Soc Choice Welf (2013) 41:579–604

DOI 10.1007/s00355-012-0694-y

ORIGINAL PAPER

Production efficiency and constraints on profit taxation

and profit distribution in economies with Ramsey
taxation

Sushama Murty

Received: 10 August 2011 / Accepted: 22 August 2012 / Published online: 29 September 2012
© Springer-Verlag Berlin Heidelberg 2012

Abstract In economies with Ramsey taxation, decreasing returns to scale, and pri-
vate ownership, we show that second-best production efficiency is desirable when the
grouping of private firms induced by the profit taxation power of the government is
at least as fine as the grouping of firms induced by the institutional rules of profit
distribution in the economy. The classic results of Dasgupta and Stiglitz (Rev Econ
Stud 39:87–103, 1972) (of firm-specific profit taxation) and Diamond and Mirrlees
(Am Econ Rev 61:8–27, 1971a; Am Econ Rev 61:261–278, 1971b) and Guesnerie
(A contribution to the pure theory of taxation, 1995) (of uniform 100 % profit tax-
ation) follow as special cases of our model. Moreover, second-best analysis shows
that optimal profit taxation is a substitute for optimal intermediate input taxation. In
smooth economies, proportional, lump-sum, and affine modes of profit taxation are
equivalent. We rework Mirrlees (Rev Econ Stud 39:105–111, 1972) counterexample,
which is posed in the context of a non-smooth economy, to show that second-best
production efficiency continues to remain desirable under an affine structure of profit
taxation.

1 Introduction

Diamond and Mirrlees (1971a,b) revisited the problem first posed by Ramsey (1927)
about alternative policy instruments that can be employed when there are informational
constraints on the implementation of the second-welfare theorem. Using a general
equilibrium model they showed that, when personalized lump-sum transfers are not
available to the government as redistributive devices, commodity (Ramsey) taxation
can be employed as an alternative, albeit second-best, means of redistribution. They

S. Murty (B)
Department of Economics, University of Exeter Business School, University of Exeter,
Streatham Court, Rennes Drive, Exeter EX4 4PU, Devon, UK
e-mail: s.murty@exeter.ac.uk

123
580 S. Murty

showed that production efficiency was desired by the second-best optimal commodity
tax system and that taxation of inter-firm transactions was not required.
The general equilibrium model they employed was one where technologies of firms
exhibited constant returns to scale. Thus, consumers received no profit incomes. An
extension of this model to decreasing returns to scale technologies (see e.g., Guesnerie
1995 and Weymark 1979) led to similar results regarding production efficiency when
private firms were subject to 100 % profit taxation with the proceeds going back to
consumers as a uniform lump-sum transfer (also called a demogrant).1
To check the robustness of the result on second-best production efficiency under
more general settings, the DM model was extended to allow consumers to receive profit
incomes in proportion to the shares that they owned in the private firms.2 Assuming
smooth technologies and preferences and employing calculus methods, Dasgupta and
Stiglitz (1972) concluded that second-best production efficiency continues to be desir-
able in such economies if the government can implement firm-specific profit taxes.
Mirrlees (1972) and Hahn (1973) work towards understanding the intuition behind DS
result by employing non calculus techniques in more general economies. There are
some technical problems with the proofs of this result in these papers, which have been
highlighted by papers such as Sadka (1977) and Munk (1980) and, more recently, by
Reinhorn (2011). Moreover, some scepticism about the DS result itself is expressed
by Mirrlees, who provides a counterexample in the context of a non-smooth economy
with an aim to demonstrate that production inefficiency may actually be desirable at a
second-best.
In general, the power of the government to implement a firm-specific system of
profit taxation is questionable. More realistic scenarios, one could argue, may be those
where the number of profit tax rates the government can peg may be smaller than the
number of private firms, i.e., profit tax rates vary only across, possibly, non-singleton
groups of private firms. Further, the rigid institutional rules by which profit incomes
are distributed among consumers may be more general than the ones considered in the
literature cited above and may also define another grouping (or a partition) of private
firms.
In this paper, we distinguish between the profit taxation power of the government
(the profit system) and the institutional rules of distribution of profits as incomes to con-
sumers in the economy. We show that the desirability of second-best production effi-
ciency depends on whether or not these two institutional constraints are well-aligned:
in particular, we show that second-best production efficiency remains desirable when
the grouping of private firms induced by the profit taxation power of the government
is at least as fine as the grouping of firms induced by the institutional distribution of
profits to consumers. We show that all the previous results on second-best production
efficiency follow as special cases of our result. On one extreme is the DS case, where
the government has full profit taxation power which induces a finest partition of the set
of private firms (each firm-group in this partition comprises of a single firm and hence
the case of firm-specific profit taxation), so that second-best production efficiency is

1 For an excellent exposition of these results see Myles (1995).

2 As in a Arrow–Debreu private-ownership economy.

123
Production efficiency and constraints on profit taxation and profit distribution 581

true no matter the partition of private sector firms induced by the rules of profit distri-
bution to consumers in the economy. On the other extreme is the DM and Guesnerie
(1995) case, where the profit taxation power of the government induces the coarsest
partition of the set of private firms (only one firm-group that contains all the firms
and government implementing a single (uniform) rate of profit taxation). This implies
that second-best production efficiency is true if the rules of profit distribution in the
economy also induce the coarsest partition of the private sector firms. Then there are
cases that lie in between these two extremes.
Our strategy of proof is different from the earlier papers. We first formalise the
profit distribution rules in the economy. Production inefficiency at a tax equilibrium
implies that the aggregate supply of the firms is not on the frontier of the aggre-
gate technology of the economy. With profit maximizing private and public-sector
firms, this implies that the producer price vectors faced by private and public firms
in such a tax equilibrium are not proportional, implying that there are differences in
the marginal rates of technical substitution across firms, i.e., there is allocative ineffi-
ciency in production. Hence, an increase in aggregate supply is technically feasible by
reallocating production across firms. In fact, at such a tax equilibrium, we construct
changes in the price vectors faced by private and public firms which ensure that an
increase in aggregate supply is also supported as profit maximizing choices of firms.
Ceteris paribus, such an increase in aggregate supply implies an increase in the aggre-
gate income in the economy. The question is can this increased aggregate income be
distributed to consumers in a manner that respects the existing institutional rules of
income distribution in the economy and improves the welfare of all consumers? In
general, the institutional rules of (profit) income distribution and the restrictions on
the power of the government to implement profit taxation can constrain severely such
welfare improvements. However, we show that, if these two institutional constraints
are aligned in a manner described above, then the increased aggregate income could
be distributed to consumers in a welfare improving manner.
Our analysis clarifies the link between intermediate input taxation and proportional
profit taxation. In our proof, we take recourse to an intermediate construct of an econ-
omy with producer price vectors varying across (groups of) private-sector firms. This
economy exhibits second-best production efficiency, albeit, unlike in the DM model,
this means that firms are subject to proportional rates of intermediate-input taxation.
It is shown that the second-best allocations of this economy are the same as those
of an economy where all private firms face the same price vector but are subject to
proportional profit taxes. This implies that optimal profit taxes are perfect substitutes
for optimal intermediate input taxes.
Almost all the papers consider the case of proportional profit taxation. Our main
proof of second-best production efficiency assumes smooth economies with propor-
tional profit taxation. However, profit taxation can also be implemented in a lump-sum
or affine manner. In the case of smooth economies these different modes of profit tax-
ation are equivalent. In non-smooth economies, this may not be the case. We show this
by reworking the Mirrlees counterexample, which is posed in the context of a non-
smooth economy. We show that, even in this case, with firm-specific producer prices,
second-best production efficiency continues to remain desirable if profit taxation can
be implemented in a lump-sum manner. However, optimal proportional profit taxation

123
582 S. Murty

is no longer a perfect substitute for optimal intermediate input taxation, rather optimal
affine profit taxation is. The intuition underlying our calculations is the same as in our
strategy of proof in the smooth case.
The results on second-best production efficiency and intermediate-input taxation
are important for three reasons: Second-best production efficiency supports the use of
producer prices as the right proxies for the generally unobservable shadow prices for
the cost-benefit evaluation of public sector projects.3 The recommendation of either
proportionate intermediate-goods taxation or profit taxation with no intermediate-
goods taxation minimizes also the practical administrative costs of implementing a
system of Ramsey taxes, e.g., it supports tax structures like VAT. Our analysis also
suggests how, by understanding the rules of profit income distribution in the economy,
the government can potentially design a system of profit taxation that can further both
its redistributive and efficiency objectives.
In Sect. 2, we lay out our general equilibrium model. We discuss constraints on
government’s profit taxation power and profit distribution in the economy and define
profit-making economies with firm-group specific profit taxes and firm group spe-
cific prices. In Sect. 3 we prove a preliminary lemma. In Sect. 4, we employ this
lemma to prove Theorem 1 that states that second-best production efficiency is desir-
able in smooth profit-making economies with firm-group specific prices. In Sect. 5,
we obtain two corollaries of Theorem 1: the desirability of second-best production
efficiency in economies with firm-group specific profit taxes and the substitutability
between optimal profit taxation and optimal intermediate input taxation. In Sect. 6, we
rework Mirrlees’ counterexample in a non-smooth economy to show that second-best
production efficiency continues to hold under affine profit taxation. We conclude in
Sect. 7.

