Note 4: Taxation and Monopoly

John Finlay

1 Introduction
This note is intended to supplement Lecture 4. Sections 2-4 contain derivations
of the results in the lecture slides on the incidence and costs of taxes, and Sec-
tion 5 derives the optimal markup charged by a monopolist. For more examples
and graphical illustrations see the lecture slides: this note focuses on providing
slightly more mathematical detail on the results in the lecture. It is not neces-
sary for you to be able to reproduce these details but I hope that seeing it will
solidify your understanding of the results in the slides.

2 Characterising Price Responses to Taxes

In the lecture slides, we presented the idea that ‘the inelastic side of the mar-
ket pays’, or more accurately, bears the burden of a tax. Here we derive the
percentage change in producer and consumer prices in response to a tax τ and
thus prove that this statement is true.
Recall that we defined the producer price as q such that
q = (1 − τ )p
where p is the consumer price. We know that in equilibrium these prices must
satisfy
D(p∗ ) = S(q ∗ ).
That is, the amount firms are willing to supply at the producer price q ∗ must
equal the amount consumers are willing to buy at the producer price p∗ . To see
how a tax affects these prices, differentiate both sides of this expression with
respect to τ to obtain
 ∗  ∗
dp dq
D′ (p∗ ) = S ′ (q ∗ ) ,
dτ dτ
where D′ (p∗ ) is just the derivative of demand with respect to the consumer
price and S ′ (q ∗ ) the derivative of supply with respect to the producer price. In
this course we focus on expressing these derivatives using elasticities. With our
constant elasticity assumption
D(p∗ ) S(q ∗ )
D′ (p∗ ) = −δ , S ′ (q ∗ ) = σ
p∗ q∗

1
We can also use the fact that we are in an equilibrium to replace supply and
demand with equilibrium quantity,
 ∗  ∗
y y
D′ (p∗ ) = −δ , S ′ ∗
(q ) = σ .
p∗ q∗
Substituting these derivatives into the expression above gives
 ∗   ∗ 
dp 1 dq 1
−δ = σ
dτ p∗ dτ q ∗
To complete the result, differentiate the definition of producer price
dq ∗
 ∗
dp
= (1 − τ ) − p∗
dτ dτ
Let’s focus on evaluating this starting from an equilibrium with zero taxes, that
is, with τ = 0. Then by substituting into the expression above
 ∗   ∗ 
dp 1 dp 1
−δ =σ −1 ,
dτ p∗ dτ p∗
or equivalently
dp∗ 1
   
σ
= .
dτ p∗ σ+δ
A slightly more intuitive way to write this would be
dp∗
 
σ
= dτ.
p∗ σ+δ
Once we have this expression we can immediately get the change in producer
prices
dq ∗
 
δ
= − dτ.
q∗ σ+δ
To get the exact expression in the lecture slides, note that if we start from
zero taxes, then the change in τ — that is, dτ — is ∗just equal to τ . And
the percentage change in the consumer price is just dp p∗ , and likewise for the
producer price. Putting all this together,
 
σ
% change in consumer price = τ,
σ+δ
 
δ
% producer in producer price = − τ.
σ+δ

It is important to understand what these expressions mean. Consider a

20% tax. Start by thinking about an extreme case: perfectly inelastic demand,
corresponding to δ = 0. We obtain
% change in consumer price = 0.20,
% producer in producer price = 0.

2
When demand is perfectly inelastic, the price facing consumers rises one-for-
one with the tax rate and the price facing producers does not change at all.
Intuitively, perfectly inelastic demand means purchase will always consume the
pre-tax quantity, regardless of its price. So the quantity supplied in the equi-
librium must not change with the introduction of the tax. This in turn means
the price facing consumers must not change, or else they would not be willing
to supply this quantity.
Now consider the opposite extreme of perfectly elastic demand, correspond-
ing to δ = ∞. We obtain

% change in consumer price = 0,

% producer in producer price = −0.20.

When demand is perfectly elastic, consumers will buy an infinite quantity of the
good if its price drops even a little, and will buy nothing if its price increases
even a little. So the only possible equilibrium is at the original, pre-tax price.
Given that we know the price consumers must face, it follows the producer price
must absorb the entire tax and fall by 20%.
Finally, consider other values of σ and δ. We can get a particularly nice
expression by taking the ratio of the two prices
% change in consumer price σ
=− .
% change in producer price δ
When demand is more elastic than supply, δ > σ and this ratio is smaller (in
absolute value) than one. The fall in producer prices is larger than the increase
in consumer prices, and producers bear most of the tax. Just the opposite is
true if demand is more elastic than supply. As claimed, the inelastic side of the
market bears the burden of taxation.

