noc19-mg15 Lec 03- Utility Analysis

IIT Kanpur July 201
20 followers
Dec 3, 2023
Important properties of Utility Functions, Risk perception of Decision Makers, Risk Aversion, Marginal Utility, Absolute Risk
Welcome back my dear friends, a very good morning, good afternoon, good evening to all
my dear friends and students for this DADM2 course which is Data Analysis and Decision
Making II.
And we are in the third lecture as you can see from the slide and this total course DADM2
is for 12 weeks which is 30 hours and each week we will have 5 lectures each being for
half an hour and as you know that after each week we will have one assignment.
So total in total we will have 12 assignments plus there would be one final end term examination
and each week assignments would be based on that week which has just been covered.
So if you remember we are discussing about utility functions, we were discussing also
that depending on the outcomes.
Now I did mention different connotations of outcomes and how the utility functions would
change depending on the outcomes, they can be the number of outcomes which are favorable
to one decisions, they can be the utility function themselves or they can be just numbers
without mentioning anything about UW.
And we then consider two different utility functions, two different decisions and considering
different combinations of 2 into 2, 4 we saw that how the decisions can change depending
on the total value of the utility, we are always trying to find out the maximum of the
utility.
And I also did mention without solving that if the utilities are same we will consider
the variance to be used to rank the decisions and if both the utilities expected value and
the variances are given we will use a combination of them where we rank the ratio of if we take
the ratio of expected value to variance we will rank them from the highest to the lowest,
if we take the inverse of that that is the variance with respect to the expected value
we will rank them from the lowest to the highest.
So they can be different ways of trying to ranking depending on what your statistic is
based on which you are trying to take a decision.
Further continuing utility analysis, so in the problem which we have considered where
there was a value of minus for the linear function and if we remember we did mention
if the utility function is negative we will purposefully consider them to be 0 even though
that may not be true, but in practicality for problem solving for this course we will
take them 0 until unless mentioned in the problem.
So now the question is that with and then we also considered a problem where you had
a government bond of 10 lakhs and the utility was W to the power half which is a power utility
and other hand you have three different options with probabilities of 20 percent, 40 percent,
40 percent and the outcomes were given the government one was not 10 lakhs it was 6 lakhs
sorry my mistake and the outcomes for the other non-deterministic decisions were 10,
5 and 1 and then we found out the corresponding expected value and found out the that the
government one was given a higher expected value than the probabilistic one hence we
take the government one.
So, they can be different variants of the problems the probabilities can change the
values can change that the rupees lakhs wealth can change the utility function change they
can be different ways of trying to understand rather than using the probability as probably
mass function we can we can have probability density function also then you need to find
out the expected value corresponding to the fact that we integrate the values rather than
sum them up.
So, the question is that would the above problem give a different answer if we use an utility
function on the form W to the power half plus C now C is basically is a positive or a negative
quantity. So, if basically it is a positive or negative quantity and if you are only considering
the linear function only not comparing against each other. So, if we take the value of 0
to be the least value then obviously, for higher negative values of C all of them would
become 0s and obviously, they would be at the same rank, but if we consider the value
of C to be to be negative highly negative and then we consider that W to the power half
plus C and the expected value is also consider in the negative terms then obviously, we can
rank them. So, the highest negative ones would be rank the least and the lowest negative
ranks more towards 0 and positive numbers would be rank the highest.
And similarly if C is positive it will would not basically change the ranking system because
they would basically in the values would come out in the positive range. So, the ranking
relative ranking remains the same.
Consider a second example or a third example. So, in a span of 6 days the price of a security
fluctuates and a person makes his or her transaction only at the following prices and
we assume the utility function to be logarithmic or utility function which is log of the price.
And in the table which is shown we have the days which is given as 1, 2, 3, 4, 5, 6 the
first column the prices are given which is in the second column. So, there are different
instances of the prices 1000 rupees, 975, 950, 1050, 925, 1025. So, these are I am considering
a random values and then in third column we find out the utility of those prices. So,
we find out the log of 1000, log of 975 so on and so forth and the utilities are given.
Now the outcomes are given. So, the outcomes here basically means the number which is N
W which means that if we take a snapshot these days now let us consider the days concept
not there if we consider the price of 35 it technically means that the number of outcomes
which were favorable to the price of 1000 out of some instances of the of the values
which I have noted they are 35 in number.
And if I consider the value of 925 rupees or dollars or yens whatever the price is then
the number of instances which are favorable for that comes out to be 5. So, if I find
out the numbers given in the second last column then the corresponding probabilities which
will find out using that outcomes would be N W divided by summation of N W that so that
it will be 35 divided by 100 because 100 is the sum of that set of values which is there
in the second last column and when we add them up it as it is 100.
