[ Music ]

>> Today we're going to

talk about unemployment.
Unemployment's a very important
topic in economic theory,
in economic policy, and also in
our own real-world experiences.
It can be difficult
to get a job.
On the other side of the market,
for firms, sometimes it's hard
to find an employee
to fill a vacancy.
That is, it's hard to find
exactly the right employee, or,
if you're a worker
looking for a job,
it's hard to find
exactly the right job.
So the classical
theory has very little
to say about unemployment.
If we represent the market by
a supply and demand diagram,
let's put the wage
on the vertical axis.
After all, the wage
is the price of labor.
Labor supply, I'm going to draw
as though it were
upward sloping.
It doesn't have to be, but let's
assume that it is upward sloping
for the sake of discussion.
I'm going to draw labor
demand as downward sloping.
That has to be downward
sloping because the idea is
when wages are higher, firms
want to employ less labor.
So the classical
economic approach is
to say the market
equilibrates at a wage
where supply equals demand.
Let's call it W-star for
the equilibrium wage,
and let's say the quantity
of labor is L-star.
It doesn't mean everybody's
working 24 hours a day
and 7 days a week, of course.
But it means that at the wage
W-star, workers are working
as much as they'd like and,
similarly, firms are employing
as much labor as
they would like.
There is no unemployment because
even though people may not be
working 24 hours a
day, they're working
as much as they would like.
Now, of course, you could
try to amend the theory.
For example, if the wage
were, for some reason,
above the equilibrium wage,
then what you can see is
supply will exceed demand.
There's excess supply
of labor at that wage.
One might call that unemployment
because workers would
like to work a lot more
than firms want to hire and,
therefore, workers, some of
them, are going to be rationed.
It's not a crazy theory,
but it misses one point.
In the real world, not
only are there some workers
who have currently no job,
there are also some employers
who currently have
unfilled vacancies.
So if the wage is above
the equilibrium wage,
you get workers who aren't
working as much as they like,
but firms are hiring
as many workers
as they would like at that wage.
Now, symmetrically, the
wage might be too low.
In that case, you can see
there's excess demand for labor.
Workers are only
working a certain amount
because the wage is low.
Firms would like to hire more,
but the supply is
not forthcoming.
So there you can have
unfilled vacancies.
What the classical
theory cannot explain is
that there are simultaneously
some workers out of work looking
for a job and some firms
with unfilled vacancies.
In equilibrium, you
have neither.
When the wage is too high or
too low, you might have one
or the other, but
you can't have both.
This is where the study of
markets with frictions comes in.
>> So while the classical theory
of the labor market is
interesting for lots of reasons,
and it has many insights that
we can use in policymaking
and understanding
the real world,
it cannot simultaneously
generate unfilled vacancies
and people looking for work.
The modern theory of the labor
market takes friction seriously
in the sense that it takes
time and other resources
to fill a vacancy,
and it takes time
and other resources
to find a job.
The branch of theory
that we're going
to use is called search theory.
Search theory, in
general, is the theory
of looking for anything.
You could be looking to buy a
house, looking to find a spouse.
You could be looking to
sell your car or buy a car.
And for this application,
we're going to talk
about people looking
to find a job.
We'll come back to the firms
trying to fill vacancies later.
For now let's focus on the side
of the market which corresponds
to labor supply, that is
workers trying to find a job.
So I'm going to give a very
simple theoretical model.
It can be generalized and
made much more intricate,
but let's not try to
run before we walk.
The model's going
to look like this.
I have some assumptions.
Assumption one.
Only the unemployed search.
The employed do not look
for better jobs while
they're currently working.
That assumption is, of course,
important to generalize,
but let's start slow.
So what we can say is
that jobs last forever.
This means that once you
take a job, you're done.
Second assumption.
While you're unemployed,
you're going to sample wages.
Now let's assume you get
one offer per period.
That offer will be denoted at
a lower case w. It's the wage,
just like in the classical
theory, but it's going
to be a random variable.
[ Pause ]
That is, any given period --
a period could correspond
to a month, or three months,
or a year, depending on
the application at hand.
So for our purposes, a period
is some length of a time,
and each period you're going
to get a random wage offer.
I'm going to call it
a wage, but of course,
jobs have many characteristics.
There could be prestige
considerations,
do you get a corner
office, perks and benefits.
That's all summarized, for
this talk, by the wage.
Now let's also assume
if you turn down a job,
you cannot later call them back
and say I've changed my mind.
We call that No Recall, an offer
which is rejected
is lost forever.
Again, that's easy
to generalize,
but let's do the
simple case first.
So what do you get
when you're working?
You get some wage.
What do you get when
you're not working?
Well, let's denote it by B,
B for unemployment
insurance benefits.
Actually it captures a lot
of other things as well.
It could be the value
of leisure.
You could be doing
some home production.
Whatever you're doing, and
whatever you're getting,
let's denote it by B. The only
other assumption we need is
that agents discount the future.
There's some discount
rate in economics.
Beta is typically what we
use for the discount rate.
And beta is a number
less than one.
What it means is a wage
next period is worth less
than a wage this period.
You'd rather have
it now than later.
How much less?
Well $1 next period is worth
beta, less than $1 today.
>> So based on those five
assumptions, what would you do
if you were looking for a job?
I'm going to describe it in
terms of a couple of equations,
which may look complicated,
but they're very easy
to understand once we talk
about them just a little bit.
So the first thing
I want to work
out is what's the
value of a job?
Let's say upper case W is what
you value a job while working.
And it's going to be a function,
of course, of the wage.
So upper case W is a function
of lower case W. What do you get
when you're employed at wage W?
Well, the first thing you get
is you get the wage, W. Now,
next period, which we
discount at rate beta,
you get something else.
Because jobs last forever,
and because this world is
not changing over time,
and you're not growing older,
what you get tomorrow is the
same as what you get today,
W of W. This says that the value
of a job is the current wage
plus the continuation value.
Next period, you
keep the same job.
This equation is pretty simple.
I can solve it.
The value of a job
is just lower case W,
the wage, over 1 minus beta.
How'd I get that?
Well, I moved upper case
W to the left-hand side,
grouped terms, and divided.
This is the present
discounted value
of earning W per period forever.
The second equation I'm going
to need is the value
to being unemployed.
Let's denote that by upper
case U. So what do you get
when you're unemployed?