2 The model

There are N commodities indexed by k, H consumers indexed by h, and I + 1 firms

indexed by i. We denote the index set of consumers as H = {1, . . . , H } and the index
set of firms as I = {0, . . . , I }. We assume that firm 0 is a public sector firm, while all
others are private firms.
For every firm i∈ I, the technology is denoted by Y i ⊂ R N . The aggregate
technology is Y = i Y i .
For all h ∈ H, the gross consumption set of consumer h is X h ⊆ R+ N and the
h
preferences over X are represented by continuous, quasi-concave, and locally non-
satiated utility functions u h : X h → R with images u h (x h ). The endowment vector of
consumer h is denoted by eh ∈ R+ N.

 For all i ∈ I define the set B of all prices

We assume that all firms are price-takers. i

p ∈ R+ N \ {0 N } such that the set { p · y  y ∈ Y i } is bounded above. Define the profit

function π i : B i → R with image4

3 See Little and Mirrlees (1974), Boadway (1972), and Drèze and Stern (1987).
4 Under Assumption 1 in Sect. 3, the technology set Y i is closed for all firms i = 0, . . . , I , and hence the
supremum can be replaced by a maximum.

123
Production efficiency and constraints on profit taxation and profit distribution 583

π i ( p) = sup{ p · y i |y i ∈ Y i }. (2.1)
yi

The supply vectors of firm i ∈ I are obtained from the solution mapping of (2.1) as
y i : B i → R N with image y i ( p).5 For all i ∈ I define the frontier of Y i as6

Ŷ i = {y i ∈ Y i | ȳ i  y i implies that ȳ i ∈
/ Y i }. (2.2)

Similarly, we can define Ŷ as the frontier of the aggregate technology Y .

The vector of consumer prices is denoted by q ∈ R+ N . Income of every consumer

h ∈ H is denoted by m ∈ R+ . Let the mapping x : R+

h h N × R → R N with image
+
x (q, m ) denote the Marshallian demand vector of consumer h and let the mapping
h h

V h : R+N × R → R with image V h (q, m h ) be the corresponding indirect utility func-

+
tion. Every consumer receives a demogrant, which is denoted by m ∈ R. The income
of consumer h ∈ H comprises of his endowment income, profit income that he receives
from the private firms, and the demogrant that he receives from the government.

2.1 Institutional constraints on distribution of profits to consumers

There are institutional constraints on how profits of private sector firms are distributed
to consumers. The pretax profit incomes of consumers are, in general, functions of the
distribution of profits of the I private sector firms. Suppose they are given by the map
r : R+I −→ R H with image
+

⎡ 1 1
⎤
  r π ,...,πI
⎢ .. ⎥
r π 1, . . . , π I ≡ ⎣ . ⎦. (2.3)
1

r π ,...,π
H I

We assume that the map r is such that

 1
I
h∈H r π , .
. . , π
I =
i=1 π ,
(i) h i

(ii) r π , . . . , π ≥ 0 ∀h ∈ H, and
h 1 I

(iii) it is homogeneous of degree one in its arguments.

A partition of I \ {0} (the index set of all private sector firms) is denoted by
P = {P1 , . . . , PT }. Members of a partition will be indexed by t or l. Given a partition

5 In general, y i ( p) need not be a singleton set, i.e., the solution mapping y i ( p) need not be a function.
6 Vector notation: for any two vectors a =
a , . . . , a  and b =
b , . . . , b  in an arbitrary Euclidean
1 n 1 n
space Rn ,

a ≥ b ⇐⇒ ai ≥ bi ∀i = 1, . . . , n,
a > b ⇐⇒ ai ≥ bi ∀i = 1, . . . , n with a  = b, and
a  b ⇐⇒ ai > bi ∀i = 1, . . . , n.

123
584 S. Murty

P = {P1 , . . . , PT } of I \{0}, we will often refer to maps of the form rP : R+

T −→ R H
+
with image:
⎡ 1 1
⎤
  rP a , . . . , a T
⎢ .. ⎥
rP a 1 , . . . , a T ≡ ⎣ . ⎦ (2.4)
1

rP a , . . . , a
H T

Let P = {P1 , . . . , PT } be the coarsest partition of I \ {0} such that there exists a
map rP : R+
T −→ R H whose image is
+
⎛ ⎞
   
rP ⎝ πi , . . . , π i ⎠ = r π 1, . . . , π I . (2.5)
i∈P1 i∈PT

P defines the minimal number of firm-groups in the economy such that the profit
incomes of consumers in the economy depend on the distribution of profits across
these groups. This implies that the profit incomes in the economy are also functions
of profits of any other set of firm-groups, where the groups are finer than in P , i.e.,
for any partition P = {P1 , . . . , PT } of I \ {0} that is a refinement of P , there exists a
map rP : R+ T −→ R H whose image is7
+
⎛ ⎞
 

rP ⎝ πi, . . . , π i ⎠ = r π 1, . . . , π I . (2.6)
i∈P1 i∈PT

P , T, r )) denote the vector of economic fun-

Let E = ((X h , u h , eh )h∈H , (Y i )i∈I , (P
damentals specified above.

2.2 Constraints on profit taxation power of government: firm-group specific profit

taxes

Suppose the government implements T profit tax rates on the I private sector firms.
This means that there exists a partition, say P = {P1 , . . . , PT }, of I \ {0} such that
all firms in each firm-group t = 1, . . . , T are subject to the same profit tax rate. In
that case we say that the government has a profit taxation power equal to T with
P = {P1 , . . . , PT } being the partition corresponding to T . Let Eπ (E, T, P) denote
a profit-making economy derived from economy E with profit taxation power of the
government equal to T and P being the partition corresponding to T .8
The firm-group specific profit taxation can be implemented in proportion to firm-
group profits or in a lump-sum manner or some combination of both:

7 A partition P of I \ {0} is a refinement of P if I ≥ T ≥ T and for all t = 1, . . . , T , we have P ⊆ P

t l
for some l = 1, . . . , T.
8 Note that, in general, it is possible that T < T or P is not a refinement of P . However, this paper shows
that second-best production efficiency is true when P is a refinement of P .

123
Production efficiency and constraints on profit taxation and profit distribution 585

• firm-group specific proportionate profit taxation involves a vector of firm-group

specific profit tax rates τ =
τ1 , . . . , τ T  ∈ R T such that the net of tax profit of
firm-group t is a t = (1 − τ t ) i∈Pt π i .
• firm-group specific lump-sum profit taxation involves a vector of firm-group spe-
 γ =
γ
cific lump-sum profit taxes 1 , . . . , γ T  ∈ R T such that the net of tax profit

of firm-group t is a = i∈Pt [π − γ t ].
t i

• firm-group specific affine profit taxation involves a vector of firm-group specific

proportional profit tax rates τ =
τ 1 , . . . , τ T  ∈ R T and a vector of firm-group
specific lump-sum profit taxesγ =
γ 1 , . . . , γ T  ∈ R T such that the net of tax
profit of firm-group t is a t = i∈Pt [(1 − τ t )π i − γ t ].
Definition A tax equilibrium with proportionate profit taxes of a profit-making econ-
omy Eπ (E, T, P) where P is a refinement of P is a configuration
q, p 0 , p, τ 1 , . . . , τ T ,
N × B 0 × ∩T B × R T +H +1 such that9
m, m 1 , . . . , m H  ∈ R+ t=1 t

 
T  
x h (q, m h ) ≤ y i ( p) + y 0 ( p 0 ) + eh and
h∈H t=1 i∈Pt h∈H

m h = rP
h
((1 − τ 1 ) π i ( p), . . . , (1 − τ T )
i∈P1

× π ( p)) + m + qeh , ∀h ∈ H.
i
(2.7)
i∈PT

Similarly, we can define tax equilibria of a profit making economy Eπ (E, T, P)

with lump-sum and affine profit taxation.10

2.3 Some examples

Consider the standard case in the literature where economy E is such that profits
are distributed to consumers in proportion to the shares that they hold in the private
sector firms. The profit shares of consumers are summarised by an exogenous H × I -
dimensional share-matrix , where the typical element θih ≥ 0 denotes the share of

consumer h in the profit of the private firm i. Thus, we have h θih = 1 for all i ∈ I.
Let O denote the set of matrices  with these properties.11
 
Example 1  ∈ O is such that P = {1}, . . . , {I } , i.e., P is the finest partition of
the set of all private firms. This implies T = I . DS study this case of distribution
of profits to consumers. They show that second-best production efficiency holds in

9 For t = 1, . . . , T , B denotes ∩ i
t i∈Pt B .
10 Recall that the income m h of each consumer h comprises of his profit income, endowment income, and
the demogrant received from the government. A tax equilibrium is tight if all the market clearing conditions
in (2.7) hold as equalities. It is non-tight otherwise. If the supply mapping y i ( p) is a correspondence for
any i = 0, . . . , I , then (2.7) needs to be modified by replacing y i ( p) with y i ∈ y i ( p).
11 The examples below are stated in terms of proportionate profit taxation. But they can be extended to
lump-sum and affine profit taxation schemes as well.

