Arrows Externalities Model
Uploaded by
Naresh SehdevArrows Externalities Model
Uploaded by
Naresh SehdevArrow’s Walrasian Model
of Public Goods and Other Externalities
Arrow showed that we can recast the public-goods allocation problem as one involving
only private goods, so that our Walrasian analysis applies. Arrow defined each individual’s
consumption of the public good as a distinct commodity, with a distinct market and price,
but with “jointness” in the production of these goods. Here’s how this works in our one-
public-good-one-private-good model with n consumers (where X is the public good and Y is
the private good):
We redefine the economy as having n + 1 goods X1 , . . . , Xn , Y , with quantities denoted by
x1 , . . . , xn , y. An allocation is therefore an n(n + 1)-tuple
n(n+1)
(x11 , . . . , x1n , y 1 ), (x21 , . . . , x2n , y 2 ), . . . , (xn1 , . . . , xnn , y n ) ∈ R+
.
However, both the production possibilities and the consumption possibilities in this economy
are assumed to have a special character:
(1) The X-goods are “joint products” in any firm’s production process: A production plan
for a firm is an (n + 1)-tuple (z, q) = (z, q1 , . . . , qn ) ∈ Rn+1
+ , where z is the amount of the
private good the firm uses as input and qi is the output of commodity Xi , but the firm has the
technological constraint q1 = q2 = · · · = qn . This is exactly like the classical joint products
mutton and wool that are produced by raising sheep.
(2) Consumer i’s consumption set is {(xi1 , . . . , xin , y i ) ∈ Rn+1
+ | j 6= i ⇒ xij = 0} — i.e.,
Consumer i can consume only the goods Xi and Y . So while Consumer i’s utility function
ui is technically defined on the domain Rn+1 i
+ , we can more intuitively write u as defined
on bundles (xi , y i ) ∈ R2+ . Therefore we can simplify the notation, defining an allocation to
consumers as a 2n-tuple (xi , yi )n1 ∈ R2n
+ .
Now a Lindahl equilibrium is just a Walrasian equilibrium of this joint-product economy.
Specifically (and assuming for simplicity that there is just a single producer/firm, which is
p1 , . . . , pbn , pby ) ∈ Rn+1
a price-taker), a Walrasian equilibrium is a price-list (b + , a consumption
xi , ybi )n1 ∈ R2n
allocation (b z , qb1 , . . . , qbn ) ∈ Rn+1
+ and a production plan (b + that satisfy
(U-max) xi , ybi ) maximizes ui (xi , yi ) subject to pbi xi + yi 5 ẙi + θi π(b
∀i : (b z, q
b)
b) maximizes π(z, q1 , . . . , qn ) = ni=1 pbi qi − pby z subject to q1 = · · · qn = f (z)
P
(π-max) (b
z, q
zb + ni=1 ybi 5 ni=1 ẙi , with equality if pby > 0.
P P
(M-Clr) ∀i : x
bi = qbi and
Therefore the First Welfare Theorem applies: if the utility functions and production functions
satisfy the usual assumptions, then the equilibrium allocation will be Pareto efficient.
But Arrow’s model also makes it clear that the Walrasian models’s price-taking assumption
for consumers is unrealistic here: for each of the distinct goods Xi there is only one person
on the demand side of the market. The only person who cares about the good Xi is person
i. It’s clearly unrealistic to assume that any of the participants will take their own price (or
Lindahl cost share) as given. This was Arrow’s motivation for modeling things this way —
to clarify this point.