By similar logic you get
your current reward, B,
that was our definition for
the value of being unemployed
within a period, and the
next period, which, as usual,
we discount at rate beta,
you get something else.
What do you get?
Well, by assumption
you get a job offer.
Now when you get a job offer,
you don't have to take it.
Why might you not take it?
You might want to hold
out for a better offer
from some other employer.
What you do is you
take the maximum payoff
to either accepting
or rejecting.
If you accept, well, upper case
W is our definition of the value
to working at that job.
If you reject, well, upper
case U is, by definition,
the value to remaining
unemployed.
It's almost the final
version of the equation.
You may notice I left
a little space there.
I want to put in
E for expectation.
What does that mean?
Well, recall that
the wage is random.
It may be high, it may be
low, it may be middling.
So you have to form
some expectations.
Technically, it's
simply the average.
So, for example, if the
wage is either 1 or 10,
each with a probability
one-half,
then the average offer is
1 plus 10 is 11 over 2.
Now I'm taking the expectation
over something more
tricky, not just the wage.
It's the expected value
of either accepting
or rejecting, whichever
is greater.
>> Okay, let's try to
represent those two equations
in terms of a simple diagram.
What I'll do is put the wage,
lowercase w on the
horizontal axis,
and now I'll draw the two
functions we just discussed.
The first one's pretty easy.
The value to working
at wage w is,
you may recall, w
over 1 minus beta.
The present discounted value
of working at wage w forever.
Well, that's just the
line through the origin.
It's a linear equation
with a slope,
1 over 1 minus beta
and 0 intercept.
So, I've depicted the value to
working at a job that pays w
as a function of w, by
this upward sloping line.
The other function's
even easier; u,
the value of being unemployed
doesn't depend on the wage,
because you don't have a
wage when you're unemployed.
What you do is you search, and
you try to find a good wage.
So, that's just a number.
That is to say, it's a
horizontal line as a function
of w. Now, from this picture,
the optimal search
strategy is quite clear.
Let's denote the
intersection of those two lines,
the horizontal line and the
upward sloping line representing
respectively the value of
being unemployed, and the value
of working at a particular
wage by the letter R. R stands
for reservation, because
that's your reservation wage
in technical jargon.
What it means is, you're
indifferent between working
at the wage R or continuing
to search for a better wage.
So, the strategy is clear, if
you get an offer less than R,
you should reject that offer;
since by definition you're
indifferent between working
at the reservation wage
and continuing to search,
you strictly prefer searching
at any wage below
the reservation wage.
And, symmetrically, for any wage
above the reservation
wage, you should accept.
Again, you're indifferent
between working between working
at wage R and continuing the
search, so if you get an offer
above R, you should accept.
The search strategy
is therefore clear;
agents should compute
the reservation wage,
the wage at which they're
indifferent between continuing
to search and starting to work,
and sample offers over time,
until they get an offer
above the reservation wage.
At that point, they
start working.
We're going to use this theory
to talk about the duration
of unemployment, and then later
about the unemployment rate.
But, we've got the
critical first piece.
It's the decision of workers
to accept or reject offers.
>> Okay we determined the
optimal search strategy
as the determination of
a reservation wage, R,
such that you're
indifferent between working
and continuing to search.
Let me show you how to figure
out in more detail
what R should be.
First off since the value
to working at any wage
by definition is W of that
wage, the value to working
at the reservation wage is W of
R. Now since you're indifferent
between working at the
reservation wage and continuing
to search, W of R must equal
U. U, after all, is the value
of continuing to search.
Now I'll re-write the
left hand side as follows.
Since the value to
working at any wage, W,
is simply W over one minus
beta, in particular the value
to working at the reservation
wage is R over one minus beta.
Now the right hand side
I'll also re-write.
The value to searching
is U and that equals B:
unemployment benefits, plus
leisure, plus other things;
plus next period you
will get a new offer.
You discount so I must put
in beta, the discount factor;
then it's going to
be a random offer -
I have to put in
your expectation.
You have to average over
different possible offers.
Then I have to put,
you may recall,
this maximization operator
because when you get an offer
next period you may accept it
or reject it.
If you accept it, I'm
going to plug in the value
of accepting that job.
As I say, that's W
over one minus beta.
Remember the W is random, that's
why we have the expectation.
If you reject the offer,
you continue to search
and I've just convinced you
of the fact that the value
to searching is R over one minus
beta, so let's plug that in.
That's basically
the final answer.
You may choose to re-write
it by multiplying both sides
by one minus beta;
it's slightly nicer.
The reservation wage will be one
minus the discount factor times
the value, the instantaneous
terms
of being unemployed including
benefits, leisure, etc..
Plus you see when I multiply
by one minus beta it's going
to cancel inside the
maximization operator?
So I'm left with
something pretty nice.
This is one equation
and one unknown.
This equation determines
the reservation wage.
Now the exact formula for the
reservation wage will depend
on parameters like B,
beta, and the distribution
of possible wage offers.
So I'm not going to try to
give you an exact answer;
it's going to depend
on many things.
This one equation and one
unknown is enough for us
to understand unemployment.
>> Okay, we've seen now what the
optimal job search strategy is.
You should choose a
reservation wage R. We've talked
about how to determine that.
And then you should
keep searching
until you get a job offer where
the wage exceeds R, then accept.
So the actual reservation
wage solves
that reservation wage equation.
It determines R but it will be
a function of the parameters
of the problem including,
for example,
beta and B. Now it's not
hard if you know calculus
to take a derivative
and show the following:
if beta is higher, the
reservation wage will be higher.
Similarly, if B is higher, the
reservation wage will be higher.
It's fairly intuitively obvious.
If beta is higher,
you put more weight
on the future - you
discount less.
So you're leaning more
towards getting a good offer.
You're willing to
hold out longer.
In fact, if beta were 0 that
would mean you don't care
about the future at all.
You only value your
current income.
In that case, go back and check
the reservation wage equation.
It's obvious that
when beta equals 0,
R equals B. That means your
reservation wage is simply equal
to the value of being
unemployed.
Now similarly, when B goes
up you can take a derivative
if you know calculus
and show logically
that the reservation
wage must be higher.
That's also fairly clear.
If unemployment insurance
for example is higher,
it's less painful being
unemployed - you're more willing
to hold out for a better job.
Let me introduce one
more new variable: gamma,
the Greek letter gamma.
And that's going to be
simply the probability
of getting a job.