So, 35 divided by 100 comes out to be 0.35 correspondingly the other probabilities are
0.15, 0.05 and 0.15. So, if we add them it definitely comes out to be 1. Now if we basically
multiply so what we are going to do is we will multiply the you want we want to find
out the expected value expected value would be the sum of utility into N W by summation
of N W. So, this is the equation we need to find out. So, I use different colors let me
use the green one. So, the utilities are given here the numbers N W's are given here
I am just marking the columns and the value of summation N W's once you find it out
it is 100. So, the probabilities which you want to have is given here. So, we can find
it out and do the calculations accordingly.
Now in case if utility function is given as function p to the power one-fourth or p to
the power 0.25 then the expected utility comes out to be 33.63 percent. So, 33.63 in value
terms. So, you can basically make a decision accordingly depending on different values
are there find out the utilities multiply them by the probabilities and find out the
expected value and rank them accordingly.
Now we will consider different properties or important properties of utilities. So,
the general properties of utility functions we will go through two main one and come repeatedly
come back to these in order to solve our problems. So, one is the concept of non-satiation. So,
the first restriction placed on the utility function is that it is consistent with the
more being preferred to less that means, more I give you more you want that means, more
it is better more the merrier. So, obviously, it would mean that in practical sense it may
not be true because say for example, in many of the consumptions problem after a certain
limit your marginal utility will start decreasing. So, but still we will consider that point
not to be true for the problems when you are trying to solve the optimization problems.
So, this means that between any two certain investments or amount of value being w value
means the wealth. So, if you take it the u w would always be less than u w plus 1 that
means, u w plus the utility based on w plus delta w would always be greater than u w which
means the first derivative the utility function with respect to w would always be positive.
Now, the question will come later on that this positive value which is there whether
in the case when we take the second derivative whether is increasing at an increasing rate
increasing as a constant rate or increasing and decreasing rate that would basically differentiate
and give us the concept of the second property which we are going to come now.
The next property is basically considered the concept of risk something due to the risk.
So, if we consider the investors or the decision makers perception of absolute risk then we
will have the concept of the property given categorized under three categories one is
the concept of risk aversion that means, I am want to avoid risk one is risk neutrality
that means, I am indifferent to risk and the third one is basically risk seeking that means,
I want to take risk. So, in this case what the point which I mentioned just few minutes
back or few seconds back that the utility function is always increasing with respect
to w and there is a property of non-satiation which means the first derivative of u w with
respect to w is always positive. Now, the second point which is there on the
slide it will mean which I mentioned that the first derivative can increase at an increasing
rate can increase at a constant rate and increase at a decreasing rate and that will give us
the properties I will come to that into details within few minutes that will give us the properties
of risk avoidance or risk aversion, risk neutrality or risk indifferent and risk seeking
property of any decision maker.
So, let us consider a very simple example. So, consider in on one table there is given
and this values which is given which you see 2 and 1 they would not make a much sense but
let us consider those values as given as values only. So, consider you if you invest an amount
2 or a unit 2 and when you basically toss the coin it is a actual coin unbiased coin
probabilities are half and half if a head comes that investment which you make would
basically yield you the value of say for example 2 and in the other case whatever investment
which you have made consider that as for the time being as 1. So, 1 you invest you get
2 and another case 1 you invest if the tail comes you do not get any money.
So, if I want to find out that and this is the unbiased coin remember. So, if I tell
you that if I keep doing that experiment tossing the coin finding out that in half
probability I get 2 half I get nothing. So, in the long run you will say the expected
value is 1 why because 2 into half that means 2 is basically the value which I am getting
remember this 2 is the value I am giving and I am not talking about utility no it is a
just a value and you consider the value as it is and that is the value which accrues
to you.
So, the probability would half would be multiplied by 2 plus half another half would be multiplied
by 0 the expected value is 1. So, leave aside this experiment come to the other experiment
on table 2 there is a Sholay coin. So, all of you must have seen the film Sholay and
those outside India may not if they have not seen that you should definitely see the film
of Sholay.
So, in Sholay coin the coin is biased in the sense both are heads consider. So, if
they are head whichever way I toss the probability of head coming is 1 that means when I toss
the probability is 1 and the investment which I get is also 1. So, if I tell you to find
out the expected value you will immediately multiply the probability of 1 into the outcome
which is 1 and the expected value is 1. So, if you consider the scenario on table 1 and
scenario on table 2 both give you the same expected value.
Now where the differentiation starts now if I ask these values to see if I ask all of
you who are taking this class that which of the decision would you take on table 1 where
is an unbiased coin or table 2 where is the Sholay coin maximum of the people or may
be all of them would definitely choose the first one which is in table 1. Because you
are even if the expected values are same in the long run you are always thinking that
your main focus is on probability half where you will win 2 units.