123
586 S. Murty

this economy when the profit taxation power of the government is at its maximum I ,
i.e., the partition associated with profit taxation power of the government is exactly
equal to P . In that case, the map of post-tax profit incomes with firm-group specific
proportionate profit taxation is given by

r h = rPh ((1 − τ 1 )π 1 ( p), . . . , (1 − τ I )π I ( p))

= θ1h (1 − τ 1 )π 1 ( p) + · · · + θ Ih (1 − τ I )π I ( p), ∀h ∈ H. (2.8)

Example 2 If  ∈ O is such that θih = θ h for all h ∈ H and i ∈ I \ {0}, then

P = {I \ {0}} and T = 1, i.e., P is a partition with only one firm group containing all
the firms. We claim that second-best production efficiency is true in this economy if
the government’s profit taxation power is at least one, i.e., T ≥ 1. If T = 1, the map
of post-tax profit incomes with proportionate profit taxation is given by
 
r h = rPh ((1 − τ ) π i ( p)) = θ h (1 − τ ) π i ( p), ∀h ∈ H. (2.9)
i∈I \{0} i∈I \{0}

A special case arises when θ h = H1 with τ = 0. This is equivalent to the case studied
in Guesnerie (1995) and Weymark (1979), where the profits of all firms are subject
to a uniform (100 %) profit tax rate (τ = 1) and the government returns its profit tax
revenue as a uniform lump-sum transfer to all consumers.12
Example 3 Suppose the share matrix  ∈ O implies P and T are such that 1 < T < I .
Then θih = θ ht for all h ∈ H, t = 1, . . . , T, and i ∈ Pt . We claim that second-best
production efficiency is true in this economy if the government’s profit taxation power
is at least T. If it is T, then the map of post-tax profit incomes with firm-group specific
proportionate profit taxes is given by
⎛ ⎞
 
r h = rPh ⎝ (1 − τ 1 )π i , . . . , (1 − τ T )π i ⎠
i∈P1 i∈PT
 
= θ (1 − τ )
h1 1
π + . . . + θ hT (1 − τ T )
i
π i , ∀h ∈ H. (2.10)
i∈P1 i∈PT

If government’s taxation power is less than T, then second-best production efficiency

may not hold.
The admissible profit distribution rules need not be confined to just linear func-
tions of firm-group profits. Theoretically, more general and complex non-linear rules,
which are homogeneous of degree one in profits and ensure that the sum of pretax
profit incomes of consumers is the sum of profits in the economy, can be constructed.
One such example, that is based on the Euclidean norm of the vector of profits, is
provided below. However, specific examples of such rules in the real world are hard

12 This is also trivially the DM case where constant returns to scale is assumed, so that profits of firms are
zero.

123
Production efficiency and constraints on profit taxation and profit distribution 587

to find. Usually, in the real world, dividends are paid as a proportion of profits (cash
flows), the proportion being determined by the number of shares held (equivalently
the amount of stock owned). Hence, typically, what we observe are profit income
distribution rules which are linear in (firm-group) profits.
 
Example 4 P = {1}, . . . , {I } (i.e., T = I ), the government can do firm-specific
profit taxation, and the post-tax profit incomes are given by

1 
I 2
π i ( p)
r =h
(1 − τ i ) ∀h = 1, . . . H − 1, and
H −1  π( p) 
i=1

I 
I
π i ( p)
2
r H
= (1 − τ )π ( p) −
i i
(1 − τ ) i
, (2.11)
 π( p) 
i=1 i=1


I i ( p)2 .13
where π( p) =
π 1 ( p), . . . , π I ( p) and  π( p) = i=1 π

2.4 Second-best production efficiency with profit taxation: a definition

A tax equilibrium
q, p, p 0 , τ 1 , . . . , τ T , m, m 1 , . . . , m H  with proportionate profit
taxes of Eπ (E, T, P), where P is a refinement of P , is production efficient if
 I
i=1 y ( p) + y ( p ) ∈ Ŷ . The second-best problem for Eπ (E, T, P) is to find the
i 0 0

mapping VP ,π :  H −1 → R with image14


VP ,π (s 1 , . . . , s H ) := max sh V h (q, m h )
q, p, p 0 ,
τ 1 ,...,τ T ,m,m 1 ,...,m H h
subject to (2.12)

q, p, p 0 , τ 1 , . . . , τ T , m, m 1 , . . . , m H  satisfying (2.7).

A solution to (2.12) is called a second-best tax equilibrium of Eπ (E, T, P).

 
∗ ∗ ∗ 0 ∗1 ∗1 ∗ ∗ 1 ∗H
Definition A second-best tax equilibrium q, p, p , τ , . . . , τ , m, m , . . . , m
with proportionate profit taxes of the profit-making economy Eπ (E, T, P), where
P is a refinement of P , is a production efficient second-best if it is production effi-
cient. If all such second-best tax equilibria of Eπ (E, T, P) are production efficient,
then Eπ (E, T, P) exhibits second-best production efficiency with proportionate profit
taxation or production efficiency is desirable at the second-best of Eπ (E, T, P) with
proportionate profit taxation.15

13 We restrict these mappings to the case where  π( p)  = 0.

14  H 1 H
H −1 is the H − 1-dimensional unit simplex in R .
s , . . . , s  ∈  H −1 can be interpreted as
a vector of welfare weights attached to consumer utilities. The second-best utility possibility frontier is
obtained by solving the second-best optimization for all possible vectors of welfare weights.
15 Similarly, we can define second-best production efficiency with lump-sum or affine profit taxation.

123
588 S. Murty

2.5 A profit-making economy with firm-group specific prices

In this paper, we show that production efficiency is desirable at the second-best when
the government’s profit taxation power is at least T with the corresponding partition
being a refinement of P . To do so, we consider a more general institutional structure
than a profit-making economy with firm-group specific profit taxes. Suppose the gov-
ernment can control private sector firms by imposing on them T vectors of producer
prices, i.e., there exists a partition, say P = {P1 , . . . , PT }, of I \ {0} such that all
firms in each firm-group t = 1, . . . , T are subject to the same producer price vector.
In that case we say that the government has the power to implement T firm-group
specific prices with P = {P1 , . . . , PT } being the partition corresponding to T . Let
E p (E, T, P) denote a profit-making economy derived from economy E with govern-
ment having the power to implement T firm-group specific prices and P being the
partition corresponding to T . We also consider the case where firm-group specific
prices are accompanied by firm-group specific lump-sum profit taxes:
• if
p 1 , . . .
, p T  is a tuple of firm-group specific prices, then the profit of firm-group
t is a = i∈Pt π i ( p t ).
t

• if
p 1 , . . . , p T  is a tuple of firm-group specific prices and the government also
has a lump-sum profit taxation power equal to T and γ =
γ 1 , . . . , γ T  ∈ R T is
a vector of firm-group specific lump-sum profit taxes, then the net of tax profit of
firm-group t is a t = i∈Pt [π i ( p t ) − γ t ].
Definition A tax equilibrium of economy E p (E, T, P), where P is a refinement of P ,
N ×B 0 ×T B ×R H +1
is a configuration
q, p 0 , p 1 , . . . , p T , m, m 1 , . . . , m H  ∈ R+ t=1 t +
such that16

 
T  
x h (q, m h ) ≤ y i ( pt ) + y 0 ( p0 ) + eh and
h∈H t=1 i∈Pt h∈H
⎛ ⎞
 
h ⎝
m h = rP π i ( p 1 ), . . . , π i ( p T )⎠ + m + qeh , ∀h ∈ H.
i∈P1 i∈PT
(2.13)

As in the previous section we can define a production efficient tax equilibrium of

economy E p (E, T, P), where P is a refinement of P . Note that the system of Eq. (2.13)
is homogeneous of degree zero in p 1 , . . . , p T , q, and m and is homogeneous of degree
zero in p 0 . Hence, it admits two normalization rules, e.g., we can adopt the normal-
ization rules p11 = 1 and p10 = 1. Under the maintained assumptions on consumers’
preferences, the budget constraints hold as equalities under utility maximization, i.e.,
for all h, we have

q · x h (q, m h ) = m h . (2.14)

16 Similarly too we can define tax equilibria when the government has a power to implement both firm-group
specific prices and firm-group specific lump-sum profit taxes.

123
Production efficiency and constraints on profit taxation and profit distribution 589

To show that at a tax equilibrium of E p (E, T, P), the government budget is bal-
anced, we multiply both sides of the first part (a vector of inequalities) of (2.13) by q
and employ (2.14) to obtain

 
T  
q· x h (q, m h ) ≤ q · y i ( pt ) + q · y 0 ( p0 ) + q · eh
h t=1 i∈Pt h∈H


T  
⇒ π i ( pt ) + H m + q · eh ≤
t=1 i∈Pt h∈H


T  
T  
pt · y i ( pt ) + q · y 0 ( p0 ) + [q − p t ] · y i ( p t ) + q · eh
t=1 i∈Pt t=1 i∈Pt h∈H


T 
⇒ Hm ≤ [q − p t ] · y i ( p t ) + q · y 0 ( p 0 ). (2.15)
t=1 i∈Pt

Condition (2.15), which is an implication of Walras law, says that the demogrant is
financed from the government’s revenue from indirect taxation and the sale of publicly
produced goods to the consumers.
The second-best problem for E p (E, T, P) is to find the mapping VP , p :  H −1 → R
with image


VP , p (s 1 , . . . , s H ) := max h sh V
h (q, m h )
q, p 0 ,
p 1 ,..., p T ,m,m 1 ,...,m H

subject to (2.16)

q, p 0 , p 1 , . . . , p T , m, m 1 , . . . , m H  satisfying (2.13).

