noc18-mg22-Lec 08-Optimization Theory and Techniques Part - 1

IIT Kanpur July 201
20 followers
Dec 3, 2023
[Music]
[Music]
welcome back to the lecture series on
microeconomics today
we are going to have a discussion on
optimization theory and some of the
associated techniques whatever we have
discussed so far it is quite evident
that all economic agents are optimizers
now we are going to move on to the case
of two variable optimization of course
this can be generalized to in dimension
you know in number of variables but let
us not get into the complications this
is not a you know advanced level
microeconomic Theory course for this
course if we know up to this that would
be good enough so in the case of two
variable optimization let us start with
this maximization problem a function
defined over two independent variables X
1 and X 2 these are the decision
variables so my maximization will be
with respect to X 1 and X 2 so what will
be the first-order condition in this
case first-order condition will be the
partial derivative of the objective
function with respect to the decision
variables in we need to set them to 0
now the second order condition is a bit
complicated compared to the first case
here we have to introduce a concept
called Hessian matrix the second-order
condition is represented through a
Hessian matrix the Hessian matrix is
defined as H F 1 1 F 1 2 F 2 1 and if -
so these are all partial derivatives
right so in general one can write these
F IJ is this partial derivative now
again there can be two types of problems
one maximization and one minimizations
we have started with the maximization
problem so let us look at the second
order condition for the maximization
case for maximization problem the
Hessian matrix must be
negative semidefinite now what do I mean
by this new term that is introduced here
negative semidefinite those who are
doing college-level matrix algebra may
be aware of this concept those who are
not for them let me define this very
quickly so let us assume a to be n cross
n symmetric matrix
and a k to be the K cross K sub matrix
so this is basically the upper left-hand
K cross K sub matrix of original matrix
a then let us also define determinant a
K the definition of determent
should be known to you then a is
negative semi definite
if
is positive for K taking values 1 2 up
to n minus 1 and determinant a should be
0 now if we are dealing with a
minimization problem then what will be
the shape of the second-order condition
again we have to construct the Hessian
of course and then this is the result
that I am going to state H matrix must
be a positive semi definite one what do
I mean by positive semi definite so let
us stick to that same definition of a
and then a is positive
semi-definite if determinant a K is
positive for k equal to 1 to 2 2 up to n
minus 1 and determinant a should be
equal to 0 now in this case let us look
at comparative static analysis
so now we are going to discuss a two
variable maximization problem so you
will start with the first order
conditions and then you know we are
going to see how from there we can
conduct our comparative statics analysis
so we will start from the first order
condition directly I said that I can
assume that the decision variables are
basically the functions of the
parameters so I am writing explicitly as
I did earlier also so this is say
equation number one then I have del F so
I am differentiating the same function
this time with respect to my decision
second decision variable which is x2 and
I am setting that equal to zero
and these one two equations they will
give me the first-order conditions now
for a comparative analysis differentiate
one with respect to the Alpha the
parameter and that will give you f 1 1
del X 1 del alpha plus F 1 2 del X 2 del
alpha plus if 1 alpha and that would be
equal to 0 now differentiate 2 with
respect to the parameter alpha which is
changing and now we will get if 2 1 del
X 1 del alpha
and you need to set that equal to zero
as well so you get two equations and you
can name them number three number four
what are the no variables in this
equations we are actually interested in
the comparative static components which
are namely del X 1 del alpha and del X 2
del alpha so these are the variables of
interest we have been discussing the
role of optimization techniques in
economic analysis now we are going to
continue with what we have learned so
far we are going to see an example of
optimization exercise in microeconomics
we are going to go back to the
discussion on the Marschallin consumer
analysis we are we have already seen how
in a Marshall Yin world consumer
optimizes now we are going to start from
there and we are going to extend that to
a two good model
in Marceline consumer analysis we have
made couple of assumptions I suggest you
go back to the previous lectures and you
know have a look at them so you know I
am NOT repeating these as options so
what we have done there we have just
worked with one consumption good namely
