Scribd
Upload
18 views

Random Utility

The document discusses random choice theory in economics. It uses the example of a person, Jenny, choosing inconsistently between French and Austrian wines on different occasions. Economists explain this using a random utility model, where utility has both a deterministic component based on product attributes and a random error term. This allows for inconsistent choices while still learning the person's underlying preferences. The model assumes independence of irrelevant alternatives, meaning removing options does not change the relative probabilities of choosing between remaining options. Logit models commonly used by researchers satisfy this assumption.

Uploaded by

kosarvardini21
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
18 views

Random Utility

The document discusses random choice theory in economics. It uses the example of a person, Jenny, choosing inconsistently between French and Austrian wines on different occasions. Economists explain this using a random utility model, where utility has both a deterministic component based on product attributes and a random error term. This allows for inconsistent choices while still learning the person's underlying preferences. The model assumes independence of irrelevant alternatives, meaning removing options does not change the relative probabilities of choosing between remaining options. Logit models commonly used by researchers satisfy this assumption.

Uploaded by

kosarvardini21
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 15

Random Choice / Random Utility

If we observe an individual long enough, she is bound to choose


inconsistently.
• May 1st. Jenny buys French wine at a price of 10 when Austrian wine
is available at price 6.
• May 10th. Jenny buys Austrian wine at price 7 when French wine is
available at price 10.
Random Choice / Random Utility
If we observe an individual long enough, she is bound to choose
inconsistently.
• May 1st. Jenny buys French wine at a price of 10 when Austrian wine
is available at price 6.
• May 10th. Jenny buys Austrian wine at price 7 when French wine is
available at price 10.

Assuming that Jenny prefers to spend as little as possible, she would be


violating WARP.
Can we learn something from her behavior?
“Though this be madness, yet there is method in ‘t”
Random Choice / Random Utility
Yes! economists consider a random version of this model.

• The individual is assumed to obtain utility

𝑢𝑢�
𝑘𝑘,𝑡𝑡 = 𝑢𝑢 𝑥𝑥𝑘𝑘 + 𝜖𝜖𝑘𝑘,𝑡𝑡

when selecting alternative 𝑘𝑘 in choice problem 𝑡𝑡.

• 𝑢𝑢 𝑥𝑥𝑘𝑘 is permanent (deterministic, the method) and it is the object that the
researcher is interested in.
• 𝜖𝜖𝑘𝑘,𝑡𝑡 is random and changes from decision problem to decision problem (the
madness).
Random Choice / Random Utility
In Jenny’s case we need the following to be true:

𝑢𝑢�
𝐹𝐹,1 = 𝑢𝑢 𝐹𝐹, 10 + 𝜖𝜖𝐹𝐹,1 > 𝑢𝑢 𝐴𝐴, 6 + 𝜖𝜖𝐴𝐴,1 = 𝑢𝑢�
𝐴𝐴,1

𝑢𝑢�
𝐹𝐹,2 = 𝑢𝑢 𝐹𝐹, 10 + 𝜖𝜖𝐹𝐹,2 < 𝑢𝑢 𝐴𝐴, 7 + 𝜖𝜖𝐴𝐴,2 = 𝑢𝑢�
𝐴𝐴,2

Which can be true if 𝜖𝜖𝐴𝐴,2 − 𝜖𝜖𝐴𝐴,1 > 𝜖𝜖𝐹𝐹,2 − 𝜖𝜖𝐹𝐹,1


Random Choice / Random Utility
It turns out that this random representation allows the researcher to
learn 𝑢𝑢 very easily.

Alternatives that are chosen more often must have a higher 𝑢𝑢 (they
cannot always be lucky!).

In our example, if I see Jenny purchasing Austrian wine 70% of the time
when the price difference is 4, but only 30% of the time when the price
difference is 3, I will infer that

𝑢𝑢(𝐴𝐴, 6) > 𝑢𝑢(𝐹𝐹, 10) > 𝑢𝑢(𝐴𝐴, 7)


In practice, researchers almost always assume a very strong form
consistency in this choice, called Independence of Irrelevant
Alternatives (IIA)

Suppose we observe the individual


choosing between these objects.

The size of the objects represents


the proportion of being chosen.
In practice, researchers almost always assume a very strong form
consistency in this choice, called Independence of Irrelevant
Alternatives (IIA)

Suppose now we eliminate the blue


circle from the choice set.
In practice, researchers almost always assume a very strong form
consistency in this choice, called Independence of Irrelevant
Alternatives (IIA)

Suppose now we eliminate the blue


circle from the choice set.

The other alternatives will be


chosen more often. How often?
In practice, researchers almost always assume a very strong form
consistency in this choice, called Independence of Irrelevant
Alternatives (IIA)

Suppose now we eliminate the blue


circle from the choice set.

The other alternatives will be


chosen more often. How often?

IIA says they will grow proportional


to their initial size.
In practice, researchers almost always assume a very strong form
consistency in this choice, called Independence of Irrelevant
Alternatives (IIA)

The intuition is that since the blue


circle is not available, it is
irrelevant.
The comparison of probabilities
between the other choices should
not change. Those that were more
popular should remain more
popular.
In practice, researchers almost always assume a very strong form
consistency in this choice, called Independence of Irrelevant
Alternatives (IIA)

Yet this is a very strong assumption.


It could be that the individual
chooses now the triangle more
often because it is blue, or the red
circle.
In practice, researchers almost always assume a very strong form
consistency in this choice, called Independence of Irrelevant
Alternatives (IIA)

This assumption is behind the so-


called Logit model.
As it turns out Logit-type models are consistent with behavior that
originates both in
1. Optimal choice given random preferences

2. Random choice given stable preferences


As it turns out Logit-type models are consistent with behavior that
originates both in
1. Optimal choice given random preferences
• There is a random component of utility but the individual fully optimizes.
• More options necessarily give more welfare.
• More volatility of preferences is also welfare-improving if there is enough
variety.

2. Random choice given stable preferences


As it turns out Logit-type models are consistent with behavior that
originates both in
1. Optimal choice given random preferences

1. Random choice given stable preferences


• The preference ranking is always the same, but the individual fails to choose
the best alternative.
• The chance that she chooses a given alternative depends on how good it is.
• More options may harm the consumer and so does the volatility.

You might also like