A solution to (2.16) is called a second-best tax equilibrium of E p (E, T, P). As

in the previous section we can define a production efficient second-best tax equilib-
rium of E p (E, T, P) and the desirability of production efficiency at the second-best
of E p (E, T, P).

3 A preliminary lemma

Assumptions 1 and 2 stated below are regularity assumptions on the technologies of

firms. They are similar to the ones employed in the previous literature on this topic.

Assumption 1 For all i ∈ I, Y i is closed, convex, contains 0, and satisfies Y i +R−

N ⊂
i
Y .

Assumption 2 For all i ∈ I, the set B i is non-empty and there exists ρ ∈ R++
N ∩ Bi .

123
590 S. Murty

Note that under Assumptions 1 and 2, for all i, π i is continuous, non-negative val-
ued, linearly homogeneous, and convex on the set B i . We first restrict our analysis to
the case of smooth and strictly convex technologies.17

Assumption 3 For all i ∈ I, there exist smooth and strictly quasi-convex functions
f i : R N → R with images f i (y) and ∇ f i (y) = 0 N such that Y i = {y ∈ R N | f i (y) ≤
0}.18

Suppose the government has the power to implement T firm-group specific prices
and P is the partition of I \ {0} that corresponds to T . The corresponding economy
is E p (E, T, P). For all t = 1, . . .
, T and p t ∈ Bt , we denote the supply of firms in
Pt ∈ P as y ( p ), i.e., y ( p ) = i∈Pt y i ( p t ). For all t = 1, . . . , T , the frontier of
t t t t
 T
Y t = i∈Pt Y i is denoted by Ŷ t .19 Let
p 1 , . . . , p T  ∈ t=1 Bt and p 0 ∈ B 0 .

Remark 1 Let Assumptions 1, 2, and 3 hold and let P = {P1 , . . . , PT } be a partition

of I \ {0}. Suppose p t ∈ Bt for all t =  1, . . . , T and p 0 ∈ B 0 . p 0 , p 1 , . . . , p T are
T
proportional to each other if and only if t=1 y t ( p t ) + y 0 ( p 0 ) ∈ Ŷ .20 This implies
 
that if
there exist t, t = 0, 1, . . . , T such that p t and p t are not proportional then
T
y = t=1 y t ( pt ) + y 0 ( p0 ) ∈
/ Ŷ (i.e., y is in the interior of Y ).

Lemma 1 shows that if there exist t, t  ∈ {0, 1, . . . , T } such that p t and p t are not
proportional,21 then there exist changes in price vectors faced by firms in groups t and
t  thatcan strictly increase the aggregate supply (i.e., lead to a higher aggregate supply
T
than l=1 y l ( pl ) + y 0 ( p 0 )) given the profit maximizing behavior of our price-taking
firms. Though Remark 1 is well-known, the construction of price vectors in Lemma 1
that induce increase in aggregate supply in a profit maximizing way is novel.22 The
proof of Lemma 1 is relegated to the Appendix.

Lemma 1 Suppose Assumptions 1, 2, and 3 hold. Suppose the government has the
power to implement T firm-group specific prices and P is the partition of I \ {0} that
corresponds to T . Suppose p̄ t ∈ Bt for all t = 1, . . . , T and p̄ 0 ∈ B 0 . Suppose there

17 The restriction to strictly convex technologies implies that the supply mappings of firms are functions.
18 A function f : R N → R is smooth if it is C ∞ , i.e., its partial derivatives of all orders exist. Note that
under 1 and 2, the set Y i has a functional representation for all i ∈ I. Assumption 3 only ensures that this
functional representation is smooth and strictly quasi-convex.
19 Note the slight abuse of notation: the technology, its frontier, and a production vector corresponding to
any firm i ∈ I are denoted by Y i , Ŷ i , and y i , respectively, while the aggregate technology, its frontier, and
a production vector obtained by summing over all firms in Pt for t = 1, . . . , T are denoted by Y t , Ŷ t , and
y t , respectively. However, in what follows, it will be clear always whether we are referring to a firm in I
or to a group of firms Pt .
20 This follows from Koopmans’ well-known result on interchangeability of set summation and optimiza-
tion. In the non-smooth case, this generalizes to the sets of support prices of y 0 ( p 0 ), y 1 ( p 1 ), . . . , y T ( p T )
having a non-empty intersection.
21 Or, in the non-smooth case, if the sets of support prices of y t ( p t ) and y t  ( p t  ) do not intersect.
22 Note Lemma 1 solves a problem in Hahn (1973) proof. Hahn’s proof aims to construct a sequence of
Pareto-superior equilibria that converge to a given production inefficient tax equilibrium. However, the
proof does not ensure that these allocations can be supported as profit maximizing choices of private firms,
i.e., individual production vectors of firms along this sequence may lie in the interior of their technologies.

123
Production efficiency and constraints on profit taxation and profit distribution 591

Fig. 1

y2 p1

y
H ( p1,0) p2

y1
H ( p 2 ,0)
y = y1 + y 2 >> 0

 T
exist t, t  ∈ {0, 1, . . . , T } such that p̄ t is not proportional to p̄ t . Let ȳ := l=0 y l ( p̄l ).
v  v 
Then there exist sequences { p } → p̄ and { p } → p̄ and an integer v̂ such that
t t t t
 t tv t t v
l=t,t  y ( p̄ ) + y ( p ) + y ( p )  ȳ for all v > v̂.
l l

We now present an example to illustrate the point made in this lemma. Exam-
ple 5 (below) considers the case of smooth production frontiers.
 It is assumed that
N = 2, there is no public production, I = 2, and P = {1}, {2} . Good two is the
output and good one is the input of these firms, so that if y =
y1 , y2  ∈ R2 is a
production vector, then y2 ≥ 0 and y1 ≤ 0. Let us also denote a hyperplane with
normal p and constant a by H ( p, a) and its lower and strictly lower half-spaces by
H≤ ( p, a) and H< ( p, a), respectively. Similarly we can define the upper and strictly
upper half-spaces of H ( p, a).
1
Example 5 Suppose technology of Firm 1 is Y 1 = {y 1 ∈ R2 | y21 ≤ (−y11 ) 2 } and
1
Y 2 = {y 2 ∈ R2 | y22 ≤ (−y12 ) 3 }. Suppose Firms 1 and 2 face price vectors p 1 =
1, 21 
and p2 =
1, 12 1
, respectively. It can be verified that the profit maximizing pro-
duction vector of Firm 1 will be y 1 ( p 1 ) =
−1, 1, while that of Firm 2 will be
y 2 ( p 2 ) =
−8, 2. Since p 1 and p 2 are not proportional, it is well-known (and can be
easily verified) that the aggregate production vector y := y 1 ( p 1 ) + y 2 ( p 2 ) =
−9, 3
must lie in the interior of Y := Y 1 + Y 2 . In fact, at these production vectors, the
marginal productivity of input in Firm 1 is 2, which is less than the marginal pro-
ductivity of input equal to 12 in Firm 2. This suggests that reallocating some of the
input from Firm 1 to Firm 2 will result in an increase in the aggregate output. The
question is whether there exist such reallocations which can also be supported as profit
maximization choices of firms. We show below that this is true.

Since p 1 and p 2 are not collinear, Fig. 1 shows that there exist changes y 1 =

21 , −2 and y 2 =
− 41 , 25  in production vectors of Firms 1 and 2, respectively, such
that p 1 · y 1 and p 2 · y 2 are non-positive and the aggregate change in production

123
592 S. Murty

Fig. 2
y 22
2
y v1
y2 p2 v2
p2

Y2
p2

y 2 ( p2 )

y12

is positive, i.e., y = y 1 + y 2 =
41 , 21   0. In particular, as seen in the figure,
y 1 ∈ H≤ ( p 1 , 0) ∩ H> ( p 2 , 0). Similarly y 2 ∈ H≤ ( p 2 , 0) ∩ H> ( p 1 , 0).23
Note however, that such changes may not be technologically feasible, e.g., for Firm
1
2, y 2 ( p 2 ) + y 2 =
−8.25, 4.5 and 4.5 > (8.25) 3 . Nevertheless, Fig. 2 shows that
y 2 can be suitably scaled so that it becomes technologically feasible with respect to
Y 2 , e.g., ȳ 2 := y 2 ( p 2 ) + 3.32347
1
y 2 =
−2.482, 1.354 lies in Ŷ 2 . By scaling down
y further, we obtain points that lie on the line segment joining ȳ 2 and y 2 ( p 2 ). Each
2

such point, when added to y 1 ( p 1 ) + y 1 , results in higher aggregate output than y.

For any such point that lies in the interior of Y 2 , Fig. 2 shows that there exist produc-
tion points in the frontier Ŷ 2 that are bigger. Since these points are on the frontier,
there will exist producer prices that support them as profit maximizing choices for
ν ν
the private-sector. p 2 1 and p 2 2 are two such price vectors. In fact, a sequence of
ν
such price vectors { p } converging to p 2 can be constructed.24 A similar sequence of
2
ν
price vectors for firm 1{ p 1 } converging to p 1 can also be constructed using  y 1 . It
ν ν
is clear from such a construction that, for all ν, we will have y 1 ( p 1 ) + y 2 ( p 2 )  y.