Now let me change the scenario of the game the coin remains as unbiased in table 1 and
the Sholay coin remains as the Sholay coin and the biased coin with probability 1 on
table 2. But the values which were 2 here which you see now increases to say for example
200 and the value remains 0 while the value 1 which is there on the other table is 1 becomes
now 100. So, again I ask you the question so again I tell you that if I want to find
out the expected value for these 2 decisions on table 1 and table 2 you can find out that
200 into half plus 0 into half gives me 100 as a unit and then 100 into 1 on the table
2 gives us 100. So, the expected values are same then again I ask you the question.
Now consider that out of all of you who are taking the class some may be a little bit
doubtful of taking decision 1 which is on table 1 because now you think that your focus
of attention is slowly shifting to the case that what if a tail comes you lose everything
that means all this 200 is gone. Let me again increase the value of 200 to say for example
20000 I am just increasing at very huge jump. So, 2 becomes 20000 0 remain 0 1 becomes 10000
then the expected values would be same 10000 10000 but now again if I ask you the question
which decision will take with table 1 and table 2 slowly you will see that as I keep
increasing the value at some point of time all people would now be slowly be indifferent
that they do not know which one decision to take and as this value 2000 20000 becomes
2 lakhs continuously 10000 becomes 1 lakh and is 2 and 1 these values basically are
multiplied by higher and higher values you will see that all of you would basically be
tempted to take a decision which is the certainty event which is on the table 2.
So this gives us their concept that at some point of time any human decision we are willing
to take can basically be of any 3 categories depending on the wealth, depending on my experience,
depending what my risk attitude is, depending on my age all these things would basically
dictate that how I take the decision. So, let me come back to what is mentioned in slide.
So, price for investing is one which I already mentioned and it is a fair gamble fair in the
sense because the probabilities are half and half in the sense that is value is exactly equal to
the decision of not investing which is there on the right hand side and table 2 and we will use
this concept of fair gamble later on. Now the equations which are given are just simple
equation and notation of whatever I have done. So, thus if I consider the investments or decisions
as I 1 and I 2 so this means my decisions are like this, this is I 1 with probability P I 1,
this is I 2 with probability I 2. So, the utility and on the other hand obviously you
will basically have the deterministic event Di. So, the probabilities here it would be half,
probability here it will be half, technically it will be half in a fair gamble.
But now I will consider the probabilities as P I 1 and P I 2. So, now let me go to the
highlighter. So, this part so obviously probability P I 1 comes here and then I 1 gets replaced by
the utility because I 1 is the value. So, once I basically assume the utility function to be
some functional form so that utility becomes U I 1 as noted down. Now if I go to I 2 this value
is basically I 2 and the probabilities are P I 2. So, you see P I 2 is here and the corresponding
utility is U I 1 U I 2. Finally consider the deterministic one D U D 1 that means utility
of the deterministic event and now basically the probability being 1 it is 2. So, that means you
have these two decisions. Now let us consider and I will highlight it with the red color the
less than sign, the equal to sign and the greater than sign. So, if you see the equations on the
left hand side are the same, equations on the right hand are the same for the three bullet
points only the inequality, equality sign is changing. If less than sign which means that
I am more inclined to take the deterministic event because I want to avoid risk. So,
you can understand from the situation. If both are equal I mean different and if the left hand
side is more as in the last bullet point which means I am willing to take the risk even if the
expected value for all the three cases may be the same is exactly the problem which I have discussed.
So, another characteristic by which to classify risk aversion proper person and this neutral
person and this seeker person we will consider this that is the second derivative would have
a sign in one case it will be less than 0 in another case it will be equal to 0 and third
case it will be greater than 0 the second derivative which means if the second derivative
is less than 0 which means that utility function is increasing because obviously the first derivative
is 0 is increasing but the increasing is happening as a decreasing rate that means I am risk averse
I want to run away from risk. If it is basically the second derivative is 0 which means the utility
is increasing that means u w by w that is d u w by d w that is the first derivative is positive
and it is increasing as a constant rate that means an indifferent and in the third case if
it so and this means I am a risk neutral person and in the third case if the second derivative
is greater than 0 that means the first derivative obviously 0 if it is and the second derivative
is increasing it means that the utility function is increasing at an increasing rate that means
I am a risk seeker. So, these are the graphs. So, if you see the green one basically has
a utility function which is increasing but increasing and decreasing rate the blue one
is basically utility function is increasing but increasing at a constant rate and the
red one is basically a utility function which is increasing but increasing and increasing
rate. So, these are these are very obvious from these graphs.