3 Calculating Deadweight Loss

Recall from the lecture slides that deadweight loss of a tax is just the reduction in
welfare relative to the pre-tax equilibrium. From the graphs in the lecture slides,
this is just the area between the (inverse) demand and supply curves between
the pre- and post-tax quantities. Denote the pre-tax equilibrium quantity by y0
and the post-tax equilibrium quantity by y1 , so that we can write
Z y0
DW L = (PD (y) − PS (y))dy.
y1

To be really explicit here, let’s write y1 as y1 (τ ), to emphasise that the post-tax

equilibrium quantity is going to depend on the size of the tax τ . Then we can
think of DW L itself as a function of τ , writing
Z y0
DW L(τ ) = (PD (y) − PS (y))dy
y1 (τ )

3
DW L(0) = 0 because when the tax rate is zero, the deadweight loss is zero; and
as τ rises we expect deadweight loss to also rise.
Now, the difficult thing here is to get a simple expression for DW L(τ ). To
build intuition let’s think about using a first-order linear approximation around
τ = 0. Here I’m relying on the idea that any function f (x) can be approximated
quite well using
f (x) ≃ f (0) + f ′ (0)x.
It is beyond the scope of this course to prove that this result is true, but it is
extremely useful — especially for delivering neat, intuitive expressions, which is
our focus in this course. Following this line of reasoning, maybe we can write

DW L(τ ) ≃ DW L(0) + DW L′ (0)τ

where we already know that DW L(0) = 0. Unfortunately, this doesn’t work

too well, because if you try to differentiate DW L(τ ) you will find

DW L′ (τ ) = −(PD (y1 (τ )) − PS (y1 (τ ))) × y1′ (τ ).

Evaluated at τ = 0, we obtain

DW L′ (0) = −(PD (y1 (0)) − PS (y1 (0))) × y1′ (0) = 0,

because at τ = 0 we are just in the pre-tax equilibrium, and so PD (y1 (0)) =

PD (y0 ) = PS (y0 ) = PS (y1 (0)). What this tells us is interesting: to a first order
approximation, deadweight loss is zero. This is actually an interesting result:
small taxes cause very small — approximately zero, in fact — reductions in
welfare.
However, this result is not so useful for thinking about larger taxes. For
those, we will need to improve on our first-order approximation with a second-
order approximation. I will now use the idea that
 
′ 1
f (x) ≃ f (0) + f (0)x + f ′′ (0)x2
2

where f ′′ (0) is the second derivative of the function f . In our deadweight loss
case  
1
DW L(τ ) ≃ DW L′′ (0)τ 2
2
where we are using the fact that we already know DW L(0) = DW L′ (0) = 0.
The challenge now is just to get the second derivative of DW L(τ ) and evaluate
it at τ = 0.
Start with our expression for the first derivative

DW L′ (τ ) = −(PD (y1 (τ )) − PS (y1 (τ ))) × y1′ (τ ).

Use the fact that in equilibrium

PS (y1 (τ )) = (1 − τ )PD (y1 (τ )),

4
so that we can write
DW L′ (τ ) = −τ × PD (y1 (τ )) × y1′ (τ ).
We are interested in differentiating this and evaluating it at τ = 0. This might
take a bit of effort, but we can make it much easier by realising that any terms
which have a τ in front of them will vanish once we evaluate the derivative at
τ = 0. Using that observation
DW L′′ (0) = −PD (y1 (0)) × y1′ (0).
That is, we only need to know the initial price and the first derivative of output
with respect to the tax. It is easy to get the first derivative of output because
we have already worked out how prices change in Section 4 of this note. In
particular
dy ∗ dD(p∗ )
  ∗ 
dD(p∗ ) p∗
   ∗ 
dp dp 1
y1′ (0) = = × = × × D(p∗ )
dτ dp∗ dτ dp∗ D(p∗ ) dτ p∗
 
δσ
=− × y∗
σ+δ
where we have also used the fact that D(p∗ ) = y ∗ in equilibrium. Plugging this
into our expression for the second derivative, we obtain
 