4 Second-best production efficiency: the case of firm-group specific prices and

smooth technologies

The following theorem proves that production efficiency is desirable at any second-
best of a profit-making economy with firm-group specific prices in the case where
technologies are smooth. Conditions (a) and (b) of the theorem ensure that local
Pareto nonsatiation, as defined in Hahn (1973), always holds. In particular, condition

23 Note that these changes imply reducing the input usage of Firm 1 and increasing the input usage of Firm
2.
24 For example, these can be support vectors for a sequence of production vectors converging to y 2 ( p 2 )
and lying on the part of the frontier Ŷ 2 that is on the left of y 2 ( p 2 ).

123
Production efficiency and constraints on profit taxation and profit distribution 593

(b) is the DM version of local Pareto nonsatiation, i.e., there exists a good that is in
positive (or negative) net demand by all consumers.25
Theorem 1 Suppose Assumptions 1, 2, and 3 hold and the government has the power
to implement T ≥ T firm-group specific prices and P, the partition of I \ {0} corre-
sponding to T , is a refinement of P . Suppose either
(a) there exists h such that u h is strictly monotonic or
(b) there exists a commodity k and a consumer h  ∈ H such that
  
(i) xkh (q, m h ) − ekh > 0 and xkh (q, m h ) − ekh ≥ 0, ∀ h ∈ H, q ∈ R+ N , and

m , m ∈ R or
h h
  
(ii) xkh (q, m h ) − ekh < 0 and xkh (q, m h ) − ekh ≤ 0, ∀ h ∈ H, q ∈ R+ N , and

mh , mh ∈ R
holds. Then production efficiency is desirable at the second-best of E p (E, T, P).
Proof Define P0 = {0} so that {P0 , P1 , . . . , PT } is a partition of I. Suppose
s̄ := 
q̄, p̄ 1 , . . . , p̄ T , m̄, m̄ 1 , . . . , m̄ H  ∈ R N +T +H +1 is a solution to (2.16) but
ȳ := t=0 T
y t ( p̄ t ) ∈/ Ŷ . Remark 1 implies that there exist t, t  ∈ {0, . . . , T } such

that p̄ = κ p̄ for any κ ≥ 0. Hence, from Lemma 1 it follows that there exist
t t
v
sequences { p t v } → p̄ t and { p t } → p̄ t such that aggregate supply is greater than ȳ,


v   v
i.e., y t ( p t ) + y t ( p t )  ȳ for all v > v̂, where v̂ is defined as in Lemma 1.
This implies that the aggregate income or the value of aggregate output measured
using consumer prices q̄ increases when producer prices p̄ t and p̄ t change to p t v and


 v
p t for all v > v̂, that is,
⎡ ⎤
 v  v
M v = q̄ · ⎣ y l ( p̄l ) + y t ( p t ) + y t ( p t )⎦ > q̄ · ȳ =: M̄, ∀v > v̂. (4.1)
l=t,t 

 
The continuity of the profit functions π i implies that i∈Pt π i ( p t v ) → i∈Pt
 
π i ( p̄ t ). If i∈Pt π i ( p̄ t ) = 0 then choose any  > 0. If i∈Pt π i ( p̄ t ) > 0 then

choose  such that 0 <  < i∈Pt π i ( p̄ t ). Then, there exists v t such that for all
 
v > v t we have | i∈Pt π i ( p t v ) − i∈Pt π i ( p̄ t )| < . Our choice of  implies that,
 
for every v > v t , the sign of i∈Pt π i ( p t v ) is the same as the sign of i∈Pt π i ( p̄ t ):
  
if i∈Pt π i ( p̄ t ) = 0 then i∈Pt π i ( p t v ) ≥ 0 and if i∈Pt π i ( p̄ t ) > 0 then
 tv t
i∈Pt π ( p ) > 0. Similarly, we can define v .
i
∗  
Pick v to be any v > max{v t , v t , v̂}. Choose scaling factors λt and λt such that
 ∗
v

λt π i ( pt ) = π i ( p̄ t ) and
i∈Pt i∈Pt
 ∗ 
t t v 
λ πi (p )= π i ( p̄ t ). (4.2)
i∈Pt  i∈Pt 

25 See Weymark (1978) for a generalization of condition (b).

123
594 S. Murty


This is possible, e.g., if i∈Pt π i ( p̄ t ) = 0, then λt = 0 needs to be chosen. If

i∈Pt π ( p̄ ) > 0, then λ is given by
i t t


i∈P π ( p̄ )
i t
λ = t
t
∗ , (4.3)
π i ( pt v )
i∈Pt

 ∗
π i ( p t v ) = 0. (Note, λt ≥ 0 and λt ≥ 0.) Define two

which is well defined as i∈Pt
∗  ∗
∗t ∗t v
new price vectors for firm-groups t and t  : p := λt p t v and p := λt p t . Then (4.2)


and the linear homogeneity of the profit functions implies that26

⎛⎛ ⎞ ⎞
∗ h   ∗ t  ∗ t 
h ⎝⎝
r := rP π i ( p̄l )⎠ , π i ( p ), π i ( p )⎠
i∈Pl l=t,t  i∈Pt i∈Pt
⎛ ⎞
 
h ⎝
= rP π i ( p̄ 1 ), . . . , π i ( p̄ T )⎠ =: r̄ h . (4.4)
i∈P1 i∈PT

Thus, the profit incomes of individual consumers do not change when p̄ t and p̄ t

∗t ∗t
change to p and p .
Summing (4.4) over all h and including the profits of the public sector, we obtain27

   ∗t  ∗t
 
T 
π i ( p̄l ) + πi (p ) + πi (p ) = π i ( p̄l ). (4.5)
l=t,t  i∈Pl i∈Pt i∈Pt  l=0 i∈Pl

 ∗t
Thus, the aggregate profits of firms do not change when p̄ t and p̄ t change to p and
  
∗t ∗t ∗t ∗t  ∗t
p . Define y := y t ( p ), and y := y t ( p ).
⎡ ⎤
∗ ∗  ∗ t ∗ t   ∗t ∗t
M v =: M = q̄ · ⎣ y l ( p̄l ) + y + y ⎦ = [(q̄ − p̄l ) · y l ( p̄l )] + [q̄ − p ] · y
l=t,t  l=t,t 
∗ t ∗ t  ∗t ∗t ∗t

∗t

+[q̄ − p ] · y + p̄l · y l ( p̄l ) + p · y + p · y
l=t,t 
 ∗t ∗t ∗t

∗t

= [(q̄ − p̄l ) · y l ( p̄l )] + [q̄ − p ] · y + [q̄ − p ] · y
l=t,t 
   ∗t  ∗t

+ π i ( p̄l ) + πi (p ) + πi (p ) (4.6)
l=t,t  i∈Pl i∈Pt i∈Pt 

26 If any one of t or t  is 0, say t  , then π t  will not be an argument of the income map r
P, p.
27 Note that this is true for both cases: (i) both t and t  are not equal to 0 and (ii) one of t or t  is 0.

123
Production efficiency and constraints on profit taxation and profit distribution 595

Similarly,


T 
T 
M̄ = [(q̄ − p̄l ) · y l ( p̄l )] + π i ( p̄l ). (4.7)
l=0 l=0 i∈Pl

∗
Since M > M̄, it follows from (4.5) that the government’s revenue from commodity

∗t ∗t
taxes is higher when we move to p , p keeping consumer prices and producer prices
for the firm groups other than t and t  unchanged, i.e.,

∗  ∗t ∗t ∗t

∗t

G := [(q̄ − p̄l ) · y l ( p̄l )] + [q̄ − p ] · y + [q̄ − p ] · y (4.8)
l=t,t 


T
> [(q̄ − p̄l ) · y l ( p̄l )] =: Ḡ. (4.9)
l=1

We show that the increase in the government’s revenue can be used to construct
another tax equilibrium where utility of at least one consumer is higher, with no loss
in utility for the others. For all h define x h (q̄, r̄ h + m̄ + q̄ · eh ) =: x̄ h , h x̄ h =: x̄,
 ∗h
and e := h eh . (4.4) implies that for all h, x̄ h = x h (q̄, r + m̄ + q̄ · eh ). Since
 ∗t ∗t

∗ ∗
x̄ ≤ ȳ + e  l=t,t  y l ( p̄l ) + y + y + e =: y + e, we have x̄ ∈ { y + e} + R−− N .

Since x h is a continuous function of qk for all h, clearly, if condition b(i) or

∗ ∗ ∗
b(ii) hold, we can apply the DM argument to find q :=
q̄−k , q k  and  such that
 ∗ ∗h ∗  ∗h ∗
(1) h x h (q, r + m̄ + q · eh ) =: h x ∈ N∗ ( x̄) ⊂ { y + e} + R−− and (2)
N

∗h ∗h
u h (x ) ≥ u h (x̄ h ) for all h and u h (x ) > u h (x̄ h ) for some h.28 This implies that
∗ ∗ ∗t ∗ t
s :=
q, ( p̄l )l=t,t  , p , p , m̄, m̄ 1 , . . . , m̄ H  is another tax equilibrium configuration
of E p (E, T, P) that Pareto dominates s̄.29 This contradicts the fact that s̄ is a solution
to (2.12).
If condition (a) holds, then we can exploit the continuity of x h in m for all h to
 ∗h 
find m̂ > m̄ and ˆ such that (1) h x h (q̄, r + m̂ + q̄ · eh ) =: h x̂ h ∈ Nˆ (x̄) ⊂

28 Note, this implies reducing (increasing) the consumer price on commodity k that every one likes and
has a non-negative net demand (dislikes and has a non-positive net demand).
29 Note, s∗ can be a non-tight tax equilibrium. Note also that it is always possible to make the new tax
equilibrium configuration conform to the normalization rules adopted, e.g., if the normalization rules are

∗ ∗t ∗t
p11 = 1 and p10 = 1 and if t  = 1, 0 and t   = 1, 0, then divide
q, ( p̄l )l=t,t  ,0 , p , p , m̄, m̄ 1 , . . . , m̄ H  by
p̄11 and p̄ 0 by p̄10 .

