Let E1,E2,…,En and F be commutative monoids written additively and let u be a mapping of E1×⋯×En into F, which is distributive with respect to all the variables. For i=1,2,…,n, let Li be a non-empty finite set and (xi,λ)λ∈Li a family of elements of Ei. Let yi=∑λ∈Lixi,λ for i=1,2,…,n. Then
(2)
u(y1,…,yn)=α∑u(x1,α1,…,xn,αn)
the sum being taken over all sequences α=(α1,…,αn) belonging to L1×⋯×Ln.
We argue by induction on n, the case n=1 following from formula (2) of § 1, no. 2. From the same reference
(3)
u(y1,…,yn−1,yn)=αn∈Ln∑u(y1,…,yn−1,xn,αn)
for yn=∑αn∈Lnxn,αn and the mapping z↦u(y1,…,yn−1,z) of En into F is a magma homomorphism. By the induction hypothesis applied to the distributive mappings (z1,…,zn−1)↦u(z1,…,zn−1,xn,αn) from E1×⋯×En−1 to F,
(4)
u(y1,…,yn−1,xn,αn)=α1,…,αn−1∑u(x1,α1,…,xn−1,αn−1,xn,αn), the sum being taken over the sequences
(α1,…,αn−1) belonging to
M=L1×⋯×Ln−1. Now
L1×⋯×Ln=M×Ln; writing
tα1,…,αn=u(x1,α1,…,xn,αn), we have (5)
α1,…,αn∑tα1,…,αn=αn∑(α1,…,αn−1∑tα1,…,αn−1,αn) by formula (7) of § 1, no. 5. (2) follows immediately from (3), (4) and (5).
Bourbaki puts the proof directly after the statement, so what is above is the statement and its proof together.