Now let us come to the concept of marginal utility functions. So, marginal utility basically
would be corresponding in the first derivative. Marginal utility functions looks like a concave
curve this is a risk aversion person that means I am avoiding risk. So, obviously the
second derivative would be negative the and I am risk averse. Marginal utility second
bullet point says the marginal utility functions looks neither concave nor convex which is
a straight line is a risk neutral person and in the last case when the marginal utility
function looks like a convex curve it is if that means it is increasing that means the
second derivative is greater than 0 which I just showed in the last graph. So, it is
a risk seeking person because the second derivative u double prime is positive.
So, which means that the marginal rate is increasing and decreasing rate for the risk
averse person. The second bullet point is marginal utility rate is increasing at a
constant rate which is a risk neutral person and the third bullet point is the basically
the marginal utility rate is increasing at an increasing rate which is the risk seeking
person. So, increasing decreasing being risk averse, increasing constant being risk neutral
and increasing increasing being risk seeker. The first part obviously would be increasing
but as per the non cessation point it will always be positive.
So, here are the graphs which I have tried to draw it as neatly as possible for you to
make understand. So, on the y axis you are measuring the utility functions on the x axis
you are measuring the wealth some amount of wealth and the green curve which you have
if you see the d y d x of that or d u w d w it basically slowly tan of that angle slowly
starts decreasing. So, hence it is a risk aversion. In a risk neutral person the d y
d x or d u w d w is straight line. So, tan of that angle does not change. So, it is basically
risk neutral person and again I am measuring u w along the y axis and w values of the wealth
values on the x axis. Similarly, if I take the utility analysis
and the marginal rate for the curve which was the red one and if I plot again y axis
being u w x axis being w and if I find out the d y d x which is d u w d w is increasing
that means, the tan of the angle is increasing hence it is basically a risk seeker person.
So, few other important concepts which we are considering the utility analysis. So,
consider the risk aversion person would basically as per the definition would be he would reject
a fair gamble because his implication would be the second derivative is negative because
that is why he is rejecting a fair gamble and going for the certainty event.
A risk neutral person would be indifferent between the fair gamble and indifferent between
certainty event. So, because in that case his or her second derivative would basically
be 0 and a risk seeking person would be the person who would select the gamble because
technically the expected value for him or her remember that that is very important.
So, the decision which is being taken by the human being him or her is based on his or
her perception of risk only. So, coming to the third point which is mentioned I am again
rereading it. So, select the person who is risk seeking would select the fair gamble
because the second derivative is basically positive.
Now we will consider some of the properties of utility analysis. So, absolute risk aversion
is the property which would be important for our discussion. So, absolute risk aversion
property of utility function where we consider the absolute functions is given by this property
where we consider the negative of the ratio of the second derivative divided by the first
derivative. So, mathematically we are not going to deal with that because there are
mathematical clues for that but that is not necessary. Why they are necessary will come
to in the practical implication of a w later on a w and a prime and similarly we will consider
another concept also which is the relative risk aversion. So, this is the absolute risk
aversion. So, if it is basically minus value of u double prime by u prime. So, u prime
is always positive and there is a minus sign there. So, u double prime being negative being
0 and being positive with a negative sign would basically dictate that how the value
of a is changing that means a changing means a prime has to be found out. So, that would
be important for us to find out the properties of the human being or decision maker who is
who is taking. So, the if you if you notice that that when you take the derivative of
a prime a which when you find out the a prime there the actual the information will be coming
out for the utility function then you will see that in the later problems.
So, I will just briefly discuss how we generally do that. We will assume basically an investor
this is just for the concept is nothing to do with the problem solving this just a overall
feel that how you can basically approach the problem. We will assume an investor has wealth
of amount w whatever the w is and a security. Security is basically a financial asset with
an outcome which is represent by Z which is basically random because security would have
basically random returns. And we will consider Z to be a fair gamble whatever the amounts
are not there fair gamble will consider the properties of such that there is an equivalent
deterministic event which can replace that. And we will consider the expected value of
that fair gamble is 0 and the variance is given by some sigma square suffix Z and the
utility is basically given by Uw. So, we will basically choose a wealth such that we can
write this as two decisions in one decisions there is the wealth Wc and another decision
is the wealth W plus the security. And based on that we will try to balance it and basically
try to expand using Taylor series expansion and then trying to find out the Taylor series
from the Taylor series expansion we will try to find out the concept of absolute discover
which I just mentioned the last slide and then similarly find out the second derivative.
We will come into the actual implications again I am mentioning as we solve the problems
later on. So, with this I will end this third class and have a nice day and thank you very
much for your attention.