123
596 S. Murty

∗ ∗h
{ y + e} + R−−
N and (2) u h ( x̂ h ) ≥ u h ( x̄ h ) for all h and u h ( x ) > u h ( x̄ h ) for some
30
h. This again leads to a (possibly non-tight) tax equilibrium that Pareto dominates
s̄, which once again contradicts the hypothesis of the theorem.31 

Theorem 1 shows that, as long as the government’s power to implement firm-group
specific prices is at least T and the corresponding partition is a refinement of P , the
government has enough degrees of freedom for unconstrained manipulation of prof-
its of firm-groups. This ensures second-best production efficiency. However, if the
power of the government to influence firm-group specific prices is less that T, then
the conclusions of the theorem may not follow.32

5 Corollaries of Theorem 1

Two results follow as corollaries of Theorem 1. Corollary 1 demonstrates second-best

production efficiency in economy Eπ (E, T, P) when the partition P is a refinement
of P . It follows because:
(1) The set of tax equilibrium allocations of economy Eπ (E, T, P) are a sub-
set of the set of tax equilibrium allocations of economy E p (E, T, P): If
ᾱτ :=
q̄, p̄ 0 , p̄, τ̄ 1 , . . . , τ̄ T , m̄, m̄ 1 , . . . , m̄ H  is a tax equilibrium of econ-
omy Eπ (E, T, P), then ᾱ p :=
q̄, p̄ 0 , p̄ 1 , . . . , p̄ T , m̄, m̄ 1 , . . . , m̄ H  with p̄ t :=
(1 − τ̄ t ) p̄ for all t = 1, . . . , T is a tax equilibrium of economy E p (E, T, P).33
Both tax equilibria result in the same allocation.
(2) Every second-best of E p (E, T, P) can be decentralized as a tax equilibrium
of Eπ (E, T, P): If
q, p 0 , p 1 , . . . , p T , m, m 1 , . . . , m H  is a second-best of
E p (E, T, P), then Theorem 1 implies that there exist positive scalars λ2 , . . . , λT
such that p t = λt p 1 for all t = 2, . . . , T . Choose p = p 1 , τ 1 = 0, and
τ t = 1 − λt for all t = 2, . . . , T . Then
q, p, p 0 , τ 1 , . . . , τ T , m, m 1 , . . . , m H 
is a tax equilibrium of Eπ (E, T, P).
Corollary 1 of Theorem 1 Production efficiency is desirable at the second-best of
economy Eπ (E, T, P) when the partition P is a refinement of P .
Corollary 2 of Theorem 1 demonstrates that optimal intermediate input taxation
and optimal proportional profit taxation are perfect substitutes: if
q, p 0 , p 1 , . . . , p T ,
m, m 1 , . . . , m H  is a second-best of economy E p (E, T, P) where the partition P is a
refinement of P , then Theorem 1 implies that all producer prices are proportional, so
that there exist positive scalars λ2 , . . . , λT such that p t = λt p 1 for all t = 2, . . . , T .

∗
30 Note, this is made possible by the fact that G > Ḡ in (4.8), so that it is possible to distribute all or a part
of this increased government budget-surplus as a higher demogrant.
31 Once again, the new tax equilibrium can be reconfigured to conform to our normalization rules.
32 See Blackorby and Murty (2009) for an example of second-best production inefficiency when the gov-
ernment lacks sufficient profit taxation power. The example there assumes P is such that T = H > 1. It
considers the extreme case where the government has no profit taxation power (0 = T < T). It can also be
re-worked also for the case of uniform profit taxation (T = 1).
33 Note that the reverse is not true, e.g., consider a production inefficient tax equilibrium of E (E, T, P).
p

123
Production efficiency and constraints on profit taxation and profit distribution 597

Thus, the wedge between price vectors of firm-groups t and t  at this second-best is

p 1 (λt − λt ). Transactions in commodities between these two firm-groups are, hence,

taxed proportionately at a rate λt − λt . Alternatively, from Corollary 1 of Theorem 1
(above) it follows that this second-best can be decentralized as a second-best tax equi-
librium of economy Eπ (E, T, P) with firm-group specific proportional profit taxes.
But this precludes intermediate goods taxation as all firm-groups face the same pro-
ducer price p 1 in this economy. Rather, firm-groups are subject to firm-group specific
profit tax rates τ t = 1 − λt for all t = 1, . . . , T .
Corollary 2 of Theorem 1 If profit taxes cannot be implemented then intermediate-
inputs should be taxed proportionately. Alternatively, firm-group specific profit taxa-
tion justifies not taxing transactions between these firm groups in profit making econ-
omies.
As a special case, the famous implication with regards to intermediate input taxa-
tion for DM and Guesnerie (1995) models follows:34 If government can implement
100 % profit taxation and redistribute proceeds to consumers as a demogrant, then
intermediate input taxation is not desirable.

6 The case of non-smooth technologies: The Mirrlees (1972) counterexample

In the case of economies with smooth technologies, proportional, lump-sum, and affine
profit taxation are equivalent policy instruments. But this need not be the case for econ-
omies with non-smooth technologies. We demonstrate this in the context of Mirrlees
counterexample with a non-smooth technology, where it is claimed that second-best
production efficiency may not be desirable in an economy with firm specific prices.35
From this one might infer that second-best production efficiency may not be desirable
with firm specific proportionate profit-taxation. Here, we rework that counterexam-
ple employing the intuition underlying the proof of Theorem 1 to show that, even in
this case, second-best production efficiency will be true if government can implement
lump-sum profit taxation in addition to firm-group specific prices.
Example 6 Suppose N = 2, there is one consumer in the economy whose preferences
are represented by indifference curves as seen in Figs. 3 and 4, the government does
not have access to a demogrant, and I = {0, 1, 2}.36 The government has 50 units
of the second commodity and none of the first. In this example, in the absence of
a demogrant, the government can provide its endowment to the consumer through
direct sale of good 2 at consumer prices or in the form of dividend incomes by using
its endowment to finance profit subsidies to the firms. In this example, we show that
a lump-sum (and not a proportional) profit subsidy results in second-best production
efficiency. The technologies of public-sector and private sector firms are (these are
also depicted in Figs. 3 and 4)

34 Recall, this case is equivalent to the case where P̂ is the coarsest partition of private firms, i.e., P̂ =
{I \ {0}} and θih = H1 for all h and i ∈ I \ {0}.
35 Reinhorn (2011) has formulated the precise condition that underlies this case.
36 In a one consumer economy, a demogrant will be a sufficient instrument to achieve the first-best transfers.

123
598 S. Murty

Fig. 3
y12 , y 22 , y 20

^
2
1 (y 2 ) 0
(y 0 )
Y y 0

~
0
Y 2 2
(y )
~ ^

Y 2 ^
2 ~ y1 1
(y1 )
y 2
y
1 2 0 ^
y ,y ,y
1 1 1 y1

^
q
y2 y2
^
Y ~
y q y
Y

^ ^
y y
^ ^
2 2

y1 y1
a b
Fig. 4

 
Y 1 = y 1 ∈ R2 | y21 ≤ 5(−y11 ), 0 ≤ (−y11 ) ,

Y 2 = y 2 ∈ R2 | y22 ≤ 3(−y12 ), if 0 ≤ (−y12 ) ≤ 4

≤ 8 + (−y12 ), if (−y12 ) ≥ 4 , and
 
Y = y ∈ R | y2 ≤ 50, 0 ≤ (−y1 ) .
0 0 2 0 0
(6.1)

so that the aggregate technology is

Y = Y 1 + Y 0 = {y ∈ R2 | y2 ≤ 50 + 5(−y1 ), 0 ≤ (−y1 )}. (6.2)

2
Any production vector y = i=0 y i with y i ∈ Ŷ i for all i ∈ {0, 1, 2} lies in Ŷ if
and only if the sets of support prices of y 0 , y 1 and y 2 have a non-empty intersection.
As in Mirrlees (1972), the best possible production-efficient tax equilibrium
2 with firm
specific prices corresponds to the aggregate production vector ȳ = i=0 ȳ i ∈ Ŷ ,

123
Production efficiency and constraints on profit taxation and profit distribution 599

where ȳ 0 =
0, 50, ȳ 1 =
−5, 25, and ȳ 2 =
0, 0 (see Fig. 4a). The support prices
of ȳ 0 , ȳ 1 , ȳ 2 are the cones

