I mentioned in a recent post that one can often improve models by increasing the detail in their microfoundations. The case of transaction costs in output models illustrates my point.
Suppose that annual output is modelled by annual output = A*capital^B where A and B are constants. Then we could include transaction costs as annual output = A*capital^B + K for K a constant depending on the expense of transaction, presumably depending on distance between buyers and sellers of the goods.
We could specify more detail: output per transaction = A2*capital^B2 + K2 for A2, B2, K2 constants, and then sum over the number n of transactions per year to get annual output = n*A2*capital*B2 + n*K2. So as the number of transactions increases, the advantage of local trading increases, other things being equal.
The observation isn't very smart, but illustrates a point.