That's simply W greater
than R. In other words,
you accept a job whenever the
offer exceeds the reservation
wage, therefore your probability
of getting a job is the
probability of sampling an offer
above R. Now whenever
you have a probability
which is the same every
period, in this case gamma,
how long will you search?
Turns out there is a very
simple formula from statistics
that says the expected
duration of search,
in other words the expected
duration of unemployment,
is simply 1 over the probability
of getting a job in that period.
For example, we now
can say things
about how policy
affects unemployment.
If B goes up, so unemployment
insurance is higher for example,
we've established that the
reservation wage will be higher.
This will reduce gamma because
it's a lower probability
of getting a W in excess of
a higher reservation wage
and this will increase
the expected duration
of unemployment.
It doesn't mean you're
worse off, but it means
that it will take you
longer to find a job.
Why? When B is higher
unemployment is not so painful
so you're more willing to
hold out for a better offer.
Okay, let's take stock.
We've derived the
reservation wage equation
which determines R, the
reservation wage, as a function
of parameters of the
problem including B:
the value of unemployment
insurance, plus leisure,
plus et cetera; and beta,
the discount factor.
We want to now talk about
individual employment histories
as agents moving into and out
of unemployment over time.
We have to generalize the model.
The first thing we'll do
is add a variable lambda.
Let's call that the layoff rate.
The Greek letter lambda
corresponding to L for layoff.
This means that every period
when you're working at a job,
with probability lambda, the job
ends, you're back on the market.
With probability 1 minus
lambda, you continue in the job.
While we're at, it let's
introduce one more new
parameter: the Greek letter
alpha, for the arrival rate.
This is the probability
of getting an offer
any given period.
The previous model had
lambda equal 0, so no layoffs
and alpha equal 1, so you
get an offer every period.
Now we're going to
generalize it.
One can derive the reservation
wage following the same
procedure as before, and I'm
not going to do it for you.
You can do it on your own.
Then we just rewrite the
equations in their new form.
Starting with the value of
unemployment, U, as usual,
you get B within the period and
next period, which you discount
at rate beta, something happens.
Before, what happened
was you got a job offer
with probability 1.
Here, you only get an offer
with probability alpha
and when you do, we just
apply what we learned before:
to write the expected maximum
of either accepting that job
and working at wage
W, or rejecting it
and remaining unemployed.
However in this generalized
version of the model,
with probability 1 minus
alpha, you do not get an offer
so you're forced to
remain unemployed.
One could see that
if alpha equals 1,
this is precisely the
value of unemployment
from the simpler
version of the model.
The next equation I'll show you
is W of W, the value to working
at wage W. As in the
simpler version of the model,
you get the wage that
period and as always,
you get the discounted value
of what happens next period.
Now what happens next
period is pretty simple.
With probability 1 minus
lambda you are not laid off,
hence you continue in your job.
But with probability
lambda, you're laid off
and you're forced to rejoin
the ranks of the unemployed.
One could see that
if lambda equals 0,
this reduces the equation
from the simpler model.
This generalized
version of the model
which allows both random
arrival rate of offers
and random layoffs is
interesting because we can talk
about labor market transitions
into and out of unemployment.
>> Okay let's take stock.
We now have a theory of
the reservation wage,
that is the wage r at which
workers are indifferent
between accepting a job
and continuing to search.
And we worked out how that
r depends on the parameters
of problem including b,
on employment insurance
plus the value of leisure,
including lambda the layoff
rate, including alpha,
the arrival rate
of offers etcetera.
Let's now put this
theory into words to talk
about individual's transition
between employment
and unemployment.
So first in this version
of the model, lambda,
the layoff rate is
the probability
that an individual
transits between employment
and unemployment, which
I'll denote this way.
Then we have the probability
of going the other way
that was called gamma before,
it's the probability of going
from unemployment to employment.
Let's write that out.
It's the product of two terms.
Alpha is the probability
of getting an offer,
I have to multiply
that by the probability
that the offer is acceptable,
which means the probability
of the random w is an access
of the reservation wage.
Sometimes we refer to this
as saying job creation is the
product of choice and chance.
Choice means you have to accept.
It is also you have to
have a chance to accept
by getting the offer
in the first place.
This theory can be used to
describe how individuals transit
between employment
and unemployment.
Next we use this theory to talk
about the aggregate
unemployment rate.
>> Okay, let's now imagine
an economy consisting
of a large number of
individuals who look just
like the individual
we studied earlier.
Sometimes they're unemployed;
they're searching to get a job
and they keep searching
until they get a W in excess
of the reservation wage.
And when they're employed
there's some probability lambda
that they're laid off.
So let me now represent
the population in terms
of the following diagram.
There is some fraction who
are unemployed, call that U;
and there's some
fraction who are employed,
we can call that E - but since
they're fractions they add
up to one, it's just
one minus U. Now those
who are unemployed end
up becoming employed
at a certain rate
that we called gamma.
Those who are employed become
unemployed at a certain rate
that we called lambda.
This economy will evolve over
time, starting with any division
of individuals into the
employed and the unemployed.
Some will get jobs,
some will lose jobs.
We could write out
the law of motion
for the unemployment rate,
describing how it changes
over time, but let's
for simplicity focus
on what's called a steady state,
which is an unemployment rate
which does not change over time.
It's the unemployment rate to
which the economy will gravitate
as time goes on - sometimes
called the natural rate
of unemployment.
To compute the natural rate
of unemployment we simply have
to multiply the number of
unemployed - or the fraction
of unemployed - times the
flow from left to right,
and equate that to the flow
in the other direction,
which is the fraction
of the population
that are employed
times the layoff rate.
This is one equation
and one unknown.
I can solve it easily
enough for the natural rate
of unemployment, which is
lambda over lambda plus gamma.
This is the rate of unemployment
to which the economy will
gravitate in the longer run,
holding constant parameters,
in particular lambda and gamma.
>> Okay. Let's wrap up.
We've developed a theory
in terms of a model
that represents how
individuals look for work.
Any individual gets
offers randomly over time.
And when they're
acceptable, they transit
from unemployment to employment.
While working, sometimes
they have the misfortune
of being laid off and transit
in the other direction.
This theory can be used to talk
about the unemployment
rate in the economy.
At any point in time,
there will be some fraction
of the population working and
some fraction not working,
and people will move
back and forth.
But the system tends
to gravitate towards the
natural rate of employment
which stays constant over time,
even though individuals
are moving back and forth.
This looks a lot different
from the classical theory.