ρ 0 ( ȳ 0 ) = { p ∈ R++
2
| p = λ1
1, 0 + λ2
0, 1, ∀λ1 ≥ 0, λ2 ≥ 0}
ρ 1 ( ȳ 1 ) = { p ∈ R++
2
| p = λ
5, 1, ∀λ ≥ 0}
ρ 2 ( ȳ 2 ) = { p ∈ R++
2
| p = λ1
1, 0 + λ2
3, 1, ∀λ1 ≥ 0, λ2 ≥ 0}. (6.3)

It is clear that ρ 0 ( ȳ 0 )∩ρ 1 ( ȳ 1 )∩ρ 2 ( ȳ 2 ) = ρ 1 ( ȳ 1 ) = ∅. However, as Mirrlees shows,

we can do better by moving to a production-inefficient tax equilibrium corresponding
to aggregate production ŷ =
−4, 62, where ŷ 2 =
−4, 12, ŷ 0 = ȳ 0 =
0, 50,
while Firm 1 shuts down (see Fig. 4a). In this case, the support prices of ŷ 2 and ŷ 1 are
the cones (see Fig. 3):

ρ 1 ( ŷ 1 ) = { p ∈ R++
2
| p = λ1
1, 0 + λ2
5, 1, ∀λ1 > 0, λ2 > 0} and
ρ ( ŷ ) = { p ∈
2 2 2
R++ | p = λ1
1, 1 + λ2
3, 1, ∀λ1 ≥ 0, λ2 ≥ 0}, (6.4)

so that ρ 1 ( ŷ 1 )∩ρ 2 ( ŷ 2 ) = ∅. For any non-zero price vector in ρ 2 ( ŷ 2 ) that is not propor-
tional to
3, 1, Firm 2 makes positive profits. Suppose the price vector is p̂ 2 =
1, 1,
so that the associated profit of Firm 2 is π̂ 2 = −4 + 12 = 8. The tax equilibrium
corresponding to ŷ is Pareto superior to the one corresponding to ȳ as the consumer
receives a positive income in the former case coming in the form of profit income from
Firm 2 (see Fig. 4a).37 In Fig. 4a, consumer price vector at ŷ and ȳ is q̂ =
4, 1.
However, if the government can implement firm-specific lump-sum profit taxation,
then we claim that the tax equilibrium corresponding to ŷ can be improved upon.
Production inefficiency of ŷ implies there are changes in the producer prices leading
to new producer prices such as p̃ 1 =
5, 1 and p̃ 2 =
3, 1 for Firms 1 and 2 which
can support ỹ 1 =
−1, 5 and ỹ 2 =
−3, 9 as profit maximizing choices of Firms
1 and 2, so that, assuming no change in public sector production from ŷ 0 , aggregate
production is ỹ =
−4, 64 > ŷ (see Figs. 3 and 4b). There is an increase in the
potential income (evaluated at the consumer price vector q̂) generated in the econ-
omy: it increases from q̂ ŷ = −16 + 62 = 46 to q̂ ỹ = −16 + 64 = 48. Question is can
this increased income be used to generate a tax equilibrium that is Pareto superior to
the tight tax equilibrium corresponding to ŷ. The answer is yes, if the government can
implement profit taxation in a lump-sum manner. Note, Firm 2 generates zero profits
at ỹ 2 . Suppose a negative lump-sum profit-tax (a profit subsidy) is imposed on Firm
2, so that Firm 2 continues to receive net-of-tax profits equal to π̂ 2 = 8. This implies
that the consumer can continue receiving the same profit income as in the tight tax
equilibrium corresponding to ŷ. If he continues facing the same consumer price vector
q̂, there is no change in the consumer demand from ŷ. The consumer demand ŷ with

37 Notice, Firm 1’s technology is constant returns to scale and hence this firm generates no profits at both
ȳ and ŷ.

123
600 S. Murty

producer supply ỹ is a new non-tight tax equilibrium that leaves consumer as well off
as in the tight tax equilibrium corresponding to ŷ. We have

q̂ ỹ ≡ q̂ ŷ 0 + (q̂ − p̃ 1 ) ỹ 1 + (q̂ − p̃ 2 ) ỹ 2 − π̂ 2 + π̂ 2 > q̂ ŷ

≡ q̂ ŷ 0 + (q̂ − p̂ 1 ) ŷ 1 + (q̂ − p̃ 2 ) ŷ 2
+πˆ2 ⇒ q̂ ŷ 0 + (q̂ − p̃ 1 ) ỹ 1 + (q̂ − p̃ 2 ) ỹ 2 − π̂ 2 > q̂ ŷ 0
+(q̂ − p̂ 1 ) ŷ 1 + (q̂ − p̃ 2 ) ŷ 2 . (6.5)

That is government revenue (even after paying the profit subsidy) at the new non-tight
tax equilibrium corresponding to ỹ is higher than the government revenue at the tight
tax equilibrium corresponding to ŷ. The government can potentially use the increase
revenue to decrease consumer price of commodity 2, which the consumer likes and
hence achieve a Pareto superior tax equilibrium compared to the one corresponding to
ŷ. In Fig. 4b the budget line of consumer pivots to π̂ 2 s  from π̂ 2 s, so that the budget
set of consumer expands. Consumer will now choose a bundle which is welfare-wise
superior to ŷ but yet less that ỹ, resulting in a Pareto-superior non-tight tax equilibrium.
The same can be repeated for any production inefficient tax equilibrium, and hence
second-best production efficiency is true in this economy with firm-specific producer
prices and lump-sum profit taxation. This also implies that second-best production
efficiency is true in an economy with firm-group specific affine profit taxation.

7 Conclusions

There is a classic literature that studies the desirability of production efficiency in

economies with Ramsey taxation where firms make positive profits which can poten-
tially be partly taxed away and partly distributed back to consumers. The results of
this classic literature are often invoked to justify the use of producer prices as proxies
for shadow prices in cost-benefit tests of marginal public sector projects. We show that
the desirability of second-best production efficiency depends on the link between the
constraints on government’s profit taxation power and the institutional rules by which
(profit) incomes are distributed in the economy. We generalize results in the literature
by showing that second-best production efficiency remains desirable when the parti-
tion of private firms induced by the profit taxation power of the government is at least
as fine as the partition of firms induced by the institutional distribution of profits to
consumers. The two cases studied in the literature of firm-specific and uniform (e.g.,
100 %) profit taxation follow as two special cases of our general result.
The result follows because, at any production inefficient status-quo of such econ-
omies, the private sector producer price vector and the prices in the public sector (the
latter reflect the true shadow prices in the economy) are not proportional. The differ-
ences in the marginal rates of substitution in the private and public sectors imply that
production can be reallocated between these sectors to increase aggregate output and
income in the economy. However, we want the increase in supply to be brought about as
a result of profit maximization by firms. Lemma 1 proves that small changes in the two
price vectors can be constructed that ensure that a potential increase in the aggregate

123
Production efficiency and constraints on profit taxation and profit distribution 601

output can be supported as profit maximizing choices of firms. If the profit taxation
power of the government is aligned with the institutional rules of profit distribution in
the economy, then the government has complete control over profit distribution in the
economy: e.g., by profit taxation the government can effectively ensure that the net of
tax firm-group profits, and hence the profit incomes of the consumers that depend on
the distribution of the firm-group profits, remain at the status-quo levels. This must
mean that the increase in aggregate income shows up as an increase in the tax and
public sector incomes of the government, which can be used to change commodity
taxes or to increase demogrant incomes of people in a Pareto improving way. Thus,
there are always Pareto improvements at any production inefficient status-quo of such
economies.
The mechanism suggests why this strategy does not work, generally, in most private
ownership economies when the profit taxation power of the government is not consis-
tent with the rules of profit distribution. This is because, while a production inefficient
status-quo suggests that there are changes in producer prices that can increase the
aggregate output in the economy, the government does not have enough instruments
to control profit distribution in the economy. Hence, all of the increased output may
not, in general, become available to the government for designing Pareto improving
changes in taxes and demogrant. Private ownership diverts some of the increased
resources from the government coffers and puts it into the hands of consumers as
profit incomes. But the private ownership structure could be such that it may lead to
an inequitable distribution of profit incomes and a decrease in welfare of some con-
sumers, which government policy is unable to correct with the remaining resources,
i.e., there may exist no directions of change in the government policy instruments that
are Pareto-improving, equilibrium preserving, and compatible with the existing pri-
vate ownership structure. Our analysis hence suggests how, by understanding the rules
of profit income distribution in the economy, the government can potentially design
profit taxation that can promote both its redistributive and efficiency objectives. Note,
however, that the analysis is silent about the causes that led to the existing distribution
of profit incomes. In this paper, the structure of private ownership, and hence the way
in which profits of firms are distributed to consumers, has been taken to be a prim-
itive of the economy, i.e., as something that is exogenously given. An analysis that
identifies the causes of and endogenizes the profit distribution in the economy (e.g.,
by introduction of a market for equity) is of interest and requires an extension of the
current model, which is an agenda for further research.
In smooth economies it is shown that optimal profit taxation is a perfect substitute
for optimal intermediate input taxation and that proportional, lump-sum, and affine
modes of taxing profits are equivalent. In non-smooth economies these different modes
of profit taxation may not be equivalent. This is shown in the context of the counter-
example provided by Mirrlees (1972), which seeks to demonstrate that production
inefficiency may be desirable at a second-best in profit making economies. Rework-
ing this example, we show that, even in this case, second-best production efficiency is
desirable if the government can implement an affine system of profit taxation. Hence,
second-best analysis of economies where profit taxation power of the government is
well-aligned with the rules of profit distribution to consumers supports tax systems
such as VAT.