x1 and then we said that the consumer
derives utility which is a function of
the x1 consumption only and there is a
concept called Nate utility say it's
denoted by V and that is basically U of
X 1 minus lambda P 1 X 1 so if you
remember the discussion we had earlier
so in the one variable model this is the
gain of the consumer and these in terms
of the utility and this is the loss of
the consumer in terms of utility as well
so a consumer maximizes net utility U of
X 1 minus lambda P 1 X 1 we have looked
at the first order condition and all and
of course the decision variable has been
x1 there now we don't want to get into
the first order condition let us look at
the graph so that we can get some
insight into it
so I am plotting you along the vertical
axis and quantity or units consumed of
commodity wall along the horizontal axis
now utility is a concave to origin
function because of the law of
diminishing marginal utility and let's
see how the consumer actually optimizes
in this given setup so as the consumer
consumes zero unit then of course there
is no loss but as the consumer started
consuming positive units the positive
units will be multiplied by the price of
the unit and the marginal utility of
money and of course you know this lambda
1 lambda P 1 that you see here will be a
constant so we are talking about
basically a positively sloping upward
sloping straight line and let us draw
such a straight line and this can be
lambda P 1 X 1 so you can see the area
that the consumer that that the area
which gives the consumers native T DT is
given by this dashed area and the net
utility is maximized when you have a
straight line drawn as parallel to this
lambda P 1 X 1 straight line and the
point of tangency of that line to the
utility function so that tangency point
at the tangency point the gap between
this straight line lambda P 1 X 1 which
is the loss function and U of X 1 which
is the gain function is highest and
that's why at this point the utility is
maximized and we can say that we have
obtained the optimized value of x 1 star
so this is where we have stopped no not
really
because we have then talked about the
first-order condition also and then we
know we have derived the first-order
condition to be marginal utility of
commodity 1 divided by the price of
commodity 1 and that was equal to the
marginal utility of money lambda now up
to here we have started now let us think
about an extension of the model
to do that we have to have some more
assumptions number one we assume
different utilities
are independent by that we mean that
when the consumer derives utility from
consumption of x1 it does not depend
upon the quantity of consumption of
commodity 2 we also assume the utility
function is additively separable what do
I mean by that I am going to show you
right now so now we are going to start
with the new utility function you this
has two components U of X 1 plus U of X
2 so you see the utility consumed from
consumption good 1 and utility consumed
from consumption good two they are
independent now if we start with that
utility function how would the consumers
maximization problem would look like so
we can talk about an eight utility
function V
remember we have to extract the loss
also so here comes the loss this is the
expenditure on community 1 this is the
expenditure on commodity 2 and you
multiply that with the marginal utility
of money so this gives you Cal loss and
of course the decision variables are X 1
and X 2 in this case so in this case you
can very well see it is very difficult
to draw a graph as you know we have two
different variables consumption goods
here so we will you know try to solve
this optimization exercise using
calculus and to start with we have to
write down the first order conditions
for maximization so we have to partially
differentiate with respect to the good
one that will give us one first-order
condition and we know at optimal this is
going to be equal to zero we have to
differentiate with respect to the
community to and at optimum this is
going to hold so these are basically
mine necessary first-order conditions
now what is the second-order condition
for second-order condition you remember
that we have to focus on Hessian matrix
and we define Hessian matrix age in this
case s
now let us try to interpret something
more out of these two first order
conditions because we have two goods now
note we can write from these phosphoryl
conditions the following stuff
and this is known as law of equal
marginal utility
so if you have n number of commodities
of course you will we have we have
you'll be having more components now how
to interpret this law of equal marginal
utility so if you have some amount of
money you are thinking to allocate this
amount of money between two goods x1 and
x2 how to split the money into
consumption of X 1 and X 2 such that
your utility is highest possible the
rule is the following so if you are
spending say 1 rupee on x1 it should
give you the same level of utility if
you decide to spend the amount of same