There's no supply
and demand diagrams,
although they are
related - workers looking
for work are supplying labor.
The interesting thing
about search theory is
that search is a
productive activity.
Holding out for a
better job is good.
Now, you don't want
to hold out forever.
At some point, you've got
to stop searching
and start working.
We determined theoretically
exactly what you should do:
choose a reservation wage and
accept when a job gives an offer
above the reservation wage.
It's still not all the way there
vis-a-vis the classical theory,
because we've got
workers looking for jobs.
We still want to build
in the other side
of having firms looking
for workers.
That's the subject
of the next lecture.
[ Music ]
[ Music ]
>> Okay, we're going to
continue our discussion
of unemployment using
search theoretic models.
Last time we talked mainly
about individuals solving
the search problem.
When they're unemployed,
they sample wages
from a distribution.
And, when they find a
sufficiently good wage,
they will stop searching
and start working.
What does it mean to be
a sufficiently good wage?
We derive the mathematical
answer to that question.
There'll be a number called the
reservation wage, and any offer
in excess of the reservation
wage will be acceptable.
We also talked about repeated
spells of unemployment.
Individuals would accept
offers when they were
above the reservation wage.
They would work on that
job, but they were subject
to stochastic layoffs.
Every so often you'll
lose your job
at which point you re-enter
the unemployment pool
and continue the cycle.
One of the critical components
of that model was the
arrival rate of offers.
We used the Greek letter alpha.
That was the probability
of getting an offer
in a given period.
Today we're going to
focus on the arrival rate,
and we're going to think of a
general equilibrium perspective.
Before that model was
basically due to Dale Mortensen.
He was concentrated on
individual search problem.
I'm going to talk
about two models today,
one due to Peter
Diamond which endogenizes
who meets whom and how.
After that, we're going to
talk about a generalization
by Christopher Pissarides
which ties together
the two approaches.
Mortensen's approach
and Diamond's approach,
and provides the
current frontier in terms
of economic theories
of unemployment.
>> Okay, let's talk about
Diamond's model of search.
It's different from
Mortensen's approach,
which was explicitly built
to study the labor market.
We're going to back up and talk
about more abstract
search models for awhile,
and that will be an ingredient
to Pissarides' model
we discuss later.
So Diamond thought it was
important to have agents trading
with each other as opposed
to simply trading along
their budget line,
as in classical economics.
The first thing you have
to ask when agents trade
with each other is,
how do they meet?
Let me show you what
Diamond did.
First, he decomposed economic
activity into two sectors,
sometimes referred
to as islands.
One would be the
production sector,
or the production island, and
the other would be the trading
or the exchange sector.
Now, agents on the production
island are going to be looking
to produce goods, and then,
in this model, they're going
to produce goods that
they are not allowed
to consume for themselves.
That's just an assumption
Diamond had to make.
We can dispense with that
later and talk about how
that endogenously happens.
So what they'll do is take their
goods over to the trading island
and try to trade for something
else that they can consume.
So let's solve these
problems one at a time.
We'll start with the
hard one, production.
The payoff or value
function of an agent,
just like in the job search
model, we'll call it V,
and we'll use the subscript 0 to
indicate the goods in inventory.
If you're a producer, you do
not yet have goods in inventory.
What Diamond assumes is
there's some arrival rate alpha,
just like in the
job search model,
but here it's not the arrival of
a wage offer, it's the arrival
of a production opportunity.
When it comes along,
you will choose
to accept it or reject it.
Now, Diamond thought about
the natural commodity
in an island economy
should be a coconut.
That's what people
produce and consume.
So he thought of a
production opportunity
as you locate a coconut tree,
and if you want to climb up,
you can get a coconut.
So here's how it
works: if you decide
to accept this opportunity,
next period your value function
will be V1, and we discount
that at rate beta, just like
in the job search model.
However, you have to pay
a cost; so subtract C,
Diamond quaintly described it
as some coconut trees are higher
and hence more costly to climb.
So this is your continuation
value minus the cost
of production, if you decide
to accept that opportunity.
Of course, you're free to
reject and continue searching
for a better opportunity,
just like a job search
agent can reject an offer
to search for a better offer.
If you reject, instead, your
continuation value will be V0
because you still do not
have any goods in inventory.
So that's your decision
problem, to accept
or reject the production
opportunity.
Since the cost C is random,
let's put the expectation
out in front of that.
Now, notice, this
is what happens
when you have an opportunity.
With probability 1 minus alpha,
you do not locate a production
opportunity so you're forced
to continue with
nothing in inventory.
This equation describes
the payoff to an agent
in the production sector looking
for a production opportunity.
>> Okay, let's move on to
talk about what happens
after an agent has produced.
So he's already finished
producing -
he's found a coconut tree
which wasn't too high,
so he climbed it
and got a coconut.
Remember, in this
particular specification,
you're not allowed to
consume your own output.
So what he's going to do,
he's going to become a trader.
He'll actually physically move
from one sector to the other,
different location, and
look for a trading partner.
So V1, we've already decided, is
the value function for an agent
with one unit of
goods and inventory.
That will be equal
to the following.
With some probability, he
meets somebody just like him,
a trading partner; that
probability is gamma.
And if he meets that person,
they're going to trade.
In this specification, all goods
are the same, so you're willing
to trade with anybody except
for goods you produced yourself.
You're not allowed
to consume those.
So with probability gamma,
he meets a counter
party, and they trade.
He gets utility from
consuming the coconut.
In case I haven't said it
explicitly, in this framework,
goods are indivisible.
You either consume
it all or none.
After consuming, you have a
continuation value, beta V0.
What happens is, after
a trade and consumption,
you go back to the production
island to renew the cycle.
On the other hand,
he may be unlucky
and not find a counter party.
That happens with
probability 1 minus gamma
in which case you're forced
to continue as a trader
which has discounted
value beta V1.
Since we're going to be
using these napkins again,
let's number them to keep track
of where they are
and where we are.
>> Okay, there's a few
things we have to do
when we analyze this model.
First, let's decide
when agents produce.
This is analogous to
the job search model.
You're sampling wages,
you're going
to have a reservation wage,
and when you draw a wage
above your reservation
wage you should accept -
stop searching and
start working.
This is different, but
similar mathematically.
It's different because here
you're a producer, not a worker.
You're looking for production
opportunities that are low cost.
So there'll still be a
reservation strategy.
Let me show you.