′′ ′ δσ
DW L (0) = −PD (y1 (0)) × y1 (0) = p∗ y ∗
σ+δ
Finally turning to our second order approximation, we have
   
1 δσ
DW L(τ ) ≃ × × (p∗ y ∗ ) × τ 2
2 σ+δ

4 Explaining Deadweight Loss

To reiterate, our expression for deadweight loss, as a fraction of revenue, is
   
DW L 1 δσ
= × × τ2
p∗ y ∗ 2 σ+δ
This has two interesting features. First, it is increasing in the demand elasticity
and the supply elasticity. To see this, first suppose either one of these is zero.
Then, we can immediately see that deadweight loss is also zero. This is because
if either elasticity is zero, the equilibrium quantity will not change with the
introduction of a tax. Then there is no scope for missing out on potentially
beneficial transactions, and there is no deadweight loss. To gain more intuition,
let’s go to the other extreme and imagine supply is perfectly elastic so that we
can ignore it. By sending σ to infinity, we obtain
 
DW L 1
= × δ × τ2
p∗ y ∗ 2

5
We can see clearly that deadweight loss rises with δ. With perfectly elastic
supply, the price facing consumers rises by exactly τ . They respond by reducing
demand by δ × τ %. The strength of their response, and thus the extent of
deadweight loss, rises with δ. We can do exactly the same exercise with the
supply elasticity to verify that deadweight loss also becomes more severe the
more sensitive is supply.
The second interesting feature of this expression is that it depends on τ 2 . In
other words, it is quadratic in the tax rate. This implies that small taxes are
not too bad; they only create very small deadweight loss. Large taxes, however,
are very damaging. To put numbers on this, let’s imagine σ = 3 and δ = 2 and
consider a tax rate of 10%. The implied deadweight loss is
   
DW L 1 3+2
= × = 0.102 ≃ 0.004 = 0.4%,
p∗ y ∗ 2 3×2
which is not so bad. Now suppose the tax rate rises to 50%. Deadweight loss
rises to    
DW L 1 3+2
= × = 0.502 ≃ 0.10 = 10%.
p∗ y ∗ 2 3×2
Although we have only increased the tax rate by 5x, deadweight loss has risen
by 25x!

5 Optimal Markups in Monopoly

In the lecture slides we claimed that the optimal pricing policy for a monopolist
facing constant elasticity demand is to charge a constant markup µ over marginal
cost, where  
δ
µ= .
δ−1
We now prove that this claim is true. The problem facing the monopolist is
max PD (y)y − c(y),
y

where PD (y) is the inverse demand curve facing the monopolist and c(y) is the
monopolist’s cost function. As we have seen repeatedly, at the optimal choice
y ∗ the derivative of this expression must be zero. Therefore
′
(y ∗ )y ∗
 
∗ ′ ∗ ∗ ′ ∗ ∗ PD
0 = PD (y ) + PD (y )y − c (y ) =⇒ PD (y ) 1 + = c′ (y ∗ ).
PD (y ∗ )
Now, we just need a convenient expression for the derivative of inverse demand
PD (y ∗ ). By definition
D(PD (y)) = y
where D is the demand curve. Differentiate both sides of this with respect to y
to obtain
1
D′ (PD (y))PD
′
(y) = 1 =⇒ PD ′
(y) = ′ ,
D (PD (y))

6
or in words, the derivative of the inverse demand curve is always one over the
derivative of the demand curve. Substituting this into our expression for the
optimal choice y ∗ , we find
′
(y ∗ )y ∗ y∗
   
′ ∗ ∗ PD ∗
c (y ) = PD (y ) 1 + = PD (y ) 1 + ′
PD (y ∗ ) D (PD (y ∗ )PD (y ∗ )
∗ ∗ −1


= PD (y ) 1 − ϵD (PD (y )) ,

where the second equality follows from the definition of the demand elasticity.
Now in the constant elasticity case we know this elasticity is just δ, so

PD (y ∗ )
   
′ ∗ ∗ −1

1 δ
c (y ) = PD (y ) 1 − δ =⇒ ′ ∗ = = ,
c (y ) 1 − δ −1 δ−1

and by definition the ratio of price PD (y ∗ ) to marginal cost c′ (y ∗ ) is just the

markup µ. So we have the desired result.