123
602 S. Murty

Acknowledgments The basis of this paper is a series of discussions with Charles Blackorby while we
were colleagues at the University of Warwick. These discussions led to the identification of some impor-
tant cases of a more general result which this paper formalizes and proves. I am very grateful to Charles
Blackorby for these discussions, which have made it possible for me to study this problem and to present
these results. I am also grateful to the Associate Editor of this journal and two anonymous referees for very
carefully going through the manuscript and for their very insightful and helpful comments, which have
gone a long way in improving this work. Should their remain errors in this work, I bear full responsibility
for them.

Appendix

Proof of Lemma 1 Smoothness of Ŷ t and Ŷ t implies that H ( p̄ t , p̄ t · y t ( p̄ t )) and
    
H ( p̄ t , p̄ t · y t ( p̄ t )) are unique supporting hyperplanes for Y t and Y t at y t ( p̄ t ) and
 
y t ( p̄ t ), respectively.

Step 1. Since p̄ t and p̄ t are not collinear, H ( p̄ t , 0) is not a supporting hyperplane
 
for H≥ ( p̄ t , 0) and H ( p̄ t , 0) is not a supporting hyperplane for H≥ ( p̄ t , 0) at 0 N . This

implies that there exist y t ∈ R N and y t ∈ R N such that the following is true:38


y t ∈ H< ( p̄ t , 0) ∩ H≥ ( p̄ t , 0),
 
y t ∈ H< ( p̄ t , 0) ∩ H≥ ( p̄ t , 0), and

y t + y t  0 N . (A.1)
  
This implies that y t ( p̄ t ) + y t ( p̄ t ) + y t + y t  ȳ. Denote y t ( p̄ t ) by ȳ t and
t    
y ( p̄ t ) by ȳ t . Since ȳ t and ȳ t belong to Ŷ t and Ŷ t , Assumption 3 implies that
 
f t ( ȳ t ) = 0 and f t ( ȳ t ) = 0.
Step 2. Recall that ∇ f t ( ȳ t ) is defined as the linear mapping such that for all
v
{h } → 0 N , we have

f t ( ȳ t + h v ) − [ f t ( ȳ t ) + ∇ f t ( ȳ t )h v ] e(h v , ȳ t )
lim ≡ lim = 0, (A.2)
hv → 0N |h v | hv → 0N |h v |

where e(h v , ȳ t ) = f t ( ȳ t + h v ) − [ f t ( ȳ t ) + ∇ f t ( ȳ t )h v ]. We show that there exists

γ t > 0 such that ∇ f t ( ȳ t ) = γ t p̄ t . Take any point y ∈ Y t such that y = ȳ t . Then,
from the convexity of Y t , f t ( ȳ t + λ(y − ȳ t )) ≤ 0 for all λ ∈ [0, 1]. Using (A.2) and
the fact that f t ( ȳ t ) = 0, we have

∇ f t ( ȳ t )(y − ȳ t ) f t ( ȳ t + λ(y − ȳ t )) − f t ( ȳ t )
= lim
| y − ȳ t | λ→0 | y − ȳ t | λ
f ( ȳ + λ(y − ȳ t ))
t t
= lim ≤ 0. (A.3)
λ→0 | y − ȳ t | λ

38 I am grateful to a referee for the following proof of (A.1): Since p̄ t and p̄ t  are not collinear, there exists

ξ ∈ R N such that ξ · p̄ t < 0 and ξ · (− p̄ t ) < 0. Take y t = ξ + 1 N for a sufficiently small , and

y t = −ξ . (The intuition becomes clear when one sees Fig. 1.)

123
Production efficiency and constraints on profit taxation and profit distribution 603

Since this is true for all y ∈ Y t , ∇ f t ( ȳ t ) is a normal to a supporting hyperplane of

Yt at ȳ t . Since, Ŷ t is smooth and H ( p̄ t , p̄ t · ȳ t ) is also a supporting hyperplane of Y t
at ȳ t , there must exist γ t > 0 such that ∇ f t ( ȳ t ) = γ t p̄ t . Similarly, we can prove that
    
there exists γ t > 0 such that ∇ f t ( ȳ t ) = γ t p̄ t .
  
Step 3. (A.3) implies that y t · ∇ f t ( p̄ t ) < 0 and y t · ∇ f t ( p̄ t ) < 0. Choose
a sequence {λv } such that λv y t → 0 and λv > 0 for all v. We now show that there
exists v  such that for all v > v  , we have y t v := ȳ t + λv y t ∈ Y t . From (A.2) we
have

f t ( ȳ t + λv y t ) − f t ( ȳ t ) − λv ∇ f t ( ȳ t )y t
lim = 0. (A.4)
v
λ →0 | y t | λv

Since f t ( ȳ t ) = 0 and ∇ f t ( ȳ t )y t < 0 (from Step 2), we have

f t ( ȳ t + λv y t ) ∇ f t ( ȳ t )y t
lim v
= < 0. (A.5)
v
λ →0 | y | λ
t | y t |

Hence, there exists a large enough v  such that for all v > v  , we have f t ( ȳ t +λv y t ) <
0, and hence y t v := ȳ t + λv y t ∈ intY t for all v > v  .39 Similarly, we can prove
v   
that there exists v  such that for all v > v  , we have y t := ȳ t + λv y t ∈ intY t .
v
Step 4. We now show that there exist sequences { p t v } and { p t }, and a positive
v   v 
integer v̂ such that for all v > v̂, we have y t ( p t ) + y t ( p t )  ȳ t + ȳ t . Define
v
v̂ = max{v  , v  }. For every v > v̂, y t ∈ Y t . It can therefore be shown that there
v
are continuous maps κ t (y t v ) := maxκ {κ ≥ 0[y t v + κ1 N ] ∈ Y t } and κ t (y t ) :=


 t v v
maxκ {κ ≥ 0[y +κ1 N ] ∈ Y t }.40 For all v > v̂, it is clear that (i) y t v +y t  ȳ t + ȳ t
 

v v
and so (y t v +κ t (y t v )1 N )+(y t +κ t (y t )1 N )  ȳ t + ȳ t , (ii) (y t v +κ t (y t v )1 N ) and
   

v v
(y t +κ t (y t )1 N ) belong to Ŷ t and Ŷ t , respectively, and (iii) {y t v +κ t (y t v )1 N } → ȳ t
 

v v v
and {y t + κ t (y t )1 N } → ȳ t . Define p t v = γ1t ∇ f t (y t v + κ t (y t v )1 N ) and p t =
   

 v  v 
∇ f t (y t + κ t (y t )1 N ). The smoothness of functions f t and f t imply that
1
γt
v v
{ p } → p̄ t and { p t } → p̄ t . Clearly, y t ( p t v ) = y t v + κ t (y t v )1 N and y t ( p t ) =
tv  

v v v
y t + κ t (y t )1 N , so that for all v > v̂, we have y t ( p t v ) + y t ( p t )  ȳ t + ȳ t .
  

v
Hence, for all v > v̂, the conclusions of the lemma follow for sequences { p } and t
v
{ p }.
t 


References

Blackorby C, Murty S (2009) Constraints on income distribution and production efficiency in economies
with Ramsey taxation. Warwick Economic Research Papers No. 908, Department of Economics,
University of Warwick
Boadway R (1972) Cost benefit rules in general equilibrium. Rev Econ Stud 39:87–103
Dasgupta P, Stiglitz J (1972) On optimal taxation and public production. Rev Econ Stud 39:87–103

39 For any set A ⊂ Rn , int A is the interior of A relative to Rn .

40 Assumptions 1 and 2 guarantee the existence of such maps.

123
604 S. Murty

Diamond P, Mirrlees J (1971a) Optimal taxation and public production I: production efficiency. Am Econ
Rev 61:8–27
Diamond P, Mirrlees J (1971b) Optimal taxation and public production II: tax rules. Am Econ Rev 61:
261–278
Drèze J, Stern N (1987) The theory of cost benefit analysis. In: Auerbach AJ, Feldstein M (eds) Handbook
of public economics II, chap 14. Elsevier Science, North Holland
Guesnerie R (1995) A contribution to the pure theory of taxation. Cambridge University Press, Cambridge
Hahn F (1973) on optimum taxation. J Econ Theory 6:96–106
Little IMD, Mirrlees J (1974) Project appraisal and planning for developing countries. Heinemann Educa-
tional Books, London
Mirrlees J (1972) On producer taxation. Rev Econ Stud 39:105–111
Munk KJ (1980) Optimal taxation with some non-taxable commodities. Rev Econ Stud 47:755–765
Myles GD (1995) Public economics. Cambridge University Press, Cambridge
Ramsey F. (1927) A contribution to the theory of taxation. Econ J 37:47–61
Reinhorn L (2011) Production efficiency and excess supply. Working Papers 2011_13, Department of
Economics and Finance, Durham University
Sadka E (1977) A note on producer taxation. Rev Econ Stud 44:385–387
Weymark J (1978) On Pareto improving price changes. J Econ Theory 19:338–346
Weymark J (1979) A reconciliation of recent results in optimal taxation theory. J Public Econ 12:171–189

123