If you look at this maximization
problem, it should be clear
that you should produce
if and only if,
beta V1 minus C exceeds beta V0.
Let me rearrange that.
You should produce
if C is less than
or equal to beta V1 minus V0.
Now let's define
the right hand side
to be K. K is the
reservation production cost,
analogous to a job search model
where there's a reservation wage
and you continue searching
until you sample a wage
above the reservation wage.
Here, you're looking to produce.
You sample opportunities
until you find an opportunity
with a cost below
the reservation cost.
So that's relatively simple.
We've now figured out when
producers will produce.
Remember that in this
theory, you're not allowed
to consume your own output.
So after production, agents will
transit to the trading sector
where they look for
a trading partner.
That decision's pretty easy.
All trading partners are
the same in this model,
all goods are the same except
the ones you produce yourself.
So the arrival rate gamma
is taken as a parameter
by the individual, and
when he has an arrival,
he trades for sure.
This describes the cycle
of production and exchange
that occurs in this economy.
The next thing we're
going to do is talk
about where the arrival
rates come from.
>> Okay, we've now decided
on the strategies of agents.
Production is the
interesting decision.
Trade is actually easy.
We'll change that later,
but for now let's focus
on the production decision.
The next thing to determine
is the arrival rates
of opportunities.
There are two.
Alpha is the arrival rate
in the production process.
Gamma is the arrival rate
in the exchange process.
Before we determine those,
let me say there's a variety
of different approaches
one can use.
For now, to fix ideas,
let's assume alpha is
an exogenous parameter.
We already have the
endogenous production decision
in the production process,
so we'll not worry too much
about the arrival rate for now.
Come back to that later.
Instead, let's focus
on the arrival rate
in the trading process, gamma.
To describe that, let's go
back to napkin #1 and talk
about the flows of
agents in both directions.
You should remember this
from the last lecture,
where we talked about flows of
individuals from unemployment
to employment, and
back the other way,
from which we derived
the steady state
or natural rate of unemployment.
This will be a very
similar exercise.
Let's assume there are N people,
that is to say a fraction N
of all the population,
in the trading process.
And the remaining fraction, 1
minus N, still in production.
Let's rate the probability
of production
as pi - pi for production.
It's simply the probability
that somebody searching comes
up with an opportunity with
cost C, less than or equal
to the reservation cost
K. Therefore, the fraction
of agents is 1 minus N.
They're each producing
with a probability pi, 1 minus
N times pi, describes the flow
of agents from production
into exchange.
The flow of the other
direction is even simpler.
That's just gamma times
the number of traders N.
So this describes the flows
of agents from production
into exchange, and
back the other way.
Just like in the
model of unemployment,
we characterize the steady
state, or natural rate
of unemployment, by equating
the flows in both directions.
So let's go back to this napkin,
and simply equate those flows;
1 minus N, times pi,
equals gamma times N.
If N satisfies this equation,
the flows are balanced
in the two directions, and
hence, N will remain constant.
This is the natural number
of traders in the economy.
>> Okay, what we've done
so far is describe the
important decision problem
of agents: when to produce.
Now let's talk about
arrival rates; there are two.
Alpha is the arrival rate
of production opportunities,
and gamma is the arrival
rate of trading partners.
Either one could be studied.
To fix ideas, let's assume
that alpha is a fixed
parameter and focus on gamma.
This is the same
assumption that Diamond made
in his original theory.
To talk about gamma,
the probability
of meeting a potential trading
partner, Diamond suggests
that we write gamma as a
function of N. He was thinking
that it may be more or
less easy to meet somebody
when there's more
people searching.
His natural inclination was
to suggest gamma's an
increasing function.
The more people there are
searching for trading partners,
the faster you meet them.
Question: does that
have to be the case?
Well, to think about that,
let's talk about
something slightly deeper -
the matching technology.
I'm going to describe a matching
function, or a meeting function,
which is analogous
to a production function
in producers theory.
In producers theory,
there's some inputs: labor,
capital, maybe others.
And given the vector of inputs,
there's a production function
telling you the output.
Modern search theorists think
about meetings the same way.
There's some inputs; in this
case, there's only one input -
people searching on
the exchange island.
There's going to be output
- the number of meetings.
So M is the number of meetings,
and that will be some function
of N. It's natural to
think there's more meetings
if there's more people.
But how many more meetings?
Let's draw a picture.
If we put the number of people
who are trading partners
on this axis, and I put the
number of meetings on that axis,
we could imagine
three possibilities.
It could be that, when you
increase the number of people,
you more than increase
the number of meetings.
So this is the case referred to
as increasing returns to scale.
If we double the number of
people in the exchange sector,
we'll more than double
the number of meetings.
I could, of course, imagine also
decreasing returns to scale.
This means if we double
the number of traders,
we increase the number
of meetings
but by less than double.
And, of course, one could
imagine the intermediate case,
constant returns to
scale, where the number
of meetings is simply
proportional
to the number of traders.
This is useful because it makes
sense to think about gamma,
your probability of a meeting,
as the total number of meetings,
M of N divided by N. So we see,
we've reduced it
to a simpler idea.
If the underlying meeting
technology is increasing
returns, then M increases
faster than N,
then it's true your
arrival rate will be higher
when there's more traders.
That was the case
Diamond focused on.
It doesn't have to be that way.
It could be decreasing
returns of scale
where your arrival rate
is decreasing the number
of trading partners.
That's often referred
to as congestion.
And, in the intermediate
case, constant returns,
your arrival rate is actually
independent of the number
of traders because
constant returns
to scale implies the number
of meetings is proportional.
Say A times N, which means
your arrival rate, gamma of N,
is A times N divided by N,
simply the constant A. You
might think that's a very
special case.
It turns out it's probably
the empirically relevant case.
We'll talk about that later.
Before we do that, let's put
the whole thing together.
I'll number this napkin 4,
and next I'll summarize.
>> So let's sum up what we have.
We have an economy where agents
are producing and trading.
They're trading with each other,
not only against
their budget line.
There are frictions
in the process.
It takes time to find
a trading partner.
It takes time to find a
production opportunity,
especially a good one.
The frictions we're focusing on
right now are the ones in terms
of finding a trading partner.
Your arrival rate gamma
depends on the number of agents.
So putting things together,
we could draw a picture just
like in Diamond's
original article.
There are two variables to be
determined in this theory, N,
the number of traders, and
K, the production cost.