amount of money on good 2 so that is
basically the law of equal marginal
utility now let us look at the following
comparative static analysis in this case
in this model the parameters are p1 and
p2 so these are the market prices of
this consumption goods 1 & 2
respectively so of course as p1 and p2
changes the demand for x1 x2 shall
change and this is what we are going to
capture through the comparative static
analysis let us talk about this exercise
how x1 star changes when p1 changes
so in symbolic sense we are looking for
remember this component or this
derivative can you identify this
derivative with what we have already
studied this is the slope of what we
call a Marshall Yin demand function
so now you can realize that one of the
beautiful example of usefulness of
computer static analysis is to find out
the slope of the demand function now let
us start from the first-order condition
itself so we have two equations there
and then
the sign of this thing which is
important to us can be given by 1 0-0
how we can get this done you know you
have two different equations there and
you have to differentiate these two
equations with respect to the parameter
which is changing here it is p1 and then
you know you will get a set of equations
and then from that set of equations you
have to apply the Crammer stool to get
you know the expression of you know del
X 1 del p1 and that expression the sine
of that expression actually is the sine
of this particular determinant and we
can say that this is going to be less
than or equal to 0 how come because of
an assumption that we made regarding
this component so we have assumed the
utility function to be concave and that
is due to the law of diminishing
marginal utility hence as the second
order derivative is negative the sine is
definitely less than or equal to 0 now
let us look at the second order
condition of this maximization exercise
so we have to write down the Hessian and
that is going to be u double prime 1 X
star 1 0 0
so for the maximization age must be a
negative semidefinite matrix right and
that means you double prime 1 X 1 star
and you double prime 2 X 2 star shall
both be
non-positive it implies less than equal
to zero so you can see the assumptions
that we make in the case of Marshall Yin
utility analysis or Marshall and even
analysis how that is playing a very
critical role to establish that we have
a downward sloping demand function and
we have the maximized utility achieved
for a consumer now let us go back to the
first-order condition say let us call
this to be equation one this to be
equation two now this u prime component
here that is the marginal utility of
commodity one and this component here is
the margin utility of commodity two now
note we got u prime one say I denote
this to be the I had good know as the
generalized case it can take value 1 2
we can have greater than and we can also
have the case of less than let me start
with the inequality so this is the first
expression for first group good and then
we have this expression for good - if
equality is not met then what will
happen first case would be
the case of this so you prime 1 divided
by P 1 is greater than u prime 2 divided
by P 2 so in that case we can very well
see that the marginal you know utility
from commodity 1 is much higher compared
to the utility gained from the commodity
2 so in that case the consumption of
good 1 should increase and consumption
of good 2 shall decrease if the reverse
is assumed and that is basically the
second case then what will happen then
we can see that as the utility from
commodity 2 is higher compared to the
utility coming from community 1 in that
case the reallocation should be such
that the consumption of X 2 shall
increase and consumption of X 1 shall
decrease now this is in a nutshell all
about martially on consumer demand
analysis but you can probably see that
there are some lapses or some
limitations in martially on consumer
demand analysis there are some we know
very strict assumptions take the case of
assumption that different utilities are
independent so you know this very
assumption of additively separable
utility function that is absurd think
about you know the consumption of you
know a cup of tea you have a cup of tea
then you have you know some sugar you
have some milk and you know there is an
optimum proportion of tea sugar and milk
will give you the highest you know
utility so you know it's very difficult
to assume that utility coming out of one
consumption good does not depend on the
level of the consumption of the other
good the another strict you know
assumption is the constant marginal
utility of money in reality these
assumptions don't take place that is why
you know in the 20th century Marshall
Lee and
analysis was heavily challenged and we
get a new stream of theory known as
Hicks aligned indifference curve theory
this is the modern approach to consumer
behavior in this course we are going to
study that and you know we are going to
continue with this discussion
[Music]
[Music]
