y</i> <i>p</i>
is actual observed choice,
= 1 if j is chosen,
= 0 otherwise
<i>ij</i>
Choices are obtained in a hypothetical situation
respondents are presented with a set of alternatives
each alternative are characterized by a set of
attributes’ levels
respondents are asked to choose among the
presented alternatives
Harper (2012) estimates WTP for the
conservation of endangered species (caribou)
Respondents were asked to choose between 2
alt
status quo: current management strategy
the proposed management strategy
<b>Resp</b> <b>Choice </b>
<b>set</b> <b>Alt</b> <b>Herd</b> <b>Cost</b> <b>Choice</b>
Pham and Tran (2005) used choice modelling to
analyze the demand for water service improvement
Each respondent were asked to make several
choices between:
the current situation (status quo)
the improved service plan
Each alternative is characterized by 3 attributes:
water quality (with 3 levels: low, medium, high)
water pressure (low; medium; high)
<b>Resp</b> <b>Choice </b>
<b>set</b> <b>Alt</b> <b>qualityWater</b> <b>pressureWater </b> <b>Cost</b> <b>Choice</b>
2 levels: water quality (2 discrete levels Low and
High)
create a dummy variable WQ
WQ = 1 if high quality; 0 otherwise
How to interpret the estimated coefficient of WQ?
3 levels (or more): water pressure (Low, Medium
and High
2 dummy variables PM and PH
PM = 1 if medium; 0 otherwise
PH = 1 if high; 0 otherwise
Similar for the case of more than 3 levels
Probability of choosing
Demand analysis
Predict the changes in
probability
quantity demanded
when an attribute changes.
1
<i>j</i>
<i>l</i>
<i>V</i>
<i>j</i> <i>J</i>
<i>V</i>
<i>e</i>
<i>p</i>
<i>e</i>
Estimate welfare changes resulted from a change
in attributes
is the attribute under consideration, p is the price
increases by 1 unit
p increases by 1 unit ($)
1 unit increase in is equivalent to ($)
increase in price
1
1
<i>x</i>
1
<i>x</i> <i>U</i>
<i>U</i>
1
<i>x</i>
Collect data
Input and organize data
Estimate the RUM using Stata
Calculate the probability of choosing a product
Illustration of how to calculate log-likelihood
value
A producer considers introducing a new product (chocolate
bar) to the market
The producer found that the following attributes are important
weight (gr): 50, 100, 200
type: milk or dark
ingredient: with or without nuts
price (1000 VND): 15, 30, 45
Target market
Questions:
How to decide the levels of attributes?
Design #: 1
You are requested to consider a chocolate bar with the characteristics presented below
Chocolate bar
Weight (gram) <b><sub>50 </sub></b>
Type (milk chocolate or dark chocolate) <b><sub>Dark </sub></b>
Ingredients (With nuts or not) <b><sub>No nuts </sub></b>
Price (thousand VND) <b><sub>15 </sub></b>
Would you buy the chocolate bar? □ Yes □ No
Please let us know:
Your gender □ Male □ Female