Now, there are these
two variables N and K,
and we have these
two relationships.
We have the relationship
that says your arrival rate
gamma is a function of N,
and N is going to be determined
by the steady state condition.
So it's intuitively
clear, the higher is K,
the more willing
agents are to produce,
the more production will
occur, the higher the flow
into the trading
sector and, therefore,
the higher the number
of traders.
So, as K goes up, we have
this increasing function
which determines N as a
function of K. To reiterate,
when K is higher, there's more
production, hence, a higher flow
from production to
exchange and, hence,
a greater number of traders.
But, if there's a
greater number of traders,
we have another relationship.
Your reservation production cost
will depend on the gain you get
when you move from
production to exchange.
Therefore, it's going to
depend on the arrival rate
in the exchange sector and,
hence, on the number of traders.
Interestingly, that's also
an increasing function.
So Diamond's diagram
looks something like this.
Both the number of
traders as a function of K
and the reservation cost K as
a function of N are increasing.
Now, Diamond made a big deal
of the fact that, therefore,
we can get multiple equilibrium.
I guess I'd better
define an equilibrium.
An equilibrium is simply a
solution to these two equations.
It tells you the number
of traders as a function
of the reservation
cost K. That comes
through the mechanical
steady state condition,
and it tells you the
reservation cost K as a function
of N. This comes through
the production decision
of individuals.
If we have a solution to
both of these equations,
then we have an equilibrium
in the model.
There are multiple
equilibria in this picture.
Makes sense.
For example, if not many people
are producing, there's going
to be very few traders; hence,
it's harder to find
a trading partner.
Hence, you're not willing to
pay a very high cost to produce,
and this rationalizes a
small number of traders.
But if only we can have more
traders, then it would be easier
to find a counter party,
people would be more willing
to produce, K would increase,
and we have another equilibrium.
So it turns out that requires
increasing returns of scale.
If there's constant returns,
you may recall then your arrival
rate does not depend on N
in which case there is
a unique equilibrium.
Why? Because K is
actually independent
of N. Let me show you
that case briefly.
I'm making a big
deal out of this
because whether it's constant
or increasing returns is
going to matter a lot.
The diagram I just completed
-- let's number that 5 --
is the case of increasing
returns to scale
where your arrival rate is
increasing and the number
of traders and, hence, your
reservation cost is too.
If we do the case of
constant returns to scale,
which recall I said is
probably the empirically
or at least maybe the
empirically relevant case,
we still have this
relationship between K and N.
When K is bigger, you get more
people in the trading sector,
but now the arrival rate
is constant and, hence,
your reservation cost is
horizontal as a function of N.
In that version of the model,
there's a unique equilibrium.
It matters what we
assume, therefore,
about the meeting technology.
If it's increasing returns,
the economy could get
stuck in a bad equilibrium.
In case I didn't mention
it, the equilibrium with N
and K higher is a
better equilibrium
than the one with N and K lower.
If there's constant
returns, on the other hand,
the economy has unique
equilibrium,
and it turns out
to be efficient.
So I don't want to make a big
deal of this, but it's important
to put it in historical context.
The difference between
increasing returns
and constant returns was
the substantive contribution
of Diamond's article.
I think, however, the more
important contribution is
writing down a model where
agents trade with each other
and not simply with
their budget line.
And once we have that, we
can talk about unemployment
in an interesting way.
[ Music ]
[ Music ]
>> So far in search theory,
we've discussed Mortensen's
basic model
of how an individual
looks for a job.
We then moved on to talk
about not only individuals
but the economy as a whole,
to think about equilibrium.
And we had an unemployment
rate which settled
down to the natural
rate of unemployment,
depending on parameters,
including the arrival rate.
The next step was to talk
about Diamond's search model,
which is a truly
equilibrium model.
Agents have interesting
interactions with each other;
they're trading with each
other and, importantly,
the arrival rate could
depend on their actions.
More production, at least
with increasing returns,
could lead to more matching and,
potentially higher
arrival rates.
Now, I want to put these
two ideas together.
This is what Pissarides did.
He said, "Let's think
about a labor market."
Now, it's important to make
this a two-sided market,
because in labor markets agents
are looking for different types.
Firms are looking for workers;
workers are looking for firms.
That's in contrast
to Diamond's model,
which is a one-sided market -
trading partners who are looking
for other trading partners,
but they're all the same.
So Pissarides decided,
"Let's write
down two separate
search problems."
One for a worker.
So a worker is going to have a
value function V. I'll subscript
it by W for "worker."
If he's unemployed, here's what
he gets: he gets some amount B,
just like in the
basic Mortensen model,
reflecting unemployment
insurance,
the value of the leisure, maybe
home production, et cetera.
Then he may get an offer.
Alpha is the notation
for the arrival rate.
I have to superscript
that by W as well,
because it's a two-sided market.
Alpha superscript W is the
arrival rate for a worker.
Now, in general, we have to take
into account offers
may be different.
But to focus on simple cases,
let's assume all job
offers are the same.
You are not searching for
a good wage in this model,
you're just searching for a job.
If you get an offer,
therefore, you should accept.
If you accept, next period
you have beta times V,
superscript W, subscript 1.
Next period you'll be working
and V1 will be the value
of being employed, with the
superscript denoting a worker.
Of course, if you don't get
an offer, 1 minus alpha W,
you're forced to
continue a search.
Beta times VW, subscript 0.
This describes the pay-off of
an agent looking for a job;
it describes it in terms of
itself and the value to working.
Let's write down the
value to having a job.
Subscript 1 is the
value function
for an agent who's
matched with a partner.
For a worker, it means he
has a firm - he has a job.
In that case, he gets the
wage, W. I know there's a lot
of Ws floating around, but
there's only so many letters.
Next period, one of two
things could happen.
It could be, with probability
lambda, that match breaks up -
the job is destroyed,
the worker is laid off.
Notice, these are the
same notations we used
in the simple Mortensen model:
alpha for the arrival rate
of offers, lambda
for the lay-off rate.
In that case, you're forced
to join the unemployment pool.
But if you're not laid off,
you continue working
at the same job.
These two equations
describe a worker's problem,
like the agent's
problem in Diamond
or the original search model
of agents looking for work.
Notice, I've superscripted alpha
by W. It's the worker's
arrival rate.
The worker's arrival
rate, in general,
differs from the
firm's arrival rate.
And I'll explain that
in a few minutes.
But notice, lambda, the
lay-off rate, is not indexed.
Pissarides and some other
people find it useful to think
about employment
relationships as bilateral,
one worker matched
with one firm.
Even though, in reality,
firms often have many workers,
and sometimes workers
have two jobs.
This is an abstraction,
and it's a good one
for the purposes at hand.
So we can think of these
two equations as describing
for any wage W, the full
value to looking for work,
and the full value
to having work.
Like in the earlier models,
there is going to be a cycle
where agents move
from employment
to unemployment throughout
their lifetime.
That's the description
of the worker's problem.
Let's number that Napkin 1.
>> Okay, we're done
with workers for now.
Let's talk about firms.
This is easy, because
it's symmetric.
Just like a worker looking for
a partner, which is a firm;
a firm could look for a
partner, which is a worker.
It'll be two equations
describing the value function
or a pay-off of a firm, both
when it does not have a partner,
V subscript 0, and when
it does, V subscript 1.
So for a firm currently looking
for a partner, there's going
to be some cost to
recruiting - let's call that K.
So the per period
pay-off to a firm
that does not have
a worker is minus K;
it has to pay to recruit.
Next period, which as
always, we discount by beta,
two things could happen.
Alpha superscript F is the
arrival rate for a firm;
that's the probability
the firm meets a worker.
Then its continuation
value, by definition,
is V superscript F subscript
1 - the pay-off to a firm
with a filled vacancy.
And if that firm does not
meet a worker, which occurs
at a probability 1 minus
alpha F, the firm continues
in the recruitment process - the
pay-off being VF subscript 0.
Looks quite a bit like the
value function for a worker,
except for the firm, it pays
K in the recruitment process.
The worker, for simplicity,
is not paying a search cost.
He simply enjoys B.
Also symmetrically,
the value to a firm with
a partner, VF subscript 1,
will be the profit
per unit time.
Let's call the real
output or the real revenue
of the firm Y. It
has to pay a wage,
so Y minus W is the flow
profit or the profit per period
to a firm which has
a worker employed.
But next period, two
things could happen.
With probability lambda,
the job is destroyed,
for whatever exogenous reason.
Of course, the generalized
version of this,
called the Mortensen
Pissarides model,
also talks about
why jobs break up.
But for this presentation,
let's take that as an
exogenous parameter.
If the job breaks up, the firm
now has a vacancy once again.
If the relationship does not
break up, the firm continues
with the worker in the vacancy.
These two equations describe
the pay-offs to firms,
both when they're
looking for workers
and when they have workers.
The model is symmetric.
It could be used to describe
men looking for women,
women looking for men.
People looking to buy a house,
people looking to sell a house.
It's a nice simplification
of Pissarides to think
about the labor market
as a two-sided market,
where workers look for firms
and firms look for workers.
It may not be 100% realistic,
but it's quite an
elegant approach
to consider them symmetric.
Once we have this, the next step
is to put everything together
and talk about equilibrium, just
like we did in the other models.
>> Now that we've described the
symmetric problems of workers
and firms, let's talk
about their arrival rates.
Remember, these are models
where frictions are important,
so we have to discuss
how people meet.
It's more interesting than
the basic Diamond model,
because the matching function,
or meeting function,
now is two arguments.
V is the vacancy rate,
that's the fraction
of firms looking
to hire a worker.
And U is the unemployment
rate, the number of workers
who currently don't have a job
and are looking to be hired.
The total number of meetings,
sometimes called the job
creation rate, is a function
of V and U. It's called
the job creation rate,
because in this model,
in its simplified version
that we're talking about,
all jobs are the same.
You're not looking
for a good wage,
you're just looking for a job.
So what we decided then is to
write the matching technology
as follows: job creation depends
on V and U. Now it's easy
to talk about the arrival rates.
Emulating what we did
with the Diamond model
of a one-sided market,
in a two-sided market,
the arrival rate for a worker
is simply total job creation
divided by the number of
workers who are looking.
Symmetrically, the arrival rate
for a firm is the number of jobs
that are created divided
by the vacancy rate.
If you think about it, these are
very similar to what we talked
about in the Diamond model,
except for this being
a two-sided market.
As we said in the analysis
of the Diamond model,
things will depend on whether
the matching function is
increasing, decreasing,
or constant returns.
In this case, there's some
good empirical evidence
to suggest it's probably
constant returns.
That's also very useful.
With constant returns,
it's a well-known property
in mathematics that I can bring
U inside of the parentheses
to write the arrival rate for a
worker as M of V over U comma 1.
Similarly, I can
write the arrival rate
for a firm as follows.
I'll do it this way:
divide top and bottom by U,
and I bring the 1 over U inside
upstairs, and downstairs I get V
over U. So what I've done is
I've derived the arrival rates
for workers and firms
as functions of V
over U. It's a very
important variable.
It's referred to as
market tightness,
the buyer-seller ratio.
Why buyer and seller?
Well, a firm is buying labor.
A worker is selling labor.
This market tightness
will determine everything
in the model.
So let's file this as napkin
3 and then proceed to talk
about equilibrium in this model.
>> Now that we have
this description
of where the arrival rates
come from as a function
of market tightness, let's
talk about steady state.
This will be very easy.
It's logically the same as
talking about steady state
in Mortensen's original
labor market setup,
or Diamond's coconut model.
U is the fraction of agents,
of workers, who are unemployed.
They become employed at rate
alpha superscript W. That's job
creation, the flow
into employment.
For it to be a steady state,
the flow out and the flow
in have to be the same.
What is the fraction
of agents flowing
from employment to unemployment?
Well, if U is the fraction of
workers who are unemployed,
1 minus U is the
fraction who are employed.
Those jobs are destroyed,
or they are laid
off, at rate lambda.
So we're now old hands at this.
It's a simple matter to
solve for the steady state
or natural rate of unemployment.
This one equation and one
unknown has the following
solution: U equals lambda
over lambda plus alpha W.
And let's catalog this
napkin as number four.
Things are falling into place.
>> We almost have the complete
Pissarides model laid out.
In fact, if we're willing
to say the number of firms
and workers are some fixed
exogenous variables, this is it.
An equilibrium for the model
would be the following objects.
Payoffs to workers and
payoffs to firms as they go
through their life cycle of
working, losing their job,
and looking for another
one on both sides
of the two-sided market.
What has to be determined is
the arrival rates, alpha W
and alpha F. We showed how
to describe those in terms
of market tightness, V
over U. Now U is determined
from the steady state condition.
V, the number of vacancies,
well, take the number of firms
who are in the market,
subtract off one minus U,
because the fraction of
employed workers is equal
to fraction firms
with filled jobs.
That's the number of vacancies.
That ratio will determine
the arrival rates.
We have a complete
description of a labor market.
We can use it to address
many policy issues,
to organize the data, and to
talk about a variety of topics.
>> So we have an internally
consistent description
of a labor market, which is
a two-sided search market.
And it describes the
unemployment rate in terms
of market tightness, and it
describes the arrival rates
in terms of market tightness,
and it describes the
payoffs to firms and workers.
This model is very
useful for many purposes:
describing policy implications
or policy predictions,
organizing the data,
and helping us
to understand worker
flows and unemployment.
Notice this model simultaneously
has unemployment and vacancies,
something that as we discussed
earlier, you cannot get
out of a classical supply
and demand analysis.
At the same time, there are
workers looking for jobs
and firms with vacancies
trying to recruit.
Why? Because of the frictions.
This is the right model to
think about the labor market.
Now we could generalize it.
One of the things
Pissarides decided
to do was introduce a
condition he called free entry.
He wanted to talk in more detail
about where this market
tightness comes from.
So just like in theories
of industrial organization
or theories of firms more
generally, we could say
if profits are positive, some
potential firms will enter.
If profits are negative,
some firms who are currently
in the market will drop out.
The free entry condition
must mean that a firm
that enters has zero
profits in the short run.
I say in the short run,
because the firm is entering
with an unfilled vacancy.
Once the firm recruits
and starts producing,
then it's going to have profits.
So the free entry
condition is very useful
because if we look back
at the firm problem,
we can see that a firm with
a vacancy pays some costs
to recruit K, and has next
period a discounted expected
value of having recruited.
So this equation can be
written K, the cost to a firm
of setting up, of trying to
recruit, must be equal to beta,
next period's discounted value
of the recruitment value,
alpha F being the
probability the firm hires,
VF 1 being the profit to a firm
that's currently a going concern
with the worker.
This equation, then, tells you
that because of free entry,
the payoff to a firm
that has a worker
in the discounted expected
sense, must be offset
by the cost of recruiting.
If the cost of recruiting
were lower, there'd be profits
to be made, more
firms would enter.
If the cost of recruiting
were higher,
firms are making a
loss, they would exit.
This equation must be
satisfied by free entry.
What equilibrates the model?
Well, you can see from our
earlier analysis that the flow
of value to a firm that actually
is a going concern is satisfying
this equation.
These two equations together,
in fact, can be solved;
it's two equations,
then two unknowns.
And we can see that the
value function to a firm
that has a worker depends
on the arrival rate.
So when we put this
all together,
this arrival rate's
going to have to adjust
to satisfy the free
entry condition.
Therefore, market tightness,
which after all is what
determines the arrival rate,
is completely pinned down by
this free entry condition.
So let's catalog this
napkin as number 5.
This is a generalized
version of Pissarides' model.
It's internally consistent,
but now we're making
endogenous the number
of firms participating in
the frictional labor market.
Free entry will say that
it's more or fewer firms,
depending on the parameters
of the problem including
policy parameters.
And this equation is going
to pin down market tightness.
Hence, it ultimately pins
down the rest of the model,
including unemployment,
and including the payoffs
to workers and firms.
That's a generalized
version of Pissarides.
>> So, we could continue
to extend the model
in a large variety of ways.
In particular, after introducing
free entry, we may notice
that there is one
variable in this model
which has not been
pinned down: W, the wage.
Where does that come from?
This is somewhat controversial.
There's many different
approaches
to determining the wage, and
it matters for the theory.
You can't just use
supply and demand
because this is a
frictional labor market.
So you can't say the wage
is going to be pinned
down to equate the supply of
labor and the demand for labor.
This is not the way
this model works.
This model is much
more realistic
and much more intricate.
One approach, which is common,
is to assume the wage
is simply negotiated
between a single worker and
a single firm when they meet.
So let's write down
the gain from trade.
The worker has some
surplus, or gain from trade,
which is the value to having a
job minus the value to continue
to search for another job.
Symmetrically, the firm has
some surplus, which is the value
to the firm to having a
filled vacancy minus the value
to the firm of having
an unfilled vacancy.
You may notice that with free
entry, this latter term is 0
because VF subscript 0 is 0.
Nonetheless, wage determination
can be discussed independent
of free entry.
So one way to determine
it is simply
to say the worker gets a
share of the total surplus.
Let that share be theta.
It's called his bargaining
power.
That simply adds up the
worker's and the firm's surplus
and gives some fraction
to the worker.
These surpluses depend
on the Vs. The Vs depend
on the W. It's both in
the worker's problem
because W is what you
get while employed,
and in the firm's problem
because Y minus W is your
profit while currently employing
a worker.
So this equation here, which
determines how the gains
from trade are split
between a firm and a worker,
suffice to give us
one more equation,
which pins down the
equilibrium wage.
So let me label that six.
That's the last of it.
We've now come to a
complete description
of the Pissarides Model
of the Labor Market.
>> This model can be
used for lots of things.
For example, we could
ask what would happen
if unemployment insurance
went up.
So unemployment insurance's
model is a component of B,
the value to being a value
worker looking for work.
It's going to have
several effects,
but the first order effect
is it makes unemployment not
so painful.
If unemployment insurance
is higher,
you sacrifice less consumption
when you don't have a job.
This is going to influence
the worker's bargaining power.
After all, the worker's
surplus is V1 minus V0.
If V0 is higher because there's
more generous unemployment
insurance, the surplus
of the worker goes down.
He's going to demand
a higher wage to work.
Higher wages, firms are
going to see less profits,
fewer firms will try to recruit,
market tightness goes
down, unemployment goes up.
We have a fully articulated
chain going
through general equilibrium of
how a certain policy impacts
on the negotiations
between workers and firms,
how this impacts the
payoff to workers and firms,
how this affects entry,
and hence market tightness,
and hence the unemployment rate.
Now, that's just one example.
There's many other examples.
We can talk about subsidizing
recruitment by firms.
We don't have taxes in the
model, those could be put in.
What we have here is a framework
which is state of the art
in modern economic theory
of a market with frictions.
There's many uses, both
in policy discussions
and in empirical work.
Here, we simply presented
the theory.
Hope you found it interesting.
[ Music ]